ANALYSIS OF FDA MEDICAL DEVICE RECALLS TIME BETWEEN RECALL INITATION DATE AND TERMINATION DATE ____________ A Thesis Presented to the Faculty of California State University, Dominguez Hills ____________ In Partial Fulfillment of the Requirements for the Degree Master of Science in Quality Assurance ____________ by Teonta Pace Spring 2020 THESIS: ANALYSIS OF FDA MEDICAL DEVICE RECALLS: TIME BETWEEN RECALL INITATION DATE AND TERMINATION DATE AUTHOR: TEONTA PACE APPROVED: ______________________________ Bob Mehta, MSQA Thesis Committee Chair ______________________________ Rochelle Cook, Ph.D., MS, PMP, CPHQ, ASQ CMBB Committee Member ______________________________ Mary McShane-Vaughn, Ph.D. Committee Member Copyright by TEONTA PACE 2020 All Rights Reserved Dedicated to the memory of my late mother Gail and my father Clay, who taught me to believe there is no glass ceiling for those of us who want to reach the top of each peak. They taught me that the resolution to any problem involves research, careful planning, and flawless execution. Each moment in every library, museum, and educational campus we’ve visited has led me to this moment. Thank you for encouraging my ambition, fostering my work ethic, and supporting every moment of courage in my life. Without you, I would not be the force that I’ve become, and many of my successful days would not have been realized. I owe everything that I am to your leadership, may I take that with me as I push forward in the rest of life’s endeavors. ACKNOWLEDGEMENTS I would like to acknowledge the wonderful people that supported my studies and the completion of this thesis. To my sister, Brittnee Booker, for her continuous support, expertise, and advice in guiding me through the program and the thesis. Thank you to my executive and senior leadership in quality and compliance for providing me with the knowledge, resources, and expertise in my area of focus. To my colleagues who always provided me valuable industry insight during my studies and encouraged and pushed me to go forward despite any challenges that surfaced during my road to completion. Lastly, I would like to thank my academic mentors and professors, with a special thanks to Bob Mehta, Dr. Rochelle Cook, Dr. Mary McShaneVaughn, and Dr. Jim Clauson for their support, encouragement, and leadership during my studies and thesis completion. v TABLE OF CONTENTS ACKNOWLEDGEMENTS ............................................................................................................V TABLE OF CONTENTS .............................................................................................................. VI LIST OF TABLES ......................................................................................................................VIII LIST OF FIGURES ...................................................................................................................... IX ABSTRACT....................................................................................................................................X 1. INTRODUCTION .......................................................................................................................1 Background ..........................................................................................................................1 Statement of the Problem ...................................................................................................10 Purpose of Study and Significance ....................................................................................10 Theoretical Bases of the Study ..........................................................................................12 Scope and Limitations........................................................................................................13 Definition of Terms............................................................................................................14 2. LITERATURE REVIEW ..........................................................................................................19 Roles and Responsibilities: Recall Regulation and Guidance ...........................................19 FDA Enforcement Actions for Software Defects and Recalls ...........................................30 Recall Timing and Effectiveness Studies ..........................................................................34 Plan-Do-Study-Act (PDSA)...............................................................................................37 Summary of Literature Review ..........................................................................................38 3. METHODOLOGY ....................................................................................................................40 Design of the Investigation ................................................................................................40 Population of Sample .........................................................................................................41 Data Collection ..................................................................................................................42 Data Analysis .....................................................................................................................47 Summary ...........................................................................................................................57 4. RESULTS AND DISCUSSION ................................................................................................58 Software Related Recall Results ........................................................................................58 Summary of Key Characteristics for Outliers ....................................................................63 Act Phase ...........................................................................................................................67 5. CONCLUSION ..........................................................................................................................71 vi Summary .............................................................................................................................71 Conclusions .........................................................................................................................71 Recommendations ...............................................................................................................72 REFERENCES ..............................................................................................................................74 APPENDIX A: FLOWCHART OR RECALL MANAGEMENT PROCESS..............................81 APPENDIX B: CODED SOFTWARE RELATED RECALL DATA SET ..................................82 APPENDIX C: DETERMINTATION OF DISTRIBUTION FITTING FOR TIME TO CLASSIFCATION ........................................................................................................................93 APPENDIX D: BOX-PLOT RESULTS: OUTLIERS FOR TIME TO CLASSIFICATION .......95 APPENDIX E: DETERMINATION OF DISTRIBUTION FITTING FOR TIME TO RECALL ..................................................................................................................................96 APPENDIX F: BOX-LOT RESULTS: OUTLIERS FOR TIME TO RECALL ...........................98 APPENDIX G: CAUSE-AND-EFFECT, FISHBONE DIAGRAM FOR ROOT CAUSE ANALYSIS OF FACTORS LEADING TO RECALL INITIATION FOR SOFTWARE RELATED DEFECTS .................................................................................................................100 APPENDIX H: CAUSE AND EFFECT, FISHBONE DIAGRAM FOR ROOT CAUSE ANALYSIS OF FACTORS LEADING PROLONGED TIME TO RECALL DURATION.................................................................................................................................101 APPENDIX I: PROPOSED FMEA FOR CONFIRMED ROOT CAUSES CONTRIBUTING TO TIME TO RECALL DELAY, INCLUDING RISK MITIGATION RECOMMENDATIONS .............................................................................................................102 vii LIST OF TABLES 1. Communication Responsibilities for Recall Notification and Corrective Action Implementation for Responsible Parties Under E.U. MDR 2017/745...............................28 2. Goodness of Fit Test for Distribution of Data for Time to Classification .........................60 3. Goodness of Fit Test for Distribution of Data for Time to Recall .....................................62 viii LIST OF FIGURES 1. FDA Triggers that Lead to Organization Product Recall ....................................................6 2. Primary and Secondary Effects of Organization Initiation of Product Recalls ...................9 3. Flow Diagram Representing an Overview of Roles and Responsibilities During Recall Lifecycle for Organization and FDA .................................................................................22 4. Recall Identification and Initiation Pathways for Immediate Recalls Demonstrating Step 1 of Uniform Recall Procedures ...........................................................................................27 5. Summary of Basic Requirements of Recall Process Combining Regulatory Requirements of US (FDA), Australia (TGA), ISO, Europe (EU), and Canada (HC) .............................30 6. The Model for Improvement ..............................................................................................38 7. Pareto Chart Demonstrating Frequency of Recall Defects by Category ...........................44 8. Fishbone Diagram Template ..............................................................................................45 9. Example FMEA Data Table...............................................................................................47 10. Pie Chart Representing Recall Initiation Modes ...............................................................51 11. Bar Chart of Recall Initiation Pattern by Region ...............................................................52 12. Pareto Chart Displaying Recall Event Defect Category ....................................................54 13. Bar Chart of Software Related Recalls by Recall Initiation Method .................................55 14. Histogram of Distribution Patterns for Software Related Recall Events ...........................56 15. Anderson-Darling Normality Test for Time to Classification ...........................................58 16. Anderson-Darkly Normality Test for Time to Recall ........................................................59 17. Identifying Outliers with Box-Plot Graph for Time to Classification ...............................61 18. Box Plot for Time to Recall ...............................................................................................63 ix ABSTRACT In recent years, the number of medical device recalls for software defects has climbed steadily, with Class II recalls accounting for the majority of total voluntary recalls. Prolonging the time it takes between recall initiation and recall termination could cause significant risk of harm to patients. By implementing the Plan-Do-Study-Act framework to examine medical device recall data, this study utilized risk management tools to analyze risks and root causes associated with prolonged recall times for software related recalls. Through identification of root cause and risks, mitigation strategies are discussed to reduce time to recall. The results of the study indicate that recall events with prolonged recall durations are linked to organizations that struggle with recall management through regulatory compliance. Organizations can benefit from analyzing industry data to identify trending areas of risk within similar systems to mitigate impact of noncompliant policies that lead to prolonged time to recall. 1 CHAPTER 1 INTRODUCTION Background Medical devices are becoming more complex with the advancement of technology leading to many revolutionary products for people navigating the treatment or cure of disease. The Food, Drugs, and Cosmetics (FD&C) Act defines medical devices as follows: (1) an instrument, apparatus, implement, machine, contrivance, implant, in vitro reagent, or other similar or related article, including a component part or accessory which is: recognized in the official National Formulary, or the United States Pharmacopoeia, or any supplement to them, (2) intended for use in the diagnosis of disease or other conditions, or in the cure, mitigation, treatment, or prevention of disease, in man or other animals. or (3) intended to affect the structure or any function of the body of man or other animals, and which does not achieve its primary intended purposes through chemical action within or on the body of man or other animals and which is not dependent upon being metabolized for the achievement of any of its primary intended purposes. (FDA, 2018) Medical devices range in style from band-aids and contact lenses to complex surgical systems and pacemakers. The FDA’s subdivision Center for Devices and Radiological Health (CDRH) maintains a database of over 6,000 diverse types of medical devices that have been classified by the agency. As technology has become more incorporated into design, medical devices often combine hardware, software, and cybersecurity components into devices to provide innovative solution to all facets of healthcare. Software has become an important part of all products, 2 integrated widely into digital platforms that serve both medical and non-medical purposes (FDA, 2018). There are three types of software related to medical devices: (1) Software that is intended to be used for at least one medical purpose, that can perform these purposes without being part of a hardware medical device (FDA, 2018). (2) Software incorporated into a medical device that is an integral part of the medical device (FDA, 2018). (3) Software that is used in the maintenance or manufacture of a medical device (FDA, 2018). Industry will continue to see an increase in the use of the diverse types of software driven medical devices due to its range of use and diversity of design. Even regulatory agencies are preparing to take on new technological advancements in medical devices. According to a speech from the FDA’s Commissioner of Food & Drugs (Gottlieb, 2018), even the FDA is making changes to the regulatory landscape for inclusion of software driven devices and digital tools. Digital tools would include wearable devices, telemedicine, telehealth, personalized medicine, health information technology, and mobile health applications. The emerging technology of digital health products has created a new lane of innovation in the medical device space combining software connectivity and consumer technology. Gottlieb highlighted the FDA’s current thoughts on the status of the agency’s perspective of digital health tools and their power to diagnose and treat diseases. In 2018, the FDA unveiled a new initiative for digital technology applications, including the following: (1) A policy to streamline FDA’s oversight of regulated components multi-function digital health products. (2) Industry Guidance that explains the regulatory approach and policy of the agency towards digital health products and provides clarification for software functions that are within scope of the FDA’s authority and those functions which are not. Functions that are within scope are treated as medical devices. 3 (3) Enhancements to the FDA’s Pre-Certification Pilot programs initiated in 2017. The pilot program allows organization to register digital health tools for FDA’s oversight to ensure the safety and effectiveness, and performance of health software. Digital tools registered within the pilot will be treated as medical devices. The FDA’s oversight of software also includes changing the focus to utilize digital health technology as part of drug development recognizing the benefits and opportunities to improve the effectiveness and safety of drug delivery systems. The agency’s goal is to create policies that will ensure that software driven technologies can be validated once incorporated into drug development systems (Gottlieb, 2018). The benefit of combining software with drug development is related to increasing the effectiveness of patient care. Technology provides the patient and provider an additional resource to help confirm that patients are taking medication, tracking and monitoring side effects, and ensuring relevant data is included within the electronic health record. The program updates from the agency highlights the necessity of ensuring adequate evaluation and analysis of software related industry trends, such as recalls, to understand the scope of digital expansion, and the potential risks exposed to patients when medical device software fails during performance. Medical device recalls protect public health by removing the defective and potentially harmful software or medical device from the market. Current State of Industry Recalls A recent analysis published by Stericycle in Q1 2018 reported that the number of medical device recalls in United States increased 126% from the previous quarter and involved more than 208 million units (Stericycle, 2018). According to a review of Stericycle Recall Index, reports conducted by the Association for the Advancement of Medical Instrumentation (AAMI), software recalls topped the recall charts for eight consecutive quarters prior to the issuance of the 2018 analysis (AAMI, 2018). The same report form, Stericycle, in Quarter 2, 2019 showed consistency, and reports that medical devices continue to lead as the top cause of industry recalls 4 at 49 recalls, followed by 32 for mislabeling, 31 for quality issues, and 18 for out of specification (Stericycle Expert Solutions, 2019). The Stericycle Index provides a quarterly industry analysis on recall trends according to their published reports. Software has been listed as the top reason for medical device recalls since 2016. Software defects in the medical devices are not rare and have the potential to negatively influence medical care (Ronquillo & Zuckerman, 2017). An increase in software related recalls can reveal that patients could be at a higher risk from medical errors caused by software defects. The analysis additionally reported that Class I (serious risk of harm or death) recalls are on the rise for medical devices, but Class II (moderate risk of harm) recalls remain consistent as the most abundant classification of medical device recalls reported to FDA. Medical device firms are responsible for ensuring that the organization is meeting regulatory requirements for a developed recall process. Such a process considers the severity of impact to patient safety when a defective product stays on-market. Products can be recalled when it is found to be defective, either when it falls out of design specification or fails to perform as intended. FDA Regulation and Enforcement The FDA governs medical device recalls under the Federal Code of Regulations, Title 21 Part 806 published in the Federal Register. Under 21 CFR Part 806 entitled medical devices; Reports of Corrections and Removals, firms can follow the provisions to promptly report actions pertaining to recalls of on-market products. The regulation includes provisions for corrective actions, market withdrawals, routine servicing, and stock recovery as identified by the United States (US) (FDA, 2018). Under Part B of the regulation, Section 806.10 provides the requirements for reporting and records in detail. Each manufacturer or importer shall submit a 5 written report to FDA of any correction or removal of a device initiated by such manufacturer or importer if the correction or removal was initiated (FDA, 2018). This action is completed for two primary reasons 1) to alert the FDA of a potential safety hazard to consumers, and 2) to allow for the firm to remedy the nonconformity that causes the health risk to the patient either through correction or removal. The regulation describes the required information that should be included within the report including: • Manufacturer information; • Product brand and classification; • Marketing status; • Unique Device Identifier; • Description of event; • Reporting of any known illnesses or adverse reactions related to device use. Recalls can be initiated voluntarily or involuntarily and notified to the FDA depending on