Clinical Engineering Engineers in the Modern Academic Medical Center Design Disasters Consequences of Blunders, Bad Luck, and Bias Patrick Norris, Ph.D. Assistant Professor of Surgery, Biomedical Engineering patrick.norris@vanderbilt.edu Clinical Engineering Why do hospitals need engineers? • Definition • Past, Present, Future • Examples – Facility Design – Biomedical Devices – Information and Technology Management • Clinical Research, Quality Improvement Biomedical Electronics Definition Clinical Technology Service Biomedical Engineering Definition The American College of Clinical Engineering: A professional who supports and advances patient care by applying engineering and management skills to healthcare technology. Definition: Hospitals need engineers when technology requires: • Special (non-trade/craft skills) customization or maintenance • Complex selection criteria • Modification of existing facilities or systems, or special design of new ones • Design and analytic skills, professional credentials, etc. differentiate engineers from technicians, craftspeople, clerical, administrators, etc… Examples: Past • Einthoven: EKG, early 1900’s • Other examples: – Day to day heat, AC, water, electricity, etc. Examples: Present • Infrastructure Design – Typical: Water, Electrical, HVAC, Telecom – Special: Medical Gas, Sample Handling – Structural: Imaging Systems • Biomedical Devices – Selection, integration, tracking – Maintenance is becoming a sophisticated trade/craft skill • Information Future • Information – Medical Informatics – 6 VUSE PhDs • Integration – People – IT Systems – Medical Devices • Regulation – Privacy, Safety, Efficacy • Across Multiple Healthcare Systems Grimes SL, IEEE Engineering in Medicine and Biology Magazine, March/April 2003 p.91-99 Clinical Research • SIMON Project – (Signal Interpretation and Monitoring) – Ongoing since 1994 • Seeks to Advance: – Medical Monitoring Technology – Critical Care – Scientific Knowledge • Clinical Engineering Component Trauma • • • • 5th Leading Cause of Death (1st Under 45) 8% of Medical Expenditures (rank: 3rd) All Age and Socioeconomic Groups VUMC – Only Level 1 Facility, 65,000 Square Miles – 3500 Annual Admissions – 800 to Trauma ICU, ~10% Mortality Patient Monitoring • Cushing, early 1900’s: – Importance of Monitoring and Recording Vital Signs • Technology Has Advanced • Fundamentally, Clinical Strategies Remain Unchanged – Intermittent Recording – Manual Interpretation Tools for Dense Physiologic Data Management Four Engineering Challenges • Data Collection – Interfaces to a Variety of Devices – Remote Locations • Storage – Clinical Applications - Short-Term – Research Applications - “Forever” • Processing – Time-Critical Tasks (Clinical Decision Support) – Research Analysis • Architecture – Integration, Reliability, Scalability, Flexibility… SIMON Data Capture • Philips CareVue – Routine, Automatic Vital Signs Capture – HR, ABP, PAP, CVP, ICP, CPP, PAP, SaO2 – Episodic Waveform Capture • Edwards Vigilance – CI, EDVI, temp, SvO2, etc. • Alaris IV Pump (near future?) SIMON Data Storage • Relational Database – Time Constraints w/ Limited Resources – Adaptive Sampling, ~0.25-1Hz Storage • 5500+ TICU Patients – Reliably Identified, Linked to Outcomes • 450,000+ Continuous Hours • Grows by: – 2 Million+ Data Points/Day – ~70 Patients/Month Daily Reports Data Display Alerts • Effective Alerting – Right Information – Right Person – Right Time • Process – Event – Alert – Notification – Response SIMON Architecture • Modular, Simple Components – Scalable – Reliable – Flexible • Time-Constrained SIMONT1 SIMONS1 Data Collection Modules (1 per device) Data Collector ODBC System Mgr. SQL 2k Bed 1 Database Mgr. Bed 2 Event Engine SIMONW1 Bed 3 (Secure WWW Server) trauma.mc.vanderbilt.edu Alert Engine Notify Engine Digi Driver • • • Simon Packet Format Census Agent Census Monitor sFTP Report Engine sFTP VUMC Census Bed 14 VUMC StarPanel Research Hypotheses