Food production diversity based on food groups is associated with minimum dietary diversity and dietary diversity among lactating mothers residing in subsistence farming households in rural Kyankwanzi district, Uganda

Study design and setting

This HH cross-sectional quantitative study was conducted in the rural Kyankwanzi district, central Uganda, an area where over 90% of food for HH consumption is produced through subsistence farming [24, 25]. Moreover, 50% of the WRA are either pregnant or lactating mothers [24]. Data collection was done during the crop harvesting season between December 2019 and March 2020.

Study participants, inclusion, and exclusion criteria

The study participants were all mothers who had infants and young children (IYC) aged 0 to 23 months in the study area during the data collection period. A woman was eligible for inclusion in the study if she was a WRA (15–49 years old) and breastfeeding. Study participants who resided in HHs without a subsistence farm, those who had mental disorders, were unable to hear or speak, and were not willing to participate in the study were excluded from the study.

Pilot study

A pilot study was conducted to inform the sample size calculation for the main study [26]. This was necessary because we found no previous studies that determined the prevalence of MFGPD and its association with MDD-W. The pilot study was conducted in the rural Kiboga district, a neighboring district to Kyankwanzi, where the main study was conducted [27]. The pilot study recruited 123 lactating mothers with infants and young children, 0–23 months old. The lactating mothers were residing in HHs with subsistence farms. The FGPD score is a sum of the number of different food groups produced by each farm. Therefore, the food groups used to classify foods on the farm were the same as those of the 10 food group classifications used in the MDD-W indicator [13]. To this end, a lactating mother was considered to have achieved MFGPD if she resided in a HH that had a farm with at least 5 out of the 10 food groups. A lactating mother was considered to have achieved MDD-W if she had consumed at least 5 out of the 10 food groups in the previous 24 h. Results from the pilot study showed that the proportion of lactating mothers who achieved MDD-W in a group that had achieved MFGPD was 19%. In contrast, the proportion of mothers who achieved MDD-W in a group that did not achieve MFGPD was 11%. These proportions were used to calculate the sample size of the main study. This pilot study was also used to train four enumerators on how to collect data concerning FPD, MFGPD, MDD-W, and WDD. The enumerators had obtained a Bachelor of Science in Human Nutrition and were experienced in collecting dietary diversity data.

Sample size determination

We used a two proportion formula,

to calculate sample size because this study tested associations between exposure variables and a binary outcome, and proportions were determined by a pilot study conducted in a neighboring district [28].

Where N = required minimum sample size.

P₀ = Proportion of mothers who achieved MDD in a group that has MFGPD = 19%.

P₁ = Proportion of mothers who achieve MDD in a group that does not achieve MFGPD = 11%.

Z_α is a standard normal value corresponding to a level of significance of 5%= 1.96.

Z_β is the standard normal value corresponding to the power of the study (80%) = 0.84. Values for P₀ and P₁ were based on the pilot study results.

After multiplying this by a design effect of 2, the final sample size was 616. Considering a non-response rate of 8.4%, 668 participants were required in the study. We calculated the design effect based on the pilot study data as the ratio of the variance of an estimate under a sampling plan to the variance of the same estimate from a simple random sample with the same number of observations. The non-response rate was also calculated based on the pilot study.

Sampling procedure for study participants

The lactating mothers were recruited by systematic sampling method. The district was divided into its administrative units of 15 sub-counties and six town councils to give a total of 21 administrative units (strata). The village health team members (VHTs) listed all HHs in each administrative unit with a subsistence farm and mothers having IYC, 0 to 23 months old. The number of lactating mothers recruited per administrative unit was based on the proportion to the size of the administrative unit. On average, 30 lactating mothers were listed per day per administrative unit. Therefore, we systematically recruited 10 of 30 lactating mothers per day per administrative unit. The following procedure was followed: Every morning, a list of all the first 30 participants was obtained. To this end, 30/10 = 3. Therefore, every third lactating mother was recruited. A number from 1 to 3 was chosen at random as a starting point for the recruitment of study participants. In this case, number 1 was chosen at random. Therefore, every 3rd participant was recruited into the study per administrative unit until the calculated sample size was achieved. That is to say, participants 1,4,7, 10, 13,16,19,22,25, and 28 were selected from the list to participate in the study. In cases where the HH had more than one lactating mother, then one lactating mother was picked by computer-generated simple random sampling to participate in the study.

