DOI: 10.1002/alz.079257 ISSN: 1552-5260

Association of contemporary dementia diagnosis rate with clinician characteristics: Evidence from Medicare claims

Ying Liu, Andrew Becker, Hankyung Jun, Christopher Wallick, Soeren Mattke
  • Psychiatry and Mental health
  • Cellular and Molecular Neuroscience
  • Geriatrics and Gerontology
  • Neurology (clinical)
  • Developmental Neuroscience
  • Health Policy
  • Epidemiology



Population‐level gap and disparities are documented for the dementia diagnosis. Little is known, however, the extent to which clinician’s characteristics, including the characteristics of the patient population they serve, are associated with their likelihood of adequately diagnosing dementia, after accounting for patient characteristics.


We used the 100% Medicare fee‐for‐service and Advantage Plan data from 2017 to 2019. Patients were attributed to primary care clinicians based on plurality of office visits. We determined each clinician’s observed number of patients with a diagnosis of dementia of any etiology and estimated the expected number of cases with a predictive model using patient characteristics (age, sex, race/ethnicity, whether dually eligible for Medicare and Medicaid). The ratio between the clinician’s observed and expected diagnosis rates (O/E ratio) serves as performance measure for diagnostic accuracy. We calculated 95% confidence intervals (CI) around those O/E ratios and categorized those whose 95% CI included 1 as diagnosing dementia at an adequate rate and those whose 95% CI did not reach 1 as underdiagnosing dementia. We then used logistic regression to predict a clinician’s likelihood of having adequate diagnosing compared to under‐diagnosing using clinician characteristics: their attributed patient panel’s composition on race/ethnicity, dually eligible, and having had annual wellness visit, urban‐rural location of the practice, the clinician’s specialty, and state fixed effects.


Among 194,699 included clinicians, those serving more White (Odds Ratio [OR] = 1.50 for 10 percentage points [pp] increase, p<0.001), fewer Black (OR = 0.94 for 10 pp increase, p<0.001) or Hispanic patients (OR = 0.98 for 10 pp increase, p = 0.003), and fewer duals (OR = 0.82 for 10 pp increase, p<.001) were more likely to have an adequate diagnosis rate than under‐diagnosing. Clinicians who practiced outside of metropolitan areas were less likely to have an adequate diagnosis rate (compared to metropolitan, OR = 0.91, p = 0.025 for micropolitan; OR = 0.75, p<0.001 for small town; OR = 0.63, p<0.001 for rural areas). No association is seen with percent of patients having annual wellness visit (OR = 1.00 for 10 pp increase, p = 0.913).


Primary care clinicians who serve racial/ethnic minority or dually eligible patients and practice outside of big cities may under‐detect dementia cases.

More from our Archive