Prevalence and correlates of pre-diabetes and diabetes among a national population-based sample of adults in Zambia: results of the first national STEPS survey in 2017

Diabetes has been on the rise in Africa. This study aimed to estimate for the first time the national prevalence and its correlates of pre-diabetes and diabetes among individuals aged 18–69 years in Zambia. Nationally representative cross-sectional data were analyzed from 3608 persons aged 18–69 years (median age: 31 years) that participated in the “2017 Zambia STEPS survey,” with complete blood glucose measurements. Results indicate that 8.8% of 18–69 year-olds had pre-diabetes and 7.2% diabetes. In adjusted multinomial logistic regression analysis, rural residence (adjusted relative risk ratio = ARRR: 2.01, 95% confidence interval = CI: 1.40–2.89), and raised total cholesterol (ARRR: 1.78, 95% CI: 1.08–2.94) were positively, and high physical activity (ARRR: 0.57, 95% CI: 0.39–0.83) was negatively associated with pre-diabetes. Being 50–69 years old (ARRR: 3.03, 95% CI: 2.03–4.52), having central obesity (ARRR: 1.90, 95% CI: 1.20–3.03), and hypertension (ARRR: 2.24, 95% CI: 1.61–3.13) were positively associated with diabetes. In addition, in the unadjusted analysis, female sex, lower education, alcohol family problems, and alcohol dependence were associated with pre-diabetes and/or diabetes. Only 8.4% of the study sample reported that they ever had their blood glucose examined by a health care professional. Having had blood glucose measured was higher among women (9.6%) than men (7.2%) were but not significant (p = 0.08). Residents in urban areas (11.8%) had significantly more often their blood glucose ever measured than residents in rural areas (5.4%) (p < 0.001). Among study participants with diabetes, 22.3% were aware, 9.4% were currently taking treatment, and 17.1% had controlled their diabetes (< 7.0 mmol/L). Almost one in ten participants had pre-diabetes and diabetes and several associated variables were detected which can aid in designing intervention strategies.


Background and purpose
Almost one-third (29%) of all deaths in 2016 in Zambia was attributed to non-communicable diseases (NCD); the mortality contribution from diabetes was 1% [1]. According to the World Health Organization (WHO), diabetes was estimated to be the seventh largest cause of mortality in 2016 worldwide [2]. Globally, the prevalence of diabetes increased significantly from 1980 to 2014 (in women from 5.0 to 7.9%, and in men from 4.3 to 9.0%) [3]. Compared to high-income countries, the prevalence of diabetes has been rising more rapidly in lowand middle-income in recent years [2]. To prevent and control diabetes, it is important that national population-based surveys are conducted periodically [3]. There is a lack of national data on the prevalence of pre-diabetes and diabetes and associated factors in Zambia, a lower-middle-income country in Southern Africa.
In a large study among adults in 16 communities from five of 10 provinces in 2010 in Zambia, the prevalence of diabetes was 3.5% [4]. In the 2008 STEPS survey in Lusaka district, Zambia, among participants 25 years or older, the combined prevalence for pre-diabetes or diabetes was 4.0% [5], and in an investigation of bank employees (N = 121) in Ndola, Zambia, the prevalence of diabetes mellitus was 15% [6]. In other African countries, the national prevalence of diabetes was 5.8% in Burkina Faso [7], 3.3% in Ethiopia [8], 5.7% in Guinea [9], 5.6% (pre-diabetes 4.2%) in Malawi [10], and 1.4% (pre-diabetes 2.0%) in Uganda [11].
The investigation aimed to estimate the prevalence and its correlates of pre-diabetes and diabetes among 18-69 year-old persons in Zambia.

