Risk-Adjusted Models of Costs Referable to General Practitioners Based on Administrative Databases in the Friuli Venezia Giulia Region in Northern Italy

Abstract

Objective: To develop risk adjustment models for cost evaluation in primary health care in Italy based on administrative databases.
Setting: The 2007 administrative databases from the National Health Service of the Friuli Venezia Giulia Region were the data source. Data referred to the general population and included information on the use of health services (inpatient, outpatient, medication, home care) as well as on the major chronic health problems. Data included persons who, for their health condition, must not pay the
contribution usually required for using health services (ticket exemption).
Design: Multilevel (multivariate) statistical analysis, where the tariff of services or the price of drugs (both summed up and separated) were the dependent variables, and the health conditions and other variables related to the citizens were the predictive variables.
Results: The analysis included 1,067,239 citizens registered with a General Practitioner (GP) and 1,129 GPs. The number of people with at least one ticket exemption was 461,532. A number of predictive models were developed, which considered the sum of all tariffs and prices, tariffs and prices of inpatient and outpatient services, and medications. The models had very robust results. The models explained a considerable share of the variation using the (R2 1) parameter: the proportional reduction of error for predicting the level-1 dependent variable with respect to the model without any predictors. The R2 1 was 44.6% for the sum of tariffs and expenditures, 30.9% for hospital tariffs, 27.1% for inpatient services and 49.3% for medications. The intra class correlation coefficient (ICC), which measures the proportion of residual variability due to the second level of analysis (GP), shows that, controlling for the Casemix, the amount of the residual variability driven by the GPs is very low or even negligible: 0.89% for total individual health care tariffs, <0.1% for inpatient services, 1.49% for outpatients services, and 2.0% for drug prescriptions.
Discussion: The health status information provided by the ticket exemption database proved to be valid beyond the Researchers expectations, despite known data quality problems. The large study sample size may have played a role in these results. Primary health care risk adjustment can be performed using administrative databases instead of GP clinical records, making it much easier and
less costly. The overall variation explained by the models was in line with the findings in the literature. The residual variation attributable to GPs was unexpectedly low. In the inpatient setting, this may be partly due to a systemic effort to reduce inappropriate hospitalisation. In the other settings this result came as a surprise, and may lead to a reconsideration of GP behaviour variation. Outcome and output measures in PHC without proper risk adjustment may lead to inaccurate findings.