Diabetes mellitus among older people in rural Sidama, Ethiopia: a two-step community-based cross-sectional study

Objective

To determine the prevalence of undiagnosed diabetes and identify associated factors using a two-step diagnostic method combining fasting blood sugar with confirmatory glycated haemoglobin A1c (HbA1c) testing among older adults in rural Sidama, Ethiopia.

Design

A community-based cross-sectional design was conducted from April 1 to July 31, 2024. Data were collected through a census of adults aged ≥ 45 years using a pretested WHO-STEPwise questionnaire. Physical and biochemical tests were performed following standard protocols. Data were analysed using Stata version 17.

Setting

Selected rural kebeles of Shebedino district, Sidama, Ethiopia.

Participants

2,875 adults aged ≥ 45 years.

Primary outcome measures

Undiagnosed diabetes confirmed by haemoglobin A1c levels (≥48 mmol/mol).

Results

The prevalence of undiagnosed diabetes confirmed by haemoglobin A1c was 1.2% (35 of 2,871; 95% CI: 0.9 – 1.7%). Previously diagnosed diabetes was found in in 0.5% (14 of 2,875; 95% CI: 0.3% - 0.8%). The total diabetes prevalence, confirmed by haemoglobin A1c or prior diagnosis, was 1.7% (49 of 2871; 95% CI: 1.3% - 2.3%). Nearly half, 46.2% (1,327 of 2,875), were undernourished. Advanced age (β = 0.18; 95% CI: 0.06, 0.30, p = 0.004), estimated annual income (β = 0.14; 95% CI: 0.01, 0.27; p = 0.039), and waist-to-body mass index (β = 0.08; 95% CI: 0.01, 0.16; p = 0.032) were significantly associated with elevated fasting blood sugar levels.

Conclusion

The prevalence of undiagnosed diabetes in this rural setting was low. Increasing age, higher income, and waist-to-body mass index were associated with elevated fasting blood sugar. Routine community-based diabetes screening, health education, and nutrition-focused interventions are recommended to sustain the low burden and address undernutrition.

Strengths and limitations of this study

• A two-step diagnostic approach (fasting blood sugar and confirmatory HbA1c) enhanced the diagnostic accuracy compared with fasting blood sugar alone.

• Census-based sampling of a large rural population aged 45 years and above improved representativeness and reduced selection bias.

• Standardised tools (WHO-STEPwise protocol) and trained field teams were used to ensure data quality.

• Diabetes types were not differentiated with additional biomarkers (e.g., autoantibody or C-peptide testing), potentially causing misclassifications.

• Self-reported behaviours (smoking, khat chewing, and alcohol use) may be prone to recall or social desirability bias.