The utility of common predictive models for detecting undiagnosed diabetes and prediabetes defined by HbA1c in a Chinese population

Abstract

Aims: To evaluate the Hong Kong Primary Care Office (HKPCO) strategy against international guidelines from China, UK, US, Canada and Australia for identifying high-risk individuals requiring diagnostic testing for diabetes mellitus (DM) and pre-DM.

Methods: A cross-sectional study involved adults aged ≥25 without prior DM diagnoses. Participants underwent HbA1c point-of-care testing, with self-reported risk factors and clinical measurements documented. DM was defined as HbA1c ≥6.5 %, and pre-DM as 5.7-6.4 %. Sensitivity, specificity, number needed to test (NNT), and net benefit of different strategies were assessed.

Results: Amongst 1810 individuals (age: 59.3 ± 9.2 years, 39.6 % men), 4.8 % had DM and 53.6 % had pre-DM. The sensitivity, specificity and NNT for various detection strategies ranged from 19.5 % to 100 %, 0.58 % to 92.8 % and 8.3 to 20.7 respectively. Over 95 % fulfilled the HKPCO guideline (sensitivity = 100 %, specificity = 0.58 %, NNT = 20.7) without missing DM cases. NICE criteria (sensitivity = 96.6 %, specificity = 19.2 %, NNT = 17.6) correctly reclassified 18.6 % of high-risk participants, providing greater net benefits compared to HKPCO guideline.

Conclusions: The community-based program detected high proportion of Chinese adults at high-risk for undiagnosed DM and pre-DM. Utilizing strategies developed by NICE, or other international strategies, can enhance net benefits and potentially improve cost-effectiveness when compared to the current HKPCO guideline.

Keywords: Diabetes; HbA1c; Prediabetes; Prediction model; Screening.