The Effect of Police Density to Improve Motorcycle Helmet Use in Thailand Suriyawongpaisal P1, Thakkinstian A2, Jiwattanakulpaisarn P3 1Professor, Department of Community Medicine, Faculty of Medicine, Ramathibodi Hospital, Mahidol University, Thailand, 2Assistant Professor, Clinical Epidemiology Unit, Faculty of Medicine, Ramathibodi Hospital, Mahidol University, Thailand, 3Transport policy analyst, Department of Highways, Ministry of Transport, Thailand. ABSTRACT While it is well established that motorcycle helmets reduce the risk of fatal injury and death, scaling up law enforcement on helmet use is still a big challenge for developing countries. These countries have a heavy burden from motorcycle-related injuries, but have limited resources for law enforcement. This report used cubic spline regression analysis on national datasets and a nationwide roadside survey of helmet use in Thailand. It has demonstrated that police density (number of policemen per square kilometer) is an independent predictor of helmet use (p<0.001) after controlling for adult literacy. This finding is consistent with evidence about effective deployment of limited numbers of police to ensure substantial and sustainable reductions of fatal road crashes in the rural road network in Queensland State of Australia. Concerning policy decisions about resource allocation, the finding provides a new parameter for allocation of limited numbers of police. Police density was also significantly associated with gross provincial product (GPP) per capita (p<0.001) and population density (p < 0.001). These additional important findings raise concerns about equitable allocation of the police force in Thailand. Keywords: Police Density, Helmet use, Law Enforcement, Thailand INTRODUCTION Globally, road traffic injuries (RTIs) resulted in 1.3 million deaths in 2004 with a major share in developing countries1. Together with the magnitude of burden, rising trends of RTIs in many developing countries have concerned public health authorities and governments. In Thailand, it is estimated that RTIs killed 28,600 in 20042. Motorcycle related injury constitutes the majority of RTIs in countries such as Thailand and Vietnam3,4. The cost of RTIs in Thailand was estimated at 232,855 million Baht or 2.81% of the Gross National Product in 20045. A meta-analysis reported that helmet use reduces the risk of head injury by 72% and are also likely to reduce the risk of death, although the effectiveness may be modified by speed6. A report of cost effectiveness modeling revealed that motorcycle helmet legislation would cost 495-784 USD/disability adjusted life year (DALY)7. Similarly, legislation for helmet use applies only on certain assigned routes and national roads in Vietnam which resulted in five times the prevalence on city roads8. Nevertheless, the helmet use rate has remained low in developing countries8,9,10. In addition to law enforcement, a study in a typical Malaysian town by Kulanthayan et al., (2000)9revealed compliance of helmet use was related to age, gender, race, formal education level, prior accident experiences and type of license held. More recently, Hongsranagon et al., (2011)11 have shown a link between helmet use and perceived risk of road crashes in Thailand. With substantial growth of motorcycle ownership and use in large cities and provincial areas of many developing countries12,13,14, it is a challenge to scale up law enforcement on helmet use nationwide. To meet this challenge, there is a need for political commitment, evidence-based planning, resource mobilisation plus allocation and effective implementation. In countries with national legislation, helmet use rates and patterns may vary considerably from one region to the next depending on educational level, penalty charged and enforcement activities15. It has been recognized that a major obstacle to attain and sustain a high level of law enforcement in low-income and middle-income countries is limited police resources15. This issue has not been clearly addressed despite the availability of the WHO manual for policy decisionmakers and practitioners on helmet use16. Recently, a study in Thailand revealed the cost of general risk-behavior monitoring per suspect was 0.69 baht, while the unit cost of detecting the use of safety Indian Journal of Public Health Research & Development. January-March 2013, Vol. 4, No. 1 203 devices, such as seat belts, was about 6–8 baht (0.2 USD) extra per offender16. Per check-point, the cost of this safety device use detection was 1,547 baht (52 USD). Making use of the unit cost knowledge requires answering 2 key questions. 1) What is the total number of police officers required to provide an acceptable level of service? 