Measurement of outcome variables

The outcome variables of this study were women’s dietary diversity (WDD) and minimum dietary diversity for women (MDD-W). A 24-hour dietary recall questionnaire was completed for each participant in a face-to-face interview. This was guided by the Food and Agriculture Organization of the United Nations (FAO) non-quantitative open 24-hour recall method used to collect data on WDD and MDD-W [13]. The trained enumerators asked a series of standard probing questions to help the lactating mothers recall all foods and beverages consumed the previous day and night and to probe for the main ingredients in mixed dishes. Thereafter, the enumerator determined to which food groups these foods belonged. According to the FAO guideline, all the food items were categorized into 10 food groups, including (i) Grains, white roots and tubers, and plantains; (ii) Pulses (beans, peas, and lentils); (iii) Nuts and seeds; (iv) Milk and milk products; (v) Meat, poultry, and fish; (vi) Eggs; (vii) Dark green leafy vegetables; (viii) Vitamin A-rich fruits and vegetables; (ix) Other vegetables; (x) Other fruits. A participant was scored either “1” if she consumed the food group or “0” if she did not consume the food group in the previous 24 h. The lactating mothers’ WDD was calculated as the total of all food group scores. The lactating mother was considered to have achieved MDD-W if she consumed at least 5 out of the 10 food groups in the previous 24 h [13].

Measurement of exposure variables

The main exposure variables were minimum farm group production diversity (MFGPD) and food group production diversity (FGPD). This study opted to use the count of “food groups” and not “food species” because, as noted above, nutritionists are expected to promote achieving MDD-W based on food groups, not food species [13, 19]. During the survey, lactating mothers were asked to report details of their current farm production. This was confirmed by the enumerators through observation of the farms. The FGPD score was a sum of the number of different food groups produced by each farm. Therefore, the food groups used to classify foods on the farm were the same as those of the 10 food group classifications used in the MDD-W indicator [13]. The lactating mother was considered to have achieved MFGPD if her farm HH had at least 5 out of the 10 food groups. The FGPD score did not take into account condiments and seasonings such as garlic, onions, and hot pepper, among others, because they are not part of the food groups for assessing either WDD or MDD-W among WRA [13].

Assessment of confounding variables

A structured questionnaire was used to collect data on confounding variables. The lactating mothers were asked to report on the socio-economic and demographic confounders such as the sex of the HH head, the mother’s age, the mother’s education, off-farm income, nutrition education to the mother, purchase of food for consumption, and sale of food produced from the HH farm because previous studies have demonstrated that they have a potential to influence dietary diversity [20,21,22, 29,30,31,32]. Categorization for: Sex of HH head was into male and female; education into no formal education, primary, secondary and tertiary education; off-farm income into yes and no; nutrition education on either WDD and MDD-W into yes and no; sale of food into yes and no; purchase of food into yes and no; lactating mothers age was categorized into adolescent (yes and no). The lactating mother was considered an adolescent and non-adolescent if she was aged 15–19 and 20–49 years, respectively [33]. Because of their higher growth spurt, adolescent mothers have higher nutritional demands compared to older non-adolescent mothers [34]; therefore, it was important to control for the age of respondents based on adolescence. Furthermore, the age of the breastfeeding child was considered as a cofounder because the infant and young child feeding guidelines recommend mothers breastfeed their children 0 to 23 months old [35]. The 0 to 23 age range was further categorized into the period of exclusive breastfeeding and complementary feeding, which target mothers with children 0 to 5 months old and 6 to 23 months old, respectively [35].

Statistical data analysis

We analyzed the data using STATA, version 15.0. A statistical test was considered significant at 95% confidence interval (CI) and probability value, P less than 0.05 (P < 0.05). The association between the WDD score and FGPD score was analyzed by bivariate and multivariable Poisson regression analysis [36]. The unadjusted and adjusted correlation coefficients (β) were reported in the former and latter, respectively.

We created a binary outcome for MDD-W (yes and no). The association between MDD-W and independent variables was analyzed by bivariate and multivariable logistic regression analysis. Crude odds ratios (COR) and adjusted odds ratios (AOR) were reported in the former and latter, respectively. Differences in the prevalence of MFGPD and MDD-W among HHs where lactating mothers resided were analyzed using McNemar’s test.

Ethical approval

This study is part of the caregiver acceptability study of a complementary food prepared from pumpkin and common bean in Uganda [37]. Permission to conduct the study was granted by the District Health Office, Kyankwanzi district, Uganda. Ethical approval was granted by The AIDS Support Organisation Research Ethical Committee (Reference number: TASO-REC/066/19-UG-REC-009) under the study “Nutrition and sensory properties of a complementary food prepared from pumpkin and common bean in Uganda.” Written and signed consent were obtained individually from the lactating mothers who participated in the study. Written/signed consent was taken from legally authorized representatives and/or guardians of all lactating mothers who were below 18 years old and those without formal education.