Estimating Underdiagnosis of Patients in Chronically Ill Populations

Diagnosis coding in administrative data is often incomplete and introduces inaccurate assessments of patients' health outcomes. The underdiagnosis of chronic conditions reduces the ability to correlate that chronic condition with associated comorbidities. We developed a novel framework to use longitudinal records of claims-based diagnoses to identify latent, or undiagnosed, members likely to have a target chronic condition before their first diagnosis of the condition is included in the records. We applied our approach to the chronic condition of adult incontinence using the 5% Medicare sample of a continuously enrolled cohort between 2014 and 2018. We considered fee-for-service Medicare beneficiaries over the age of 65 years. We identified observed and latent members with the chronic condition and then examined its correlation with four comorbidities commonly associated with incontinence: urinary-tract infections (UTIs), dermatitis, slips and falls, and behavioral disturbances. We increased our identification of incontinence from the 11.2% with an observed diagnosis in 2018 to an estimated prevalence in 2018 of 34.7%. Considering the four comorbidities, our approach increased from 38% to 68% the share of UTIs experienced by those with incontinence, from 20% to 41% for dermatitis, from 22% to 54% for slips and falls, and from 26% to 57% for behavior disturbances. Our methodology highlights the importance of more accurately estimating the prevalence of chronic conditions for understanding the true correlation between the chronic condition and comorbidities.