the scope and severity of the recall and compliance to the regulations. Due to increased enforcement by the FDA, many medical device recalls are initiated voluntarily by the manufacturing organization. Overall, the FDA’s focus on industry correction and removal compliance has contributed to a 50% increase in the annual number of voluntary recalls reported since 2009 (CDRH, 2018). This can be attributed to a new FDA focus on identifying reporting deficiencies for medical device under 21 CFR Part 806 during routine compliance inspections. According to a recent enforcement report published by the CDRH in November 2018, firms are 8 times more likely to report a recall in the year after inspection, than the entire average for industry (CDRH, 2018). The increasing trend can serve as a safety net for patients and consumers as additional security allows monitoring and enforcement more visibility to hazardous 6 defects. A recall due to safety hazard occurs because one of three main triggers (Wood, Wang, Olesen, & Reiners, 2017). The first trigger happens when the firm detects a safety hazard after the discovery of a product defect. The second trigger stems from the Customer Feedback loop when patients or Health Care Providers inform the firm of a hazardous product. The third trigger stems from patients, competitors, or other consumers report a potentially hazardous product to the regulatory agency. Figure 1 illustrates these 3 FDA triggers that lead to organizational product recall. This prompts investigation by the firm and the agency to investigate the cause of the defect and initiate actions to eliminate the hazard or recover the product. Customer Feedback Safety Hazard Report of Hazard to FDA Recall Initiation Figure 1. FDA triggers that lead to organizational product recall. Figure created by the author of this thesis. Types of Recalls Recalls occur when the product is found to be defective causing harm to patient safety or a deterioration to product quality and performance. The FDA uses the term recall when a manufacturer takes a correction or a removal action to address a problem with a medical device that violates FDA law. Recalls occur when a medical device is defective, when it could be a risk 7 to health, or when it is both defective and a risk to health (FDA, 2018). Organizations have the option to perform a correction, by addressing the defect of the medical device, in the place the device is being sold. In instances of correction, the medical device may be returned to use once the devices have been assessed and fixed for known issues. The risks associated with use should be determined by the organization and communicated to customers. An alternative option is removing the product from the market where the medical device is sold. During removals, it is important to ensure that the defective device is fully recovered from the market to prevent any harm to patients by eliminating the opportunity for exposure. The FDA classifies recalls based on the impact to patient health and safety, there are three classifications for FDA recalls denoted as follows: • Class I (serious threat of harm) • Class II (moderate risk of harm) • Class III (unlikely risk of harm) Class I recalls are categorized based on the risk of serious adverse health consequences or even death upon exposure to the defective product. Class II recalls are initiated when a product may cause a reversible or temporary adverse health event, or there is a slight chance that exposure could cause serious injury or death of a patient. Class III recalls are initiated when the defective product is not likely to cause any serious health concerns or death of a patient. Once the FDA has decided on the classification of the recall, the agency monitors the recall progress to ensure that the appropriate action plans are in place to ensure effectiveness of product correction or removal (FDA, 2014). 8 Primary and Secondary Effects of Recall Impact The primary and most important effect of recall is its immediate impact to patient safety. Patient safety from the regulatory and firm perspective is to always protect the patient and reduce the likelihood and exposure to hazard and harm. Design and manufacturing controls build in hazard detection and harm reduction while product is still under the control of the organization. Following distribution, recall protocols should be designed and executed to keep the patient from using known defective products and cause an adverse reaction or health hazard. Recalls triggered by safety concerns imply that the device poses a considerable risk to the patient. The recall announcement can serve as a benefit to patient because it advises discontinuation of the recalled device. For example, for an implantable device, if a recall is initiated it could mean that the patient could be subject to removal or re-insertion which could require surgery. Delays in recalling critical to care could cost a patient their life or cause significant injury. The primary impact is a positive benefit for patients and regulators who become aware of a potentially hazardous situation. A secondary effect of recall follows the public announcement and notification to consumers. Recall announcements can negatively impact patient and customer confidence and brand image. In the age of social media, news can go viral within an instant, compromising the public image of an organization. Recalls are becoming an increasingly common notification and may reduce public confidence in the reputation of manufacturers (Wood et al., 2017). Manufacturers of healthcare products must be in tune with how the effectiveness of the recall program impacts public perception on the ability of an organization to ensure its products are safe on-market. Figure 2 illustrates the relationship of primary and secondary recalls and their impact on recall initiation. Previous research discusses the reasons behind why a product should 9 be recalled. However, the recent spate of product recalls has shifted attention from why products are recalled to why is takes so long to recall a defective product that poses a potential safety hazard (Hora, Bapuji, & Roth, 2011). Recall Initiation Effects Primary Ensures patient safety Reduces exposure of risk to harm Secondary Patient and consumer confidence decreases Figure 2. Primary and secondary effects of organization initiation of product recalls. Figure created by the author of this thesis. Time to Recall One component of the recall process that has a significant impact to recall effectiveness and patient safety is the time to recall. Time to recall occurs in three different reference points in the recall process as follows: • Time between date of product launch, date first sold, and date of the recall announcement. • Time between issuing the recall following notification of product hazards by patients. • Time between the announcement of recall to the recovery of recalled product from on-market distributors and patients. The time it takes for firms to recall their products depends on the following factors: 1) their recall strategy (proactive vs. reactive); 2) type of defect (manufacturing defect vs. design flaw); and 3) the recalling firm (manufacturer, distributor, retailer) (Wowak & Boone, 2015). 10 This data supports the thinking that more complex medical devices, complex distribution, and supply chains will have a larger gap in recall timeliness. Statement of the Problem The medical device industry has seen a consistent increase in software related recalls in recent years according to the Stericycle Recall Index (Stericycle, 2018). By the first quarter, 2018 software related recalls accounted for 23% of all recalls and have been the leading factor in medical device recalls every quarter since 2016. According to Fu, Guo, Zhang, Jiang, and Sha (2017), one out of every three software based medical devices has been recalled due to defects and errors within the software itself. Software defects or coding errors could be difficult to detect prior to distribution, and thus more likely to cause harm to patient when they occur. Software defects can become critical for medical devices whose operational functionality relies on the effectiveness and adequacy of the software component. Once aware of a software defect, it becomes critical to patient safety to quickly cease distribution and work to recover the product from the market. Prolonging the time, it takes between recall initiation and recall termination and product recovery can lead to a higher risk of harm to patient. The timeliness of recall recovery from initiation to termination will have a significant impact to patient safety. When not adequately controlled, the following three factors will have a negative impact on the timeliness of recall recovery potentially taking more time to fully recover on-market products (1) product traceability, (2) recall protocols, and (3) distribution pattern. Purpose of Study and Significance The purpose of this study is to examine industry data to support current literature and industry thinking on the importance of recall effectiveness, timeliness, and its impact to patient 11 safety and product quality. This thesis will include an evaluation of the current industry state for medical device recalls and provide recommendations and considerations for the implementation of recall management systems based on these trends. The study of recall management has implications for manufacturers and their supply chain partners to ensure that they are following the regulatory requirements. A quality culture that includes full transparency and traceability within the supply chain is important to ensuring the success of multifaceted recalls even within the most complex systems. This thesis aims at using industry data to benchmark best strategies supported by literature on reduction of time to recall, and its success for an organization. The focus on software related medical device recalls has positive benefits to data trending and retrospective learning in industry. Many medical devices rely heavily on software systems as a critical component of functionality. Additionally, there are software systems that standalone as medical devices of their own without hardware systems. The tech-trends in industry have created more options for technology driven data applications, wearable and wireless technology, and smart medical devices. Quality professionals, senior leaders, and regulators must understand the importance of recognizing areas of risks to better understand how to investigate industry trends and recommend quality solutions to industry that address risk to device design, technology in manufacturing and distribution, and patient safety. Additionally, time to recall effectiveness has not been significantly studied for medical devices. Considering the implications medical device hazards have on patient safety, understanding the parameters around recall recovery is important to ensure adequate controls are in place. Time to recall metrics could demonstrate the impact and the ability to retrieve products from consumers to eliminate risk of harm. Organizations could use data for continuous 12 improvement opportunities in the recall process, but also within supply chain traceability practices. This study’s focus highlights the potential for process improvements, industry benchmarking, and reinforcing the continuous improvement lifecycle for software-related medical devices. This study aims to resolve inquiries into reducing time to recall duration, measuring effectiveness of recall strategy using data trends, and evaluate areas of risks and mitigation strategies to incorporate into a standard recall process. Theoretical Basis and Organization This study will conduct a retrospective analysis of industry data to provide clear recommendations to industry for strategies to reduce recall timeliness to reduce risk of harm to patient. Analysis of data and information involves multiple phases of planning, execution, evaluation, and conclusion to draw relevant perspective and encourage risk-based decision making. The Plan-Do-Check-Act (PDSA) theoretical framework allows for a continuous cycle of learning and improvement through multi-phase-based implementation. W. Edwards Deming (1982), in the book Out of The Crisis, introduced industry to the concept of process improvements occurring within a continuous cycle of learning. PDSA is a four-phase theoretical framework towards quality improvement that is used to identify a problem and plan a resolution, check the resolution in a small population, evaluate the results, and implement the resolution into the system. For this thesis, PDSA represents a process improvement framework that has been widely used to support organization learning from quality improvement processes. The PDSA framework can be used to repeatedly build on opportunities for improvement within a quality management system and will guide the identification and analysis of data to present findings for areas of improvement within the recall process: 13 • Plan phase. Identify gaps within the recall execution protocol that would identify areas of improvement for reduction of time to recall duration for software related recalls. Develop a plan for mitigation of risks associated with identified gaps. • Do phase. Carry out data analysis on recall data and develop the risk mitigation plan for improvement for reduction of time to recall duration for software related recalls. • Study phase. Evaluate of the results by understanding what can be learned from the data analysis. Understand how risk mitigation strategies are supported by the results and literature. • Act phase. Decide if the results lead to adopting risk mitigation strategies and updating protocols for recall execution if the results of the study phase are in alignment with the goal of the plan phase. If the results are not successful, this phase will consider additional cycling through the PDSA loop to gather new insights and learning. This thesis explores the use of the PDSA learning cycle in identifying risk mitigation pathways to reduce the time between FDA recall initiation and product recovery by examining potential risk factors associated with delays in total time to recall. Risk mitigation and process mapping strategies are explored to take advantage of the FDA data sets to shorten the time between (1) FDA recall initiation and (2) FDA recall termination and product recovery to reduce opportunities for risk to patients. Additionally, this thesis will use the results of data analysis and risk mitigation, supported by the literature to propose a protocol for recall execution, and data trending based on the evaluation of recall triggers found using common cause codes within the data set. Scope and Limitations For the purpose of this thesis, the study was limited to Class II medical device recalls reported to the FDA between 2016 through the end of 2018. Additionally, the FDA enforcement report data only includes recalls that have been determined by FDA as classified recalls and is updated as recalls evaluated to meet requirements. The FDA publishes recalls as information is classified and made available, but the number of recalls featured in these lists differ significantly 14 from the numbers shown in FDA's medical device recalls database, as some Class I recalls are not included in the annual lists and some of the entries on the list account for more than one database entry when the recall affects multiple models made by a company (Mezher, 2018). This duplication or lack of addition of data will not account for 100% of the recall population within the reporting period but will include all data that the FDA has made publicly available. The data set will be even further restricted following categorical coding to identify and analyze records specific to software related recalls. The study will additionally restrict the scope of distribution patterns reported in the enforcement report in order to analyze larger subgroups such as national and international to account for multiple variables that could have an impact on the measurable metrics. Recalls without termination dates will not be analyzed within the context of this study, as open recalls are out of the scope of time to recall calculations. Additionally, following data analysis, the selection quality tools utilized, and risk mitigation strategies will serve as a recommendation for recall management supported by current literature. Definition of Terms Association for the Advancement of Medical Instrumentation (AAMI): a non-profit organization, devoted to the management, development, safety, and effectiveness of health technology. AAMI is a source of medical device industry knowledge, and support for healthcare industry professionals (AAMI, 2019). Authorized Representatives: In the context of the European Medical Device Regulation (E.U. MDR), any natural or legal person established within the Union who has received and accepted a written mandate from a manufacturer, located outside the Union, to act on the manufacturer's behalf in relation to specified tasks with regard to the latter's obligations under the Medical Device Regulation 2017/745 (European Parliament, 2017). 15 Center for Devices and Radiological Health (CDRH): a subdivision of the FDA, responsible for protecting public health, by facilitating medical device innovation through enhancements to regulatory science. The CDRH provides industry with guidance for regulatory compliance to ensure the safety and effectiveness of medical devices (FDA, 2019). Class I Recall: a situation in which there is a reasonable probability that the use of or exposure to a violative product will cause serious adverse health consequences or death (FDA, 2017). Class II Recall: a situation in which use of or exposure to a violative product may cause temporary or medically reversible adverse health consequences or where the probability of serious adverse health consequences is remote (FDA, 2017). Class III Recall: a situation in which use of or exposure to a violative product is not likely to cause adverse health consequences (FDA, 2017). Code of Federal Regulations (CFR): is an annual codification of the rules, both general and permanent, published in the Federal Register by the agencies of the federal government and their executive departments (National Archives and Records Administration, 2019). Digital Health Tool: A scope of devices and technologies designed to better manage and track health and wellness related activities to improve patient health (FDA, 2019). Distributor: Organization responsible for distribution of medical devices from the manufacturer to consumers and/or consumer markets. Distributors: In the context of the E.U. MDR, any natural or legal person in the supply chain, other than the manufacturer or the importer, that makes a device available on the market, up until the point of putting into service (European Parliament, 2017). Economic Operators: In the context of the E.U. MDR, means a manufacturer, an authorized representative, an importer, a distributor (European Parliament, 2017). 16 Effectiveness Check: During the effectiveness check, the FDA works to evaluate if all reasonable efforts that have been made to remove a product from the market. Once the recall is complete, the FDA will also investigate the reason a product was found defective and the adequacy of product destruction or recondition (FDA, 2010). Enforcement Report: A library of medical device recalls monitored and classified by the FDA Firm: used interchangeably with Manufacturer, an organization responsible for producing, packing, and shipping medical devices. Food and Drug Administration (FDA): the federal agency that has the authority to enforce regulation for Medical Devices, Cosmetics, Food, Drugs, and Animal Products. Harm: hurt or mischief, physical damage or mental damage caused to a person. Hazard: a source of danger, or an effect of harm. Health Canada (HC): Health Canada is the regulatory body that governs and enforces regulatory requirements in Canada. Importers: In the context of the E.U. MDR, any natural or legal person established within the Union that places a device from a third country on the Union market (European Parliament, 2017). International Organization for Standardization (ISO): is an international organization that serves as a global network of national standard bodies. ISO is responsible for developing and publishing international standards for a variety of industries including medical devices (ISO, 2019). Outside of the United States (oUS): oUS is used within the context of this thesis to describe geographical locations that are outside of U.S. mainland and individual territories such as Puerto Rico, Guam, Virgin Islands, Hawaii, and any other U.S. territories. The acronym oUS categorizes all international locations outside of the United States. 