METHODS

Inclusion criteria

We included all published and unpublished reports in all languages describing systematic reviews or meta-analyses of the effects of any mode of transport or transport policy on health. Health effects included social, psychological, and physical effects that could be measured on humans.

Exclusion criteria

We excluded non-systematic literature reviews, descriptions of environmental or physical effects that did not include human responses to them, behavioural interventions without objectively measured outcomes, predicted but not empirical health impacts, and reviews that described the effects of intermediate mechanisms by which transport affects health—such as exercise, walking or cycling—without evaluating the effectiveness of policies or programmes to bring about these changes.

Search methods

We searched the following electronic databases plus the world wide web using www.Google.com: Ovid Medline (1966–1/2001), EMBASE (1980–1/2001), CINAHL (1982–1/2001), DARE (1/2001), ERIC (1966–1/2001), Cochrane Database of Systematic Reviews (2000, Issue 4) and PsycINFO (formerly PSYCLIT) (1971–1/2001) using the search terms meta-analys$, metaanalys$, systematic AND review, evaluation synthesis or research synthesis, limited to human subjects, where $ denotes all suffixes. Keywords were transport, car, cars, bus, automobile$, traffic or vehic$. TRANSPORT—which comprises the three databases TRIS, ITRD and TRANSDOC—was electronically searched from its first entries in the 1960s to February 2001 for all meta-analyses and systematic reviews without any additional restrictions on the initial search because all entries should be relevant to transport. Bibliographies of selected papers were searched and experts in the field were asked to identify other relevant reviews.

Evaluation of included reviews

Two reviewers independently screened abstracts then scored papers using Oxman and Guyatt’s index for quality assessment of reviews,²¹ using the nine criteria given in figure 1. A quality index, based on these criteria, could range from 1 (major flaws) to 7 (minimal flaws).

We have presented brief summaries of the most important findings of each review in the Results.

RESULTS

Altogether 3183 reports were identified in the first electronic search and all abstracts read to select those that met inclusion criteria. A total of 127 candidate papers and reports were selected. All were obtained and considered for suitability. Twenty papers were agreed to meet the inclusion criteria. A further eight eligible reviews were identified through contact with experts and bibliographies giving a total of 28 systematic reviews and meta-analyses. All reviews that appeared in the same special edition of the American Journal of Preventive Medicine were considered with an introductory paper²² that described the common methodology of all their searches.

We classified reviews into four categories of intervention: health promotion, engineering, environmental, and legislative. Where reviews covered more than one type of intervention, the dominant area determined its classification in our summary. A brief summary of the results of each review is given in tables 1–4. Only six reviews, all of which described health promotion interventions, had minimal flaws.

Health promotion interventions

Systematic reviews with minimal flaws identified beneficial effects of primary care based counselling to prevent childhood injury and efforts to increase bicycle and motorcycle helmets,^{24, 29, 30} and raising the minimum drinking age above 18 years.^{29, 30} Programmes to increase car seat and seatbelt use by children are effective but their benefits seem to be comparatively shortlived.³⁷ There are mixed effects of drink driving remediation, with some interventions, such as the combination of rehabilitation and probation, being associated with a potential increase in the risk of injury.³⁸ Driver improvement and education courses are associated with an increase in crash involvement and violations.³²

Reviews with more methodological flaws, in addition to providing further information on topics already covered by better quality reviews, described beneficial effects of road safety and education campaigns.^{26–28, 33, 34}

Engineering improvements

Five reviews were identified although four^{40, 41, 43, 44} had major methodological flaws (table 2). Elvik’s review of traffic calming schemes⁴² had a Quality Index of 6. It found that traffic calming schemes had a mean effect of reducing accidents by 15% and that similar effect sizes were found in different decades and in different countries.

There is some evidence from less methodologically rigorous reviews for the effects of ignition interlock devices, studded tyres, daytime running lights, and measures to reduce vehicle speeds. Ignition interlock devices, which require drivers to record a legal breath alcohol level before the car engine can be started, were associated with reductions in re-conviction and re-arrest rates for driving while intoxicated of at least a third.⁴⁰ Studded tyres had mixed effects, on some road surfaces increasing accident rates and in others reducing them.⁴¹ Laws to enforce the use of studded tyres were associated with statistically significant increases in accident rates of 3%–10%. The use of sidelights or dipped headlights during the day (daytime running lights) was associated with reductions in accidents involving more than one vehicle of about 14%, but this effect was much greater in more northerly countries.⁴³ Speed limit zones are effective in reducing personal accidents and material damage.⁴⁴ Creating raised road surfaces at crossroads may increase accidents, while noisy road surfaces (rumble strips) before crossroads are associated with reductions in accidents. Road humps and differential speed limits may reduce accidents locally but increase them in surrounding areas.

Environmental interventions

Three reviews considered environmental interventions (table 3) and all had more than minor methodological flaws.

Elvik’s review^{45, 46} found that public lighting reduced night time accidents in all cases but depended on the baseline risk and the proportion of night time accidents.⁴⁵ Guard rails and crash cushions were found to increase numbers of accidents but decrease their severity.⁴⁷ One review,⁴⁸ with a Quality Index of 3, aimed to describe the effects of modifiable risk factors on child pedestrian injuries. Age, behaviour, race, and sex were considered to be among the strongest risk factors although it is not clear how any but behaviour might be modified.

