Multilevel analysis of noncommunicable diseases and associated factors among Ethiopian adults: evidence from the 2024 STEPS survey

Abstract

Background

Noncommunicable diseases (NCDs) such as cardiovascular disease, hypertension, and diabetes are a major and growing global health burden, particularly in low- and middle-income countries including Ethiopia. However, evidence on their distribution and associated individual- and community-level factors remains limited. Therefore, this study aimed to address this gap and generate evidence to inform prevention strategies and public health policy.

Methods

Nationally representative cross-sectional study was conducted. Data were collected using the standardized WHO STEPwise approach. A multistage, stratified cluster sampling method was employed. In the first stage, 728 enumeration areas (EAs) were selected as primary sampling units. In the second stage, a new listing of households was collected from each EA. From these updated lists, 15 households were randomly selected, and one individual aged 18–69 years was subsequently selected at random from each household. Among the 34,594,270 weighted households included in the study, 29,546,369 completed interviews and were included in the analysis. Multilevel logistic regression analysis was employed to identify factors associated with noncommunicable diseases. A series of four models were fitted, and the model with the lowest Bayesian Information Criterion (BIC), and Akaike’s Information Criterion (AIC) values was selected as the best-fitting model for interpretation.

Results

In this study, the prevalence of self-reported history of cardiovascular disease, hypertension, and diabetes was 5.4%, 17.3%, and 4.2%, respectively. Age ≥44 (adjusted OR (AOR) 1.6; 95% CI 1.13–2.22)), Female (AOR 1.7; 95% CI 1.17–2.38)), obese (AOR 2.6; 95% CI 1.34–4.98)), normal weight (AOR 0.6; 95% CI 0.42–0.91), and high altitude (AOR 2.3; 95% CI 1.02–5.16), were significantly associated with self-reported history of cardiovascular disease. Besides, aged 30–44 (AOR 1.9; 95% CI 1.57–2.31), age 45–59 (AOR 3.9; 95% CI 3.09–4.84), age ≥60 (AOR 7.2; 95% CI: 5.54–9.41), divorced/widowed (AOR 1.6; 95% CI: 1.21–2.20), informal employment (AOR 0.8; 95% CI 0.62–0.96), normal BMI (AOR 1.7; 95% CI 1.37–2.07), overweight (AOR 3.1; 95% CI 2.39–4.04), obesity (AOR 5.2; 95% CI 3.61–7.39), urban residence (AOR 1.5; 95% CI 1.17–1.82), moderate altitude (AOR 1.4; 95% CI 1.08–1.70), and high altitudes (AOR 1.4; 95% CI 1.01–2.01) were significantly associated with hypertension. Aged 30–44 (AOR 2.1; 95% CI 1.40–3.11), age 45–59(AOR 3.6; 95% CI 2.35–5.63), age ≥ 60(AOR 5.1; 95% CI 3.11–8.40), overweight (AOR 1.7; 95% CI 1.08–2.60), obese (AOR 3.1; 95% CI 1.78–5.45), urban residence (AOR 1.5; 95% CI 1.02–2.10), and those live in high altitude (AOR 0.5; 95% CI 0.28–0.96) were significantly associated with diabetes.

Conclusion

The findings indicate a substantial burden of noncommunicable diseases in Ethiopia, particularly hypertension, followed by self-reported history of cardiovascular disease and diabetes. The study identified many core individual- and community-level factors influence noncommunicable diseases. Therefore, targeted interventions focusing on factors such as age, sex, obesity, and place of residence may contribute to reducing the burden of noncommunicable diseases. Strengthening national noncommunicable diseases prevention and control programs, including routine screening for cardiovascular disease, hypertension, and diabetes across all levels of the health system, may improve early detection and management.

Introduction

According to the World Health Organization (WHO), noncommunicable diseases (NCDs) account for approximately 41 million deaths annually, representing about 74% of all global deaths, with nearly three-quarters occurring in low- and middle-income countries [1]. It is further estimated that around 17–18 million people die from noncommunicable diseases before the age of 70 years each year, of whom approximately 86% are in low- and middle-income countries, which bear the greatest burden of NCD mortality. Noncommunicable diseases constitute the leading cause of death globally, with the greatest impact observed in low- and middle-income countries [2–4]. Cardiovascular diseases remain the leading cause of noncommunicable diseases–related deaths, followed by cancers and chronic respiratory diseases. Diabetes and hypertension are also major noncommunicable diseases, affecting approximately 34.2 million and 1.28 billion people worldwide, respectively [5–10]. Diabetes is a chronic metabolic disease characterized by elevated blood glucose (blood sugar) levels, which over time can cause serious damage to the heart, blood vessels, eyes, kidneys, and nerves [11, 12]. Therefore, examining these outcomes together with a history of cardiovascular disease provides a more complete picture of noncommunicable diseases risk and the overall disease burden. The burden of noncommunicable diseases is driven by multiple interacting determinants, including behavioral risk factors, biological changes, and social and environmental conditions [13–15]. In addition, understanding noncommunicable diseases requires evidence that can capture both individual-level exposures and the broader context in which people live, since community and environmental factors may shape health behaviors and biological risk in addition to personal characteristics [16].

In addition to the high prevalence of infectious diseases [17], noncommunicable diseases are increasingly recognized as major contributors to national health challenges in Ethiopia [18–20]. However, the distribution and determinants of noncommunicable diseases are unlikely to be uniform across the country. Differences across regions and between urban and rural areas can shape health behaviors, access to health care, nutritional practices, physical activity patterns, and exposure to locally prevalent factors [21, 22]. Such variations suggest that noncommunicable disease risks may cluster at the community level, meaning that the same individual risk factors may produce different effects depending on the environment and context. A key challenge in existing evidence is that many analyses focus only on individual characteristics and do not adequately account for clustering or community-level influences for noncommunicable diseases. Multilevel analytical approaches can address this limitation by simultaneously modeling individual and community determinants, allowing researchers to distinguish how much of the variability in noncommunicable diseases outcomes is attributable to individual characteristics versus the setting in which individuals reside. In addition to commonly examined demographic factors such as age and education, both biological and behavioral determinants play a key role in shaping noncommunicable disease (NCD) outcomes [23–25]. Body mass index (BMI), for example, is a well-established biological indicator strongly associated with metabolic and cardiovascular risk [26–28]. Moreover, contextual factors such as altitude and place of residence (urban or rural) may also influence health outcomes by shaping environmental exposures, lifestyle behaviors, and access to healthcare services. To generate nationally representative and standardized evidence, this study uses data from the 2024 WHO STEPS Survey conducted across Ethiopia. The STEPS approach supports reliable measurement of behavioral risk factors, physical indicators, and biochemical markers using comparable methods across regions. The study aimed to estimate the prevalence of self-reported cardiovascular disease, hypertension, and diabetes, and to identify their individual- and community-level determinants. The analysis evaluates the associations of selected individual-level factors (age, education level, marital status, occupation, and BMI) while accounting for community-level influences (including residence type and altitude) using multilevel modeling. By identifying individual and community determinants of NCD outcomes, this study seeks to support evidence-based prevention and control strategies in Ethiopia. The results may help policymakers and public health stakeholders’ better target high-risk groups and tailor interventions to the multi-layered reality in which NCD risks develop and persist.

Methods

Study design

Nationally representative cross-sectional study was conducted between July 1 and October 30, 2024, encompassing all 12 regional states and 2 city administrations of Ethiopia. Data were collected using version 3.2 of the noncommunicable disease (NCD) STEPS risk factor survey instrument, implemented in accordance with the World Health Organization (WHO) STEPwise approach to NCD surveillance.

Study population, inclusion and exclusion criteria

Adults aged 18 to 69 who lived in the household for at least 6 months prior to the survey were eligible for random selection to be included in the study. Pregnant women were not included in physical assessments such as measurements of weight; height, hip circumference, and spot urine collection because pregnancy involves physiological and behavioral changes (including differences in diet, healthcare utilization, medication use, and stress) that may influence the outcomes of interest and introduce variability. Additionally, certain indicators including symptoms, biomarkers, and medication use, may require pregnancy-specific interpretation, potentially affecting consistency and comparability. Their exclusion therefore helps to reduce misclassification, improve the clarity of findings, and ensure better comparability across the nonpregnant adult population. Moreover, individuals with physical conditions that prevented accurate physical measurements were also excluded from the study.

