Harnessing EHR for goals of care: The role of electronic alerts for care optimization in gynecologic cancer patients at risk of death in 6 months.

5571 Background: This study employed the validated Surprise Question to evaluate the ability of an electronic health record (EHR) alert to predict 6-month mortality, prompt goals of care (GOC) documentation, and facilitate high-quality end of life (EOL) care for patients with gynecologic cancer. Methods: EHR coding identified patients in an outpatient academic gynecologic oncology practice seen more frequently than annually with scheduled imaging, CA125 values, or chemotherapy. Patients without a diagnosis of primary gynecologic cancer were excluded. An outpatient EHR alert posed the modified Surprise Question: “Would you be surprised if the patient passed away in the next 6 months?” Choices were “Yes” and “No” (both considered meaningful responses) and “Show me next time” (deferral response; excluded). “No” responders were instructed on-screen to document a GOC discussion and consider palliative care referral. We analyzed EHR alert data from August 1, 2021-December 31, 2022, and clinical events up to 22 months after this period. Continuous variables were compared using Wilcoxon rank-sum tests, categorical variables were compared with chi-square tests, and both Cox proportional hazards regression and log-binomial models were used to predict death/time to death. Results: Meaningful responses were elicited for 804 unique patients, of which 130 (16.2%) were “No” (not surprised). Among patients whose providers replied “No”, 35.4% died within 6 months of that encounter, compared to 4.2% of “Yes” replies (p<0.001). Among patients with a documented GOC conversation, the median time from first “No” response to documented conversation was 0 days (IQR 0.0, 84), compared to 255 days (IQR 81, 501) for “Yes”. Among patients who died, subjects in the “No” group were more likely to be enrolled in hospice at EOL than those in the “Yes” group (79.0% vs 63.5%, p=0.014). A “No” response to the Surprise Question had a higher RR for death at 6 and 12 months compared to age, race, or cancer type. Conclusions: For patients with gynecologic cancer, responses to a Surprise Question EHR alert effectively predict 6-month mortality and are associated with increases in both GOC discussions and hospice utilization. Log-binomial predictors of death at 6 months (univariate model). Risk Ratio (RR) 95% CI of RR p-value % deaths Age at 1 st alert <4950-5960-6970+ Reference1.000.930.93 0.95-1.050.87-0.980.88-0.98 0.010 4%4%11%11% Race CaucasianBlack/Other Reference0.95 0.91-1.00 0.063 8%12% Cancer Type OvarianOther Reference1.03 0.99-1.08 0.16 8%10% Surprise Question Response YesNo Reference1.48 1.31-1.69 <.0001 4%35%