PMC patient severity scale: derivation and validation - Patient Management Category

Health Services Research, August, 1994 by Wanda W. Young, Susan Kohler, Jeanne Kowalski

Analysis of Severity Level Combinations, with Duplication

Next, all combinations of patients receiving two or more PMC assignments of the same severity level, within each of the ten statistically distinct combinations from the previous phase of testing, were analyzed. Specifically, using the multiplicative rule within these ten combinations and allowing duplication of the same severity level, 66 severity level combinations were identified. Duplication of the same severity level was limited to three;(6) that is, if a patient had four PMCs, each with the severity level 2, the patient is represented in the combination {2&2&2}. To test for homogeneity with respect to LOS, distributions of LOS were examined and the same statistical analyses as in the previous phase of testing (Kruskal Wallis H-test and Scheffe's test) were conducted between and within all combinations of various sizes.

Based on this analysis, all combinations were reordered into seven severity levels. Table 2 illustrates the final composition of each of these seven severity levels. The result is a PMC Severity Scale which is an ordinal scale, with Level 7 representing the greatest likelihood of death and major disease burden. The scale quantifies the severity of each of the patient's disease(s) and accounts for the impact of all coexisting conditions and complications that are typically treated in general acute care hospitals.

VALIDATION OF PMC SEVERITY SCALE

Although the development of the PMC Severity Scale incorporated both clinical judgment and statistical analyses, its overall content validity is primarily derived from the clinical framework of the Patient Management Category (PMC) Classification System. PMCs define patients' clinical conditions accurately relative to other diagnosis-based classifications (Thomas, Holloway, and Guire 1993; Young, Macioce, and Young 1990), identify specific TABULAR DATA OMITTED comorbid conditions and complications of each patient (Young 1984), and can uncover the clinical heterogeneity within other statistically homogenous patient groups (Charbonneau, Ostrowski, Poehner, et al. 1988).

Because the distinct concepts of clinical specificity, comorbidity, intensity, and severity of illness are separately operationalized in the PMC system, it has been possible to assess the impact of comorbidity on severity in the derivation of the PMC Severity Scale. The PMC system identifies the variety of comorbid combinations that are present, and the PMC Severity Scale quantifies the additive, hierarchical, or interactive impact of each patient's specific combination of diseases and complications. That is, in the PMC system, whether a patient with comorbid conditions is more severely ill than a single-disease patient depends on the patient's specific comorbid combination.

In addition to the clinical basis of the PMCs and related content validity of the PMC Severity Scale, a number of empirical analyses have been conducted to test the validity of the PMC Severity Scale and its reliability over time and across geographic regions. To test the construct validity of the resultant scale, the death rate and average length of stay were examined by severity level for all patients in the database used for development (Maryland 1989). Mortality rates and length of stay (as an indirect measure of morbidity) are frequently used as surrogates of severity in large population databases. Table 3 illustrates the direct relationship of both mortality rates and average lengths of stay with the severity scale using the developmental 1989 Maryland database. Results of the Kruskal Wallis H-test indicate a significant difference (p !is less than^ .01) in both LOS and mortality across severity levels.