2) How should a specified total number of these officers be allocated by geographical regions or time periods to maximise helmet use15. Concomitantly, a national policy decision was made in Thailand to implement an extensive motorcycle helmet use campaign across the country starting in 2010. This could potentially have been another ad hoc campaign without sustained effect similar to previous ones since the helmet law came into effect on a national scale in 1994. Out of this concern, a large scale national survey of helmet use in all 76 provinces was undertaken to monitor the policy implementation17. This survey revealed a two fold discrepancy in the percentage of helmet use across geographical regions. A larger discrepancy of 4.3 folds was observed between provinces in Thailand. Making use of the survey findings and other relevant datasets, this report aimed to shed some light on the association between law enforcement related factors and rate of helmet usage in order to better guide the allocation of police manpower with respect to this policy decision. MATERIALS AND METHOD A. Datasets According to a study using multiple correspondence analysis, the several potential predictors (adult literacy, population density and gross domestic product) were associated with road safety performance of 23 selected E.U. countries18. To our best knowledge, the potential predictors which have been identified in this report are the best available evidence in this subject area. For the purpose for this report, the following datasets were obtained: a) projected population, estimated gross provincial product (GPP) and province-specific adult literacy in 2010 were retrieved from the websites of the National Economic and Social Development Board (www.nesdb.go.th); b) number of police in 2010 for each province from the website of the Royal Thai Police ( http://www.personnel.police.go.th/index.php); c) province specific area in square kilometers from the website of the Department of Provincial Administration (http://www.dopa.go.th/padmic/jungwad76/ jungwad76.htm); and d) helmet use from the nationwide roadside survey17. The source of data on motorcycle helmet use was the database from the nationwide roadside survey20. Detailed sampling design and data analysis were presented in the report. In brief, the nationwide survey, undertaken from May to December in 2010, employed direct observation about helmet use among 945,956 randomly chosen motorcyclists (71.4% of which were drivers). These were identified at 3,252 selected sites comprising road intersections or road sections with slow traffic in urban and rural municipalities of varying sizes. The number and proportionate distribution of the selected sites (as percentages of the total sites for each category) of large, medium and small municipalities was: 1276 sites (32.9%), 560 sites (17.2%), and 1416 sites(43.5%) respectively. Between provinces, the number of sites varied from 22 to 84 according to area and number of population. For Bangkok, the capital city with a population of 6.9 million19, 100 sites were chosen (two for each of the total of 50 districts). Data about the number and percentage of helmet uses for each province were summarised and used for analysis (full detailed tables are available on request). B. Statistical analysis A Pearson correlation matrix was applied to estimate the correlation between variables. Since the data were highly skewed, a cubic spline regression was applied to assess the relationship between predictors and the outcome20. The interested outcome was percentage of motorcycle helmet use whereas the predictors were police density and adult literacy. GPP, population density, and the number of policemen were not included in the model because these were highly correlated with police density which would cause multicollinearity. The percentage of helmet use, police density and adult literacy were included in the spline regression with degrees of 0 and 3 (2 knots) for police density, and 0 (linear) for adult literacy. Goodness of fit of the model was checked. The predicted percentage of helmet use was then estimated and described for each province. Apart from these relationships, we were also interested in assessing the relationship between police density and other predictors i.e., population density, adult literacy and GPP. In this regard, police density was treated as a dependent variable in spline regression analysis with population density, adult literacy and GPP included as independent variables. All analyses were performed using STATA version 12.0. A p value <0.05 was considered as statistically significant. 204 Indian Journal of Public Health Research & Development. January-March 2013, Vol. 4, No. 1 RESULTS For the whole country, the percentage of helmet use in drivers (53.3%) was higher than that in passengers (19.3%) (Table1). Regional comparison revealed the highest percentage of helmet use by both drivers (93.0%) and passengers (45.2%) in Bangkok, which was two folds of the lowest figure in the North for drivers and almost five folds of the lowest figure in the South for passengers. Except for the population to police ratio (562.9 persons per police), Bangkok has the highest population density (1330.4 persons per square kilometer), the highest GPP per capita (361,243 baht per capita) and the highest police density (13.09 policemen per square kilometer) compared to other regions in 2010. Using province specific data for the whole country, the percentages of