Indian Journal of Public Health

: 2018  |  Volume : 62  |  Issue : 4  |  Page : 265--270

Risk behaviors contributing to recent serious unintentional injuries among school-going adolescent boys in Kolkata: Application of zero-inflated count model

Arup Chakraborty1, Arista Lahiri2,  
1 Assistant Professor, Department of Community Medicine, Medical College and Hospital, Kolkata, West Bengal, India
2 Post-Graduate Trainee, Department of Community Medicine, Medical College and Hospital, Kolkata, West Bengal, India

Correspondence Address:
Dr. Arista Lahiri
4th Floor, Department of Community Medicine, Medical College and Hospital, MCH Building (Near Gate No. 2), 88, College Street, Kolkata - 700 073, West Bengal


Background: Unintentional injuries have become a major noncommunicable disease burden, especially among the adolescents. Objective: The current study was conducted to estimate the effect of different aspects of daily activities of adolescence for sustaining serious unintentional injuries in the past 1 year. Methods: A cross-sectional survey with multistage sampling with validated pretested questionnaire was done among the school-going adolescent boys in Kolkata. Poisson regression was used to model the counts of serious injuries. To account for the excess of zero in the outcome, zero-inflated Poisson regression was performed. Results: Among the participants, 73.5% did not report any serious unintentional injury sustained in the past 1 year, 11.9% reported to have sustained serious unintentional injury once in the past 1 year, and rest had more than one count. Statistically significant higher chance of sustaining an episode of injury was found among frequent users of motorbike (incidence rate ratio [IRR]: 1.183), frequently walking on roads (IRR: 1.910), and frequently crossing major roads on bicycle (IRR: 2.181) were observed. A statistically significant protective rate ratio was also obtained for those frequently obeying traffic signals while crossing roads (IRR: 0.493) and frequent users of bicycles (IRR: 0.384). Significantly lower rate ratio for sustaining a serious injury was observed with frequently getting into fight at home (IRR: 0.343) and getting beaten up at school (IRR: 0.595). Conclusions: The study revealed traveling in a car and obeying traffic rules were protective from sustaining serious injury. However, walking and participation in sports appeared to be risky, especially for sustaining another episode of serious injury.

How to cite this article:
Chakraborty A, Lahiri A. Risk behaviors contributing to recent serious unintentional injuries among school-going adolescent boys in Kolkata: Application of zero-inflated count model.Indian J Public Health 2018;62:265-270

How to cite this URL:
Chakraborty A, Lahiri A. Risk behaviors contributing to recent serious unintentional injuries among school-going adolescent boys in Kolkata: Application of zero-inflated count model. Indian J Public Health [serial online] 2018 [cited 2019 Jan 16 ];62:265-270
Available from:

Full Text


With regard to attaining a society at optimum health, adolescent age group remains perhaps the most important, contributing 35% of disease burden globally.[1] Unintentional injuries are large contributors in adolescent death and disability.[2] Globally, adolescent deaths due to unintentional injuries are comparable to deaths due to various infectious causes in terms of the burden, with the majority occurring in low- and middle-income countries.[3] Deaths being merely the tip of the iceberg, several morbidity and disability associated constitute the grave scenario surrounding sustaining severe unintentional injuries among the young population.[4],[5]

In the Indian perspective of demographic and epidemiologic transition, injuries become cardinal to the noncommunicable disease burden, especially among the adolescent population. Although a global push in research related to unintentional injuries among adolescents, studies regarding the same are relatively rare in the national context. However, the Indian studies demonstrate a higher incidence of unintentional injuries among adolescent with road traffic injuries being a leading cause.[6],[7],[8],[9],[10],[11] Interestingly, there are several studies regarding sports-related injuries among the school and college students among the Indian population.[12],[13],[14],[15]

Unintentional nonfatal injuries leading to hospitalization, disability, or significant difficulty in daily activity are usually considered serious in nature.[16] It can be conceptualized that this particular subset of injuries is cardinal in causing a higher burden of morbidity. Common daily activities such as using protection such as helmet, seatbelt, or abiding by the traffic rules while crossing roads or even carefulness during daily sports activities can be useful in addressing the issue of adolescent injuries.[2],[3],[10],[12],[17] The current study was conducted to estimate the effect of different aspects of daily activities of adolescence for sustaining serious unintentional injuries in the past 1 year.