New measurements, available through techniques of dense data capture and analysis, will: • Identify failure of communication pathways (uncoupling) • Linking systems, organs, cells, proteins, and genes • Illuminate underlying control mechanisms • …especially in the critically ill Short-Term HRV - Survival Percent Time, entire stay 1.5 N = 825 1.0 0.5 0 0 0.5 1.0 1.5 2.0 2.5 3.0 5-minute HR Standard Deviation 3.5 Short-Term HRV - Death Percent Time, entire stay 1.5 N = 98 1.0 0.5 0 0 0.5 1.0 1.5 2.0 2.5 5-minute HR Standard Deviation 3.0 3.5 Short-Term HRV - Combined Percent Time, entire stay 1.5 N = 923 1.0 0.5 0 Time normalized within outcome group 0 0.5 1.0 1.5 2.0 2.5 3.0 5-minute HR Standard Deviation 3.5 Design Disasters Consequences of Blunders, Bad Luck & Bias • What is a Design Failure? • Why Do They Happen? • Examples – “Recipes for Design Disasters” – Space Program – Transportation – Medical What is a Design Failure? • Elements of Establishing Defect: – Identify the design defect – Establish a causal link to harm or cost – Identify alternate designs (correctable) – Compare to similar products • “A product does not have a design defect when it is safe for any reasonably foreseeable use and meets all applicable functional specifications.” Geddes, Medical Device Accidents With Illustrative Cases Example Design Defect (probably from urban legend) Nurses in Pelonomi Hospital, South African hospital were baffled that every Friday morning the patient in one particular bed would be found dead! Investigation revealed that the cleaning person would unplug that bed’s life support equipment, in order to plug in her floor polisher when she did the floors each Friday. When finished, she would plug the equipment back in unaware that the patient was now dead. Example Design Defect Definition • Identify defect • Causal link • Alternate designs • Comparison Pelonomi Hospital Legend • Life support equipment could be unknowingly unplugged • Staff were not alerted when machine unplugged, patient died • Alarms and batteries • All life-critical equipment offered by vendors X,Y,Z have alarm & battery backup What is a Design Failure? • There are plenty of definitions • Numerous example cases • In the end, failures are debatable – Ultimately, court may have to decide – With testimony from experts – Sometimes difficult to separate liability from design flaw – Negligence is a legal, not technical, term Why Do Designs Fail? At least three types of factors • Blunders (Human Error) – “Everyone makes mistakes” • Bad Luck (Random Effects) – “S*** happens” • Bias – People sometimes believe what they want to, irrespective of facts – Especially when money, power, relationships are involved Example: $125M Blunder • 1999 Mars Orbiter • JPL, Lockheed • Metric vs. English units • Erroneous orbital entry calculation – engine burn time Example: Bad Luck (?) • Weather: A random effect • Dense fog on I-75 – 99 vehicle pile-up in TN – Killing 12, injuring 56 • Initially weather blamed • Then local paper mill – $13.5M settlement – Once = bad luck – Many times = negligence? Example: Bad Luck • Tacoma-Narrows bridge • Unforeseeable consequence of lightweight design, wind profile • No human deaths • $5.2M in 1940, ~$70M today • (Insurance paid) Types of Bias • “Statistical” – Sampling – Multiple comparisons – Repeated measurements • Psycho-Social – Groupthink – “Corpthink” Examples: Statistical Bias • More people die in hospitals than anywhere else, therefore don’t go to the hospital! (unfair sampling) • Similar situation: A medical device designed only for the critically ill • Randomized, controlled trials are part of the answer Examples: Statistical Bias • Suppose you design a device that will roll a six every time – how many times do you need to test it? • Which results do you report? • Increasingly an issue in medical drug and device trials • 95% significance (p<.05) means that 1 in 20 studies is a false-positive Psycho-Social