Sample size and sampling procedure

The sample size was calculated using the age- and sex-specific approach recommended in the updated 2020 WHO STEPS Surveillance Manual. Participants were stratified into four age groups (18–29, 30–44, 45–59, and 60–69 years) and by sex. A single population proportion formula was applied, assuming a prevalence (P) of 50% to maximize variability (except for Oromia and Amhara regions). The calculation used a 95% confidence level (CI) (Z = 1.96) and a margin of error of 5%. To account for clustering, a design effect of 1.5 was applied, and the sample size was further adjusted to ensure adequate precision across age-sex groups and regional estimates. A 15% nonresponse rate was added in line with WHO STEPS recommendations. For the more populous regions of Oromia and Amhara, the same assumptions were retained, except that the margin of error was reduced to 4% and the design effect increased to 2 to improve precision.

A multistage, stratified cluster sampling design was employed using the 2019 sampling frame developed by the Ethiopian Statistical Service. The data follow a hierarchical sampling structure of enumeration areas (EAs), households, and individuals. In the first stage, enumeration areas (EAs) were selected as primary sampling units. A total of 540 EAs were selected from 12 regions (45 per region), with additional sampling from Oromia and Amhara regions (94 EAs each), resulting in a total of 728 EAs. Each EA was delineated using Q-field software. In the second stage, a new listing of households was collected from each EA. From these updated lists, 15 households were randomly selected per EA, yielding a total planned sample of 10,924 households. Within each selected household, all eligible individuals were listed using the eSTEPS application, and one participant was randomly selected per household for inclusion in the survey. Because exactly one adult was sampled per household, the individual-level and household-level variances are structurally confounded at the analysis stage, meaning distinct household-level clustering is not analytically relevant or identifiable. Consequently, the analytical hierarchy for the multilevel models was specified as a two-level structure: individuals (Level 1) nested within enumeration areas (Level 2). In this framework, enumeration areas (EAs) were treated as random effects. Broader geographic strata (regions) were included as fixed effects to capture community level variations. This analytical specification precisely aligns the modeling framework with the empirical distribution of the data.

Data collection procedure

Data collection followed the standardized WHO STEPwise approach using the STEPS survey instrument (version 3.2) [29], adapted to the national context to ensure cultural relevance and applicability. The adaptation included a country-specific module on khat use and optional modules on cervical cancer, mental health (depression), and chronic respiratory diseases, as endorsed by the Technical Working Group in consultation with the Ministry of Health. The STEPS approach comprises three sequential components: (1) a structured questionnaire on behavioral risk factors, (2) physical measurements, and (3) biochemical assessments. Each participant was assigned a unique identification number and a quick response (QR) code to enable linkage of data across all three steps, including laboratory results.

Step 1: interview questionnaire

Data were collected through face-to-face interviews using a structured questionnaire to capture demographic and behavioral characteristics. Data collectors were equipped with digital maps and electronic questionnaires to support data collection. Key risk factors assessed included tobacco and alcohol use, dietary practices (fruit, vegetable, and salt intake), and physical activity. To enhance understanding and accuracy of responses, visual show cards were used during interviews (e.g., types of tobacco and alcohol products, portion sizes, and salt quantities). The questionnaire incorporated core, expanded, and optional modules, including mental health (depression), chronic respiratory diseases, and khat use. Interviews were conducted in six local languages: Afarigna, Afan Oromo, Amharic, Sidamigna, Somaligna, and Tigrigna.

Step 2: physical measurements

Physical measurements were conducted immediately after the interview at the household level. Blood pressure and heart rate were measured three times at 3-min intervals using a digital BOSO Medicus blood pressure monitor, and the average of the three readings was used for analysis. Anthropometric measurements included height, weight, and waist circumference. Height was measured using an ultrasonic height device, while weight was measured using a digital scale. Waist circumference was measured using a standard tape. These measurements were used to calculate body mass index (BMI) and assess overweight and obesity.

Step 3: biochemical measurements

Biochemical assessments involved the collection of blood and urine samples. Participants were instructed to fast for at least 8 h prior to testing. Blood samples were collected via finger prick in the morning and analyzed using a CardioChek point-of-care device to measure blood glucose and lipid levels. Spot urine samples were also collected. Participants on diabetes medication were advised to bring their medications and take them after sample collection. Data for Steps 1, 2, and 3 were collected at different time points and later merged using unique participant identifiers and QR codes. The survey was implemented by 32 field teams, each comprising a laboratory technologist and a public health professional. Data collection was conducted in three phases: interview and physical measurements by both team members, followed by biochemical assessment by the laboratory technologist. Field activities were supervised by eight coordinators. All personnel received one week of standardized training prior to data collection. Data quality was ensured through regular supervision, use of standardized checklists, and weekly centralized monitoring with feedback provided to field teams.

Study variables

Response Variables: The dependent variables of this study are self-reported history of cardiovascular diseases, raised blood glucose (diabetes), and raised blood pressure (hypertension).

History of cardiovascular disease: This variable was assessed using self-reported information. Participants were asked a single yes/no question on whether they had ever been diagnosed with or experienced any of the following conditions: myocardial infarction (heart attack), angina pectoris (chest pain due to heart disease), or stroke (cerebrovascular accident).

Hypertension (raised blood pressure): In this study, participants were classified as having raised blood pressure (hypertension) if systolic blood pressure (SBP) ≥ 140 mmHg and/or diastolic blood pressure (DBP) ≥ 90 mmHg, or if they were currently using antihypertensive medication (Yes). Participants with SBP < 140 mmHg and DBP < 90 mmHg who were not using antihypertensive medication were classified as not having raised blood pressure (No). Blood pressure was measured using a validated automatic sphygmomanometer after the participant had been seated and rested for at least 5 min; three measurements were taken 1–2 min apart and the mean of the last two readings was used. This process is based on the WHO guideline [30].

Diabetes mellitus (raised blood glucose): Participants were classified as having raised blood glucose (diabetes) according to WHO guideline values [31] if fasting plasma glucose (FPG) ≥ 126 mg/dL or if they reported current use of glucose-lowering medication (Yes). Participants with FPG < 126 mg/dL and not reporting current use of glucose-lowering medication were classified as not having raised blood glucose (No).

Explanatory variables: The predictor variables of this study were sex, age, educational level, marital status, BMI, residence type, and altitude.

Operational definitions

Individual-level variables

In this study, individual-level variables are characteristics measured on each participant that capture personal demographics, behaviors, clinical status, biological measures, and psychosocial factors relevant to the risk, presence, management, or outcomes of noncommunicable diseases. These variables are analyzed as exposures, outcomes, confounders, or effect modifiers in noncommunicable diseases and are distinct from area- or system-level variables. These are:

Age was treated as a categorical variable grouped into four categories: less than 30 years, 30 to 44 years, 45 to 59 years, and 60 years or older. However, given the low prevalence of self-reported cardiovascular disease, age was dichotomized (< 45 vs. ≥45 years) to improve the robustness and stability of the analysis.

Sex was coded as a binary variable, with males and females.

Educational status was categorized into three levels: no formal education, primary education, and secondary and higher education.

Marital status was categorized into three groups for analysis: single (never married), currently in union (married or cohabiting), and currently not in union (separated, divorced, or widowed). This grouping is consistent with previous epidemiological studies of sociodemographic determinants of hypertension and allows for adequate sample sizes within each category for multivariable and multilevel modeling.

Occupation was categorized according to employment type, typically grouped into unemployed, informal employment, and formal employment.

