HbA1c Performance in African Descent Populations in the United States With Normal Glucose Tolerance, Prediabetes, or Diabetes: A Scoping Review

Introduction African descent populations in the United States have high rates of type 2 diabetes and are incorrectly represented as a single group. Current glycated hemoglobin A1c (HbA1c) cutoffs (5.7% to <6.5% for prediabetes; ≥6.5% for type 2 diabetes) may perform suboptimally in evaluating glycemic status among African descent groups. We conducted a scoping review of US-based evidence documenting HbA1c performance to assess glycemic status among African American, Afro-Caribbean, and African people. Methods A PubMed, Scopus, and Cumulative Index to Nursing and Allied Health Literature (CINAHL) search (January 2020) yielded 3,238 articles published from January 2000 through January 2020. After review of titles, abstracts, and full texts, 12 met our criteria. HbA1c results were compared with other ethnic groups or validated against the oral glucose tolerance test (OGTT), fasting plasma glucose (FPG), or previous diagnosis. We classified study results by the risk of false positives and risk of false negatives in assessing glycemic status. Results In 5 studies of African American people, the HbA1c test increased risk of false positives compared with White populations, regardless of glycemic status. Three studies of African Americans found that HbA1c of 5.7% to less than 6.5% or HbA1c of 6.5% or higher generally increased risk of overdiagnosis compared with OGTT or previous diagnosis. In one study of Afro-Caribbean people, HbA1c of 6.5% or higher detected fewer type 2 diabetes cases because of a greater risk of false negatives. Compared with OGTT, HbA1c tests in 4 studies of Africans found that HbA1c of 5.7% to less than 6.5% or HbA1c of 6.5% or higher leads to underdiagnosis. Conclusion HbA1c criteria inadequately characterizes glycemic status among heterogeneous African descent populations. Research is needed to determine optimal HbA1c cutoffs or other test strategies that account for risk profiles unique to African American, Afro-Caribbean, and African people living in the United States.


Results
In 5 studies of African American people, the HbA 1c test increased risk of false positives compared with White populations, regardless of glycemic status. Three studies of African Americans found that HbA 1c of 5.7% to less than 6.5% or HbA 1c of 6.5% or higher generally increased risk of overdiagnosis compared with OGTT or previous diagnosis. In one study of Afro-Caribbean people, HbA 1c of 6.5% or higher detected fewer type 2 diabetes cases because of a greater risk of false negatives. Compared with OGTT, HbA 1c tests in 4 studies of Africans found that HbA 1c of 5.7% to less than 6.5% or HbA 1c of 6.5% or higher leads to underdiagnosis.

Introduction
People of African descent in the United States have a disproportionate burden of type 2 diabetes; prevalence is higher in African descent populations, 14%, compared with White populations of European descent (White populations), 9% (1). Additionally, African descent populations are represented as a single group, despite being comprised of African American (91%), Afro-Caribbean (4.7%), and African (3.7%) people (2,3). Limited evidence examines how intraethnic differences in cardiometabolic risk criteria, social determinants of health, and genetic admixture affect diabetes risk in these 3 populations (4,5). Current glycated hemoglobin A 1c (HbA 1c ) cutoffs (HbA 1c 5.7% to less than 6.5% for prediabetes; HbA 1c of 6.5% or higher for type 2 diabetes), determined from predominantly White population cohorts (4-8), may perform suboptimally in evaluating glycemic status in this diverse population of African American, Afro-Caribbean, and African populations (9-12). African American people may have higher HbA 1c values across the glycemic spectrum (9,13), and African immigrants may have lower HbA 1c values compared with White people (14). To ensure accurate detection of type 2 diabetes, there is a need to understand the ability of HbA 1c to correctly classify type 2 diabetes status and to evaluate intraethnic variation among African American, Afro-Caribbean, and African people (15)(16)(17).
Compared with random glucose, fasting plasma glucose (FPG), and the 2-hour oral glucose tolerance test (OGTT), HbA 1c has multiple benefits. It does not require fasting, tracks plasma glucose over the preceding 2 to 3 months, and better predicts complications such as cardiovascular disease (4,18). The HbA 1c test is stable, unaffected by external variables (eg, exercise, recent meals, and environmental stressors), and easily added to blood tests (19,20). However, interpretation of HbA 1c results is affected by the reduced lifespan of red blood cells in patients with type 2 diabetes, anemia, and hemoglobinopathies, conditions which disproportionately affect African descent populations (21)(22)(23)(24)(25).
The goal of our study was to conduct a scoping review of USbased peer-reviewed evidence documenting HbA 1c performance in African American, Afro-Caribbean, and African populations in the United States with the objectives of 1) summarizing evidence on HbA 1c performance in each subethnic group; 2) demonstrating variations in HbA 1c performance by each subethnic group; and 3) identifying potential future areas of research.

