Table 3 Performance of the machine learning predictions when patients are targeted according to the calibration cutoff
From: Predicting control of cardiovascular disease risk factors in South Asia using machine learning
Outcome definition | Not achieving control | Not achieving improvements | ||||
|---|---|---|---|---|---|---|
CVD risk factor | HbA1c | SBP | LDL | HbA1c | SBP | LDL |
Chosen specification | Logistic (M) | Logistic (S) | Logistic (M) | Logistic (L) | Tree (S) | Logistic (M) |
Detection prevalence | 36% | 16% | 27% | 69% | 74% | 67% |
[95% CI] | [31%,41%] | [12%,20%] | [23%,32%] | [64%, 73%] | [70%, 78%] | [63%,72%] |
Precision | 75% | 30% | 23% | 71% | 71% | 74% |
[95% CI] | [67%,82%] | [19%,41%] | [15%,30%] | [66%, 77%] | [66%, 76%] | [68%,79%] |
Sensitivity | 53% | 26% | 42% | 82% | 88% | 86% |
[95% CI] | [46%,60%] | [16%,37%] | [29%,55%] | [76%, 86%] | [83%, 92%] | [82%,90%] |
