Research Article

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 4

Chi-squared test for binary variables between complicated patients versus uncomplicated patients.

PredictorPatients test
ComplicatedUncomplicated

Smoking4956 (15.04%)2823 (9.16%)<0.001
Antidiabetic Tx: acarbose1230 (3.73%)439 (1.42%)<0.001
Antidiabetic Tx: phenformin9 (0.03%)7 (0.02%)0.907
Antidiabetic Tx: glibenclamide12816 (38.90%)5007 (16.24%)<0.001
Antidiabetic Tx: glimepiride220 (0.67%)48 (0.16%)<0.001
Antidiabetic Tx: glipizide21 (0.06%)8 (0.03%)0.040
Antidiabetic Tx: metformin18271 (55.45%)6072 (19.70%)<0.001
Antidiabetic Tx: pioglitazone992 (3.01%)203 (0.66%)<0.001
Antidiabetic Tx: rosiglitazone391 (1.19%)145 (0.47%)<0.001
Antidiabetic Tx: sitagliptin1506 (4.57%)157 (0.51%)<0.001
Antidiabetic Tx: tolbutamide36 (0.11%)13 (0.04%)0.004
Self-care: attached to the treatment21706 (65.88%)10808 (35.06%)<0.001
Self-care: self-medication6142 (18.64%)2805 (9.10%)<0.001
Self-care: diabetes education23160 (70.29%)20710 (67.18%)<0.001
Self-care: physical activity21791 (66.14%)19597 (63.57%)<0.001
Ketoacidosis302 (0.92%)611 (1.98%)<0.001
Hyperglycemia6200 (18.82%)1640 (5.32%)<0.001
Hyperosmolar225 (0.68%)71 (0.23%)<0.001
Hypoglycemia635 (1.93%)283 (0.92%)<0.001
MIDE follow-up: discharge2774 (8.42%)992 (3.22%)<0.001
MIDE follow-up: continue20625 (62.60%)8472 (27.48%)<0.001
MIDE follow-up: death39 (0.12%)9 (0.03%)<0.001
MIDE follow-up: abandonment257 (0.78%)133 (0.43%)<0.001
MIDE follow-up: secondary care968 (2.94%)178 (0.58%)<0.001
Nutrition consultation9565 (29.03%)3082 (10.00%)<0.001

Complicated: total number of patients and percentage in whom physicians registered one or more of the five events of interest in their medical record. Uncomplicated: total number of patients and percentage in whom physicians registered none of the five events of interest in their medical record.