helmet use in both drivers and passengers were significantly correlated to the number of police (r=0.3613), police density (r=0.5121), population density (r=0.5864) and GPP (r=0.4944) (Table 2). In the same table, police density has been shown to be highly correlated to population density (r=0.9515) and GPP (r=0.9074) at a statistically significant level (p<0.001). Using spline regression analysis, it was found that police density was independently the strongest predictor of helmet use at p < 0.001 (Table 3). Again with spline regression, population density and GPP were found to be independent and significant predictors of police density at p < .01(Table 4). DISCUSSION Evidence has shown that careful planning in policing, could reduce road crashes by an accumulated effect of 12% of all severities and 15% of all fatal crashes with low levels of police enforcement and sustained effects21. Numerous attempts have been made to increase the effectiveness of police enforcement in reducing traffic accidents22,23,24. Most attempts involved increasing police presence. Even for program designs for which accident reductions have been convincingly demonstrated, the drawback has been that the increased police presence required tends to be difficult to sustain, especially in an environment of public expenditure restraint. For this reason, any increase in police presence tends to be only short term (a ‘blitz’) and any accident reduction achieved is not maintained25,26. Unsustainable presence of police activity might reflect an inadequate number of police officers, which are required to serve a wide range of police services including traffic policing. For police services in general, the UN recommended population to police ratio is 450:127. Based on this figure, the ratio in Thailand as a whole, (541.3:1 as shown in Table1) could be considered inadequate. Nevertheless, our finding did not support the association between helmet use and population to police ratio (r=0.1886, p = 0.1028). We found police density as defined by the number of police per square kilometer to be independently and significantly associated with the percentage of motorcycle helmet use after controlling for the confounding effect of adult literacy. The association between police density and helmet use could be interpreted as an association between spatial and temporal coverage of police surveillance activity and helmet use. This interpretation was consistent with findings from experimental studies linking speed reduction with deployment of single stationary police vehicles at random times to randomly chosen sectors of specific rural highway sections28. The result of the experiment was an average speed reduction spread of effect per patrol of 22 km/hr which was some four times greater than those demonstrated by Smith (1962, cited in Edwards and Bracket, 197828). Further trials, with a similar approach, in Queensland Australia, demonstrated the average crash reduction for fatal crashes was 31%21. With an average programwide fatal crash coverage of 51%, the program reduced the aggregate of the seven regions’ fatal crashes by some 16%, and that of the entire state of Queensland by 15%. For a macro policy decision, the linkage between police density and motorcycle helmet use implies that allocation of the police force should take into account deployment strategies to ensure spatial and temporal coverage of police services. In contrast, seeking to meet the UN’s recommended ratio could hardly be achieved, especially for resource poor countries. Ideally, distribution of police should be guided by a need to ensure an adequate level of law enforcement on helmet use. With this argument, provinces with a low helmet use rate would require higher numbers of police in terms of police density. In fact, the opposite situation seemed to be the case. There was a huge disparity in police density between the highest figure (13.09 per square kilometer) in Bangkok and the lowest (0.19 per square kilometer) in the central region with a ratio of 68.9:1(Table 1). This ratio was much larger than that (2.1:1) between the highest helmet use rate of 93.0% in Bangkok and the lowest of 44.8% in the northern region. Since police density is positively linked to population density and GPP (Table 5), it is understandable that the existing allocation of the police force in Thailand has been biased toward the rich urban communities. In addition to the highest percentage of helmet use, the lowest case-fatality rate (1.8%) of road traffic injuries has been documented in Bangkok since 19953. These facts have strongly indicated a need for more equitable Indian Journal of Public Health Research & Development. January-March 2013, Vol. 4, No. 1 205 distribution of the police force in Thailand using an evidence-based approach. Based on our findings, the level of motorcycle helmet use (93%) in Bangkok might be adopted as a protective threshold for head injuries to guide better distribution of the police force, so that sufficient coverage of law enforcement on helmet use could be met throughout the country. This does not mean that the