A Framework for the Development and Interpretation of Different Sepsis Definitions and Clinical Criteria

Abstract

Although sepsis was described more than 2,000 years ago, and clinicians still struggle to define it, there is no "gold standard," and multiple competing approaches and terms exist. Challenges include the ever-changing knowledge base that informs our understanding of sepsis, competing views on which aspects of any potential definition are most important, and the tendency of most potential criteria to be distributed in at-risk populations in such a way as to hinder separation into discrete sets of patients. We propose that the development and evaluation of any definition or diagnostic criteria should follow four steps: 1) define the epistemologic underpinning, 2) agree on all relevant terms used to frame the exercise, 3) state the intended purpose for any proposed set of criteria, and 4) adopt a scientific approach to inform on their usefulness with regard to the intended purpose. Usefulness can be measured across six domains: 1) reliability (stability of criteria during retesting, between raters, over time, and across settings), 2) content validity (similar to face validity), 3) construct validity (whether criteria measure what they purport to measure), 4) criterion validity (how new criteria fare compared to standards), 5) measurement burden (cost, safety, and complexity), and 6) timeliness (whether criteria are available concurrent with care decisions). The relative importance of these domains of usefulness depends on the intended purpose, of which there are four broad categories: 1) clinical care, 2) research, 3) surveillance, and 4) quality improvement and audit. This proposed methodologic framework is intended to aid understanding of the strengths and weaknesses of different approaches, provide a mechanism for explaining differences in epidemiologic estimates generated by different approaches, and guide the development of future definitions and diagnostic criteria.

Figure 1

The “zone of rarity” problem: ideal and typical distributions of surface phenomena (clinical and biologic features) among patients with and without disease. Panels A and B illustrate situations in which a surface phenomenon (e.g., a single blood test) or set of phenomena (e.g., a combination of clinical features and blood tests) is used to separate a population into those who do and those who do not have a particular disease. Ideally (Panel A), there would be a large zone of rarity where few individuals would exhibit the test result or constellation of features at the border between health and disease. However (Panel B), most tests or combinations of tests and features are expressed on a continuum, with no zone of rarity. For example, the distribution of white blood cell count values across a population of hospitalized patients will not exhibit a zone of rarity near the upper limit of normal. Rather, many patients will have borderline-elevated values. Panel C and D show the corresponding distributions for sepsis, where surface phenomena classify patients with both infection and organ dysfunction. Although the ideal criteria (Panel C) for both infection and organ dysfunction would have clear zones of rarity, neither domains have such criteria (Panel D). For example, most organ dysfunction measures, like measures of infection, are expressed on a continuum with many patients exhibiting borderline values.