Study design and procedures
Cross-sectional nationally representative data from the "2017 Zambia STEPS Survey" were analyzed [20]. Using a multistage cluster sampling technique, a nationally representative sample of adults (18-69 years) in Zambia was produced [21]. "In the first stage of sampling, Standard Enumeration Areas (SEAs) were selected from each province using a probability proportional to size (PPS). In the second stage, 15 households in rural SEAs and 20 households in urban SEAs were selected systematically using an appropriate sampling interval based on the number of households in that SEA" [21], and in the third stage, one member from the eligible household members (18-69 years, residing in household) was selected by simple random sampling [21]. Data collection followed the WHO three STEPS methodology: step 1 included administration of a structured questionnaire, step 2 consisted of blood pressure and anthropometric measurements, and step 3 included biochemical tests (blood glucose and blood lipids) [21]. A pilot study was conducted to check the content validity of the questions after translation [21]. The main fieldwork started on July 22, 2017, and ended on October 15, 2017 [21]. Each field investigator team included one supervisor for planning and checking the completeness of questionnaires [21]. The "overall response rate was 74.3%" [21].

Measures
Outcome variable: pre-diabetes and diabetes Fasting (≥ 10 h) blood sugar measurements were conducted and the history of diabetes assessed (see Supplementary file 1) [21]. "Testing was performed using a portable rapid diagnostic device (Cardiochek™) machine which used test strips for both blood glucose and lipid profiles (total cholesterol and HDL cholesterol)." [21]. Blood samples were collected using a finger prick [21]. Pre-diabetes was defined as "fasting plasma glucose levels 6.1 to< 7 mmol/L and diabetes as fasting plasma glucose levels ≥ 7.0 mmol/L, and/or currently taking insulin or oral hypoglycemic drugs and/or having been diagnosed with diabetes by a health care professional." [3]. Diagnosed diabetes was defined as selfreported health care provider diagnosis and/or currently taking insulin or oral hypoglycemic drugs, and undiagnosed diabetes was defined as fasting plasma glucose levels ≥ 7.0 mmol/L and no self-reported health care provider diagnosis and/or currently taking insulin or oral hypoglycemic drugs.
Sociodemographic information included sex, age, work status, education, ethnic affiliation, residence status, and marital status.
Psychosocial distress variables included having alcohol family problems in the past 12 months, family members ever died from suicide, suicidal ideation in the past 12 months and passive smoking (at home and/or at work) in the past 30 days (details in Supplementary file 1) [21].
Health status variables included measured central obesity (waist circumference > 88 cm in females and > 102 cm in males); body mass index (measured < 18.5 kg/m 2 underweight, 18.5-24.4 kg/m 2 normal weight, 25-29.9 kg/m 2 overweight and ≥ 30 kg/m 2 obesity); hypertension based on blood pressure (BP) measurements (average of the last two of three readings) defined as systolic BP ≥ 140 mmHg and/or diastolic BP ≥ 90 mmHg or currently on antihypertensive medication; raised total cholesterol (TC) ("fasting TC ≥ 5.0 mmol/L or currently on medication for raised cholesterol") [21].
Health risk behavior variables included daily tobacco use (smoking and/or smokeless tobacco, alcohol dependence, inadequate fruit and vegetable intake) (< 5 servings/day), and based on the "Global Physical Activity Questionnaire" low, moderate, or high physical activity and sedentary behavior (≥ 8 h/day) [21]. Alcohol dependence was assessed with three questions of the "Alcohol Use Disorder Identification Test (AUDIT)" (items 4-6), e.g., "How often during the last year have you found that you were not able to stop drinking once you had started?" Response options ranged from "0 = never to 4 = daily or almost daily"; total scores of 4 or more indicate alcohol dependence [22]. Physical activity was categorized by the median metabolic equivalent (METs) of performed activities as low ("total physical activity METs minutes per week is < 600"), moderate ("3 or more days of vigorous-intensity activity of at least 20 min per day OR; 5 or more days of moderate-intensity activity or walking of at least 30 min per day OR; 5 or more days of any combination of walking, moderate or vigorous intensity activities achieving a minimum of at least 600 MET-min per week") and high ("vigorous-intensity activity on at least 3 days achieving a minimum of at least 1500 MET-min per week OR; 7 or more days of any combination of walking, moderate or vigorous intensity activities achieving a minimum of at least 3000 MET-min per week.") physical activity [23,24].

