Evidence for factors associated with diet and physical activity in African and Caribbean countries

Abstract Objective To identify and describe summarized evidence on factors associated with diet and physical activity in low- and middle-income countries in Africa and the Caribbean by performing a scoping review of reviews. Methods We searched the Medline®, LILACS, Scopus, Global Health and Web of Science databases for reviews of factors associated with diet or physical activity published between 1998 and 2019. At least 25% of studies in reviews had to come from African or Caribbean countries. Factors were categorized using Dahlgren and Whitehead’s social model of health. There was no quality appraisal. Findings We identified 25 reviews: 13 on diet, four on physical activity and eight on both. Eighteen articles were quantitative systematic reviews. In 12 reviews, 25–50% of studies were from Africa or the Caribbean. Only three included evidence from the Caribbean. Together, the 25 reviews included primary evidence published between 1926 and 2018. Little of the summarized evidence concerned associations between international health or political factors and diet or associations between any factor and physical activity across all categories of the social model of health. Conclusion The scoping review found a wide range of factors reported to be associated with diet and physical activity in Africa and the Caribbean, but summarized evidence that could help inform policies encouraging behaviours linked to healthy diets and physical activity in these regions were lacking. Further reviews are needed to inform policy where the evidence exists, and to establish whether additional primary research is needed.


Introduction
Almost three quarters of deaths from noncommunicable disease occur in low-and middle-income countries, particularly in Africa and the Caribbean. 1 Moreover, the burden of noncommunicable disease in the World Health Organization's (WHO) African Region is expected to exceed that of communicable disease by 2030. 2 Premature death from noncommunicable disease in these regions is relatively common; for example, the probability of dying between the ages of 30 and 70 years from noncommunicable disease is 12% in the United Kingdom of Great Britain and Northern Ireland, whereas, in Kenya, Cameroon and South Africa, it is 18%, 20% and 27%, respectively, and in the Caribbean, it ranges from 17% in Jamaica to 37% in Guyana. 1 Studies consistently show that an unhealthy diet and physical inactivity are the leading modifiable behavioural risk factors for the four primary noncommunicable diseases: type 2 diabetes, cardiovascular disease, cancer and chronic respiratory disease. 3 Clear recommendations have been made by WHO for a healthy diet (i.e. high intake of fruit, vegetables and fibre and low intake of fat, sugar and salt) and physical activity (e.g. at least 150 minutes of moderate-intensity activity per week for adults). 4,5 According to 2019 Global Burden of Disease data, 6 the percentage of deaths from noncommunicable disease directly attributable to diet was 15.6% in Africa and 15.3% in the Caribbean; the percentage directly attributable to low physical activity was 2.2% in Africa and 3.7% in the Caribbean.
In both Africa and the Caribbean there are ongoing regional and national policy initiatives on noncommunicable disease, consistent with WHO's Global action plan for the prevention and control of noncommunicable diseases 2013-2020. 4 In Africa, these include the regional 2011 Brazzaville Declaration and national policy initiatives. 7 In the Caribbean, the 2007 Port of Spain Declaration on noncommunicable diseases was a first for lower-middle-income regions. This declaration provided a framework for the development and implementation of policies on the prevention and control of noncommunicable disease, both regionally and nationally. An evaluation of the Port of Spain Declaration in 2018 found that taking effective measures to address the distal (or upstream) determinants of an unhealthy diet and physical inactivity (e.g. cultural and environmental conditions) remained challenging, 8 although new initiatives, such as taxing sugar-sweetened beverages, were being implemented.
Behaviours associated with a healthy diet and physical activity are core contributors to good health and, thus, the ability to participate in these behaviours can be viewed as a universal right. These behaviours are shaped by a range of factors, including: (i) international policies and politics; (ii) socioeconomic, cultural and environmental conditions; (iii) living and working conditions; (iv) social and community networks; and (v) more proximal individual factors (e.g. age and sex). 4,5,9 Evidence on factors associated with these behaviours, on their distribution across different population groups and on whether they are modifiable is important for understanding the drivers of disease burden, for predicting future trends and for identifying targets for interventions and policy changes.
Most existing research summaries on the determinants of diet and physical activity come from high-income countries.
Consequently, the generalizability of their findings to Africa and the Caribbean is questionable and evidence is needed from low-and middle-income countries to inform research, interventions and policy development. 10,11 Scoping reviews adopt a systematic approach to map published evidence on a topic, summarize the main themes and highlight knowledge gaps. 12 We chose to conduct a scoping review of reviews because systematic reviews and meta-analyses provide the highest level of evidence on which to draw evidencebased conclusions.
The principle aim of our study was to identify and summarize existing reviews on a broad range of factors associated with diet and physical activity in low-and middle-income countries in Africa and the Caribbean. A secondary aim was to identify gaps in the current evidence. Our review was conducted as part of an initial scoping exercise for the Global Diet and Activity Research Network, 13 which is a collaboration of researchers in the Caribbean, Cameroon, Kenya, South Africa and the United Kingdom. The overall goal of the network is to generate evidence on the determinants of diet and physical activity to inform noncommunicable disease prevention in Africa and the Caribbean.

