- 137 - 5. MOVING QUALITY MEASUREMENT FORWARD Annual health care expenditures in the United States exceed 13% of gross domestic product (GDP), or more than $1.5 trillion (Heffler, Smith et al. 2002). Despite a growing number of efforts to measure and report levels of health care quality resulting from these expenditures, useful information is neither uniformly nor widely available (Institute of Medicine 2001; McGlynn and Brook 2001). However, the information that is available suggests that millions of Americans fail to receive adequate quality care (Schuster, McGlynn et al. 1998). Consequently, there are large potential gains in welfare associated with improving the performance of the health care system. Quality of care measurement can serve both as a catalyst and a tool for such improvement. In 1999 Congress mandated that the Agency for Healthcare Research and Quality (AHRQ) develop and publish an annual national quality report on health care delivery, with the first report to be published in 2003. To respond to this mandate, AHRQ commissioned the IOM to conduct a study of how best to measure the overall quality of health care in the nation. In its resulting report, Envisioning the National Health Care Quality Report, the IOM suggested that quality measurement should address four components of health care – safety, patient centeredness, timeliness, and effectiveness. Effectiveness refers to the technical quality of care – that is, “providing services based on scientific knowledge to all who could benefit, and refraining from providing services to those not likely to benefit” (Institute of Medicine 2001). Among its many findings and recommendations, the report specifically highlights that data sources presently available to measure effectiveness are deficient. Clinically detailed information, such as that found in medical records, is needed for comprehensive measurement of technical quality. However, paper medical records are not a long-term solution for quality measurement (Institute of Medicine, 2001). The costs associated with locating and abstracting paper medical records are too large to sustain quality measurement on a routine, timely, and broad basis. Recognizing these constraints, this dissertation analyzes whether we are utilizing - 138 - claims data, which are widely available and affordable, to their fullest capacity for quality measurement. The findings from this study can be used to develop a road map to close the gap in information needed to measure technical quality. Untapped Potential of Claims Data for Quality Measurement The analysis in Chapter 3 determined that 186 of the QA Tools indicators (i.e., about one-third of the indicators) could be constructed with claims data, albeit some more accurately than others. This is a significant finding. As a comparison, HEDIS, the most common set of measures now used to measure performance, includes only 26 indicators of technical quality. Thus, the capacity of claims data for quality measurement is not being fully recognized. These missed opportunities for measurement with claims data could offer significant value to the health care system. Increasing the breadth of quality measurement has the potential to affect the behavior of payers, providers, and consumers in ways that would improve the quality of health care. There may even be a business case for expanding quality measurement with claims data for some types of care. In particular, it could be profitable for health plans or providers to do so to the extent that improving quality reduces costs (e.g., costs of treatment related to injuries that result from sub-standard quality, liability for sub-standard care), increases revenues from qualityconscious payers and consumers, or both. Additional measures of quality would reduce the risk of payers and providers shifting resources to a small set of indicators on which they know they will be measured. When using the leading indicator approach, such as represented by HEDIS, there is an opportunity for performance to look good on a small selection of indicators, while having unsatisfactory performance on the multitude of care processes not being assessed (Bodenheimer and Casalino 1999). By increasing the number of indicators that are routinely measured from 26 to 186, or even a substantial subset of the 186 selected on the basis of accuracy, the chance of this undesirable shifting of resources is minimized. - 139 - Consumers must find the information in performance reports salient for it to affect their behavior (Hibbard and Jewett 1997). By increasing the number of indicators routinely constructed with claims data, consumers are more likely to identify with indicators that are relevant to their health care needs and subsequently use comparative reports to select health plans and providers that offer the best value. What can and cannot be Measured with Claims Data? Using claims data for quality measurement is inherently limited to services that are reimbursed and by the standardized coding systems used in the claims files. Despite these limitations to quality measurement with claims data, this study found that one or more indicators could be constructed in 35 of the 37 clinical areas included in QA Tools.39 The ability to identify whether patients receive the indicated care is primarily determined by whether the services involved are reimbursed. This means that claims data can be used to measure quality when billable modes of care such as admissions, immunizations, laboratory