number of police should be increased in regional provinces to the density (13.09 per square kilometer) in Bangkok, since the magnitude of difference in percentage of helmet use between Bangkok and regional provinces, is a lot smaller than that in police density between the two groups as discussed. Apart from considering the redistribution of the police force, there is a need to consider investment in police training to make use of state-of-the-art techniques in effective deployment of traffic policing activity as demonstrated in Queensland, Australia21. A unique feature of the Queensland approach was the deployment of a small number of police force randomly scheduled to randomly chosen road sectors of specific rural highway sections to provide widespread coverage resulting in long term (up to 3 years) reduction of road crashes of all severities. Hence, it is an interesting and promising approach to maximize the capacity of the existing limited numbers of police. There are two major limitations of this report worth consideration. The first was the cross sectional design of the study could not support a conclusion of causal linkage between motorcycle helmet use or police density and the predictors. The other was that the relationship between the predictors and helmet use or police density might be confounded by other known confounding factors not included in this report. Nonetheless, this report provides a unique finding on the relationship between police density and helmet use from the results of a nationwide direct observation of the behavior of motorcycle helmet usage and of the existing national datasets of potential predictors. The findings might be useful to guide a policy decision on the allocation of police force to meet the need for better enforcement of the helmet law and to address the equity concerns with resource allocation. Further studies are needed to better ascertain the relationship between helmet use and police density making use of a longitudinal study design with the inclusion of other relevant confounding factors. CONCLUSIONS Using cubic spline regression analysis on national datasets, and the nationwide roadside survey of motorcycle helmet use in Thailand, this report has demonstrated that police density is an independent predictor of helmet use. Concerning policy decisions on resource allocation, the finding provides a new parameter for the allocation of limited numbers of police. Police density was also significantly associated with gross provincial product (GPP) per capita and population density. This additional important finding raises a concern about the equitable allocation of the police force in Thailand. ACKNOWLEDGEMENTS The authors are grateful to Stephen Pinder for correction of the grammar. Source of Funding Faculty of Medicine,Ramathibodi Hospital, Mahidol University. Conflicts of Interest The authors declared no conflicts of interest. ETHICAL CLEARANCE This report made use of secondary datasets hence there is no need for ethical clearance. Table 1 Percentage (range*) of helmet use of motorcycle drivers and passengers and related factors by region in 2010 Region(survey samples) % helmet Population Population to GPP per Police density use density police ratio capita** (per sq km) driver passenger (per sq km) whole country (954,956) 53.30 19.30 131.2 541.3 150,118 0.24 Bangkok (27,647) 93.00 45.20 1330.4 562.9 361,243 13.09 Central (289,654) 63.90(28.6-79.9) 24.40(6.2-31.0) 63.5 336.2 264,285 0.19 Northeast (247,821) 47.60(31.7-74.1) 19.80(9.5-56.6) 135.5 654.0 49,092 0.21 North (150,888) 44.80(33.2-61.4) 17.20(5.8-32.4) 129.9 427.7 79,158 0.30 South (238,946) 47.00(21.5-79.7) 9.40(1.8-50.6) 132.6 344.2 104,738 0.39 * figures of province with the lowest and the highest percentage in each region ** at current market prices in 2010 206 Indian Journal of Public Health Research & Development. January-March 2013, Vol. 4, No. 1 Table 2. Pearson correlation matrix for factors associated with motorcycle helmet use (p value) Helmet use No. of police Police Population to Population GPP Adult literacy density police ratio density Seat belt 1.0000 No. of police 0.3613 1.0000 (0.0013) Police density 0.5121 0.9005 1.0000 (0.0000) (0.0000) Population to police ratio -0.1886 0.1569 0.1729 1.0000 (0.1028) (0.1758) (0.1353) Population density 0.5864 0.7976 0.9515 0.0371 1.0000 (0.0000) (0.0000) (0.0000) (0.7502) GPP 0.4944 0.8802 0.9074 0.0404 0.8603 1.0000 (0.0000) (0.0000) (0.0000) (0.7287) (0.0000) Adult literacy 0.1654 0.0656 0.1001 -0.2887 0.1478 0.1288 1.0000 (0.1533) (0.5733) (0.3894) (0.0114) (0.2028) (0.2674) Table 3. Factors associated with percentage of helmet use according to spline regression analysis. Predictors coef SE t P>|t| [95% CI] Police density* 6.17 1.21 5.12 <0.001 3.77 8.58 (Police density-0.253)3 2.36 1.21 1.95 0.055 -0.05 4.77 Adult literacy 1.12 1.22 0.92 0.362 -1.31 3.54 *number of policemen/area, CI=Confidence Interval, SE= Standard Error Table 4. Factors associated with police density: A spline regression Predictors coef SE t P>|t| [95% CI] Population density 0.28 0.08 3.560 0.001 0.12 0.43 (Population density- 164.7)3 0.38 0.05 7.297 0.000 0.28 0.48 GPP 1.23 0.08 15.194 0.000 1.07 1.39 (GPP- 8.7x104)3 -0.67 0.05 -14.051 0.000 -0.76 -0.57 (GPP- 5.6x104)3 0.05 0.02 2.777 0.007 0.01 0.09 CI = Confidence Interval, SE=Standard Error (Endnotes) REFERENCES 1. World Health Organisation. The global burden of disease: 2004 update.2008. 