 Materials and Methods

Study type and population

With proper permissions from the respective authorities, a descriptive, cross-sectional survey was carried out among the school-going adolescent boys studying at Government higher secondary boys' schools in Kolkata. Students pursuing higher secondary level education (Classes IX and X) were included who gave consent for participation in the study. Students aged lesser than 14 years and more than 19 years and who gave partial or nonresponse to the administered survey questionnaire was excluded from the study. Students of residential schools were also excluded. The data collection for the survey was carried out during January–August 2017, after obtaining approval from the Institutional Ethics Committee of Medical College and Hospital, Kolkata. Consent from the head of the schools was taken, and informed consent of the parents obtained during parents–teachers meeting, and no participant was included without consent.

Method of data collection

Multistage sampling was used with schools being sampled at first, and finally, students were surveyed from the selected section(s) of each class. The scheme of sampling is shown in [Figure 1]. Partial and nonresponse rate was considered acceptable if lesser than 10%. For calculation of optimum sample size, a prevalence of 50% was taken, with 5% precision and absolute error of 10%. Now, a design effect of 2 was applied (for multistage design), which yielded a minimum required sample size of 192.{Figure 1}

A validated and pretested survey questionnaire was administered in a single section to all the students present in a day. All the completed responses were considered for analysis after meeting the minimum sample size required (calculated by probability proportional to size) in each section. Data collection from other class (sections) in a school was done at a gap of minimum 6 weeks, allowing recall period to be over. Finally, a total of 219 completed questionnaires were included in the study.

Construct of the bilingual (Bengali and English simultaneously) questionnaire was developed based on the several validated questionnaires such as Global School Health Survey questionnaire Indian version[18] and Bangladesh version,[19] and World Health Organization steps Violence and Injury questionnaire module.[20] Statistical techniques were used in validating the questionnaire (internal consistency reliability: Cronbach's alpha value 0.61), but that is beyond the scope of the present article.

Variables in the study

Data on age, religion, area of residence, type of family, and number of family members of the respondents were collected to set up baseline information.

Serious unintentional injuries[2],[16] were defined for the purpose of this study as presenting to physician with complaints pertaining to the injury sustained without “intentionality” and/or hospitalization for more than 48 h for the event and/or residual disability and/or missing school for 1 week or more immediately after sustaining an injury sustained without “intentionality.” Count or number of times a participant sustained such injury in the last 1 year was considered the outcome of the several risk behaviors (factors) taken as predictor variables.

Such episodes suffered due to causes related to sports were predicted with the help of location of outdoor sports, daily duration of outdoor sports, and involvement in fight. Modeling injuries in relation automobile use were done with traveling on a motorbike and car, use of helmet and seatbelt. Attributes related to traffic rules were not included since it was presumed that a participant is likely to travel on an automobile rather than driving it. Walking and cycling habits were measured in terms of duration per day, average number of times road crossed and maintaining traffic signals. Usual daily time spent on outdoor sports, involvement in fights and frequency of getting beaten up at home, in school, and outside both school and home were among the remaining predictors of the count of serious injury episode(s).

Methods for statistical analysis

Data were compiled and analyzed with the help of EpiInfo 7 (Centers for Disease Control, Atlanta, USA) and STATA 14.2 (StataCorp LLC, College station, Texas, USA). Having the outcome as count outcome, count data models were used.[21] Separate multivariate models were built for automobile use, walking and using bicycle, sports, and getting into fight.

Likelihood ratio (LR) test was used to understand goodness-of-fit for negative binomial model over Poisson regression. Since in consonance with the result of LR test, Poisson model was chosen.[21],[22] Now, to account for the excess of zero count in the outcome (participants not sustaining injury meant zero count of injury) zero-inflated count regression analyses were performed.[23] Since from the nature of the dependent variables, it was evident that the inflation of zeroes were contributed by both sampling (i.e., not sustaining injury due to the effect and distribution of predictors) and structural parts (i.e., not sustaining injury due to any unobservable factor[s]), a zero-inflated Poisson model was chosen over hurdle-at-zero model.[22],[24] Vuong's test[25] was used to objectively test for appropriateness of zero-inflated Poisson model over Poisson model. Robust standard errors were used in the models with presumed level of two-tailed statistical significance at 5%. The estimates of risk for the factors contributing to the outcome were reported in terms of incidence rate ratios and its 95% confidence interval (CI) obtained from the count regression analyses taking the predictor variables in a dichotomous form. Predictor variables were dichotomized depending on their median response. The items were set with a Likert-type response options. For example, let's say the use of motorbike had responses “always,” “often,” “sometimes,” “rarely,” and “Never.” Say “sometimes” came out to be the median response then from “never” to “sometimes” it is categorized “infrequent,” above that meaning “often” and “always” was considered “frequent.” In this way, responses above the median were considered frequent.