17 Recall Enterprise System (RES): An electronic system utilized by the FDA for the management of recall data. Responsible personnel use the RES to submit and update information about recall classification, and status. Recall Initiation: A recall is initiated when the organization starts to notify the public or their direct accounts of the recall (FDA, 2017). Recall Termination: A recall is terminated when the FDA determines that all reasonable efforts have been made to recall or correct the product to the degree of the hazard, and that all corrective actions were deemed appropriate. Recall termination is requested by the organization with objective evidence submitted to the FDA (FDA, 2017). Serious Injury: means an injury or illness that: (1) is life threatening, (2) results in permanent impairment of a body function or permanent damage to a body structure, or (3) necessitates medical or surgical intervention to preclude permanent impairment of a body function or permanent damage to a body structure. Permanent means irreversible impairment or damage to a body or to a body structure or function, excluding trivial impairment or damage (FDA, 2018). Therapeutic Good: The Therapeutic Good Administration defines a Therapeutic Good as a broad range of products for use in humans in connection with: (1) preventing, diagnosing, curing or alleviating a disease, ailment, defect or injury; (2) influencing inhibiting or modifying a physiological process; (3) testing the susceptibility of persons to a disease or ailment influencing, controlling or preventing conception; (4) testing for pregnancy (TGA, 2017). Therapeutic Goods Administration (TGA): TGA is the regulatory body that governs therapeutic goods in Australia. 18 Unique Device Identification (UDI): a coding system used to ensure traceability and adequate identification of medical devices. It includes the assignment of a Unique Device Identifier to each device in order to improve safety and post market surveillance (FDA, 2019). Wearable Technology: Wearable technology are multi-function consumer devices designed to track and provide data related to health and fitness. Wearable technology can include mobile application as a component of usability. The devices may contain motion sensors, monitors, and GPS. 19 CHAPTER 2 LITERATURE REVIEW The literature review presented the opportunity to develop a clear and concise understanding of applicable regulatory requirements, guidelines for industry, and industry expectations of the roles and responsibility of the organization during the recall process. Current regulatory requirements and policies are used for the identification of key elements of a compliant recall process and assist with general understanding of a procedural baseline. The review of FDA enforcement actions provided context in connection to understanding vulnerabilities that exist in software driven medical devices and their impact to risk of harm. Sources reviewed included well-established and emerging regulations as well as guidance documents, international standards, and enforcement actions that signify the importance of recall controls and the need continuous improvements. Roles and Responsibilities: Recall Regulation and Guidance Ensuring the safety and effectiveness of medical devices is vital to human health, and the medical devices must be managed by strict regulations according to their assigned levels of risk. Current industry perspectives have changed the regulatory landscape of the organizations develop and distribute medical devices around the globe and medical device regulatory reform has been promoted by global harmonization and streamlining (Chen et al., 2018). Relevant international organizations have been established such as the International Medical Device Regulators Forum (IMDRF). Member countries of the IMDRF include Australia, Brazil, Canada, China, the European Union, Japan, Russia, Singapore, and the US. 20 Currently the IMDRF has discussed several topics for Harmonization including UDI implementation for traceability, and guidance for medical devices with software (Chen et al., 2018). Global harmonization aims to reduce regulatory differences worldwide, eliminate immoderate or country-specific requirements, and build up a consistent and transparent international regulatory management system (Chen et al., 2018). Given the global context of the supply chain, many organizations distribute medical devices on a global scale, impacting the scope of regulatory responsibility for recalls. This literature review provides a comprehensive overview of regulatory requirements within key global regions to provide insights on the similarities and differences between regulatory enforcement impacting the recall process for global organizations. The quality system regulation that governs and outlines industry responsibilities for recalls is managed under 21 CFR Part 7 Enforcement Policy (FDA, 2019). The regulation is divided into five subparts labeled A-E; subpart A covering the scope of the regulation and relevant definitions, subpart B and D, which are reserved for text, and subpart C covering recalls, subpart E covering criminal violations. Subpart C provides detailed requirements for recall policy, health hazard evaluation and recall classification, recall strategy, recall communications, notifications, status reports, and termination of recall (FDA, 2019). Subpart C also includes provisions for roles and responsibilities of the FDA and the organization during the recall process from initiation to termination. Industry guidance has been written and published by the FDA to provide additional support to organizational understanding of the FDA’s current expectations and interpretation of 21 CFR Part 7. According to the current regulation, the organization, either medical device manufacturer or distributor may initiate a voluntary recall at any time by issuing a communication about the 21 voluntary recall directly to consumers or to the public. When an organization initiates a recall, they notify the FDA and submit all information as required by regulation. The organization may notify the of a recall FDA by phone, letter, email, or a combination of communication practices when the recall is voluntary (FDA, 2019). Following initial notice to the FDA, the organization is responsible for the following elements of the recall as show in Figure 3. 1) Create a recall strategy detailing the course of action that needs to be taken to properly conduct a recall including the scope of the recall, issuance of public notification, and measuring effectiveness checks during the recall. The strategy also includes a risk assessment of the health hazard. Recall root cause analysis must be conducted and communicated to FDA. 2) Supply information to the users identify the product and include information for users to minimize the consequences of exposure. 3) Ensuring that establishes producers for implementing corrective actions are in place and effective to aid leaders and responsible parties in ensuring that corrective and preventive measures are in place. 4) Submit status reports to the FDA District office on the progress of the progress of the recall. Once actions and implementation are complete, the organization must submit a recommendation for termination of recall. Status reports will discontinue at FDA recall termination. 22 Organization Process begins. FDA Notifies consumers and direct accounts Organization notifies FDA of voluntary recall Initiate Recall in RES Creates Recall Strategy Organization submits recall strategy to FDA for review Review Recall Strategy T1 Time to Recall Corrective Action & status reports Continuous monitoring cycle Organization requests recall termination Monitoring & Audit of Recall Termination T2 Process ends. Figure 3. Flow diagram representing an overview of roles and responsibilities during recall lifecycle for organization and FDA. Figure created by the author of this thesis. Once the FDA receives notice of recall initiation, the FDA is responsible for the following various aspects of the recall process as defined in 21 CFR Part 7 Subpart C. FDA roles and responsibilities displayed in Figure 3 are defined as followed: 1) Initiation of the recall to the Recall Enterprise System within 24 hours of notification. 2) Determine if the action is a recall by determining if the action is in alignment with the definition of a recall per regulation. 3) Review the organization’s recall strategy, classifies recall, and assesses the health hazard posed by the product defect or non-conformity. 23 4) Following review of the firm’s recall strategy, the FDA issues a public announcement of the recall and notifies the firm of the classification of the recall (i.e., Class I, II, or III). 5) The FDA continuously monitors the recall and develops a recall audit strategy to evaluate the effectiveness of the recall strategy. 6) Recall termination is assigned when the FDA determines that all reasonable efforts were made to correct or remove the defective product according to the recall strategy and that all defective product has been removed with appropriate disposition. The regulation set by the FDA CFR provides a basic template for the required elements of a recall process that meets the requirements of the FDA regulation. Given the global context of product movement and traceability in today’s current context, many organizations develop recall programs that encompass traceability exercises that extend outside of the FDA’s jurisdiction. It is important that global organizations consider international standards and regulations to develop a comprehensive recall program that can meet the requirements of several regulatory authorities. Additional review into the requirements of other regulatory agencies provide similarities to the FDA requirements of recall program management. The International Organization for Standardization (ISO) published the latest revision of standards quality management systems (QMS) for medical devices in 2016. ISO 13485:2016 all requirements for the management of the QMS including requirements for management, measurement and analysis, product conformity, traceability, product realization, and nonconformances. As part of its comprehensive structure, ISO 13485:2016 covers requirements for recalls under the definition of advisory notices and adverse events (ISO, 2016). Per the standard definition, advisory notices are issued by the organization after the delivery of the medical device and includes the return of medical devices to the organization or the destruction of the medical device. 24 Within the standard, clauses 7.2.3 and 8.2.3 outline that the organization should have a plan and documented procedure to notify the agency of adverse events and communicate advisory notices to regulatory authorities. Additionally, clause 8.3.3 requires an organization to implement a procedure for reporting adverse events and issuance of advisory notices at any time as an action in response to nonconforming products. In summary, ISO provides standards for requiring organizations to establish a procedure for the notification and communication of adverse events to regulatory authorities, maintenance of records, traceability, and corrective and preventive actions. The Canadian regulatory authority Health Canada (HC) provides regulation to industry for the establishment of a recall process under the Canadian Medical Device Regulation (CMDR), specifically SOR/98-282 Food and Drugs Act (Government of Canada, 2019). The regulation includes general requirements for all quality system activities, but specifics pertaining to recalls includes provisions for distribution records and establishing a recall procedure outlining steps for how to recall product. Under the Canadian regulation, section 63 to 65.1 cover the specifics for the recall policy, requiring manufactures and importers are required to report recalls to the regulatory authority. According to the guidance document published by HC to assist in the interpretation of SOR/98-282, the recall process for HC is divided into five stages, recall initiation, recall strategy, notification and correction, follow-up, review and recall closure. Similar to the FDA’s requirements, following initiation of a product recall HC requires an evaluation of risk, timelines of action plan execution, recall communications, effectiveness checks, and notification to the authority of recall completion and request to closure (HC, 2016). 25 Therapeutic Goods Administration, the regulatory authority in Australia, develops and regulates policies for quality system management under the Therapeutic Goods Act 1989. The regulation covers provisions for medical device recalls under chapter 4, parts 4-9. When a product does not comply with requirements and cannot be lawfully supplied, the Secretary can request the initiation to recall. The regulation details provisions for recall initiation, public notification, product recovery, notification to the regulatory authority, and criminal penalties for violations to the act. In order to support industry understanding of the regulation, TGA published a Uniform Recall Procedure (URPTG), to provide a consistent approach for undertaking recall actions of therapeutic goods that have been made available to market through import into or export from Australia (TGA, 2019). The URPTG outlines the action steps to complete a recall outlining the requirements for immediate versus all other recalls. For immediate recalls of medical devices, Figure 4 represents the high-level process flow of recall initiation for immediate recalls under TGA authority. Immediate recalls are listed as Step 1 of the recall process because it is important to contact the Australian Recall Coordinator and customers immediately. The flow chart represents the order in which all concerned parties need to be contacted. The process starts at the highest level by identifying the type recall, denoted by the blue fields represented in Figure 4. Following identification of recall type, the process continues within Step 1 notifying customers and contact the Australian Recall coordinator of the recall event. Both forms of communication may occur simultaneously or in any order. Once appropriate contacts are notified, the recall strategy must be implemented and agreed to by relevant stakeholders. Following completion of this step, identified in the URPTG as Step 1, the recall process continues with Step 2 to Step 11. Step 2 is the initiation point for all 26 other recalls and flows through the closure of the recall process. The TGA URPTG lists the remainder of the recall process as follows: • Obtaining stock status and distribution status. • Conducting a risk analysis. • Deciding the type, class, and level of recall. • Developing a recall strategy. • Drafting a communication strategy. • Submitting recall information to TGA. • TGA Assessment of the recall, to provide advice and assistance in relation to letters, consumer recall notices, and recall strategy. • Implementing the recall. • Reporting on the recall via progress and status reports, including root cause analysis and corrective and preventive actions. • TGA will review the recall for verification of completion comparing action plans to objective evidence, determining satisfaction of recall implementation, and assesses the effectiveness of the recall. 27 Figure 4. Recall identification and initiation pathways for immediate recalls demonstrating Step 1 of Uniform Recall Procedure. Adapted from TGA. (2019, Feb 7). Uniform recall procedure for therapeutic goods (URPTG). Retrieved from Therapeutic Goods Act: https://www.tga.gov.au/publication/uniform-recall-procedure-therapeutic-goods-urptg The European Union recently published a new revision to the old Medical Device Directive in late 2017, now moving medical device regulation from directive to law under the new Medical Device Regulation (MDR) 2017/745. The massive overhaul of the former requirements has new implication for importers, distributors, and economic operators for roles and responsibilities to carry out a medical device recall. Under MDR 2017/745, a recall is any measure undertaken with the aim to achieve the return of a device that has already been made available to the end user (European Parliament, 2017). Medical device recall regulation is covered in Articles 94 and 95 of the regulation. Manufacturers of the medical device hold the responsibility to inform distributors, authorized representatives, and importers of a product recall once the product is found out of conformity. In situations where the product poses serious risk to health, manufacturers must immediately inform the competent authority of the non-compliance and any corrective action that is to be taken. 28 Importers are responsible for maintaining a list of product recalls that have been initiated, and providing manufacturers, distributors, and authorized representatives of any product information requested of them. Table 1 illustrates the responsibilities for each concerned party, either manufacturer, distributor, or importer holds the responsibly of notifying the competent authority of recall initiation if a nonconformity poses health risks to patients. Additionally, each party is responsible for contributing to the execution and closure of corrective actions associated with medical device recalls. Table 1 Communication responsibilities for recall notification and corrective action implementation for responsible parties under E.U. MDR 2017/745 Action in Recall Process Manufacturer Initiate recall X Notification to Competent Authority of hazard product X Competent Authority Importer Distributor X X X X X X X X Notification to manufacturer Notification to Authorized Representative X Notification to Importer X Notification to Distributor X Notification to notified body X X Corrective Action planning and implementation X X Restrict or withdraw product from market Maintain register of all recalls Note. Table created bt the author of this thesis. X X X X X 29 There are key similarities between global regulations and standards in terms of the requirements and elements of a basic recall process. As many organizations are operating on a global scale, it is necessary to develop a truly comprehensive recall process that broadens the reach to include critical components of the regulatory authorities governing the markets of their distribution channels. It becomes clear that the regulators see the recall process as an opportunity to set-standards in resolving nonconforming product defects. Notification to the regulatory is a key step in ensuring public safety and having support in adequately addressing recall strategy execution. Figure 5 provides a summary of expectations for a basic recall process based on the compilation of regulatory requirements from the US, Australia, Canada, European Union, and ISO. The basic understanding of the recall process will be relevant to this study to provide an understanding of the baseline for recalls. Using this baseline, this study provides analysis for opportunities for improvements where gaps in standard understanding based on basic requirements exist. 