Body mass index (BMI) is a continuous measure derived from an individual’s weight and height, calculated as weight in kilograms divided by height in meters squared (kg/m²). In this study, body mass index (BMI) was categorized in accordance with established classifications as follows: underweight (< 18.5 kg/m²), normal weight (18.5–24.9 kg/m²), overweight (25.0–29.9 kg/m²), and obese (≥ 30.0 kg/m²), consistent with previous studies [32].

Community-level variables

In this investigation, community-level variables included type of residence (urban or rural) and altitude of residence.

Place of residence: This was categorized based on administrative classification as urban or rural. Altitude was included as a contextual environmental factor reflecting the elevation of the participant’s place of residence.

Altitude: This refers to the height above sea level at which a community is located, affecting environmental conditions and potentially influencing health outcomes due to variations in air quality and temperature. In our analysis, we categorized altitude into four groups: low (< 1500 m), moderate (1500–2500 m) and high (> 2500 m).

Data analysis and procedures

Traditional regression models may underestimate standard errors and produce biased estimates when applied to hierarchically structured data. To address this limitation, a multilevel logistic regression model was employed. The data had a hierarchical structure, with households nested within primary sampling units (enumeration areas), which were further organized within strata. Consequently, observations within the same cluster are likely to be more similar to each other than to those from different clusters, violating the assumptions of independence and homoscedasticity underlying standard regression models. The multilevel approach accounts for this clustering and between-group heterogeneity by incorporating both fixed effects (individual- and community-level covariates) and random effects (unobserved cluster-level variation), thereby producing more reliable standard error estimates and improved model accuracy for analyzing noncommunicable disease outcomes within hierarchical data structures. The data were checked for completeness and subsequently imported into Stata version 16 for analysis. Descriptive statistical analyses, including the computation of frequencies and percentages, were performed to summarize the characteristics of the study variables. The presence of multicollinearity among independent variables was assessed using the Variance Inflation Factor (VIF) as supported by many books [33, 34]. All independent variables demonstrated acceptable independence, with individual VIF values below the threshold of 5 and a mean VIF of 1.76. This confirmed the absence of significant multicollinearity in the final model. Furthermore, model fit was assessed using statistical criteria, including Akaike’s Information Criterion (AIC), Bayesian Information Criterion (BIC) [35–37]. To examine the associations between individual- and community-level factors and noncommunicable diseases, binary and multilevel logistic regression analyses were performed. Variables with a p-value < 0.05 in the bivariable analysis were subsequently included in the multilevel regression model. The data were weighted to ensure representativeness of the national population, thereby enabling reliable statistical estimation. Sampling weights were constructed through a multistage process in collaboration with WHO experts. First, a base sampling weight (SW) was calculated as the inverse of the probability of selection of each participant through the multistage cluster sampling design. Second, nonresponse adjustment weights were computed separately for Step 1 (NRW1), Step 3 (NRW3), and laboratory participation (NRW_Lab) to compensate for differential participation across survey components. Third, population distribution weights (PDW1, PDW3) were derived to align the age-sex distribution of respondents with that of the Ethiopian target population. Final analysis weights were then calculated as the product of the sampling weight, the relevant nonresponse adjustment weight, and the corresponding population distribution weight. Thus, Wstep1 = SW × NRW1 × PDW1 for questionnaire analyses, Wstep3 = SW × NRW3 × PDW3 for biochemical measurements, and Wsteplab incorporated the laboratory-specific nonresponse adjustment for analyses based on laboratory biomarkers. This weighting procedure ensured that estimates were representative of the Ethiopian population and accounted for unequal selection probabilities and differential nonresponse.

Four multilevel logistic regression models were fitted. Model 1 (null model) included no covariates and estimated the random intercept at the cluster level to assess variation in noncommunicable diseases across communities. Model 2 incorporated individual-level variables (sex, age, education, occupation, marital status, and BMI), while Model 3 included community-level factors (altitude and residence). Model 4 combined both individual- and community-level variables. All models accounted for the intraclass correlation (ICC) of the outcome variables [38], ensuring valid standard error estimates. ICC estimates were calculated using the standard formula for multilevel logistic regression models as follows.

where $={{\mathrm{\sigma }}^2}_u\text{represents}\text{between}(\text{level}-2)\text{variance},$ and $\frac{{\pi }^2}{3}$ (≈ 3.29) is the assumed level-1 variance. The variability on the odds of noncommunicable diseases explained by successive models was calculated by proportional change in variance (PCV) as:$=\frac{{{\delta }^2}_{\mathit{null}}-{{\delta }^2}_{\mathit{sm}}}{{{\delta }^2}_{\mathit{null}}}\times 100\%,where{{\delta }^2}_{\mathit{null}}$: is the variance in cardiovascular disease, hypertension and diabetes in the null model, and ${{\delta }^2}_{\mathit{sm}}$ is the variance in the successive models.

Results

Demographic characteristics of participants

Of the 34,594,270 weighted households included in the study, 29,546,369 were interviewed, resulting in a response rate of 85.4%. Above one-third (43.4%) of participants were younger than 30 years. Participants aged 45–59 years and ≥60 years accounted for 18.5% and 6.2% of the population, respectively. In terms of educational attainment, 36.8% of participants had primary education and 36.2% had no formal education. The majority of participants had a normal body mass index (BMI) (68.9%). Underweight individuals accounted for 19.8% of the population, whereas 8.6% were overweight and 2.8% were obese. Nearly three-fourth (73%) of participants resided in rural areas (Table 1).

Table 1.

Demographic characteristics of participants

Study variables	Unweighted n (%)	Weighted N (%)
Individual-level characteristics
Sex
Men	3,785 (40.6)	16,430,038(55.6)
Women	5,545 (59.4)	13,116,331(44.4)
Age
< 30	2,905 (31.1)	12,813,546(43.4)
30 – 44	3,709 (39.8)	9,422,660(31.9)
45 – 59	1,800 (19.3)	5,480,018(18.5)
>= 60	916 ( 9.8)	1,830,145(6.2)
Education
No formal education	3,447 (37)	10,695,586(36.2)
Primary education	3,022 (32.4)	10,870,171(36.8)
Secondary and higher education	2,858 (30.6)	7,970,010(27)
Marital status
Single	1,282 (13.8)	5,667,511(19.2)
Currently in union	6,675 (71.9)	21,737,899(73.7)
Currently not in union	1,333 (14.3)	2,104,370(7.1)
Occupation
Formal employment	1,729 (18.8)	3,344,948(11.4)
Informal employment	3,911 (42.5)	16,028,622(54.7)
Unemployed/not in active employment	3,556 (38.7)	9,910,577(33.8)
BMI category
Underweight	1,416 (15.7)	5,745,483(19.8)
Normal	6,035 (66.9)	19,992,017(68.9)
Overweight	1,090 (12.1)	2,483,219(8.6)
Obese	480 ( 5.3)	803,150(2.8)
Residence
Rural	5,659 (60.7)	21,570,319(73)
Urban	3,671 (39.3)	7,976,050(27)
Altitude
Low (< 1500 m)	2,642 (32.2)	3,185,583(11.9)
Moderate (1500–2500 m)	4,700 (57.3)	18,928,588(70.8)
High (> 2500)	860 (10.5)	4,623,816(17.3)

National prevalence of noncommunicable diseases among Ethiopian adults

In this study, the national self-reported prevalence of cardiovascular disease among the participants was 5.4%. In addition, the prevalence of hypertension and diabetes was 17.3% and 4.2%, respectively (Table 2 and Fig. 1).

Table 2.

Prevalence of noncommunicable diseases among Ethiopian adults

Study variables	Unweighted sample size	Missing data n (%)	Target population Size	Weighted prevalence (%)	95% CI
Hypertension	9,330	0 (0.0%)	29,546,368	17.3	17.20–19.93
Self-reported history of cardiovascular disease	9,330	0 (0.0%)	29,546,368	5.4	4.60–6.38
Diabetes mellitus	8,510	820 (8.8%)	27,258,390	4.2	3.19–4.28

Fig. 1.

