Extensive phenotype data and machine learning in prediction of mortality in acute coronary syndrome – the MADDEC study

Figures & data

Table 1. Characteristics of patients undergoing coronary angiography for acute coronary syndrome between January 2007 and October 2017 in Tays Heart Hospital.

	All (n = 9.066)
Age, mean (SD)	68.1 (11.9)
Gender (female)	32.7% (2965)
Hypertensive	58.7% (5322)
Diabetic (any)	22.6% (2049)
Valvular heart disease (any)	6.4% (580)
Peripheral artery disease	5.7% (517)
Cancer	4.4% (403)
Clinical diagnosis of Heart failure	19.4% (1759)
Chronic obstructive pulmonary disease	2.1% (189)
Atrial fibrillation (any type)	9.0% (819)
Previous stroke	6.5% (589)
Previous myocardial infarction	14.1% (1278)
Previous percutaneous coronary intervention	10.7% (970)
Previous coronary artery bypass grafting	7.8% (707)
Creatinine, μmol/L	87.3 (56.1)
Haemoglobin, g/L	129.8 (16.0)
Mean corpuscular volume, fL	89.9 (4.8)
Thrombocytes, 10^9/L	224.3 (69.3)
Leukocytes, 10^9/L	8.6 (3.8)
Troponin T, ng/L (maximum recorded value)	590 (118−2220)
C-reactive protein mg/L	8.1 (2.4−8.1)
Left ventricular ejection fraction, percentage^a	51.7 (11.9)
Active smoker^a	25.3% (1235)

Continuous variables are presented as mean values with corresponding standard deviation except for Troponin T and C-reactive protein, which are presented by population median value with corresponding interquartile range.

^a Data available < 50% of the study population.

Table 2. Significant factors associating with six-month mortality in multivariable analysis by logistic regression (p < .05) in the training set of the MADDEC study population.

	OR	95%CI		p-value
Age	1.79	1.59	2.01	7.6e^−22
Mean Leukocyte value	1.58	1.41	1.78	8.2e^−15
Maximal measured creatinine value	1.43	1.29	1.59	4.7e^−11
Cardiac arrest (Resuscitation)	3.10	2.17	4.42	4.2e^−10
Hemodynamic instability at angiography	2.58	1.88	3.54	4.9e^−9
Complications in PCI	3.43	2.27	5.19	5.6e^−9
Minimum measured INR value	1.45	1.28	1.65	7.2e^−9
Number or previous angiographies	0.54	0.44	0.67	1.4e^−8
Type 1 diabetes mellitus	3.75	2.14	6.55	3.6e^−6
Mean measured haemoglobin value	0.76	0.67	0.85	6.0e^−6
Prevalent Cancer	2.35	1.56	3.54	4.7e^−5
Mean value for CRP measurements	1.31	1.14	1.50	1.2e^−4
Maximal measured INR value	1.27	1.12	1.44	1.4e^−4
Maximal measured mean corpuscular volume	1.21	1.09	1.34	2.7e^−4
Heart Failure	1.53	1.21	1.93	3.8e^−4
Use of oral anticoagulant	0.58	0.42	0.79	6.3e^−4
TIMI flow at culprit artery at angiography	0.88	0.81	0.95	.002
Maximal measured blood glucose concentration	1.17	1.06	1.30	.003
TIMI flow at culprit artery after PCI	0.81	0.70	0.94	.004
Prevalent peripheral artery disease	1.64	1.16	2.32	.005
History of stroke	1.53	1.11	2.12	.094
History of ST-elevation myocardial infarction	2.81	1.19	6.67	.019

OR: Odds Ratio; CI: Confidence Interval; PCI: Percutaneous Coronary Intervention; INR: International Normalized Ratio; CRP: C-reactive protein.

Figure 1. Comparison of model performance by receiving operating characteristic curves for different risk prediction models for six month mortality among patients undergoing coronary angiography in Tays Heart Hospital for acute coronary syndrome during years 2015 and 2016 (n = 1.722 with n = 122 fatalities during a six-month follow-up).