The epidemiological transition is accompanied by the nutrition transition towards foods rich in saturated fat and sugar( Reference Olshansky and Ault 1 , Reference Drewnowski and Popkin 2 ), which, combined with lower levels of physical activity and sedentary behaviour( Reference Popkin 3 ), lead to a rise in overweight, obesity and nutrition-related non-communicable diseases (NR-NCD) including type 2 diabetes and CVD( Reference Reddy 4 ). These transitions occur more rapidly in low- and middle-income countries( Reference Abrahams, McHiza and Steyn 5 ) and consequently the pace of the rise in obesity and NR-NCD in such countries appears faster( Reference Popkin 6 ). Over 80 % of NR-NCD deaths now occur in low- and middle-income countries( 7 ) where, notably, obesity is expected to increase proportionally by a larger amount compared with high-income countries( Reference Kelly, Yang and Chen 8 ).
Obesity/overweight prevalence is higher among women in Africa( Reference Kelly, Yang and Chen 8 ); almost twice as high as in men( 9 ). Urbanization has been cited as an important driver for NR-NCD and overweight/obesity increases in women( Reference Popkin, Adair and Ng 10 , Reference Neuman, Kawachi and Gortmaker 11 ), but factors other than urban residence seem to contribute towards overweight and obesity, particularly socio-economic status (SES)( Reference Neuman, Kawachi and Gortmaker 11 ). The gender disparity in overweight/obesity, in addition to that in urban v. rural settings, explains the present review’s specific focus on urban African women.
The WHO’s response to the rising burden of NR-NCD includes global prevention strategies( 12 , 13 ) focusing on modifiable behaviours and social determinants through creating healthier environments. Some research exists into the various factors driving obesity in Africa( Reference Maletnlema 14 – Reference Vorster, Venter and Wissing 16 ) but the available research hails predominantly from high-income countries( Reference Reddy 4 ) and greater insight into the specific determinants of dietary behaviour would guide culturally appropriate and hence more effective interventions.
An ‘ecological’ framework of the drivers of food choice has been developed to visually assemble the various factors into four levels( Reference Story, Kaphingst and Robinson-O’Brien 17 ). The individual level includes demographics, behaviours and cognitions (e.g. attitudes, preferences, knowledge and values), which affect food choice via characteristics including self-efficacy, motivation and behavioural capacity. Environmental factors are separated into social (family and peers), physical (where food is consumed and procured) and macro (social norms, marketing, etc.) environmental levels. This visual map serves to illustrate the relationships between the various levels and, notably, between people and their environment. This model complements the vast discourse surrounding the social determinants of health inequalities; the so-called ‘causes of the causes’( Reference Marmot 18 ). It of particular interest to the present review as it focuses on the relationship between levels (and the factors therein) and subsequent eating behaviours, as well as possible areas for effective policy intervention( Reference Story, Kaphingst and Robinson-O’Brien 17 ); however, most of the source data hail from high-income countries( Reference Story, Kaphingst and Robinson-O’Brien 17 ).
The study objective was to determine the factors influencing dietary behaviour among women living in urban Africa, map them visually, and identify any gaps and/or priority areas for future research.
Methods
This research topic is novel for this population; therefore a mapping review is ideal for conceptualizing the determinants in the literature and identifying neglected areas( Reference Grant and Booth 19 ). Systematic searches of the literature were conducted according to a predefined protocol published on PROSPERO( Reference Gissing, Holdsworth and Saeed 20 ). Preliminary scoping searches yielded studies with dietary diversity as the outcome, which is desirable as research has determined dietary diversity to be a key component of healthy diets( Reference Ruel 21 ). The SPIDER (Sample, Phenomenon of Interest, Design, Evaluation, Research type) tool was utilized to develop the eligibility criteria (see online supplementary material, Supplemental Table 1) and the review question as it is more suited to qualitative and mixed-methods research( Reference Cooke, Smith and Booth 22 ). The eligible population comprised women aged ≥18 and <70 years living in urban Africa. Language was limited to English and French. No publication type limit was applied to yield an adequate representation of the relevant available literature. Following scoping searches, the date was limited to 1971 onwards as studies concerning health behaviour in the context of the epidemiological transition began at that time( Reference Omran 23 ). The search strategy (see Supplemental Table 2) was then used on six online databases in April 2015: EMBASE, MEDLINE, PsychINFO, CINAHL, ASSIA and African Index Medicus. The WHO International Clinical Trials Registry (ICTRP) was searched for ongoing trials and the University of Sheffield Library Catalogue (STARPlus) for relevant theses. The reference lists of included literature and records citing them were also searched alongside data extraction and quality assessment.
