Risk stratification of patients who present with chest pain and have normal troponins using a machine learning model

Table 1 Demographics

Characteristics	Abnormal cardiac stress test requiring cardiac catheterization and/or CABG		Degree of association (95%CI)	P value
	Yes, n = 245	No, n = 2083
Age, mean ± SD	63.02 ± 11.67	61.99 ± 12.46	Mean different: 1.03 (-0.53, 2.59)	0.200
Sex male	153 (62.4%)	965 (46.3%)	RR: 1.8 (1.41, 2.30)	< 0.001
Race
Caucasian	163 (66.5%)	1182 (56.7%)	RR: 1.8 (1.08, 3.00)	0.020 (combined)
African American	65 (26.5%)	650 (31.2%)	RR: 1.35 (0.79, 2.32)
Asian	2 (0.8%)	43 (2.1%)	RR: 0.66 (0.16, 2.79)
Other	15 (6.1%)	208 (10.0%)	Reference

CABG: Coronary artery bypass graft; CI: Confidence interval; RR: Risk ratio; SD: Standard deviation.

Table 2 Bivariate analysis of the risk factors

Risk factors	Abnormal cardiac stress test requiring cardiac catheterization and/or CABG		Risk ratio (95%CI)	P value
	Yes, n = 245	No, n = 2083
CAD history, yes	216 (88.2%)	1063 (51.0%)	6.11 (4.18, 8.92)	< 0.001
Obesity, yes	136 (55.5%)	1121 (53.8%)	1.06 (0.84, 1.35)	0.620
Diabetes mellitus, yes	112 (45.7%)	760 (36.5%)	1.41 (1.11, 1.78)	0.005
Hypertension, yes	232 (94.7%)	1783 (85.6%)	2.77 (1.61, 4.78)	< 0.001
Hyperlipidemia, yes	236 (96.3%)	1602 (76.9%)	6.99 (3.62, 13.50)	< 0.001
CKD history, yes	88 (35.9%)	540 (25.9%)	1.52 (1.19, 1.94)	< 0.001
Smoking history, yes	153 (62.4%)	1133 (54.4%)	1.35 (1.05, 1.72)	0.020

CABG: Coronary artery bypass graft; CAD: Coronary artery disease; CI: Confidence interval; CKD: Chronic kidney disease.

Table 3 Comparison between the high-risk group and the low-risk group

Risk category	Abnormal cardiac stress test requiring cardiac catheterization and/or CABG		Risk ratio (95%CI)	P value
	Yes, n = 245	No, n = 2083
High risk	236 (96.3%)	1700 (81.61%)	5.31 (2.75, 10.24)	< 0.001
Low risk	9 (3.7%)	383 (18.39%)

CABG: Coronary artery bypass graft; CI: Confidence interval.

Table 4 Adjustment of risk factors association with the abnormal cardiac stress test via binomial regression

Covariate	Risk ratio	95%CI	P value
Sex, male	1.52	1.2, 1.94	< 0.001
CAD history, yes	4.46	3.08, 6.72	< 0.001
Hypertension, yes	1.35	0.82, 2.44	0.280
Hyperlipidemia, yes	3.87	2.12, 8.12	< 0.001
Diabetes mellitus, yes	1.06	0.83, 1.34	0.650
CKD history, yes	1.02	0.80, 1.30	0.860
Smoking history, yes	1.06	0.84, 1.35	0.620

CAD: Coronary artery disease; CI: Confidence interval; CKD: Chronic kidney disease.

Table 5 Comparison of the models for the prediction of an abnormal cardiac stress test

Feature	BR	RF	XGBoost
Prediction cutoff value	0.20	0.18	0.27
Sensitivity (95%CI)	45.06 (44.23, 45.88)	13.92 (13.50, 14.33)	30.54 (29.30, 31.79)
Specificity (95%CI)	80.46 (80.14, 80.79)	93.66 (93.53, 93.80)	88.51 (88.15, 88.86)
PPV (95%CI)	21.34 (21.09, 21.60)	20.55 (20.05, 21.04)	24.33 (23.46, 25.20)
NPV (95%CI)	92.55 (92.42, 92.69)	90.24 (90.14, 90.35)	91.34 (91.12, 91.56)

BR: Binomial regression; CI: Confidence interval; NPV: Negative predictive value; PPV: Positive predictive value; RF: Random forest.