Session number: 1128
Abstract title: Charlson index instability among veterans with diabetes
Author(s):
M Brimacombe - VA New Jersey Health Care System, East Orange; University of Medicine and Dentistry of New Jersey, New Jersey Medical School, Newark NJ
M Safford - VA New Jersey Health Care System, East Orange; University of Medicine and Dentistry of New Jersey, New Jersey Medical School, Newark
M Rajan - VA New Jersey Health Care System East Orange, NJ
L Pogach - VA New Jersey Health Care System, East Orange, NJ; University of Medicine and Dentistry of New Jersey, New Jersey Medical School, Newark NJ
Objectives: The Charlson comorbidity index is a common co-morbidity adjuster in health services research. Based on weighting of recorded ICD-9-CM codes, stability of Charlson scores should exist across inpatient and outpatient assessments within limited time periods for the same subjects. Our objective was to examine the stability of the Charlson index for a large cohort of diabetes patients in the VHA both cross-sectionally and longitudinally.
Methods: We used Austin data for FY98 and FY99. We constituted a cross-sectional sample of individuals who met the HEDIS definition of diabetes, and who had been hospitalized in the VHA. We constructed Charlson index scores based on inpatient only, outpatient only and combined inpatient/outpatient ICD-9-CM codes. We then evaluated inpatient/outpatient correlation across age, gender and race. Also, we evaluated how Charlson scores changed on the same individuals longitudinally.
Results: For FY99, we studied 129,026 individuals. Inpatient and outpatient Charlson scores had Pearson correlations of .25-.50. These correlations decreased linearly (R2 >.95) as a function of age, with similar age patterns within racial and gender stratifications. Outpatient Charlson scores (OCS) were higher relative to inpatient. For those with inpatient Charlson scores (ICS)=0 (n=22,154), overall 71% had higher OCS (White and Black=71%). For ICS=1 or 2 (n=64,911), overall 40% had higher OCS (White: 41%, Black: 39%). For ICS=3 or 4 (n=27,182), overall 28% had higher OCS (White: 27%, Black: 31%). For those with ICS and OCS in both years, the correlations and actual Charlson scores themselves were stable.
Conclusions: Charlson scores derived from inpatient administrative data source were consistently lower than scores derived from outpatient or combined inpatient/outpatient sources. This pattern is most prominent in older individuals. For a given patient, inpatient and outpatient Charlson scores were stable across two years longitudinally.
Impact statement: Use of combined inpatient and outpatient files for ICD9CM codes may be necessary to capture all co-morbidities that comprise the Charlson index. Development and validation of a disease-specific index that incorporates age may be needed.