Weekly Scoping Review Due Items: Group 2 (Kate, Clay, Shivani, Jon) - Natural language processing in opioid related disorders research: a scoping review

Weekly Scoping Review Due Items: Group 2 (Kate, Clay, Shivani, Jon) - Natural language processing in opioid related disorders research: a scoping review

by Shivani Mehta -

Title: Natural language processing in opioid related disorders research: a scoping review

Authors: Shivani Mehta, Kate Chirikova, Clay Carter

Structured summary:

Background

Detection of opioid related disorders (ORD) is important but challenging. Natural language processing (NLP) techniques applied to clinical notes and other kinds of unstructured data (e.g. social media, health research databases) have a great potential for identifying cases of ORD but this approach is relatively new and not widely used. This scoping review explores literature in which NLP techniques  were used to identify evidence of opioid related disorders. We assess the advantages and shortcomings of this analytical approach based on the available literature.

Methods

This review followed the Preferred Reporting Items for Systematic Reviews and Meta-Analysis (PRISMA) guidelines. The literature search was performed in four databases: PubMed/Medline, Embase, Web of Science, and CINAHL. Two authors screened the articles for inclusion. Articles discussing identification of ORD and application of natural language processing were included in this review.

Results - TBD

Conclusions  - TBD

Introduction

Rationale 

NLP is considered the mainstream approach to handle text and speech data. NLP algorithms are able to identify health outcomes through electronic medical record (EMR) clinical notes data. Semi-automated NLP methods can accurately identify evidence of opioid related disorders in vast amounts of patient-level EMR clinical notes1. Applying these scalable methods may increase surveillance and help clinicians  identify patients with symptoms of opioid-related disorders1. Thus, is it vital for this scoping review to assess the landscape of NLP methods in opioid-related disorders.   

The intended audience of this scoping review includes bioinformaticists, epidemiologists, and clinicians. It is expected that the findings will be useful to understand the landscape of NLP methods in opioid-related disorders. To our knowledge, this is the first scoping review of the application of NLP methods in opioid-related disorders research. 

Objectives

The primary aim of this scoping review was to summarize the studies where NLP had been applied to identify evidence of opioid-related disorders. We assessed the types of data (i.e., structured, unstructured, electronic medical records, social media data) and types of open-source NLP software (i.e. cTAKES, Apache OpenNLP) that have been used to implement NLP methods. Where possible, we report sensitivity and specificity of NLP methods used in the studies in comparison to the gold standard (i.e. diagnosis by a clinician).



(Edited by Sepehr Hashemi - original submission Wednesday, January 26, 2022, 6:58 PM)

(Edited by Sepehr Hashemi - original submission Wednesday, January 26, 2022, 6:58 PM)

(Edited by Sepehr Hashemi - original submission Wednesday, January 26, 2022, 6:58 PM)

(Edited by Sepehr Hashemi - original submission Wednesday, January 26, 2022, 6:58 PM)

(Edited by Sepehr Hashemi - original submission Wednesday, January 26, 2022, 6:58 PM)

(Edited by Sepehr Hashemi - original submission Wednesday, January 26, 2022, 6:58 PM)

Re: Weekly Scoping Review Due Items: Group 2 (Kate, Clay, Shivani) - Natural language processing in opioid related disorders research: a scoping review

by William Brown III -
The introduction needs more substance. examples of usage, examples of the challenges, etc. See the others' posts. See the other group's posts. This is clearly a method scoping review, which will be highly valuable in informatics and other data methods journals. I like that you mentioned that you were using PRISMA as your scoping review method and I understand your focus topic-wise. NLP is the compelling part but you might also want to highlight "unstructured data" in your title if that is your data application focus in these reviews.

Re: Weekly Scoping Review Due Items: Group 2 (Kate, Clay, Shivani) - Natural language processing in opioid related disorders research: a scoping review

by Shivani Mehta -
Methods
This scoping review was conducted following the Preferred Reporting Items for Systematic Reviews and Meta-Analysis (PRISMA) guidelines.

