Article type
Year
Abstract
Background: The synthesis of evidence through systematic reviews and meta-analytic methods are reliant on the unbiased publication and complete reporting of outcomes and results from the primary studies. Many systematic reviews restrict their inclusion criteria to randomised trials; however, for many research questions randomisation is not ethical or feasible (e.g. examining the effect of coronavirus control measures on transmission rates). In these instances, the interrupted time series (ITS) design is a suitable alternative. For this design, a series of measurements on groups of individuals are collected at regular time points before and after an interruption. The period before the interruption can be used to estimate the underlying time trend and, if modelled correctly, when projected into the post-interruption period, can be used to obtain an estimate of what would have occurred in the absence of the interruption. Biases in the evidence base can arise from the publication of outcomes and/or results that favour the investigators’ hypothesis, or if greater prominence is given to results with large effect sizes and small p-values. The issues associated with selective inclusion of outcomes, and results from randomised trials have been well documented. However, we are not aware of studies examining these issues for ITS studies.
Objectives: To assess the extent of selective reporting of outcomes and results in ITS studies.
Methods: We used a previously curated sample of 200 ITS studies with a public health outcome and located the corresponding published protocols and/or statistical analysis plans (SAPs). We then identified any discordance between the details defined in the protocols/SAPs to those reported in the final publications with respect to: the design (e.g. timeframe for which data are sought, timing of the interruption(s), time interval of data aggregation), outcome (e.g. description and classification at the individual and aggregate level), statistical methods (e.g. estimation method, handling of autocorrelation and seasonality), and effect measures (e.g. level change and slope change).
Results/Conclusions: We will quantify the extent of selective reporting of outcomes and results in ITS studies. This is critical to understanding the validity of meta-analytic findings of ITS studies, on which important public health funding decisions are often based.
Patient or healthcare consumer involvement: No patients/consumers were involved in the design/reporting of this study.
Objectives: To assess the extent of selective reporting of outcomes and results in ITS studies.
Methods: We used a previously curated sample of 200 ITS studies with a public health outcome and located the corresponding published protocols and/or statistical analysis plans (SAPs). We then identified any discordance between the details defined in the protocols/SAPs to those reported in the final publications with respect to: the design (e.g. timeframe for which data are sought, timing of the interruption(s), time interval of data aggregation), outcome (e.g. description and classification at the individual and aggregate level), statistical methods (e.g. estimation method, handling of autocorrelation and seasonality), and effect measures (e.g. level change and slope change).
Results/Conclusions: We will quantify the extent of selective reporting of outcomes and results in ITS studies. This is critical to understanding the validity of meta-analytic findings of ITS studies, on which important public health funding decisions are often based.
Patient or healthcare consumer involvement: No patients/consumers were involved in the design/reporting of this study.