Effectiveness of an Automated Falls-Risk Screening and Referral Tool in the Emergency Department (ED)

The specific aim of this retrospective cohort study is to test the effectiveness of the automated screening and referral intervention on completed referrals to the UW Health fall prevention clinic and rates of injurious falls, using a limited dataset created from EHR and Medicare claims data. The investigators hypothesize that ED patients referred using the falls risk CDS tool will have decreased healthcare use due to fall-related injuries, and that the intervention will have similar levels of effectiveness across different types of patient characteristics. The investigators will also systematically examine barriers to patients completing their clinic referrals, as well as clinic scheduling and pre-visit planning protocols that may have excluded patients from receiving Falls Clinic services.

Effectiveness will be assessed based on examination of a limited dataset consisting of EHR and Medicare claims data measuring rates of (1) patient referrals at each ED, (2) completed referrals to the Mobility and Falls Clinic (i.e., a completed clinic appointment), and (3) healthcare visits for fall-related causes occurring within the six months following the initial ED visit.

The primary analysis to evaluate effectiveness of the fall-risk CDS tool will include data from older adult patients (age ≥65) who visit study EDs during the intervention period at each site, have a UW Health System-affiliated primary care provider, and are discharged from the ED or ED observation unit (not admitted). Members of the UWH Applied Data Science team (not part of the study team) will extract data from patient EHR to create a limited dataset including: past falls and fall-related injuries (in the 12 months pre-visit, including the index ED visit), post-visit falls and fall-related injuries (in the 6 months following the index ED visit), patient demographics (age, gender, race/ethnicity, insurance), comorbidities, active medications, and utilization (e.g., primary and specialty care visits). Education and income levels will be approximated using census track data. Area Deprivation Index will also be employed based on patient address. These variables will be extracted retrospectively and stored on secure servers.