SEARCH
We searched for published studies in PubMed/Medline, Embase, Web of Science, CINAHL, and Google Scholar on January 28, 2022. The references of studies selected for full text review were hand searched. We included controlled vocabulary terms, along with free-text words, related to “Opioid related disorders”, “Opioid use disorder”, “Opioid dependence”, “Natural language processing” (Table 1). We did not apply other filters related to language and time of publication. A detailed description of the search strategy and a Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) study selection flow diagram are in Table 1 and Figure 1, respectively. Search results were exported into Zotero and duplicates were removed.

Table 1 


ELIGIBILITY CRITERIA
 We defined study eligibility criteria for the RQs a priori using the population, exposure, comparator, outcomes, study designs, and timing (PECODT) framework2 (Table 2). For all RQs, the population of interest were patients with opioid related disorders. We defined opioid related disorders as broad term disorders related to or resulting from abuse or misuse of opioids. We use the word opioid related disorders to describe individuals with opioid use disorders, opioid addiction, opioid dependence, prescription opioid misuse and abuse. We excluded studies of patient populations with non-opioid substance use/abuse.

We considered the following study designs to be eligible: randomized clinical trials, nonrandomized comparative studies (NRCSs), observational cohort studies, case-control studies, cross-sectional, and single group studies. We excluded cost-effectiveness studies, economic evaluations, systematic reviews, and narrative reviews. We do not have any minimal sample size requirement for any of the study designs. Additional details are found in Table 2.

Table 2

STUDY SELECTION
Three independent researchers (SM, EC, CC) applied the eligibility criteria and independently screened all titles and abstracts returned by the search. Each study was screened by the two reviewers. We did this using Abstrackr (http://abstrackr.cebm.brown.edu). Discrepancies were resolved through discussion and, where necessary, through consultation with a third researcher. This process was repeated using the same eligibility criteria for all full-text articles deemed to be potentially eligible based on abstract screening.

DATA EXTRACTION AND ANALYSIS
 The data from the included studies were extracted in a standardized form created using Microsoft Excel software. The information extracted was the first author’s name, year of publication, country of origin, study design, study population, sample size (comparator groups where applicable), type of NLP method used, purpose of NLP, type of data structure for application, NLP method efficacy estimates (where applicable). Additional details are found in Appendix 1. Three independent researchers (SM, EC, CC) performed data abstraction. Data from each study were charted by two reviewers and any discrepancies were resolved through discussion.

The extracted data were summarized and categorized based on type of NLP method and data structure. This scoping review provides the overview of the literature with no critical appraisal of the research rigor.

Re: Weekly Scoping Review Due Items: Group 2 (Kate, Clay, Shivani) - Natural language processing in opioid related disorders research: a scoping review

by Shivani Mehta -
Submitting a PRISMA diagram for this week. We're in the process of screening full-text articles.

Re: Weekly Scoping Review Due Items: Group 2 (Kate, Clay, Shivani) - Natural language processing in opioid related disorders research: a scoping review

by Shivani Mehta -
RESULTS

Literature search results
The searches of PubMed, EMBASE, Web of Science, and CINAHL, yielded 143 records. Out of the 143 records, 79 (55%) were duplicates removed before screening. We screened 64 unique abstracts, of which we excluded 27 (42%). We screened 30 articles in full, of which we excluded 20. The most frequent reasons for exclusion of full-text articles were that they retrieved data from social media (10 articles; 50%) and the articles did not report an outcome of interest (8 articles; 40%). A detailed description of search strategy and a Preferred Reporting Items for Systematic Reviews
and Meta-Analyses (PRISMA) study selection flow diagram are in Appendix Figure XX.

Re: Weekly Scoping Review Due Items: Group 2 (Kate, Clay, Shivani) - Natural language processing in opioid related disorders research: a scoping review

by Kate Chirikova -
Link to the google doc of the draft:
https://docs.google.com/document/d/175w7I0TAQCTm8bwmpLsStdgHj1h9VX5PXRA74nPr0jM/edit