Individual‐level drivers of dietary behaviour in adolescents and women through the reproductive life course in urban Ghana: A Photovoice study

Abstract Evidence on the individual‐level drivers of dietary behaviours in deprived urban contexts in Africa is limited. Understanding how to best inform the development and delivery of interventions to promote healthy dietary behaviours is needed. As noncommunicable diseases account for over 40% of deaths in Ghana, the country has reached an advanced stage of nutrition transition. The aim of this study was to identify individual‐level factors (biological, demographic, cognitive, practices) influencing dietary behaviours among adolescent girls and women at different stages of the reproductive life course in urban Ghana with the goal of building evidence to improve targeted interventions. Qualitative Photovoice interviews (n = 64) were conducted in two urban neighbourhoods in Accra and Ho with adolescent girls (13–14 years) and women of reproductive age (15–49 years). Data analysis was both theory‐ and data‐driven to allow for emerging themes. Thirty‐seven factors, across four domains within the individual‐level, were identified as having an influence on dietary behaviours: biological (n = 5), demographic (n = 8), cognitions (n = 13) and practices (n = 11). Several factors emerged as facilitators or barriers to healthy eating, with income/wealth (demographic); nutrition knowledge/preferences/risk perception (cognitions); and cooking skills/eating at home/time constraints (practices) emerging most frequently. Pregnancy/lactating status (biological) influenced dietary behaviours mainly through medical advice, awareness and willingness to eat foods to support foetal/infant growth and development. Many of these factors were intertwined with the wider food environment, especially concerns about the cost of food and food safety, suggesting that interventions need to account for individual‐level as well as wider environmental drivers of dietary behaviours.


| INTRODUCTION
Sub-Saharan Africa is rapidly urbanizing and is experiencing changing dietary behaviours as food habits and food environments become increasingly linked to marketization, industrialization and globalized food supplies (Agyemang et al., 2016;Holdsworth & Landais, 2019;Rousham et al., 2020). Changing nutrition landscapes, often referred to as the nutrition transition, have shifted the global disease burden from communicable to non-communicable diseases (NCDs) (Baker et al., 2020). In 2021, estimates indicated that 77% of NCDs were found in low-and middle-income countries (LMICs) (World Health Organization [WHO], 2021). Increased prevalence of NCDs in these settings is further compounded by multiple burdens of malnutrition (micronutrient deficiencies, undernutrition as well as overweight and obesity), often present within the same individual, household or population (Popkin et al., 2020). Unhealthy diets, propelled by shifts in food environments and dietary changes, are one of the major drivers of this emerging phenomenon. Adolescent girls and women in Africa are more vulnerable to overweight/obesity than men and adolescent boys (Case & Menendez, 2009;Kanter & Caballero, 2012;Muthuri et al., 2014), partly because of the consumption of energy-dense, nutrient-poor foods (Sedibe et al., 2014;Trubswasser et al., 2020).
Ghana is a highly urbanized country (~60% of the population lives in urban areas) (Ghana Statistical Services [GSS], 2021) that has reached an advanced stage of the nutrition transition (Agyemang et al., 2016;Ecker & Fang, 2016). Dietary behaviours in urban Ghana have been modified by urbanized lifestyles and increased preference for imported food (Food and Agriculture Organization [FAO], 2009[FAO], , 2021, which may contribute to the increased prevalence of overweight/obesity among women (34.4% in 2006 to 39.2% in 2016) and school-aged and adolescent girls (12.6% in 2006 to 17.5% in 2016) (Global Nutrition Report, 2021). NCDs account for 43% of total deaths in Ghana (WHO, 2018(WHO, , 2020. In addition, poor health outcomes from diet-related NCDs (DR-NCDs) are particularly common among Ghanaian women (Agyei-Mensah & de-Graft Aikins, 2010;GSS, 2015;Ofori-Asenso et al., 2016;Ofori-Asenso et al., 2017). Given this nutritional context, identifying factors that drive dietary behaviours is essential, especially as the Ministry of Health (MoH) of Ghana has placed integrated interventions to promote healthy diets at the core of its public health policies (MoH, 2012(MoH, , 2020. A range of models and frameworks have been developed to understand the drivers of food choice and how food environments can influence individual-level dietary behaviours (Marijn Stok et al., 2018;Osei-Kwasi et al., 2021;Story et al., 2008;Turner et al., 2018). This paper will contribute to the growing evidence on the influence of individual-level factors on dietary behaviours in Africa across the different dimensions of the food environment. Individual-level factors are important to investigate as they may influence food consumption through different pathways, such as self-efficacy and skills (Story et al., 2008). As adolescent and young adult populations increase worldwide (Norris & Richter, 2016), alongside rapidly changing food environments (Holdsworth & Landais, 2019;Turner et al., 2018), interventions targeting adolescent girls and women of reproductive age are needed as they have the potential to promote positive lifelong and intergenerational nutrition outcomes (Norris et al., 2022;Wells et al., 2020) as they progress into different stages of the reproductive life course. Ensuring good nutrition among all these age groups, coupled with female empowerment, can help improve dietary diversity and overall diet quality in Ghana (FAO, 2021).
This study, therefore, aims to identify the individual-level drivers of (un)healthy dietary behaviours of adolescent girls and women at different time points during their life course, among socioeconomically deprived urban neighbourhoods in Ghana.
More specifically, the study investigates (i) the individual-level (biological, demographic, cognitions, practices) drivers of (un) healthy food consumption and (ii) whether there are any differences in the factors influencing dietary behaviours between women at different stages of life course (i.e., early adolescence, pregnancy or lactating status).

