Study design
Between June 2016 and January 2017, we conducted a cross-sectional, community-based survey of adolescent girls in West Singhbhum district, Jharkhand, a state of eastern India. This study was the baseline survey to inform the design of a cluster-randomised controlled trial (ISRCTN17206016) of an intervention to improve adolescent health with Ekjut, a civil society organisation based in Jharkhand, in partnership with University College London.
Setting
The size of the study area was chosen for logistical reasons and comprised 50 villages and their attached hamlets in Khuntpani block, West Singhbhum district, southern Jharkhand. Over a quarter of Jharkhand’s population is from Scheduled Tribes [10]. The majority of the population living in our study villages were from the Ho tribe, a tribal group comprising over 1 million people, around 930,000 of whom reside in Jharkhand [12]. The Ho families in our study areas were mainly engaged in cultivation and seasonal migration for wage labour.
Because this study was the baseline survey for a cluster randomised controlled trial, the 50 villages and hamlets were grouped into 38 population clusters. Each cluster was a purposively selected geographic area with a population of around 1000 (range: 723–1962). All of the 50 study villages had an Anganwadi centre providing basic healthcare for mothers and children. Eleven out of 50 had a sub-centre usually staffed by an Auxiliary Nurse Midwife (female) and a multipurpose worker (male). Most villages (33/50) had a primary school and 19 had a secondary school.
Survey methods
Inclusion criteria
We aimed to interview all married and unmarried adolescent girls aged 10–19 living in the study area.
Recruitment process
An interviewer visited each household outside school hours to identify eligible girls, obtain their consent for participation as well as the consent of their caregiver, and to conduct face-to-face interviews with girls. All interviewers were female. Depending on the participant’s preference, interviews were conducted in Hindi, Ho or Oriya by interviewers who were fluent in these languages. If an eligible girl was unavailable due to e.g. attending a boarding school or living in a hostel, the reason was documented. Interviewers did not make a repeat visit to the household.
Survey instrument development
Our survey instrument is provided as an additional file (Additional file 1). The instrument included questions on physical and mental health, disability, nutrition, sexual and reproductive health, gender norms, decision-making, education and violence. Interviewers also conducted short interviews with girls’ caregivers (or girls themselves if aged 18 or older) to collect household-level socioeconomic data.
For some areas of adolescent health, we were able to adopt questions used in previous national surveys of adolescents in India and adolescent health indicators from the international literature [13,14,15]. For example, we assessed literacy by asking girls to read a standard sentence, as in India’s National Family Health Survey. We asked about disability using the Washington Group Short Set of Disability Questions [16]. These assess whether participants have difficulty with walking, seeing, hearing, cognition, self-care or communication, are suitable for children aged 5 years and above, and have been used extensively in India [17].
Similarly, to assess girls’ nutritional status, we used standard international indicators and methods. We measured girls’ height using a Seca 213 stadiometer, their weight using a Seca 874 scale, and Mid Upper Arm Circumference (MUAC) using a standard adult tape procured from UNICEF. We defined stunting (height for age) and thinness (BMI for age) as − 2 Standard Deviations (SD) below the mean sex-specific WHO Reference 2007 [18]. We defined overweight as + 2 SD above the mean. In reporting BMI-for-age statistics, we excluded girls who were pregnant for the first time, and girls who had already been pregnant if it was unclear whether or not they were pregnant at the time of the survey. We calculated the proportion of girls aged 15–19 with a BMI less than 18.5. There is no standardised MUAC cut-off to identify adolescent thinness, so we used a cut-off of < 160 mm among girls aged 10–14 years based on nutrition guidelines for HIV-infected children [19]. Interviewers received anthropometry training and participated in a standardisation exercise with 10 adolescent girls to assess Technical Error of Measurement. The inter-observer coefficient of reliability was 0.98 for height and 0.99 for weight and MUAC. Each measurement was repeated and we calculated a mean of the two measurements. We used the Food and Nutrition Technical Assistance (FANTA) tool to measure minimum dietary diversity, i.e. the proportion of girls who had consumed five or more of 10 key food groups in the last day or night [20].
