that contribute little to outcomes. Finally, for many clinical areas, practitioners and researchers may be unable to reach consensus on best practices.
Measuring the outcomes of care is the third approach to quality assessment in the classic paradigm. Outcomes are the end results of care—the effect of the care process on the health and well-being of patients and populations. One outcomes list comprises “the five Ds”—death, disease, disability, discomfort, and dissatisfaction (Elinson, 1987). More positively they may be thought of as survival, states of physiologic, physical, and emotional health, and satisfaction (Lohr, 1988).
The philosophical shift in orientation from process measurement to outcomes has been dramatic in the late 1980s.4 For instance, the Joint Commission has launched an ambitious and far-ranging restructuring of its survey and accreditation procedures—the Agenda for Change—that is expected to rely heavily on outcome measurement (Joint Commission, 1987; Roberts, 1988). Ellwood (1988) has proposed a radical shift to “outcomes management,” by which he means collaborative action involving patients, payers, and providers in national information gathering and analysis to facilitate better, more “rational” decision making. He defines outcomes management as a “permanent national medical data base that uses a common set of definitions for measuring quality of life to enable patients, payers, and providers to make informed health choices” (p. 1555), and he puts particular emphasis on understanding the relationships between medical interventions and health outcomes and between health outcomes and money.
This emphasis on outcomes is a critical one. It will not, however, be easy to put into action, and it should never be seen as fully displacing process-of-care assessment. Greenfield (1989), for instance, cautions about the importance of choosing the right kinds of measures.
Outcomes offer a vast range of units of measurement: for example, death; hospital complication rates; functional capacities and performance; emotional health; cognitive functioning; and patient satisfaction, knowledge, and compliance.5 Such outcomes are measured directly and indirectly. Direct assessments rely on patient examination, physiologic data, patient self-report, and physician or other professional reports or records. Indirect evaluations can be based on utilization records (e.g., encounter forms and insurance claims), vital and health statistics, surveys, and the like; data sources can be the same as those used for process measures in some instances. Almost none of the outcomes mentioned above is amenable to measurement exclusively by insurance claims data, so researchers and policy makers have given substantial attention lately to nonintrusive outcomes, such as mortality or readmissions to hospitals. These proxy measures, typically taken