Health plan innovations and health care costs in the commercial health insurance market

Abstract

The US has been increasingly seeking solutions in health insurance designs to control health care costs. Despite an ongoing debate about whether demand-side cost sharing or supply-side restrictive provider choice is more effective at reducing costs, there is little work to guide this debate due to challenges in causal inference, estimation, and measurement. This dissertation aims to: (1) understand the role of health plan designs in health care cost containment using a large, multiple-employer, multiple-insurer panel dataset; (2) develop a new estimation algorithm for models with multiple high-dimensional fixed effects; and (3) design a new statistical method for inferring consumer choice of providers using claims level data.

Chapter 1 examines the effects of health plan designs at a high level, by looking at variations in health care treatment across plan types that differ in their cost sharing and choice of providers. It finds that narrow provider choice may be more effective than high cost sharing at reducing health care utilization. This chapter speaks to insurance benefit design, and contributes to the literature by developing a new “treatment spells” approach that improves on episode or calendar interval analyses. Chapter 2 deals with challenges in estimating causal inference models. We present a new estimation algorithm for models that entail multiple high-dimensional fixed effects, large unbalanced panels, instrumental variables, and clustered standard error corrections. Applying the algorithm to a sample of over 1.4 million patients using more than 150,000 distinct primary care doctors over a 47-month period, we find that provider network breadth dominates cost sharing in influencing consumers’ monthly utilization of care. Chapter 3 examines the consequences of narrow provider plans, namely how the breadth of consumer choice of providers affects individual health care utilization and spending. Since providers are not observed when their services are unused, I select plans with high enrollment/low provider density where provider network breadth can be more reliably inferred. Using an instrumental variable strategy, I find that narrow provider plans redirect patients from in-network to out-of-network services but only modestly lower expected health care costs.