30 •Issue a notification to direct accounts, consumers, and regulatory authorities once it has been identified that product is hazardous •Notify distributors, importers, manufacturers as applicable Initiation and Notification •Evaluate product for hazard risk and risk of harm to patient. Strategy and Corrective Actions •Develop a recall strategy including distribution status of product, and stock recovery plan. •Review strategy with regulatory authority for feedback and guidance. Ensure that controls include root cause analysis, and corrective actions to address the root cause. •Define preventive action plan to reduce reoccurrence of root cause •Implement recall stragey and product recovery plan. •Submit periodic updates to regulatory authorities including feedback of all responsible parties (distributors, importers, and manuacturers as applicable). •Request verification from regulatory authority upon completion of corrective actions, Effectiveness preventive actions, and product recovery. and Closure •Recall Closure Figure 5. Summary of basic requirements of recall process combining regulatory requirements of the United States (FDA), Australia (TGA), ISO, Europe (EU), and Canada (HC). Figure created by the author of this thesis. FDA Enforcement Actions for Software Defects and Recalls In 2018, Phillips Medical Systems (PMS), a medical device manufacturer, received a Notice of Inspectional Observations FDA Form 483 following an FDA inspection in July 2017 at the manufacturing facility in Cleveland, Ohio. The results of the inspection revealed that that the organization lacked adequate controls for handling and processing thousands of customer complaints, in addition to various other nonconformities, that lead to more than 20 different product recalls. The month-long investigation exposed that over 129,000 complaints had been closed based solely on pre-defined hazard codes without requiring further investigation. During the investigation, the FDA also found that a corrective action plan (CAPA) was opened in January 2014 for software anomalies was closed without further investigation even though 59 customer complaints were related to the same issue noted in the nonconformance (FDA, 2017). The Form 483 reports that between February 2014 to the time of inspection, PMS had issued 30 31 Class II recalls related to software defects where the cause was contributed to software design controls. Many of those software related recalls were linked to a single supplier that was identified as on probation between 2014 and 2016. The company was additionally cited for failure to initiate any corrective action to identify the problem with software design controls to ensure that testing is adequate to ensure defects are detected prior to market release. Further inquiry revealed that the citation was a repeat observation from the FDA, and that every version of their software has been recalled. As of 2018, Phillips is still in communication with the FDA on the status of their progress towards addressing software errors as a result of the inspectional findings (Mulero, 2018). Warning letters are issued by the FDA following the issuance of the 483 listing inspection findings. Every organization is allowed a period of time to respond to the FDA with their intended corrective action plans, and periodic status updates in 30 intervals. Once the progress updates are submitted, the FDA reviews the response from the organization for adequacy and effectiveness in addressing the citation. If the FDA finds that the actions are ineffective or fail to address the issue, the FDA issues a warning letter to the organization. Warning letters serve as a final notice from the FDA to correct the nonconformance before the issuance of a consent decree. In 2015, Merge Healthcare received a warning letter from the FDA following an inspection at their facility in Wisconsin. Merge Healthcare is a manufacturer of software used in clinical settings to manage patient data for medical images and patient vitals. The organization received a 483 following the inspection noting violations in software validation and management of software related recalls. For software validation, the citation was issued because the organization’s process allowed for the shipment of devices prior to full completion of software 32 validation activities (Public Health Service, 2015). Furthermore, the organization was issued a citation for failure to notify the FDA of two Class II software related recalls as per the regulatory requirements of 21 CFR 806.10 Corrections and Removals. According to the warning letter, the organization has not provided the FDA with a plan on how to address the determination of the defects risk to health and the need to report the removal or correction to the agency. Additionally, no timeline was provided for the completion of recall activities. Following response to the warning letter from the organization, the FDA issued a close out letter confirming receipt of responses from the organization and confirming that the organization has addressed the concerns raised in the warning letter. Between 2018 and 2019, the FDA issued public notices for several software related Class I (high risk) recalls. In August 2018, Akron anesthesia delivery systems recalled their delivery system devices due to a software defect that caused the device to go into a failed state while the machine was in use (FDA, 2018). During the failed state, the device stops the mechanical ventilation system from functioning resulting in an increased risk of harm to patient. In September 2018, Medtronic recalled their Synergy Cranial Software and Stealth Cranial Station S7 software due to inaccuracies displayed during surgical procedures. The software is intended to be used in surgery to produced three-dimensional (3D) images of the patient’s brain to allow surgeons to effectively navigate surgical tools and implants during surgery (FDA, 2019). The software was recalled due to reports to the organization of incorrect information appearing on the display during biopsy procedures. The defect was noted to cause serious to lifethreatening risks of harm to the patient. In April 2019, Brain Lab AG recalled their Brain Spine and Trauma 3D Navigation Software due to the potential for incorrect information to appear on the display during surgery 33 (FDA, 2019). The defect would prevent the surgeon from being able to accurately navigate surgical tools inside the patient potentially resulting in brain damage, or serious to lifethreatening injury to the patient. The Class I examples show the dangers and severity of undetected software defects and their impacts on the risk of harm to patient. The public notifications provide consumers, hospitals, and health care practitioners with relevant product information, information of the defect, potential impacts to patient safety, and instructions on what to do if the device in their possession is affected by the recall. Summary of Enforcement Actions Circumstances leading to the issuance of warning letters, inspection citations, and public recall notices provide direct evidence of the necessity to maintain controls for the recall process to ensure that patient safety is always at the forefront of the organizations focus once defective product is known on the market. The combination of these various modes of enforcement provide insights to the vast scope of issues that can trigger recall, but also highlight inadequacies in the recall process managed by an organization. Customer complaints trends, established protocols and procedures, software validation during design are all aspects of the product, and quality system lifecycle that have the potential to impact the product once made available to the market. It becomes critical to regulators to understand the scope of their responsibility in keeping industry on track with compliance. It is equally important for organizations to understand the varying components within their systems that may impact the likelihood of recall, and the effectiveness of recall execution. Recall Timing and Effectiveness Studies In 2011, Hora et al. conducted an empirical study to identify and test three key factors that can be associated with the time to recall a product from the date the product is sold until 34 recall initiation. The primary focus of the study was to use those key factors to explain why it takes so long to recall a product that poses a safety hazard. The study focused its in depth analysis on the relationship between time to recall against three critical characteristics (1) recall strategies, (2) source of the defect, and (3) supply chain position of the recalling firm. Recall strategy was involves the initial decision to voluntarily or involuntarily recall a product, and whether the strategy to execute the recall is preventive or reactive. Hora et al. (2011) discussed the impacts of recall strategy for preventive vs. reactive approached to recalls and their impact in recall timelines. In a preventive recall strategy, the organization elects to notify customers of initiation of recall based on the organizations ability to identify a defect. Signaling to customers that the organization is diligent and proactive about quality issues. Reactive strategy, incidents ,and injury have already been publicly reported and the organization is responding by initiating a recall. Source of defect relates to the point in the process the defect surfaces either in manufacturing and production, design flaws, or inadequate instructions and warnings. The study evaluates which sources of defect would cause increases in time to recall duration for the toy industry. Lastly, the final characteristic is supply chain position relating to whether the product return is handled by the manufacturer or distributor and where in the supply chain the product might be between the manufacturers and the end user. Supply chain position evaluates the logistical aspect of product movement and estimates implications on its impact to time to recall duration. Time to recall was used as the dependent variable in the study while the critical characteristics (1) recall strategies, (2) source of the defect, and (3) supply chain position of the recalling firm represented the independent variables. Using statistical analysis, the researchers conducted the experiment using correlations between all variables. The study results in a 35 confirmation of the original hypothesis for each characteristic. Time to recall was longer for product recalls that utilized preventive strategy over reactive. Time to recall was longer for product design flaws than manufacturing defects. Time to recall was longer for supply chain players that had closer proximity to the end customer than those that do not. Although this study was conducted on the toy industry, which has an established set of regulations that are vastly different in comparison to medical devices, the study has strong implications for the context of this research. This study is unique as its available research on time to recall evaluation is limited in scope across many industries, including medical devices. Similarities exist between the selection of data set, Hora et al. (2011) conducted this evaluation using available industry data on toy recalls and considered recall strategy as an aspect that will impact time to recall duration. Differences exist not only for recall context, but for additional characteristics as well. The explored medical device recalls, between a shorter period, and the time to recall parameter is set from recall initiation to recall termination. The Hora et al. (2011) study does not take recall termination into account in the study, limiting recall analysis from product availability date or date first sold to recall initiation. Food recalls are used as an effective and preventive tool to limit exposure to illness by removing contaminated product from the market. Yu et al. (2018), conducted a research study to evaluate recall effectiveness indicators in the US meat and poultry industry in order to offer a framework of relationships between (1) discovery time, (2) completion time, and (3) recovery rate. Using three indicators, the research aimed to verify the effectiveness of the recall in relation to recall timing and product recovery. Discovery time represents the number of days from earliest production until the defect is discovered and recall is initiated. Completion time represents the time between opening and 36 closing of the recall. Recovery rate is the proportion of recalled product recovered by the organization (Yu & Hooker, 2019). The underlying assumption of the study is that these three indicators are clear measures of recall effectiveness. The researchers hypothesized the following: • Recalls with shorter discovery times have higher recovery rates than those with longer discovery times. • Recalls with longer completion times have higher recovery rates in contrast to those with shorter completion times. • Recalls with higher recovery rates have longer completion times than those with longer recovery rates. Yu et al. (2018) utilizes data made publicly available, and conducts the analysis using descriptive statistics for continuous vs binary variables. The research results in confirmation of the hypothesis supporting that recalls with shorter discovery times and longer completion times have higher recovery rates (Yu & Hooker, 2019). This encourages the theory that recall effectiveness is directly associated with both completion and recovery times. Additional implications were made during the discussion highlighting that the results could be attributed to having better technological resources for product recall and recovery allowing for more effective recall management. The research study is similar to the scope of this thesis as both will use publicly available information for data analysis to evaluate recall effectiveness and product recovery. The timing parameters vary as this thesis will evaluate time to recall between initiation and termination of recall, against distribution pattern, recall protocols, and product traceability. Plan-Do-Study-Act (PDSA) The PDSA learning cycle provides a theoretical framework for iterative testing to implement, evaluate, and measure changes within a quality system structure (Taylor et al., 2014). The PDSA learning cycle has been used industry wide in a variety of applications from manufacturing to healthcare improvements. In suggesting improvements to processes based on 37 established regulations, it is important to understand the context of historical data, and the impact the data has had on the quality of the medical device and the safety of the patient. The PDSA learning cycle is utilized within this thesis, to drive improvements to the recall process based on a risk analysis of historical recall data in terms of time to recall duration and effectiveness. According to Deming (1982), the reason to study the results of change is to try and learn how to improve tomorrow’s product. The PDSA learning cycle requires careful planning and the development of predictions based on assumptions, testing the plan, studying the results, and evaluating the impact of the change. Studying the results involves evaluating what we learned from the change. Did the changes lead to improvements in the quality system, the process, or the product? In the case of this thesis, does the change lead to lower risk of harm to patients. The PDSA learning cycle allows for an adoption or a rejection of change based on the results, when results are favorable, the PDSA learning cycle can be repeated under different conditions to learn whether favorable results are false or are valid over a range of different conditions. Taylor et al. (2014) adapted a theoretical framework for the PDSA learning cycle highlighted by Figure 6 representing the four stages of the PDSA learning cycle, and key critical considerations at each phase. Limited research has been conducted utilizing PDSA for quality improvements within quality system functions, specifically recall analysis, however, PDSA has been utilized to support research in quality improvements associated with patient care practices. Furthermore, Wowak et al. (2015) noted that although the literature on product recalls is informative, our understanding of recall events is still in its infancy. 38 Figure 6. The model for improvement. Adapted from Taylor, M. J., Mcnicholas, C., Nicolay, C., Darzi, A., Bell, D., & Reed, J. E. (2014). Systematic review of the application of the plan–do study–act method to improve quality in healthcare. BMJ Quality & Safety, 290-298. Summary of Literature Review The literature review highlights the need for ensure compliance to global regulation while considering areas of weakness within the recall communication chain. To identify the underlying factors that contribute to an organization’s inability to reduce time to recall, it’s important to understand the entire recall process. The current execution of this process is based on what is required versus areas of flexibility between the perspective of the organization and regulatory authorities. Industry enforcement data creates a communication channel between regulators and organizations that identifies quality system areas of noncompliance that suggest opportunities for organizational learning and improvement related to recall management. Review of the regulatory 39 requirements against enforcement actions when systems do not meet regulation demonstrates how regulatory compliance gaps influence prolonged time to recall. 40 CHAPTER 3 METHODOLOGY Design of the Investigation This study utilized regulatory requirements, industry guidance, and international standards to establish the structure for process improvement as the scope of inquiry. The PDSA learning cycle was developed as the theoretical framework that served as the basis of investigation. Using the PDSA learning cycle, this thesis explored four phases of investigation execution to identify areas of risks discovered from historical data analysis, and to provide an improvement strategy for implementation, evaluation, and discovery. The PDSA learning cycle was executed as follows: • Plan Phase. Identified key characteristics of a basic recall process and predict areas where improvements might be possible based on literature. Created a flow-chart highlighting key characteristics and identifying proposed areas of improvements as demonstrated in Appendix A. Plan phase was included in Data Analysis Procedures. • Do Phase. Conducted data analysis on recall data set employing the use of quality tools such as Pareto charts, Fishbone diagrams, and box plots using coded data sets. The Pareto chart was developed to identify display all defect categories and highlight software defects, verifying software as one of the top causes of recalls. Developed Fishbone diagram to represent root cause analysis to identify factors contributing to prolonged time to recall duration as demonstrated in Appendix G and H. The Anderson-Darling normality test, and the boxplots were presented to map the data results for software related recalls to confirm which recall events were found to have prolonged time to recall duration. Statistical plots were created based on analysis of software related recalls time to recall and time to classification data points, means, standard deviations, and control limits were calculated using Minitab®. Additionally, Failure Modes and Effects Analysis (FMEA) was developed to identify failure modes stemming from potential root causes that contribute to an increase in time to recall duration and a decrease in recall effectiveness. Do Phase was included in Data Analysis Procedures. • Study Phase. Presentation and discussion of results discovered during the Do Phase. Reported results of root cause identification and discussed confirmation of gaps from Plan Phase. The FMEA tool developed within the Do Phase, was updated to reflect 41 risk mitigation strategies for confirmed root causes and their failure modes ad demonstrated in Appendix I. Study phase was reported under Chapter 4, Results and Discussion. • Act Phase. Referenced literature to support the proposed implementation of risk mitigation strategies for process and product failure modes discussed during the Study Phase. Provided recommendations to industry to consider implementation of risk management strategies following implementation of proposed tool for recall data trending and risk mitigation for recall execution. Act phase was discussed under Chapter 4, Results and Discussion. Population or Sample The data utilized for analysis within this study included publicly available data that has previously been published within the FDA medical device recall database. Recall data collection was performed via request of recall data from the FDA using the Freedom of Information Act (FOIA) request from on the FDA website. Request response was provided by the Division of Freedom of Information in the format of an excel spreadsheet including a compilation of FDA reported recalls for Class I, II, and III medical device recalls initiated between 2016 through the end of 2018. The data set includes data related to the duration of recall, location of recall, distribution pattern, organization name, and address, product description, and reason for recall. All data was previously published for public notification and does not include any information related to human subjects. This data set encompassed all recalls to allow for application of Treatment procedures to the data set to identify software related recall events. Using stratified random sampling procedures, the initial data set of 9,647 records was truncated after treatment to represent 1,605 specific recall events. Those 1,605 recall events were reduced by defect category identify software related recalls, this treatment resulted in 282 recall events, detailed in Data Analysis Procedures. The final sample size is represented by the 282 software related recall events which 42 were used to develop the quality tools referenced under data analysis. Results were reported in Chapter 4. Data Collection Treatment This thesis provided an in-depth analysis of Class II medical device recalls filtered from the data provided by the FDA FOIA request. Of the Class II medical devices, this study analyzed software related recalls that have both initiation and termination dates. Recalls without termination dates were not analyzed through statistical analysis, as open recalls are out of the scope of time to recall calculations. Distribution patterns were coded according to global and national recalls as indicated within the dataset. Microsoft® ExcelTM and Minitab® software applications were utilized to perform statistical and graphical data analysis on the sample data. It is important to analyze data for identifiable characteristics and remove sensitive organizational information. The data set was coded to categorize key characteristics for statistical analysis within the control chart. The state of descriptive text within the data set does not allow for a clean review of data and coding will provide increased flexibility and provide a valuable benefit in facilitating the data analysis. Data coding by categorizing quantitative and qualitative responses converts the data for suitable investigation with the assistance of statistical software. The following columns of the FOIA data set provided by the FDA were modified using a numerical coding system: 1) Recall Event IDs were recoded from assigned FDA numbers to sequential numerical codes labeled beginning with the number one, ending at the last row of the data set. Duplicate entries were removed using Microsoft® ExcelTM data tools to code multiple entries for a single event. 