Sociodemographic background

Of the 219 adolescent boys, who participated in the study, majority were Hindu (92.7%), residing in the corporation area (63.9%) which signifies subjects belonged from urban area and coming from a nuclear family background (76.7%). The mean age was 15.05 years (standard deviation 0.725 years) with minimum and maximum age being 14 and 17 years, respectively. Majority (73.9%) of the participants were aged below the mean age.

Injury in last 1 year

[Figure 2] demonstrates the number of episodes of serious unintentional injury reported by the respondents. Among the participants, who properly responded to the study questionnaire, 73.5% did not report any serious unintentional injury sustained in the last 1 year, and 11.9% reported to have sustained serious unintentional injury once in last 1 year. The highest counts of serious injury sustained by a participant were found to be 5 and were reported by 0.5% of the respondents.{Figure 2}

Factors related to sustaining injury

The relationship between sustaining serious accidental injury and behaviors related to automobile use is presented in Model (A) of [Table 1]. Frequent use of motorbike had a greater chance (1.183, 95% CI: 0.441–3.175) of sustaining an episode of serious unintentional injury, but infrequent use of a helmet while riding had a lesser chance (0.765, 95% CI: 0.232–2.528). Both of them did not show any statistical significance. Statistically significant protective rate ratio of sustaining serious injury was observed among frequent users of car (0.498, 95% CI: 0.290–0.855). Although not statistically significant, infrequent use of seatbelt had a rate ratio of 0.932 (95% CI: 0.436–1.996). The zero-inflation model showed frequent use of motorbike had a rate ratio of 13.701 (95% CI: 1.008–186.262) of not sustaining any serious injury in the last 1 year. This means motorbike users had less chance of getting injured overall, but if seriously injured once, then the chance of subsequent injury is high. Use of helmet, use of car and seatbelt did not have any statistical significance in the zero-inflated model. Infrequent use of helmet and frequent use of car was associated with a lesser chance (0.650 [95% CI: 0.087–4.870] and 0.469 [95% CI: 0.137–1.602], respectively) of not sustaining injury. Interestingly, infrequent use of seatbelt had a rate ratio of 1.195 (95% CI: 0.259–5.519) of not sustaining injury.{Table 1}

The count model in Model (B) of [Table 1] showed statistically significant higher chance of sustaining an episode of injury among those frequently walking on roads (1.910, 95% CI: 1.166–3.128). Frequently crossing major roads on foot and crossing through zebra lines had a higher risk of sustaining injury but was not significant statistically. A statistically significant protective rate ratio was obtained for those frequently obeying traffic signal while crossing roads (0.493, 95% CI: 0.258–0.941) and frequent users of bicycle (0.384, 95% CI: 0.247–0.598). The results can be depicted in a comparative manner with walking on roads being riskier in having subsequent injuries compared to cycling. Those frequently crossing major roads on bicycle had a statistically significant increase in risk of sustaining injury (2.181, 95% CI: 1.204–3.954), while increased risk for frequently obeying traffic signal while cycling was not significant statistically. Decrease in the chance of not sustaining any injury among those frequently walking on roads, obeying traffic signals while walking and riding a bicycle, crossing major roads on bicycle did not attain statistical significance in the zero-inflated model, neither did the higher rate ratio among those crossing major roads on foot. Frequently crossing roads through zebra lines was associated with a significantly high rate ratio of having not sustained any unintentional injury in last 1 year (26.849, 95% CI: 2.030–355.027). Frequently riding a bicycle on major roads had a low chance of not sustaining injury (0.036, 95% CI: 0.001–0.877).

