Extended Data Fig. 9: Pattern of distribution of event probabilities derived from competing risk models constructed with continuous variables as covariates.

a. Predicted probability of insulin initiation (use of insulin for more than 6 months or a clinical requirement for insulin, indicated as two or more HbA1c reading > =8.5% more than three months apart while taking two or more oral antidiabetic agents) at 5- year period from the diagnosis of T2D (n = 22595). b. Probability of incident diabetic retinopathy (R3/R4) at 5-year period (n = 22759). c. Probability of incident major adverse cardiac events (identified from SMR and GRO based on ICD 9 and ICD 10 codes) at 5-year period (n = 18239). d. Probability of incident chronic kidney disease (eGFR < = 60 ml/min/1.73m2 on at least 2 readings which were 90 days apart) at 5-year period (n = 19956). For all outcomes (A-D) probabilities were generated from a competing risk model constructed with continuous variables (age of diagnosis, sex, HbA1c, BMI, HDL-C, TG, TC, ALT, BP, and Creatinine) and competing risk of death. e. Linear regression estimates (with 95% CI) between the DDRTree dimensions and the four diabetes outcomes probability f. Spatial autocorrelation diabetes outcome probability; The Moran’s I statistic is shown on the X-axis, with higher values representing phenotypes that are more strongly autocorrelated; all values are p < 0.001.