A population-based, electronic health record-guided approach to improve the quality of dementia care.

Background: The quality of care of the 6.5 million Americans living with dementia has been suboptimal, leading to worse outcomes and higher costs. Few health systems have formal systems in place to guide the care of these patients. To help improve the care of persons living with dementia, we developed and preliminarily evaluated the effectiveness of electronic health record (EHR)-generated recommendations for patients with dementia.

Methods: This quality improvement study was conducted from October 2020 through June 2022 at a single academic healthcare system and included patients identified as having dementia on their problem list and their physicians. Ten (seven outpatient and three inpatient) algorithms based on clinical logic and evidence were embedded in an EHR system to generate specific recommendations based on combinations of utilization, diagnosis, and medications. The number of each type of recommendation generated, and orders for each type of recommendation were recorded, as well as physician's perceptions of this approach.

Results: Three thousand six hundred and nine recommendations on 763 patients were triggered by the algorithms in the outpatient setting, and 185 referrals were placed. The most common recommendations were for ongoing care through the UCLA Alzheimer's and Dementia Care program, Palliative Care, the Extensivist Clinic, Urogynecology, and Clinical Pharmacy. The most commonly acted upon by providers were recommendations for referral to Pharmacists and the UCLA Alzheimer's and Dementia Care program. The most common reason for not responding to specific recommendations was that these were not perceived as relevant to the patient. Compared to general medicine physicians, geriatricians felt more comfortable managing dementia care without a referral to a service (23% (95% CI 15%-34%) versus 3% (95% CI 0%-17%), p = 0.012) and less commonly felt the recommendation was appropriate (1% (95% CI 0%-7%) versus 13% (95% CI 4%-30%), p = 0.02).

Conclusions: EHR-generated algorithms can help guide patients with dementia to appropriate clinical services.