Methods
This scoping review of reviews was conducted according to a previously described method 12 and followed reporting guidance in the preferred reporting items for systematic reviews and metaanalyses extension for scoping reviews. 14 A review protocol was developed beforehand and was consistent with the scoping review method. 15 The review question and the study selection criteria were developed iteratively as familiarity with the literature increased.
We searched the Medline®, LILACS, Scopus, Global Health and Web of Science databases for reviews of factors associated with physical activity and dietary behaviour in Africa and the Caribbean that were published between January 1998 and December 2019. A search was carried out in April 2018 and, again, in December 2020 to include literature to the end of 2019. No author was contacted to provide additional information and no grey literature was included because our aim was to identify peer-reviewed evidence syntheses.
Our full search strategy is detailed in Box 1 (available at: http:// www .who .int/ bulletin/ volumes/ 99/ 6/ 20 -269308). In brief, we combined search terms in sets: (i) diet (e.g. diet, nutrition, food intake, fruit, vegetables, fat, sugar, salt and junk food); (ii) physical activity (e.g. walking, manual labour and screen time); (iii) determinants (e.g. risk factors, correlations and demographic factors); (iv) low-and middle-income countries, with specific terms for African and Caribbean countries; and (v) reviews (i.e. reviews of quantitative or qualitative studies).
Reviews were eligible for inclusion if they provided summaries of primary research on factors associated with diet and physical activity and at least 25% of studies included were conducted in lowor middle-income African or Caribbean countries. Reviews could include quantitative or qualitative evidence from observational or interventional studies. We excluded literature reviews that: (i) explored how diet or physical activity shaped health outcomes or disease burden; (ii) reported only health outcomes; (iii) dealt primarily with health-system care models; (iv) focused on migrant groups or ethnic minorities in high-income countries; (v) related to humanitarian crises or natural disasters; (vi) considered only nutritional biomarkers, without an accompanying assessment of diet; (vii) addressed breastfeeding as a determinant of diet in infants; or (viii) were not published in English.
Citations identified by the search strategy were first imported into Rayyan QCRI systematic review software (Qatar Computing Research Institute, Doha, Qatar) and any duplicates were removed. Working in two pairs, we double-screened the titles and abstracts of all citations. If there was a conflict, all authors conferred, with two authors acting as arbiters. Then, the full texts of selected reviews were retrieved and, again working in two pairs, we doublescreened all texts. Any disagreement was discussed with reference to the eligibility criteria, with any one of three authors acting as arbiter.
We examined the final set of selected reviews to determine which data items should be abstracted and their format. The abstraction fields identified included: (i) the citation; (ii) the type of review; (iii) the number and type of studies in the review, including spe-cifically the number and type of studies conducted in Africa and the Caribbean; (iv) the review setting (e.g. urban or rural); (v) the target population group (e.g. children or adults); (vi) factors associated with diet or physical activity; (vii) outcomes; and (viii) the main findings. Three authors entered data into a pretested data abstraction form. Where possible, data and conclusions specifically relevant to African and Caribbean countries were abstracted separately. We used Dahlgren and Whitehead's social model of health to categorize and conceptualize both distal factors (e.g. international policies and politics, and socioeconomic, cultural and environmental conditions) and proximal factors (e.g. living and working conditions, social and community networks and individual factors such as age and sex). 9 The abstracted data were summarized and tabulated.
Our scoping review of reviews describes the results, discussions and conclusions of the selected reviews, not of the primary studies underlying them. Moreover, as is common in scoping reviews, there was no appraisal of the quality of the reviews. Hypothesized or putative explanations for relationships identified in the reviews were included in our summary only if supported by a synthesis of the underlying primary studies.

