Original Research••

National and provincial prevalence of self-reported diabetes: results from the cross-sectional Demographic and Health Survey in Sri Lanka–2016

Abstract

Introduction Diabetes is a major global public health burden. The objective of this study was to estimate the national and provincial-level prevalence of self-reported diabetes and establish the demographic risk factors of diabetes in Sri Lanka.

Research design and methods A nationally representative sample of the population from all nine provinces in Sri Lanka was obtained from the cross-sectional Demographic and Health Survey 2016. Those aged 20 years and older were included in this study. Age, sex, ethnicity, religion, education level, marital status, sector of the resident (urban/rural/estate) and province of residence were included as potential exposures. The outcome was defined as self-reported diabetes.

Results The age-adjusted national prevalence of self-reported diabetes was 10.60% (95% CI: 10.59% to 10.602%). Provinces with higher gross domestic product (GDP) per capita showed a higher prevalence of diabetes. The Western province with the highest GDP of 734 094 Sri Lankan rupee showed the highest prevalence of diabetes (13.3%; 95% CI: 13.32% to 13.34%) compared with 8.1% (95% CI: 8.54% to 8.58%) in the Northern province with lowest GDP (321 289 Sri Lankan rupee). Prevalence of diabetes was higher among residents living in urban areas (14.39%; 95% CI: 13.72% to 15.06%) compared with their counterpart. Other risk factors of self-reported diabetes included age, gender, ethnicity, education level and marital status with those aged 55–64 years, females, who belong to Moor ethnicity, had secondary or higher education, and divorced or widowed had higher risk of diabetes compared with their counterparts.

Conclusions Sri Lanka has a high prevalence of self-reported diabetes and it differs by province, sector of residence, sex, education level, ethnicity, age and marital status. There is a need for risk stratifying adult Sri Lankan population to help target those at high risk of diabetes so that cost-effective approaches of screening, preventive and treatment options can be initiated.

What is already known on this topic

  • According to existing literature, Sri Lanka has a high prevalence of diabetes. However, there is a lack of evidence generated using nationally representative data particularly on diabetes prevalence and risk factors associated with diabetes among Sri Lankans.

What this study adds

  • The age-adjusted national prevalence of self-reported diabetes was 10.6% in 2016, higher than the national-level studies conducted prior to 2016. The risk factor analysis shows that those who are in age group 55–64 years, female, belong to Moor ethnicity, had secondary or higher education, living in urban areas and divorced or widowed had higher risk of diabetes.

How this study might affect research, practice or policy

  • Our study highlights that the estimated prevalence of self-reported diabetes in Sri Lanka differs by province, sector of residence, sex, education level, ethnicity, age and marital status. It emphasises the need for risk stratifying the adult Sri Lankan population for diabetes so that a more cost-effective approach of screening and optimised preventive and treatment options for diabetes in the country can be initiated.

Introduction

Diabetes is a major global public health burden.1 According to the latest figures from the International Federation of Diabetes (IDF), there are 537 million people aged 20–79 years living with diabetes globally.1 Approximately 90% of people with undiagnosed diabetes reside in low-income and middle-income countries.1

Sri Lanka is a lower middle-income country in South Asia (SA) with significant public health burden of non-communicable diseases.2 Sri Lanka also has many healthcare challenges pertaining to diabetes prevention and management. Recent studies have identified that shortage of physician, lack of health education and poor technology are the main challenges in the Sri Lankan healthcare system.3 4 However, the primary care infrastructure in Sri Lanka is still in its infancy,2 and community-level optimisation of medical care and screening for conditions such as diabetes is required to identify people with undiagnosed diabetes to initiate treatment early. Given the lack of resources in the healthcare system, there is a need to identify the strata of the population with higher risk of diabetes that would benefit most from the preventive measures and optimisation of diabetes care in Sri Lanka. Prevalence of diabetes seems to vary by the level of urbanisation and ethnic differences in different regions.5 6 There are regional variations in sociodemography in Sri Lanka with some areas being more urbanised than others. The urban/rural classification in Sri Lanka has three sectors as urban, rural and estate where estate sector is defined as tea plantation areas that are 20 acres or more and with 10 or more resident labours.7 Estate sector is the lowest socioeconomic region with poor infrastructure compared with urban and rural sectors.8 There are only 18.2% of the population who reside in urban areas of the country and the majority of the population live in rural sector (77.4%) and the rest in the estate sector (4.4%).9 Moreover, Sri Lanka has regional variation in its ethnic composition with some areas populated more by Sinhalese and others mainly by Tamils, and some areas have a mix of all three main ethnic groups that include Sinhalese, Tamils and Moors.10

Existing literature on the prevalence of diabetes and its complications in Sri Lanka is mainly restricted to certain areas or communities.2 11–15 Due to the heterogeneous distribution of ethnicities and urbanisation across the provinces in Sri Lanka, these studies do not accurately reflect the national prevalence of diabetes and it is not possible to extrapolate the public health burden of diabetes in the country. Furthermore, previous studies were conducted at different time spans using various methods of sampling.2 11 A recent study conducted using the Sri Lanka Health and Ageing Survey (SLHAS) 2018/2019 has collected data from all the nine provinces in Sri Lanka; however, it has not explored the risk factors of diabetes among Sri Lankans.12 In this context, there is an emerging need to identify the national and regional prevalence of diabetes and risk factors for diabetes among Sri Lankans using a nationally representative sample of population.