2) Voluntary/Mandated responses were recoded numerically. Responses marked Voluntary: Firm Initiated were coded using the number one. Responses marked FDA Mandated will be coded using the number two. 43 3) Country responses were recoded initially by global region (i.e., U.S., Europe, Oceana, North America, East Asia, South America, and the Middle East) based on country location. Regions were recoded numerically starting with the number one, ending at the last row of the data set. 4) Recall Categories were reviewed manually for 100% of the dataset. Reason for recall responses will be recoded categorically by type of defect based on the description within the response. Categories for coding include the following: Labeling, Software, Component, Product, Process, Registration, and Unknown. 5) Distribution Pattern responses were reviewed manually for 100% of the dataset. Distribution pattern responses were recoded categorically by the geographical region identified. Categories for coding include the following: Domestic—US Only, International—outside US (oUS) only, Domestic & International, and Unknown. Regions will be recoded numerically starting with the number one, ending at the last row of the data set. Quality Tool Implementation Methods Several quality tools were used within the implementation of the PDSA learning cycle for data analysis. The first tool introduced within the study is a Pareto chart represented as a bar graph that represents frequency of recall defects within the data set. Figure 7 demonstrates an example Pareto chart for visual representation displaying six categories of defects, denoted numerically. The purpose of the Pareto chart was to visually demonstrate significant recall defect categories to highlight the frequency of software related recalls in comparison to other recall related defect by category within the same period. For this study, the resulting Pareto chart was constructed using recoded values for the reason for recall by category. The Pareto chart was configured using following application of data treatment method using Minitab® software application. 44 20 18 16 14 12 10 8 6 4 2 0 87% 96% 100% 76% 63% 39% 120% 100% 80% 60% 40% 20% 0% Category Category Category Category Category Category 5 3 1 4 2 6 Figure 7. Pareto chart demonstrating frequency of recall defects by category. Adapted from ASQ. (2019, September 17). THE 7 BASIC QUALITY TOOLS FOR PROCESS IMPROVEMENT. Retrieved from American Society for Quality: https://asq.org/quality resources/seven-basic-quality-tools The second tool introduced within the data analysis and PDSA implementation is the cause-and-effect diagram, otherwise known as the fishbone diagram. Figure 8 is provided as an example fishbone diagram to illustrate the identification of specific issues that contribute to software recalls. This quality tool was utilized to assist in identifying potential causes for software related recalls based on common themes that are noted as responses within the recall event data set. Additionally, the diagram assisted in identifying causes related to increased duration in time to recall following the initial analysis. The fishbone diagram was created manually within Microsoft® PowerPointTM. 45 Figure 8. Fishbone diagram template. Adapted from ASQ. (2019, September 17). THE 7 BASIC QUALITY TOOLS FOR PROCESS IMPROVEMENT. Retrieved from American Society for Quality: https://asq.org/quality-resources/seven-basic-quality-tools Several tools were used to identify trends and patterns within the software recall data set in order to understand characteristic variations in time to recall duration. This study utilized mathematical calculations to compute the time between initial time and final time based the following parameters. 1) Recall duration measured the timing between recall initiation to recall termination, measured in calendar days, as reported within the data set. 2) Recalls with multiple recall event codes were condensed into one recall record for count of total medical device recalls by organization. 3) Classification duration measured the timing between recall initiation to recall classification, measured in calendar days, as reported within the data set. The results of this calculation provided the numbers for the variables within the final data set used for statistical analysis. Statistical analysis of time to recall duration and time to classification duration from the 282 software related recall events should reveal inconsistencies 46 and deviations from the average that require further investigation for risk mitigation using FMEA. This study employed the Anderson-Darling normality test to determine if the data for time to recall and time to classification followed a normal distribution. Following the normality test, each variable will be analyzed for outliers using boxplots generated within Minitab®. Once the outliers are identified, this study will analyze each outlier for reason for recall and root cause of recall and recommend mitigation strategies using FMEA analysis. Finally, this study utilizes tools available for risk analysis using FMEA. FMEA is a risk approach that is utilized to identify failures within the recall process that contribute to delays in time to recall duration, and the initiation of software related recalls. Failure modes represent errors or gaps within the process that need improvement or monitoring in the context of this study. Effects analysis evaluates the impact of those failures to the process. Failures were categorized based on their risk which includes frequency, severity, and the ability to detect the issue. Figure 10 illustrates the basics included in a traditional FMEA. In terms of software related recalls, the failure modes were classified based on common themes found within the data set. The purpose of using the FMEA in this study to recommend actions to eliminate or reduce identifiable failures to increase their impact on time-to-recall duration, and software related recall initiation. FMEA developed the risk analysis strategies listed below: 1) Process Failure Mode and Effects Analysis (pFMEA) was utilized to identify known factors that contribute to the initiation of a Class II medical device recall. pFMEA was utilized to identify known factors that contribute to an increase of time to recall duration of a Class II medical device recall. 2) Product FMEA was utilized to identify known and predicted factors of software related medical device components that contribute to the initiation of a Class II medical device recall. 47 Figure 9. Example FMEA data table. Adapted from ASQ. (2019, September 17). THE 7 BASIC QUALITY TOOLS FOR PROCESS IMPROVEMENT. Retrieved from American Society for Quality: https://asq.org/quality-resources/seven-basic-quality-tools Data Analysis Procedures Plan Phase. Recall Process Proposal The literature review of regulatory requirements in Chapter 2 provided a general outline for the basic elements of a recall process. The key takeaways within the context of FDA recalls is to think globally as many organizations have adopted a global supply chain including third party manufacturers and distributors around the world. Combining the regulations led to the identification of three high-level phases of the recall process (1) initiation and notification, (2) recall strategy and corrective actions, and (3) effectiveness and closure. Each phase contains elements of the recall process that have been highlighted by at least one of the global agencies and combined to propose an outline that is inclusive of strict global requirements. Appendix A provides a detailed flow chart to explain the following elements of a recall process. Initiation and Notification begins with first identification of the hazardous medical device, and an immediate notification to direct accounts, consumers, and regulatory authorities as required (FDA, 2019). Additionally, the organization is responsible for notifying distributors, importers, and manufacturers taking into consideration third party manufacturers per the 48 applicable requirements under the medical device regulation in the EU (European Parliament, 2017). Once notifications have been issued, it is the responsibility of the organization to begin a hazard analysis for risk of harm to patient. This may be performed through investigation of previously identified hazards from the design controls process or working to identify new hazards and update the design FMEA. Recall strategy represents a critical area of focus for this thesis, as it could directly impact time to recall duration. The organization is responsible for the development of the recall strategy, and the regulatory authorities are responsible for reviewing and advising prior to implementation. Developing a recall strategy should include the following steps as included in Appendix A. 1) Data collection for status of device in stock in inventory, and collection of distribution history. 2) Development of a stock recovery plan based on the size of distribution and the scope of geographic location of distribution channels. 3) Identification of root cause that led to hazard and development of a corrective action plan to address root cause of occurrence. A preventive action plan should also be developed and implemented to prevent recurrence of the issue. 4) Recall strategy execution. The final phase of the recall process involves monitoring the recall through effectiveness checks, review with regulatory authorities through status updates about the recall recovery plan, and incorporation of feedback from all parties. It is important during the monitoring phase for the organization to hold status updates and reviews with distributors, importers, manufacturers to understand the progress of the investigation and ensure each element of the strategy is fully implemented. Upon completion of all activities within the plan, and relevant milestones, including records of recall completion are maintained, the organization must request final review 49 and closure with the regulatory authority. This step is vital to ensure that the recalled product has been fully recovered and corrective action plans have addressed the root cause. Once all activities are verified as complete, the recall is led to closure. The purpose of this study was to incorporate retrospective review of recall data for considerations in minimizing the risk of recall initiation, and to benchmark categorical defects to minimize risk of hazard and time to recall duration. This part of the proposal is not currently a regulatory requirement; however, it is listed within Appendix A as a proposed tool for inclusion within Phase 0 of the recall process. By identifying defects from recalls of similar devices, organizations can benefit in two ways: 1) Mitigate risks of known industry hazards by implementing new hazards into design FMEAs or product improvement FMEAs based on industry data. This practice can be supported using the guidance under ISO 14971—Risk Management of Medical Devices. 2) Gain better control of identification of root cause following recall initiation to decrease time to recall duration and minimize risk of harm to patient from prolonged product exposure. Review of the flow chart allowed for analysis of opportunities for areas of improvement with the recall process. Potential gaps could exist in three key areas (1) recall strategy errors, potentially suggesting that the organization could not adequate plan and execute an efficient recall strategy, (2) Delays in monitoring the effectiveness of the recall during strategy execution because this is the responsibility of the FDA suggesting process delays in reviewing milestones within the recall plan, and (3) Root cause identification delay, which could lead to classification delays and potentially slow down the speed of time to recall. Chapter 4 includes an overview of the results confirming the identification of process gaps. 50 Do Phase: Application of Treatments to Class II FDA Recalls Data provided by the FDA Division of Freedom of Information was condensed via manual coding to reduce the entire dataset from all medical device recalls to software related recalls within the two-year timeframe. The initial data set provided by FDA contained 9,647 records related to medical device recalls encompassing all recall events from 2016 to 2019. Implementation of treatments for data coding resulted in a remainder 1,605 recall events, this is the data that will be used to identify the specific number of software related recall events for data analysis. For the identification of software related recalls, and relevant parameters for measurement, treatments were fully implemented to reduce the data set. The analysis continued with applying the treatments discussed in Chapter 3 to the entire working data set. The following procedural steps were explained in detail to demonstrate the pathway for the identification of software related recalls. Recall initiation methods were analyzed for comparative analysis; voluntary/mandated responses were recorded numerically and represented graphically in Figure 11 to demonstrate the number of firm initiated versus FDA mandated recalls for Class II medical devices. According pie chart, 1,593 recall events of all Class II recall events were initiated by the organization voluntarily and 12 recall events were mandated by the FDA. Number one represents voluntary, firm-initiated recalls, and number two represents FDA mandated recalls in the figure below. 51 Figure 10. Pie Chart representing Recall Initiation Modes. Figure created by the author of this thesis. Recall initiation location was marked by Country responses within the initial data set representing the geographical location of the organization responsible for recall initiation to the FDA. It is not clear if the organizations represent manufacturers or distributors but both parties may be responsible for recall initiation. Treatment was applied to all responses, with initial coding marking individual countries by their regions (i.e., United States., Europe, Oceana, North America, East Asia, South America, and the Middle East). Additionally, regions were recorded numerically starting with the number one and ending at the last row of the dataset. Figure 12 represents a pie chart to visually display the geographical regional distribution of organizations initiating recalls within the two-year time frame. Categories are represented numerically as assigned during data coding as follows: • Category 1: 1455 recall events were initiated within the US. • Category 2: 110 recall events were initiated in Europe. • Category 3: 4 recall events were initiated in Oceana (i.e., Australia, New Zealand). • Category 4: 17 recall events were initiated in North America (i.e., Mexico, Canada). 52 • Category 5: 14 recall events were initiated in East Asia. • Category 6: No recalls were reported in South America. • Category 7: 5 recall events were initiated in the Middle East. Figure 11. Bar chart of recall initiation pattern by region. Figure created by the author of this thesis. Recall categories were assigned to the 1,605 recall events listed within the data following a 100% manual review of Reason for Recall responses reported to the FDA. Categories for reason for recall were listed by type of defect based on the wording reported to the FDA. Issues related to labeling (i.e., incorrect labeling, misinformation on labels, IFUs) were recoded as Labeling. Issues related to packaging (i.e., unit carton issues, breakage of seal, incorrect coding on outer packaging) were coded as Packaging. Issues related to product defect related to components or functionality issues were coded as Product or Component as applicable. Issues related to process (i.e., service, maintenance, manufacturing errors) we coded as Process. Issues related to design registration or misclassification based on regulation were coded as Registration. 53 Issues that have not been identified or undetermined by FDA, or not yet determined by the organization were coded as Unknown/other. Issues related to software were coded as Software as applicable. Figure 13 is a Pareto chart that visually represents the result of the manual coding of reason for recall by defect category. The Pareto graph displays the two largest defect categories as Component and Software, with recall events at 521 and 282 respectively. Identification of the leading categories of recall defects supports the necessity to focus efforts on software related recalls as they have one of the highest impacts to industry recall initiations. A summary of all categorized events was provided as follows: 1) 521 Component related recall events. 2) 282 Software related recall events. 3) 207 Labeling related recall events. 4) 207 Product related recall events. 5) 168 Packaging related recall events 6) 131 Design related recall events. 7) 69 Process related recall events. 8) 11 Registration related recall events. 9) 9 Unknown/other causes for recall. 