12.07.2015 Views

Deliverable D 1.3 RESULTS OF META-ANALYSIS ... - cast-eu.org

Deliverable D 1.3 RESULTS OF META-ANALYSIS ... - cast-eu.org

Deliverable D 1.3 RESULTS OF META-ANALYSIS ... - cast-eu.org

SHOW MORE
SHOW LESS

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

Campaigns and awareness raising strategies in traffic safety — <strong>Deliverable</strong> D-1.1Table 7.24. Overall effect of road safety campaigns on road accidents and seatbelt useaccording to use of control group, after accounting for publication bias. Each overall effect isbased on n individual effects (excluding any artificial effects generated by trim-and-fill).AccidentsSeatbelt usedesigneffect(%)95%confidenceinterval (%)nSig.?(p < 0.05)effect(%)95%confidenceinterval (%)nSig.?(p < 0.05)control -7 (-12;-2) 87 yes +17 (+9;+25) 48 yesnocontrol -6 (-8;-4) 28 yes +31 (+21;+41) 84 yesThere is little overall difference between those accident effects reported bycontrolled studies (-7 per cent) and those reported by non-controlled studies (-6 per cent). At 31 per cent, the overall seatbelt use effect for non-controlledstudies is higher than that for controlled studies (17 per cent), but thisdifference is only apparent after accounting for publication bias. The actualoverall effect reported by controlled studies before adjustment was 51 percent (with a 95 per cent confidence interval of 43 to 60 per cent), comparedwith 48 per cent for non-controlled studies (with a 95 per cent confidenceinterval of 38 to 58 per cent). In other words, before accounting for publicationbias, overall campaign effects were similar in size and variability whether theywere based on changes in accident counts or seatbelt use. This gives usconfidence to use the effects derived from non-controlled studies.A further aspect of evaluation design that may influence the apparent effect ofcampaigns is whether the outcome measure is recorded before and during orbefore and after the road safety campaign. Some studies took outcomemeasurements before, during and after the campaign, in which case anaverage of during and after measurements was taken for the database, andthe effect allocated greater weight accordingly. An analysis was carried out tosee if there are differences for effects derived from these three different typesof study. This analysis informs whether the campaigns in this sample aremore effective while they are ongoing or after they are complete.75

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!