Study setting and research design
This study was conducted within Narok County where 40% of girls aged 15–19 years have begun child-bearing. In this region of study, 7.4% of adolescents are pregnant with their first child and 33% have ever given birth as compared to the national levels of 3.4 and 14.7%, respectively. These statistics are supported by the risks facing adolescents in Kenya which include but not limited to: high HIV infections, particularly among girls (16% of people living with HIV are aged 10–24 years); high teenage pregnancies (18%); early marriages (11%) for older adolescents (15–19 years); persistent female genital mutilation (11%); high rates of anaemia (41%) among pregnant adolescents; high number of adolescents exposed to sexual violence (11%) and physical violence (50%) as well as low secondary school attendance with a net ratio of 47%. All these risks perpetuate further the vulnerability of this age group to a healthy life.
The study was carried out in Trans Mara East Sub-County within Narok County. Trans Mara East Sub-County was purposively selected since it is the smallest in size (275.4 km2), among the four sub-counties in Narok County and had the highest prevalence of teenage pregnancies [9]. To achieve the objectives of this formative study, a cross-sectional study employing concurrent nested quantitative priority mixed methods approaches with both quantitative and qualitative research techniques was applied as previously reported in our work [10]. In this case, quantitative data were given priority while qualitative data were applied to triangulate some relevant quantitative components.
Study population and sampling technique
Study population
The primary study population comprised of pregnant and lactating adolescent girls (aged 10–19 years old) resident in Trans Mara East Sub-County.
Sample size determination for quantitative approach
Sample size was determined using the Cochran formula [11], which allowed for calculation of an ideal sample size given a desired level of precision, desired confidence level, and the estimated proportion of the attribute present in the population. A total sample size of 292 was applied as previously reported in our work [10]. As previously indicated, proportionate distribution was done across 25 clusters equivalent to villages and by adolescent status (i.e. pregnant or lactating) to select study participants.
Test for sample size adequacy
Given the nature of the questionnaire where 90% of key variable measures were based on 5 point-Likert scale, descriptive test for sample size adequacy was performed using Kaiser-Mayor Olkin and Batt-test of sphericity as previously described [10].
Sampling procedure
Sampling was carried out as described in our work [10]. In brief, probability sampling techniques using cluster and simple random methods was used to practically access pregnant and lactating adolescents. An enumerator covered at least one village in a day to administer at least eight (8) questionnaires at random. Each enumerator moved to the center of a village selected for the day and began by facing North direction. After that, eight papers representing North (N), North East (NE), East (E), South East (E), South (S), South West and West (W) were shaken to give each direction equal chance of being selected and one piece finally picked to inform the direction to walk. Once direction was picked, an enumerator walked on a straight line to the next household until he/she reached a household with an eligible adolescent. Once the first adolescent was interviewed the enumerator again stood at the door of the just completed house facing North again and picked a direction from the pieces of papers randomized. The enumerator again walked to the next household. This process was repeated throughout the day until all eight adolescents were interviewed. An enumerator that reached the end of the village before completing the numbers required would go back to the centre of the village and randomly select a new direction to walk to. In case an enumerator double-selected the previous household, that household was passed until another eligible adolescent was reached. Each time an enumerator strived to interview three adolescent who were pregnant against seven who were lactating. This was to allow for proportionate distribution as per the target group listing ratio.
Methods of data collection tools and process
Quantitative data was collected using adolescent questionnaire targeting critical indicators of access, utilization and individual power dynamics (See Supplementary File I). Focus Group Discussion guide was administered to adolescents and Community Health Volunteers (CHVs). For quality control purposes, the data enumerators were trained on the procedures and ethical issues related to the data collection and the instruments were pre-tested prior to use. The collection of data was performed under the supervision of the principal investigators. In each case, Kipsigis (local language), Kiswahili or in exceptional cases, English, was used as medium of communication.