| Study setting
This study was part of a wider project, the Dietary Transitions in Ghana project (datalink: https://dataverse.ird.fr/dataverse/diet_trans_ ghana;jsessionid=d8c3c605c1c1bf3125e01476d0f6), conducted in Accra (Greater Accra region) and Ho (Volta region), as we were interested in capturing cities with different levels of urbanization and prevalence of overweight/obesity (as a proxy for nutrition transition). In 2015 (study conception), overweight/obesity prevalence among women of reproductive age (WRA) was 57.3% and 31.1% in Greater Accra and the Volta region, respectively (GSS, 2015).

Key messages
• Time constraints, eating at home, eating out, cooking skills, food preferences and food safety concerns were identified as key factors influencing dietary behaviours at the individual level in urban Ghana.
• Ability to eat nutritious, safe food was largely mediated by income and wealth.
• Biological factors, such as pregnancy/lactating status influenced behaviours through medical advice, nutrition knowledge and willingness to promote foetal/infant development.
• Many individual-level factors were intertwined with the wider food environment.
• Factors influencing dietary behaviour at different levels need to be considered together when developing interventions/policies for healthier diets.

| Study design
A qualitative study was conducted among young adolescent girls (13-14 years) and WRA (15-49 years) living in socioeconomically disadvantaged neighbourhoods in Accra and Ho. The study was designed to identify a range of factors at the individual, social, physical and macro-levels that influence dietary behaviours (Story et al., 2008). This paper reports the findings on the individual-level factors that emerged. The findings on the influence of the physicallevel (accessibility, affordability, convenience, etc.) food environment on dietary behaviours have been previously published .
Photovoice, a community-based participatory photography method, was used to allow participants to document influences on their dietary behaviours in their daily lives. This method facilitates in-depth exploration, stimulates reflection and enables discussion among participants and policymakers to foster change in a community (Wang, 1999).
While Photovoice has largely been used in high-income countries (Belon et al., 2016;Díez et al., 2017;Gravina et al., 2020;Heidelberger & Smith, 2016), recent studies have used this method in Africa, to assess factors influencing adolescents' dietary behaviours in urban Ethiopia (Trubswasser et al., 2020), among women in rural/urban Uganda  and balancing work and childcare in Kenya (Hani Sadati et al., 2019). The Photovoice methodology was selected as it places the research participant at the centre of the research process, opening up a pathway for dialogue between the researchers and the participants in a way that face-to-face interviews or focus group discussions alone do not.
Photographs allow access to the participants' world and can help to break down power dynamics between the researcher and researched, encouraging reflection, recall and discussion (Auma et al., 2021).