Biases • Individual – Primacy: The first option mentioned seems best – Recency: The last option seems best • Group – Groupthink: Consensus rules – “Corpthink”: Desire to please those higher in the chain of command NASA: Ripe for Disaster? • Huge shift in corporate culture – Space race: Do it at any cost – Increasing cost concerns, cuts, downsizing, resource pressure, etc. • Feynman, Challenger Disaster Report – Engineer estimate of catastrophic failure: 1 in 100 – Management: 1 in 100,000 – “What is the cause of management’s fantastic faith in the machinery?” More Design Failures • “Recipes for Disaster” – Ignition Source + Flammable Material –… • More Examples – Transportation – Space Program – Software Hindenburg • German airship • Caught fire while landing in 1937 • Design defect: – Hydrogen? – Skin? http://www.youtube.com/watch?v=F54rqDh2mWA Apollo 1 • Pad fire during test • Killed 3 astronauts • Design defects: – 31 miles of electrical wire – Pressurized pure oxygen environment – Flammable materials – Substandard wiring Medical Devices & Fire Ignition Source • Electrocautery • Nerve stimulators • Short-circuit • Electrostatic discharge • Cigarettes • … Flammable Materials • Anesthetic gas – not so much today, ex. O2 • Gases in the body, especially GI system – Geddes reports ~10 cases of GI explosions during procedures, some lethal! • Bedding, clothing • Bandages • Cleaning solutions, solvents, etc. Medical Software Design • What type of [medical] technology is least regulated? – Software – There is no professional-level (i.e. PE) certification for software engineering – Less regulation than devices/drugs Medical Software Design • Design failures are being publicized • Computerized Physician Order Entry – Cedars-Sinai software rollout – Multi-million dollar project scrapped – Software “endangered patient safety” – This story is not unique • Privacy issues • Will software design failures increase? Summary – Clinical Engineering • Definition of clinical engineering • Engineers’ role in the hospital? – Technology design, management – Increasingly, information management – Clinical research, i.e. VUMC Trauma • Differences between engineering and trade/craft skills (design & analysis) Summary – Design Disasters • Geddes definition of design failure: – Identified defect – Causal link to harm – Available alternative – Deficiency w/ respect to other products • 3 factors in design disasters: – Human error (blunder) – Random effects (bad luck) – Bias Sample Questions Which is not an aspect of establishing design failure (according to Geddes)? What factor best differentiates engineers from trades/craftspeople? A. B. C. D. A. Design and analytic skillset B. Professional ethics C. Ability to work in highly regulated fields D. Salary Identified defect Causal link to harm Negligence Feasible alternative design Sample Questions What kinds of bias is most likely encountered by an individual doing statistical analysis of complex data? A. B. C. D. Unfair sampling Groupthink Recency All of the above According to Feynman’s appendix to the Challenger disaster report, NASA engineers estimate probability of failure at about 1 in ________, compared to management’s 1 in ________ . A. 10, 10000 B. 1000, 1000 C. 100, 100000 D. 10000, 100 References/Sources • • • • • • • • • • • • • • • • clinicalengineering.duhs.duke.edu/ cms.clevelandclinic.org/anesthesia/body.cfm?id=124 www.healthsystem.virginia.edu/internet/clinical-eng/ www.wikipedia.org www.ceasa-national.org.za/ www.mc.uky.edu/clinicalengineering/ cms.clevelandclinic.org/anesthesia/body.cfm?id=156 www.uams.edu/ClinEng/default.aspx simon.project.vanderbilt.edu/ tafkac.org/medical/hospital_cleaning_lady.html www.cnn.com/TECH/space/9909/30/mars.metric.02/ mars.jpl.nasa.gov/msp98/orbiter/ www.douglasjfeeslaw.com/achievements.jsp gtresearchnews.gatech.edu/reshor/rh-ss01/fog.html www.ralentz.com/old/space/feynman-report.html youtube.com patrick.norris@vanderbilt.edu