Therefore, we aimed to identify and describe the demographic risk factors of diabetes among Sri Lankans and to establish the national and provincial-level prevalence of self-reported diabetes using a nationally representative sample obtained from the Sri Lanka Demographic and Health Survey (SLDHS)–2016.

Method

Study design and participants

Data were extracted from three sources in this study. Data on self-reported diabetes status and the sociodemography were extracted from SLDHS 2016.16 Detailed sampling procedure for this survey is explained elsewhere.16 In brief, this survey used a multistage stratified area probability sampling with two-stage stratified sampling design.16 Population was first stratified by district and then categorised into urban, rural and estate sectors within each district.16 A total of 28 800 housing units were selected for the survey.16 A household questionnaire was completed from each selected household to collect information on demographic and socioeconomic parameters and behavioural factors. A total of 106 466 individuals were included in this survey. Among them, 71 130 individuals were aged 20 years and above. We mirrored the age group defined by the IDF Diabetes Atlas.1 The self-reported diabetes status in the questionnaire was used to define people with self-reported diabetes. Individuals whose diabetes status was recorded as unknown were excluded from the sample. This resulted in a sample of 71 066 individuals (age 20 years and above) with self-reported diabetes status distributed across all nine provinces in Sri Lanka (figure 1). The gross domestic product (GDP) data were obtained from the economic and social statistics reports of Sri Lanka 2017 and 2018 released by the Central Bank of Sri Lanka.7 ,17 The population data for each province were obtained from the reports of the Census of Population and Housing–2012 Sri Lanka (CPH) as they are the most recent census data available in Sri Lanka.

Figure 1
Request permissionrequest permission icon
Figure 1

Sample selection criteria.

Patient and public involvement

In this project, patients and the public were not directly involved.

Outcome

The outcome was defined as self-reported prevalence of diabetes status. We used the SLDHS question ‘Did any member of your household suffer from diabetes during the last 12 months?’ to evaluate the diabetes status, with possible answers yes, no and do not know (64; 0.06%).

Demographic risk factors

Individual-level data on age, sex, ethnicity, religion, marital status, education status, self-reported sector of the resident, province of residence and diabetes status were obtained from SLDHS 2016.

Age was recorded as an integer between 20 and 99. Education level was recorded as none/primary or secondary and education above secondary level. Province was recorded as one of the nine provinces in Sri Lanka. Marital status was classified as married, unmarried, divorced/widowed and unknown. Ethnicity of the individuals was recorded as Sinhala, Moor, Tamil and other categories. Religion was recorded as Buddhist, Christian/Catholic, Islam, Hindu or other categories.

Population-level data on economic status were obtained from economic and social statistics reports of Sri Lanka released by the Central Bank of Sri Lanka.7 These data were reported as GDP per capita data for each province, and we used 2016 data for the analysis.

Statistical methods

Age-adjusted provincial and national-level prevalence of self-reported diabetes was calculated by sex, using the CPH 2012 Sri Lanka data. Age-adjusted prevalence was obtained by multiplying the crude age-specific prevalence of diabetes by age-specific weights. Weights were calculated as the proportions of the CPH 2012 population data for the respective age categories. Then, province-wise, district-wise and sector-wise diabetes prevalence was calculated using the sample data. As the data satisfied the normality condition, the estimates were expressed as means with SDs or proportions, as appropriate. Initially, all the variables were considered as categorical variables and Χ2 test and Fisher’s exact test were used to assess the relationship with diabetes status and other socioeconomic variables as appropriate. First, we examined the univariable association between each covariate and diabetes status. Next, we fitted a multivariable logistic regression model using stepwise backward and forward procedure using all the covariates. Then, we used all the possible interaction terms and built models with the covariates. We checked the accuracy of the fitted models using the area under the receiver operating characteristic curve (AUC). Interactions did not reveal any significant difference in AUC. Therefore, we used the model without interactions as the optimal model. We also checked for multicollinearity using variance inflation factor and goodness of fit using Hosmer-Lemeshow test to confirm that there was no multicollinearity and good fit of the final model to the data. ORs derived from this optimal model were then used to examine the relationship between the covariates and outcome (diabetes status). We considered p values less than 0.05 to be significant. R V.4.1.1 was used for all statistical analyses.

