Data-Driven Healthcare Management: Predictive Analytics to Improve Access, Quality, and Cost Efficiency

Abstract

The United States healthcare system is still stuck with a trilemma of access, quality and cost. In 2024, total national healthcare spending was more than $4.9 trillion (17.6% of GDP), but an estimated 25 million Americans are uninsured, Medicare alone spends about $30 billion on avoidable hospitalizations, and there is a wide variation in quality outcomes across the country. In 2020-25, U.S. health systems have accelerated the use of predictive analytics, big data infrastructure, and AI/ML clinical decision support as the operational stack for a data-driven approach to the trilemma. This study conducts an empirical analysis of an integrated model that posits that three technology-capability constructs, along with data governance and staff/organizational capacity, influence two performance outcomes: healthcare access and quality and cost efficiency and utilization. Between February and September 2025, N = 400 healthcare managers and executives from across the United States in eight organization types (academic medical centers, integrated delivery networks, community and specialty hospitals, ambulatory groups, federally qualified health centers, long-term-care facilities, and health-insurance payers) completed cross-sectional survey data. OLS structural-model estimation, moderation and Sobel-tested mediation were employed to test six hypotheses. Measurement-model reliability was strong (Cronbach's α range 0.86–0.91; CR range 0.80–0.88; AVE range 0.50–0.59) and structural-model fit was good (χ²/df = 1.40; CFI = 0.961; TLI = 0.957; RMSEA = 0.033 [0.026, 0.039]; SRMR = 0.041). AI/ML clinical decision support dominates the access-and-quality outcome (β = 0.375, p < .001); big data infrastructure dominates the cost-efficiency outcome (β = 0.364, p < .001); and predictive analytics capability has strong direct effects on both outcomes (β = 0.194, 0.178, p < .001); staff and organizational capacity partially mediates the analytics-to-quality path (Sobel z = 6.65, p < .001; mediation ratio ≈ 35%). Moderation by data-governance of the relationship between analytics and quality was not supported (β_int = 0.002, p = .965). Five of six hypotheses are supported. The final models account for 52.8% and 47.8% of the variance in the two outcomes. Illustrative extrapolation of within-sample association suggests that patient-access target achievement increased from 61% (low-analytics quartile) to 84% (high-analytics quartile), and avoidable-cost reduction increased from 3% to 24%. Implications for the order of investment in health-management and for federal Value-Based Care policy are applied.