| Sampling
A list of all deprived neighbourhoods in Accra and Ho from the Accra Poverty Mapping Exercise (CHF International, 2010) and United Nations Human Settlements Programme urban profiling report (UN-HABITAT, 2009) were used to select two neighbourhoods: James Town (Accra) and Dome (Ho) (see further detail in Supporting Information 1). To ensure diversity, participants were purposively selected using quota sampling based on age/reproductive life course stage, gender, body mass index (BMI), education, occupation, maternal status and socioeconomic status (SES) (Supporting Information 2). A subsample (i.e., a third) of the overall study population was randomly invited to partake in the Photovoice study, resulting in 32 participants in Accra and Ho (n = 64 total). Recruitment took place through the communities, schools and health facilities (see Supporting Information 3 for additional information).
Before the project began, initial formal meetings with community leaders were held to explain the study and establish community entry. These meetings encouraged community mobilization and engagement with the study and facilitated data collection. RP led the qualitative fieldwork training for seven Ghanaian research assistants. Fieldwork was conducted by native speakers, who were not members of the targeted communities.

| Data collection
Data for the Photovoice study were collected between May and December 2017. The Photovoice interview guide was adapted from the original format proposed by Wang (1999) (Supporting Information 4). Initial community engagement activities revealed that women in these urban areas had busy schedules outside of the home setting, making it difficult to organize group discussions at a time suitable to all participants. Therefore, individual interviews were conducted instead of focus group discussions. The Photovoice interview guide was piloted in Accra (n = 3) and Ho (n = 3) and then amended, accordingly, thus excluding them from the analysis stage.
The Photovoice study took place in three stages. The first stage was comprised of an initial home visit, where participants were trained on: (i) the consent process (because they potentially would photograph people); (ii) the Photovoice methodology; (iii) the use of a camera to take photographs; (iv) photography ethics, including the 'no face or identification details' protocol to ensure the anonymity of people or places (Supporting Information 5). Participants were asked to take photographs that identify factors driving their dietary behaviours. Specifically, they were asked to take five photographs on the following themes: (i) a place where you eat food and/or drink; (ii) Something that makes eating healthy difficult for you; (iii) something that makes eating healthy easy for you; (iv) something that influences what you eat in your area/neighbourhood; (v) a person that influences your food or drink choices in your area/ neighbourhood. During the second stage, two follow-up visits were made to check on progress. The third stage consisted of an in-depth interview that lasted 45-60 min. Interviews were conducted with participants in their preferred language: Ga (n = 24); Twi (n = 5); English (n = 3) in Accra and Ewe (n = 28); English (n = 3); Twi (n = 1) in Ho, respectively. During the interviews, participants told the 'stories' related to their five selected photographs. When data collection was complete, a photography exhibition was held to raise awareness of drivers of unhealthy food and beverage consumption in the targeted communities. Photographs from the data collection stages were used as a tool to facilitate dialogue between study participants, the media and local government officers. The photography exhibition also promoted community dialogue and stakeholder engagement by sharing results with the wider community.

| Data analysis and synthesis
In-depth interviews were transcribed and translated verbatim into English for analysis. All coders, RP/AT/SL, used an agreed-upon codebook in NVivo version 11 to ensure consistency and accuracy, with blind double coding of 25% of the transcripts (Fonteyn et al., 2008). Interviews were coded using deductive (a priori themes) and LIGUORI ET AL. | 3 of 16 inductive (data-driven codes) schemes, to allow for emerging themes (Supporting Information 6). Existing socioecological models of dietary behaviours and systematic review evidence from Africa (Gissing et al., 2017;Story et al., 2008) were used to identify factors, biological, demographic, cognitions (e.g., knowledge and preferences) and practices (e.g., skills and behaviours), influencing dietary behaviours at the individual level. The African Food Environment framework, an expert validated framework created to help prioritize research and intervention development in Africa, was also consulted and used to structure the reporting of our results .
Data were synthesized by creating a framework matrix with nodes for different themes and subthemes (Gale et al., 2013). Nodes were then broken down into four populations at different stages of the life course: early adolescents, WRA who were neither pregnant nor lactating, pregnant WRA and lactating WRA. Similarities and differences were highlighted between the different stages and the factors influencing dietary behaviour.

