Should we include dementia diagnosis or cognitive impairment to predict home care cost? An observational study using...

The reintroduction of dementia diagnosis into Medicare Advantage risk-adjustment models suggests that the condition independently influences Medicare costs. Similarly, Switzerland is considering specific tariffs for dementia patients in long-term care. However, the question of whether including dementia diagnosis meaningfully improves predictions of home nursing care costs is still under debate. Therefore, we examined whether including dementia diagnosis and/or symptoms of cognitive impairment, in addition to standard predictors, meaningfully improves the performance of home care cost predictions.

Data from an observational multicentre study of eight Swiss home care providers across all language regions was analysed. The study included all patients undergoing the routine interRAI-Home Care Switzerland assessment of patient characteristics. We used administrative data from months two and three post-assessment including home nursing care costs covered by health insurers as well as generalised linear models with a gamma log-link function to predict costs. We assessed whether including a dementia diagnosis and/or a measure of cognitive impairment (cognitive performance scale, CPS) improved prediction in comparison to models with only standard predictors such as impairments in activities of daily living (ADL, e.g. bathing, dressing), instrumental activities of daily living (IADL, e.g. cooking, managing finances), and living situation (living alone).

Among the 1,035 patients in our dataset, 176 (17%) had a dementia diagnosis. Mean monthly home care costs for dementia patients were CHF 362 (Swiss francs) higher (mean: CHF 1,082; median: CHF 812) than for non-dementia patients (mean: CHF 720; median: CHF: 457). ADL and IADL were the most important predictors of home care costs. Introducing dementia diagnosis or cognitive impairment or both alongside ADL, IADL, living alone, age and gender did not materially change prediction performance (AIC remains unchanged, BIC changes + 1.09 to + 6.36 in the main model).

Our findings give no support for the implementation of differentiated reimbursement tariffs or risk adjustment models for dementia patients or patients with cognitive impairment. More general predictors, such as ADL and IADL impairments seem to predict home nursing care costs adequately enough. However, further research with larger and more representative samples in other countries is necessary to validate these findings.

Not applicable.