Surgical Site Infection Surveillance Following Total Hip and Knee Arthroplasty Using California Administrative Data

Background: Routine surgical site infection (SSI) surveillance largely relies on hospital-based detection during the surgery hospitalization or readmissions to the same facility. This may underestimate SSI burden if a substantial proportion of patients with SSI are readmitted to other hospitals. Comprehensive evaluation of pre- and post-discharge SSI using ICD-9 diagnosis code-based criteria has been previously shown to have high sensitivity and positive predictive value for detecting SSI following total hip and knee arthroplasty (THA and TKA) compared to routine surveillance [Bolon, et al. CID 2009;48:1223-9].

Objective: Utilizing a comprehensive state-wide administrative dataset, estimate the proportion of THA and TKA SSI missed by limiting surveillance to the index surgery hospital.

Methods: We conducted a retrospective cohort study of all patients in CA undergoing primary THA or TKA between 3/1/06 and 12/31/07 to assess SSI within 365 days of surgery using mandatory CA hospital data which allows tracking of patients between hospitals. We identified all subsequent hospital admissions within 365 days of surgery with ICD-9 diagnosis codes previously shown to be highly predictive of SSI (998.5, 998.51, 998.59, or 996.66). We assessed the fraction of SSI identified through admissions to the index hospital vs. other hospitals and the time (in days) to SSI.

Results: 52,106 THA and 65,540 TKA procedures were identified. 1,255 (2.4%) THA and 1,410 (2.2%) TKA received an SSI diagnosis code within 365 days of surgery. Of those with an SSI code, 61% of THA and 53% of TKA SSI were detected within 30 days, and 76% of THA and 69% of TKA SSI were detected within 60 days (Table). Time to SSI readmission was significantly longer for patients readmitted to hospitals other than the index hospital (p<.0001). Overall, 18% of SSI would have been missed by index hospital surveillance alone. The per hospital proportion of SSI detected at non-index hospitals ranged from 0-100% across CA hospitals with a median of 8% for THA and 9.5% for TKA.

Conclusions: Estimated SSI rates based on previously validated SSI diagnosis code criteria were 2.4% (THA) and 2.2% (TKA). Nearly one-fifth of SSI would have been missed using surveillance limited to the index hospital, disproportionately affecting SSI rates for some hospitals. The majority of SSI would be detected using a surveillance period of 60 days, but this timeframe would miss a higher fraction of SSI detected at non-index hospitals. Large administrative datasets with the ability to track healthcare encounters for specific patients across multiple facilities can provide a more complete picture of SSI burden following arthroplasty than surveillance limited to the hospital where the surgery was performed.