| Sociodemographic characteristics of the study sample
The Photovoice study was conducted with 64 female participants across the two cities (n = 32 in Accra; n = 32 in Ho). The age range of the study sample was 13-49 years, with 75.0% of participants aged 15-49 years (Table 1). Overall, 37.5% of participants were in work, 12.5% in education and 50.0% were not in work/education. Half of the participants were either pregnant or lactating. Almost half (48.4%) had a BMI ≥ 25 kg/m 2 (overweight or obese). a Household socioeconomic status (SES) was measured using the EquityTool (Chakraborty et al., 2016). SES scores were derived using proxy indicators of the household environment (ownership of consumer durables; source of drinking water and type of toilet facilities; type of materials used for the floors and walls; and land ownership). SES quintiles were subsequently derived. Participants were further classified into three groups: lowest SES (first quintile); low to middle SES (second and third quintiles) and high SES (fourth and fifth quintiles). For this project, only participants in the first and second tertiles, representing the lowest and low to middle SES, respectively, were selected.

| Individual factors influencing dietary behaviours
willingness to eat what is good for the baby's growth-that is, increased homemade consumption and diversified diet and willingness to consume foods that increase breast milk production: 'Because if I don't eat a lot or eat healthy food, they [young children] will not get the breast milk to feed on and they need to grow well […] At first, anything I get, I will eat it. much oil as a cause of malaria' and the nutrient-rich composition of malted beverages, frequently consumed during lactation, were held.

Risk perceptions around food safety
The majority of participants referenced food safety risks, indicating that there was a lot of anxiety and a good level of knowledge surrounding food safety. In addition, some participants discussed the need to limit the consumption of foods and beverages with additives sold in shops, like sugar-sweetened beverages and stock cubes. Participants deployed mitigation strategies to avoid falling ill when risk was perceived. Individual hygiene practices such as hand washing before meals were common among pregnant/lactating participants. Washing utensils after eating and preparing food in clean areas were also listed. One 13-14-year-old participant mentioned her ability to positively influence food hygiene within

Preferences
In all groups, food preferences were important. Many participants reported a preference for nutrient-rich foods such as fish, plantain, fruit and chicken. However, affordability was frequently mentioned as a barrier. Only three participants referenced eating well-liked foods that were specifically beneficial to their pregnancy. Stable income was also linked with personal and family-related preferences.
This was echoed by some younger and older adolescents, who described eating food available at home and using pocket money to buy sweets and other preferred foods.
Food characteristics (texture, taste, aroma, food appearance) Plain rice and small portions of purchased food were considered bland and participants added meat, eggs or fish to create 'fine' meals full of flavour. Food appearance, notably in terms of quantity or portion size, was frequently mentioned. Another participant mentioned that freshness was a key element to make a meal taste good.
In the pregnant/lactating group, two participants mentioned that sewage inside the home or from toilets or manholes out of the home, made eating difficult and caused loss of appetite.

Hunger and satiety
Eating well was associated with feeling full. Eating a filling meal was preferable for the majority of participants, as a way to stay satiated

Cooking skills
Participants in all reproductive stages described practices such as cooking skills, eating at home or out and time constraints (Table 4). In regard to cooking skills, only one participant (13-14 years) mentioned her ability to cook, with younger participants frequently describing involvement in food preparation or learning cooking skills at home.
Cooking at home between one and three times per day was mentioned as a common practice. Many culinary skills were described, such as cooking over charcoal, frying fish, boiling yams and grinding cassava/corn/nuts into flour. Among the 15-49-yearold participants, a lack of cooking skills led to unhealthy eating behaviours despite food safety concerns as there were few alternative solutions available: '[…] after that experience, you may decide not to buy food there again but because you don't know how to cook at home, you will still go and buy food there again' [Accra, 19 years, low-middle SES].

