Impact of alternative coding schemes on incidence rates of key complications after total hip arthroplasty: a risk-adjusted analysis of a national data set.

BACKGROUND: Administrative data are commonly used to examine orthopedic outcomes including total hip arthroplasty (THA), but little is known about how minor analytic decisions impact results. Our objective was to examine how the rates of 3 adverse outcomes (deep vein thrombosis [DVT], pulmonary embolism [PE], and hemorrhage) varied with subtle adjustments to our analytic method.
METHODS: We used Medicare Part A data to identify all beneficiaries who underwent primary or revision THA during 2007 to 2008. We used 2 published algorithms (Katz/Cram and Patient Safety Indicators [PSIs]) to identify cases of DVT, PE, and hemorrhage occurring at 3 different points in time; index admission; 30-day readmission; and index admission plus readmission. We used the kappa statistic to compare the agreement between methods. We examined variation in complication rates across hospitals using regression models that adjusted for differences in patient demographics and comorbidity.
RESULTS: Among 202 773 primary and 40 973 revision THA patients, the agreement between the Katz/Cram and PSI methods was excellent for DVT and PE at all time points (kappa 0.95-1.0) but poor for hemorrhage (kappa 0.07-0.29). The incidence of DVT during the index admission among the primary THA cohort was 0.40% using the Katz/Cram method and 0.37% using the PSI method. The incidence of hemorrhage during the index admission among the primary THA cohort was 1.29% using the Katz/Cram method and 0.05% using the PSI method. We found significant variation in hospital rates of all 3 complications (DVT, PE, and hemorrhage). For example, the mean rate of hemorrhage at index admission or readmission for revision THA was 5.7% (standard deviation: 12.8%); we found 137 hospitals with hemorrhage rates of 25% or higher among their revision THA patients. DISCUSSION: We found important differences in the rates of THA complications depending upon the coding algorithms and time frame employed. Our results suggest that administrative data can be used to evaluate THA complications but that methodology should be carefully considered.