Outcomes
Dietary outcomes summarized in the reviews included: (i) subsistence skills, such as food gathering, hunting and food preparation; 22 (ii) child feeding complementary to breastfeeding; 25,32,34,40 (iii) school meals or nutrition policies; 26,29 (iv) access to and choice of food; 35 (v) food security; 17,23,37,38 (vi) diet diversity or quality; 17,21 (vii) adherence to a prescribed diet; 27 (viii) calorie or food group consumption (e.g. fruit and vegetables, animal protein or processed food); 21,24,28,33,36,39 and (ix) macro-and micro-nutrient intake. 30,32 Physical activity outcomes included: (i) active travel (e.g. walking or cycling for transport); 19 (ii) total physical activity; (iii) domains of physical activity (e.g. occupational or leisure); 20,26,28 (iv) total sedentary behaviour; (v) domains of sedentary behaviour (e.g. television watching); 20,26 and (vi) physical inactivity (e.g. not meeting physical activity guidelines). 16,18,21,33 Several reviews also reported physical fitness. 20,28 The reviews reported a range of hypothesized and demonstrated relationships between various factors and diet and physical activity. These were categorized using Dahlgren and Whitehead's social model of health ( Table 2). Little of the summarized evidence was related to distal factors in the category of international health, policy and politics in the social model of health and there were relatively few reported associations with physical activity in any category. 9 A wide range of associations were described, particularly for diet ( Table 2). Several reviews reported that the shift to an urban, westernized lifestyle and diet and the threat of a competitive, globalized market were permeating influences. 23,24,29 On diet, reviews that considered factors in the category of international health, policy and politics mentioned: the historic influence of colonization; humanitarian and development aid; the epidemiological transition; the transition to a western lifestyle and diet; the dual burdens of over-and undernutri-     24 Sex; 16,18-20 ethnicity; 16 and age 18,19 HIV/AIDS: human immunodeficiency virus and acquired immunodeficiency syndrome. a Factors reported in reviews as associated with diet or physical activity were categorized using Dahlgren and Whitehead's social model of health. 9 Systematic reviews Diet and physical activity in Africa and the Caribbean Eleanor Turner-Moss et al.
tion; infectious and chronic disease; and the impact of climate change. In addition, associations were described with: socioeconomic, cultural and environmental conditions, including access to food, the availability of food, prices, food security, deagrarianization and urbanization; living and working conditions, including education, poverty, household composition, land rights, skills, assets, rurality, and agricultural and school-based interventions; social and community networks, involving for example social capital, skills acquisition, peer support, key influencers, taboos and norms; and constitutional factors, particularly age and sex. On physical activity, only one review described evidence on determinants in the most distal category of the social model of health (i.e. international health, policy and politics; Table 2). As expected, there were similarities and differences between the associations described for diet and physical activity. For example, both featured urbanization, socioeconomic status and gendered roles. In contrast, certain associations were described only for physical activity: (i) topography and climate; (ii) aspects of the built environment; (iii) dangerous traffic; (iv) fear of violent crime; (v) access to leisure facilities and green spaces; and (vi) restrictions on girls' mobility after puberty.
Many reviews reported the heterogeneity and lack of standardization of the assessment methods used in the primary studies. For example, one review on food insecurity reported that the studies included used 26 distinct indicators of food insecurity and that many studies neither directly measured food insecurity nor adequately reported the measures they used. 23 On physical activity, reviews typically reported that the primary studies tended to use selfreport assessments and not objective assessments or measuring tools. [18][19][20]28