Records yielded by the search strategy underwent duplicate removal, title and abstract screening, and full-text screening by a single reviewer; 10 % of excluded records at the title/abstract and full-text screening stages were checked for adherence to the protocol by another reviewer. Complete concordance between reviewers was apparent at both checking stages. Reasons for exclusion at the full-text stage were recorded.
Included studies also underwent quality assessment to increase internal validity( Reference Kmet, Lee and Cook 24 ). This process was undertaken by two reviewers independently, compared, and differences resolved via discussion. The Standard Quality Assessment Criteria for Evaluating Primary Research Papers From a Variety of Fields was used because it is applicable to both the quantitative and qualitative studies yielded in the present review, having a separate checklist and scoring process for each( Reference Kmet, Lee and Cook 24 ). All checklist items had a defined ‘yes’, ‘no’, ‘partial’ or ‘N/A’ grading( Reference Kmet, Lee and Cook 24 ).
Data were extracted from included studies initially by a single reviewer with a standardized form which was initially piloted and modified appropriately. Extracted data comprised that presented in Tables 1 and 2, with the addition of a column for ‘effect of determinant(s) on dietary behaviour’ and how each determinant and dietary behaviour was measured (the latter being reported in the online supplementary material, Supplemental Tables 3–6). Income level was categorized according to the World Bank( 25 ), and the level of each determinant was guided by the ecological framework( Reference Story, Kaphingst and Robinson-O’Brien 17 ). Extracted data were then checked by another reviewer as double data extraction is more valid( Reference Buscemi, Hartling and Vandermeer 26 ). Extracted data informed the iterative synthesis of a visual ‘map’ of drivers of dietary behaviour to gain a sense of the nature of research available within the topic area. Drivers identified by the review were matched where appropriate to the aforementioned ecological model( Reference Story, Kaphingst and Robinson-O’Brien 17 ), which was adapted to account for any novel drivers. Reporting of the review followed the PRISMA (Preferred Reporting Items for Systematic reviews and Meta-Analyses) checklist where applicable( Reference Moher, Liberati and Tetzlaff 27 ).
Table 1 Characteristics of included studies
* QAS, quality assessment score.
Table 2 Determinants and dietary behaviours measured in the included studies
Results
Search results
The search strategy yielded 4722 title and abstract records after duplicates were removed (Fig. 1). One hundred and twenty-one records remained for full-text retrieval, at which stage 108 records were excluded. The resulting twelve included studies were reported in seventeen records( Reference Batnitzky 28 – Reference Waswa 44 ).
