A multi-task machine learning approach for comorbid patient prioritization

With the advent of Internet of Things, remote patient monitoring and diagnosis is becoming the more advanced and accessible feature in healthcare industry regardless of age and disorder. With the increase in patient population, remotely prioritizing and diagnosing patients becomes a major problem. Research has been conducted to address this issue. However, many models fail to support the condition of comorbidity. In this paper, we developed a multi-task machine learning model that supports remote patient prioritization based on disease severity. This model not only considers patients diagnosed with a single disease, but also acknowledges comorbidity of a patient to calculate their severity utilizing physiological data. Classifying a comorbid patient based on their disease(s) severity is a challenging task. We achieve this by exploiting the commonalities and differences between the diseases using multi-task machine learning. Based on the correlation between diseases, we classify the patient into appropriate severity level. In addition to this, our model also supports continuous patient monitoring based on their severity levels, which is made possible by implementing in Apache Spark. Apache Spark also serves as a real-time streaming processor suitable for remote patient monitoring and diagnosis.