Research That Matters (January 17 - 20, 2008)

Background & Purpose: This study uses data from a three-year longitudinal panel study to assess whether different methods of measuring coverage lead to disparate conclusions about the impact of uninsurance on low-income adults' access to needed health care.

Methods: Participants were adults enrolled in Oregon's Medicaid program, the Oregon Health Plan (OHP), in February 2003. This study used a series of multi-wave mail surveys over a period of 30 study months. Insurance coverage was conceptualized in three ways --a traditional "static point in time" measure, a series of static measures spread across multiple data collection points, and a "total amount of time uninsured" measure. These three methods are applied to the same underlying dataset in a logistic regression model evaluating the likelihood of having unmet healthcare needs. Results from each model are compared to determine relative strengths and weaknesses of each conceptual approach to defining insurance status.

Results: Coverage 1 (Static, Point in Time): When a static/cross sectional coverage variable is regressed on unmet need, results indicate that being uninsured is associated with 3 ˝ times greater odds (OR = 3.497) of reporting unmet need relative to being insured. Coverage 2 (Static, Multiple Points in Time): When coverage was conceptualized as three “point in time” measures across the study period, results show each time a person reported being uninsured was associated with just over twice (OR = 2.193) the odds for unmet need relative to someone who reported no points of uninsurance during the study period. Thus, those who were uninsured at two of the three data points had four times (OR = 4.386), and those uninsured at all three data points had six times (OR = 6.579), greater odds for unmet need relative to those who were not uninsured at any of the three data collection points. Coverage 3 (Actual Time Uninsured): When coverage was conceptualized as the number of three-month spans spent uninsured (out of ten three-month spans that elapsed during the study period), results show each three-month span of uninsurance is associated with a 25% increase in the odds of reporting unmet need (OR = 1.25) relative to those who had no uninsurance during the study period. Thus, someone experiencing two spans of uninsurance (6 months out of 30) during the study had 2 ˝ times greater odds (OR = 2.512) of experiencing unmet need, while someone experiencing ten spans of uninsurance (the entire length of the study) had 12 times greater odds (OR = 12.56) relative to a continuously insured person.

Conclusions & Implications: These results suggest how we measure insurance coverage definitely matters. Our data indicate static measures of coverage overestimate the effect of short-term uninsurance and underestimate the effects of longer uninsurance periods. The tendency of static coverage measures to underestimate the impacts of long-term uninsurance may be of critical importance for researchers hoping to understand the true “net effect” of coverage loss. Policy-makers should be aware that extensive uninsurance requirements for public insurance might expose individuals to increasing risk for unmet healthcare needs.