Objectives: to assess at population level if healthcare administrative data can be suitable to identify variability and determinants of the prescribing rates for some diagnostic imaging procedures at high-risk of inappropriateness.
Design: population-based observational study.
Setting and participants: Piedmont Region (Northern Italy) adult population (>20 years) in year 2013. The subjects included were 3,566,147, referring to 3,016 general practitioners (GPs), in 12 Local Health Authorities (LHAs).
Main outcome measures: prescription rates of 4 diagnostic imaging procedures at high-risk of inappropriateness (vertebral and joint magnetic resonance - MR, vertebral computed tomography - CT, and bone densitometry). The following factors have been investigated: sociodemographic and clinical data of the subjects, prescribing GPs' characteristics, and organizational environment of the LHA for whom the GP is working. For each procedure, prescription rate determinants have been estimated by a logistic regression modelling of the probability of receiving at least one prescription during the year. Hierarchical data structure has been managed. The Intraclass Correlation Coefficient (ICC) and the Median Odds Ratio (MOR) were used as measures of variability due to GPs and LHAs.
Results: the degree of variability among LHAs and, within each LHA, among GPs was moderate (ICC for LHA: 0.009 and 0.004; ICC for GP: 0.015 and 0.014, respectively, for vertebral and joint MR). Prescribing probability was mainly explained by the demographic and clinical characteristics of the subjects. Most of the GPs' characteristics did not show any relevant association. Results suggest also a potential role of the private technology availability in the LHA where the GP is working.
Conclusions: administrative healthcare data did not describe a relevant degree of variability in prescribing diagnostic imaging procedures in Piedmont Region, except for some GPs and LHAs with a different rate from the regional average. To explain differences in prescribing rates of some GPs or LHAs and to assess their appropriateness, outpatient databases should be enriched with clinical data (mainly, diagnosis or clinical suspicion) or different tools, such as clinical audit, should be used.