National prevalence of self-reported history of cardiovascular, diabetes, and hypertension among participants

Prevalence of self-reported history of cardiovascular disease, diabetes, and hypertension with respect to type of residence

Hypertension, diabetes, and self-reported history of cardiovascular disease showed differing prevalence patterns across rural and urban settings. The highest prevalence of hypertension was observed in urban areas (22.1%), compared to rural areas (17.3%). Similarly, diabetes prevalence was higher in urban settings (4.5%) than in rural areas (3.4%). In contrast, self-reported history of cardiovascular disease was more prevalent in rural areas (5.7%) (Fig. 2).

Fig. 2.

Prevalence of self-reported history of cardiovascular disease, diabetes, and hypertension with respect rural and urban

Self-reported history of cardiovascular disease with respect to individual- and community-level characteristics

The weighted prevalence of self-reported cardiovascular disease (CVD) varied across individual- and community-level characteristics. self-reported history of CVD prevalence increased with age, rising from 5.5% among participants aged <30 years to 6.5% among those aged ≥60 years. Participants with no formal education had the highest prevalence of self-reported history of CVD (7%), followed by those with primary education (5.4%), while those with secondary or higher education had the lowest prevalence (3.4%). Occupational differences in self-reported history of CVD prevalence were observed, with the highest proportion among participants engaged in informal employment (6.7%), followed by those not actively employed (4%). The lowest prevalence was recorded among participants in formal employment (3.5%). Body mass index (BMI) showed a pronounced relationship with self-reported history of CVD. The prevalence was highest among underweight (10.7%) and obese individuals (9.9%), while it was lower among overweight (4.2%) and lowest among those with normal BMI (3.6%). In terms of residence, self-reported history of CVD prevalence was higher in rural areas (5.7%) compared to urban areas (4.7%) (Table 3).

Table 3.

Distribution of self-reported history of cardiovascular disease with respect to individual- and community-level characteristics

Characteristics	Cardiovascular disease
	Unweighted n (%)		Weighted N (%)
	No	Yes	No	Yes
Individual-level characteristics
Sex
Men	3,696 (97.6)	89 (2.4)	15,671,700(95.4)	758,337(4.6)
Women	5,341 (96.3)	204 (3.7)	12,272,540(93.6)	843,791(6.4)
Age
< 30	2,822 (97.1)	83 (2.9)	12,106,639(94.5)	706,907(5.5)
30 – 44	3,602 (97.1)	107 (2.9)	8,967,459(95.2)	455,200(4.8)
45 – 59	1,735 (96.4)	65 (3.6)	5,159,181(94.2)	320,837(5.9)
>= 60	878 (95.9)	38 (4.1)	1,710,961(93.5)	119,184(6.5)
Education
No formal education	3,306 (95.9)	141 (4.1)	9,946,055(93)	749,531(7)
Primary education	2,935 (97.1)	87 (2.9)	10,284,922(94.6)	585,248(5.4)
Secondary and higher education	2,793 (97.7)	65 (2.3)	7,702,661(96.7)	267,349(3.4)
Marital status
Single	1,262 (98.4)	20 (1.6)	5,509,559(97.2)	157,952(2.8)
Currently in union	6,458 (96.7)	217 (3.3)	20,423,633(94)	1,314,266(6.1)
Currently not in union	1,277 (95.8)	56 (4.2)	1,974,460(93.8)	129,910(6.2)
Occupation
Formal employment	1,686 (97.5)	43 (2.5)	3,227,779(96.5)	117,169(3.5)
Informal employment	3,772 (96.4)	139 (3.6)	14,960,153(93.3)	1,068,469(6.7)
Unemployed/not in active employment	3,451 (97)	105 (3)	9,516,489(96)	394,087(4)
BMI category
Underweight	1,341 (94.7)	75 (5.3)	5,132,671(89.3)	612,813(10.7)
Normal	5,898 (97.7)	137 (2.3)	19,269,865(96.4)	722,152(3.6)
Overweight	1,057 (97)	33 (3)	2,378,574(95.8)	104,645(4.2)
Obese	447 (93.1)	33 (6.9)	723,732(90.1)	79,417(9.9)
Residence
Rural	5,482 (96.9)	177 (3.1)	20,341,362(94.3)	1,228,957(5.7)
Urban	3,555 (96.8)	116 (3.2)	7,602,878(95.3)	373,172(4.7)
Altitude
Low (< 1500 m)	2,586 (97.9)	56 (2.1)	3,065,121(96.2)	120,462(3.8)
Moderate (1500–2500 m)	4,544 (96.7)	156 (3.3)	17,893,445(94.5)	1,035,143(5.5)
High (> 2500 m)	821 (95.5)	39 (4.5)	4,377,014(94.7)	246,802(5.3)

Distribution of hypertension with respect to individual- and community-level variables

Hypertension prevalence increased markedly with age, rising from 11.1% among participants aged <30 years to 36% among those aged ≥60 years. Educational status showed minimal variation, with prevalence ranging from 17.6% among individuals with primary education to 20.9% among those with secondary and higher education. A clear gradient was evident across BMI categories, with hypertension prevalence increasing from 12.6% among underweight individuals to 29.9% among obese participants. At the community level, hypertension was more prevalent in urban areas (22.1%) than in rural areas (17.3%). Similarly, prevalence increased with altitude, from 15.4% in low-altitude areas to 21.3% in high-altitude areas (Table 4).

Table 4.

Distribution of hypertension with respect to individual- and community-level characteristics

Characteristics	Hypertension
	Unweighted n (%)		Weighted N (%)
	No	Yes	No	Yes
Individual-level characteristics
Sex
Men	3,046 (80.5)	739 (19.5)	13,493,212(82.1)	2,936,825(17.9)
Women	4,398 (79.3)	1,147 (20.7)	10,567,026(80.6)	2,549,305(19.4)
Age
< 30	2,605 (89.7)	300 (10.3)	11,393,263(88.9)	1,420,283(11.1)
30 – 44	3,056 (82.4)	653 (17.6)	7,654,810(81.2)	1,767,850(18.8)
45 – 59	1,247 (69.3)	553 (30.7)	3,841,681(70.1)	1,638,336(29.9)
>= 60	536 (58.5)	380 (41.5)	1,170,484(64)	659,661(36)
Education
No formal education	2,698 (78.3)	749 (21.7)	8,781,897(82.1)	1,913,689(17.9)
Primary education	2,448 (81)	574 (19)	8,960,797(82.4)	1,909,374(17.6)
Secondary and higher education	2,295 (80.3)	563 (19.7)	6,306,943(79.1)	1,663,067(20.9)
Marital status
Single	1,117 (87.1)	165 (12.9)	4,956,477(87.5)	711,034(12.6)
Currently in union	5,386 (80.7)	1,289 (19.3)	17,638,989(81.1)	4,098,910(18.9)
Currently not in union	909 (68.2)	424 (31.8)	1,431,605(68)	672,765(32)
Occupation
Formal employment	1,343 (77.7)	386 (22.3)	2,663,134(79.6)	681,814(20.4)
Informal employment	3,202 (81.9)	709 (18.1)	13,320,604(83.1)	2,708,018(16.9)
Unemployed/not in active employment	2,807 (78.9)	749 (21.1)	7,880,442(79.5)	2,030,135(20.5)
BMI category
Underweight	1,232 (87)	184 (13)	5,020,200(87.4)	725,283(12.6)
Normal	4,887 (81)	1,148 (19)	16,225,514(81.2)	3,766,502(18.8)
Overweight	736 (67.5)	354 (32.5)	1,792,524(72.2)	690,695(27.8)
Obese	309 (64.4)	171 (35.6)	562,885(70.1)	240,265(29.9)
Residence
Rural	4,692 (82.9)	967 (17.1)	17,849,549(82.8)	3,720,770(17.3)
Urban	2,752 (75)	919 (25)	6,210,690(77.9)	1,765,360(22.1)
Altitude
Low (< 1500 m)	2,196 (83.1)	446 (16.9)	2,693,572(84.6)	492,011(15.4)
Moderate (1500–2500 m)	3,715 (79)	985 (21)	15,539,923(82.1)	3,388,666(17.9)
High (> 2500)	665 (77.3)	195 (22.7)	3,640,432(78.7)	983,385(21.3)

Distribution of diabetes with respect to individual- and community-level characteristics

In this study, the prevalence of diabetes mellitus (DM) increased with age, from 1.9% among participants aged <30 years to 8.4% among those aged ≥60 years. Marital status showed variation, with the highest prevalence observed among individuals not currently in union (6.8%), followed by those in union (3.7%), while single participants had the lowest prevalence (2.7%). Occupational disparities in the prevalence of diabetes were evident, with the highest prevalence among participants in formal employment (6.6%), compared to those in informal employment (3.5%) and those not in active employment (3%). A strong gradient was observed across BMI categories, with diabetes prevalence increasing from 3.5% among underweight individuals to 21% among obese participants. At the community level, diabetes prevalence was higher in urban areas (4.5%) compared to rural areas (3.4%). In contrast, prevalence decreased with increasing altitude, from 4% in low-altitude areas to 2.5% in high-altitude areas (Table 5).

