Imagine a hospital announcing that it has bought new equipment, followed its treatment protocols, and sent patients home healthier. All three statements sound reassuring. But they answer different questions: what the hospital could do, what its staff actually did, and what happened afterward. Which statement would justify calling the care good?
Avedis Donabedian approached that problem in 1966, in Evaluating the Quality of Medical Care. This reading uses the journal's 2005 reprint, which preserves the original wording. Under “Approaches to Assessment,” he discusses outcomes first, then process, then structure—the reverse of the familiar sequence in which his framework is usually taught.[1]
That order is worth lingering over. My reading is that he starts with the thing we want to explain and works backward toward the evidence that might explain it. The categories become questions to investigate, rather than steps that automatically deliver success.
A vocabulary assembled under pressure
Donabedian's own account of the commission is refreshingly uncertain. In a recollection dated 29 November 1982, published in Current Contents in February 1983, he described being asked to review the limited literature on assessing care. His essay belonged to a US Public Health Service project commissioning 15 papers across health services research. He did not recall being treated as an established authority.[2]
The recollection locates his achievement in giving a scattered field an organized language. It also identifies deliberate omissions: the original analysis assumed a relationship between good care and its results, and set monetary cost aside. His later work enlarged that model.[2]
This makes a difference to how the paper should be read. A framework assembled to organize inquiry can be useful without being a complete theory of a health system. Remembering the commission keeps the subsequent fame from flattening the document into a universal answer.
The result still needs an explanation
On page 694 of the reprint, Donabedian calls outcomes the “ultimate validators” of care, while warning that a result does not locate the strengths or failures that produced it.[1]
The Institute of Medicine's 1999 statement develops the distinction clearly: good care increases the probability of a desirable result; it cannot guarantee that result. Recovery can occur despite poor care, and illness can prevail despite good care. Comparisons therefore need attention to differences in patients' starting risks and illness severity.[3]
Consider two imaginary wards, one receiving relatively straightforward cases and another accepting difficult referrals. A difference in their outcomes would be a reason to investigate. It would not, by itself, settle which team performed better. The practical question becomes what happened to comparable patients, over a comparable period, under comparable definitions of recovery.
The same statement insists that a useful process measure must connect to an outcome people value. Counting an activity is insufficient if the activity has no demonstrated bearing on the benefit being sought.[3] A hospital could become excellent at completing a form while learning very little about the experience the form supposedly represents.
Capacity, action, consequence
The 1990 Institute of Medicine report on Medicare describes structural measures as evidence of a provider's presumed capacity: facilities, qualifications, staffing arrangements, and organizational authority. Process assessment looks at the care delivered. Outcomes concern the effects on health and well-being. It treats process and outcome assessment as complementary.[4]
Return to the imaginary hospital. Owning equipment establishes availability. Showing that staff used it appropriately establishes something about practice. Discovering whether patients benefited asks a further question. Moving from one statement to the next requires evidence; attaching three labels does not supply it.
The distinction can also identify the next investigation. If equipment is available but indicated care is repeatedly missed, another purchase might leave the failure untouched. If the expected actions occur but patients fare poorly, the inquiry has to widen. The treatment assumptions, the implementation, the patient group, and the chosen outcome all become candidates for scrutiny. These are illustrations of how to use the categories, not findings about any particular hospital.
The measurement has a working life
In 1987, Donald Berwick and Marian Gilbert Knapp argued that quality assessment remained too detached from everyday management. Writing from their work at Harvard Community Health Plan, they asked how measurement could become useful to people making and changing decisions about care.[5]
They also described a tension within review itself. Expert judgment can accommodate complicated cases, but its reasoning may be difficult to see. Explicit criteria make assessment clearer and allow trained reviewers to apply common standards; poorly designed criteria can oversimplify clinical circumstances.[5]
A neat score consequently has two histories worth examining: the care that produced the record and the review that produced the score. In the original paper's discussion of clinical records, Donabedian likewise distinguishes documentation as an aspect of good practice from documentation as evidence about other aspects of care.[1] An empty field may identify a recording failure without establishing everything that happened at the bedside.
What remains after the score
The 1990 Medicare report separates assessment from assurance. Measuring a problem begins the work; a response must investigate causes, implement changes, and check whether those changes helped or introduced other problems.[4] The score needs someone with both the responsibility and the practical means to act on it.
Looking back in 2016, Berwick and Daniel Fox warned against reducing Donabedian to a technical classification. Their appraisal emphasizes his interest in governance and management while acknowledging how subsequent thinking about patient participation, information systems, and care as a whole system exceeded his original account.[6]
That is a productive way to preserve a classic: use its distinctions, keep its boundaries visible, and continue the inquiry. A hospital can report its resources, its actions, and its results. The serious conversation begins when it explains how those observations fit together—and what it will change because of them.
Sources
- Avedis Donabedian, “Evaluating the Quality of Medical Care” (1966), reprinted in The Milbank Quarterly 83(4), 2005—especially pp. 691–696: scope, assessment categories, outcomes, and clinical records.
- Avedis Donabedian, Citation Classic commentary, Current Contents, 7 February 1983; recollection dated 29 November 1982—commission, intellectual purpose, and acknowledged exclusions.
- Institute of Medicine, Measuring the Quality of Health Care: A Statement by the National Roundtable on Health Care Quality (1999)—probabilistic outcomes, fair comparisons, and the connection between process and benefit.
- Institute of Medicine, Medicare: A Strategy for Quality Assurance, Volume 1 (1990), chapter 2, “Concepts of Assessing, Assuring, and Improving Quality”—assessment categories and the cycle from measurement to corrective action.
- Donald M. Berwick and Marian Gilbert Knapp, “Theory and Practice for Measuring Health Care Quality,” Health Care Financing Review (1987 supplement), pp. 49–55—management relevance and implicit versus explicit review.
- Donald Berwick and Daniel M. Fox, “Evaluating the Quality of Medical Care: Donabedian's Classic Article 50 Years Later,” The Milbank Quarterly 94(2), 2016—governance, later developments, and limits of a purely technical reading.
- University of Michigan, archival portrait of Avedis Donabedian, photograph undated; image reproduced by EMS World in 2022—photographic asset.