This multi-centric community-based cross-sectional study was conducted at 11 different sites across eight states in India between March, 2018 and September, 2020. A purposive selection of study sites was made, ensuring adequate geographical representation. The study population comprised children aged six months to <18 years from both rural and urban areas viz. Siddlaghatta, Bangalore, Karnataka; Pauri Garhwal, Srinagar, Uttarakhand; Vellore, Tamil Nadu; Perundurai, Erode, Tamil Nadu; Mangalore, Karnataka; Bardhman Sadar North, Bardhman, West Bengal; Dhoraji, Rajkot Gujarat; East Sikkim, Sikkim; Dholka, Ahmedabad, Gujarat; East Ludhiana, Ludhiana, Punjab; Huzur, Bhopal, Madhya Pradesh (Fig. 1).

Fig. 1 Map of India with location of participating institutions and study sites.

Sample size and sampling strategy: The sample size was calculated considering the overall prevalence of childhood injuries to be 11.0% (including minor injuries) as per the guidelines for conducting community surveys on injuries and violence by WHO [4], with a relative precision of 13% and 95% desired confidence level, with increase in sample size by 10% to allow for non-responses, design effect of 2 to account for cluster sampling and relative precision of 13%. Hence, 2341 households from rural and urban areas (combined) were selected from each site proportionately based on the population’s rural-urban distribution as per 2011 census [5]. In each site, the district predominantly served by the participating institution was selected. Subsequently, one taluk was selected through a simple random sampling technique. Applying probability proportionate to size (PPS) sampling, within each taluk, clusters of households in rural areas and urban areas were selected. Each cluster consisted of 16 houses (estimated based on the number of households to be covered by four field workers in a day) both in urban and rural areas.

Further, within the rural areas, all villages in the selected taluk were in the sampling frame. In each village, each cluster consisted of 16 households. The number of clusters to be surveyed to meet the required sample was arrived at.

In the case of urban areas, one town was selected using a simple random sampling technique from the total number of towns in the taluk. Like rural areas, 16 households made up one cluster. The required number of clusters were selected from a randomly chosen locality in the town.

The inclusion criterion was to have all six months to <18-year permanent resident children, and visiting children living in that area for a minimum duration of previous six months. Additionally, information pertaining to deaths in the above age group due to unintentional injuries was sought from the government sources. Birth injuries or injuries consequent to intra-natal complications, and disabilities due to other conditions were excluded.

After clearance from the institutional ethics committee (IEC), permissions were obtained from the district and taluk administrative authorities at all sites, the zilla panchayat CEO, and the Child Development Project Officer for accessing information of village panchayats and community health workers.

Data collection: Each site recruited four field workers (medico-social workers) who underwent standardized training of 5 days, through a workshop conducted at each site. The workshop oriented the field investigators to undertake an initial pilot study about child injuries, the tool, rapport building in the community, data collection procedures and hands-on instructions on using the electronic handheld device.

Subsequently, field investigators prepared a spot map of each selected village (rural area) or locality (urban area) and numbered the houses serially. After explaining the study’s purpose in the local language, informed written consent was obtained from the respondent (parent/primary caregiver). Verbal assent was taken from children more than or equal to 7 years of age. In addition to interviewing the adult, children 7 years and above were also interviewed to substantiate parents’ information. A history of previous three months was used to collect information relating to all injuries whereas information about fatal injuries was collected for the previous 12 months and verified in the register maintained in the panchayat office. At the end of every interview, the caregivers were verbally provided education regarding injury prevention among children.

Tool employed for data collection: The World Health Organization (WHO) child injury questionnaire adapted to the Indian settings was used [4]. Definitions for types (road traffic accidents, falls, burns, poisoning, drowning, animal-related injuries) and severity of injuries (mild, moderate and severe) was adapted from the WHO standard definitions [6]. Injuries other than these types were categorized as miscellaneous. A cloud-based soft-ware was developed through an external vendor and validated. Each center received four handheld devices for the field workers and one for the supervisor. Quality assurance of data was built into the software with features of valid entries, skip logic and consistency checks.

A robust management information system and dashboard was developed through which each site could visualize their respective electronic data on the web and download a copy of their data. Data were analyzed centrally based on the approved statistical analysis plan.

Quality assurance: An operation manual and a training manual for field workers were prepared by the coordinating team and shared with the participating sites to ensure uniformity. Online monthly meetings were organized by the central and national coordinating sites with other site investigators and field workers for centralized monitoring and supportive guidance to ensure regular interaction and quality of data collected. The coordinating center continuously monitored data through the dashboard. A team from the coordinating center and ICMR team visited each of the sites during data collection for supportive supervision. The investigators in each site revisited 5% of households randomly and collected information independently to check data quality and discrepancies if any, were resolved.

Statistical analyses: Data from all sites were analyzed by the coordinating team using Statistical Package for the Social Sciences 16.0 (SPSS Inc.). Data were coded according to severity of injuries. Minor injuries were excluded for subsequent estimates. The association between factors such as prevalence rate with age, gender, and other factors was tested for statistical significance by Chi-square test or Fisher’s exact test. The difference in mean values between two groups was tested for statistical significance by Student’s t-test. Probability value <0.05 was considered as cutoff for statistical significance. Prevalence rates (period prevalence for 3 months) with 95% confidence intervals were estimated.

The overall prevalence (95% CI) of unintentional childhood injuries, including minor and trivial was noted to be 14.5% (14.1-14.9). Fig. 2 describes the samples selected and the distribution of the types of injuries in all sites combined. However, after excluding minor and trivial injuries, it was noted that 1452 children reported 1535 events of unintentional injuries. The prevalence rate of injuries in various sites ranged from 0.6-12.0% (rural areas) and 1.1-8.4% (urban areas) (Fig. 3). The prevalence was higher among males as compared to females (5.9% vs 3.3%, P<0.001). The differences in the prevalence rates between different age groups was found to be statistically significant (P=0.01), with the lowest prevalence rate among children below 1 year. The prevalence rate showed a decreasing trend with increasing socioeconomic status. The difference in prevalence based on the number of children in the family was minimal (Table I).