PT - JOURNAL ARTICLE AU - Celestin Hategeka AU - Hinda Ruton AU - Mohammad Karamouzian AU - Larry D Lynd AU - Michael R Law TI - Use of interrupted time series methods in the evaluation of health system quality improvement interventions: a methodological systematic review AID - 10.1136/bmjgh-2020-003567 DP - 2020 Oct 01 TA - BMJ Global Health PG - e003567 VI - 5 IP - 10 4099 - http://gh.bmj.com/content/5/10/e003567.short 4100 - http://gh.bmj.com/content/5/10/e003567.full SO - BMJ Global Health2020 Oct 01; 5 AB - Background When randomisation is not possible, interrupted time series (ITS) design has increasingly been advocated as a more robust design to evaluating health system quality improvement (QI) interventions given its ability to control for common biases in healthcare QI. However, there is a potential risk of producing misleading results when this rather robust design is not used appropriately. We performed a methodological systematic review of the literature to investigate the extent to which the use of ITS has followed best practice standards and recommendations in the evaluation of QI interventions.Methods We searched multiple databases from inception to June 2018 to identify QI intervention studies that were evaluated using ITS. There was no restriction on date, language and participants. Data were synthesised narratively using appropriate descriptive statistics. The risk of bias for ITS studies was assessed using the Cochrane Effective Practice and Organisation of Care standard criteria. The systematic review protocol was registered in PROSPERO (registration number: CRD42018094427).Results Of 4061 potential studies and 2028 unique records screened for inclusion, 120 eligible studies assessed eight QI strategies and were from 25 countries. Most studies were published since 2010 (86.7%), reported data using monthly interval (71.4%), used ITS without a control (81%) and modelled data using segmented regression (62.5%). Autocorrelation was considered in 55% of studies, seasonality in 20.8% and non-stationarity in 8.3%. Only 49.2% of studies specified the ITS impact model. The risk of bias was high or very high in 72.5% of included studies and did not change significantly over time.Conclusions The use of ITS in the evaluation of health system QI interventions has increased considerably over the past decade. However, variations in methodological considerations and reporting of ITS in QI remain a concern, warranting a need to develop and reinforce formal reporting guidelines to improve its application in the evaluation of health system QI interventions.