Improved Risk Adjustment for Comorbid Diagnoses in Administrative Claims Analyses of Orthopaedic Surgery.

Abstract

BACKGROUND The accurate inclusion of patient comorbidities ensures appropriate risk adjustment in clinical or health services research and payment models. Orthopaedic studies often use only the comorbidities included at the index inpatient admission when quantifying patient risk. The goal of this study was to assess improvements in capture rates and in model fit and discriminatory power when using additional data and best practices for comorbidity capture. METHODS Hip fracture care was used as an exemplary case of an inpatient condition in a population typically having multiple comorbidities. Cohorts were built from 3 administrative resources: (1) Medicare, (2) all-payer, and (3) private-payer. Elixhauser comorbidities were calculated first using only the index admission and subsequently by adding inpatient and outpatient data from the previous year. Comorbidities identified on outpatient records required 2 instances occurring ≥30 days apart. Model fit and discriminatory power for in-hospital metrics (death, length of stay, and costs or charges) and post-discharge metrics (90-day readmission and surgical site infection, and 90-day and 1-year death) were compared among capture strategies. RESULTS The index admission missed 9.3% to 65.6% of individual Elixhauser comorbidities for the Medicare cohort, 2.9% to 39.0% for the all-payer cohort, and 14.1% to 57.9% for the private-payer cohort compared with data from the index admission plus the previous year. Using prior inpatient and outpatient data provided substantial improvements in model fit and explanatory power for post-discharge outcomes, whereas information from the index admission was sufficient for in-hospital death and length of stay. The utility of outpatient data was greatest when complete outpatient claims were captured compared with only ambulatory surgery claims. CONCLUSIONS The comorbidity capture strategies demonstrated in this study, namely including all available data for post-discharge outcomes, using a 1-year lookback period, and requiring outpatient codes to appear on 2 claims ≥30 days apart, are relevant for improved risk adjustment in orthopaedic clinical or health services research and quality improvement and payment models. LEVEL OF EVIDENCE Prognostic Level III. See Instructions for Authors for a complete description of levels of evidence.