Practices related to sports among the participants and the relationship with serious unintentional injury have been depicted in the zero-inflated Poisson model in Model (C) of [Table 1]. Frequent participation in sports, playing on road(s), and fighting on issues related to sports had a higher rate ratio (1.244 [95% CI: 0.702–2.203], 1.440 [95% CI: 0.664–3.123], and 1.211[95% CI: 0.633–2.317], respectively) of sustaining a serious unintentional injury. However, a higher maximum duration of daily outdoor sports had a protective effect (observed rate ratio: 0.839). None of these were found to be statistically significant in the count part of the ZIP model. The zero-inflated part of the model yielded a statistically significant lower rate ratio of not sustaining an injury among those frequently participating in sports (0.172, 95% CI: 0.047–0.632). Higher maximum duration of sports activity in a day had lesser chance (observed rate ratio: 0.495, 95% CI: 0.191–1.280) of not sustaining any injury, although not significant statistically. Frequently playing on roads and fighting over sports-related issues had a higher chance of not sustaining injury (1.036 [95% CI: 0.329–3.259] and 1.739 [95% CI: 0.635–4.763], respectively), again not statistically significant.


It was observed that the zero-inflation model sometimes demonstrated the opposite effect for some variables in count and zero-inflation part. It should be kept in mind that while the count part reveals the effect of the variable on the count of the outcome, the zero-inflated part typically like a logistic model predicts the effect of the variable on whether the outcome will yield a count of zero or otherwise. Therefore, even though the effect may appear opposite in some cases, the interpretation is otherwise. While zero-inflated part tells us about the chance of not sustaining serious injury (injury count = 0), the count part reflects the risk of sustaining injury.

In the first model, among the variables, only traveling in a car was found to be protective. Riding in a motorbike had a higher chance of not getting injured in the zero-inflation model. This may be due to better safety practices of those riding on a bike. However, in a study by Shrestha et al. in Nepal, the occurrence of serious injury was comparable for the use of two- and four-wheeled vehicles.[26] As presumed conceptually frequent walking on the road had a higher chance of sustaining injury. Obeying safety measures such as using zebra lines was protective as evident from count and zero-inflated analysis. Similarly, the use of bicycle was also protective but not on zero-inflated part. It was conceptually interpreted as the use of bicycle will decrease the chances of not sustaining serious injury, but the chances of subsequent injury after suffering at least an episode will be lowered. Tetali et al. in their cross-sectional study on students of Hyderabad showed that cycling was riskier compared to walking.[10] This is comparable to the current study, especially with respect to the zero-inflation model results. In the sports-related model, it was evident statistically that participation in sports increased the chance of getting injured seriously, which was in consonance with the conceptual framework. In their study conducted in Bengaluru, Shekhar et al. also identified that injury is common with sports, but seriousness was not considered. Still, they reported fight during sports to be an important factor.[27] Sunitha and Gururajalso recognized the burden of sports-related injuries among adolescents.[6]


The study supported the generalized fact that simple interventions such as safety habits during walking, cycling, and traveling on an automobile among the school-going adolescents are of utmost importance for injury prevention. Behavioral training, anger management among adolescents, also appears to be important in view of the results. The importance thus derived is the strengthening of the conceptual background in decreasing burden of serious injury among the school-going adolescent population at the expense of efficient yet simple training and behavior-change practices. It is well understood that the practices reported by the adolescents may have been biased, the risk of which is always there for any self-report questionnaire. The recent practices were taken so that effective intervention areas could be suggested. Unlike other studies[6],[7],[8],[9],[10],[11] done among those who sustained serious injuries, the current study was conducted among the presumed at-risk population who might or might not have sustained serious unintentional injury, and the behavior and practices were modeled. However, the incidence of serious injury in past year was not classified according to types such as road traffic injury or sports injury as such since the objective was to study the simple behaviors and practices and relating them to sustaining episodes of serious unintentional injury. Further studies can be undertaken to study the relationship of these variables with sustaining serious unintentional injury and the different types. A study with larger sample size can help in building a model with all the variables together in contrast to compartmentalized models produced in the current study. This may indeed help in a better understanding of the behavior and practice-related factors.


The authors would like to acknowledge the participants in the study and the staffs and faculties of the institutions.

Financial support and sponsorship

This study is self-funded by the authors.

Conflicts of interest

There are no conflicts of interest.


