An Improved Supervised Classification Algorithm in Healthcare Diagnostics for Predicting Opioid Habit Disorder

Introduction

Opioid habit disorder is characterized as dependence on opioids, a medication present in many legal prescription medications, and illicit drugs such as heroin. The World is in the middle of an outbreak of drug addiction, and it has now become a problem for public health. Growing the risk of relapse can be due to many causes, including mental disorders, intimate and social interaction difficulties, exposure to other opioid users, and previous stressful experiences (National Institute on Drug Abuse, 2020). Contemporary research has correlated those demographic, social, physical, and psychological factors with drug use diseases. However, several of these experiments aim to understand drug use disorders by considering each function in isolation or one at a time. For example, the Department of Health and Human Services (DHHS) reports that the risk of drug misuse is raised by living with a disability (National Rehabilitation Information Center, 2011). Similarly, a correlation between mental illness and drug abuse (Saffer & Dave, 2002) is documented by the National Bureau of Economic Research. To decide who is likely to acquire OHD, it is essential to step from clarification to prediction, where intricate correlations can be considered between these characteristics, which will assist public health authorities in preparing and providing appropriate plans for action and assistance.

This study investigates supervised machine learning to construct a classifier based on an interconnected consideration of social, social, physical, and psychological characteristics to identify individuals at risk for OHD. Ιt aims to improve algorithm performance using a general framework for Opioid Habit Disorder (OHD) Diagnostic Model in order:

• •

  To investigate and examine massive data to discover the hidden patterns in any disease to deliver adapted dealing and predict the disease in any patient.

• •

  To propose a generalized model (pro-IDT) for forecasting a disease in the healthcare sector.

• •

  To compare the proposed model with machine learning algorithms such as ID3, Random Forest, and Support Vector Machine in Python programming.

• •

  To access the performance analysis of proposed work and other machine-learning algorithms in accuracy, precision, misclassification rate, recall, specificity, and F1 score.

Background

To analyze factors impacting the use of opioids over a long period, multiple studies on long-term patterns in opioid use have been carried out. ε.g., risk factors such as preoperative opioid use along with younger age, anxiety or depression, females, and other intrinsic factors are reported in (Bedard et al., 2017) with a higher number of refills or use at 12 months post-TKR. In (Martel et al., 2013), identical conclusions are drawn. It is seen in (Goesling et al., 2016) that the risk factors include more extreme pain, poor functioning, signs of depression, and a higher preoperative dosage of opioids. Good predictors for chronic opioid use are indicated for the form of surgery, prolonged hospital visits, discharge to the recovery unit, preoperative opioid use, higher comorbidity score, back pain, migraine, and smoking at baseline (Miotto et al., 2016). Also, previous use of opioids is often recognized as a significant risk factor for prolonged opioid use (Rozell et al., 2017). Stress or anxiety and discomfort are summarized in (Li et al., 2018) to be significantly linked to chronic pain following surgery or drug utilization postoperatively.