496 tions. Hundreds of variables related to patient factors (physiologic and psy- chosocial abnormalities presented at each admission), medical care process factors, and outcome factors can be ex- amined by using a comprehensive study method called Clinical Practice Improvement (CPI). 1 Clinicians exam- ine resulting associations to objectively evaluate the effects of treatments given to similarly ill patients. Without all 3 types of data, clinicians cannot tell if the outcomes achieved are the result of process steps or differences in patient severity levels. 1-15 Determination of effective care prac- tices requires a method to assess rela- tionships among patient variables, treatments, and subsequent outcomes. From ISIS, Inc, the Institute for Clinical Outcomes Research, and the Department of Pediatrics, Primary Children’s Medical Center, Salt Lake City, Utah; the Department of Pediatrics, Arkansas Children’s Hospital, Little Rock; and the Department of Pediatrics, University of Virginia Health Sciences Center, Charlottesville. Supported through two contracts (Phase I and Phase II) with the Center of General Health Ser- vices Extramural Research of the Agency for Health Care Policy and Research (AHCPR), from October 1994 to July 1998. ISIS, Inc, entered into these two contracts, received the funds, and performed the study thereunder. Susan D. Horn is a full-time employee and stockholder of ISIS. Julie Gassaway and Randall Smout are full-time employees and stock option holders of ISIS. Submitted for publication Sept 24, 2001; revision received Mar 6, 2002; accepted May 30, 2002. Reprint requests: Susan D. Horn, PhD, Institute for Clinical Outcomes Research, 2681 Parleys Way, Ste 201, Salt Lake City, UT 84109-1630. Copyright © 2002, Mosby, Inc. All rights reserved. 0022-3476/2002/$35.00 + 0 9/21/126925 doi:10.1067/mpd.2002.126925 Development of a pediatric age- and disease-specific severity measure Susan D. Horn, PhD, Adalberto Torres, Jr, MD, Douglas Willson, MD, J. Michael Dean, MD, Julie Gassaway, MS, RN, and Randall Smout, MS ACSI Admission CSI discrete score ACSIC Admission CSI continuous score CPI Clinical Practice Improvement CSI Comprehensive Severity Index DORA Dynamic Objective Risk Assessment DRG Diagnosis-related group ICD-9-CM International Classification of Diseases, 9th ed–Clinical Modification LOS Length of stay MCSI Maximum CSI discrete score MCSIC Maximum CSI continuous score NACHRI National Association of Children’s Hospitals and Related Institutions NOS Not otherwise specified OR Operation PICU Pediatric intensive care unit PIM Pediatric Index of Mortality PRESS Sum of squares of predicted residual errors PRISM Pediatric Risk of Mortality PSI Physiologic Stability Index ROC Receiver operating characteristic SPrR Sum of PRESS residuals SSR Sum of squares residuals TISS Therapeutic Intervention Scoring System Severity scoring systems attempt to quantify patient variables to allow ob- jective comparison of treatment and outcomes in similar patient popula- Objectives: To adapt the adult Comprehensive Severity Index (CSI) for hospitalized pediatric patients and evaluate the ability of the CSI to predict common outcomes. Study design: Adult CSI was modified by a panel of pediatric subspecialists from 10 children’s hospitals. Predictive power was evaluated by using retro- spective data collected from 16,495 randomly selected children admitted to these hospitals from April 1995 through September 1996. Outcomes were mortality, length of stay (LOS), and cost. Results: Admission CSI score predicted mortality well (Hosmer-Lemeshow tests: P = .41-.98) and discriminated well (area under receiver operating char- acteristic [ROC] curve range = 0.80-0.99) within 9 case-mix groups with ≥10 deaths (P < .0001). Maximum CSI score explained the variation in LOS (r 2 = 0.13-0.67) and cost (r 2 = 0.08-0.73) within 32 case-mix groups (P < .005). Significant differences existed in admission and maximum average CSI scores across sites in 26 and 29 of 32 case-mix groups, respectively (P < .05). CSI had better predictability than Pediatric Risk of Mortality. Conclusions: The age- and disease-specific pediatric CSI score correlates highly with LOS, cost, and mortality in hospitalized children and can help determine the best clinical practices for specific diseases and adjust for differ- ences in severity of illness across providers. (J Pediatr 2002;141:496-503) See editorial, p 463.