The conceptual framework for a combined food literacy and physical activity intervention to optimize metabolic health among women of reproductive age in urban Uganda | BMC Public Health

Step I: Needs assessment

Findings from our systematic review [16] were used to design a theoretical framework for the qualitative study [17]. Notable determinants identified in the systematic review were financial and time limitations, health/beauty paradox (= overweight/obesity as a sign of beauty and wealth), and lack of knowledge, self-efficacy, and skills. Qualitative study findings re-affirmed the systematic review findings concerning health/beauty paradox, knowledge, self-efficacy, and skills gaps. In addition, the qualitative study showed socio-cultural misconceptions around lifestyle PA, fruits, vegetables, and habitual orientation towards carbohydrate foods. We also found that there is a high trust in nutrition information shared on social and mass media, yet skills to evaluate this nutrition information are limited. Figure 1 below shows the logical model of needs assessment, summarises the determinants of dietary and PA in urban Uganda [16, 17].

Logical model of needs assessment, summarizing the personal and environmental determinants of dietary and PA behavior in urban Uganda. Adapted from Yiga et al., [16] and Yiga et al., [17]

Step II: Formulation of behavioral intervention, performance, and change objectives

We hypothesised that changing the overall existing behaviours towards WHO healthy lifestyle guidelines in one intervention may meet strong resistance and thus may not be effective. For example, the planning group hypothesised that due to the existing health/beauty paradox and habitual orientation towards carbohydrate rich foods, interventions focusing directly on weight loss and reduction of portion sizes of foods rich in carbohydrates may meet strong resistance. Therefore, we decided to go for more feasible gradual changes able to enact clinically relevant metabolic improvements. We hypothesised that increased consumption of vegetables and fruits will indirectly translate into reduction of portion sizes of carbohydrate rich foods. In line with WHO health recommendations, the intervention aims to stimulate WRA to consume at least 400 g fruits and vegetables [13]. Moderate intensity PA that can be incorporated in daily life activities may be the achievable type of PA among WRA compared to structural high intensity PA [26]. Non-factual nutrition information influences dietary and PA behaviors in urban Uganda [17]. Thus, we decided to supplement the intervention with a component on information evaluation; to enact ability to distinguish evidence-based information from nonfactual information.

Accordingly, three behavioural intervention objectives were formulated.

1. 1.

   Women evaluate the accuracy of food, nutrition, and PA information.

2. 2.

   Women engage in moderate intensity PA for at least 150 min a week.

3. 3.

   Women consume at least one portion of vegetables and one portion of fruit every day.

Table 1 shows the behavioral intervention objectives, subdivided into POs providing the answer to the question; “what do the participants of the intervention need to do to achieve the behavioural objectives”. The model of food literacy [27] guided the formulation of POs. Food literacy is the interrelated combination of knowledge, skills and self-efficacy to (i) plan, (ii) select, (iii) prepare, (iv) eat food with the ultimate goal of developing a lifelong healthy, sustainable and gastronomic relationship with food within the prevailing environment [27, 28]. The POs were based on the above mentioned four components of food literacy (plan, select, prepare, and eat). For PA, a similar model was adopted, where “eat” was replaced with “do”, that is; plan, select, prepare, and do. The model of food literacy was chosen as it is a holistic behavior change model focusing on a “how to do approach” to initiate and sustain healthy eating habits [27, 28]. Evidence shows a positive association between food literacy and healthy dietary behaviors, particularly increased intake of vegetables and fruits [29, 30]. Table 2 shows the determinants considered to have a strong influence on accomplishing the created POs. Matrices of change objectives are presented in Additional file 3.

Step III: Selection of theory-based methods and practical strategies

We aimed to create an intervention capable of initiating and sustaining behaviour change. Eleven BCTs scientifically shown to enact changes in knowledge, skills, self-efficacy, subjective norms, and social support were selected, Additional file 4. The selected BCTs are supported by the self-regulation theory and self-determination theory which specifies the need for autonomy, competence, and relatedness to attain a positive behaviour change [33, 34]. Accordingly, our intervention aims to create behavioural change through enacting autonomy, competence, and relatedness. Providing information coupled with motivation interviewing creates a positive intention [35]. Implementation intentions can be achieved through goal setting [24, 34, 35]. Goal setting necessitates competence, which we hypothesised to be attained through a combination of (i) action planning; (ii) guided practice; ii) self-monitoring; iv) feedback on performance and v) planning of coping plans [24, 26, 34,35,36]. To sustain the behavioural goals requires relatedness, which can be achieved using a combination of social support, role modelling, feedback, planning coping responses and motivation interviewing [20, 24, 34].

The selected BCTs were then operationalised into practical strategies. BCTs; motivational interviewing, role modelling, feedback, guided practice, social support through exchanging ideas and planning coping responses were translated into interactive group-based sessions. Brainstorming workshops with planning group II and FGDs with target group revealed that group sessions may be the best strategy to deliver the intervention in this setting.

