Improving prediction of opioid use disorder with machine learning algorithms

Opioid Use Disorder (OUD) refers to a concerning pattern of opioid use that results in considerable impairment or distress. It is a chronic and treatable condition that can impact individuals irrespective of their race, gender, income, or social standing. Therefore, accurately predicting individuals at high risk of developing opioid use disorder is of utmost significance. Extensive research has been undertaken to forecast opioid use disorder and determine the most effective predictors that can enhance accuracy levels. This research proposes an improvement to these studies by utilizing two feature selection techniques, Boruta and Recursive Feature Elimination (RFE), to identify the best predictor through a majority voting system. The selected feature, combined with 10 demographic, socioeconomic, physical, and psychological predictors (gender, age, employment Status (Full time, Part time,..), income, Level of the impact of the disorder on an adults life in 4 role domains (home management, work, social life, close relationships with other, education, age when first drink alcoholic beverage, any mental illness, age when first used Marijuana, overall health), resulted in better prediction accuracy using various machine learning classification algorithms. To build a labeled dataset, responses from the 2018 and 2019 edition of the National Survey on Drug Use and Health (NSDUH) were collected. This dataset was used to train and test several classification models (Artificial Neural Network, Naïve Bayes and XgBoost), with the average recall, precision, f1 score, accuracy and AUC compared using 50 iterations. The results showed that considering “psychotherapeutic dependence or abuse” or “illicit drug other than marijuana dependence or abuse” have significant impact on improving accuracy metrics of used machine learning algorithms. Both Xgboost and naïve bayes was able to predict patients at risk for OUD with an AUC of 0.995. This work can aid healthcare providers in determining appropriate preventive care and resources for at-risk patients.

Name thesis:degree_name

M.S. (Master of Science)

Level thesis:degree_level

Masters

Discipline thesis:degree_discipline

Computer Science (UMKC)

Grantor

University of Missouri--Kansas City

Year dc:date.issued

2023

Author dc:creator

• Razzazi, Sahar

Advisor dc:contributor.advisor

• Nait-Abdesselam, Farid

Handle dc:identifier.uri

https://hdl.handle.net/10355/96229

OAI identifier oai:identifier

oai:mospace.umsystem.edu:10355/96229