Automated and code-free development of a risk calculator using ChatGPT-4 for predicting diabetic retinopathy and macular edema without retinal imaging

Table 5 Prediction performance of developed algorithms and diabetes indicators for diabetic macular edema

	ROC-AUC [95% CI]	Accuracy (%), [95% CI]	Sensitivity (%), [95% CI]	Specificity (%), [95% CI]	P-value*
Internal validation
LR (ChatGPT-4)	0.940 [0.914, 0.960]	85.6 [82.0, 88.7]	100.0 [59.0, 100.0]	85.4 [81.7, 88.5]	Reference
RF (R)	0.918 [0.888, 0.942]	84.0 [80.3, 87.3]	100.0 [59.0, 100.0]	83.8 [79.9, 87.1]	0.261
GBM (Orange)	0.955 [0.931, 0.972]	86.5 [82.9, 89.5]	100.0 [59.0, 100.0]	86.3 [82.7, 89.3]	0.463
SVM (MATLAB)	0.892 [0.859, 0.919]	78.2 [74.1, 81.9]	85.7 [42.1, 99.6]	78.1 [73.9, 81.8]	0.368
HbA1c	0.896 [0.865, 0.924]	68.5 [64.0, 72.8]	100.0 [59.0, 100.0]	68.0 [63.4, 72.4]	0.328
DM duration	0.765 [0.724, 0.804]	49.9 [45.1, 54.6]	100.0 [59.0, 100.0]	49.1 [44.3, 53.8]	0.007
External validation
LR (ChatGPT-4)	0.835 [0.780, 0.881]	66.3 [59.8, 72.4]	100.0 [29.2, 100.0]	65.9 [59.3, 72.1]	Reference
RF (R)	0.851 [0.798, 0.895]	69.4 [63.0, 75.3]	100.0 [29.2, 100.0]	69.0 [62.5, 74.9]	0.414
GBM (Orange)	0.851 [0.798, 0.895]	71.6 [65.3, 77.3]	100.0 [29.2, 100.0]	71.2 [64.8, 77.0]	0.805
SVM (MATLAB)	0.757 [0.695, 0.811]	82.5 [76.9, 87.2]	66.7 [9.4, 99.1]	82.7 [77.2, 87.4]	0.308
HbA1c	0.821 [0.765, 0.869]	72.5 [66.2, 78.2]	100.0 [29.2, 100.0]	72.1 [65.8, 77.8]	0.819
DM duration	0.770 [0.710, 0.824]	60.3 [53.6, 66.6]	100.0 [29.2, 100.0]	59.7 [53.0, 66.2]	0.609

CI confidence interval, DM diabetes mellitus, GBM gradient boosting machine, LR logistic regression, ROC-AUC area under the receiver operating characteristic curve, SVM support vector machine
^*Differences in ROC-AUC values compared to the logistic regression performed using ChatGPT-4

ISSN: 2056-9920