A Study of the Relationship Between Severity of Illness and Hospital Cost in New Jersey Hospitals; Critical Reaction; The Authors Respond; Commentary: Are We Splitting Hairs Over Splitting DRGs?

• Published in: Health services research Vol. 27; no. 5; p. 587
• Main Authors: Averill, Richard F, McGuire, Thomas E, Manning, Beatrice E, Fowler, Daniel A, Horn, Susan D, Dickson, Pamela S, Coye, Molly J, Knowlton, David L, Bender, Judith A, Feldman, Roger
• Format: Journal Article
• Language: English
• Published: Chicago Blackwell Publishing Ltd 01-12-1992
• Subjects: Data collection; Diagnosis related groups; DRGs; Health care expenditures; Health care industry; Hospital costs; Hospitals; Illnesses; Medical records; Patients; Statistical analysis; United States > US; New Jersey
• ISSN: 0017-9124; 1475-6773

The New Jersey Department of Health conducted a severity of illness evaluation study in which severity of illness, diagnosis related group (DRG), and uniform cost information were collected for 76,798 patients in 25 hospitals. Severity of illness was measured using the Computerized Severity Index (CSI) and was found to be a significant determinant of hospital cost in 76 DRGs that accounted for 41.4% of the total direct hospital patient care costs and 27% of the patients. The addition of CSI severity levels to the 76 DRGs reduced the coefficient of variation of cost in these DRGs by 17.4% and improved the overall reduction in variance of cost within the 76 DRGs by 38.2%. It is not recommended that the study be published in its present form due to the built-in bias of the "data-splitting" technique. This can result in an improvement although the split makes no medical sense. A simulation shows that the expected reduction in variance from random data-splitting is, in general, small and that it declines as the size of the population increases. It is shown that the expected reduction in variance is positive, even when a population is randomly split into several groups.