Results

General description of study participants

A total of 71 066 individuals with data on diabetes status were included in the study and their mean age was 45.6 (SD 16.4) years. Table 1 shows the number of individuals with and without diabetes with respective covariate. A high prevalence of self-reported diabetes was observed in the age group 55–64 years (5476 of 8785, 38.4%) in females (5432 of 32 911, 14.2%), in those categorised as divorced/widowed (2319 of 6189, 27.3%), Moors (783 of 4883, 13.8%), Muslims (835 of 6116, 13.7%) and in individuals with low education attainment (2810 of 11 909, 19.1%). Table 2 shows the distribution of diabetes by ethnicity among the nine provinces. Among Sinhalese (1931 of 15 398, 12.54%), Tamils (622 of 8377, 7.43%) and Moors (228 of 6153, 3.71%), Western province, Northern province and Eastern province showed the highest prevalence of self-reported diabetes.

Table 1
|
Characteristics of the self-reported diabetes population
Table 2
|
Distribution of the self-reported diabetes by ethnicity and province

General characteristics of the study population by province

The online supplemental table 1 shows the provincial distribution of sociodemographic characteristics of the study population. Western province was the most urban region and North Central province was the most rural region. In terms of ethnic composition, majority of population in Northern and Eastern provinces were Tamil, and majority of the Moors were distributed in Eastern province while Sinhalese were the majority in all other regions.

National and provincial prevalence of self-reported diabetes

The overall age-adjusted prevalence of self-reported diabetes in people aged 20 years or above was 10.6%. When considering the age-adjusted provincial prevalence of self-reported diabetes, the Western province showed the highest prevalence (13.3%), and this province has the highest GDP per capita (734 094 Sri Lankan rupee). Northern province had the lowest GDP (321 289 Sri Lankan rupee) and had the lowest prevalence of diabetes (8.16%) (figure 2A,B).

Figure 2
Request permissionrequest permission icon
Figure 2

Prevalence of diabetes compared with the gross domestic product (GDP) of each province in Sri Lanka. (A) Distribution of weighted prevalence of diabetes by province; (B) distribution of GDP by province; (C) distribution of weighted diabetes prevalence by females; (D) distribution of weighted diabetes prevalence by males.

The distribution of diabetes by sex in each province is shown in figure 2C,D and online supplemental table 2. The highest prevalence of self-reported diabetes for both sexes was observed in Western province with around 11.0% men and 16.7% women having diabetes, while the lowest prevalence was observed in Sabaragamuwa and North Central provinces for females (12.1%) and males (6.2%), respectively. In all provinces, self-reported diabetes prevalence was higher in females by approximately 5% compared with males (online supplemental table 2). Online supplemental table 3 shows the prevalence of self-reported diabetes by province, district and sector. In general, prevalence of diabetes was higher in urban sectors than in rural and estate sectors.

Figure 3 shows the sex comparison of diabetic prevalence in each age group. Prevalence of self-reported diabetes steadily increases with the age in both genders. The prevalence of self-reported diabetes showed a higher trend in females than in males. For both groups, the highest prevalence of self-reported diabetes was reported among the individuals aged 55 years and older.

Figure 3
Request permissionrequest permission icon
Figure 3

Prevalence of diabetes by age and sex in Sri Lanka.

Risk factors associated with self-reported diabetes

The multivariable logistic regression analysis showed that age, ethnicity, sex, level of education, urban/rural status and province of residence were independent risk factors for self-reported diabetes (table 3). People living in Western province were 64% more likely to have diabetes compared with other provinces. Compared with residents living in rural areas, those who are in urban areas had 19% increased odds of diabetes (OR 1.19, 95% CI 1.11, 1.28), whereas estate sector individuals had 16% lower odds of having diabetes compared with residents in rural areas (OR 0.84, 95% CI 0.72, 0.97). The odds of having diabetes among females were 72% higher than males (OR 1.72, 95% CI 1.62, 1.82). Moors had 43% increased odds of diabetes compared with Sinhalese (OR 1.43, 95% CI 1.30, 1.58). Tamils had 5% lower odds of having diabetes compared with Sinhalese (OR 0.95, 95% CI 0.85, 1.06). Individuals who had a secondary education qualification or above had 10% increased odds of having diabetes (OR 1.10, 95% CI 1.04, 1.17) than those with low education attainment.

Table 3
|
Multivariable logistic regression with diabetes status as the dependent variable

Discussion

This study used the data from one of the most recent national surveys to estimate the prevalence of self-reported diabetes among Sri Lankan adults. There are three major findings. First, the national age-adjusted prevalence of self-reported diabetes among Sri Lankan adults in 2016 was 10.6%. Second, we report the province-specific age-adjusted prevalence and we observed a high prevalence of self-reported diabetes in the provinces with a high GDP value, and a low prevalence in the provinces with a low GDP value with Eastern province showing a high prevalence in self-reported diabetes despite the low GDP value. This discrepancy is likely to be due to the majority of the Moor community, who have a higher risk of diabetes compared with other ethnicities.18 Third, the risk factors of self-reported diabetes included age group 55–64 years, female sex, living in urban area, Moor ethnicity, having a higher education level, and divorced or widowed.

