|Title||Medicare claims data underestimate hallucinations in older adults with dementia|
|Publication Type||Journal Article|
|Year of Publication||2022|
|Authors||Hamedani, AG, Weintraub, D, Willis, AW|
|Journal||The American Journal of Geriatric Psychiatry|
|Keywords||Dementia, Hallucinations, Medicare, sensitivity, specificity, validity|
Objective Administrative claims data are used to study the incidence and outcomes of dementia-related hallucinations, but the validity of International Classification of Diseases (ICD) codes for identifying dementia-related hallucinations is unknown. Methods We analyzed Medicare-linked survey data from two nationally representative studies of U.S. older adults (the National Health and Aging Trends Study and the Health and Retirement Study) which contain validated cognitive assessments and a screening question for hallucinations. We identified older adults who had dementia or were permanent nursing home residents, and we combined this with questionnaire responses to define dementia-related hallucinations. Using Medicare claims data, we calculated the sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) of ICD codes for dementia-related hallucinations overall and within prespecified strata of age, neurologic comorbidity, and health care utilization. Results We included 2,337 older adults with dementia in our cohort. Among 3,789 person-years of data, 1,249 (33.0%) had hallucations, and of these 286 had a qualifying ICD code for dementia-related hallucinations or psychosis (sensitivity 22.9%). Of 2,540 person-years of dementia without hallucinations, 284 had a diagnosis code for hallucinations (specificity 88.8%). PPV was 50.2%, and NPV was 70.1%. Sensitivity was greatest (57.0%) among those seeing a psychiatrist. Otherwise, there were no significant differences in sensitivity, specificity, PPV, or NPV by age, neurologic diagnosis, or neurologist care. Conclusion Dementia-related hallucinations are poorly captured in administrative claims data, and estimates of their prevalence and outcomes using these data are likely to be biased.