Discussion
We identified 25 reviews published between 1998 and 2019 that described factors associated with diet and physical activity in Africa and the Caribbean. Although our scoping review considered only evidence from these regions, our findings confirm that evidence is generally lacking from such settings on which to base policy and design interventions for improving diet and physical activity. Moreover, our findings are consistent with those of a previous study, 10 which carried out a systematic review of research from low-and lowermiddle-income countries published between 1990 and 2015 on the effect of interventions aligned with WHO's "best buy" interventions on noncommunicable disease. 41 They identified 36 studies, which covered only nine of the 83 low-and lower-middle-income countries. Only two of the 36, both from Pakistan, concerned diet and physical activity. In our study, we found no review from Africa or the Caribbean that summarized evidence relevant to WHO's "best buy" interventions. Similarly, none of the literature we identified assessed primary research relevant to WHO's global action plan targets on noncommunicable diseases or to targets set for the relevant sustainable development goals (SDGs). 4,42 Although there may be research from Africa and the Caribbean that has not yet been reviewed, our findings suggest that, to date, policies on diet and physical activity are not informed by summarized research evidence on their determinants from these settings. This conclusion has two clear implications: (i) relevant primary research that has not yet been reviewed should be identified and evaluated; and (ii) new research should be undertaken to fill gaps in the evidence.
The policy responses and types of intervention required to improve health outcomes associated with diet and physical inactivity may be quite different in Africa and the Caribbean than in higher-income settings. In the absence of evidence indicating how different they need to be, current international guidance (e.g. WHO's "best buy" interventions and recommendations in the global action plan on noncommunicable diseases) should be followed, so long as the interventions employed are robustly evaluated and can subsequently contribute to the evidence available from Africa and the Caribbean. 43 Research funding bodies could help fill knowledge gaps and encourage the production of evidence summaries to guide policy. It would help if the terminology and definitions used for outcomes and their hypothesized determinants were much more consistent than we found in our study. In addition, international research networks that cover a range of different settings across Africa and the Caribbean could help develop and promote the high-quality, multidisciplinary research needed to address the complexity inherent in understanding how behavioural determinants vary between different contexts. 44 In choosing to carry out a broad scoping review of factors associated with diet and physical activity and by adopting the review as the unit of analysis, our intention was to highlight gaps in the summarized literature (rather than in the primary literature) as an aid to policy-making. We did not directly look for primary research on the determinants of diet and physical activity, nor did we summarize policy documents. Consequently, our review does not indicate, for example, whether or not there exists a large number of primary research studies that have not yet been included in systematic appraisals of the evidence. Nor can we evaluate the degree to which existing policies are evidencebased; we can only comment on whether there is sufficient summarized evidence to inform those policies.
Our search strategy and the study's conclusions were limited to factors that had a hypothesized or demonstrated association with behaviours affecting diet or physical activity. It is likely that, in some settings, academic research investigated factors associated with obesity or noncommunicable disease but did not explicitly categorize behaviour. Consequently, given that we were primarily interested in factors associated with behaviour rather than disease, our search strategy -though broad -could have missed some reviews of the determinants of diet and physical activity. Moreover, the cut-off date for inclusion in our review was 2019, which was just 4 years into the period covered by the SDGs. Most of the research included was, therefore, conducted during the era of the millennium development goals, which focused on undernutrition and did not stipulate any targets or indicators for noncommunicable disease.
Another limitation was that we did not appraise the quality of the reviews or the robustness of their evidence because our scoping review was intended primarily to map work in this area. Moreover, we identified papers only in English and may have missed reviews in other languages. We did not search the grey literature as our focus was on peerreviewed academic journals. However, having identified gaps in the literature,

Systematic reviews
Diet and physical activity in Africa and the Caribbean Eleanor Turner-Moss et al.
we plan to include both academic and grey literature in a range of languages in future reviews.

Pruebas de los factores asociados a la dieta y la actividad física en los países de África y el Caribe
Objetivo Identificar y describir las pruebas resumidas sobre los factores asociados a la dieta y la actividad física en los países de ingresos bajos y medios de África y el Caribe mediante la realización de una revisión del alcance de las revisiones. Métodos Realizamos búsquedas en las bases de datos Medline®, LILACS, Scopus, Global Health y Web of Science de revisiones de factores asociados a la dieta o la actividad física publicadas entre 1998 y 2019. Al menos el 25% de los estudios de las revisiones debían proceder de África o el Caribe. Los factores se clasificaron utilizando el modelo social de salud de Dahlgren y Whitehead. No hubo evaluación de la calidad. Resultados Identificamos 25 revisiones: 13 sobre la dieta, cuatro sobre la actividad física y ocho sobre ambas. Dieciocho artículos eran revisiones sistemáticas cuantitativas. En 12 revisiones, entre el 25 y el 50% de los estudios eran de África o el Caribe. Solo tres incluyeron pruebas del Caribe. En conjunto, las 25 revisiones incluyeron evidencia primaria publicada entre 1926 y 2018. Pocas de las pruebas resumidas se referían a las asociaciones entre los factores políticos o de salud internacionales y la dieta o las asociaciones entre cualquier factor y la actividad física en todas las categorías del modelo social de salud. Conclusión En la revisión del alcance encontramos una gran variedad de factores que, según los informes, se asocian con la dieta y la actividad física en África y el Caribe, pero carecemos de pruebas resumidas que puedan ayudar a informar las políticas que fomentan los comportamientos relacionados con las dietas saludables y la actividad física en estas regiones. Es necesario realizar más revisiones para informar a las políticas sobre los puntos en los que existen pruebas y sobre la necesidad de realizar investigaciones primarias adicionales.