Table 5.

Distribution of diabetes with respect to individual- and community-level characteristics

Characteristics	Diabetes
	Unweighted n (%)		Weighted N (%)
	No	Yes	No	Yes
Individual-level characteristics
Sex
Men	3,286 (94.8)	181 ( 5.2)	14,670,514(96.7)	501,136(3.3)
Women	4,742 (94)	301 ( 6)	11,580,534(95.8)	506,207(4.2)
Age
< 30	2,561 (97.1)	77 ( 2.9)	11,724,053(98.1)	228,539(1.9)
30 – 44	3,193 (95)	169 ( 5)	8,214,630(96)	344,411(4)
45 – 59	1,521 (91.5)	142 ( 8.5)	4,796,419(94.2)	295,219(5.8)
>= 60	753 (88.9)	94 (11.1)	1,515,946(91.6)	139,173(8.4)
Education
No formal education	2,921 (93.1)	215 ( 6.9)	9,359,923(95.2)	470,415(4.8)
Primary education	2,658 (95.3)	131 ( 4.7)	9,764,245(96.9)	316,707(3.1)
Secondary and higher education	2,447 (94.8)	135 ( 5.2)	7,116,459(97)	220,040(3)
Marital status
Single	1,125 (96.1)	46 ( 3.9)	5,143,434(97.3)	143,771(2.7)
Currently in union	5,734 (94.7)	323 ( 5.3)	19,285,199(96.4)	731,101(3.7)
Currently not in union	1,133 (91.1)	111 ( 8.9)	1,789,843(93.2)	131,037(6.8)
Occupation
Formal employment	1,368 (91)	136 ( 9)	2,732,948(93.4)	194,027(6.6)
Informal employment	3,461 (95.4)	166 ( 4.6)	14,331,364(96.5)	519,700(3.5)
Unemployed/not in active employment	3,097 (94.8)	170 ( 5.2)	8,981,494(97)	276,319(3)
BMI category
Underweight	1,244 (95.3)	62 ( 4.7)	5,103,002(96.5)	186,381(3.5)
Normal	5,353 (95.9)	226 ( 4.1)	18,059,746(97.2)	526,659(2.8)
Overweight	909 (91.1)	89 ( 8.9)	2,171,056(94.3)	132,232(5.7)
Obese	328 (78.8)	88 (21.2)	563,393(79)	149,421(21)
Residence
Rural	4,948 (94.9)	265 ( 5.1)	19,345,300(96.6)	678,809(3.4)
Urban	3,080 (93.4)	217 ( 6.6)	6,905,747(95.5)	328,534(4.5)
Altitude
Low (< 1500 m)	2,281 (94.9)	122 ( 5.1)	2,818,289(96)	116,605(4)
Moderate (1500–2500 m)	4,148 (95.2)	210 ( 4.8)	17,045,868(96.8)	562,654(3.2)
High (> 2500)	780 (97.1)	23 ( 2.9)	4,222,684(97.5)	107,827(2.5)

National prevalence of behavioral risk factors and screening history for noncommunicable diseases

The findings indicate that 27.5% of participants had ever consumed alcohol, with 23.9% reporting alcohol use in the past 12 months. Physical activity was relatively common in daily routines, with over half of participants engaging in walking or cycling for transport (56.5%) and moderate-intensity work (56.4%), while participation in recreational physical activities remained low. Blood pressure screening was limited, as only 33.5% had ever had their blood pressure measured by a health professional. Dietary habits were suboptimal, characterized by low fruit and vegetable intake and a high prevalence of salt consumption, with 75.7% of participants reporting always adding salt or salty sauces to their food. Among individuals with hypertension-related conditions, 39.4% reported taking antihypertensive medication, whereas 18% relied on traditional or herbal remedies (Table 6).

Table 6.

National prevalence of behavioral risk factors and screening history for noncommunicable diseases

Study variables	Response category	Unweighted sample size	Weighted Prevalence n(%)	95% CI
Alcohol consumption
Ever consumed any type of alcohol	Yes	9,330	2,570 (27.5)	25.88–29.27
Alcohol use in the past 12 months	Yes	9,330	2,233 (23.9)	22.32- 25.62
Physical activity
Heavy, vigorous-intensity work labor	Yes	9,330	2,149 (23)	21.28–24.88
Moderate-intensity work labor	Yes	9,330	5,258 (56.4)	54.59–58.12
Walking or using bicycle to get to and from places	Yes	9,330	5,270 (56.5)	54.69–58.25
Vigorous recreational sports/fitness	Yes	9,330	397 (4.3)	3.47–5.23
Moderate recreational sports/fitness	Yes	9,330	566 (6.1)	5.24–7.02
Blood pressure screening and treatment
Ever had blood pressure measured by a professional	Yes	9,330	3,128 (33.5)	31.92- 35.17
In the past 2 weeks took any drug for blood pressure	Yes	717	283 (39.4)	33.56- 45.56
Currently taking traditional/herbal remedies	Yes	717	130 (18)	13.84–23.39
Currently using any traditional healer for high blood pressure	Yes	717	130 (18)	13.84–23.39
Having family with a history of blood pressure	Yes	717	622 (6.7)	5.90 – 7.61
Dietary habits
Number of days eating fruit per week	0 days 1 day 2 days 3 days 4 to 7 days	8,235	1,714 (20.8) 1,820 (22.10%) 2,562 (31.11%) 1,207 (14.66%) 932 (11.32%)*	19.43–22.26 (20.59% to 23.69%) (29.17% to 33.10%) (13.28% to 16.15%) (10.15% to 12.60%)*
Number of days eating vegetables per week	0 to 1 days 2 days 3 days 4 days 5 days $\ge$ 6 days	8,545	3,134 (36.7) 2,757 (32.27%) 1,313 (15.37%) 444 (5.20%) 438 (5.12%) 459 (5.37%)	35.03–38.35 (30.47% to 34.12%) (14.08% to 16.76%) (4.61% to 5.86%) (4.52% to 5.79%) (4.83% to 5.96%)
Adds salt or salty sauce to food	Always Often Sometimes Rarely/Never	9,195	6,962 (75.7) 857 (9.32%) 782 (8.50%) 594 (6.45%)*	74.30–77.09 (8.50% to 10.22%) (7.54% to 9.58%) (5.73% to 7.25%)*
Eats high-salt processed foods	Always/Often Sometimes Rarely Never	7,503	368 (4.9) 792 (10.55%) 931 (12.41%) 5,412 (72.13%)	4.01- 5.99 (9.40% to 11.81%) (11.12% to 13.84%) (70.30% to 73.89%)

Multilevel logistic regression analysis of noncommunicable diseases

Self-reported history of cardiovascular disease

This part of multilevel logistic regression analysis identified both individual- and community-level factors associated with cardiovascular disease. In the final model (Model 4), which adjusted for both household- and community-level variables, several predictors remained statistically significant.