Data sources
In early January 2020, we searched PubMed, Scopus, and Cumulative Index to Nursing and Allied Health Literature (CINAHL) for peer-reviewed studies published between January 1, 2000, and January 1, 2020, by using complex search strings that were tested and developed in partnership with our institution's health sciences librarian (L.A.F.). The search string included medical subject headings (MeSH) terms and key words such as "African continental ancestry group," "African Americans," "Caribbean," and "West Indian" to describe population groups and "Glycated Hemoglobin A," "hemoglobin A 1c ," and "hba 1c " to describe the testing indicator of interest for type 2 diabetes (Appendix).

Study selection
Throughout the review process, we screened articles for studies meeting the following inclusion criteria: Articles were original studies published between January 2000 and January 2020, that evaluated HbA 1c performance in African descent groups.

2.
Study participants were living in the United States. 3.
Study was a database analysis, epidemiologic study, or clinical study. 4.
HbA 1c performance was reported specifically in one or more of the African descent groups.

5.
HbA 1c performance was assessed in healthy populations or for screening or diagnosis of prediabetes or type 2 diabetes.

6.
HbA 1c performance was assessed by statistical methods (eg, sensitivity, specificity, and positive predictive value), compared with other tests in the same population, or compared African descent populations to other racial groups.

7.
During the study selection process, we included studies that compared various diabetes screening tests against HbA 1c , including the OGTT, FPG, and glycated protein tests, to avoid excluding major findings. Although the OGTT is considered optimal for comparison, it is far more costly, resource intensive, and time consuming than the FPG and glycated protein tests (6-8); additionally, research supports the use of other tests along with OGTT or in place of OGTT to enhance detection of diabetes (7,(18)(19)(20)(21)(22). Because African descent populations are less likely to be adequately represented in clinical research and simultaneously experience health care inequities (4,19), we wanted to be inclusive of all the data, in comparison to HbA 1c , that were available for the populations.
On the basis of the title and abstract review, we excluded articles that did not match the set inclusion criteria above ( Figure). Two authors (L.K. and S.B.) conducted independent title and abstract reviews. In the full-text review, we excluded articles with insufficient data (eg, case studies), narrative reviews, and articles that fell under a previously set exclusion criterion not detected during the title and abstract review. Full-text articles for potential studies were reviewed by 2 authors (L.K. and S.B.) independently. When multiple exclusion criteria were met, we categorized the article by the exclusion criterion that appeared first in title, abstract, or full text review. A third author (M.H.R.) verified that the exclusion criteria were relevant throughout the article. Figure. Flow diagram of the study selection process for glycated hemoglobin A 1c (HbA 1c ) testing performance in African descent populations in the United States, using PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses). Studies were published January 1, 2000, to January 1, 2020.
During the identification process, 3,238 records were identified through database searching. In the screening phase, 3,081 records were screened after 157 duplicates were removed. Records were excluded by using a title and abstract review (n = 3,035) by the following exclusion criteria: topic was type 1 diabetes (n = 98); age was exclusively less than 19 years or greater than 64 years (n = 217); topic was animals or objects (n = 22); study was conducted outside the United States (n = 422), study generalized African descent populations as one group (n = 58); study did not report HbA 1c performance in African descent populations (n = 631); topic was a dietary study (n = 30); topic was other diseases, disorders, complications (including diabetes-related complications), or illnesses (eg, kidney) (n = 933); topic was a treatment or interventional study (n = 527); topic was gestational diabetes (n = 13); or topic was a genetic study (n = 84). After this screening process, the remaining 46 full-text articles were assessed for eligibility. Of these, 34 full-text articles were excluded based on the following exclusion criteria: insufficient data (n = 6); being a narrative review (n = 10); or for a reason not previously detected in the title or abstract (n = 18). The inclusion stage yielded 12 studies to be included in qualitative synthesis.