Fig. 1 PRISMA (Preferred Reporting Items for Systematic reviews and Meta-Analyses) flow diagram showing the selection of studies for the present systematic mapping review
Description of studies
Table 1 presents the characteristics of the twelve included studies. Two were qualitative( Reference Batnitzky 28 , Reference Rguibi and Belahsen 38 ), one used mixed methods( Reference Charlton, Brewitt and Bourne 29 ) and the remaining nine were quantitative( Reference Charlton, Brewitt and Bourne 29 – Reference Mbochi, Kuria and Kimiywe 33 , Reference Hattingh, Walsh and Bester 35 – Reference Hattingh, Le Roux and Nel 37 , Reference Savy, Martin-Prével and Danel 39 – Reference Waswa 44 ). One abstract was included as sufficient data were obtained by an author( Reference Landais 31 , Reference Landais 45 ). Two studies were set in low-income countries: Burkina Faso( Reference Savy, Martin-Prével and Danel 39 , Reference Becquey, Savy and Danel 40 ) and Benin( Reference Sodjinou, Agueh and Fayomi 42 ). Seven were from lower-middle-income countries; four from Morocco( Reference Batnitzky 28 , Reference Jafri, Jabari and Dahhak 30 – Reference Landais, Bour and Gartner 32 , Reference Rguibi and Belahsen 38 ), two were Kenyan( Reference Mbochi, Kuria and Kimiywe 33 , Reference Waswa 44 ) and one Ghanaian( Reference Mogre, Atibilla and Kandoh 34 ). Three studies were situated in South Africa; an upper-middle-income country( Reference Charlton, Brewitt and Bourne 29 , Reference Hattingh, Walsh and Bester 35 – Reference Hattingh, Le Roux and Nel 37 , Reference Van Zyl, Steyn and Marais 43 ).
Quality assessment
Nine of the twelve studies achieved scores ≥75 % (Table 1), indicating that the overall quality of the studies was fairly high. Strengths comprised sufficient description of the study question/objective, sample selection/characteristics, data analysis and appropriate design/sample size. The main weaknesses were lack of adjustment for confounders and insufficient description of sampling strategy, data collection and analysis.
Data synthesis
The specific determinants measured by the twelve studies span all levels of the existing ecological model (see Table 2 and Fig. 2). Components of dietary behaviour included meal content( Reference Batnitzky 28 ), food consumption( Reference Jafri, Jabari and Dahhak 30 – Reference Mbochi, Kuria and Kimiywe 33 , Reference Van Zyl, Steyn and Marais 43 ), micro/macronutrient intake( Reference Mbochi, Kuria and Kimiywe 33 , Reference Hattingh, Walsh and Bester 35 – Reference Hattingh, Le Roux and Nel 37 , Reference Waswa 44 ), dietary diversity( Reference Landais 31 , Reference Landais, Bour and Gartner 32 , Reference Becquey, Savy and Danel 40 ), diet quality( Reference Batnitzky 28 , Reference Sodjinou, Agueh and Fayomi 41 , Reference Sodjinou, Agueh and Fayomi 42 ) food choice( Reference Charlton, Brewitt and Bourne 29 ), eating patterns( Reference Mogre, Atibilla and Kandoh 34 , Reference Becquey, Savy and Danel 40 ), adequacy of food intake( Reference Waswa 44 ) and diet practices( Reference Rguibi and Belahsen 38 ) (see online supplementary material, Supplemental Tables 3–6).
Fig. 2 Map of determinants of diet/dietary behaviour in women living in urban Africa (adapted from Story et al.( Reference Story, Kaphingst and Robinson-O’Brien 17 )). Underlined determinants are those for which statistically significant results were reported; in parentheses are the numbers of studies finding these determinants significant
Individual- and household-level determinants
Nine studies( Reference Charlton, Brewitt and Bourne 29 – Reference Hattingh, Le Roux and Nel 37 , Reference Savy, Martin-Prével and Danel 39 – Reference Sodjinou, Agueh and Fayomi 42 , Reference Waswa 44 ) reported the effects of individual- or household-level determinants (online supplementary material, Supplemental Table 3).
