LAM, LO KUEN CINDY (Principal Investigator), Professor and Head, Department of Family Medicine and Primary Care, The University of Hong Kong FONG, YEE TAK DANIEL, Associate Professor, School of Nursing, The University of Hong Kong KWOK, LAI PING RUBY, Senior Manager, Department of Primary & Community Services, Hospital Authority CHAO, VAI KIONG DAVID, Chief of Service and Consultant, Department of Family Medicine and Primary Health Care, Kowloon East Cluster, Hospital Authority TAN, CHOON BENG KATHRYN, Clinical Professor, Department of Medicine, The University of Hong Kong HUI, MING TUNG ERIC, Chief of Service and Consultant, Department of Family Medicine, New Territories East Cluster, Hospital Authority TSUI, WING SZE WENDY, Consultant, Family Medicine & Primary Healthcare, QMH, Hong Kong West Cluster, Hospital Authority CHAN, KING HONG, Consultant (Family Medicine and GOPC), Kowloon Central Cluster, Hospital Authority FUNG, SIU CHEUNG COLMAN, Honorary Clinical Assistant Professor, Department of Family Medicine and Primary Care, The University of Hong Kong WAN, YUK FAI ERIC, Assistant Professor, Department of Family Medicine and Primary Care & Department of Pharmacology and Pharmacy, The University of Hong Kong DONG, WEINAN DOVEY, PhD Candidate, Department of Family Medicine and Primary Care, The University of Hong Kong

Background Diabetes Mellitus (DM) is a leading global disease burden with rising prevalence in China. We need accurate models to predict the risk of long-term complications or mortality to facilitate cost-effective individualized interventions for Chinese DM patients. Aims and Objectives: To develop and validate 10-year risk prediction models for total cardiovascular diseases (CVD), CHD, heart failure, stroke, ESRD and mortality in primary care Chinese DM patients. To develop simplified nomograms and charts for the prediction of 10-year risk of CVD and mortality. Study Design and Subjects 10-year retrospective cohort study. 141,516 patients who had a clinical diagnosis of T2DM without complication managed in public (HA) primary care clinics between January and December 2008 were included and followed up until December 2017. Methods: 2/3 subjects were randomly selected for development of sex-specific 10-year risk prediction models for each outcome by Cox regressions. The models were validated on the remaining 1/3 subjects by Harrell’s C statistic and ROC. Up to seven most important predictors were used to construct the nomograms and charts. Results 10-year cumulative incidence of CVD, ESRD, and mortality was 22.9%, 6.0% and 19.8%, respectively. In addition to traditional risk factors, variabilities of SBP and HbA1c were significant predictors of CVD, ESRD and mortality. The use of transformation terms (e.g. SBP2) and interaction terms (e.g. age*WHR) significantly improved predictive power. The models performed well in the validation sample (Harrell’s C for female/male CVD, ESRD and mortality was 0.748/0.709, 0.889/0.889, and 0.857/0.841, respectively). The CVD and mortality nomograms and charts differentiated different risk groups effectively. Conclusions 10-year risk of CVD, ESRD and mortality of primary care Chinese T2DM patients can be accurately predicted by routinely available parameters. Implications These 10-year risk prediction models will enable accurate risk stratification of Chinese T2DM patients to guide clinical decision and patient activation.

Health and Medical Research Fund, Hong Kong Ref. No. 14151181.

The ethics of this study was approved by the IRB of the University of Hong Hong/HA HKWC IRB, Reference number UW 15-258.

US ClinicalTrials.gov: Identifier NCT03299010, Protocol Record HKUCTR-2232. HKU Clinical Trial Registry: Identifier HKUCTR-2232..

We would like to thank the Central Panel on Administrative Assessment of External Data Requests of the Hospital Authority for approval of the data extraction for our study. Special thanks to Mr. Peggo Lam and staff of the Statistics and Workforce Planning Department of the Strategy and Planning Division in Hospital Authority Head Office for their help in data extraction.