54 Figure 12. Pareto chart displaying recall event defect category. Figure created by the author of this thesis. For software related recall events, the resulting sample size for data analysis includes 282 recall events identifiable software characteristics that led to software errors. This represents the second highest category of defects for all recalls initiated from 2016 through 2018. Relevant keywords used to identify software recalls include factors such as coding, algorithm, display errors, error messages, improper calculations, user interface, software errors, software versions, software updates, digital systems, data, modules, modalities, electronic systems, or workstations. Software recalls accounted for 17.57% of all recall events within the time frame between the beginning of 2016 to the end of 2018. Data Analysis of Software Recalls Software recalls were further analyzed to identify recall events that had prolonged durations from time to recall and time to classification, and the factors that potentially influenced 55 the delay. The resulting data consisted of tabulated, coded data for software recalls including region of initiation, distribution pattern, time to classification duration, and time to recall duration. Of the 282 software recall events, analysis found that in terms of recall initiation method 275 recall events or 97.5% were initiated by the organization voluntarily while 7 recall events or 2.5% were FDA mandated. Figure 14 provides a visual representation of this comparison for software recall events. Figure 13. Bar chart of software related recalls by recall initiation method. Figure created by the author of this thesis. A review of region initiation patterns show consistency between software related versus all recalls in terms of origination location of the recalling organization. Most recalls initiated were within the United States and its territories with a total of 252 recall events initiated. Europe leads second at 24, followed by East Asia with four recall events, and North America (i.e., Canada and Mexico) with two recall events. Distribution patterns were analyzed as shown in Figure 15, and the following patterns resulted: (1) 141 recalls were distributed domestically within the United States only, (2) 131 recalls were distributed in both the United States and 56 internationally, (3) Nine software related recalls were distributed internationally only, and (4) One recall distribution patterns were unknown at the time of recall termination. Figure 14. Histogram of Distribution Patterns for Software related recall events. Figure created by the author of this thesis. Statistical Analysis Parameters Time to recall duration and recall classification duration times were calculated using standard formulas. Time to recall is measured as the time between recall initiation to recall termination date, and reported in total calendar days. Recall classification duration times were calculated between recall initiation to recall classification date, and reported in total calendar days. For the sample of 282 recall events, the average for time to recall duration was 429 days with a standard deviation of 244 days. For a sample of 282 recall events, the average time to classification duration was 78 days with a standard deviation of 86 days. The results of this data were used to conduct statistical analysis using Anderson-Darling normality test, Goodness of Fit and box plots to test for normality, find the relevant distribution pattern, and identify outliers within the data set. 57 Summary The PDSA learning cycle was used to explore the identification, analysis, evaluation, and recommendations for risk mitigation of software related recalls and increases in time to recall duration. Using quality tools such as pareto charts, fishbone diagrams, and FMEA reveals important aspects of the key factors that contribute to the issuance of recalls and prolonging time to recall related to software defects. The evaluation of data yields new insights for understanding the recall process, and the contributing characteristics to ensure that recalls are effective and managed in a timely and effective way to reduce risk of harm to patient from exposure to defective product. The results of the analysis led to recommendations to industry organizations and the FDA on how to mitigate risks to better manage recall requirements based on regulatory and retrospective data analysis. 58 CHAPTER 4 RESULTS AND DISCUSSION Software Related Recall Results Test for Normality of Distribution Software related recall events accounted for 282 recall events identified from the data set, and this is the final sample size used for the results of this study. In order to determine normality of distribution, Minitab® software application was used to conduct an Anderson-Darling Normality Test. Under the Anderson-Darling test, H0 and H1 represent two phases of normality where H0 is normal, and H1 is not normal. If the resulting p-value for the test is less than 0.05, then H0 is rejected and the data distributions is classified as non-normal. Figures 15 and 16 represent the Anderson-Darling results for Time to Recall and Time to Classification Variables. Figure 15. Anderson-Darling Normality Test for Time to Classification. Figure created in Minitab®. 59 Figure 16. Anderson-Darling Normality Test for Time to Recall. Figure created in Minitab®. The resulting Anderson-Darling normality test concluded that time to classification and time to recall data followed a non-normal distribution. For time to classification, Minitab® calculated the mean of data at 77.73 days with a standard deviation of 86.41 days. Additionally, the p-value is less than 0.005, confirming that the distribution is not normal. For time to recall, Minitab® calculated the mean of data at 429.5 with a standard deviation of 243.8 days. Additionally, the p-value for time to recall is less than 0.005, confirming the distribution is not normal. Since non-normality for distribution of data has been established for both variables, the data for time to recall and time to classification were evaluated using a distribution fitting procedure to determine which distribution each variable follows. Distribution fitting was utilized to select the which statistical distribution bet fits the data in order to determine the probability of the data producing results that fall outside of the expectation. 60 Time to Classification Identification of Outliers Distribution fitting for Time to Classification resulted in four statistical plots, representing fifteen different tests for goodness of fit as shown in Appendix C. According to the p-values calculated by Minitab®, time to classification did not produce any distribution or transformation that could be utilized to normalize the data for analysis. Table 2 represents complied results from the analysis for distribution fitting, and the associated p-values that were calculated. From the results, there were no p-values listed as high enough to be significant. Time to Classification data was further evaluated for outliers following non-normal distribution. Table 2 Goodness of Fit Test for distribution of data for Time to Classification Distribution AD P LRT P Normal Box-Cox Transformation 29.984 4.203 <0.005 <0.005 Lognormal 3-Parameter Lognormal 4.203 4.097 <0.005 * 0.716 Exponential 10.803 <0.003 2-Parameter Exponential Weibull 9.846 10.773 <0.010 0.000 <0.010 3-Parameter Weibull Smallest Extreme Value 9.801 42.628 <0.005 0.000 <0.010 Largest Extreme Value 18.539 <0.010 Gamma 11.098 <0.005 3-Parameter Gamma 10.189 * Logistic Loglogistic 22.837 3.612 <0.005 <0.005 3-Parameter Loglogistic 3.247 * 0.005 0.114 In order to find outliers for time to classification data, a box plot was utilized to identify points within the data set that were beyond the threshold of acceptability. For the box plot shown in Figure 18, the threshold is represented by 1.5 times the Interquartile range (IQR), calculated in Minitab®. The IQR is represented in Figure 18 as whiskers, any points beyond the edge of that 61 line were identified as outliers, resulting in the recall events that are out of range for time to classification. Figure 17. Identifying outliers with Box-Plot graph for Time to Classification. Figure Created in Minitab®. Using descriptive statistics calculated in Minitab®, an outlier rule was created to identify the points within the data set outside of 1.5 times the IQR, where IQR was calculated to be 58.25 days. The outlier rule applies to any recall event where time to classification was found to be greater than or equal to 87.375 days. A total of 69 of 282 recall events were identified as outliers for time to classification based on the outlier rule using Box-Plot. A summary of the associated recall events is listed in Appendix D. Time to Recall Identification of Outliers Distribution fitting for Time to Recall resulted in four statistical plots, representing fifteen different tests for goodness of fit as shown in Appendix E. According to the p-values calculated by Minitab®, time to recall data was identified for Gamma distribution, according to 62 the p-values calculated during goodness for fit testing. Table X represents a summary of the pvalues calculated within Minitab®. Gamma distribution resulted in the most significant with a pvalue of greater than 0.25. Minitab® calculated the shape and scale of Gamma distribution to be 2.9939. and 143.4546 respectively. Table 3 Goodness of Fit Test for distribution of data for Time to Recall Distribution AD P Normal 4.991 <0.005 LRT P Box-Cox Transformation Lognormal 0.455 0.267 1.137 0.006 3-Parameter Lognormal 0.567 Exponential 23.594 * <0.003 0.002 2-Parameter Exponential 16.883 <0.010 0.000 Weibull 1.004 0.012 3-Parameter Weibull 0.540 0.176 Smallest Extreme Value 12.054 Largest Extreme Value 1.006 <0.010 0.012 Gamma 0.433 >0.250 3-Parameter Gamma 0.385 * Logistic 3.908 <0.005 Loglogistic 1.070 3-Parameter Loglogistic 0.892 <0.005 * Johnson Transformation 0.278 0.649 0.014 0.526 0.174 In order to find outliers for time to classification data, a box plot was utilized to identify points within the data set that were beyond the threshold of acceptability. For the box plot shown in Figure 18, the threshold is represented by 1.5 times the Interquartile range (IQR), calculated in Minitab®. The IQR is represented in Figure 18 as whiskers any points beyond the edge of that line were identified as outliers resulting in the recall events that are out of range for time to classification. 63 Figure 18. Box Plot for Time to Recall. Figure created in Minitab®. Using descriptive statistics calculated in Minitab®, an outlier rule was created to identify the points within the data set outside of 1.5 times the IQR, where IQR was calculated to be 319.8 days. The outlier rule applies to any recall event where time to recall was found to be greater than or equal to 479.70 days. A total of 96 of 282 recall events were identified as outliers for time to classification based on the outlier rule using Box-Plot. A summary of the associated recall events is listed in Appendix F. The IQR is represented as the end of the whiskers within the boxplot in Figure 18. Summary of Key Characteristics for Outliers FDA Recall Monitoring The results of study suggest that FDA overview process for recall termination additionally impacts time to recall duration. According to a literature search, a report by the Government Accountability Office (GAO) reported in 2011 that the FDA should review recall termination requests within three months of submission by the organization (GAO, 2011). FDA termination timeline could not be determined from the data set provided by the Division of 64 Freedom of Information or the recall records made available online, this proved to be a noteworthy limitation of the data analysis for publicly available information. Currently, there are no target timelines from the FDA to suggest what the time to recall duration should be in an ideal scenario, and additionally does not suggest that the FDA will provide an immediate review for termination. It is probable that the FDA requires a longer review period for organization’s that have been found out of compliance with key protocols that impact time to recall. Kramer, Tan, Sato, and Kesselheim (2014) stated that Class II recalls tend to involve strict FDA oversight, including follow-up and auditing of communications provided to consumer and key stakeholders, and recall documentation and reporting. Monitoring of recall implementation strategy and effectiveness could take longer time for review to ensure that actions are effective and consistent for issues as serious and high-risk as medical device recalls. This is outside of the organization’s control. Enforcement and Inspection History The enforcement history data for the organizations that had the highest number of time to recall duration confirms were reviewed to identify pattern similarities between enforcement data and recall initiation. Specifically, procedures and policies dealing with the handling of nonconforming product and addressing known issues through corrective and preventive action. This was identified initially through root cause analysis and confirmed through review of special events found in Table 2. This finding is significant and supported a study conducted by Kramer et al. (2014) suggesting that protocols for corrective and preventive actions collect reports of problems and implementing solutions that include the initiation of recalls. Notification of correction or removal protocols also ensure that regulatory requirements are met for informing the FDA of a recall including correction of the defect or removal of the medical device. 65 Additionally, recalls initiated within the US have implication for other global markets. In Japan, if a recall is initiated elsewhere, it is likely to be recalled there signaling an organization’s need to have protocols in place to ensure all marketed territories are considered (Kramer et al., 2014). In the case of Company ABC, all of the recalls found in this study were dated within the year after an inspection at the facility which resulted in the issuance of an Inspection Citation FDA Form 483. Company ABC was initially cited for lack of adequate implementation of procedures related to Corrective and Preventive Action, Nonconforming Product, and Corrections and Removals. Company ABC failed to adequately address the root causes and implement effective solutions, which led to the issuance of the warning letter. The organization was previously cited against 21 CFR 820 Quality System Regulation and 21 CFR 806 Corrections and Removals from inspections in 2009, 2010, and 2011. Three citations were issued for corrective and preventive action activities, one citation was related to failure to establish procedures for nonconforming product, and one was for failure to report device malfunction. These citations reveal insights as to how the lack of control an organization has over their systems via the implementation of protocols can lead to recall initiation and prolong time to recall. An additional review of citations from the FDA Inspectional Citation Database website for citations issued between 2009 and 2019 led to the identification of several recurring citations for Company DEF. An overview of inspection citations between 2012 and 2018 revealed that Company DEF has been consistently cited on eight separate inspections for failure to report risk of health or risk of malfunction to the FDA as required by regulation. Additionally, the company was cited six times for failure to establish adequate procedures for corrective and preventive action between 2012 and 2018. A warning letter was also issued for the citations related to the 66 inspection in 2017 detailing failure to address observations, especially corrective and preventive actions. Company DEF received one citation in 2017 for failure to validate software as used within the quality system. A review of Company BCD product history on the FDA’s recall website listed several recalls from 2014 to 2018 for the medical devices citing software related defects in several of the recalls initiated within that period. Additional review of citations revealed that Companies CDE, EFG, FGH, GHI, and HIJ have also been cited for inadequacies in their process for controlling non-conforming material. Regulatory enforcement and inspection history highlight characteristics of the organization in terms of responsibility and ability to comply to global regulations. Nearly every Company identified in Table 2 was cited for inadequacies in addressing nonconforming product. Since recall strategy is a large part of the recall process, this study suggests that if an organization is struggling to maintain key quality system processes in a normal scenario, the organization will struggle with rapidly responding to issues identified within the market. Available data shows that companies cited for these inadequacies most recently had longer time to recall durations. This could suggest that software related defects that triggered recalls were identified during the inspection, or during resolution of inspection findings. Root Cause Identification and Risk Mitigation Strategy Risk mitigation must begin and end with identifying the true root causes that contribute to time to recall duration. In the initial assessment of cause-and-effect analysis, four root causes were identified for prolonged time to recall duration. These included software defects, FDA overview, organization protocols, and organization resources. After studying the results, the confirmed root causes for time to recall duration remain at organization resources and FDA 67 overview. Software defects is identified as a contributing factor for recall initiation and should be explored as part of Phase 0 of the proposed recall process. A manual review to recall events as listed in Appendices C and E, lead to the creation and development of two fishbone diagrams for root cause analysis. Key themes emerged in reviewing the reason for recall and total product lifecycle data provided on the FDA website as part of the public recall record. The root cause diagram identified three primary root causes for software defects that lead to software related recalls, and an additional root cause assessment of potential factors contributing to prolonged time to recall duration. Each branch added factors that were found during the manual review and assumed to contribute to prolonged time to recall. The fishbone diagrams are shown in Appendices G and H. Software defects can be mitigated using the FMEA process to trend recalls from industry for defects that could be applicable to all or similar medical devices. The new FMEA suggests risk mitigation strategies for three of the four originally proposed root causes. Appendix I provides an updated FMEA for confirmed root causes with suggested risk mitigation strategies to ensure process issues are addressed to reduce time to recall duration. Act Phase The organizational learning process can be developed by preparing a review of the recall requirements against industry data using the PDSA framework to propose a recall process that will mitigate risks associated with longer recall times. The decision to adopt the proposed process for risk and root cause mitigation of the recall process to decrease time to recall duration has important implications for the prevention and management of software related medical device recalls. Using quality tools, this study identified factors associated with the initiation of the recall such as the software defect, and process gaps that contribute to prolonged time to 68 recall. Through root cause analysis and FMEA, steps to mitigate prolonged time to recall and address software defects were proposed for implementation by an organization. By addressing the software defects through continuous feedback evaluation and risk mitigation, the organization and take a proactive position on reducing recall initiation. By addressing the process failure modes through continuous risk analysis based on industry data trends, the organization can develop and implement plans and protocols in place to ensure that recalls are managed at rapidly and effectively, reducing time to recall duration. The decision to adopt is supported by research on the product recall process from various industries. According Kumar and Schmitz (2011), performing a risk assessment prior to the recall is a proactive measure, and it puts the organization in a better position to react quickly and effectively during a recall.The study conducted a recall analysis to identify failure points that required risk mitigation from both a process and product perspective. Similar to the results in this thesis, Kumar and Schmitz (2011) concluded that organizations need to understand their risks and be prepared to defend their decision of whether or not to initiate a recall. This not only includes the method of communication about the recall, but also includes selecting the appropriate recall strategy. The study suggests that organizations need to adopt a proactive approach for preventing recalls. Proactive activities include four basic types of learning listening, testing, studying, and tracking information. These activities encompass all aspects of the product lifecycle from design to customer feedback incorporating external data back into the system for quality improvements. Kumar and Schmitz (2011) states that organizations that participate in the global supply chain need to work on improving their product recall process. Additionally, the research proposed that improvements can be achieved through the use of technology tracking data throughout the entire 69 supply chain and include more diligent analysis of failure modes that result in a product recall. This study confirms the relevance of the purpose of this thesis by connecting the use of risk management tools into process improvements for the recall process to track and trend relevant data to support a more effective and timely recall. Furthermore, GAO reported in the 2011 analysis, which was the second the GAO had produced since 1998, that there were several gaps in the medical device recall process and oversight from the FDA. The gaps limited the organization’s and the FDA’s ability to ensure that the highest-risk recalls were implemented in a timely manner (GAO, 2011). The GAO’s study states that the FDA does not routinely analyze recall data to identify whether or not there are systemic problems affecting recalls. Instead the FDA is trending recall data to support their inspection resources, and compliance and enforcement actions. The GAO’s study confirms the necessity of the analysis conducted within this thesis. By examining recall data for software related defects, organizations can learn to identify systemic trends affecting industry recalls that might be applicable to their product type. Additionally, discovering how the FDA is using the recall data is important for organization understanding and risk management. The FDA has not analyzed the data to educate or inform organization learning, yet instead uses the data to enforce and inspect organizations. This observation can be used to confirm that the trend observed within recall initiation from organizations that had been recently inspected holds validity. By using the data to support risk management activities for recall initiation and time to recall duration, organizations can also maintain compliance and reduce regulatory risk while maintaining patient safety. Simultaneously, the FDA can rely on the organization’s risk management strategy to manage recalls. 