1World Health Organization. Adolescent Health Epidemiology WHO. Available from: [Last accessed on 2017 Nov 11].
2World Health Organization. Injuries WHO. Available from: [Last accessed on 2017 Nov 11].
3World Health Organization. Injury Prevention and the Attainment of Child and Adolescent Health WHO. Available from: [Last accessed on 2017 Nov 11].
4Chandran A, Hyder AA, Peek-Asa C. The global burden of unintentional injuries and an agenda for progress. Epidemiol Rev 2010;32:110-20.
5Peden M, Oyegbite K, Ozanne-Smith J, Hyder A, Branche C, Rahman A, et al. World report on child injury prevention. Geneva: World Health Organization; 2008.
6Sunitha S, Gururaj G. Health behaviours & problems among young people in India: Cause for concern & call for action. Indian J Med Res 2014;140:185-208.
7Jagnoor J, Suraweera W, Keay L, Ivers RQ, Thakur J, Jha P. Unintentional injury mortality in India, 2005: Nationally representative mortality survey of 1.1 million homes. BMC Public Health 2012;12:487.
8Sahu A, Satapathy D, Tripathy R. Epidemiological study of road traffic accident cases: A study from South Odisha. Int J Interdiscip Multidiscip Stud 2014;1:202-9. Available from: [Last accessed on 2017 Dec 01].
9Aeron-Thomas A, Jacobs G, Sexton B, Gururaj G, Rahman F. The Involvement and Impact of Road Crashes on the Poor: Bangladesh and India Case Studies. Available from: [Last accessed on 2017 Dec 08].
10Tetali S, Edwards P, Murthy GV, Roberts I. Road traffic injuries to children during the school commute in Hyderabad, India: Cross-sectional survey. Inj Prev 2016;22:171-5.
11Epidemiology of Road Traffic Accident Deaths in Children in Chandigarh Zone of North West India-Science Direct. Available from: [Last accessed on 2017 Nov 11].
12Dorje C, Gupta RK, Goyal S, Jindal N, Kumar V, Masih GD, et al. Sports injury pattern in school going children in union territory of Chandigarh. J Clin Orthop Trauma 2014;5:227-32.
13Singh G, Garg S, Damle SG, Dhindsa A, Kaur A, Singla S, et al. A study of sports related occurrence of traumatic orodental injuries and associated risk factors in high school students in North India. Asian J Sports Med 2014;5:e22766.
14Kumar V, Mangal A, Yadav G, Raut DK, Singh S. Prevalence and pattern of sport injuries among college students in Delhi, India. Saudi J Sports Med 2014;14:109.
15Department of Community Medicine, Rajah Muthaiah Medical College, Annamalai University, Chidambaram, Tamil Nadu, India, N.C I. Prevalence and pattern of sports injuries among the university students of physical education, Southern India. J Med Sci Clin Res 2016;4:13434-40.
16Adolescents' Health-Related Behaviours – Violence-injuries. Available from: [Last accessed on 2018 Apr 13].
17Norton R, Hyder AA, Bishai D, Peden M. Unintentional injuries. In: Jamison DT, Breman JG, Measham AR, Alleyne G, Claeson M, Evans DB, et al. editors. Disease Control Priorities in Developing Countries. 2nd ed. Washington (DC): World Bank; 2006. Available from: [Last accessed on 2018 Apr 13].
182006 India, Central Board of Secondary Education (CBSE) GSHS Questionnaire. Available from: [Last accessed on 2017 Sep 03].
19World Health Organization. Global School-Based Student Health Survey (GSHS). WHO. Available from: [Last accessed on 2017 Nov 11].
20World Health Organization. STEPS Optional Modules. WHO. Available from: [Last accessed on 2017 Nov 11].
21Cameron A, Trivedi P. Regression Analysis of Count Data. Cambridge: University Press; 1998.
22Mullahy J. Specification and testing of some modified count data models. J Econ 1986;33:341-65.
23Lambert D. Zero-inflated Poisson regression, with an application to defects in manufacturing. Technometrics 1992;34:1-14.
24Hu MC, Pavlicova M, Nunes EV. Zero-inflated and hurdle models of count data with extra zeros: Examples from an HIV-risk reduction intervention trial. Am J Drug Alcohol Abuse 2011;37:367-75.
25Vuong QH. Likelihood ratio tests for model selection and non-nested hypotheses. Econometrica 1989;57:307-33.
26Shrestha VL, Bhatta DN, Shrestha KM, Gc KB, Paudel S. Factors and pattern of injuries associated with road traffic accidents in Hilly district of Nepal. J Biosci Med 2017;5:88-100.
27Shekhar M, Suresh S, Gautam A, Mahalaxmi V, Chakarborty R, Purbey S. Knowledge, attitudes, and behavior of physical education teachers in Bengaluru, India, regarding sports-related orofacial injury and its prevention. Kumar A, editor. Int J Oral Care Res 2016;4:196-200.