2. Bundhamcharoen, K., Odton, P., Phulkerd, S., Tangcharoensathien, V. Burden of disease in Thailand: changes in health gap between 1999 and 2004. BMC Public Health.2011: 53-61. 3. Suriyawongpaisal, P., Kanchanasut, S. Road traffic injuries in Thailand: trends, selected underlying determinants and status of intervention. Injury Control and Safety Promotion.2003;10(1-2): 95-104. 4. Le L., Pham, C., Linnan, M., Dung, V.D., Phuong, N.L., Hanh, H.L. Vietnam profile on traffic related injury: facts and figures from recent studies and their implication for road traffic injury policy. In: Road traffic injury and health equity conference, 10-12April2002, Cambridge, MA.2002. 5. Taneerananon, P. The cost of road accidents in Thailand. In: Technology and Innovation for Sustainable Development Conference (TISD2008), KhonKaen University, KhonKaen. 2008. 6. Lui, B., Ivers, R., Norton, R., Blows, S., Lo, S.K. Seat belts for preventing injury in motorcycle riders. Cochrane Database Syst Rev. 2004. 7. Bishai, D.M., Hyder, A.A. Modeling the cost effectiveness of injury interventions in lower and middle income countries: opportunities and challenges. Cost Effectiveness and Resource Allocation. 2006; 4(2): 1-11. 8. Hung, D.V., Stevenson, M.R., Ivers, R.Q.Prevalence of helmet use among motorcycle riders in Vietnam. Injury Prevention.2006;12(6): 409-413. 9. Kulanthayan, S., Umar, RS.,Hariza, H.A., Nasir, M.T., Harwant, S. Compliance of proper safety helmet usage in motorcyclists. Med. J. Malaysia. 2000; 55(1): 40-44. Indian Journal of Public Health Research & Development. January-March 2013, Vol. 4, No. 1 207 10. Pitaktong, U., Manopaiboon, C., Kilmarx, P.H., et al. Motorcycle helmet use and related risk behaviors among adolescents and young adults in Northern Thailand. Southeast Asian J. Trop. Med. Public Health.2004;35(1): 232-241. 11. Hongsranagon, P., Khompratya, T., Hongpukdee, S.,et al. Traffic risk behavior and perceptions of Thai motorcyclists: A case study. IATSS Research.2011; 35(1): 30-33. 12. Li, L.P., Li. G.L., Cai, Q.E., et al. Improper motorcycle helmet use in provincial areas of a developing country. Accident Analysis & Prevention. 2008; 40(6): 1937-1942. 13. Sirathranont, J.K., Kasantikul, V. Mortality and injury from motorcycle collisions in Phetchaburi Province. J. Med. Assoc. Thai.2003; 86(2): 97-102. 14. Liberatti, C.L., de Andrade, S.M., Soares, D.A., et al. Helmet use by motorcyclists injured in traffic accidents in Londrina, southern Brazil.Pan American Journal of Public Health.2003;13(1): 33-38. 15. World Health Organization, 2006. Helmets: A Road Safety Manual for Decision-Makers And Practitioners, Geneva, Switzerland. 16. Singweratham, N., Podang, J.Direct cost and net costs of road traffic enforcement in Thailand. In: Proceedings of the 13th Biennial SMDM European Meeting, 2010 May 30-Jun 2, Hall/Innsbruck, Austria. 2010. 17. ThaiRoads Foundation. Helmet use rate of motorcyclists in Thailand B.E.2553. Bangkok.2011. 18. Wegman F., Commandeur J.,Doveh E., et al. SUNflowerNext: Towards a composite road safety performance index. SWOV Institute for Road Safety Research, the Netherlands. 2008. 19. Social and quality of life database system. Estimated annual number of population by sex and province for the year 2000-2020. Office of the National Economic and Social Development Board.http://social.nesdb.go.th/SocialStat/ StatReport_Final.aspx?reportid=180 & template = 2R2C&yeartype=M&subcatid=32 20. Royston, P. Choice of scale for cubic smoothing spline models in medical applications. Stat Med.2000; 19(9): 1191-1205. 21. Newstead, S.V., Cameron, M.H., Leggett, M.W. The crash reduction effectiveness of a network wide traffic police deployment system. Accident Analysis & Prevention.2001; 33(3): 393-406. 22. Armour, M. A review of the literature on police traffic law enforcement. Australian Road. 1984 14(1): 17-24. 23. Shinar, D., McKnight, J.A. The effects of enforcement and public information on compliance. In: L Evans, R.C Schwing (Eds.), Human Behaviour and Traffic Safety. Plenum, New York. 1985 24. Zaal, D. Traffic law enforcement: a review of the literature. Report Number 53. Monash University Accident Research Centre, Melbourne, Australia. 1994 25. Ross, H.L. Deterring the drink driver, Lexington Books. Lexington, MA.1981 26. Bjornskau, T., Elvik, R. Can road traffic law enforcement permanently reduce the number of accidents? Accident Analysis and Prevention.1992;24(5): 507-520. 27. Mwenda, A.A review of the Kenya police force budget, and its effect on crime management. Commonwealth Human Rights Initative. 2005. http://www.humanrightsinitiative.org/publications/ police/kenya_budgetary_report. pdf. Accessed July 15, 2012. 28. Edwards, M.L., Brackett, R.Q. The management of speed. Traffic Safety.1978; October, 18-30.