Random-effects analysis

Based on the Akaike Information Criterion (AIC) and Bayesian Information Criterion (BIC), Model 4 demonstrated the best overall fit, as it had the lowest values (AIC = 1838 and BIC = 1935.5. Therefore, all interpretations and conclusions regarding cardiovascular disease were based on Model 4. The intraclass correlation coefficient (ICC) was 53.3% in the null model, suggesting that a substantial proportion of the total variation in cardiovascular disease was attributable to differences between clusters. This value slightly declined to 50.2% in the final model. This could be due to strong clustering effects at the community level and possible homogeneity within clusters. The proportional change in variance (PCV) in Model 4 was 11.7%, indicating that both individual- and community-level variables explained a meaningful proportion of the between-cluster variation. The median odds ratio (MOR) decreased from 6.4 in Model 1 to 5.7 in Model 4, suggesting a reduction in unexplained cluster-level heterogeneity, although considerable variation persisted.

Fixed-effects analysis

Findings from Model 4 demonstrated that cardiovascular disease was significantly influenced by both individual- and community-level factors. Age was significantly associated with cardiovascular disease. Individuals aged ≥44 years had higher odds of cardiovascular disease (AOR = 1.6; 95% CI 1.13–2.22) compared to those aged <44 years. Female participants also had higher odds of cardiovascular disease (AOR = 1.7; 95% CI 1.17–2.38) than men. Regarding BMI, obese individuals had significantly higher odds of cardiovascular disease (AOR = 2.6; 95% CI 1.34–4.98) compared to underweight individuals. Individuals with normal weight had significantly lower odds of cardiovascular disease (AOR = 0.6; 95% CI 0.42–0.91) compared to underweight individuals. At the community level, altitude was significantly associated with self-reported history of cardiovascular disease. Participants residing at high altitude (> 2500 m) had higher odds of cardiovascular disease (AOR = 2.3; 95% CI: 1.02–5.16) compared to those living at low altitude (< 1500 m) (Table 7).

Table 7.

Multilevel logistic regression results for predictors of self-reported history of cardiovascular disease

Characteristics	Model 1 (AOR with 95% CI)	Model 2 (AOR with95% CI)	Model 3 (AOR with 95% CI)	Model 4 (adjusted AOR with 95% CI)
Household-level characteristics
Age in year
< 44		Ref		Ref
≥ 44 years		1.5(1.09, 2.04)**		1.6 (1.13, 2.22)***
Sex
Male		Ref		Ref
Female		1.7 (1.18, 2.30)***		1.7 (1.17, 2.38)***
BMI category
Underweight		Ref		Ref
Normal		0.6(0.44, 0.91)**		0.6 (0.42, 0.91)**
Overweight		0.8 (0.45, 1.28)		0.7(0.41, 1.27)
Obese		2.2 (1.24, 4.00)***		2.6 (1.34, 4.98)***
B. Community-level characteristics
Altitude
Low (< 1500 m)			Ref	Ref
Moderate (1500–2500 m)			1.5 (0.89, 2.67)	1.6(0.92, 2.77)*
High (> 2500 m)			2.1 (0.93, 4.84)*	2.3 (1.02, 5.16)**
Random Effects (measure of variation)
Variance (SE)	3.8 (0.62)	3.3(0.57)	3.7(0.66)	3.3 (0.63)
P-value	< 0.001	< 0.001	< 0.001	< 0.001
ICC (%)	53.3	50.2	52.9	50.2
PCV (%)	Ref	11.6	1.7	11.7
MOR	6.4	5.6	6.3	5.7
Model fit statistics
AIC	2230.8	2089.1	1952.3	1838
BIC	2245.1	2167.1	1987.3	1935.5

SE, standard error; ICC, intraclass correlation coefficient; PCV, Proportional change in variance; MOR, Median odds ratio; AIC, Akaike’s information criterion; BIC and Bayesian information criteria. Model 1 (Empty model) was fitted without determinant variables; Model 2 is adjusted for household-level variables; Model 3 is adjusted for community-level variables; Model 4 is the final model adjusted for household- and community-level variables. *** p<0.01, ** p<0.05, * p<0.1

Hypertension

This section of multilevel analysis demonstrates that both individual-and community-level factors play an important role in shaping the distribution of hypertension. The consistent improvement in model fit statistics (AIC and BIC) from the null to the full models indicates that the included predictors meaningfully explain variation in these outcomes.

Random-effects analysis

Model comparison indicated that Model 4 provided the best fit to the data, as evidenced by the lowest AIC (6856) and BIC (6974.5) values. Therefore, Model 4 was selected for interpretation. The random-effects results demonstrated significant clustering of hypertension across communities. The variance remained relatively stable across models (ranging from 0.80 to 0.91), indicating persistent between-cluster variability. The random effects were statistically significant in all models (p < 0.001), confirming that the observed clustering was not due to chance. The intraclass correlation coefficient (ICC) in the null model was 21.7%, suggesting that approximately one-fifth of the total variation in hypertension was attributable to differences between clusters. This value slightly decreased in Model 4 (20.9%), indicating that community-level variables explained part of the variability. The proportional change in variance (PCV) showed that Model 3 explained the largest proportion of variability (12.4%), while Model 4 explained 4.3% compared to the null model. This suggests that the included predictors explain little of the between-cluster variation. The median odds ratio (MOR) in Model 4 (2.44) further indicated substantial heterogeneity between clusters, implying that individuals moving between clusters with different risk levels could experience notable differences in hypertension risk.

Fixed-effects analysis

In the final model (Model 4), both individual and community-level variables were significantly associated with hypertension. Age showed a strong and graded association with hypertension. Compared to individuals aged <30 years, the odds of having hypertension was significantly higher among those aged 30–44 years (AOR = 1.9; 95% CI 1.57–2.31), 45–59 years (AOR = 3.9; 95% CI: 3.09–4.84), and ≥60 years (AOR = 7.2; 95% CI 5.54–9.41). Divorced or widowed individuals had higher odds of hypertension (AOR = 1.6; 95% CI 1.21–2.20) compared to single individuals. Regarding occupation, individuals engaged in informal employment had lower odds of hypertension (AOR = 0.8; 95% CI 0.62–0.96) compared to those in formal employment, while unemployment was not significantly associated. Body mass index (BMI) demonstrated a strong positive association with hypertension. Compared to underweight individuals, those with having normal BMI (AOR = 1.7; 95% CI 1.37–2.07), overweight (AOR = 3.1; 95% CI 2.39–4.04), and obesity (AOR = 5.2; 95% CI 3.61–7.39) had progressively higher odds of hypertension. At the community level, urban residence was associated with increased odds of hypertension (AOR = 1.5; 95% CI 1.17–1.82) than those who live in rural. Similarly, individuals living at moderate (AOR = 1.4; 95% CI 1.08–1.70) and high altitudes (AOR = 1.4; 95% CI: 1.01–2.01) had higher odds of hypertension compared to those residing at low altitude (Table 8).

Table 8.

