Abstract
Hypertension or high blood pressure is a critical health condition, that if it is not detected or treated in advance could cause sever to mortal cardiovascular diseases. Moreover, this type of health condition does not generate symptoms that the patient or medical expert could perceive or that are associated with it. Due to the use of electronic health records and the available medical data, it could be possible to monitor or generate an early diagnosis of hypertension by employing data processing techniques to enable the prediction or the presence of hypertension in an individual.
Introduction
Problem Statement
According to the World Health Organization (WHO), hypertension also referred to as high blood pressure, is a global public health issue \cite{hypertension}. Having a bad or no constant monitoring of blood pressure could lead to developing hypertension with associated diseases such as heart attack, heart failure, kidney disease, coronary heart disease, diabetes, or strokes. Furthermore, high blood pressure has been associated with at least 45% of deaths due to heart disease and 51% of deaths due to stroke in 2013 as established by the WHO \cite{hypertensiona}.
Hypertension also tends to be clinically costly, difficult to manage, and often leads to severe and mortal health conditions like cardiovascular diseases. Besides, hypertension affects more than 85 million people in the USA, and 1.1 billion around the world according to the information of WHO. Countries with weak health systems such as middle-income countries are the most affected by this type of health condition. Some of the most common risk factors associated with hypertension include age, gender, body mass index, obesity, stress, lipoproteins, cholesterol, physical activity, smoking, and family history. On the other hand, there is evidence from clinical trials that suggest that early prevention of hypertension, life habits, and adequate treatment can reduce the development of hypertension. Nevertheless, this type o health condition has the characteristic of rarely producing noticeable symptoms that a medical expert or patient can recognize without and explicit analysis of his health state, that is why hypertension is often referred to as a silent killer. Sometimes, hypertension could produce symptoms such as dizziness, chest pain, headaches, breathing difficulties, nosebleed, or heart palpitations, nonetheless, these symptoms could not signify hypertension \cite{Ye2018}.
On the other hand, the increase in the available data contained in electronic health records (EHRs) provides historical information that can be used to monitor the progression of diseases. EHRs have a great potential for accelerating clinical research and predictive analysis of a population. The versatility of EHR-based datasets provides more generalized prediction results with high levels of confidence, the clinical construction of EHR-based risk prediction models has become more effective and is in higher demand. (
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC5811646/). The availability of data has picked up the interest of using machine learning methods to classify or making predictions based on the current amount of clinical data to monitor or predict different types of health conditions or diseases. The machine learning algorithms used in the medical field go from traditional methods such as logistic regressions or linear regression to more complex techniques related to the subfield of deep learning known as artificial neural networks and its diverse type of topologies. Nonetheless, the development of machine learning models to predict or monitor disease does not try to substitute an actual medical expert but provides tools that could help to determine a proper verdict of the health conditions of patients. Moreover, According to Rashidi [8], it is important to count on a multidisciplinary team to apply efficiently the concepts of machine learning to medical situations.