Data extraction
We created a data extraction sheet to record the study author and name, populations (sample size, male/female breakdown, race/ethnicity distribution, age, and study location), HbA 1c laboratory methods, study design, HbA 1c evaluation methods, findings, and HbA 1c performance. We successfully retrieved any missing information by 1) searching through cited articles from which the studies retrieved data; 2) identifying parent studies and protocol descriptions given in prior publications; and 3) emailing corresponding authors. HbA 1c performance was classified using 2 labels: 1) greater risk of false positive (GRFP) label indicated that the HbA 1c test may result in overdetection of glycemic status (eg, type 2 diabetes) that the study is measuring or 2) greater risk of false negatives (GRFN) label indicated that the HbA 1c test may result in underdetection of glycemic status. This classification system (GRFP or GRFN) was based on text analysis of the language used by the authors of each study in the way they interpreted their results (eg, lower sensitivity, lower specificity, more misdiagnoses). This allowed for standardization of labeling findings from different study designs. GRFP was assigned if studies reported 1) higher HbA 1c values in African descent participants compared with other ethnic groups (eg, White participants) at the same glycemic level; 2) lower sensitivity because of less true positives; or 3) lower specificity because of more false positives. GRFN was assigned if studies reported 1) lower HbA 1c values in participants compared with other ethnic groups at same glycemic level; 2) lower sensitivity because of more false negatives; or 3) or lower specificity because of less true negatives. Discrepancies in the review process and data extraction were resolved with input from a third author (M.H.R.).
Included studies were grouped based on study population (African American, Afro-Caribbean, and African) and then organized in alphabetical order by the first author's last name. Studies were labeled numerically as 1 through 12 based on this ordering.

Results
Of the 12 articles that met the inclusion criteria, studies numbered 1 through 7 analyzed HbA 1c performance among African American people (26-32), study number 8 analyzed HbA 1c performance among Afro-Caribbean people (33), and studies numbered 9 through12 analyzed HbA 1c performance among African people PREVENTING CHRONIC DISEASE   [34][35][36][37]. All studies were conducted with people living in the United States (Table 1).
The population size of the studies varied from 83 to 16,056 participants, with the sex representation ranging from 69% male/31% female to 0% male/100% female ( Table 1). The study cohorts consisted of 20.2% to 100% African descent populations. The overall age range across the different studies was 18 to 92 years and the mean age was between 37 and 64 years when reported (Table 1).
The study designs included either clinical data collection (studies 1 and 5 through 12) (26,(30)(31)(32)(33)(34)(35)(36)(37)  In these studies, HbA 1c performance was evaluated by comparing HbA 1c results in African descent populations to HbA 1c results in other ethnic groups (eg, White people) (studies 1, 2, 5, and 7) (26,27,30,32), evaluating HbA 1c test results against the 2-hour OGTT, FPG, glycated plasma proteins test results, or previous diagnosis in the same participants (studies 3, 6, and 8-12) (28,31,33-37), or both (study 4) (29) ( Table 2). Studies conducted among African American people showed that the HbA 1c test almost always had a GRFP in this population. Studies 1, 2, 4, 5, and 7 demonstrated that HbA 1c values were higher in African Americans when compared with Whites across a range of glycemic states (26,27,29,30,32). Additionally, Study 7 showed that HbA 1c values were higher in African American people when compared with both White people and Hispanic people, leading to the potential of overdiagnosis of type 2 diabetes in African American people (32). Using OGTT as a standard test for diagnosis of glycemic status, studies 4 and 6 demonstrated that using HbA 1c results in overdiagnosis of type 2 diabetes when HbA 1c is 6.5% or higher (29,31). Study 3 showed that African American people may experience an overdiagnosis of prediabetes or type 2 diabetes at HbA 1c of 5.7% to less than 6.5%; however, study 6 showed that an HbA 1c cutoff of less than 5.7% does not eliminate the possibility of a type 2 diabetes diagnosis (28,31) ( Table 2).
In the Afro-Caribbean population, HbA 1c testing at the 6.5% or higher cutoff has a GRFN (33). Using FPG as a standard for dia-gnosis of type 2 diabetes, study 8 showed that more participants were correctly diagnosed as having type 2 diabetes if the cutoff was lowered to 6.26% or higher, suggesting that HbA 1c values are generally lower in Afro-Caribbean people ( Table 2).
The Africans in America studies 9 through 12 all showed that HbA 1c has a GRFN in African people at the HbA 1c cutoff of 5.7% to less than 6.5% for prediabetes and HbA 1c cutoff of 6.5% or higher for type 2 diabetes (34-37). Using OGTT as a diagnostic standard for glycemic status, studies 9 through 12 demonstrated that using an HbA 1c cutoff of 5.7% to less than 6.5% will lead to underdiagnosis of prediabetes in Africans. Additionally, study 9 showed that using an HbA 1c cutoff of 6.5% or higher will lead to an underdiagnosis of type 2 diabetes in Africans (34) ( Table 2).