Data analysis
Statistical analyses were done with "STATA software version 15.0 (Stata Corporation, College Station, Texas, USA)," taking into account the complex study design. The data were weighted "to make the sample representative of the target population (adults in Zambia aged 18 to 69 years)." [21]. Pearson Chi-square tests are used to calculate differences in proportions. Unadjusted and adjusted multinomial logistic regression was used to assess predictors of pre-diabetes and diabetes (with no pre-diabetes/diabetes as the reference category). Variables significant in the unadjusted analysis were included in the multivariable logistic regression model. The missing values were not included in the analysis. p < 0.05 was accepted as significant.

Sample and diabetes status characteristics
The sample comprised of 3657 18-69 year-old persons (31 years median age, 18 years interquartile range) with complete blood glucose measurement. More than half of the participants (61.8%) were female, 48.0% had more than primary education, 41.0% were never married, separated, divorced, or widowed, 50.4% were employed, 33.8% were Tonga by ethnicity, and 64.0% lived in rural areas. More than one in seven participants (14.7%) reported alcohol family problems, 6.2% had a close family member who died from suicide, 7.8% had past 12-month suicidal ideation, and 26.8% were exposed to passive smoking.
In addition, in the unadjusted analysis, female sex, lower education, alcohol family problems, and alcohol dependence were associated with pre-diabetes and/or diabetes (see Table 2).

Diabetes awareness, treatment, and control
Only 8.4% of the study sample reported that they ever had their blood glucose measured by a health care professional. Having had blood glucose measured was higher among women (9.6%) than men (7.2%) were but not significantly (p = 0.08). Residents in urban areas (11.8%) had significantly more often their blood glucose measured than residents in rural areas (5.4%) (p < 0.001). Among the study participants with diabetes, 22.3% were aware, 9.4% were currently taking treatment, and 17.1% had controlled their diabetes (< 7.0 mmol/L). Awareness, treatment, and control status of diabetes did not significantly differ by sex. Urban dwellers with diabetes were significantly more often aware, treated, and controlled their diabetes than rural dwellers. Awareness, treatment, and control of diabetes increased with age, but this was only significant for the treatment of diabetes (see Table 3).

