A randomized trial examining the effect of predictive analytics and tailored interventions on the cost of care

The COVID-19 pandemic underscored the importance of accelerating digital medicine, particularly telehealth, to alleviate strain on healthcare systems. The US telehealth market is projected to grow significantly, with virtual visits and remote patient monitoring leading the expansion. The World Health Organization's Global Strategy on Digital Health encourages broader telehealth deployment for equitable access. Mass General Brigham (MGB), a leader in telehealth, uses Personal Emergency Response Systems (PERS) in population health management programs. PERS, enhanced with predictive models to identify patients at high risk of emergency department (ED) visits, facilitates proactive interventions. This study focuses on a Stepped-Care intervention incorporating such a predictive model, triggering nurse-driven interventions. Unlike many programs targeting the highest-cost patients (top 5%), this study centers on the middle segment of the cost pyramid, identified in prior research as a persistently high-cost group with increasing annualized costs. The study aimed to evaluate the impact of this Stepped-Care intervention on healthcare utilization and costs in this middle segment.

Several interventions have shown promise in improving care for older adults with multiple chronic conditions, including the Guided Care Model, Project BOOST, the Transitional Care Model, and Mobile Integrated Healthcare. While some demonstrated improvements in clinical outcomes, cost savings have been inconsistent. Previous research by the authors demonstrated that the middle segment of the cost pyramid, rather than the highest-cost segment, consistently incurred the highest and most rapidly increasing healthcare costs. This highlighted the need for interventions targeting this specific group.

A two-arm RCT was conducted with 370 patients (aged 65+) receiving home care from Partners HealthCare at Home (PHH). Patients were selected from the middle segment of the cost pyramid based on their prior-year healthcare expenditures. After a 3-month observation period, patients were randomized (1:1) to either a control group (CG, n=189) receiving standard care or an intervention group (IG, n=181) receiving the Stepped-Care intervention. The IG received daily risk assessments via a predictive model (AUC=0.78) for ED transport. High-risk patients received tailored interventions from a nurse, including triage, needs assessments, and personalized care plans (e.g., phone follow-up, patient education, home visits, or outpatient visits). Healthcare costs (inpatient and outpatient) were extracted from the MGB Enterprise Data Warehouse and analyzed using Poisson regression (for encounters) and linear regression (for costs). Intention-to-treat analysis was used.

The study enrolled 370 patients, with 172 and 159 in the CG and IG, respectively, included in the final analysis. Baseline characteristics were similar between groups. The IG experienced a statistically significant 31% reduction in annualized inpatient costs per patient ($3,700 lower than CG, p=0.02) and a 20% reduction in annualized total cost per patient ($3,500 lower than CG, p=0.04). The reduction in total costs was primarily driven by decreased inpatient costs, while outpatient costs were similar across groups (p=0.10). Inpatient encounters comprised a small percentage of all encounters (5.9% in CG, 4.8% in IG) but accounted for a disproportionately large share of total costs (66% in CG, 57% in IG). The Stepped-Care intervention was most effective in reducing inpatient costs among high utilizers (patients with multiple hospitalizations). Analysis of inpatient costs by ICD-10 diagnostic categories revealed that top 10 costliest categories contributed to about half of the total inpatient costs. Heart failure was a common condition among both study groups.

This RCT demonstrates that a telehealth-based Stepped-Care intervention effectively reduces healthcare costs, particularly inpatient costs, in older adults in the middle segment of the cost pyramid. The intervention’s success in reducing hospitalizations among high utilizers highlights its potential for improving care management and resource allocation. The findings align with other studies showing the effectiveness of interventions for older adults with complex conditions, however, this study uniquely demonstrates this effectiveness within the understudied middle segment of the cost pyramid. The disproportionate impact of inpatient encounters on healthcare costs is notable and underscores the importance of preventing hospitalizations. The study's results support the adoption of telehealth-integrated, proactive care models. This is especially relevant in the context of the ongoing COVID-19 pandemic, where such models can enhance care delivery, mitigate resource strain, and improve patient outcomes.

This study provides strong evidence for the effectiveness of a Stepped-Care intervention combining predictive analytics and tailored interventions in reducing healthcare costs among older adults. The intervention's success in targeting high-cost utilizers, specifically those requiring multiple hospitalizations, offers valuable insights for population health management. Future research should explore generalizing these findings to more diverse populations and investigate the long-term sustainability and cost-effectiveness of such interventions.

The study population was predominantly older, female, white, living alone, and highly educated, potentially limiting the generalizability of the findings. The analysis did not include healthcare costs incurred outside the MGB system, which might have influenced the overall cost reduction. These factors could have affected the magnitude of observed cost reductions.