Cognitions
Six studies looked at the cognitive element of individual-level factors( Reference Batnitzky 28 , Reference Landais 31 , Reference Mogre, Atibilla and Kandoh 34 , Reference Savy, Martin-Prével and Danel 39 , Reference Becquey, Savy and Danel 40 , Reference Waswa 44 ). One study found that individual perception of taste was rated as having a ‘large influence’ as opposed to a ‘small influence’ or ‘no influence’ on food choice by 75 % of participants( Reference Charlton, Brewitt and Bourne 29 ). Another study found that lack of appetite and not feeling hungry were among the most common reasons for skipping breakfast( Reference Mogre, Atibilla and Kandoh 34 ). Hunger and mood were not associated with adequacy of food intake in another study( Reference Waswa 44 ). Perception of body image did not influence adequacy of food intake( Reference Waswa 44 ), except women with a positive body image reported an adequate food intake and consumed more protein than women with a negative body image( Reference Waswa 44 ). One comparatively high-quality study (95 %) looked at perceptions of quality and quantity of diet and found that dietary diversity was significantly higher in those who rated the quantity of their diet as ‘sufficient’ v. ‘insufficient’ and dietary diversity was significantly higher in those who rated diet quality as ‘rather varied’ v. ‘not varied enough’( Reference Savy, Martin-Prével and Danel 39 , Reference Becquey, Savy and Danel 40 ). One study found an association between knowledge about fruit and vegetables and diversity of consumption of both (those with low knowledge ate a mean of 181 g of fruit and vegetables daily, compared with 255 g/d in those with high knowledge, P<0·01)( Reference Landais 31 ). A second study of comparatively lower quality (87 % v. 100 %) found that those with better knowledge of food choice were twice as likely to have inadequate protein intake (OR=2·37, 95 % CI 1·23, 4·53)( Reference Waswa 44 ).
Lifestyle
Seven studies looked at individual lifestyle as determinants of dietary behaviour( Reference Charlton, Brewitt and Bourne 29 , Reference Landais, Bour and Gartner 32 , Reference Mogre, Atibilla and Kandoh 34 – Reference Hattingh, Le Roux and Nel 37 , Reference Savy, Martin-Prével and Danel 39 – Reference Sodjinou, Agueh and Fayomi 42 , Reference Waswa 44 ). Skipping at least one daily meal was associated with eating fewer vegetables (152 g/d v. 187 g/d in those not skipping meals, P<0·01) and fruit and vegetables (305 g/d v. 343 g/d in those not skipping meals, P<0·05)( Reference Landais, Bour and Gartner 32 ). ‘Lack of time’ was the commonest reason for skipping breakfast( Reference Mogre, Atibilla and Kandoh 34 ). Those who skipped meals in another study were found to be three times more likely to have inadequate energy intake than those who did not (OR=3·12, 95 % CI 1·21, 8·06)( Reference Waswa 44 ). Snacking often was associated with higher dietary diversity compared with those who never snacked in Burkina Faso (P=0·01)( Reference Savy, Martin-Prével and Danel 39 , Reference Becquey, Savy and Danel 40 ). One study found that a majority of women rated ‘trying to eat a healthy diet’ and ‘habit or routine’ as a ‘large influence’ on food choice, and ‘slimming foods’ was given ‘low influence’ by a majority( Reference Charlton, Brewitt and Bourne 29 ). There was no association between physical activity level and energy intake in one study( Reference Hattingh, Walsh and Bester 35 – Reference Hattingh, Le Roux and Nel 37 ) and another also found no association between physical activity and diet quality( Reference Sodjinou, Agueh and Fayomi 41 , Reference Sodjinou, Agueh and Fayomi 42 ). Eating from a shared bowl had no effect on fruit and vegetable consumption( Reference Landais, Bour and Gartner 32 ). Smoking was not associated with diet quality in one study( Reference Sodjinou, Agueh and Fayomi 41 , Reference Sodjinou, Agueh and Fayomi 42 ); however, lower alcohol consumption and healthier overall lifestyle (less alcohol/smoking and more physical activity) were significantly associated with higher diet quality( Reference Sodjinou, Agueh and Fayomi 41 , Reference Sodjinou, Agueh and Fayomi 42 ).
