Development and Validation of a Prediction Model for Incident Hypothyroidism in a National Chronic Kidney Disease Cohort

Abstract Context Hypothyroidism is a common yet under-recognized condition in patients with chronic kidney disease (CKD), which may lead to end-organ complications if left untreated. Objective We developed a prediction tool to identify CKD patients at risk for incident hypothyroidism. Methods Among 15 642 patients with stages 4 to 5 CKD without evidence of pre-existing thyroid disease, we developed and validated a risk prediction tool for the development of incident hypothyroidism (defined as thyrotropin [TSH] > 5.0 mIU/L) using the Optum Labs Data Warehouse, which contains de-identified administrative claims, including medical and pharmacy claims and enrollment records for commercial and Medicare Advantage enrollees as well as electronic health record data. Patients were divided into a two-thirds development set and a one-third validation set. Prediction models were developed using Cox models to estimate probability of incident hypothyroidism. Results There were 1650 (11%) cases of incident hypothyroidism during a median follow-up of 3.4 years. Characteristics associated with hypothyroidism included older age, White race, higher body mass index, low serum albumin, higher baseline TSH, hypertension, congestive heart failure, exposure to iodinated contrast via angiogram or computed tomography scan, and amiodarone use. Model discrimination was good with similar C-statistics in the development and validation datasets: 0.77 (95% CI 0.75-0.78) and 0.76 (95% CI 0.74-0.78), respectively. Model goodness-of-fit tests showed adequate fit in the overall cohort (P = .47) as well as in a subcohort of patients with stage 5 CKD (P = .33). Conclusion In a national cohort of CKD patients, we developed a clinical prediction tool identifying those at risk for incident hypothyroidism to inform prioritized screening, monitoring, and treatment in this population.