The recalibration and redevelopment of a model to calculate patients’ probability of completing a colonoscopy following an abnormal fecal test

Abstract

Abstract Background: Fecal immunochemical testing (FIT) is an effective screening tool for colorectal cancer. If a FIT is abnormal, a follow-up colonoscopy is necessary to remove polyps or find cancers. Identifying patients who have a low probability of obtaining follow-up colonoscopy after an abnormal fecal test could help deliver early interventions that increase colonoscopy adherence (e.g., patient navigation) to patients who need them most. We sought to develop a usable risk prediction model to identify patients unlikely to complete a colonoscopy following an abnormal FIT test. Methods: We recalibrated and then redeveloped a prediction model created in a group of federally qualified health centers (FQHCs) to be used in a single large FQHC. The models were created from a retrospective cohort of patients aged 50-75 with an abnormal FIT test. The models used clinical data. Logistic and Cox regressions were used to recalibrate the group of FQHC prediction model and then redevelop it in the single large FQHC. Results: The initial risk model used data from 8 FQHCs (26 clinics) and included eight variables including race, clinic system, prior missed appointments, insurance, prior flu shots, age, indication of anticoagulation use, and income inequality. The first model included 1723 patients. However, when we applied the model to a single large FQHC (34 clinics, n=884), the model did not recalibrate successfully (C-statistic dropped more than 0.05, from 0.66 to 0.61). The model was redeveloped in a cohort of 1401 patients and contained 12 variables including age, race, language, insurance, county, a composite variable for sex and mammogram screening, number of prior missed appointments, Gagne’s comorbidity score, number of prior encounters, BMI, marital status, and prior screening with a c-statistic of 0.65. Conclusions: The original model developed in a group of FQHCs did not adequately recalibrate in the single large FQHC. Health system, patient or specialty care characteristics, or differences in data captured in the electronic health record may have led to the inability to recalibrate the model. However, the redeveloped model provides an adequate model for the single FQHC. Precision medicine is best applied when risk is understood in context and interventions are tailored for specific populations’ predictors.