Biology
Biological factors were investigated in five studies( Reference Jafri, Jabari and Dahhak 30 , Reference Landais, Bour and Gartner 32 , Reference Mogre, Atibilla and Kandoh 34 – Reference Hattingh, Le Roux and Nel 37 , Reference Savy, Martin-Prével and Danel 39 , Reference Becquey, Savy and Danel 40 , Reference Waswa 44 ). One study found no differences between age groups for energy intake( Reference Hattingh, Walsh and Bester 35 – Reference Hattingh, Le Roux and Nel 37 ) and another found no age differences for fruit and vegetable diversity or consumption( Reference Landais, Bour and Gartner 32 ). However, dietary diversity was reported to be higher in younger women (<25 years v. ≥50 years, P<0·05)( Reference Savy, Martin-Prével and Danel 39 , Reference Becquey, Savy and Danel 40 ). The prevalence of fattening product consumption was also higher in younger women (17·6 % in <25 years v. 7·4 % in those >55 years, P<0·05)( Reference Jafri, Jabari and Dahhak 30 ). Overall health had no influence on adequacy of self-reported food intake in one study( Reference Waswa 44 ).
Socio-economic status
Three studies looked at socio-economic factors as determinants of food choice( Reference Landais 31 – Reference Landais, Bour and Gartner 32 , Reference Savy, Martin-Prével and Danel 39 , Reference Becquey, Savy and Danel 40 ). Two studies looked at SES: one found that higher status was significantly associated with higher fruit and vegetable consumption and diversity compared with lower status( Reference Landais 31 , Reference Landais, Bour and Gartner 32 ); a second found that higher SES was significantly associated with higher dietary diversity score( Reference Savy, Martin-Prével and Danel 39 , Reference Becquey, Savy and Danel 40 ). One Kenyan study indicated that women of higher SES may have unhealthier diets as they consumed significantly more energy-dense foods, cholesterol and alcohol( Reference Mbochi, Kuria and Kimiywe 33 ). There was no association between employment and dietary diversity or consumption( Reference Jafri, Jabari and Dahhak 30 , Reference Landais, Bour and Gartner 32 , Reference Savy, Martin-Prével and Danel 39 , Reference Becquey, Savy and Danel 40 ). Three studies looking at education were inconclusive as to whether education had a positive impact on diet( Reference Jafri, Jabari and Dahhak 30 , Reference Savy, Martin-Prével and Danel 39 , Reference Becquey, Savy and Danel 40 ). Two of these studies found lower dietary diversity with low education level( Reference Savy, Martin-Prével and Danel 39 , Reference Becquey, Savy and Danel 40 ); however, a third study reported no association between education level and fruit and vegetable consumption, but that the effect of education was modified by economic level, and in the lowest economic group the highest educated had higher food diversity scores (β=0·59 (se 0·08) for no education v. β=1·05 (se 0·14) for some education, P<0·01)( Reference Landais 31 ). One study from Burkina Faso found that dietary diversity was higher with a greater monthly household food expenditure (β=6·0 (se 0·14) for no expenditure v. β=7·0 (se 0·25) for expenditure >30 000 CFA francs, P<0·01)( Reference Savy, Martin-Prével and Danel 39 , Reference Becquey, Savy and Danel 40 ).
Social-level determinants
Seven of the included studies investigated social-level determinants( Reference Charlton, Brewitt and Bourne 29 – Reference Landais 31 , Reference Mbochi, Kuria and Kimiywe 33 , Reference Savy, Martin-Prével and Danel 39 – Reference Sodjinou, Agueh and Fayomi 41 , Reference Landais 45 ) (online supplementary material, Supplemental Table 4). Four studies investigated the relationship between marital status and dietary behaviour( Reference Jafri, Jabari and Dahhak 30 , Reference Landais, Bour and Gartner 32 , Reference Rguibi and Belahsen 38 – Reference Becquey, Savy and Danel 40 ). One found no difference in marital status and use of fattening products( Reference Jafri, Jabari and Dahhak 30 ) and a second found no association with fruit and vegetable consumption or diversity( Reference Landais, Bour and Gartner 32 ). Conversely, another study found that dietary diversity score was higher for single or married women compared with widowed/divorced women (6·7 for single, 6·3 for married and 5·8 for widowed/divorced women, P=0·01)( Reference Savy, Martin-Prével and Danel 39 , Reference Becquey, Savy and Danel 40 ). A qualitative study reported a fattening practice undertaken by Moroccan women of the Saharawi culture in preparation for marriage( Reference Rguibi and Belahsen 38 ). Another qualitative study found that married women ate diets of lower nutritional value than men( Reference Batnitzky 28 ). Women reported to snack during food preparation, which the authors used to suggest how household social roles may influence energy intake( Reference Batnitzky 28 ). In addition, that study looked at household composition, reporting that women were served in order of age and those with the highest status (e.g. mother-in-law) would get the largest piece of meat( Reference Batnitzky 28 ).
