Using Artificial Intelligence to Develop a Multivariate Model with a Machine Learning Model to Predict Complications in Mexican Diabetic Patients without Arterial Hypertension (National Nested Case-Control Study): Metformin and Elevated Normal Blood Pressure Are Risk Factors, and Obesity Is Protective

Table 2

Summary of the ten most important predictors for the model of chronic complications in patients with diabetes.

Predictor	Value	SHAP	OR	CI
Years smoking	(30.0, 38.0]	0.269083	0.388	[0.38, 0.396]	<0.001
Continues in the MIDE program	Yes	0.202862	1.756	[1.725, 1.787]	<0.001
Metformin	Yes	0.184438	3.908	[3.792, 4.029]	<0.001
Age	(31.999, 50.0]	0.134289	0.641	[0.629, 0.654]	<0.001
Maximum weight	(97.0, 116.5]	0.124975	0.560	[0.55, 0.57]	<0.001
Age	(67.0, 104.0]	0.085687	1.447	[1.417, 1.477]	<0.001
Nutrition consultation	Yes	0.079658	1.662	[1.626, 1.699]	<0.001
HbA1c	(2.999, 6.0]	0.075796	0.719	[0.705, 0.734]	<0.001
Deliveries	(3.0, 4.0]	0.069664	0.729	[0.717, 0.741]	<0.001
Attached to treatment	Yes	0.062306	1.351	[1.323, 1.379]	<0.001

OR: odds ratio; CI: 95% confidence interval; : the value of statistical significance. This table presents the results obtained by the XGBoost algorithm when trained to predict the dependent variable of chronic complication. The order of the variables is descending according to their importance value (SHAP) to predict said variable. The value column presents the value used for each intervening variable during the training. In addition, the odds ratio (OR) values, confidence interval (CI), and statistical significance value () are presented. The notation (#, #] represents a range that does not include the first value and includes the second value. For calculating odds ratios, values outside this range are used as the reference value, such as lower or higher values in the database.