This subproject is one of many research subprojects utilizing the resources provided by a Center grant funded by NIH/NCRR. The subproject and investigator (PI) may have received primary funding from another NIH source, and thus could be represented in other CRISP entries. The institution listed is for the Center, which is not necessarily the institution for the investigator. Quasi-experimental interrupted time series (with comparison series) studies are the design of choice to study the effects of health policies because randomized controlled policy evaluation trials are rarely feasible. Interrupted time series studies allow health policy researchers to estimate whether and how much a policy intervention changed an outcome of interest, given the level and trend of the outcome before the policy;whether the effects occurred immediately after the intervention or with delay;whether the effects were transient or long-term;and whether other factors than the policy intervention could explain observed changes. Time series data require segmented regression analysis methods that account for correlations of the data across time points with autoregressive error models. To design interrupted time series studies, it is necessary to know how many time points and how many observations at each time point are needed to obtain stable estimates of policy effects. To date, no satisfactory method exists to calculate the power and necessary sample size for interrupted time series studies. We propose to develop and test a new method to estimate power and sample size for interrupted time series studies of policy effects.
The specific aims of the proposed research are to: 1. Develop a simulation-based statistical method and computation algorithms to estimate power and calculate the sample size for segmented regression analyses using a likelihood ratio test statistic. 2. Compare the power and sample size calculations of the simulation-based approach to an existing Wald-type statistic-based approach of a Simple Intervention Analysis (SIA) model, which only allows for testing the behavior of a single parameter (level or trend change, not both). 3. Describe how event (outcome) rates, their variability, and correlations over time (autoregressive behavior) and interventions of different effect size influence power and the sample size estimates. 4. Explore the possibilities to extend the approach to more complex models of interest to the statistical community (generalized autoregressive conditional heteroskedasticity [GARCH] models, which allow error variability to change over time, for example between a pre- and post intervention period). The proposed research would contribute in the following three ways: (1) It would enable pharmaceutical policy and other researchers to design time series studies with sufficient power to detect effects and document power calculations as required in grant applications;(2) it would fulfill the task of world health organization (WHO) to improve methods for pharmaceutical policy and other intervention research globally;and (3) it would provide preliminary data for future grant applications to further develop methods to analyze longitudinal observational data. We request 30,000 service units.
Showing the most recent 10 out of 292 publications