One study reported that ‘what the rest of my family will eat’ was rated by a majority as having a ‘large influence’ on food choice( Reference Charlton, Brewitt and Bourne 29 ). Number of children was not associated with fruit and vegetable consumption or diversity in another study( Reference Landais, Bour and Gartner 32 ). One study created a ‘Care for Women’ index for the attention and support women receive from other household members, but this was not associated with dietary diversity( Reference Savy, Martin-Prével and Danel 39 , Reference Becquey, Savy and Danel 40 ). Another study found no association between parental influence on food choice and self-reporting of food intake as ‘adequate’ or ‘inadequate’( Reference Waswa 44 ). One study reported that women who indicated that peers had a ‘large influence’ on their food intake were almost seven times more likely to self-report that their food intake was ‘inadequate’ instead of ‘adequate’ (OR=6·54, 95 % CI 3·08, 13·89)( Reference Waswa 44 ).
Physical-level determinants
Three studies investigated the association between aspects of the physical environment and dietary behaviour( Reference Landais, Bour and Gartner 32 , Reference Savy, Martin-Prével and Danel 39 , Reference Becquey, Savy and Danel 40 , Reference Waswa 44 ) (online supplementary material, Supplemental Table 5). Household sanitation was found to have no association with dietary diversity( Reference Savy, Martin-Prével and Danel 39 , Reference Becquey, Savy and Danel 40 ). One study looked at living area and found no association with fruit and vegetable intake( Reference Landais, Bour and Gartner 32 ). No relationship was found between the influence of food availability on food choice and whether food intake was self-reported as ‘adequate’ or ‘inadequate’( Reference Waswa 44 ).
One study found that women eating outside the home more often were significantly more likely to consume fewer vegetables( Reference Landais, Bour and Gartner 32 ). However, another study found that eating outside the home was associated with higher dietary diversity compared with never eating outside the home (6·6 v. 6·0; P<0·01)( Reference Savy, Martin-Prével and Danel 39 , Reference Becquey, Savy and Danel 40 ).
Macro-level determinants
Six studies examined macro-level determinants of dietary behaviour( Reference Charlton, Brewitt and Bourne 29 , Reference Mogre, Atibilla and Kandoh 34 , Reference Rguibi and Belahsen 38 – Reference Becquey, Savy and Danel 40 , Reference Van Zyl, Steyn and Marais 43 , Reference Waswa 44 ) (online supplementary material, Supplemental Table 6). Two studies investigated cultural beliefs; one of which found that women stating that cultural beliefs affected their food intake were three times more likely to report an ‘inadequate’ food intake (OR=3·07, 95 % CI 1·18, 7·99)( Reference Waswa 44 ). A second study found that two-thirds of Saharawi women had used fattening practices and 71·4 % reported overfeeding at the time of the study( Reference Rguibi and Belahsen 38 ). The researchers cited an association of beauty with overweight as the motivation, as well as the cultural acceptance of unhealthy fattening practices( Reference Rguibi and Belahsen 38 ).
