*3.2. Cross-Validation*

For the cross-validation analysis, 66% of the database was used to train the models, using a number of 10 folds as parameter. For this, the used classifiers were: tree classification, random forest, SVM, logistic regression and kNN. Considering the results of the FRCV decision tree, of the previous analysis, these secondary factors were used as targets, of which results are presented in Table 3.

#### **Table 3.** Metric's formulas.


In Table 4, the variables determined as secondary factors to consider when a patient arrives in the emergency room with chest pain are shown. The research suggests a close relation between these diseases and habits, since one can be caused by another. Among these variables, according to the results of the machine learning analysis, dyslipidemia may be considered as the main disease responsible for possible thoracic pain with cardiac origin, followed by hypertension, smoking habits, diabetes, chronic kidney disease, and PPT ranges. In the case of the variable dyslipidemia, the best obtained result was using logistic regression with an accuracy of 0.969, F1 of 0.938, precision of 0.937, and recall of 0.940. In hypertension, we found an accuracy of 0.994, F1 of 0.966, precision of 0.966, and recall of 0.966. For smoking, we found an accuracy of 0.918, F1 of 0.799, precision of 0.796, and recall of 0.803. Lastly, for diabetes, we found an accuracy of 0.986, F1 of 0.961, precision of 0.963, and recall of 0.961. For the variable of PPT ranges, the Random Forest model showed better results, with an accuracy of 0.977, F1 of 0.880, precision of 0.881, and recall of 0.891.


**Table 4.** Cross-Validation results using a target FRCV decision tree results.

#### **4. Discussion**

In emergency rooms, between 5% and 15% of the patients report thoracic pain, whereby 23.8% of patients with thoracic pain are related to cardiovascular pathologies [16]. Another case that was found to be alarming in the Hospital de la Línea de la Concepción in Cádiz is that 25% of the patients present an AMI (Acute Myocardial Infarction) after they left the emergency room due to a normal electrocardiogram [4], which can be construed as 1 out of 4 patients had a wrong diagnosis, which could lead to a sudden death. For this reason, it is important to find new methods to efficiently classify the origin of thoracic pain, since it can be related to cardiogenic factors either ischemic or not ischemic; and not cardiogenic factors being of gastrointestinal, pulmonar, neuromuscular, or psychological origin [17]. Due to the multiple risk factors for CVD, it is critical to find the nearest linked factor to a sudden death caused by a cardiomyopathy with thoracic pain as a symptom, considering that health conditions and lifestyle, including alimentation, have a considerable impact in CVD development. For this, machine learning techniques and tools are proposed to predict cardiopathies that could lead to sudden death [4,18]. Furthermore, studies found that when a person presents various risk factors, the probabilities to develop a CVD in a 10-year range increases significatively [19]. Hence, it is recognized as widely important

to identify the risk factors present when the patient arrives at the emergency room with thoracic pain. This study suggests the use of the risk factors obtained as results in the tree classification analysis and validated by cross-validation method, in the evaluation of the thoracic pain in order to classify it as cardiac or not cardiac, considering these as secondary factors alongside those currently used in the emergency rooms. However, these results must be interpreted with caution and a series of limitations must be taken into account since the study was carried out only with elderly patients and with less than 30 variables; to achieve more precise results in future studies, the use of a database with more variables to consider and a population with different age ranges is proposed, and with this, better training in machine learning models would be achieved, which would allow for finding greater differentiation between variables.

#### *4.1. Relationship of Secondary Factors Variables with IAM*

#### 4.1.1. Patients with Smoking Habits

Tobacco consumption increases the oxidative stress due to the free radicals generation for both passive and active consumers, and for this reason it is known as the main factor in the development of different diseases, including CVD. Among the adverse effects in health caused by tobacco consumption besides oxidative stress, studies found a relationship in the increase of the arterial pressure and cardiac frequency, increase in inflammation, developments of atherosclerosis, thrombosis, and damage in both arterial coronary systems [20,21].

Regarding chest pain, a study done with 70,208 participants, which mostly have smoking habits, discusses an experimentation using methods as pain tolerance testing and surveys, which concluded that people with smoking habits tend to have lower pain tolerance; this information is important to know regardless that the intensity has not relation with a cardiac origin pain. Moreover, it was found that the chest pain in smoking patients can be originated by, inter alia, the frequency in tobacco consumption, chronic cough and shortness of breath [22].