Discussion
We assessed 12 studies that evaluated the ability of HbA 1c to correctly identify African American, Afro-Caribbean, and African people with prediabetes or type 2 diabetes. Studies among African American people found that HbA 1c of 5.7% to less than 6.5% or HbA 1c of 6.5% or higher led to overdiagnosis. In one study of Afro-Caribbean people, HbA 1c of 6.5% or higher had a greater risk of false negatives (GRFN). Among African people, HbA 1c of 5.7% to less than 6.5% or HbA 1c of 6.5% or higher led to greater risk of underdiagnosis.
Overdiagnosis of diabetes was likely among African American people in 3 ways. African American people had consistently higher HbA 1c levels than White people regardless of glycemic status (26,27,29,30,32). Furthermore, half of normoglycemic African American people had HbA 1c values greater than 5.7% (28); and lastly, African American people were more likely to be diagnosed with type 2 diabetes by HbA 1c of 6.5% or higher alone but not by OGTT (29,31). Although study 6 did suggest a GRFN at HbA 1c less than 5.7%, by misdiagnosing some participants as having normal glycemic status if their HbA 1c was less than 5.7% (31), the finding is limited by the smaller sample size of 83 participants when compared with the other studies. This finding must be investigated further.
In Afro-Caribbean people, the HbA 1c cutoff of 6.5% is likely to result in underdiagnosis of type 2 diabetes because study 8 showed that more participants were correctly diagnosed as having type 2 diabetes if the cutoff was lowered to 6.26% (33). However, this finding may not be generalizable to other Afro-Caribbean populations because of the smaller sample size and limitation of the study population to Haitian American people. Additionally, because only 1 study provided this conclusion, generalizability is further limited. For African people, underdiagnosis of prediabetes and PREVENTING CHRONIC DISEASE VOLUME 18, E22 PUBLIC HEALTH RESEARCH, PRACTICE, AND POLICY

