Interrupted time series (ITS) analysis is a valuable study design for evaluating the effectiveness of population-level health interventions that have been implemented at a clearly defined point in time. It is increasingly being used to evaluate the effectiveness of interventions ranging from clinical therapy to national public health legislation. Whereas the design shares many properties of regression-based approaches in other epidemiological studies, there are a range of unique features of time series data that require additional methodological considerations. In this tutorial we use a worked example to demonstrate a robust approach to ITS analysis using segmented regression. We begin by describing the design and considering when ITS is an appropriate design choice. We then discuss the essential, yet often omitted, step of proposing the impact model a priori. Subsequently, we demonstrate the approach to statistical analysis including the main segmented regression model. Finally we describe the main methodological issues associated with ITS analysis: over-dispersion of time series data, autocorrelation, adjusting for seasonal trends and controlling for time-varying confounders, and we also outline some of the more complex design adaptations that can be used to strengthen the basic ITS design.

/pdf/interrupted-time-series-regression-for-the-evaluation-of-m3n2tlgxpo.pdf

Interrupted time series regression for the evaluation of public health interventions: a tutorial

https://purehost.bath.ac.uk/ws/files/190174563/The_need_for_a_complex_systems_model_of_evidence_for_public_health_Rutter_et_al_Lancet_2017_1.pdf

The need for a complex systems model of evidence for public health

Objective To evaluate the association between physical distancing interventions and incidence of coronavirus disease 2019 (covid-19) globally. Design Natural experiment using interrupted time series analysis, with results synthesised using meta-analysis. Setting 149 countries or regions, with data on daily reported cases of covid-19 from the European Centre for Disease Prevention and Control and data on the physical distancing policies from the Oxford covid-19 Government Response Tracker. Participants Individual countries or regions that implemented one of the five physical distancing interventions (closures of schools, workplaces, and public transport, restrictions on mass gatherings and public events, and restrictions on movement (lockdowns)) between 1 January and 30 May 2020. Main outcome measure Incidence rate ratios (IRRs) of covid-19 before and after implementation of physical distancing interventions, estimated using data to 30 May 2020 or 30 days post-intervention, whichever occurred first. IRRs were synthesised across countries using random effects meta-analysis. Results On average, implementation of any physical distancing intervention was associated with an overall reduction in covid-19 incidence of 13% (IRR 0.87, 95% confidence interval 0.85 to 0.89; n=149 countries). Closure of public transport was not associated with any additional reduction in covid-19 incidence when the other four physical distancing interventions were in place (pooled IRR with and without public transport closure was 0.85, 0.82 to 0.88; n=72, and 0.87, 0.84 to 0.91; n=32, respectively). Data from 11 countries also suggested similar overall effectiveness (pooled IRR 0.85, 0.81 to 0.89) when school closures, workplace closures, and restrictions on mass gatherings were in place. In terms of sequence of interventions, earlier implementation of lockdown was associated with a larger reduction in covid-19 incidence (pooled IRR 0.86, 0.84 to 0.89; n=105) compared with a delayed implementation of lockdown after other physical distancing interventions were in place (pooled IRR 0.90, 0.87 to 0.94; n=41). Conclusions Physical distancing interventions were associated with reductions in the incidence of covid-19 globally. No evidence was found of an additional effect of public transport closure when the other four physical distancing measures were in place. Earlier implementation of lockdown was associated with a larger reduction in the incidence of covid-19. These findings might support policy decisions as countries prepare to impose or lift physical distancing measures in current or future epidemic waves.

https://www.bmj.com/content/bmj/370/bmj.m2743.full.pdf

Physical distancing interventions and incidence of coronavirus disease 2019: natural experiment in 149 countries.