“Through education sessions, like you come in this group and give us a health talk, like the way you have come, you teach us and then us we can go and teach our other friends out there. Like for us every Tuesday we be meeting here, very many of us, so if you say you will give us one Tuesday in a week or month, or the last Tuesday of a month and you come and teach us”. “It would be very nice, because literally I share the information with others, so it will move, it moves much faster, because these groups are not only here, but also have these groups in other dioceses, so we can go visit them, and the teach them, but in health centers you only visit when you’re sick”. “Yes it helps, what I know is good, I wish it for my friends and we act as a support for each, and we as well spread it to other groups, example of myself, I used to never eat pumpkin, but I got it from these ladies, that this pumpkin is good and with time I gradually started to eat it until it become part of my diet”, participants in FGD 4 and 6.

Additionally, a recent systematic review shows that diet and PA interventions delivered through group sessions are effective in promoting clinically relevant weight loss [34]. These groups provide opportunities for social support, experience sharing, and may create a motivating atmosphere [22, 34]. Our needs assessment as well revealed that the community and church small groups are an opportunity to share dietary and PA counselling [16, 17]. Our environmental asset assessment revealed existence of women groups within religious structures. Existing groups boosts social cohesion, a facilitator for behavioural change [22].

The reading culture of Ugandans is low.

“We need more of practical, and also the pamphlet, some of us don’t really understand so much, but if it brings out the picture very well, even I can pick interest in it”. “Pamphlets, some people are lazy to read”, participants in FGD 5.

So, the BCT of “providing information through imagery” was translated into infographics with less text and more locally recognisable visuals. Evidence as well shows that visuals increase attention, interest, and credibility of the messages [20].

During FGDs with the target group, participants emphasised the need for practical vegetable preparation skills.

“like we are trying to reduce cooking oil and other stuff from our daily life, so maybe we meet in a group, there is a demonstration whereby some food stuffs are prepared in the best possible way which is to the taste, and people learn how to prepare them, because most of us, do not know how to cook, that is the truth, but somebody may not even fry food, but it tastes so good, if you know how to mix the ingredients and so on. Yes, include cooking demonstrations”, participants in FGD 2.

Hence, BCT of “guided practice” was specifically translated into a practical vegetable group cooking session. We also included vegetable recipes based on locally available vegetables in the intervention infographics. Intervention strategies linked to personal metabolic health and lifestyle needs, and environmental opportunities may help drive behaviour change and positively influence health outcomes [37]. Thus, BCT of; implementation intentions, goal setting and action planning were translated in to; (i) creating “if then plans” in line with metabolic health, (ii) SMART fruit/vegetable/PA goals, detailed action plans to achieve set SMART goals drawn considering environmental opportunities. Figure 2 below shows the hypothesised intervention logical model (conceptual framework) of behavioural change. Practical strategies built from BCT are hypothesized to effect changes in the change objectives, which in turn translate in changes in the determinants. Changes in the determinants in turn result in attainment of POs and corresponding behavioural intervention objectives.

hypothesised intervention logical model for behavioural change (conceptual framework for the intervention)

Step IV: Development of the intervention programme

The practical strategies were built into the intervention scope and sequence, Additional file 5. The intervention consists of five interactive group sessions, 150 min each, Fig. 3. A booklet (infographics); on benefits/recommendations, local vegetable recipes, and practical tips to eat more fruits, vegetables and do more PA is included as a guide, Additional file 6. Tools to assess PA and food environment for opportunities were included, Additional file 7. As well a self-monitoring tool for PA, fruit and vegetable intake was included for participants to track their behaviour daily goals for use in the feedback sessions, Additional file 8. The infographics were designed with locally recognisable images as cultural relevance of health promotion materials is vital for the success of an intervention [20]. Messages on the infographics were framed in a positive and active tone as evidence shows that positively framed messages are more acceptable [20].

Showing delivery timeline of the intervention sessions, intervention content (organised practical strategies from step III), role of participants, and anticipated outcome per session

Brain storming workshop with planning group I and FGDs with the target group identified religious institution women group structures as an appropriate potential delivery channel. The women group structures boosts established social networks, community reach (85{e4f787673fbda589a16c4acddca5ba6fa1cbf0bc0eb53f36e5f8309f6ee846cf} Ugandans are Christians) and trust. The channel offers an opportunity for assessing the intervention effectiveness in an unrestricted real-life community setting.

“Religious institutions because they are transparent, religious organizations because they reach out to a bigger community and then they are transparent. The health centers, there is that rudeness, and still for health centers will only meet those people who come to them, but the church, you get a bigger audience”, “Come to churches like this, people really belong to this communities, then you say every third Saturday or Sunday of the month, from 4 to 5 pm, there will always be a nutritional class, for the first-time people may not come, but eventually they come, if it is a free class”, participants in FGD 4.

STEP V: Adoption and implementation plan

The intervention will be delivered through institutional religious women groups (results of environmental asset assessment framework – see step IV). Through meetings with the strategic community leaders, a collaboration was established with Our Lady of Africa Catholic Parish, Mbuya. Mbuya Catholic Parish has six sub parishes. Within these sub parishes they are existing women groups, and these groups will be utilized for face-to-face intervention group sessions. FGDs with target group and meetings with planning group II pointed at the importance of opinion peer leaders being part of the implementation team.