Legislative interventions

Six reviews considered the effectiveness of legislation on accidents, injuries, and drinking and driving (table 4). A review that had a quality score of 6 found that laws for a maximum legal blood alcohol concentration of 0.02% reduced night time injuries and fatal crashes.⁵¹ The introduction of random breath alcohol testing is associated with a reduction in alcohol related hospital admissions, deaths, injuries, night time crashes, and charges for drink driving by around a fifth.⁵³ A review of laws against drink driving in the absence of any other offence (“administrative per se”)⁵² had minor flaws and found inconclusive results although no harmful effects. Other studies had more methodological flaws and their results should be interpreted with caution. Legislation to deter drinking and driving was effective in most cases although several studies on mandatory jail sentences showed increases in crashes after legislation.^{49, 50} Lighter sanctions against drink drivers increased subsequent crash rates by 7%³⁹ (see table 1). Foss⁵⁴ found that graduated driver licensing (where there is progressive freedom after passing the driving test to drive unaccompanied and at night) and night time curfews were associated with variable reductions in deaths and accidents. These were confounded to some extent by reductions in the rate of licensure among teenagers. Laws to enforce car seat belt use by adults increased belt use and reduced serious and fatal injuries.⁵⁵

DISCUSSION

Evidence is available in well conducted systematic reviews both to support a range of transport related interventions that will benefit health and to indicate interventions that are intended to improve health but are in fact harmful and should not be implemented. Beneficial interventions include health promotion campaigns to prevent childhood injuries, efforts to increase bicycle and motorcycle helmet use, children’s car seat and seatbelt promotion, traffic calming, and specific legislation against drink driving. Driver improvement and education courses may increase accidents by encouraging greater numbers of inexperienced drivers on to the roads. We agree with comments made in a review⁵⁶ published during the preparation of this paper that they cannot be recommended. Evidence for other health effects of transport interventions has also been identified in systematic reviews that have more methodological flaws. Ignition interlock devices, daytime running lights, public lighting, graduated driver licensing, and laws to enforce seatbelt use may all be effective in improving health. The evidence is equivocal on the benefits and harms of guard rails, crash cushions, and interventions to reduce vehicle speeds. Some modes of drink driving remediation, including mandatory jail sentences and laws to enforce studded tyres, are associated with harmful effects on health. Further research, beginning with more rigorously conducted systematic reviews, is needed to determine whether these findings are valid.

The scope of this review

All 28 systematic reviews and meta-analyses on transport and health that we identified were concerned with injury prevention and all but four were concerned with preventing motorcar injuries. We did not use accident prevention as a primary search term because published and ongoing reviews exist in this area.^57–59 We found no evidence to suggest that any intervention will bring about a shift in the use of different modes of transport and as a result improve health in the broader ways suggested in the Introduction. Such evidence may exist in primary studies that have not been synthesised in systematic reviews. For those who support public policies to promote walking, cycling, and public transport, experimental evidence may be superfluous and the intrinsic value of reducing our dependence on the car may be self evident. But this review has identified several counterintuitive effects of apparently beneficial transport programmes. We therefore believe that evidence of outcomes should guide transport interventions that are intended to benefit health and that good intentions are not enough.

Information from qualitative and quantitative research that describes important relations between transport use and broader determinants of health, for example exercise⁶ or severance of communities by roads⁶⁰ may be helpful both in designing more effective new interventions and in understanding how the social and environmental context of an intervention may influence its effectiveness. It was not our aim to review the entire body of literature on how transport affects health, however, but to identify where population based interventions have been shown to impact upon health.

Because of the loss of information in summarising reviews we would recommend that anyone considering using their findings obtains the original papers.

Quality of included reviews

Oxman and Guyatt’s quality scores²¹ for each study indicate that only six reviews achieved a score of 7, indicating minimal flaws. Six studies had a score of 3 or less, indicating major flaws. Failure to demonstrate that selection bias had been avoided and inappropriate assessment of validity were the commonest methodological problems overall. Oxman and Guyatt scores mark down failure to report review methods, even if they have been carried out. It is therefore important for systematic reviews to make their methodology explicit. Improvements in database search engines are also needed so that, for example, controlled trials can be more readily identified.⁶¹

One implication of these results is that systematic reviews and meta-analyses have a useful contribution to evidence based policymaking for transport. As evidence of the health effects of other interventions continues to accrue, regularly updated systematic reviews will be required. Some of this work may fall within the remit of the newly established Campbell Collaboration (http://campbell.gse.upenn.edu/). Systematic reviews do not, however, obviate the need for high quality primary studies as important sources of evidence to improve health through transport choices.

Conclusions

Systematic reviews are able to provide evidence about effective ways of improving health through transport related interventions. The best evidence indicates that health promotion campaigns to prevent childhood injuries, increase bicycle and motorcycle helmet use, and children’s car seat and seatbelt use, plus traffic calming, and specific legislation against drink driving are all beneficial, while driver improvement and education courses may be harmful. A systematic review of primary studies that embraces a wider range of possible health effects of transport—including social and environmental effects—is required and work on this is underway by the authors. This should determine if evidence is available to support claims for a spectrum of health effects, whether this evidence is of acceptable quality, and if not, where new primary studies should be directed. In the meantime, we suggest that policymakers, planners and health professionals put the available evidence into practice and monitor its benefits on health.