70 A study conducted by Fu et al. (2017) concluded with a presentation of the procedure to collect software related medical devices from the FDA to share important data about software recalls. The analysis revealed important insights about limitations of data made available by the FDA. The study notes that industry needs to collect more information about software-related recalls to confirm detailed root causes that contribute to software failure modes in medical devices. This learning would allow industry to adequately identify and address potential safety hazards and risk mitigation strategies to prevent recurrence of similar defects in current and future product designs. By addressing software defect root causes through industry data trending, organizations can take a proactive approach at planning for defects to (1) reduce the likelihood of recall initiation, and (2) reduce time to recall by rapidly identifying and mitigating software defects with a pre-planned risk mitigation strategy. All three studies provided information about the use of a recall management tool that encompasses consideration of risk management prior to and during the recall process supporting the purpose and results of this thesis. The proposed process improvements including recall data benchmarking and risk analysis will not only provide learning within the organization but best practices for industry. The results of this study provides evidence to support that using data analysis and risk mitigation strategy an organization can (1) plan to address software defects as they are identified, (2) incorporate systematic trends in similar devices into the design FMEA for future considerations, and (3) reduce time to recall duration by planning ahead for hazardous situations and having appropriate processes in control for management of nonconforming product to increase patient safety by reducing risk of harm. 71 CHAPTER 5 CONCLUSION Summary The medical device industry has seen an increase in software related recalls in recent years, and the consistent trend has raised some concerns for organizational awareness of recall management. Prolonging the time to recall duration negatively impacts patients’ risk of harm by prolonging exposure to hazardous software defects that could pose risk of serious injury to end user. This thesis examined the recall process, including relevant regulatory considerations, to understand the factors that contribute to prolonged time to recall duration. Initial hypothesis suggested that product traceability, recall protocols, and distribution pattern were causes for prolonged time to recall duration that needed to be addressed. Using the PDSA framework this thesis produced an analysis of FDA recall data to evaluate the impact these factors had on software related recalls. Comparison of these factors was made against similar metrics for all recalls to determine if any differences exist between the management of software related defects versus all other defects that lead to initiation of the product recall. Quality tools were utilized to uncover root cause analysis for prolonged time to recall and confirm gaps within the process that can be attributed to timing delays. Conclusions The results of this study refuted the hypothesis for the original factors suggested to contribute to prolonged time to recall. This study provided a connection between literature and data to support that lack of compliance to regulation and ensuring processes for recall management were not adequately implemented, disabling an organization’s ability to efficiently 72 recall defective software embedded medical devices. Additionally, the FDA recall overview process contributes to prolonged time to recalls suggesting communication gaps between the FDA and organization on timelines and roles and responsibilities of timely recall management. Analysis of enforcement data confirmed that prolonged time to recalls existed for organization’s who consistently struggled to maintain full compliance for procedures related to corrective and preventive actions, control of nonconforming product, and notification of product action. This suggests that the organization’s recall strategy would not ensure a reduced time to recall since proper controls were not already in place. Recommendations The following recommendations were made to industry, and the FDA for using risk mitigation and root cause analysis to address gaps within the software related recall process. In order to reduce software defects leading to the initiation of recall, the organization and the FDA must trend previous recall data to identify trends in recall defects. Using this data, the organization should propose risk identification through FMEA and explore risk mitigation strategies for defects that are applicable to their device type to reduce recall likelihood. Data trending should occur on a semi-annual to annual basis. Using data from this thesis, the organization should consider addressing gaps with the recall management process by evaluating relevant procedures for notification of adverse event or device malfunction, handling nonconforming on-market product, and developing a process for recall management. The process improvement recommendation is to include Phase 0 for data trending to understand time to recall delays in conjunction with analysis of inspection observations for processes related to recall management. 73 The FDA should also make changes to their data monitoring and trending process to address limitations discovered within data analysis to allow for better use of public data. The FDA should trend root cause of recall by relevant quality system category to highlight the nonconformity and additionally, include subcategories for deeper review. The recall database should include figures on total product recovery percentage to alert industry and the public on the status of remaining recall inventory on market. Within the recall database, there should be a field to include when the organization requested termination versus when the FDA determines termination to understand the time delay between request and response. For distribution patter, the FDA should include additional field for region, and the number of sites impacted to make meaningful correlations to product location impact throughout the recall process. With changes made per these recommendations, future research can be explored to understand additional underlying factors for prolonged time to recall duration. 74 REFERENCES AAMI. (2018, May 22). Medical Device Recalls Surge in 2018 with Software to Blame. Retrieved from Advancing Safety in Health Technology: https://www.aami.org/newsviews/newsdetail.aspx?ItemNumber=6475 AAMI. (2019, Aug 20). About AAMI. Retrieved from Association for the Advancement of Medical Instrumentation: https://www.aami.org/membershipcommunity/content.aspx?ItemNumber=1292&navItem Number=4603 ASQ. (2019, September 17). THE 7 BASIC QUALITY TOOLS FOR PROCESS IMPROVEMENT. Retrieved from American Society for Quality: https://asq.org/qualityresources/seven-basic-quality-tools CDRH. (2018). Medical Device Enforcement and Quality Report. U.S. Food and Drug Administration. Chen, Y.-J., Chiou, C.-M., Huang, Y.-W., Tu, P.-W., Lee, Y.-C., & Chien, C.-H. (2018). A Comparative Study of Medical Device Regulations: US, Europe, Canada, and Taiwan. Therapeutic Innovation & Regulatory Science, 62-69. Deming, S. N. (2016). Statistics in the Laboratory: Control Charts, Part 3. American Laboratory, 12-15. Deming, W. E. (1982). Out of The Crisis. Cambridge: Masschusetts Institute of Technology, Center for Advanced Educational Services. Deming, W. E. (1994). The New Economics. Cambridge: Massachusetts Institure of Technology, Center for Advanced Educational Services. 75 European Parliament. (2017, April 5). REGULATION (EU) 2017/745 OF THE EUROPEAN PARLIAMENT AND OF THE COUNCIL. Official Journal of the European Union . The European Parliament and the Council of the European Union. FDA. (2010, May 18). FDA 101: Product Recalls. Retrieved from U.S. Food and Drug Administration: https://www.fda.gov/consumers/consumer-updates/fda-101-productrecalls FDA. (2014, July 31). Recall Background and Definitions. Retrieved from Food and Drug Administration: https://www.fda.gov/safety/industry-guidance-recalls/recallsbackground-and-definitions FDA. (2017, July 25). Enforcement Report Information and Definitions. Retrieved from U.S. Food and Drug Administration: https://www.fda.gov/safety/enforcementreports/enforcement-report-information-and-definitions FDA. (2017, August 18). Philips Medical Systems (Cleveland) Inc, Cleveland, OH, 483. Retrieved from Food and Drug Administraion: https://www.fda.gov/media/111401/download FDA. (2018, August 10). Arkon Anesthesia Delivery System recalled due to unexpected failed state while in use or idle. Retrieved from Food and Drug Administration: https://www.fda.gov/medical-devices/medical-device-recalls/arkon-anesthesia-deliverysystem-recalled-due-unexpected-failed-state-while-use-or-idle FDA. (2018, April 1). CFR - Code of Federal Regulations Title 21 Part 803 Medical Device Reporting. Retrieved from Food and Drug Administration: https://www.accessdata.fda.gov/scripts/cdrh/cfdocs/cfcfr/CFRSearch.cfm?fr=803.3 76 FDA. (2018, September 14). Medical Device Overview. Retrieved from https://www.fda.gov/industry/regulated-products/medical-device-overview: https://www.fda.gov/industry/regulated-products/medical-device-overview FDA. (2018, April 1). Medical Devices; Reports of Corrections and Removals. Retrieved from Code of Federal Regulations Titles 21 Part 806: https://www.accessdata.fda.gov/scripts/cdrh/cfdocs/cfcfr/CFRSearch.cfm?CFRPart=806 FDA. (2018, August 31). Software as a Medical Device (SaMD). Retrieved from U.S. Food and Drug Administration: https://www.fda.gov/medical-devices/digital-health/softwaremedical-device-samd FDA. (2018, 09 26). What is a Medical Device Recall. Retrieved from U.S. Food & Drug Administration: https://www.fda.gov/MedicalDevices/Safety/ListofRecalls/ucm329946.htm FDA. (2019, Sep 04). 21 CFR Part 7 Enforcement Policy. Retrieved from Food and Drug Administration: https://www.accessdata.fda.gov/scripts/cdrh/cfdocs/cfcfr/CFRSearch.cfm?CFRPart=7 FDA. (2019, April 15). Brainlab AG Recalls Spine & Trauma 3D Navigation Due to Inaccurate Display That May Result in User Misinterpretation. Retrieved from Food and Drug Administration: https://www.fda.gov/medical-devices/medical-device-recalls/brainlabag-recalls-spine-trauma-3d-navigation-due-inaccurate-display-may-result-user FDA. (2019, July 12). Center for Devices and Radiological Health. Retrieved from U.S. Food and Drug Administration: https://www.fda.gov/about-fda/office-medical-products-andtobacco/center-devices-and-radiological-health 77 FDA. (2019, July 19). Digital Health. Retrieved from Medical Devices: https://www.fda.gov/medical-devices/digital-health FDA. (2019, Jan 08). Medtronic Recalls Synergy Cranial Software and Stealth Station S7 Cranial Software Due to Inaccuracies Displayed During Surgical Procedures. Retrieved from Food and Drug Administration: https://www.fda.gov/medical-devices/medicaldevice-recalls/medtronic-recalls-synergy-cranial-software-and-stealth-station-s7-cranialsoftware-due-inaccuracies FDA. (2019, June 05). Unique Device Identification System (UDI System). Retrieved from U.S. Food and Drug Administrarion: https://www.fda.gov/medical-devices/device-advicecomprehensive-regulatory-assistance/unique-device-identification-system-udi-system Fu, Z., Guo, C., Zhang, Z., Jiang, Y., & Sha, L. (2017). Study of Software-Related Causes in the FDA Medical Device Recalls. International Conference on Engineering of Complex Computer Systems, 60-69. GAO. (2011). Medical Devices FDA Should Enhance Its Oversight of Recalls. United States Government Accountability Office. Gottlieb, S. (2018, April 26). Transforming FDA's Approach to Digital Health. Retrieved from Food and Drug Administration: https://www.fda.gov/news-events/speeches-fdaofficials/transforming-fdas-approach-digital-health-04262018 Government of Canada. (2019, June 17). Medical Devices Regulations (SOR/98-282). Retrieved from Justice Laws Website: https://laws-lois.justice.gc.ca/eng/regulations/sor-98282/page-1.html Health Canada. (2016, November 3). Medical devices recall guide. GUI-0054. Canada. 78 Hora, M., Bapuji, H., & Roth, A. V. (2011). Safety hazard and time to recall: The role of recall strategy, product defect type, and supply chain player in the U.S. toy industry. Journal of Operations Management, 766-777. ISO. (2016, March 01). ISO 13485. Medical devices — Quality management systems Requirements for regulatory purposes. Geneva, Switzerland: International Organization for Standardization. ISO. (2019, 16 Sep). About Us. Retrieved from International Organization for Standarization.: https://www.iso.org/about-us.html Kramer, D. B., Tan, Y. T., Sato, C., & Kesselheim, A. (2014). Ensuring Medical Device Effectiveness and Safety: A Cross - National Comparison of Approaches to Regulation. Food and Drug Law Journal, 69(1), 1-i. Kumar, S., & Schmitz, S. (2011). Managing recalls in a consumer product supply chain - root cause analysis and measures to mitigate risks. International Journal of Production Research, 235-253. Mezher, M. (2018, Jan 10). Device Recalls in 2017: Making Sense of the Numbers. Retrieved from Regulatory Affairs Professionals Society: Device Recalls in 2017: Making Sense of the Numbers Mulero, A. (2018, March 08). Philips Medical Systems Draws Lengthy FDA 483 Over Issues with Thousands of Complaints. Retrieved from Regulatory Affairs Professionals Society: https://www.raps.org/news-and-articles/news-articles/2018/3/philips-medical-systemsdraws-lengthy-fda-483-over National Archives and Records Administration. (2019, July 22). Federal Register. Retrieved from National Archives: https://www.archives.gov/federal-register/cfr/subjects.html 79 Public Health Service. (2015, September 30). Warning Letter Merge Healthcare, Inc. Retrieved from Food & Drug Administration: https://www.fda.gov/inspections-complianceenforcement-and-criminal-investigations/warning-letters/merge-healthcare-inc-09302015 Ronquillo, j. G., & Zuckerman, D. M. (2017). Software‐Related Recalls of Health Information Technology and Other Medical Devices: Implications for FDA Regulation of Digital Health. The Milbank Quarterly, 535-553. Stericycle. (2018). Stericycle Recall Index Q1 2018. Stericycle. Stericycle Expert Solutions. (2019). Recall Index Q2 2019. Stericycle Expert Solutions. Taylor, M. J., Mcnicholas, C., Nicolay, C., Darzi, A., Bell, D., & Reed, J. E. (2014). Systematic review of the application of the plan–do–study–act method to improve quality in healthcare. BMJ Quality & Safety, 290-298. TGA. (2017, June 9). Is my product a therapeutic good? Retrieved from Therapeutic Good Administration: https://www.tga.gov.au/sme-assist/my-product-therapeutic-good#1 TGA. (2019, Feb 7). Uniform recall procedure for therapeutic goods (URPTG). Retrieved from Therapeutic Goods Act: https://www.tga.gov.au/publication/uniform-recall-proceduretherapeutic-goods-urptg Wood, L. C., Wang, J. X., Olesen, K., & Reiners, T. (2017). The effect of slack, diversification, and time to recall on stock market reaction to toy recalls. . International Journal of Productions Economics, 244-258. Wowak, K. A., & Boone, C. A. (2015). So many recalls, so little research: A review of the literature and road map for future research. Journal of Supply Chain Management, 54-72. 80 Yu, J., & Hooker, N. H. (2019). Exploring relationships among recall effectiveness indicators in the US meat and poultry industry. International Food and Agribusiness Management Review, 97-106. 