In one study, culture and religion were said to have ‘no influence’ on food choice by a majority( Reference Charlton, Brewitt and Bourne 29 ). Of the three studies which looked at religion, one reported that ‘religious reasons’ had no association with skipping breakfast( Reference Mogre, Atibilla and Kandoh 34 ) or with dietary diversity in another study( Reference Savy, Martin-Prével and Danel 39 , Reference Becquey, Savy and Danel 40 ). However, those who said that religious beliefs affected their food choice in a third study were more likely to report their food intake as ‘inadequate’ compared with ‘adequate’ (OR=0·09, 95 % CI 0·03, 0·24)( Reference Waswa 44 ).
Two studies investigating food prices had contradictory findings: one found that price had no influence on whether food intake was self-reported as ‘adequate’ or ‘inadequate’( Reference Waswa 44 ), and one reported that a majority said price had a ‘large influence’ on food choice( Reference Savy, Martin-Prével and Danel 39 ).
One study investigating ‘quality or freshness of food’ and ‘quick and easy to make foods’ reported that most women said both had a ‘large influence’ on food choice( Reference Charlton, Brewitt and Bourne 29 ). The same study also found that presentation or packaging had a reportedly ‘small influence’ on food choice( Reference Charlton, Brewitt and Bourne 29 ).
Two studies looked at the role of media( Reference Van Zyl, Steyn and Marais 43 , Reference Waswa 44 ). One found that those who cited the media as affecting their food choice were more likely to report their food intake as ‘inadequate’ (OR=0·12, 95 % CI 0·04, 0·32)( Reference Waswa 44 ). One study which examined specific types of media found that 20·6 % of women said adverts on billboards, television, radio and flyers ‘always’ result in them buying fast food( Reference Van Zyl, Steyn and Marais 43 ); and 83·2 % said television adverts encouraged them to buy fast food( Reference Van Zyl, Steyn and Marais 43 ).
Map of determinants
A visual map was constructed of all determinants of dietary behaviour in women living in urban Africa identified by the present review (Fig. 2), adapted from the ecological model of food environments which illustrates the interaction of the levels and the factors therein( Reference Story, Kaphingst and Robinson-O’Brien 17 ). The present review identified additional factors not specifically mentioned in the model including mood, health, education and religion, representing potentially important areas for future research to prevent unhealthy dietary behaviours. In addition, there were also factors in the ecological model not identified by the current review presenting possible areas for future research: the influence of friends; physical environments such as worksites, fast-food outlets, childcare or stores; and macro-level determinants such as food/agriculture policy, government/politics, land use and transportation.
Discussion
The present systematic mapping review has identified and visually mapped determinants of dietary behaviour in women living in urban Africa from twelve included studies and six countries( Reference Batnitzky 28 , Reference Charlton, Brewitt and Bourne 29 , Reference Landais 31 – Reference Waswa 44 ). The majority of available research related to the individual and household levels. The primarily cross-sectional nature of the included studies makes determining causality challenging. However, some clear associations have been found between ‘determinants’ measured and certain dietary behaviours.
There was substantial heterogeneity between the settings, participants’ age, determinants investigated and their measurement, and dietary behaviour assessed. Due to the wide age range of participants, it is reasonable to generalize the results to a similar age range of adult urban African women as focused on here (18–70 years). All income levels in Africa are represented by the included studies, giving the findings more generalizability. In terms of country, South Africa and Morocco were well represented in the available research and hence the findings can be reasonably applied to women in these contexts. However, generalizations to other African nations may be more challenging.
Notably, some study findings are based on self-reported beliefs and this subjectivity could have been influenced by interviewer bias, hence reducing the internal validity of the present review’s findings.
It is important to note that the lack of research into determinants of dietary behaviour in this population makes significant evaluations with existing research challenging. The vast heterogeneity of determinants and tools used for measuring behaviour identified in the review furthers the difficulty in forming meaningful conclusions concerning specific determinants and their effects on particular dietary behaviours. However, ten of the seventeen determinants of dietary behaviour in women living in urban Africa are similar to those found in high-income countries( Reference Story, Kaphingst and Robinson-O’Brien 17 ): perceptions, food knowledge, education, food expenditure, lifestyle, age, family (marital status), eating environments outside the home, culture and the media. It is therefore reasonable to suggest that many determinants of dietary behaviour are perhaps similar between urban Africa and high-income countries.
