Comorbidity Measures for Use with Administrative …

1 Comorbidity Measures for Use with Administrative DataAuthor(s): Anne Elixhauser, Claudia Steiner, D. Robert Harris and Rosanna M. CoffeySource: Medical Care, Vol. 36, No. 1 (Jan., 1998), pp. 8-27 Published by: Lippincott Williams & WilkinsStable URL: .Accessed: 29/09/2013 07:16 Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at ..JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range ofcontent in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new formsof scholarship.

2 For more information about JSTOR, please contact .Lippincott Williams & Wilkins is collaborating with JSTOR to digitize, preserve and extend access to This content downloaded from on Sun, 29 Sep 2013 07:16:33 AMAll use subject to JSTOR Terms and ConditionsMEDICAL CARE Volume 36, Number 1, pp 8-27 01998 Lippincott-Raven Publishers Comorbidity Measures for Use with Administrative Data ANNE ELIXHAUSER, PHD,* CLAUDIA STEINER, MD, MPH,t D. ROBERT HARRIS, PHD,: AND ROSANNA M. COFFEY, PHD,? OBJECTIVES. This study attempts to develop a comprehensive set of Comorbidity Measures for use with large Administrative inpatient datasets.

3 METHODS. The study involved clinical and empirical review of Comorbidity Measures , development of a framework that attempts to segregate comorbidities from other aspects of the patient's condition, development of a co- morbidity algorithm, and testing on hetero- geneous and homogeneous patient groups. Data were drawn from all adult, nonmaternal inpatients from 438 acute care hospitals in California in 1992 (n = 1,779,167). Outcome Measures were those commonly available in Administrative data: length of stay, hospital charges, and in-hospital death. RESULTS. A comprehensive set of 30 comor- bidity Measures was developed.

4 The comor- bidities were associated with substantial increases in length of stay, hospital charges, and mortality both for heterogeneous and homoge- Measures of the overall medical condition of patients are essential for health care research, whether collecting data prospectively or using data that have been collected for another purpose. This is true for testing new treatments, assessing established ones, evaluating health plans and providers, or studying the impact of health care policies. Biomedical evaluations that have used randomized controlled trials usually have ex- cluded patients with certain preexisting condi- *From MEDTAP International, Inc.

5 , Bethesda, Maryland. tFrom the Agency for Health Care Fblicy and Research Cen- ter for Organization & Delivery Systems, Rockville, Maryland. tFrom WESTAT, Inc., Rockville, Maryland. ?From The MEDSTAT Group, Inc., Washington, DC. This work was conducted while the authors were em- ployees of the Agency for Health Care Policy and Re- search (AHCPR). The views expressed in this article are neous disease groups. Several comorbidities are described that are important predictors of outcomes, yet commonly are not measured. These include mental disorders, drug and al- cohol abuse, obesity, coagulopathy, weight loss, and fluid and electrolyte disorders.

6 CONCLUSIONS. The comorbidities had inde- pendent effects on outcomes and probably should not be simplified as an index because they affect outcomes differently among dif- ferent patient groups. The present method ad- dresses some of the limitations of previous Measures . It is based on a comprehensive ap- proach to identifying comorbidities and sepa- rates them from the primary reason for hospitalization, resulting in an expanded set of comorbidities that easily is applied with - out further refinement to Administrative data for a wide range of diseases. Key words: Comorbidity ; Administrative data; hospital resources; in-hospital mortality.

7 (Med Care 1998;36:8-27) tions, although even experimental research in- creasingly has been using statistical controls on more heterogeneous Outcomes as- sessments of clinical procedures applied to large populations of patients typically have used statis- tical techniques to control retrospectively for clinical differences among Compari- sons of health care providers have attempted to adjust for the medical and financial risk of serving different Health policy studies, such those of the authors and do not necessarily reflect those of AHCPR. Additional materials are available from the authors.

8 Address correspondence to: Anne Elixhauser, PhD, MEDTAP International, Inc., 7101 Wisconsin Ave., Suite 600, Bethesda, MD 20814; e-mail: Received April 10, 1997; initial review completed May 20, 1997; final acceptance July 29, 1997. 8 This content downloaded from on Sun, 29 Sep 2013 07:16:33 AMAll use subject to JSTOR Terms and ConditionsCOMORBIDITY Measures as evaluations of the effects of payment policies or assessments of the performance of plans and providers, have used statistical techniques to con- trol for the medical conditions of the heterogene- ous patient populations that inevitably must be '7 When using Administrative data, preexisting conditions, or comorbidities, should always be controlled.

9 In general, comorbid conditions have been handled analytically by: (1) stratifying pa- tients into groups-those with a Comorbidity and those without; (2) using separate binary indica- tors for discrete conditions; or (3) summarizing Comorbidity information into an index or score that provides a single parameter for measuring multiple ,8-15 One of the most commonly used indexes was developed by Charlson et Although it was de- veloped for the express purpose of prospectively predicting 1 year mortality among patients being considered for breast cancer clinical trials, it has been applied to discharge abstract and claims data to predict short-term outcomes such as in- hospital mortality, blood transfusions, hospitali- zation charges.

10 And length of Romano et al21'22 explored the Charlson Index and pointed out a number of precautions when using indexes to control for comorbidities. First, the complexity of ICD-9-CM coding and coding idiosyncrasies must be taken into account in de- fining Second, the weights for particular comorbidities should be estimated separately for different populations and different outcomes because their predictive values differ by patient groups. Finally, there is no evidence that the comorbidities included in the Charlson Index are comprehensive, given that it includes only those conditions that happened to occur in a nar- rowly defined clinical population of fewer than 600 patients.