#### 4.1.2. Patients with Hypertension

Hypertension is one of the most important risk factors on CVD. Worldwide, hypertension is responsible for 54% of strokes, and 47% of ischemic cardiopathy [23]. It has also been observed that after a decade of presenting hypertension, the risk of contracting any CVD has increased from 15% to 30% [24]. On the other hand, evidence of a study made in 1997 in Chile found interesting results in records of people with obesity, which suggest that obesity increases blood pressure with 6.5 mmHg, plasma cholesterol with 12 mg/dL, and 2 mg/dL of blood glucose for each 10% of accession in the patient's weight [25].

A study described in the Cuban Magazine of Health compares their findings done between 2007 and 2011 with findings made in Spain on 2011, and both results agree with the fact that Hypertension is strongly related with sudden death by a cardiac event; this, due to the development of an adaptive process initiated by blood pressure causing hypertrophy as a result of left ventricular injure. It is also stated that a combination of hypertension with smoking habits or any other risk factor as diabetes, dyslipidemia, and obesity can lead to an increase in the left ventricular hypertrophy expanding the probabilities of suffering a cardiac event [26].

#### 4.1.3. Patients with Diabetes

Diabetes is a disease that is also tightly related with CVD and obesity, when there are no other risk factors involved it is called Diabetic Heart Disease (DHD). Amidst the possible factors of the relationships between these conditions, insulin resistance, hyperglycemia, and hyperinsulinemia were found to be responsible for the decrease in elasticity of the tissue generated by an impact in the production of collagen, which provokes myocardial damage leading to hypertrophy and fibrosis [27].

Despite the fact that in a study carried out in the Grama region, it was determined that patients with DM II who present other cardiovascular risk factors, compared to those without Diabetes, did not present chest pain as a symptom. However, the study suggests that those with DM II are exposed to cardiac failure by a factor of 2.8, since it has also been found that patients with this disease suffer from alterations in diastolic function without having any history of cardiovascular disease [28].

The chemical reactions generated by cardiac metabolism are oxidative in nature, so that as there is a lack of biological contribution to the region of cardiac tissue, ATP is stopped in cardiomyocytes, which in turn, causes a metabolic change due to the lack of oxygen and nutrients directly affecting cardiac functionality [29].

#### 4.1.4. Patients with Chronic Kidney Disease

Chronic Kidney Disease (CRD) is another risk factor linked to CVD. Findings from a study made with dialysis patients revealed that CVD patients start their development in precocious phases of the CRD, causing problems such as left ventricular hypertrophy, atherosclerosis, and vascular calcifications [30]; therefore, early detection and treatment of this disease can reduce the chances of death from CVD, as well as decrease kidney damage, since it was revealed in a study carried out using patients with advanced ECR with and without dialysis, which those with an AMI have a very low chance of survival [31,32]. On the other hand, CKD is found in some cases related to diabetes, which is called diabetic nephropathy, which develops hypertension and kidney damage [33].

#### 4.1.5. Patients with Dyslipidemia

Among the distributions related to patients with dyslipidemia and pain in the database from Medical Norte, 22.87% of the patients with dyslipidemia presented soft pain, while 26.74% presented moderate pain and 14.34% presented severe pain. Of the remaining individuals without dyslipidemia, only 5.81% presented severe pain. Despite these results, it is important to know, beyond pain, how dyslipidemia would affect the cardiovascular system.

Dyslipidemia is a disease where the regulation of lipids in blood is affected by the augmentation of cholesterol and triglycerides, which in turn produces the accumulation of lipids in the arterial walls causing ischemic heart disease, which can lead to death; the main reason of this disease is due to obesity, even though it can be also a genetic disease [34]. The most known disease in Mexico is obesity since, in 2012, 71.3% of the population was diagnosed with obesity, while in Baja California, 74.9% of the population presented obesity and overweight [35]. Obesity is one of the main factors for various diseases, including CVD. The relationship between dyslipidemia and obesity is very close due to the excess of fatty tissue, which produces an insulin resistance [36]; also, it is related to diseases such as Diabetes Mellitus II (DM II). According to the WHO in 2012, 44% of the people living in Baja California developed DM II due to obesity and overweight, pathology related with hypertension, dyslipidemia, CDV, osteoarthritis, and different types of cancer [37].

