Skip to main content

Table 3 Summary table on risk of bias of the included interrupted time series studies (excludes two studies (35, 36) that could not be assessed)

From: A systematic review: effectiveness of mass media campaigns for reducing alcohol-impaired driving and alcohol-related crashes

ITS Intervention independent of other changes Shape of the intervention effect pre-specified Intervention unlikely to affect data collection Knowledge of the allocated interventions adequately prevented during the study Incomplete outcome data adequately addressed Study free from selective outcome reporting Study free from other risks of bias
Whittam 2006 [18] Low risk (ARIMA model used and had comparator site) Low risk Low risk Low risk Low risk High risk (crash data only for 16–19 year olds) Low risk (has comparator site, using ARIMA model)
Good quality study
Murry 1993 [19] Low risk (authors state that data were transformed to isolate the experimental effect from any extraneous influences) Low risk Low risk Low risk Low risk High risk (using proxy indicator, using certain age group only) Low risk (using comparator site, using model)
Good quality study
Newstead 1995 [20] Low risk (regression model used to account for other factors) Low risk Low risk Low risk Low risk Unclear risk (using proxy indicator) Low risk (using regression model, but no comparator site)
Good quality study
Tay 2002 [21] Low risk (used regression models to exclude other factors) Low risk Low risk Low risk Unclear risk (States that some inconsistencies may exist in the reporting as done by local police) Unclear risk (used proxy measures) Low risk
Good quality study
Fell 2008 [23] High risk (other enforcement measures took place) Low risk Low risk Low risk Low risk Low risk Low risk
Intermediate quality study
Zwicker 2007a [24] High risk (enforcement also took place) Low risk Low risk Low risk Low risk Low risk Low risk (used ARIMA model and applied parameters to model periodic fluctuations in the crash rates)
Intermediate quality study
Zwicker 2007b [25] High risk (enforcement also took place) Low risk Low risk Low risk Low risk Low risk Low risk (contiguous county data were used to remove factors that may have obscured the effect of the campaign on the trend)
Intermediate quality study
Lacey 2008 [26] High risk (law enforcement activities also took place) Low risk Low risk Low risk Low risk Low risk Low risk (ARIMA model used)
Intermediate quality study
Epperlein 1987 [29] Unclear risk (no comparator site, but daytime crashes used to account for other changes) Low risk Low risk Low risk Low risk Unclear risk (using proxy indicator) High risk, not using model
Low quality
Miller 2004 [30] High risk (media campaign done together with other enforcements, though model was used to look at each interventions) Low risk Low risk Low risk Low risk Unclear risk (using proxy indicator, fatal nighttime crashes) High risk (ARIMA model was used, but evaluation of mixed approaches in different areas over different period)
Low quality study
Agent 2002 [27] High risk (enforcement activities also took place as part of the campaign) Low risk Low risk Low risk Low risk Unclear risk (documentation of alcohol use is dependent on the reporting officer) High risk (no model used)
Low quality study
Solomon 2008 [28] High risk (enforcement measures also in place) Low risk Low risk Low risk Low risk Low risk High risk (Only looking at changes in absolute numbers, no application of models, no comparator site)
Low quality study
Beck 2009 [31] High risk (enforcement also took place) High risk (point of analysis not clear) Low risk Low risk Low risk Low risk High risk (only looking at the absolute number of alcohol-related crashes, not using any models or accounting for rates in comparator sites)
Low quality study
NHTSA 2007 [32] High risk (enforcement also took place) Low risk Low risk Low risk Low risk Low risk High risk (compared with non-intervention sites, but no model used. Unclear if other factors accounted for)
Low quality study
Suriyawongpaisal 2002 [33] High risk (enforcement measures also used) High risk (point of analysis is not the point of intervention, and not clearly stated why the data points were selected) Low risk High risk (hospital staff of the study sites were not blinded, and could have affected how they collected data) High risk (not sure what proportion of cases were missed in each period, data collection dependent on hospitals enrolled) Unclear risk High risk (the study did not account for other changes that could have affected the outcome)
Low quality study
  1. ARIMA, Autoregressive Moving Average Model; ITS, Interrupted Time Series; NHTSA, National Highway Traffic Safety Administration