Eating at home or out
Eating food prepared at home was most common. Eating out or getting a takeaway were practised with varied frequencies. Homemade food was the overall preference, as ingredients, flavour, portion size and hygiene level could be monitored. Younger adolescents ate out at school canteens, while adults described eating at work or on the way to or from work. Adolescents and WRA reported eating out more often on weekdays and eating more at home on the weekends, influenced by busy school/work schedules and income. Among the 13-14-year-old participants, eating homemade food was thought of as a positive, enjoyable practise that facilitated healthy eating. One participant even refused to eat school food. Several participants brought homemade food to work as it was convenient and affordable. Many individual-level factors, such as the cost of food, overlapped with the wider social and physical food environments Wanjohi et al., 2022). This suggests that interventions need to account for multiple levels and wider drivers of food consumption. This supports findings from previous studies in Kenya (Downs et al., 2022) and Ghana (Boatemaa et al., 2018), showing the need to target multiple levels of the food environment to help women negotiate factors such as food safety, nutrition, time, cost trade-offs that prevent them from adopting healthy diets.
Despite low levels of education, there appears to be an overall high level of knowledge and awareness of food safety and food hygiene and the impact these may have on diets. Food safety was also observed as a key factor influencing adolescent dietary behaviours in Ethiopia (Trubswasser et al., 2020) and women of reproductive age in Uganda (Yiga et al., 2020). Participants used several risk mitigation strategies, such as preparing and consuming homemade food and eating in clean environments, to ensure the hygienic preparation of food consumed. Despite food safety concerns, participants continued to eat out of home, buying hot meals, eating in cleaner environments and from familiar vendors to reduce perceived food safety risks where finances allowed. Continued eating out practices, influenced by affordability, preference and taste and lack of alternatives were observed in a recent systematic review on food safety concerns in LMICs .
Actions to improve food safety among food vendors have emerged as a core concern among participants in several research studies in Ghana (Boatemaa et al., 2018;FAO, 2016;Pradeilles et al., 2021;Rheinlander et al., 2008). Individuals' primary concern related to food hygiene is observed in street food practices in LMICs Alimi, 2016;Omari & Frempong, 2016), with concern increasing among participants that have experienced prior episodes of food-borne illness (Adam et al., 2014), which supports the need to implement healthy food environment policies in these cities  nutrient-poor foods, which are widely consumed among this group, without creating a feeling of satiety (Drewnowski & Darmon, 2005).
Among low SES groups, there appears to be an emphasis on consuming maximum calories, rather than nutritional quality (Darmon & Drewnowski, 2008). A study in the same communities found that time allocated to a meal was usually <30 min for the vast majority of study participants . This finding supports increased incentives and subsidies targeting local food vendors to provide healthy foods that are convenient and can be consumed quickly .
This study includes several strengths, namely, the sampling method used to achieve diversity across the life course, application of the African Food Environment framework and the use of Photovoice (i.e., added value over commonly used methods like in-depth interviews and focus group discussions only). Participants were asked to tell their stories and to engage with a research topic that sought to better understand the current situation within their community. Using photography allowed participants in low-income communities to have an additional means of communication to identify, capture and discuss challenges and facilitators to eating healthily. While individual Photovoice interviews were conducted in place of group discussions, participants were able to discuss their concerns with a larger audience directly during a community-based photography exhibition . Nevertheless, as participants were only selected from two neighbourhoods in Accra and Ho, additional or differing factors may also be a concern within urban neighbourhoods and rural areas in Ghana. It is also important to consider the potential for limited success when individual-level approaches do not account for the wider food environment factors that influence the individual level (Allender et al., 2015;Doak et al., 2006;Mackenbach et al., 2014;Osei-Kwasi et al., 2020;Story et al., 2008;Tanentsapf et al., 2011).

| CONCLUSION
In