Access to Primary Care and Primary Care Treatable Emergency Department Visits Among Medicare Beneficiaries

The emergency department (ED) can be a difficult place to receive primary care (PC), particularly for elderly people. Older people can be more difficult to diagnose and appropriately treat in a setting like the ED where their health history may be less accessible and time in shorter supply than in a PC setting. Nevertheless, studies have found high proportions of PC treatable ED visits among elderly people. Improving access to PC could reduce these visits by providing preventive care to avoid emergencies or by providing timely access to lower acuity care.

We explore this effect in a nationally representative 5% sample of claims data for fee-for-service Medicare beneficiaries, using two approaches to determine whether increased use of PC is associated with reduced rates of PC treatable ED visits, while attempting to correct for selection effects that could bias the result. One approach looks at the intensive margin: the effect of more frequent PC on PC treatable ED visits; while the second approach looks at the extensive margin: the effect of having versus not having a PC relationship.

We use geographic variation at the hospital referral region (HRR) level in access to and use of PC to estimate the extensive marginal effect. This variation is hypothesized to be due to differences in patient preferences and physician practice style across geographies, but not with underlying differences in population health beyond those accounted for by observable differences in age, race, sex, etc.

The second approach assesses the intensive marginal effect at the beneficiary level of having a PC relationship. We use an endogenous treatment Poisson model with a plausibly exogenous HRR-level variable included in the treatment equation, but not the outcome equation: the supply of PC physicians. We would expect that living in an area with more PC physicians would increase the likelihood that a beneficiary has a PC relationship, but would not affect the likelihood of a PC treatable ED visit except through that relationship.

Both the individual level and the HRR analyses showed an association between PC use and decreased rates of PC treatable ED visits was observed in this data both at a hospital referral region level and at an individual level. At the HRR level, we estimated a substantial reduction of nearly 15 percent in rates of PC treatable ED visits associated with an increase in the number of PC visits per beneficiary per year (IRR 0.852; 95% CI [0.805, 0.901]). The individual level analysis estimated a 28%  reduction in PC treatable ED visits associated with having a PC relationship (marginal effect -3.16 visits per 100 beneficiary-years; 95% CI [-4.22, -2.10].

This study presents evidence from observational data that receipt of primary care is associated with a reduction in primary care treatable ED visits, though residual selection effects cannot be ruled out. This study adds to the evidence from small-scale randomized trials of primary care interventions, suggesting that this beneficial effect is detectable in observational data at a population level.