Table 2. Performance of logistic regression models for prediction of prevalent diabetes with and without the DNN score in the Test Dataset.
All models are shown without (“a”) and with (“b”) inclusion of the DNN score as a predictor. Models 1-3 are nested models, containing incrementally more demographic predictors and BMI. Model 4 adjusts for heart rate and heart rate variability. Model 5 adjusts for common cardiovascular comorbidities. Sample sizes reported indicate individual people.
| Predictor | Multivariable-adjusted OR without DNN Score (95% CI) |
p- value† |
Multivariable- adjusted OR with DNN Score (95% CI) |
p-value† |
|---|---|---|---|---|
| LogReg Model 1a: Age: AUC=0.691 (95% CI: 0.672–0.710) (n=11,313) | LogReg Model 1b: 1a + DNN score: AUC=0.770 (95% CI: 0.754-0.786) (n=11,313) | |||
| Age, years | 1.04 (1.04–1.05) | <0.001 | 1.01 (1.01–1.02) | <0.001 |
| DNN Score, per SD | - | - | 2.69 (2.41-2.99) | <0.001 |
| LogReg Model 2a: Age, gender and race: AUC=0.698 (95% CI: 0.674 – 0.722) (n=7,851) | LogReg Model 2b: 2a + DNN score: AUC=0.777 (0.757 – 0.798) (n=7,851) | |||
| Age, years | 1.04 (1.04–1.05) | <0.001 | 1.01 (1.00–1.02) | 0.013 |
| Gender | ||||
| Males | Ref. | - | Ref. | - |
| Females | 0.99 (0.82-1.21) | 0.996 | 0.65 (0.53–0.79) | <0.001 |
| Race/Ethnicity | 0.003 | 0.17 | ||
| Non-Hispanic White, n (%) | Ref. | - | Ref. | - |
| Black or African American, n (%) | 1.87 (1.11-3.15) | 0.001 | 1.40 (0.82 - 2.38) | 0.213 |
| Hispanic, Latino or Spanish origin/ancestry, n (%) | 0.73 (0.49-1.07) | 0.106 | 0.69 (0.46 – 1.01) | 0.058 |
| Asian, n (%) | 1.86 (1.30-2.67) | 0.001 | 1.46 (1.01 – 2.12) | 0.047 |
| Multi-ethnic, n (%) | 1.27 (0.78-2.07) | 0.344 | 1.29 (0.78 – 2.13) | 0.314 |
| Other, n (%) | 0.974 (0.49-1.95) | 0.941 | 0.86 (0.42 – 1.74) | 0.674 |
| DNN Score, per SD | - | - | 2.88 (2.51-3.31) | <0.001 |
| LogReg Model 3a: Age, gender, race and BMI: AUC 0.801 (95% CI: 0.752 – 0.850) (n=1,033) | LogReg Model 3b: 3a + DNN score: AUC: 0.830 (95% CI: 0.787 – 0.873) (n=1,033) | |||
| Age, years | 1.04 (1.02 - 1.06) | <0.001 | 1.01 (0.99 – 1.04) | 0.189 |
| Males | Ref. | - | Ref. | - |
| Females | 0.67 (0.39–1.13) | 0.130 | 0.51 (0.30 – 0.88) | 0.015 |
| Race | 0.232 | 0.415 | ||
| Non-Hispanic White | Ref. | - | Ref. | - |
| Black or African American | 0.33 (0.04 –2.63) | 0.294 | 0.32 (0.04 – 2.63) | 0.291 |
| Hispanic, Latino or Spanish origin or ancestry | 1.22 (0.45 –3.35) | 0.696 | 1.08 (0.38 – 3.05) | 0.884 |
| Asian or Pacific Islander | 2.82 (0.97-8.22) | 0.058 | 2.36 (0.77 – 7.24) | 0.135 |
| Multi-ethnic | 0.416(0.12-1.45) | 0.168 | 0.46 (0.13 – 1.59) | 0.218 |
| Other/prefer not to disclose | 0 (0) | 0.999 | 0 (0) | 0.999 |
| BMI | 1.15 (1.11–1.19) | <0.001 | 1.08 (1.04-1.12) | <0.001 |
| DNN Score, per SD | - | - | 2.12 (1.53 – 2.94) | <0.001 |
| LogReg Model 4a: Heart Rate and Heart Rate Variability: AUC 0.586 (95% CI: 0.565 – 0.606) (n=11,313) | LogReg Model 4b: 4a + DNN score: AUC: 0.765 (0.748 – 0.782) (n=11,313) | |||
| HR, bpm | 1.02 (1.01–1.02) | <0.001 | 1.01 (1.00–1.01) | 0.024 |
| HRV - RMSSD, per 10 ms | 0.97 (0.94-0.99) | 0.027 | 1.02 (1.00 – 1.05) | 0.068 |
| DNN Score, per SD | - | - | 2.92 (2.65 – 3.21) | <0.001 |
| LogReg Model 5a: Comorbidities: AUC 0.784 (0.766 – 0.802) (n=11,313) | LogReg Model 5b: 5a + DNN score: AUC 0.830 (0.815 – 0.844) (n=11,313) | |||
| Hypertension, n (%) | 3.49 (2.93-4.16) | <0.001 | 2.57 (2.15 – 3.07) | <0.001 |
| Hypercholesterolemia, n (%) | 2.44 (2.05-2.89) | <0.001 | 1.97 (1.66 – 2.34) | <0.001 |
| Coronary artery disease, n (%) | 1.35 (1.04-1.76) | 0.024 | 1.22 (0.94 – 1.59) | 0.144 |
| Prior MI, n (%) | 1.04 (0.74-1.48) | 0.815 | 1.06 (0.74 – 1.50) | 0.765 |
| CHF, n (%) | 2.39 (1.67-3.42) | <0.001 | 2.09 (1.46 – 2.98) | <0.001 |
| PVD, n (%) | 1.49 (1.00 -2.21) | 0.051 | 1.43 (0.97 – 2.11) | 0.075 |
| Prior Stroke, n (%) | 1.91 (1.39-2.61) | <0.001 | 1.74 (1.27 – 2.38) | 0.001 |
| Sleep apnea n (%) | 2.06 (1.72-2.46) | <0.001 | 1.85 (1.54 – 2.22) | <0.001 |
| DNN Score, per SD | - | - | 2.22 (2.00 – 2.46) | <0.001 |
Abbreviations: LogReg: logistic regression; OR: diagnostic odds-ratio; BMI: body mass index; CI: Confidence Interval; DNN: Deep neural network; SD: standard deviation; AUC: area under the receiver operating characteristic curve; HR: Heart Rate; HRV: Heart Rate Variability; RMSSD: Root Mean Square of Successive RR interval differences; y: years. Independent variables were standardized using the Z-score.
p-value is calculated using the Wald test for the multivariable adjusted odds ratio (two-sided).