Factors associated with hypertension identified through multilevel logistic regression models

Characteristics	Model 1 (AOR with 95% CI)	Model 2 (AOR with 95% CI)	Model 3 (AOR with 95% CI)	Model 4 (adjusted AOR with 95% CI)
Individual-level characteristics
Age
< 30		Ref		Ref
30 – 44		2 (1.68, 2.43)**		1.9 (1.57, 2.31)**
45 – 59		4.5 (3.67, 5.58)**		3.9 (3.09, 4.84)**
≥60		7.7 (5.98, 9.82)**		7.2 (5.54, 9.41)**
Marital status
Single		Ref		Ref
Married		1.1 (0.87, 1.37)		1.2(0.89, 1.46)
Divorced or widowed		1.5(1.15, 1.98)**		1.6 (1.21, 2.20)**
Occupation
Formal employment		Ref		Ref
Informal employment		0.8 (0.63, 0.93)**		0.8 (0.62, 0.96)*
Unemployed/not in active		1.03 (0.85, 1.24)		0.98 (0.80, 1.22)
BMI category
Underweight		Ref		Ref
Normal		1.6 (1.35, 2.01)**		1.7 (1.37, 2.07)**
Overweight		3.2 (2.46, 4.03)**		3.1 (2.39, 4.04)**
Obese		3.8 (2.78, 5.18)**		5.2 (3.61, 7.39)**
B. Community-level characteristics
Residence
Rural			Ref	Ref
Urban			2(1.61, 2.37)**	1.5 (1.17, 1.82)**
Altitude
Low (< 1500 m)			Ref	Ref
Moderate (1500–2500 m)			1.4 (1.14, 1.72)**	1.4 (1.08, 1.70)**
High (> 2500 m)			1.5(1.08, 2.07)*	1.4 (1.01, 2.01)*
Random-effects metric/measure of variation
Variance (SE)	0.9(0.09)	0.9 (0.10)	0.8 (0.09)	0.9(0.10)
P-value	<0.0001	<0.0001	<0.0001	<0.0001
ICC (%)	21.7	21.7	19.5	20.9
PCV (%)	Reference	-0.29	12.4	4.3
MOR	2.48	2.49	2.34	2.44
Model fit statistics
AIC	8933.6	7855.4	7728.3	6856
BIC	8947.8	7954.6	7756.5	6974.5

SE, standard error; ICC, intraclass correlation coefficient; PCV, Proportional change in variance; MOR, Median odds ratio; AIC, Akaike’s information criterion; BIC and Bayesian information criteria. Model 1 (empty model) was fitted without determinant variables; Model 2 is adjusted for household-level variables; Model 3 is adjusted for community-level variables; Model 4 is the final model adjusted for household- and community-level variables. ** p<0.01, * p<0.05

Diabetes mellitus

This multilevel logistic regression analysis identified both individual- and community-level factors associated with diabetes. In the final model (Model 4), which adjusted for both household- and community-level variables, several predictors remained statistically significant.

Random-effects analysis

Model comparison showed that Model 4 had the best fit, with the lowest AIC (2541.1) and BIC (2665.2) values. Therefore, it was selected as the final model for interpretation and conclusion of the data. The random-effects analysis revealed substantial clustering of diabetes mellitus across communities. The variance decreased markedly from 2.6 in the null model to 1.3 in Model 4, indicating that included predictors explained a significant portion of the variability. All random effects were statistically significant (p < 0.001). The ICC in the null model was high (44.4%), suggesting that nearly half of the total variability in diabetes was attributable to between-cluster differences. This value declined to 27.7% in Model 4, indicating that both individual- and community-level variables substantially explained the observed variability. The PCV further supported this, with Model 4 explaining 52.1% of the variance compared to the null model. The MOR decreased from 4.7 in the null model to 2.9 in Model 4, indicating reduced but still meaningful between-cluster heterogeneity.

Fixed-effects analysis

In Model 4, both individual- and community-level variables were significantly associated with diabetes mellitus. Age was a strong predictor of diabetes. Compared to individuals aged <30 years, the odds of having diabetes was significantly higher among those aged 30–44 years (AOR = 2.1; 95% CI 1.40–3.11), 45–59 years (AOR = 3.6; 95% CI 2.35–5.63), and ≥60 years (AOR = 5.1; 95% CI 3.11–8.40). Body mass index (BMI) was significantly associated with diabetes status. Compared with underweight individuals, the odds of diabetes were higher among overweight participants (AOR = 1.7; 95% CI 1.08–2.60) and markedly higher among obese individuals (AOR = 3.1; 95% CI 1.78–5.45). In contrast, individuals with a normal BMI did not show a statistically significant difference in odds of diabetes relative to the underweight participants. At the community level, urban residence was associated with higher odds of diabetes (AOR = 1.5; 95% CI 1.02–2.10) than those who reside in rural areas. In contrast, residing at high altitude was associated significantly with lower odds of diabetes (AOR = 0.5; 95% CI 0.28–0.96) compared to those who live at low altitude (Table 9).

Table 9.

Factors associated with diabetes mellitus identified through multilevel logistic regression analysis

Characteristics	Model 1 (AOR with 95% CI)	Model 2 (AOR with95% CI)	Model 3 (AOR with 95% CI)	Model 4 (adjusted AOR with 95% CI)
Individual-level variables
Age
< 30 years		Ref		Ref
30 – 44 years		1.8(1.26, 2.54)*		2.1(1.40, 3.11)*
45 – 59 years		3.3(2.24, 4.87)*		3.6 (2.35, 5.63)*
$\ge$ 60 years		4.8 (3.03, 7.46)*		5.1 (3.11, 8.40)*
BMI category
Underweight		Ref		Ref
Normal weight		0.9(0.64, 1.26)		1 (0.67, 1.37)
Overweight		1.8(1.19, 2.74)*		1.7 (1.08, 2.60)*
Obese		4.7 (2.92, 7.60)*		3.1 (1.78, 5.45)*
Community-level variables
Residence
Rural			Ref	Ref
Urban			1.8(1.28, 2.40)*	1.5(1.02, 2.10)*
Altitude
Low (< 1500 m)			Ref	Ref
Moderate (1500–2500 m)			0.9 (0.68, 1.33)	0.9 (0.64, 1.29)
High (> 2500 m)			0.5 (0.29, 0.96)*	0.5 (0.28, 0.96)*
Random-effects metrics (measure of variance)
Variance (SE)	2.6(0.36)	2.01(0.31)	1.3 (0.23)	1.3 (0.24)
P-value	<0.001	<0.001	<0.001	<0.001
ICC (%)	44.4	37.9	27.9	27.7
PCV (%)	Reference	23.5	51.5	52.1
MOR	4.7	3.9	2.9	2.9
Model fit statistics
AIC	3347.2	3039.3	2742	2541.1
BIC	3361.3	3144.4	2776.6	2665.2

SE, standard error; ICC, intraclass correlation coefficient; PCV, Proportional change in variance; MOR, Median odds ratio; AIC, Akaike’s information criterion; BIC and Bayesian information criteria. Model 1 (empty model) was fitted without determinant variables; Model 2 is adjusted for household-level variables; Model 3 is adjusted for community-level variables; Model 4 is the final model adjusted for household- and community-level variables. * p<0.05

Discussion

This study utilized nationally representative data from the 2024 Ethiopian WHO STEPwise (STEPS) Survey to assess the prevalence and multilevel determinants of three major noncommunicable diseases namely self-reported history of cardiovascular disease, hypertension, and diabetes mellitus among Ethiopian adults. It further examined associations between these outcomes and both individual- and community-level factors. Multilevel logistic regression modeling was applied to account for clustering of individuals within communities, enabling simultaneous evaluation of determinants at multiple levels. The findings revealed a substantial burden of noncommunicable diseases in Ethiopia, with hypertension being the most prevalent condition (17.3%), followed by self-reported history of cardiovascular disease (5.4%) and diabetes mellitus (4.2%). These results highlight the continuing epidemiological transition in Ethiopia, where noncommunicable diseases are becoming increasingly important public health challenges. The pattern aligns with the epidemiological understanding that blood pressure related conditions are often widespread in adult populations, while diabetes prevalence tends to be lower but still clinically important due to its strong contribution to cardiovascular morbidity and mortality [39–41]. Furthermore, the prevalence of hypertension observed in this study is consistent with global and regional evidence indicating that raised blood pressure is the most common noncommunicable disease risk condition in adult populations. It is widely recognized as a “silent epidemic,” often remaining undiagnosed until complications occur, thereby contributing significantly to cardiovascular morbidity and mortality [42]. The observed prevalence of hypertension was also comparable with previous Ethiopian national estimates [43], suggesting a relatively stable burden over time. Similarly, the prevalence of diabetes mellitus (4.2%) observed in this study aligns closely with previous national-level evidence [43]. Although lower than hypertension, diabetes remains a critical public health concern due to its strong association with cardiovascular complications and premature mortality [44]. The relatively lower prevalence compared to hypertension may be explained by differences in disease progression, diagnostic coverage, and awareness levels. In many low- and middle-income settings, including Ethiopia, diabetes is often underdiagnosed, suggesting that the true burden may be higher than reported [45–47]. The prevalence of self-reported history of cardiovascular disease (5.4%) in this study is consistent with findings from a systematic review and meta-analysis conducted in Ethiopia, which reported a pooled prevalence of 5% [48]. This consistency suggests a stable burden of self-reported cardiovascular disease across populations and time periods. However, the reliance on self-reported history may underestimate the true burden, as undiagnosed cases of stroke, angina, or myocardial infarction may not be captured, particularly in rural and underserved populations with limited access to healthcare services.