According to the Sri Lanka Diabetes Cardiovascular Study (SLDCS) conducted among 5000 adults between 2005 and 2006, the age-adjusted national prevalence of diabetes in Sri Lanka was 10.3%,19 which is not significantly different to our current estimate of 10.6%. Although these studies cannot be compared as SLDCS data are from only four provinces of Sri Lanka, this finding suggests that overall prevalence of diabetes has not increased significantly between the years 2006 and 2016. This is likely to be due to 20% of Sri Lankans living in urban areas and urbanisation is still confined to the Western province that is home to 15% of the population of Sri Lanka. With further urbanisation, a higher prevalence is anticipated; as reflected in the most recent study of SLHAS, the age-adjusted prevalence of self-reported diabetes in Sri Lankan adults in 2019 is 14.3%.12 Hence, when the prevalence of self-reported diabetes is compared, it has increased by 3.7% between 2016 and 2019. Therefore, as suggested by both studies, it is evident that diabetes prevalence is increasing in Sri Lanka. Hence, it is timely to initiate diabetes prevention and treatment programmes at least in urban areas of Sri Lanka.

In our study, the multivariable logistic regression model reported significantly higher risk of self-reported diabetes among Moor ethnicity who is also known as Muslims and significantly lower risk of diabetes among Tamils compared with Sinhalese after adjusting for sex, age, marital status, education level, province of residence and sector of residence. It is difficult to compare these results directly with existing literature from Sri Lanka as they have used smaller samples that are not nationally representative and/or focused only on reporting diabetes prevalence without further analysis on risk factors. Therefore, considering literature from India, the state closest to Sri Lanka and one with considerable cultural and ethnic similarities shows similar results of Muslims having higher risk of diabetes compared with Hindu known as Tamils in Sri Lanka.20 In general, the difference of risk of diabetes among different ethnicities is likely to be due to different lifestyles, cultural influences, dietary habits and genetic predisposition.21–23

The SLDCS also showed a female predilection for diabetes. Increased frequency of diabetes in females has also been observed in other SA populations.24 As most of the females are housewives, this higher prevalence may reflect the sedentary nature of their roles. Similarly, we found that 55–64 years age group is at risk of diabetes. This may also be partly explained by the transition to sedentary retired life.19 In addition, other lifestyle changes such as dietary habits may also contribute to these findings and explain the preponderance of diabetes in those with higher education level, divorcees and widowers and Moor ethnicity.25 Similar findings can be observed as diabetes prevalence was more dominant in females, urban regions, among more affluent inhabitants and Muslims according to results of the SLHAS.12 SLHAS has also reported that prevalence of diabetes increased with age until age ≥70 years.12

Our study highlights the need for policymakers in Sri Lanka to plan initiatives for prevention of diabetes and initiate prevention programmes or optimise diabetes control among those at risk of diabetes which will be a cost-effective way of preventing diabetes and its complications. The findings in this study recommend initiating screening in the Western province.

This study has many strengths. It used a nationally representative sample to estimate the national prevalence of self-reported diabetes in Sri Lanka, and the sample of this study is the largest sample that has been considered in a cross-sectional study in the country to date. We also included rural, urban and estate populations to ensure generalisability. Except for the SLHAS, there is only one previous study that has illustrated the inequality in prevalence of diabetes in estate sector compared with rural and urban sectors.12 26 The trend of increasing prevalence of diabetes and the associated risk factors identified in our study mirror the reports from other SA countries.27

This study has some limitations. We have used the data from latest version of SLDHS which was conducted in 2016 and this might not reflect the current situation in the country. Also, the cross-sectional nature of the study does not permit for inferences of causality to be made. In addition, we only had access to self-reported diabetes status; therefore, undiagnosed diabetes was not considered in this study. Also, some important covariates, such as occupation, income level, dietary patterns and physical activity level, were not available in the DHS dataset. Hence, we could not assess the individual-level relationship between diabetes and these risk factors. We were also not able to report prevalence by type of diabetes due to the lack of information on this dataset.

Conclusion

Sri Lanka has a high prevalence of self-reported diabetes and it differs by province, sector of residence, sex, education level, ethnicity, age and marital status. There is a need for risk stratifying adult Sri Lankan population to help target those at high risk of diabetes so that cost-effective approaches of screening, preventive and treatment options can be initiated.

Overview

Article metrics
Altmetric data not available for this article.
Dimensionsopen-url