MARCH 2021
The opinions expressed by authors contributing to this journal do not necessarily reflect the opinions of the U.S. Department of Health and Human Services, the Public Health Service, the Centers for Disease Control and Prevention, or the authors' affiliated institutions.
type 2 diabetes is also likely at the standard HbA 1c cutoffs because diagnosis was missed by HbA 1c despite being detected by OGTT (34-37). The findings among African people hold true regardless of hemoglobin variant or obesity status (35,36).
Genetics are often thought to be responsible for the differences of HbA 1c performance in African descent populations (24,(40)(41)(42)(43). In fact, genetic analysis in study 5 shows that the HbA 1c difference was primarily because of the genomic principal component analysis (PCA) factor in African American people when compared with White people (30). The study demonstrated that the PCA factor was associated with increased HbA 1c values in African American people. However, genetics do not fully explain HbA 1c differences among African American people (44), because increases in HbA 1c may be mediated by social determinants of health (eg, chronic financial strain as seen in study 3) or chronic inflammation (sIL-6R) (28,45). Additionally, G6PD variant or deficiency is often correlated with lower HbA 1c values in various populations (40), especially in African American people and African people because of its higher prevalence in these groups (14,46,47). Similarly, the sickle cell trait is associated with lower HbA 1c values in African descent populations (21,25). However, study 1 showed that the sickle cell trait may not actually correlate to changes in HbA 1c values for African American people (26). Findings regarding associations of genetics with HbA 1c are still being researched in this population. Research accounting for genetically linked HbA 1c differences in Afro-Caribbean people is also lacking. Genetic polymorphisms between African American people and Haitian people have been researched and show that differences in the PPARGC1A gene will correlate to risk of type 2 diabetes in African American people as opposed to protective associations with type 2 diabetes in Haitian people, suggesting that other genetic associations may explain differences in diabetes for Haitian people (48). Although little research explains the role of genetics in HbA 1c differences for Haitian people, one likely contributor to lower HbA 1c values may be the G6PD variant because of its higher prevalence in populations of African descent (47). Nevertheless, opposing findings regarding the role of genetics in influencing HbA 1c values (eg, PCA factor is associated with higher HbA 1c whereas the sickle cell trait is associated with lower HbA 1c ) make it difficult to ascertain the overall impact genetics has in causing the differences in HbA 1c that were found for the African descent populations and therefore require further evaluation.
Socioeconomic factors and health behaviors such as diet, smoking, and exercise may explain some differences in glycemic control and HbA 1c values among the 3 groups. Higher income and educational attainment appear to decrease the odds of diabetes among African immigrants, whereas only higher education lowers the odds for African American people (5). Neither education nor in-come appear to affect diabetes risk among Afro-Caribbean people (5,49). Additionally, study 3 found that financial stress and chronic inflammation were associated with higher HbA 1c . Chronic inflammation resulting from social and environmental stressors, including experiences of racism, correlate to higher HbA 1c in nondiabetic adults (50). In terms of health behaviors, compared with African American people, African and Afro-Caribbean people are less likely to smoke. As African and Afro-Caribbean immigrants settle in the United States, they are affected by dietary acculturation often characterized by increased caloric intake and diets higher in refined carbohydrates, animal protein, fat, and sodium (5). Although diet may affect glycemic control, it is unlikely that diet explains the differences in HbA 1c performance illustrated in this study. These socioeconomic factors highlight the diversity of experience within African descent groups, which is often overshadowed by perceived homogeneity of the "Black" experience in the United States. Since immigration to the United States presents unique socioeconomic circumstances that can affect factors like HbA 1c (4), impacts of these circumstances are important to analyze distinctly from global concerns.
With these factors affecting HbA 1c performance, results must be interpreted with caution. Some alternative diagnostic tests are suggested to aid or replace HbA 1c for classification of glycemic status. For example, FPG in combination with HbA 1c increases the sensitivity for type 2 diabetes diagnosis in African people (study 10) (35). A stronger relationship between HbA 1c and FPG at higher FPG levels in most ethnic groups has been suggested as well (51). Study 8 suggests that FPG may be a better measure of glycemic status than HbA 1c in Afro-Caribbean people (33). At the same time, studies 3, 6, and 9 through 12 suggest that OGTT more accurately measures glycemic status than HbA 1c in both African American and African people (28,31,(34)(35)(36)(37). Comparisons between HbA 1c and OGTT in Afro-Caribbean people are lacking and should be studied further.
Convenient nonfasting alternatives for type 2 diabetes testing are other glycated proteins (eg, glycated albumin, fructosamine, and other advanced glycation end products) either in combination with or in place of HbA 1c (36,37,(52)(53)(54)(55). Although this approach is supported in multiethnic studies, these glycated proteins should be evaluated specifically in African descent groups.
Several limitations exist for the findings of our review. Despite constructing a comprehensive search, articles published in peer reviewed journals that were not indexed in PubMed, Scopus, and CINAHL may have been missed. The search contained nouns and adjectives as identification for African descent countries and regions of origin and HbA 1c testing. However, study participant groups may be based on self or researcher categorization rather than actual region, country, or ethnic group of the participant. Findings must be interpreted with caution because of this subjective labeling within studies. Additionally, we did not use a specific protocol to evaluate the quality of the included studies, as this is not a part of scoping review methodologies and can increase risk of bias (56,57). Another limitation that must be considered is that time may pass between HbA 1c testing and alternate testing in some studies and glycemic status of individuals can change in that time; this limitation will usually exist in this nature of clinical research methodology and therefore must be recognized when evaluating the conclusions from those studies.
According to our review process, there is only 1 study protocol in the United States that examines performance of diabetes screening tests among African immigrants to the United States (34-37). However, studies 9 through 12 demonstrate distinct comparisons within this cohort that illustrate significant conclusions about HbA 1c performance. This is because the protocol is ongoing, and the number of participants increased over time. In turn, this also lends strength to the findings, because the similarity in protocol is balanced by the increasing diversity of the sample for each study design.
Finally, the lack of existing studies for Afro-Caribbean people in the United States presents a substantial limitation; our findings for this group must be interpreted cautiously. Further research is needed to understand the performance of HbA 1c and evaluate alternate tests in place of the HbA 1c in specific African descent populations, especially Afro-Caribbean people. Unique settings like New York City, where 32% of the African descent population is Afro-Caribbean and 4% is African (58), may serve as key locations for public health researchers to investigate type 2 diabetes screening and diagnostics.
Our review also has several strengths. In partnership with our institution's research librarian, we tested several search constructions and selected the searches that provided the broadest selection within the scope of our topic. Additionally, we searched 3 databases without limiting article type or study designs on title and abstract review and had 2 reviewers independently screen the articles. This improved the selection of articles available for review and reduced selection bias. Finally, we were able to provide clear findings by constructing a label categorization scheme (GRFP/ GRFN) that allowed for grouping of studies that used different comparative analytic and statistical methods to analyze HbA 1c .
In African descent populations in the United States, the utility of HbA 1c is limited in screening for glycemic status, determining care methods, assessing risk of type 2 diabetes complications, or analyzing health disparities. Current HbA 1c cutoffs for prediabetes and type 2 diabetes may overestimate glycemic status in African American people and underestimate glycemic status in Afro-Caribbean and African people. Reasons for variations in HbA 1c have been attributed to genetic, biochemical, and socioeconomic factors. Alternate testing such as OGTT, FPG, and other glycated blood proteins in place of or in combination with HbA 1c may better assess glycemic status in African descent populations. Intraethnic HbA 1c heterogeneity within the African descent groups must be recognized, and identification of more reliable type 2 diabetes screening and diagnostic tests is urgent.   and glycated albumin) Among subjects with prediabetes by OGTT, HbA 1c of 5.7% to less than 6.5% had 37% sensitivity in nonobese African immigrants and 64% sensitivity in obese African immigrants (36).
negatives Additional finding: For HbA 1c of 5.7% to less than 6.5% combined with glycated albumin ≥13.77%, sensitivity increased to 72% for nonobese African immigrants.
• Abbreviations: OGTT, 2-hour oral glucose tolerance test; FPG, fasting plasma glucose; IGT, impaired glucose tolerance; PCA, principal component analysis; GRS, genetic risk score; SCT, sickle cell trait; ROC, receiver operating characteristic. a Exact temporality between the previous diagnosis and HbA 1c testing was not provided within the study, with an estimate of less than 12 months extrapolated from the study design. Findings from this study may represent new onset diabetes. This provides a limitation in the conclusive findings for HbA 1c performance in this study.
PREVENTING CHRONIC DISEASE VOLUME 18, E22 PUBLIC HEALTH RESEARCH, PRACTICE, AND POLICY

