Purpose: Geriatric diabetes care guidelines encourage individualized glycemic targets (i.e., A1C goal) for older patients based on patient life expectancy (LE) and preferences. We pilot tested a web-based decision support tool which provides individualized prognostic information from a geriatric diabetes complication model, and elicits patient preferences.
Method: We randomized physicians and their patients to the decision support tool, with a 3:1 recruitment ratio. Patients were ≥ 65 years, had A1C ≥ 6.5%, and no dementia. Prior to a clinic visit, intervention patients interacted with the tool, which generated a summary for their physician that included individual patient's LE estimates, risks of developing complications, treatment preferences, and screening for geriatric conditions. Control patients received an educational pamphlet about A1C. Physicians and patients were surveyed before and after the visit.
Result: Intervention (N=75) and control patients (N=25) were similar by gender (77% female), age (mean 74 years), ethnicity/race (89% black) and diabetes duration (mean 16 years). Baseline knowledge of A1C goals by patients was low (35%). Compared to controls, intervention patients were more likely to have their physician report an A1C discussion during a visit (91% vs. 76%, p=0.06) and were more likely to have their physician report that the patient knew their A1C goal (81% vs. 60%, p=0.03). Patient decisional conflict scores declined more for intervention (52.7→24.5, p<0.01) than control patients (51.2→36.6, p=0.03). Compared to controls, more intervention patients had their physician shift their A1C goal by at least 0.5% (49% vs. 28%, p=0.06). Among the intervention patients, we found that the percentage with an intensive goal (A1C goal ≤ 7.0%) increased from 49% to 56% (p=0.23). The movement towards intensive goals occurred only in patients with longer LE (53% to 62%); no change occurred in patients with shorter LE (41%). Among the control patients, we found that the percentage with an intensive goal declined from 68% to 60%. This occurred in patients with both shorter and longer LE.
Conclusion: A personalized decision support tool incorporating prognostic information and patient preferences encouraged active discussion regarding A1C goal selection, decreased patients' decisional conflict, and had a tendency to increase appropriate personalization of A1C goals based on LE estimates. A larger, longitudinal clinical trial is needed to evaluate the intervention effects over time.