services, medications, surgery, and visits are being assessed. In contrast, modes of care that are not typically billable such whether appropriate education was provided, history was taken, or physical exam was performed cannot be measured with claims data. For most clinical conditions, the indicated services will include modes of care that can be assessed with claims data. Determining which patients satisfy an indicator’s eligibility criteria is largely affected by whether the standardized codes in claims data sufficiently describe the clinical characteristics of patients. Since clinically detailed information such as signs and symptoms, laboratory results, and the findings of procedures, are not coded in claims data, for many indicators it is not possible to identify the patients who would be eligible for the indicated care. Although claims data can be used to assess a wide variety of clinical conditions and processes, the depth of measurement is limited. In the remainder of this chapter, the value of capturing health information electronically ___________ 39 No QA Tools indicators assessing the care for Cesarean Delivery or Osteoarthritis could be constructed with claims data. - 140 - and the types of information of high value to quality measurement are discussed. Building a Health Information Infrastructure There have been many calls to develop and implement electronic data systems to facilitate quality measurement and satisfy other information needs in the health care system (Institute of Medicine 1991; The President's Advisory Commission on Consumer Protection and Quality in the Health Care Industry 1998; Institute of Medicine 2001; Institute of Medicine 2001). However, payment issues, rather than clinical needs, have driven investment in health information technology. Therefore, electronic billing systems are generally much more prevalent and relatively more standardized than clinical information systems. Although the benefits of electronic clinical data are difficult to quantify, evidence has begun to accumulate about the ability of connected and automated information to improve the efficiency and effectiveness of clinical care. The availability of information across providers and settings, for example, allows access to longitudinal patient information that can help reduce duplicate tests and avoid errors associated with decisions based on missing information (Schiff and Rucker 1998; Schiff, Aggarwal et al. 2000). Using automated clinical information in the form of decision support tools and reminder systems also improves care and reduces errors (Bates, Leape et al. 1998; Dexter, Perkins et al. 2001), and health plans and providers are beginning to seriously consider the implementation of electronic medical records (EMRs). In 2001, 13% of providers had a fully operational EMRs in place, and another 53% reported they were either beginning to install the hardware and software for EMRs or have planned EMR implementation (National Committee on Vital and Health Statistics 2001). Only a few private and public sector provider organizations have made strategic moves toward fully integrated information systems. For example, Kaiser Permanente is developing a Web-based system that includes a nationwide clinical information system, patient communication with doctors and nurses for advice, online guidelines and protocols for providers, and all administrative functions. Intermountain Health Care - 141 - has a well-integrated hospital-based information system with clinical decision-support capabilities (Gardner, Pryor et al. 1999). The Department of Veteran Affairs and the Department of Defense’s TRICARE40 program have the ability to measure and improve quality through continuous feedback and the application of computerized decision support systems (Institute of Medicine 2002). Fully integrated health information systems have the potential to facilitate quality measurement and improve the delivery of efficient and effective care. However, a giant leap is not practical for most health care providers and delivery systems. A series of incremental steps are required to make widespread automation of health information a reality. Substantial effort has gone into addressing two of the greatest impediments to electronic clinical information -- privacy concerns41 and data standards.42 There have also been attempts to identify the types ___________ 40 TRICARE is the health care delivery system for the US military. 41 Information security technologies, such as encryption, authentication of both the sender and receiver of data, and audit trails to detect unauthorized users, are available to protect the privacy of individually identifiable health information (U.S. General Accounting Office 1999; Detmer 2000). In addition to the technical capacity to protect privacy, DHHS finalized rules about the confidentiality of individually identifiable health information in 2002 (Code of Federal Regulations 2002). 