Causal Inference
Association between referral to specialty care and patient-centered outcomes among patients with Alzheimer’s disease and related dementias: an analysis leveraging physician preference instruments Erin Ferguson* Erin Ferguson Silvia Miramontes Katherine Possin Justin S White Anna Chodos Fan Xia Alexander Smith Eva Raphael M Maria Glymour
Background: Primary care providers (PCPs) must often decide whether to refer patients with Alzheimer’s disease and related dementias (ADRD) to specialists. It is unclear whether specialist care improves patient outcomes. Leveraging random variation in PCPs’ preference for specialty referral, we evaluated whether referrals are associated with patient-centered outcomes using observational and instrumental variable (IV) methods.
Methods: Analyses included 3,577 individuals diagnosed with ADRD by San Francisco Bay Area PCPs from 1998 to 2022. Referral to specialists (neurologists and memory care) was determined from electronic health records. The instrument (PCP preference) was defined as the proportion of prior patients a PCP referred to specialists. We identified 9 outcomes using ICD codes: hip fractures, depression, urinary or fecal incontinence, weight loss, hospitalizations, falls, and benzodiazepine or antipsychotics use. We modeled observational associations between referral to care and cumulative incidence of outcomes 5 years post-diagnosis using linear regressions adjusted for age, sex, race, and insurance type; two-stage least squares models were used for IV analyses.
Results: The sample was an average of 79 years (SD: 6.6) at ADRD diagnosis; 8.9% received a referral. Preference strongly predicted actual referral (F=183.7). At 5 years, referral was observationally associated with worse outcomes in every domain except fecal incontinence (Figure 1). In IV estimates, referral was associated with absolute increases in antipsychotics use (overall risk=17%; RD associated with referral=0.25, 95% CI:0.06-0.56), depression (overall risk=24%; RD=0.27, 95% CI:0.07-0.48), and falls (overall risk=30%; RD=0.31, 95% CI:0.06-0.56) (Figure 1).
Discussion: Referral to specialty care was adversely associated with some patient-centered outcomes. However, this IV has limitations to interpretation and viability, which we will discuss and explore in our future work.