This documental research confirms the correlation between the proposed secondary risk factors related with possible thoracic pain with cardiac origin. In Figure 1, the diagram above shows graphically the relation between these variables, which was confirmed by both the assessment with machine learning and bibliography. The figure is divided into three main components, the blue navy hexagon in the center indicates the target, which is thoracic pain with cardiac origin, the second level with blue hexagons shows the six main conditions proposed as factors to consider in the determination of a cardiac event with thoracic pain as symptom, and the last level with light blue hexagons shows some effects that the main factors have in health. The orange lines used in Figure 2 express the relationship between the conditions.

**Figure 2.** Relationship between secondary risk factors variables.

#### **5. Conclusions**

Among the main health problems presented in the country are deaths from obesity problems and cardiovascular diseases, which in turn are related to each other, sharing other risk factors. When considering cardiovascular problems as diseases that can cause sudden events involving a person's life, it is important to learn to recognize the patterns that these cardiac events present and to take into account the factors that have the greatest impact on their development. It is known that in emergency rooms, there are a limited number of patients to attend, and since thoracic pain is a symptom of a future cardiac event, but also a symptom of different diseases, it is important to learn to recognize when thoracic pain is of cardiac origin and non-cardiac.

Nowadays, there are different computer tools such as machine learning, deep learning, and artificial intelligence, which, through algorithms, can find patterns and classify a large number of data. This is why it was decided to carry out a machine learning analysis of a database provided by Clinic Medical Norte in Tijuana, Baja California, Mexico. The results of this analysis suggest variables that can be considered secondary conditions to classify thoracic pain as cardiac in addition to those already established in the emergency department, such as Troponin levels, smoking habits, and diseases such as dyslipidemia, chronic kidney disease, diabetes, and hypertension.

**Author Contributions:** Conceptualization, C.Z. and C.C.-O.; Methodology, C.C.-O., V.R.-M. and C.Z.; Software, A.G.-S. and C.C.-O.; Validation, C.C.-O., V.R.-M. and C.Z.; Formal Analysis, A.G.-S., C.C.-O. and V.R.-M.; Investigation, V.R.-M., C.C.-O.; Resources, C.Z.; Data Curation A.G.-S., C.C.-O. and V.R.-M.; Writing-Original Draft Preparation, C.C.-O., C.Z., V.R.-M. and A.G.-S.; Writing-Review & Editing, C.C.-O., C.Z., V.R.-M. and A.G.-S.; Visualization, C.Z.; Supervision, C.C.-O. and C.Z.; Project Administration, C.C.-O. and V.R.-M.; Funding Acquisition, V.R.-M. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research was funded by the 1st Internal Call for Research Projects from CETYS University. The authors gratefully acknowledge CETYS University research coordination for the support for the realization of this project.

**Institutional Review Board Statement:** The study was conducted according to the guidelines of the Declaration of Helsinki and approved by the Ethics Committee of Clinic Medical Norte (protocol code 0051 and date 16/12/2019 of approval).

**Informed Consent Statement:** Informed consent was obtained from all subjects involved in the study. **Acknowledgments:** The authors would especially like to express their gratitude to Clinic Medical Norte. **Conflicts of Interest:** The authors declare no conflict of interest.

#### **Appendix A**



#### **Appendix B**

#### **Table A2.** Variables used by methods.


Note. "Framingham risk score for estimation of 10-years of cardiovascular disease risk in patients with metabolic syndrome" by Jahangiry, L., Farhangi M.A. and Rezaei, F., 2017 (https://www.ncbi.nlm.nih.gov/pmc/articles/PMC5682637/, accessed on 17 February 2021). Copyright 2017 by Jahangiry, L., Farhangi M.A. and Rezaei, F. "ACC/AHA ASCVD Risk Calculator" by ACC/AHA, 2013 (http:// www.cvriskcalculator.com/, accessed on 17 February 2021). Copyright 2013 by ACC/AHA. "SCORE Risk Charts" by European Society of Cardiology, 2020 (https://www.escardio.org/Education/Practice-Tools/CVD-prevention-toolbox/SCORE-Risk-Charts, accessed on accessed on 17 February 2021). Copyright 2020 by European Society of Cardiology.

#### **References**

