國立虎尾科技大學 |

Leveraging Machine Learning Framework to Predict Buprenorphine/Naloxone Treatment Discontinuation.

Record Type:	Language materials, printed : Monograph/item
Title/Author:	Leveraging Machine Learning Framework to Predict Buprenorphine/Naloxone Treatment Discontinuation./
Author:	Mohite, Prathamesh .
Published:	Ann Arbor : ProQuest Dissertations & Theses, : 2020,
Description:	23 p.
Notes:	Source: Masters Abstracts International, Volume: 81-06.
Contained By:	Masters Abstracts International81-06.
Subject:	Operations research. -
Online resource:	http://pqdd.sinica.edu.tw/twdaoapp/servlet/advanced?query=27669114
ISBN:	9781392659496

Leveraging Machine Learning Framework to Predict Buprenorphine/Naloxone Treatment Discontinuation.
Mohite, Prathamesh .

Leveraging Machine Learning Framework to Predict Buprenorphine/Naloxone Treatment Discontinuation. - Ann Arbor : ProQuest Dissertations & Theses, 2020 - 23 p.

Source: Masters Abstracts International, Volume: 81-06.

Thesis (M.S.)--Northeastern University, 2020.

This item must not be sold to any third party vendors.

Opioid use disorder (OUD) is a growing public health concern in the United States, causing nearly 130 deaths every day. The negative economic health burden attributed to OUD is substantial that exceeds roughly $56 billion annually. Buprenorphine/naloxone is an evidence-based and cost-effective opioid agonist therapy most commonly prescribed for treating OUD. Retaining in the treatment for long-term treatment with better continuity is proven to produce better treatment outcome in terms of reduced relapse rate and abstinence. However, premature discontinuation from the buprenorphine/naloxone treatment often times leads to unfavorable and negative health outcomes such as higher rate of relapse and increased risk of other substance use disorder. Hence, there exists a critical need to proactively identify individuals who are at risk of discontinuing buprenorphine/naloxone treatment prematurely. Till date, there has been little in-depth research that focuses on predicting buprenorphine/naloxone treatment discontinuation. To that end, in this study, we propose a framework leveraging state-of-the-art machine learning algorithms to predict premature treatment discontinuation, in particular for naive patients on buprenorphine/naloxone treatment. We used All Payer Claims Data (APCD), a database that records pharmacy claims for state’s commercially insured population. We designed a retrospective longitudinal study to investigate the buprenorphine refill pattern for those naive patients. In addition, we identified several patient-level demographic and prescriberlevel characteristics to be used as features in the predictive models. We then implemented several binary classification algorithms including logistic regression, decision tree, random forest, neural network, boosted trees classifier, support vector machine, naive Bayes and soft voting ensemble models to predict treatment discontinuation. Models are compared on the basis of accuracy, precision, recall and area under receiving operating characteristics curve (C-statistics). Results showed that random forest model outperforms other models with a high C-statistics of 82.16% and a recall score of 93.75%. Features having significant importance on predicting treatment discontinuation were identified from feature importance plots. Stratification of new patient records based on their discontinuation risk level was done using Youden index to optimize the threshold on predicted propensities. This study has the potential to help clinicians improve existing treatment guidelines in order to reduce the premature discontinuation from buprenorphine/naloxone treatment.

ISBN: 9781392659496Subjects--Topical Terms:

573517
Operations research.
Subjects--Index Terms:

Buprenorphine/Naloxone

Leveraging Machine Learning Framework to Predict Buprenorphine/Naloxone Treatment Discontinuation.
LDR:03832nam a2200385 4500 001 951882
005 20200821052217.5
008 200914s2020 ||||||||||||||||| ||eng d
020 $a 9781392659496
035 $a (MiAaPQ)AAI27669114
035 $a AAI27669114
040 $a MiAaPQ $c MiAaPQ
100 1 $a Mohite, Prathamesh . $3 1241378
245 1 0 $a Leveraging Machine Learning Framework to Predict Buprenorphine/Naloxone Treatment Discontinuation.
260 1 $a Ann Arbor : $b ProQuest Dissertations & Theses, $c 2020
300 $a 23 p.
500 $a Source: Masters Abstracts International, Volume: 81-06.
500 $a Advisor: Noor-E-Alam, Md.
502 $a Thesis (M.S.)--Northeastern University, 2020.
506 $a This item must not be sold to any third party vendors.
520 $a Opioid use disorder (OUD) is a growing public health concern in the United States, causing nearly 130 deaths every day. The negative economic health burden attributed to OUD is substantial that exceeds roughly $56 billion annually. Buprenorphine/naloxone is an evidence-based and cost-effective opioid agonist therapy most commonly prescribed for treating OUD. Retaining in the treatment for long-term treatment with better continuity is proven to produce better treatment outcome in terms of reduced relapse rate and abstinence. However, premature discontinuation from the buprenorphine/naloxone treatment often times leads to unfavorable and negative health outcomes such as higher rate of relapse and increased risk of other substance use disorder. Hence, there exists a critical need to proactively identify individuals who are at risk of discontinuing buprenorphine/naloxone treatment prematurely. Till date, there has been little in-depth research that focuses on predicting buprenorphine/naloxone treatment discontinuation. To that end, in this study, we propose a framework leveraging state-of-the-art machine learning algorithms to predict premature treatment discontinuation, in particular for naive patients on buprenorphine/naloxone treatment. We used All Payer Claims Data (APCD), a database that records pharmacy claims for state’s commercially insured population. We designed a retrospective longitudinal study to investigate the buprenorphine refill pattern for those naive patients. In addition, we identified several patient-level demographic and prescriberlevel characteristics to be used as features in the predictive models. We then implemented several binary classification algorithms including logistic regression, decision tree, random forest, neural network, boosted trees classifier, support vector machine, naive Bayes and soft voting ensemble models to predict treatment discontinuation. Models are compared on the basis of accuracy, precision, recall and area under receiving operating characteristics curve (C-statistics). Results showed that random forest model outperforms other models with a high C-statistics of 82.16% and a recall score of 93.75%. Features having significant importance on predicting treatment discontinuation were identified from feature importance plots. Stratification of new patient records based on their discontinuation risk level was done using Youden index to optimize the threshold on predicted propensities. This study has the potential to help clinicians improve existing treatment guidelines in order to reduce the premature discontinuation from buprenorphine/naloxone treatment.
590 $a School code: 0160.
650 4 $a Operations research. $3 573517
650 4 $a Artificial intelligence. $3 559380
650 4 $a Health care management. $3 1148454
650 4 $a Public health. $3 560998
653 $a Buprenorphine/Naloxone
653 $a Machine learning
653 $a Opioid
653 $a Risk stratification
653 $a Treatment discontinuation
690 $a 0796
690 $a 0800
690 $a 0769
690 $a 0573
710 2 $a Northeastern University. $b Mechanical and Industrial Engineering. $3 845717
773 0 $t Masters Abstracts International $g 81-06.
790 $a 0160
791 $a M.S.
792 $a 2020
793 $a English
856 4 0 $u http://pqdd.sinica.edu.tw/twdaoapp/servlet/advanced?query=27669114