MARCH 2021
The opinions expressed by authors contributing to this journal do not necessarily reflect the opinions of the U.S. Department of Health and Human Services, the Public Health Service, the Centers for Disease Control and Prevention, or the authors' affiliated institutions.
Cumulative Index to Nursing and Allied Health Literature (CINAHL) (hba1c OR glycosylated hemoglobin A OR glycated hemoglobin OR "hemoglobin A1c" OR "glycated hemoglobin A") AND (africa OR african OR africans OR afro OR black OR african americans OR blacks OR caribbean OR Angola OR Angolan OR Benin OR Beninese OR Botswana OR Motswana OR Batswana OR "Burkina Faso" OR Burkinabe OR Burundi OR Burundian OR Cameroon OR Cameroonian OR "Cape Verde" OR "Cape Verdean" OR "Central African Republic" OR "Central African" OR Chad OR Chadian OR Comoros OR Comorian OR "Republic of the Congo" OR Congolese OR Djibouti OR Djiboutian OR "Equatorial Guinea" OR "Equatorial Guinean" OR Equatoguinean OR Eritrea OR Eritrean OR Ethiopia OR Ethiopian OR Gabon OR Gabonese OR Gambia OR Gambian OR Ghana OR Ghanaian OR Guinea OR Guinean OR "Guinea-Bissau" OR "Bissau-Guinean" OR "Ivory Coast" OR Ivorian OR Kenya OR Kenyan OR Lesotho OR Mosotho OR Basotho OR Liberia OR Liberian OR Madagascar OR Malagasy OR (continued on next page) PREVENTING CHRONIC DISEASE VOLUME 18, E22 PUBLIC HEALTH RESEARCH, PRACTICE, AND POLICY

MARCH 2021
The opinions expressed by authors contributing to this journal do not necessarily reflect the opinions of the U.S. Department of Health and Human Services, the Public Health Service, the Centers for Disease Control and Prevention, or the authors' affiliated institutions.