81 APPENDIX A: FLOWCHART OF RECALL MANAGEMENT PROCESS 82 APPENDIX B: CODED SOFTWARE RELATED RECALL DATA SET Recall Event ID Region of Origin Voluntary/ Mandated 11 1 1 18 1 1 26 1 1 28 1 1 30 1 1 55 1 1 68 1 1 75 1 1 77 1 1 83 1 1 92 1 1 96 1 1 112 1 1 118 1 1 124 1 1 125 1 1 127 5 1 132 1 1 133 4 1 137 1 1 138 1 1 140 1 1 141 1 1 145 1 1 Distribution Pattern Domestic - US Only Domestic - US Only Domestic & International Domestic & International Domestic & International Domestic - US Only Domestic - US Only Domestic - US Only Domestic & International Domestic - US Only Domestic & International Domestic - US Only Domestic & International Domestic & International Domestic & International Domestic & International Domestic - US Only Domestic - US Only Domestic & International Domestic & International Domestic - US Only Domestic - US Only Domestic - US Only Domestic & International Defect Category Time to Classification Time to Recall Software 25 271 Software 67 407 Software 32 653 Software 61 543 Software 36 427 Software 49 494 Software 32 474 Software 21 384 Software 51 127 Software 13 635 Software 63 878 Software 63 180 Software 13 241 Software 46 693 Software 44 408 Software 65 348 Software 24 409 Software 56 87 Software 28 922 Software 140 429 Software 80 563 Software 52 886 Software 28 358 Software 31 580 83 Recall Event ID Region of Origin Voluntary/ Mandated 160 1 1 165 2 1 169 1 1 172 1 1 174 1 1 179 1 1 180 1 1 185 1 1 190 1 1 191 1 1 207 1 1 217 1 1 218 1 1 226 1 1 229 5 1 230 1 1 235 1 1 238 1 1 240 1 1 242 1 1 254 1 1 256 1 1 258 1 1 265 1 1 266 1 1 267 1 1 Distribution Pattern Domestic & International Domestic & International Domestic & International Domestic - US Only Domestic - US Only Domestic - US Only Domestic - US Only Domestic - US Only Domestic & International Domestic & International Domestic - US Only Domestic & International Domestic - US Only Domestic - US Only Domestic - US Only Domestic & International Domestic & International Domestic & International International oUS Only Domestic - US Only Domestic - US Only Domestic & International Domestic - US Only International oUS Only Domestic - US Only Domestic & International Defect Category Time to Classification Time to Recall Software 24 352 Software 104 439 Software 41 819 Software 48 286 Software 23 169 Software 41 395 Software 20 616 Software 30 384 Software 41 827 Software 18 266 Software 43 596 Software 76 309 Software 49 771 Software 32 98 Software 34 244 Software 17 335 Software 138 830 Software 42 850 Software 34 543 Software 51 59 Software 21 381 Software 24 285 Software 10 37 Software 27 541 Software 69 354 Software 61 786 84 Recall Event ID Region of Origin Voluntary/ Mandated 269 1 1 279 1 1 280 1 1 285 1 1 318 1 1 321 1 1 324 1 1 334 1 1 366 1 1 372 1 1 378 1 1 380 1 1 385 1 1 392 1 1 395 1 1 396 1 1 400 1 1 404 1 2 414 1 1 415 1 1 417 1 1 419 1 1 420 1 1 421 1 1 423 1 1 426 1 1 Distribution Pattern Domestic - US Only Domestic & International Domestic & International Domestic & International Domestic & International Domestic & International Domestic & International Domestic & International Domestic & International Domestic & International Domestic - US Only Domestic - US Only Domestic & International Domestic - US Only Domestic - US Only Domestic - US Only Domestic & International Domestic - US Only Domestic & International Domestic - US Only Domestic & International Domestic & International Domestic - US Only Domestic - US Only Domestic - US Only Domestic & International Defect Category Time to Classification Time to Recall Software 20 340 Software 53 815 Software 38 221 Software 30 617 Software 103 794 Software 47 956 Software 167 770 Software 60 381 Software 56 742 Software 40 154 Software 42 1016 Software 51 315 Software 107 803 Software 24 225 Software 90 816 Software 34 318 Software 36 182 Software 145 426 Software 46 809 Software 162 361 Software 40 179 Software 57 259 Software 160 370 Software 36 168 Software 82 102 Software 50 890 85 Recall Event ID Region of Origin Voluntary/ Mandated 432 1 1 447 1 1 449 1 1 453 2 1 458 2 1 464 1 1 472 1 1 474 1 1 481 2 1 485 1 1 489 1 1 493 1 1 497 1 1 499 2 1 508 1 1 518 1 1 522 1 1 523 1 1 526 1 1 527 1 1 528 1 2 530 1 1 537 1 2 539 1 1 540 1 1 541 2 1 Distribution Pattern Domestic & International Domestic & International Domestic - US Only International oUS Only Domestic - US Only Domestic & International Domestic & International Domestic - US Only Domestic & International Domestic & International Domestic & International Domestic & International Domestic - US Only Domestic - US Only Domestic & International Domestic & International Domestic - US Only International oUS Only Domestic & International Domestic & International Domestic - US Only Domestic - US Only Domestic - US Only Domestic - US Only Domestic - US Only Domestic - US Only Defect Category Time to Classification Time to Recall Software 34 325 Software 18 792 Software 47 174 Software 56 447 Software 14 451 Software 49 146 Software 10 298 Software 29 159 Software 36 62 Software 22 173 Software 15 136 Software 27 928 Software 39 190 Software 136 480 Software 185 263 Software 32 150 Software 36 712 Software 35 314 Software 41 687 Software 52 217 Software 129 135 Software 41 195 Software 164 444 Software 12 470 Software 113 247 Software 23 266 86 Recall Event ID Region of Origin Voluntary/ Mandated 545 1 1 549 1 1 551 1 1 553 1 1 556 1 1 564 1 1 570 1 1 584 1 1 593 1 1 596 1 1 598 1 1 605 1 1 610 1 1 612 1 1 618 1 1 619 1 1 624 1 1 630 1 1 632 1 1 634 1 1 635 1 1 636 1 1 637 1 1 638 2 1 643 1 1 647 1 1 Distribution Pattern Domestic & International Domestic - US Only Domestic & International International oUS Only Domestic - US Only Domestic & International Domestic - US Only International oUS Only Domestic - US Only Domestic - US Only Domestic & International Domestic & International Domestic & International Domestic & International Domestic - US Only Domestic & International Domestic - US Only Domestic - US Only Domestic & International Domestic & International Domestic - US Only Domestic & International Domestic - US Only Domestic - US Only Domestic - US Only Domestic & International Defect Category Time to Classification Time to Recall Software 101 768 Software 24 154 Software 47 827 Software 25 264 Software 205 376 Software 13 194 Software 53 76 Software 25 158 Software 205 562 Software 235 685 Software 50 638 Software 50 238 Software 32 200 Software 2 699 Software 59 653 Software 16 366 Software 43 230 Software 47 382 Software 59 557 Software 40 677 Software 34 149 Software 45 646 Software 21 244 Software 33 1013 Software 27 888 Software 31 613 87 Recall Event ID Region of Origin Voluntary/ Mandated 661 1 1 662 1 1 670 1 1 672 1 1 685 1 1 691 1 1 692 1 1 697 5 2 717 1 1 721 1 1 728 1 1 729 1 1 730 1 1 735 2 1 744 1 1 747 1 1 752 1 1 753 1 1 754 1 1 761 2 1 769 1 1 773 1 1 781 1 1 782 2 1 783 1 1 788 1 1 Distribution Pattern Domestic - US Only Domestic - US Only Domestic - US Only Domestic & International Domestic - US Only Domestic - US Only Domestic - US Only Domestic - US Only Domestic & International Domestic & International Domestic - US Only Domestic - US Only Domestic - US Only Domestic & International Domestic - US Only Domestic & International Domestic - US Only Domestic - US Only Domestic - US Only Domestic - US Only Domestic & International Domestic - US Only Domestic & International Domestic - US Only Domestic & International Domestic & International Defect Category Time to Classification Time to Recall Software 63 265 Software 33 258 Software 127 777 Software 19 232 Software 303 1145 Software 43 297 Software 38 244 Software 114 491 Software 21 208 Software 15 167 Software 25 207 Software 14 241 Software 24 201 Software 33 247 Software 29 322 Software 51 188 Software 43 222 Software 41 825 Software 15 232 Software 57 973 Software 48 594 Software 15 371 Software 30 371 Software 28 944 Software 26 371 Software 19 595 88 Recall Event ID Region of Origin Voluntary/ Mandated 789 1 1 793 1 1 801 1 1 802 1 1 804 1 1 823 1 1 841 1 1 843 1 1 844 1 1 855 1 1 857 1 1 860 1 1 881 1 1 899 1 1 911 1 1 917 1 1 926 1 1 927 1 1 929 1 1 931 2 1 940 1 1 945 2 1 946 1 1 954 1 1 955 1 1 959 1 1 Distribution Pattern Domestic - US Only Domestic - US Only Domestic - US Only Domestic & International Domestic & International Domestic & International Domestic & International Domestic - US Only Domestic - US Only Domestic - US Only Domestic - US Only Domestic & International Domestic - US Only Domestic & International Domestic & International Domestic & International Domestic - US Only Domestic - US Only Domestic - US Only Domestic - US Only Domestic & International Domestic - US Only Domestic & International Domestic & International Domestic - US Only Domestic - US Only Defect Category Time to Classification Time to Recall Software 23 204 Software 27 391 Software 27 528 Software 30 378 Software 22 151 Software 20 181 Software 78 555 Software 296 1106 Software 71 72 Software 32 852 Software 39 161 Software 26 563 Software 48 118 Software 37 259 Software 33 112 Software 59 106 Software 311 1099 Software 20 322 Software 20 309 Software 34 229 Software 28 574 Software 84 875 Software 25 189 Software 62 483 Software 51 301 Software 47 307 89 Recall Event ID Region of Origin Voluntary/ Mandated 960 1 1 961 1 1 963 1 1 964 1 1 968 1 1 972 1 1 980 1 1 981 1 1 985 1 1 986 1 1 991 4 1 993 1 1 1001 1 1 1009 1 1 1018 1 1 1023 1 1 1025 1 1 1026 1 1 1027 1 1 1029 1 1 1035 1 1 1037 1 1 1038 1 1 1041 1 1 1057 2 1 1058 1 1 Distribution Pattern Domestic & International Domestic & International Domestic - US Only Domestic - US Only International oUS Only Domestic & International Domestic & International Domestic - US Only Domestic & International Domestic & International Domestic - US Only Domestic & International Domestic - US Only Domestic - US Only Domestic - US Only Domestic - US Only Domestic & International Domestic & International Domestic & International Domestic & International Domestic - US Only Domestic - US Only Domestic - US Only Domestic & International Domestic & International Domestic & International Defect Category Time to Classification Time to Recall Software 54 510 Software 81 559 Software 350 350 Software 74 806 Software 87 93 Software 34 262 Software 408 812 Software 327 332 Software 44 517 Software 49 505 Software 90 357 Software 48 208 Software 335 343 Software 48 572 Software 38 489 Software 30 168 Software 29 445 Software 33 239 Software 25 733 Software 28 359 Software 21 98 Software 34 307 Software 28 244 Software 28 202 Software 27 547 Software 32 367 90 Recall Event ID Region of Origin Voluntary/ Mandated 1070 1 1 1071 1 1 1073 1 1 1074 1 1 1079 1 1 1090 1 1 1091 1 1 1095 2 1 1097 1 1 1118 1 1 1129 1 1 1137 1 1 1145 1 1 1147 1 2 1171 1 1 1173 1 1 1181 1 2 1184 1 1 1188 1 2 1202 1 1 1206 1 1 1208 1 1 1209 1 1 1214 1 1 1215 2 1 1216 2 1 Distribution Pattern Domestic - US Only Domestic & International Domestic - US Only Domestic - US Only Domestic - US Only Domestic & International Domestic & International Domestic - US Only Domestic & International Domestic & International Domestic - US Only Domestic & International Domestic - US Only Domestic - US Only Domestic - US Only Domestic - US Only Domestic - US Only Domestic - US Only Domestic & International Domestic - US Only Domestic & International Domestic & International Domestic & International Domestic & International Domestic & International Domestic & International Defect Category Time to Classification Time to Recall Software 372 1053 Software 13 538 Software 17 132 Software 48 308 Software 13 119 Software 31 126 Software 39 423 Software 36 312 Software 25 721 Software 21 501 Software 404 460 Software 30 393 Software 36 280 Software 118 490 Software 45 148 Software 434 1062 Software 29 425 Software 51 463 Software 24 436 Software 183 406 Software 38 378 Software 24 421 Software 6 385 Software 601 831 Software 411 996 Software 203 850 91 Recall Event ID Region of Origin Voluntary/ Mandated 1218 1 1 1219 1 1 1229 1 1 1236 1 1 1237 1 1 1262 1 1 1267 1 1 1284 2 1 1290 1 1 1291 1 1 1301 1 1 1302 1 1 1303 1 1 1321 2 1 1325 1 1 1326 1 1 1331 1 1 1334 1 1 1335 1 1 1339 1 1 1349 1 1 1354 1 1 1364 1 1 1372 1 1 1373 1 1 1377 1 1 Distribution Pattern Domestic & International Domestic & International Domestic - US Only Domestic & International Domestic - US Only International oUS Only Domestic & International Domestic - US Only Domestic - US Only Domestic - US Only Domestic - US Only Domestic - US Only Domestic - US Only Domestic - US Only Domestic - US Only Domestic - US Only Domestic - US Only Domestic & International Domestic - US Only Domestic & International Domestic - US Only Domestic & International Domestic & International Domestic & International Domestic - US Only Domestic & International Defect Category Time to Classification Time to Recall Software 245 463 Software 181 424 Software 93 275 Software 47 602 Software 22 632 Software 9 538 Software 44 583 Software 87 121 Software 166 404 Software 253 470 Software 197 456 Software 211 664 Software 189 251 Software 183 408 Software 182 363 Software 68 247 Software 176 364 Software 19 316 Software 252 414 Software 144 549 Software 166 377 Software 200 605 Software 152 349 Software 231 642 Software 221 288 Software 50 494 92 Recall Event ID Region of Origin Voluntary/ Mandated 1378 1 1 1379 1 1 1391 1 1 1411 1 1 1417 1 1 1424 1 1 1440 2 1 1444 2 1 1445 5 1 1448 1 1 1453 1 1 1474 2 1 1486 1 1 1495 1 1 1515 1 1 1521 1 1 1524 1 1 1525 1 1 1540 1 1 1542 2 1 1562 2 1 1586 1 1 1591 1 1 1602 2 1 Distribution Pattern Domestic & International Domestic - US Only Domestic - US Only International oUS Only Domestic & International Domestic & International Domestic - US Only Domestic - US Only Domestic - US Only Domestic - US Only UNKNOWN Domestic - US Only Domestic & International Domestic - US Only Domestic & International Domestic & International Domestic & International Domestic & International Domestic & International Domestic & International Domestic - US Only Domestic & International Domestic - US Only Domestic & International Defect Category Time to Classification Time to Recall Software 177 345 Software 152 370 Software 182 396 Software 244 466 Software 42 431 Software 162 534 Software 63 386 Software 133 471 Software 144 489 Software 217 469 Software 77 265 Software 86 414 Software 62 456 Software 124 437 Software 96 259 Software 219 224 Software 92 233 Software 166 300 Software 57 380 Software 56 153 Software 50 302 Software 54 160 Software 48 112 Software 39 155 93 APPENDIX C: DETERMINATION OF DISTRIBUTION FITTING FOR TIME TO CLASSIFICATION 94 95 APPENDIX D: BOX-PLOT RESULTS: OUTLIERS FOR TIME TO CLASSIFICATION Recall Event ID Time to Classification 137 165 235 140 104 138 Recall Event ID 1214 1215 1216 Time to Classification 601 411 203 318 324 385 395 404 415 103 167 107 90 145 162 1218 1219 1229 1290 1291 1301 245 181 93 166 253 197 420 499 508 528 537 160 136 185 129 164 1302 1303 1321 1325 1331 211 189 183 182 176 540 545 556 593 596 113 101 205 205 235 1335 1339 1349 1354 1364 252 144 166 200 152 670 685 697 843 926 963 127 303 114 296 311 350 1372 1373 1378 1379 1391 1411 231 221 177 152 182 244 980 981 991 1001 1070 408 327 90 335 372 1424 1444 1445 1448 1495 162 133 144 217 124 1129 1147 1173 1202 404 118 434 183 1515 1521 1524 1525 96 219 92 166 96 APPENDIX E: DETERMINATION OF DISTRIBUTION FITTING FOR TIME TO RECALL 97 98 APPENDIX F: BOX-PLOT RESULTS: OUTLIERS FOR TIME TO RECALL Recall Event ID Time to Recall Recall Event ID Time to Recall Recall Event ID Time to Recall 26 28 55 653 543 494 618 632 634 653 557 677 1216 1236 1237 850 602 632 83 92 118 133 138 635 878 693 922 563 636 638 643 647 670 646 1013 888 613 777 1262 1267 1302 1339 1354 538 583 664 549 605 140 145 169 180 190 886 580 819 616 827 685 697 753 761 769 1145 491 825 973 594 1372 1377 1424 1445 642 494 534 489 207 218 235 238 240 265 596 771 830 850 543 541 782 788 801 841 843 855 944 595 528 555 1106 852 267 279 285 318 321 786 815 617 794 956 860 926 940 945 954 563 1099 574 875 483 324 366 378 385 395 770 742 1016 803 816 960 961 964 980 985 510 559 806 812 517 414 426 447 493 499 522 809 890 792 928 480 712 986 1009 1018 1027 1057 1070 505 572 489 733 547 1053 526 545 687 768 1071 1097 538 721 99 Recall Event ID 551 Time to Recall 827 593 596 598 612 562 685 638 699 Recall Event ID 1118 1147 1173 1214 1215 Time to Recall Recall Event ID 501 490 1062 831 996 Time to Recall 100 APPENDIX G: CAUSE-AND-EFFECT, FISHBONE DIAGRAM FOR ROOT CAUSE ANALYSIS OF FACTORS LEADING TO RECALL INITIATION FOR SOFTWARE RELATED DEFECTS 101 APPENDIX H: CAUSE AND EFFECT, FISHBONE DIAGRAM FOR ROOT CAUSE ANALYSIS OF FACTORS LEADING PROLONGED TIME TO RECALL DURATION 102 APPENDIX I: PROPOSED FMEA FOR CONFIRMED ROOT CAUSES CONTRIBUTING TO TIME TO RECALL DELAY, INCLUDING RISK MITIGATION RECOMMENDATIONS. Revision : 02 Purpose: To identify failure modes for root cause categories identified through cause-and-effect analysis to mitigate patient risk of prolonged time to recall duration. Instructions: Use the following template in conjunction with cause-and-effect analysis to identify failure modes for root cause. Recommend risk mitigation based on failure mode and calculate risk priority numbers prior to risk mitigation and after to determine if resulting conditions were implemented and effectiveness. Use this FMEA as a combined tool for product FMEA to understand software defects that lead to recall initiation and time to recall duration and develop FMEA to identify vulnerabilities in organization or FDA oversight of the recall process to recommend and adapt continuous process improvements. Notes: Risk Priority columns have been removed. Process Name: Time to Recall for Software Related Recalls Product FMEA FM EA Step # Root Cause Potential Failure or Defect Mode Report Errors 1 Potential Causes of Failure(s ) Design or validatio n error. User error. Current Conditio ns Potential End (Patient/ User) Effect Patient/User Current Controls and Assumpt ions Resulting Conditions Recomme nded Mitigation Error not identified during design review. Reevaluate software design files and customer complaints to identify source of the issue. Correct with software upgrade. Review design file and perform software validation to verify. Review design file and Software Defects Trends Freeze Mode Design or validatio n error. User error. User Error not identified during design review. Interface Malfunction Algorith m error. User/Patient Error not found during Verificat ion or Propose d Mitigati on Respons ible Depart ment Verify report errors with inhouse testing or on-site maintena nce. Design Verify report errors with inhouse testing or on-site maintena nce. Monitor complain ts for Design Design 103 Revision : 02 Process Name: Time to Recall for Software Related Recalls validatio n or end user trial testing. Unable to view images Design or validatio n error. User error. iOS and Android application error Lack of validatio n controls during design. Incorrect results Algorith m error for data calculatio n. User Error not identified during design review. Patient/User Error not found during validatio n or end user trial testing. Patient/User Error not found during validatio n or end user trial testing. perform software validation to verify. Review design file and perform software validation to verify. Update software code and interface. Perform mobile application testing, validate new version and issue new software. Reevaluate software design files and customer complaints to identify source of the issue. Correct with software upgrade. recurring issues. Monitor complain ts for recurring issues. Design Monitor complain ts for recurring issues. Design Verify report errors with inhouse testing or on-site maintena nce. Design Process FMEA FM EA Step # 2 Root Cause Potential Failure or Defect Mode Potential Causes of Failure(s ) Potential End (Patient/ Organization ) Effect Current Conditio ns Current Controls and Assumpt ions Resulting Conditions Recomme nded Mitigation Verificat ion or Propose d Mitigati on Respons ible Depart ment Quality 104 Revision : 02 Process Name: Time to Recall for Software Related Recalls Inadequate Corrective Action procedures Organiza tion Protocol Inadequate complaints procedure Inadequate process for reporting of adverse events Inability to adequatel y address root cause identifica tion and remediati on. Inability to address customer feedback for customer resolutio n into design for patient safety. Inability to meet regulator y requirem ents. Correctiv e action procedur e may be in place but does not address all regulator y requirem ents. External complianc e assessment of procedure against CFR. Patient/Organi zation Complai nts procedur e in place to receive, but not fully investigat e reported defects. Review customer complaints process with customer service, quality, and design for gaps. Ensure process aligns to regulatory expectatio ns. Patient/Organi zation Adverse Events procedur e inadequat e to ensure reports are submitte d timely. External complianc e assessment of procedure against CFR. Patient/Organi zation Monitor CAPA reports for effective ness. Internal audit CAPA process at one year from complian ce assessme nt. Create metrics for quarterly review of complain ts data to ensure relevant issues are identified and escalated to design and recall team for trending and review. Monitor adverse events reports for timelines s and correct notificati on. Internal audit this process at one year from complian Quality and Custome r Service Quality 105 Revision : 02 Process Name: Time to Recall for Software Related Recalls Lack of process for notification of product action Inability to notify relevant stakehold ers of product action. Failure to follow procedure/pr ocess to meet regulatory requirement s Inability to maintain effective quality system to ensure patient safety and reduction of risk to harm. Patient/Organi zation Current process failing to notify regulator y authoritie s and key stakehold ers. External complianc e assessment of procedure against CFR. Establish new procedure including all stakeholde rs in recall. Patient/Organi zation Failure to comply to regulatio n could lead to regulator y enforcem ent action. Quality system not fully effective. External complianc e assessment of quality system. Identify relevant gaps and use quality plans and CAPA process or improvem ents. ce assessme nt. Monitor recall notices for timelines s and correct notificati on. Internal audit this process at one year from complian ce assessme nt. Track status updates quarterly through managem ent review. Quality and Regulato ry Senior Leadersh ip