Association Between a Co-Designed Dashboard and Use of Costly Health Services in Patients With Chronic Kidney Disease and Advanced Cancer: Propensity Score-Adjusted Difference-in-Differences Study.

Abstract

The US health care system faces escalating costs, increasing emphasis on patient autonomy, and a regulatory shift toward patient-centered care and patient-reported outcomes (PROs). Leveraging PROs to support shared decision-making has the potential to improve outcomes and reduce health care use for patients with advanced chronic conditions. This study aims to evaluate the impact of a PRO-based clinical dashboard on the use of costly health services among patients with advanced cancer and chronic kidney disease (CKD). We conducted a quasi-experimental, propensity score-weighted, difference-in-differences analysis using routinely collected data (June 2020 to January 2022) from a large US academic health system. Dashboard users were compared with contemporaneous nonexposed patients matched on clinical criteria. The primary outcomes were unplanned all-cause hospital admissions, potentially avoidable emergency department visits, excess days in acute care within 30 days of discharge, and 7-day readmissions. Cancer-specific secondary outcomes included acute encounters during outpatient chemotherapy, oncology triage use, advance directive completion, and hospice use. CKD-specific outcomes were CKD-related acute care use and disease progression. In the advanced cancer cohort (dashboard users: n=284; dashboard nonusers: n=917), dashboard use was associated with significantly fewer chemotherapy-related emergency department or hospital encounters (ratio-in-odds ratios 0.35, 95% CI 0.16-0.75) and a nonsignificant 1.7-percentage point reduction in unplanned admissions (β=-0.017, 95% CI -0.107 to 0.072). Using Firth penalized logistic regression to reduce small sample bias, dashboard use was also associated with significantly higher odds of 7-day readmissions (ratio-in-odds ratios 8.58, 95% CI 2.28-32.32). Among readmissions in the dashboard user group, most (13/14, 93%) were scheduled by clinicians. Excess days in acute care increased by 4 percentage points (β=0.040, 95% CI -0.001 to 0.089). Advance directive completion declined significantly (β=-0.009, 95% CI -0.039 to 0.020). In the CKD cohort (dashboard users: n=365; dashboard nonusers: n=2137), no significant differences were observed for any primary or CKD-specific outcome. In routine oncology practice, a PRO dashboard was associated with fewer acute care encounters during chemotherapy but more planned early readmissions. The dashboard had no measurable effect on patients with CKD. These disease-specific mixed results highlight the need to tailor dashboards to the clinical context and embed them within workflows that can act on real-time PRO information.