Countries in recession experience high unemployment rates and a decline in living conditions, which, it has been suggested, negatively influences their populations’ health. The present review examines the recent evidence of the possible association between economic recessions and mental health outcomes. Literature review of records identified through Medline, PsycINFO, SciELO, and EBSCO Host. Only original research papers, published between 2004 and 2014, peer-reviewed, non-qualitative research, and reporting on associations between economic factors and proxies of mental health were considered. One-hundred-one papers met the inclusion criteria. The evidence was consistent that economic recessions and mediators such as unemployment, income decline, and unmanageable debts are significantly associated with poor mental wellbeing, increased rates of common mental disorders, substance-related disorders, and suicidal behaviours. On the basis of a thorough analysis of the selected investigations, we conclude that periods of economic recession are possibly associated with a higher prevalence of mental health problems, including common mental disorders, substance disorders, and ultimately suicidal behaviour. Most of the research is based on cross-sectional studies, which seriously limits causality inferences. Conclusions are summarised, taking into account international policy recommendations concerning the cost-effective measures that can possibly reduce the occurrence of negative mental health outcomes in populations during periods of economic recession.

/pdf/mental-health-outcomes-in-times-of-economic-recession-a-21rictdx6w.pdf

Mental health outcomes in times of economic recession: a systematic literature review.

Signal of increased opioid overdose during COVID-19 from emergency medical services data.

Interrupted time series analysis differs from most other intervention study designs in that it involves a before-after comparison within a single population, rather than a comparison with a control group. This has the advantage that selection bias and confounding due to between-group differences are limited. However, the basic interrupted time series design cannot exclude confounding due to co-interventions or other events occurring around the time of the intervention. One approach to minimizse potential confounding from such simultaneous events is to add a control series so that there is both a before-after comparison and an intervention-control group comparison. A range of different types of controls can be used with interrupted time series designs, each of which has associated strengths and limitations. Researchers undertaking controlled interrupted time series studies should carefully consider a priori what confounding events may exist and whether different controls can exclude these or if they could introduce new sources of bias to the study. A prudent approach to the design, analysis and interpretation of controlled interrupted time series studies is required to ensure that valid information on the effectiveness of health interventions can be ascertained.

/pdf/the-use-of-controls-in-interrupted-time-series-studies-of-iu3roln2dw.pdf

The use of controls in interrupted time series studies of public health interventions.

Dear Editor,

In their letter, Librero et al. have highlighted one of the main issues affecting studies of the financial crisis (and one that we highlighted ourselves in the ‘Strengths and Limitations’ section of the original article), that is, ‘how do we define the financial crisis?’. As recommended when undertaking interrupted time series analyses of imperfectly identifiable events, we do this as objectively as possible by using an independent indicator—in this case gross domestic product (GDP).1 This is the most commonly used indicator to define a financial crisis; we use it in the same way as in similar studies. There is considerable debate among economists as to how to define a recession, …

The effect of the late 2000s financial crisis on suicides in Spain: an interrupted time-series analysis.

© 2020 Oxford University Press. All rights reserved. The originally published version of this article contained an algebraic definition of the regression model for interrupted time series (ITS) that could lead to erroneous interpretations of the estimated parameters. This model was presented in the equation at page 351, right column, and the following text. We provide here a more accurate definition.

Corrigendum to: Interrupted time series regression for the evaluation of public health interventions: a tutorial.

We thank Benmarhnia and Rudolph [REF] for their critical appraisal on our recent article on the use of controls in interrupted time series (ITS) studies. This offers the opportunity to clarify some important issues related to ITS and controlled ITS (CITS) designs and their comparison with other methods applied for public health evaluation. In particular, we argue that Benmarhnia and Rudolph based their assessment on three incorrect premises: that ITS without control is not a valid design for assessing causal relationships; that CITS is just another name for the difference-in-difference (DID) design that they advocate; that the synthetic control methodology represents an alternative to CITS.

James Lopez Bernal

Papers

Interrupted time series regression for the evaluation of public health interventions: a tutorial

The use of controls in interrupted time series studies of public health interventions.

The effect of the late 2000s financial crisis on suicides in Spain: an interrupted time-series analysis.

Corrigendum to: Interrupted time series regression for the evaluation of public health interventions: a tutorial.

Difference in difference, controlled interrupted time series and synthetic controls.