Objective: It is difficult to analyze medical and financial outcomes of health programs when randomized control study designs are not employed. This paper evaluates the ability of three statistical techniques to control for observed and unobserved differences between treatment groups when non-randomized studies are conducted.
Data: Data from an evaluation of a telephone-based nurse triage system are used to evaluate the ability of statistical techniques to control for differences between user groups.
Study Design: Two study designs are employed in the analysis: a pre/post study design and an intervention/comparison group study design. Three types of statistical techniques are used with the intervention/comparison study design to control for differences between treatment groups. The statistical techniques include direct adjustment, propensity model, and switching model with endogenous switching. ACGs are used to control for observed health status differences.
Results: The pre/post analysis indicates utilization of physician office services decreased after implementation of the nurse-triage service by 10.6%. The results of the intervention/comparison group study design, for all three statistical methods, suggest utilization increased. It appears the statistical techniques do not sufficiently control for unobserved health status and other characteristics that differ between the treatment groups and the results of the intervention/comparison study are not valid.
Conclusions: Recent improvements in statistical software, the quality of medical claims, and health status risk-adjustment tools enable researchers to use more sophisticated statistical techniques. The ability of these techniques to meet research objectives must be carefully evaluated before their results are considered valid.
Learning Objectives: Evaluate the ability of different statistical techniques to control for observed and unobserved differences between treatment groups when non-randomized studies are conducted
Keywords: Health Care Utilization, Health Care
Presenting author's disclosure statement:
Organization/institution whose products or services will be discussed: Access Health Group
I have a significant financial interest/arrangement or affiliation with any organization/institution whose products or services are being discussed in this session.
Relationship: Former employer