“Our women group leader has helped us a lot, she taught us the dangers of cooking in polyethene bags and taught us the use of banana leaves, us we had got so much used to using the polyethene bags, she can’t eat the food you have prepared in polyethene bags, even if she visits you and if you have cooked like that, she can’t eat that food. “We have musawo (village health team) in our group, she usually brings for us education sessions on how to eat, she goes a lot for these education sessions and what she learns she brings them back to us”, participants in FGD 6.

Scientific evidence shows that the efficacy and acceptability of health promotion interventions increases if peer opinion leaders within the target group are part of the implementation team [38]. Peer opinion leaders provide entry and legitimacy to the external change agents and may help drive changes in social norms. Selection of peer opinion leaders: the intervention will be delivered within existing women groups. Leaders of these existing groups will be selected to work as peer opinion leaders on the implementation team. The main role and responsibilities peer opinion leaders will be to (i) mobilize fellow women to participate in the intervention, (ii) follow up and (iii) give social support to participating women to attain set intervention goals. Women leaders will be given a two – day refresher training on mobilization and leadership skills, as mobilization is the routine responsibility for women leaders in their usual group meetings. The planning group I designed the sessions to be moderated by health behavior coach (PhD researcher) following the techniques of motivational interviewing [39]. A general guide (scope & sequence) will ensure consistency during the group sessions.

Step VI: Development of an evaluation plan

Study design, setting and timing

The effectiveness of the intervention will be evaluated through a cluster-randomized controlled trial. The intervention will be evaluated in Kampala, the capital city of Uganda. The six sub parishes of Mbuya catholic parish will be randomized to treatment and control arms, Fig. 4. The treatment arm will be exposed to both the developed intervention infographics and face to face group sessions while the control arm will only receive the developed intervention infographics. An awareness session will be organized to distribute the infographics to the control arm. Within the sub parishes, there are existing women groups. These existing groups will be utilized for face-to-face intervention group sessions. For the intervention purposes, each group will be limited to a maximum of 14 members. The study period is divided into two phases: a three-month intervention and a three-month post-intervention follow-up phase.

Recruitment

The PhD researcher and women leaders of existing groups will make presentations about the intervention during one of the routine meetings. Flyers with details of the intervention will be distributed for sharing with members who are absent during the briefing. At the end of the presentations, interested participants will be invited for the first session to test their eligibility to participate in the study. Eligible participants will be provided with an informed consent form to endorse.

Inclusion criteria

1. i)

   Sex (women),

2. ii)

   Age (18 to 45 years),

3. iii)

   Central obesity [waist circumference ≥ 80 cm]),

4. iv)

   Fluent in either Luganda or English (sessions will be conducted in Luganda/English).

5. v)

   Willingness to follow the three-months intervention and three months follow-up,

6. vi)

   Willingness to sign the informed consent.

Exclusion criteria

1. i.

   Being treated for diabetes Mellitus Type 1 or Type 2, hypertension, high cholesterol, or any other cardio-metabolic related disease.

2. ii.

   Pregnancy.

Outcomes

Primary outcome is reduction in waist circumference. Decreases in waist circumference are recommended as critically important treatment target for reducing adverse cardiometabolic health risks [15]. Secondary outcomes include optimisation of, fasting blood glucose, total cholesterol, HDL, LDL, triglycerides, body composition, food literacy, PA, and fruit and vegetable intake.

Sample size calculation

Sample size calculation is based on waist circumference.

To calculate the sample size, we used the formula described by Rutterford, Copas [40], Table 3.

Randomization

The six sub parishes (clusters) will be listed alphabetically. A cluster randomization with a 1:1 allocation will then be applied to randomize the sub parishes to either the treatment or control arm. In the sub parishes, women group leaders and participants will be blinded about the study arms.

Data collection

Table 4 gives an overview of the different measurements and time points during the study.

Data analysis

Data will be analysed using R software. To evaluate the effects of the intervention, multilevel analysis will be used. Using this technique, regression coefficients will be adjusted for the clustering of observations within sub parishes. We will define two levels in our multi-level analysis: (1) participant and (2) sub parishes. Linear mixed effect models will be used to examine the effect of the intervention on each of the outcome values. All analyses will be performed according to the intention-to treat-principle [42]. To assess changes in metabolic health between the intervention and control groups, a linear mixed effect model will be built where “time” (end line measurement (M₂) will be compared with base-line measurement (M₁) and post-follow up measurement (M₃)), treatment (and interaction of time and treatment) as well as age will be specified as fixed effects, and sub parishes and participants as random factors. For all linear mixed models, compatibility with mixed-model assumptions will be checked by inspection of residual plots and Q-Q plots. In the case of heteroscedastic residuals, data will be log transformed. Tukey or Benjamini–Hochberg procedures will be applied when performing post hoc analyses to further identify differences within treatments as well as between time points. Statistical outliers will be defined as any observation which has an absolute residual exceeding 3 times the residual standard deviation. p < 0.05 will be considered significant in all analyses.