We assessed knowledge and practices related to sexual and reproductive health among all girls aged 15–19 (married and unmarried) as well as girls aged 10–14 who were married. We used questions from the Youth in India: Situation and Needs Study 2005–6 [14]. We asked girls whether they had heard about contraception, what kind, from where, and about whether they thought abortion was legal. We also asked about previous pregnancies. Married girls were asked whether they or their husband had used any methods to prevent or delay pregnancies.
For other areas of adolescent health however, there were no validated instruments or internationally recognised indicators. For example, there are no appropriate brief screening tools for adolescent mental health validated in India. We decided to use the Brief Problem Checklist (BPC), which comprises a six-item Internalising scale and a six-item Externalising scale, as well as a Total Problems scale. This tool was developed and validated for a US clinical population and was shown to be both easily administered and psychometrically strong [21]. Interviewers in our study found the BPC simple to use and participants understood the items easily. In our sample, Cronbach’s alpha, a measure of internal consistency, was 0.77. The BPC has now been replaced by the Brief Problem Monitor (BPM) [22].
Collecting data on violence was equally challenging. We asked about girls’ exposure to emotional, physical and sexual violence using a translated and modified version of the Child Abuse and Neglect Screening Tool – Child Version (ICAST-C) [23]. The tool was adapted to shorten administration and to reflect local forms of violence such as being made to work or look after siblings, and exposure to ‘societal violence’ defined as witch hunting, communal violence, social boycott, being prevented from accessing public facilities or common resources, or being subjected to a community-imposed penalty.
Data management and statistical methods
We collected data using smartphones programmed with CommCare software [24]. We coded and analysed data in Stata version 14 [25]. For nutrition data, we computed height for age and BMI for age z-scores with a Stata macro for the WHO Reference 2007.
We report descriptive statistics (mean, SD, range) for each area of adolescent health need. We present results for younger (10–14 years) and older (15–19 years) adolescents as well as for the total sample because these age groups differ and there is relatively little information on the younger age group in particular.
Ethical considerations
The study was approved by an independent ethics board convened by Ekjut in Ranchi, Jharkhand, and by the Research Ethics Committee of University College London. Ekjut had been working in the study districts and collecting data on maternal and child health in tribal communities and with local data collectors for 10 years, so had a good rapport with the study communities. However, the fact that the study involved asking adolescent girls questions about potentially sensitive issues presented substantial ethical challenges.
Consent
We sought consent for each village’s participation from the local village governance institutions (Panchayat and headmen) and opinion leaders after explaining the study’s purpose and processes. Interviewers explained to all adolescents and any parents or husbands that the survey would include questions on health, nutrition, as well as potentially more sensitive subjects such as alcohol and tobacco use, feelings and worries about growing up, about safety, or about experiences at school and at home. Interviewers said that participation was voluntary, that choosing not to participate would not disadvantage the family or adolescent in any way, and that participants could stop the interview at any time or skip any questions that they did not want to answer. We obtained consent (a witnessed thumbprint) from all girls who participated in the study. For girls younger than 18 years, we also sought consent from their caregiver. For married adolescent girls younger than 18 years, we sought consent from the husband, though married adolescent girls were interviewed privately. We did not collect data from any girls if they had not themselves provided informed consent, regardless of whether or not their caregiver or husband had.
Confidentiality
Data collectors needed to be able to speak Hindi and Ho, but could not be from the villages in which interviews were being conducted in order to protect participants’ confidentiality. We followed WHO ethical and safety recommendations for research on domestic violence against women by always asking questions about violence in a private space, with no third-party present, and within the bounds of a more general survey [26]. All data collected on smartphones were anonymised following downloads, and interviewers had password protected smartphones.
Referrals
Ekjut convened a local multidisciplinary child protection committee to help coordinate the referral of vulnerable adolescents identified through the research. The committee included the Child Protection Officer and Probation Officer from the District Child Protection Unit, all members and the Chairperson from the Child Welfare Committee, the Office Legal Assistant from the District Legal Services Authority, and NGO representatives. We used the data collection software (CommCare) to flag participants who had faced sexual or physical violence, as well as those with severe mental health problems or severe undernutrition to each interviewer. The interviewers and their supervisors then confidentially offered help from a trained psychosocial counsellor to visit these adolescents, discuss their needs and organise any onward referral with explicit consent from the adolescent.