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

Summary Background Intensive care units (ICUs) are high-risk areas for transmission of antimicrobial-resistant bacteria, but no controlled study has tested the effect of rapid screening and isolation of carriers on transmission in settings with best-standard precautions. We assessed interventions to reduce colonisation and transmission of antimicrobial-resistant bacteria in European ICUs. Methods We did this study in three phases at 13 ICUs. After a 6 month baseline period (phase 1), we did an interrupted time series study of universal chlorhexidine body-washing combined with hand hygiene improvement for 6 months (phase 2), followed by a 12–15 month cluster randomised trial (phase 3). ICUs were randomly assigned by computer generated randomisation schedule to either conventional screening (chromogenic screening for meticillin-resistant Staphylococcus aureus [MRSA] and vancomycin-resistant enterococci [VRE]) or rapid screening (PCR testing for MRSA and VRE and chromogenic screening for highly resistant Enterobacteriaceae [HRE]); with contact precautions for identified carriers. The primary outcome was acquisition of resistant bacteria per 100 patient-days at risk, for which we calculated step changes and changes in trends after the introduction of each intervention. We assessed acquisition by microbiological surveillance and analysed it with a multilevel Poisson segmented regression model. We compared screening groups with a likelihood ratio test that combined step changes and changes to trend. This study is registered with ClinicalTrials.gov, number NCT00976638. Findings Seven ICUs were assigned to rapid screening and six to conventional screening. Mean hand hygiene compliance improved from 52% in phase 1 to 69% in phase 2, and 77% in phase 3. Median proportions of patients receiving chlorhexidine body-washing increased from 0% to 100% at the start of phase 2. For trends in acquisition of antimicrobial-resistant bacteria, weekly incidence rate ratio (IRR) was 0·976 (0·954–0·999) for phase 2 and 1·015 (0·998–1·032) for phase 3. For step changes, weekly IRR was 0·955 (0·676–1·348) for phase 2 and 0·634 (0·349–1·153) for phase 3. The decrease in trend in phase 2 was largely caused by changes in acquisition of MRSA (weekly IRR 0·925, 95% CI 0·890–0·962). Acquisition was lower in the conventional screening group than in the rapid screening group, but did not differ significantly (p=0·06). Interpretation Improved hand hygiene plus unit-wide chlorhexidine body-washing reduced acquisition of antimicrobial-resistant bacteria, particularly MRSA. In the context of a sustained high level of compliance to hand hygiene and chlorhexidine bathings, screening and isolation of carriers do not reduce acquisition rates of multidrug-resistant bacteria, whether or not screening is done with rapid testing or conventional testing. Funding European Commission.

Interrupted time series is a strong quasi-experimental research design to evaluate the impacts of health policy interventions. Using simulation methods, we estimated the power requirements for interrupted time series studies under various scenarios. Simulations were conducted to estimate the power of segmented autoregressive (AR) error models when autocorrelation ranged from -0.9 to 0.9 and effect size was 0.5, 1.0, and 2.0, investigating balanced and unbalanced numbers of time periods before and after an intervention. Simple scenarios of autoregressive conditional heteroskedasticity (ARCH) models were also explored. For AR models, power increased when sample size or effect size increased, and tended to decrease when autocorrelation increased. Compared with a balanced number of study periods before and after an intervention, designs with unbalanced numbers of periods had less power, although that was not the case for ARCH models. The power to detect effect size 1.0 appeared to be reasonable for many practical applications with a moderate or large number of time points in the study equally divided around the intervention. Investigators should be cautious when the expected effect size is small or the number of time points is small. We recommend conducting various simulations before investigation. Copyright © 2011 Elsevier Inc. All rights reserved.

There has been a recent increase in interest in alternatives to randomisation in the evaluation of public health interventions. We aim to describe specific scenarios in which randomised trials may not be possible and describe, exemplify and assess alternative strategies. Non-systematic exploratory review. In many scenarios barriers are surmountable so that randomised trials (including stepped-wedge and crossover trials) are possible. It is possible to rank alternative designs but context will also determine which choices are preferable. Evidence from non-randomised designs is more convincing when confounders are well-understood, measured and controlled; there is evidence for causal pathways linking intervention and outcomes and/or against other pathways explaining outcomes; and effect sizes are large. Non-randomised trials might provide adequate evidence to inform decisions when interventions are demonstrably feasible and acceptable, and where evidence suggests there is little potential for harm, but caution that such designs may not provide adequate evidence when intervention feasibility or acceptability is doubtful, and where existing evidence suggests benefits may be marginal and/or harms possible.