This thesis addresses health-policy relevant questions regarding quantity and quality of service delivery in primary healthcare using health administrative data from the province of Ontario. It is comprised of four chapters that explore the following questions: (1) What is the impact of switching from an enhanced fee-for-service (EFFS) payment model to a blended capitation payment model on the specialist referral rates of primary care physicians? (2) What are the rates of inappropriate laboratory testing in the province of Ontario? (3) What are the costs and determinants (physician and practice characteristics) of these inappropriate tests? (4) What is the impact of primary care payment structure on the quantity (number and cost) and the quality (appropriateness) of clinical laboratory testing? Fee-for-service (FFS) payment systems give physicians an incentive to treat patients on the margin of being referred, whereas in capitation systems physicians do not have a financial incentive to treat such marginal patients. Chapter 1 empirically examines how these two payment systems affect referral rates. The results show an increase in specialist visits upon a switch from an EFFS model to a blended capitation model when the physician is listed as the referring physician in the data, but no change in total specialist visits for these physicians’ patients. This change is not observed immediately upon switching payment models. Physicians paid by blended capitation who practice in an interdisciplinary health team have fewer specialist visits per rostered patient compared to EFFS physicians, despite an increase in their patients’ specialist visits after joining the interdisciplinary team. Using a definition of inappropriateness that quantifies ordering clinical laboratory tests too often or too soon following a previous test, Chapter 2 examines the rates of inappropriate laboratory testing for nine selected analytes in Ontario. The chapter finds that the percentage of inappropriate tests ranges from 6% to 20%. Moreover, between 60% and 85% of the time, the physician ordering an inappropriate test is the same physician who ordered the previous test. The findings also show that specialists are more likely than primary care physicians to order repeat tests too soon. Chapter 3 examines the costs and determinants associated with the rates of inappropriate laboratory utilization. The associated costs of inappropriate/redundant laboratory testing for the selected analytes ranges between 6 – 20% of the total cost of each test. Statistical analyses of the association of physician and practice characteristics with inappropriate testing are done using a logit model. Conditional upon the variables within the model, male physicians, physicians trained outside of Canada, older physicians, and a younger patient population are all shown to be associated with less inappropriate testing. Primary care physicians in group practices and in payment models with pay-for-performance (P4P) incentives are less likely to order inappropriate tests and specialist physicians are twice as likely to order inappropriately compared to FFS primary care physicians. Differences in physician, practice and patient characteristics, however, explain only a small amount of the variation in inappropriate utilization. Chapter 4 examines how physicians’ laboratory test ordering patterns change following a switch from an FFS payment model enhanced with P4P to a blended capitation payment model, and the differences in ordering patterns between traditional staffing and interdisciplinary teams within the blended capitation model. Using a propensity score weighted fixed-effects specification to address selection, the chapter estimates that a mandatory switch to capitation would lead to an average of 3% fewer laboratory requisitions per patient. Patients’ laboratory utilization also becomes more concentrated with the rostering physician. More importantly, using diabetes-related laboratory tests as a case study, physicians order 3% fewer inappropriate/redundant tests after joining the blended model and 9% fewer if they joined an interdisciplinary care team within the blended model. / Thesis / Doctor of Philosophy (PhD)

Laboratory testing is necessary when it contributes to the overall clinical management of the patient. Redundant testing, however, is often unnecessary and expensive and contributes to overall reductions in healthcare system efficiency. The purpose of this study is two-fold. First, to evaluate the frequency of ordering duplicate laboratory tests in hospitalized patients and the costs associated with this practice. Second, it was designed to determine if the use of a computerized alert or prompt will reduce the total number of unnecessarily duplicated Acute Hepatitis Profile (AHP) laboratory tests. This two-phase study took place in an inpatient facility that was part of a large tertiary care hospital system in Florida. A retrospective descriptive design was used during Phase 1 was to evaluate six laboratory tests and the frequency of ordering duplicate laboratory tests in hospitalized patients and to determine the associated costs of this practice for a 12-month time period in 2010. A test was considered a duplicate or an unnecessarily repeated test if it followed a previous test of the same type during the patient’s length of stay in the hospital and one in which any change in their values likely would not be clinically significant. A quasi-experimental pre- and post-test design was used during phase 2 was to determine the proportion of duplication of the AHP test before and after the implementation of a computerized alert intervention implemented as part of a system quality improvement process on January 5th, 2011. Data were compared for two 3-month time periods, pre- and post-alert implementation. The AHP test was considered redundant if it followed a previous test of the same type within 15 days of the initial test being final and present in the medical record. In phase 1, including each of the six tests examined, there were a total amount of 53, 351 test ordered, with 10, 375 (19.4%) of these cancelled. Out of the total amount of result final tests iv (n = 42,976), including each of the six tests examined, 4.6-8.7% were redundant. Results of the proportion of duplication of the six selected tests are as follows: AHP 196/2514 (7.8%), Antinuclear Antibody (ANA) 120/2594 (4.6%), B12/Folate level 396/5874 (6.7%), Thyroid Stimulating Hormone (TSH) 1893/21595 (8.7%), Ferritin 384/5171 (7.4%), and Iron/Total iron binding capacity (TIBC) 316/5155 (6.1%). The overall associated yearly cost of redundant testing of these six selected tests was an estimated $419, 218. The largest proportion of redundant tests was the Thyroid Stimulating Hormone level, costing a yearly estimated $300, 987. In Phase 2, prior to introduction of the alert, 674 AHP tests were performed. Of these, 53 (7.9%) were redundant. During the intervention period, 692 AHP tests were performed, of these 18 (2.6%) were redundant. The implementation of the computerized alert was shown to significantly reduce the proportion of AHP tests (Chi-Square: χ2 = df 1, p ≤ 0.001). The differences in the associated costs of duplicated AHP were $5238 dollars in 2010 as compared to $1746 in 2011 post-alert and these differences were significant (Mann Whitney U, Z = -4.04, p ≤ 0.001). Although the proportions of unnecessarily repeated diagnostic tests that were observed during Phase 1 of this study were small, the associated costs could adversely affect hospital revenue and overall healthcare efficiency. The implementation of the AHP computerized alert demonstrated a drop in the proportion of redundant AHP tests and subsequent associated cost savings. It is necessary to perform further research to evaluate computerized alerts on other tests with evidence-based test-specific time intervals, and to determine if such reductions postimplementation of AHP alerts are sustained over time.