Many studies have looked at which variables determine medical utilization rates. Specifically, most of the studies have focused on primary care or general practitioner (GP) visits. A recent study by Nolan (2007) employs panel data from Ireland to reach some new conclusions on this topic.
Ireland is an interesting location to preform the study. All individuals are eligible for the universal public health insurance, but the coverage insists on significant out-of-pocket payments for GP visits. Only the poor and the unemployed are eligible to receive a ‘medical card’ which entitles them to free medical treatment. Also, 50% of Irish hold private insurance. However, private insurance is used mostly to pay for private or semi-private hospital care and does not cover the large copayments for primary care visits.
Generally, studies using cross-sectional data find that non-need factors such as education level, employment status, income, and marital status are important determinants of GP visitation. Causality, however, is difficult to ascertain from these findings. Does low income and education cause poor health or does poor health inhibit wage growth and the accumulation of human capital? Dolan also finds that 30% of individuals do not see a GP each year. However, when we look at the entire panel, only 2.5% of individuals had not visited a GP during sample time period.
Since the dependent variable, GP visits, is a count variable, a poisson regression is used. The function Î»() is altered from the standard procedure to account for the panel nature of the data.
- Î»=Î±i*exp[zitÎ³+g(yit-1)Ï?] (1a)
- Î±i=ci*exp[Î±0+ri0Î±1+Zia2] (1b)
The variable zit represents a variety of covariates and is Zi is the within-individual means of the covariates. The vector of initial conditions is represented by ri0. Five dummies for lagged values of the number of patient visits (0, 1, 2, 3-5 and 6+ visits) is represented by g(yit-1). The Î» equation includes an individual effect, Î±i, and the lagged dependent variables. The individual effects are parameterized by equation (1b). Although the above is the preferred regression, Nolan also uses a static random effects model (2) and a static random affects model with correlated individual affects (3a, 3b).
- Î»=Î±i*exp[zitÎ³] (2)
- Î»=Î±i*exp[zitÎ³] (3a)
- Î±i=ci*exp[Î±0+Zia2] (3b)
The data set the paper uses is the Living in Ireland Survey between 1995 and 2001. The survey is part of European Community Household Panel (ECHP).
Using the panel framework, the author concludes that “the only significant non-need factors are medical card eligibility and employment status.” Nolan finds that more physician visits in the prior year lead to more GP visits in the current year. Also, the initial level of GP visits (in 1995) also had a significant impact on current primary care visits. Other findings include:
- Unemployment lead to 0.18 fewer GP visits
- As expected, health-related covariates such as old age, ill-health, stress, and a birth in a family lead to higher GP usage.
- Having a “medical card,” which entitles the patient to completely free care, leads to an increase of 0.33 GP visits per year.
- Covariates such as income, education level, and marriage status had no impact on GP visits.
- Having private insurance actually decrease GP visits by 0.17 per year.
This paper supports the case of both economists and patient advocates. Economists will point to the medical card finding that giving away medical care for free will increase utilization rates. Nolan point out the RAND health insurance found that decreasing copayment lead to an increase of both ‘appropriate’ and ‘inappropriate’ medical services. On the other hand, patient advocates will point to the fact that the most important covariates determining GP visits are health-related.
This paper answers important questions regarding the determinants of GP utilization. The panel nature of the data allows the researcher to control for individual heterogeneity as well as condition on past GP visitation rates.
- Nolan (2007) “A dynamic analysis of GP visiting in Ireland: 1995-2001,” Health Economics, vol. 16, pp. 129-143.