There is limited evidence available concerning some specific determinants influencing dietary behaviour identified in the current review in similar populations. The review found that higher SES was associated with greater consumption of ‘Westernized’ food (e.g. more fat- and sugar-rich and fewer traditional foods( Reference Mbochi, Kuria and Kimiywe 33 )), as well as increased diversity and fruit and vegetable consumption( Reference Landais, Bour and Gartner 32 , Reference Savy, Martin-Prével and Danel 39 , Reference Becquey, Savy and Danel 40 ). Interestingly, the review included a study reporting that knowledge had a relationship with inadequate protein intake, suggesting that food knowledge may actually be detrimental( Reference Waswa 44 ). However, the sample used was specifically students at a single university aged 20–25 years who may not be representative of the target population of urban-dwelling adult women( Reference Landais 45 ). Moreover, a second study found that education positively influenced dietary diversity( Reference Savy, Martin-Prével and Danel 39 , Reference Becquey, Savy and Danel 40 ) and consumption of fattening products( Reference Jafri, Jabari and Dahhak 30 ), and another found that its effect was modified by economic level( Reference Landais, Bour and Gartner 32 ). Additionally, a fourth study included found a positive relationship between education level as a measure of SES and protein and fat consumption( Reference Abidoye, Izunwa and Akinkuade 46 ). This corroborates the current review’s findings by providing further evidence for the nutrition transition’s occurrence in urban Africa and education/SES as determinants of such dietary changes. Second, included studies concerning food prices were contradictory( Reference Charlton, Brewitt and Bourne 29 , Reference Waswa 44 ); however, a study found a significant reduction in consumption of foods high in salt and sugar, and an increase in consumption of fruit and vegetables and whole grains in South Africans if healthy foods were cheaper( Reference An, Patel and Segal 47 ). An economical modelling study also reported that a tax on sugar-sweetened beverages in South Africa would reduce consumption and hence reduce obesity( Reference Manyema, Veerman and Chola 48 ). Food price has the potential to influence food consumption and further research into this and other macro-level determinants is needed.
Conclusion
The findings from the present review identified seventeen determinants of diet and related behaviour in African women spanning all food environments, comprising the individual, social, physical and macro levels. Most of the available research in African settings related to individual and household food environments, and included determinants such as perceptions of body image and diet, food knowledge, meal skipping, snacking, alcohol consumption, healthier overall lifestyle, age, SES, education and food expenditure. Marital status, eating outside the home, culture, religion and the media were also found to affect dietary behaviour. There were numerous similarities between the identified determinants in this population and those found in high-income countries, which could represent effective target areas for prevention. The differences found could highlight factors specific to Africa, or represent areas of insufficient or non-existent research. The neglected areas of physical and macro-level environments indicate important areas of future investigation. Therefore, the review’s findings could support context-specific interventions for preventing obesity and NR-NCD in African women.
Acknowledgements
Financial support: This research received no specific grant from any funding agency in the public, commercial or not-for-profit sectors. E.C. is supported by the South African DST/NRF Centre of Excellence in Human Development. Conflicts of interest: None. Authorship: S.C.G. designed the research question with the help of M.H., R.P. and H.A.O.-K. S.C.G. conducted the research and analysed the data with oversight and guidance from M.H., H.A.O.-K., R.P. and E.C. H.A.O.-K. checked 10 % of excluded records at title/abstract and full-text screening stages. M.H., R.P. and E.C. checked data extraction and quality assessment. S.C.G. led the writing of the manuscript with the help of M.H., R.P., E.C. and H.A.O.-K. All authors took responsibility for the final editing, reviewing and approval of the manuscript. Ethics of human subject participation: Not applicable.
Supplementary material
To view supplementary material for this article, please visit https://doi.org/10.1017/S1368980017000970