42 A number of US standard development organizations have developed standards for the exchange and management of data. Some of these transaction standards, such as Health Level Seven (HL7) and Digital Imaging and Communications in Medicine (DICOM), are in widespread use in the US and other countries. HL7 is a Standards Developing Organization (SDO) that produces standards, or protocols, for the exchange, management, and integration of clinical and claims data. HL7 does not develop software for the exchange of data, but rather develops a messaging standard that enables disparate health care applications to exchange key sets of clinical and claims data (http://www.hl7.org; Accessed 2/18/2002). The DICOM Standards Committee creates and maintains standards for communicating medical images and their associated information across disparate health care applications (http://medical.nema.org/dicom/geninfo/dicom_strategy/Strategy_2002-0809b.htm; Accessed 1/1/2003). Standards for codes that give specific meanings to the content of these messages have also been developed. As described in Chapter 3, standardized coding systems such as SNOMED and LOINC have been developed, but adoption of these standards has been limited. Although LOINC is in the public domain, using SNOMED codes in any application requires a licensing fee that might prohibit widespread adoption. - 142 - of data elements that should constitute the core of electronic data collection efforts in health care (National Committee on Vital and Health Statistics 1996; Schneider, Riehl et al. 1999). However, the relative values of these data elements have not been evaluated. Therefore, as health care delivery systems and providers look to make investments in clinical information systems, there is no guide to the types of information that should be obtained first. What Additional Data Elements Contribute the Most to Quality Measurement? The analyses in this dissertation were framed by RAND’s QA Tools system, which has been recognized by the IOM as a comprehensive approach to measuring the quality of care (Institute of Medicine, 2001). The QA Tools system allowed me to consider a large set of indicators that represent what we would like to know about the delivery of care without the bias of what information is typically considered to be unavailable from claims data. By identifying the data elements required to assess quality, this dissertation described what can be measured currently with electronic data and identified the types of data elements that are required to improve the capacity for quality measurement. The analyses from Chapter 3 suggest that about 40% of the information needed for comprehensive quality measurement is not available from claims data. Information that is typically unavailable includes: • Counseling and educational discussion • Elements of physical exam performed during an encounter and findings • Laboratory results • Patient history • Procedure findings • Signs and symptoms These types of missing information should be included in health information systems. But, which data elements would have the most important effects on quality measurement? The findings in Chapter 3 suggest that the addition of no one type would provide a major - 143 - contribution to the capacity for quality measurement. However, among the five types of information investigated in Chapter 3 –laboratory results, procedure findings, vital signs, and signs and symptoms- access to information about patients’ signs and symptoms would leverage quality of care measurement more than other types of information, however. Further, Chapter 4 highlights that using diagnostic criteria to identify who should receive care is the most significant source of inaccuracy. These findings suggest that as health care delivery systems and providers consider their next set of investments in health information systems, they should consider the electronic storage of problem lists with standardized clinical codes such as SNOMED. The capacity to classify people into clinically meaningful groups would improve our ability to assess whether appropriate care is being delivered to those most likely to benefit. This information would also help providers better manage their patients’ clinical care and provide information to implement decision support systems. The capacity for individual health care delivery systems and providers to measure their performance is not sufficient. The power of quality measurement to inform choice and motivate change comes from having performance information reported across the entire health care system. As investments are made in clinical information systems, the ability to integrate data across delivery systems will be central to broad and routine quality measurement. A coordinating structure to standardize health information is needed. A health care information infrastructure could evolve through a Federal mandate to support the National Healthcare Quality Report or market pressure from health care purchasers, such as that used to develop HEDIS. Regardless, resources and participation from both the public and private sectors are needed to move quality measurement forward and thereby improve the quality of care delivered in the US. In sum, while widespread quality measurement relying extensively on medical records would provide more broad-based and accurate information, the associated costs currently appear prohibitive to most organizations. This economic reality provides the key motivation for this dissertation, which has explored less costly, albeit less - 144 - informative, ways to expand the scope of quality measurement. There are four broad options over the next several years: (1) the status quo; (2) expanded use of claims data, possibly augmented with the kinds of information analyzed in chapter 3; (3) expanded use of medical records; and (4) expanded use of both claims data and medical records. The benefits and costs of various versions of the last three options relative to the status quo, and relative to each other, are important, but open, questions.