Can Quality Indicators for Colonoscopy Be Derived from Administrative Data? Identifying Predictors of New or Missed Cancers After Colonoscopy

Abstract

Background: There is wide-spread interest in deriving quality indicators from administrative data, however most indicators are process measures and little research has examined their relationship to meaningful outcomes. We designed this study to determine if endoscopist characteristics defined using administrative data were associated with development of a new or missed cancer (NMC) within 3 years of colonoscopy. Methods: We identified individuals ≥ 20 diagnosed with colorectal cancer (CRC) in Ontario from Jan 2000 - Dec 2005 using the Ontario Cancer Registry and categorized patients as having proximal (proximal to splenic flexure) or distal (splenic flexure and distal) cancers. We excluded those with prior diagnosis of CRC or unknown primary site. We determined performance of colonoscopy from 1992 using Ontario Health Insurance Plan data. Patients who had complete colonoscopy 7 - 36 months prior to diagnosis were defined as having a NMC, patients who had complete colonoscopy within 6 months of diagnosis (but not between 7 and 36 months) had a detected cancer. We determined if endoscopist factors (colonoscopy volume, polypectomy rate, colonoscopy completion rate over the proceeding 2 years, specialization and setting) were associated with NMC using logistic regression controlling for patient age, sex and comorbidities, stratified for proximal and distal cancer and adjusted for clustering. Results: 14,064 patients with CRC had a complete colonoscopy within 36 months of diagnosis; 67.1% were diagnosed with a distal cancer. 584 patients (6.8%) with distal and 676 patients (12.4%) with proximal cancers had a NMC. In multivariate modeling, for all patients, endoscopist specialty (non-gastroenterologist / non-general surgeon) and setting (office-based colonoscopy) were associated with an increased risk of NMC. Conversely, patients who underwent colonoscopy by an endoscopist with a high colonoscopy completion rate in the previous 2 years were less likely to have a NMC (p = 0.03, OR for >95% [vs, <80%] completion rate for distal NMC = 0.73, 95% CI 0.54-0.97 and p=0.002 for proximal NMC OR = 0.72, 95% CI 0.53-0.97). Patients with proximal (but not distal) cancers who underwent colonoscopy by an endoscopist with a high polypectomy rate had a lower risk of NMC (p<0.0001, OR for >30% [vs. <10%] polypectomy rate for proximal NMC =0.61, 95% CI 0.42-0.89). Procedure volume was not significantly associated with the risk of NMC. Conclusions: Provider characteristics derived from administrative data including completion rate and polypectomy rate are associated with development of a NMC after colonoscopy and may have potential use as quality indicators.