Questionnaire-interview method
The questionnaire was administered to each respondent by an enumerator for a period of about 45 min. The focus was given to participants’ value label attached to critical items such as Collection and use of Iron and Folic Acid Supplements (IFAS), Regular nutrition assessment, Practice of quality of diet, Use of Ready to-Use Therapeutic Supplements/Ready to Use Supplementary Feeds (RUTS/RUSF), Vitamin A supplementation for the child, Use of Insecticide Treated Nets (ITNs), Regular visit for Nutrition education and counselling and overall adherence to utilization. These were treated generally as proxy quality indicators. Utilization pattern associated with nutrition services was assessed in such a way that participants who scored 4 or more items against a scoring rating between 4 and 5 were labelled ‘good utilizers’ while those who scored between 1 and 3 were labelled ‘bad utilizers. Good utilizers were assumed to have high chances of accessing quality of nutrition services. Validation of all proxy quality indicators were done based on principle component factoring and all the indicators revealed uni-dimentionality of measures.
Focus group discussions (FGDs)
Three focus groups targeting Community Health Workers, Parents (a group of biological parents of adolescents who either pregnant or lactating) and Mother-to-Mother Support Group (a group of adolescent pregnant or lactating) were conducted to understand issues surrounding nutrition needs of the adolescent girls who are pregnant or lactating. Focus was given to quality of the services received. Each target group was made up of an average of eight (8) members to engage in free discussions. Focus group discussions (FGDs) were conducted by a trained facilitator who also acted as moderator and a note-taker. All discussions were audio-recorded using digital recorder and transcribed verbatim. Major questions of FGDs included who provides nutrition advice and services for adolescent pregnant/lactating for adolescent pregnant/lactating at the health facility? A mention of some of the facilities’ nutrition pieces of advice/services provided for adolescent pregnant/lactating mothers; how accessible are these facilities and how are the nutrition advice information conveyed to the adolescent pregnant/lactating whenever they visit a facility to seek services? Finally, we asked them to gauge the level of satisfaction with the nutrition and health information provided to the adolescents.
Statistical analyses
Quantitative data analysis adopted use of descriptive and inferential statistics. Descriptive statistics was used to characterize different frequencies. Z-test for single proportions was used to test for significant difference between the actual frequencies and expected frequency. Expected frequency was set at 50% for dichotomized data and 100/n percent for data that had more than two options. Principal Axis Factoring was used to establish the access pattern as well as generating Batt-scores for further modeling especially for indicators that were fitted into access and utilization models to determine cause and effect. Bivariate analysis based on odds ratio were done to determine how each of the socio-demographic and facility-based variables relate with perceived quality of nutrition services among pregnant and lactating adolescent girls. Expected frequencies for categorical data proportions were estimated as a ratio of 100% to the number of possible categories for each variable.
Diet diversity score
A simple dietary diversity score was used to predict micronutrient adequacy of diets of women of reproductive age. The food groups considered in the score for the Women Diet Diversity Score (WDDS) put more emphasis on micronutrient intake [10] than on economic access to food. A score based on nine food groups was chosen. Consumption recall was within 24 h food consumption period. The final score grouped the scores into highest category (> 6 food groups), medium (4 and 5 food groups) and Low (at most 3 food groups).
Qualitative Data Analysis on the other hand adopted the use of Framework analysis [12] for Focus Group Discussions. One key advantage with this framework analysis is that although it uses a thematic approach, it allows themes to develop both from the research questions and from the narratives of research participants. The process of data analysis began during the data collection, by skillfully facilitating the discussion and generating rich data from the interviews and FGDs, complementing them with the observational notes and typing the recorded information. This stage was followed by familiarization with the data, which was achieved by listening to voice records, reading the transcripts in their entirety several times and reading the observational notes taken during and after the interview and/or FGDs. The aim was to immerse in the details and get a sense of the interview as a whole before breaking it into parts. The next stage involved identifying a thematic framework, by writing narrative memos in the margin of the text in the form of short phrases, ideas or concepts arising from the texts and beginning to develop categories. At this stage, descriptive statements were formed and an analysis carried out on the data under the questioning route. The third stage, indexing, comprised sifting the data, highlighting and sorting out quotes and making comparisons both within and between cases. The fourth stage, charting, involved lifting the quotes from their original context and re-arranging them under the newly-developed appropriate thematic content.