Discussion
The investigation aimed to estimate the prevalence and correlates of pre-diabetes and diabetes in a national populationbased survey among 18-69 year-old persons in Zambia. The prevalence of diabetes (overall 7.2%, 7.5% in women and 6.9% in men) and pre-diabetes (8.8%) was similar among women globally (7.9%) and lower among men globally (9.0%) [3], and was higher than in local studies in Zambia (3.5% diabetes [4] and pre-diabetes or diabetes 4.0% [5]) and in Malawi (5.6% diabetes and prediabetes 4.2%) [10], in Uganda (1.4% diabetes and 2.0% pre-diabetes 2.0%) [11], Burkina Faso (5.8%) [7], Ethiopia (3.3% diabetes) [8], and in Guinea (5.7%) [9]. The increased rate of diabetes found in Zambia (a lower-middle income country) may be explained by a greater change of lifestyle, older age structure, and greater urbanization than in low-income other African countries (Burkina Faso, Ethiopia, Guinea, and Malawi) and older studies in Zambia [13,25]. The investigation showed a high prevalence of undiagnosed diabetes (77.3%), which seems to be higher than in Guinea (56%) [9], in Uganda (70.5%) [12], in Benin (68.0%) [12], and in Zambia (34.5%) [4] and lower than in Burkina Faso (91.7%) [12]. The prevalence of treated diabetics in this study (9.4%) was lower than in most other African countries, e.g., Benin (32.0%), Kenya (27.7%), and South Africa (40.1%), except for Burkina Faso (7.3%) [12]. RRR relative risk ratio, CI confidence interval; ***p < 0.001, **p < 0.01, *p < 0.05 The prevalence of controlled diabetes among diabetics (17.1%) in this study was lower than in Benin (21.7%) and South Africa (21.4%), similar to Kenya (18.4%) and higher than in Burkina Faso (6.9%) [12]. The study found that urban dwellers had greater awareness, treatment, and control of their diabetes than rural dwellers, while there were no sex differences. The lack of awareness, treatment, and control among rural dwellers may be attributed to poorer health services access. "Most primary care facilities in Zambia do not routinely screen for cholesterol or diabetes." [21]. By enhancing primary facilities to conduct blood glucose tests, especially in rural Zambia [21], diabetes awareness, treatment, and control may improve. Consistent with former research [4,[8][9][10][11], in this investigation, pre-diabetes and diabetes increased with age. In unadjusted analysis, the study showed that female sex was associated with pre-diabetes, but no significant sex differences were found in the adjusted analysis for pre-diabetes and diabetes. In a systematic review on sex differences of the prevalence of diabetes in Africa, in most countries, no sex differences were identified [26]. Residing in rural areas and in unadjusted analysis, lower education increased the odds for pre-diabetes. Some previous studies confirmed the association between rural residence [27], and lower education in high-income and not low-or middle-income countries [25,27] with diabetes, while some other studies [8,28,29] found a higher prevalence in urban areas. It appears, however, that diabetes has penetrated into both urban and rural areas in Zambia. Another possible reason for the higher rate of pre-diabetes in rural compared to urban areas in Zambia is that rural dwellers are significantly older than urban residence (p < 0.001), as pre-diabetes increases with age. A diabetes-screening program may be introduced, particularly targeting the older age high-risk groups [29].
Some studies found an association between psychosocial distress, such as suicidal behavior [15,16], stress [17,18], and passive smoking [19], increased the likelihood of diabetes, while in this study, only in unadjusted analysis alcohol family problem was associated with diabetes, while the stress of family members that died from suicide, suicidal ideation, and passive smoking were not significantly different with prediabetes and/or diabetes.
This survey found an association between hypertension, central obesity, and raised total cholesterol with pre-diabetes or diabetes. These findings are consistent with previous investigations [4,6,8,9,11], showing major "modifiable cardiometabolic risk factors" [9]. This "combination of cardiometablic risk factors calls for a multiple rather than a single risk intervention approach in this population." [13].
Several health risk behaviors, such as unhealthy diet, sedentary behavior, physical inactivity, tobacco use, and alcohol misuse, have been found to increase the risk for diabetes [6,11,[30][31][32][33], while in this study, only physical inactivity and in the unadjusted analysis alcohol dependence were associated with pre-diabetes, and in unadjusted analysis, physical inactivity was associated with diabetes, and no significant association between sedentary behavior, daily tobacco use, and inadequate fruit and vegetable intake and pre-diabetes and/or diabetes was found.

Study limitations
Diagnosed diabetes was based on participant's recall and not medical records, which may have led to underreporting. Participants were also not asked to specify whether they had type 1, type 2, or gestational diabetes. Only fasting capillary blood glucose was used for diagnosis of pre-diabetes and diabetes which in the absence of HbA 1C and a post glucose load value would result in an underestimate. The variable of household income had many missing values and could therefore not be included in the analysis.

Conclusion
The study found among a nationally representative population of 18 to 69 years in Zambia that almost one in ten participants had pre-diabetes and diabetes. Less than one in five Zambians were aware, treated, and controlled their diabetes. Several risk factors for pre-diabetes and/or diabetes were identified, including older age, rural residence, central obesity (or Table 3 Diabetes awareness, treatment, and control (N = 296)

Variable
Of diabetics aware Of diabetics treated Of diabetics controlled (< 7.0 mmol/l) N (%) N (%) N (%) overweight or obesity), hypertension, raised total cholesterol, and physical inactivity, and in unadjusted analysis, female sex, lower education, alcohol family problems, and alcohol dependence, which can assist in guiding interventions to prevent pre-diabetes and diabetes in the Zambian population.