The multilevel analysis revealed substantial clustering of noncommunicable disease outcomes across communities, as evidenced by significant intraclass correlation coefficients (ICC) in all models. This finding indicates that individuals living within the same geographic or contextual environment share similar risks for noncommunicable diseases beyond individual characteristics alone. Such clustering underscores the importance of contextual and environmental determinants in shaping NCD risk in Ethiopia. Even after adjusting for individual-level factors, a considerable proportion of variance remained attributable to community-level differences, particularly for diabetes and self-reported history of cardiovascular disease, suggesting persistent contextual inequality in health outcomes. At the individual level, age emerged as a strong and consistent predictor of all three NCD outcomes. Older adults had significantly higher odds of hypertension, diabetes, and self-reported history of cardiovascular disease compared to younger individuals. This finding is consistent with the biological processes of aging, including vascular stiffening, metabolic changes, and cumulative exposure to risk factors over time. The observed age gradient reinforces the importance of targeted prevention strategies for older populations. Sex differences were also observed, with females showing higher odds of self-reported history of cardiovascular disease in the final model. This may be partly explained by differences in health-seeking behavior, hormonal factors, and differential exposure to risk factors [49]. Body mass index (BMI) demonstrated a strong and graded association with all noncommunicable disease outcomes, particularly hypertension and diabetes. Obesity significantly increased the odds of both conditions, consistent with established evidence linking excess adiposity to metabolic dysfunction, insulin resistance, and elevated blood pressure [50–52]. The strong association between BMI and noncommunicable diseases highlights the growing impact of nutritional transition in Ethiopia, characterized by increased consumption of energy-dense foods and reduced physical activity. At the community level, place of residence and altitude were important determinants of noncommunicable disease outcomes. Urban residence was associated with higher odds of hypertension and diabetes, reflecting the influence of urban lifestyles, including sedentary behavior, dietary changes, and increased stress levels [21]. Conversely, self-reported history of cardiovascular disease showed a relatively higher prevalence in rural areas, which may reflect delayed diagnosis and limited access to preventive and curative services. Altitude was also significantly associated with NCD outcomes, although the direction of association varied by disease. Individuals residing at higher altitudes had increased odds of self-reported history of cardiovascular disease and hypertension but lower odds of diabetes in some models. This may be explained by physiological adaptations to hypoxia, differences in lifestyle patterns, dietary practices, and physical activity levels across ecological zones. However, further research is needed to clarify the mechanisms underlying these associations. The findings from the multilevel model highlight the importance of individual- and community-level interventions in addressing noncommunicable diseases in Ethiopia. The reduction in intraclass correlation coefficients across models suggests that incorporating contextual variables improves model fit and explains part of the observed variability. However, persistent clustering even in the final model indicates that unmeasured community-level factors, such as healthcare access, environmental exposures, and socioeconomic inequality, may still play a significant role. Despite the relatively stable prevalence of noncommunicable diseases compared with previous studies, the findings emphasize the need for strengthened prevention strategies. The observed consistency may reflect the impact of national policies targeting noncommunicable diseases prevention, increased public awareness, and improved healthcare access [53]. However, ongoing exposure to behavioral risk factors and rapid urbanization may counteract these gains, potentially leading to increased noncommunicable diseases burden in the future. Additionally, recent policy measures targeting restriction on promotion of tobacco and alcohol taxes may have contributed to stabilizing noncommunicable disease trends. Furthermore, the integration of noncommunicable diseases into national health policies and strategies may have contributed to more stable detection and reporting.

Limitations of the study

Although this study has several strengths, it is also subject to certain limitations. First, the cross-sectional design of the 2024 WHO STEPS Survey limits the ability to establish causal relationships between noncommunicable disease outcomes and their associated factors. Second, the study relied on self-reported data for cardiovascular disease, which may be subject to recall bias and underreporting, particularly among individuals with undiagnosed conditions. This may lead to underestimation of the true burden of cardiovascular disease. Third, although key behavioral variables such as tobacco use, alcohol consumption, and physical activity were considered during the initial analysis, they were not retained in the final multivariable model due to lack of statistical significance in the bivariable analysis. The exclusion of these variables may result in residual confounding, as they are well-established risk factors for noncommunicable diseases and could still influence the observed associations.

Conclusion

This study provides nationally representative evidence on the prevalence and multilevel determinants of major noncommunicable diseases in Ethiopia. Hypertension was the most prevalent condition, followed by self-reported history of cardiovascular disease and diabetes mellitus. The study identified many core individual- and community-level factors significantly influence disease distribution. Overall, the study highlights the need for integrated prevention strategies that address both individual behavioral risks and broader environmental determinants. While the cross-sectional nature of the study limits causal inference, the results may help inform public health planning and prioritization. Targeted interventions focusing on factors such as age, sex, obesity, and place of residence may contribute to reducing the burden of noncommunicable diseases. Strengthening national NCD prevention and control programs, including routine screening for cardiovascular disease, hypertension, and diabetes across all levels of the health system, may improve early detection and management. Regional and local health authorities may also consider implementing community-based screening and early detection programs, particularly for high-risk groups such as adults aged 44 years and above. In addition, improving the availability of essential diagnostic tools, including blood pressure apparatus, glucometers, and anthropometric measurement devices at primary healthcare facilities may support timely diagnosis and effective management. In addition, improving the availability of essential diagnostic tools, including blood pressure apparatus, glucometers, and anthropometric measurement devices at primary healthcare facilities may support timely diagnosis and effective management.

Abbreviations

AOR

Adjusted odds ratio

BMI

Body mass index

CI

Confidence interval

DBP

Diastolic blood pressure

FPG

Fasting plasma glucose

NCDs

Noncommunicable diseases

CVD

Cardiovascular disease

SBP

Systolic blood pressure

WHO

World Health Organization

Author contributions

AMG was responsible for conceptualization, methodology, and drafting the original manuscript. AMG, GTZ, STT, and AB were involved in the resources and investigation. AMG, GTZ, FC,YBT, AZ, DA, HA,TG, WT,TD, SAS, BTK, AAL, MSZ, AL, SAB, HT, DAC, DZ, AW, GG, KBA, WBG, YL, FG, GM, TT, and AT contributed to data collection and supervision. AMG, and ESG conducted the formal analysis and software. AMG, GTZ, MGN, AW, TM, ABW, SA, ASK, MH, HST, and GT contributed to investigation, resources, and visualization, validation. HST, GT, AMG, and GTZ contributed to project administration and funding acquisition. AMG, ASK, HTS, and GT participated in investigation, reviewing, editing, and approving the final manuscript.

Funding

This research was funded by the Ministry of Health of Ethiopia. However, the funding body had no role in the study design, data collection, analysis, interpretation of results, manuscript preparation, or the decision to publish the findings.

Data availability

The dataset used to conclude the study is available from the corresponding author and provided on a reasonable request.

Declarations

Ethics approval and consent to participate

Ethical approval for this study was obtained from the Ethiopian Public Health Institute Institutional Review Board with Ref. No. EPHI-IRB-534-2023. The study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments. Written informed consent was obtained from all participants prior to data collection. Participants were informed of the voluntary nature of their participation, the confidentiality of their responses, and their right to withdraw from the study at any time without consequence. All physical measurements and biochemical tests were conducted free of charge and without financial compensation. Study results were communicated to participants on-site to facilitate early identification of noncommunicable disease risk factors, and individuals with abnormal findings were referred to nearby health facilities for further evaluation and management.

Competing interests

The authors declare no competing interests.