Exploring the latent trait of opioid use disorder criteria among frequent nonmedical prescription opioid users

A Comparison of Veterans with Problematic Opioid Use Identified through Natural Language Processing of Clinical Notes versus Using Diagnostic Codes

Opioid use disorder is known to be under-coded as a diagnosis, yet problematic opioid use can be documented in clinical notes, which are included in electronic health records. We sought to identify problematic opioid use from a full range of clinical notes and compare the demographic and clinical characteristics of patients identified as having problematic opioid use exclusively in clinical notes to patients documented through ICD opioid use disorder diagnostic codes. We developed and applied a natural language processing (NLP) tool that combines rule-based pattern analysis and a trained support vector machine to the clinical notes of a patient cohort (n = 222,371) from two Veteran Affairs service regions to identify patients with problematic opioid use. We also used a set of ICD diagnostic codes to identify patients with opioid use disorder from the same cohort. The NLP tool achieved 96.6% specificity, 90.4% precision/PPV, 88.4% sensitivity/recall, and 94.4% accuracy on unseen test data. NLP exclusively identified 57,331 patients; 6997 patients had positive ICD code identifications. Patients exclusively identified through NLP were more likely to be women. Those identified through ICD codes were more likely to be male, younger, have concurrent benzodiazepine prescriptions, more comorbidities, and more care encounters, and were less likely to be married. Patients in both these groups had substantially elevated comorbidity levels compared with patients not documented through either method as experiencing problematic opioid use. Clinicians may be reluctant to code for opioid use disorder. It is therefore incumbent on the healthcare team to search for documentation of opioid concerns within clinical notes.

Opioid use disorder (OUD) and treatment for opioid problems among OUD symptom subtypes in individuals misusing opioids

Background

In 2021, approximately 60 million individuals worldwide and 9 million individuals in the United States (US) reported opioid misuse. In the US, 2.5 million have OUD, of which only about a third receive any substance abuse treatment. OUD is often regarded as a monolithic disorder but different opioid problem subtypes may exist beyond DSM-IV/5 criteria. Understanding the characteristics of these subtypes could be useful for informing treatment and intervention strategies.

Methods

Latent class analysis was used to identify OUD symptom subtypes among persons in the US who reported misusing prescription opioids or heroin in the 2015-2018 National Survey on Drug Use and Health (n=10,928). Regression analyses were utilized to determine associations between class membership and treatment receipt, as well as demographic characteristics and other comorbid conditions.

Results

Five classes were identified with unique OUD symptom patterns: Class 1: Asymptomatic (71.6%), Class 2: Tolerance/Time (14.5%), Class 3: Loss of Control/Pharmacological (LOC/Pharmacol) (5.7%), Class 4: Social Impairment (2.6%), and Class 5: Pervasive (5.6%). Nearly all persons in the LOC/Pharmacol, Social Impairment, and Pervasive classes met criteria for OUD (98-100%); however, they differed in receipt of past-year treatment for substance use (28%, 28%, 49%, respectively). Age, race, education, insurance status, and criminal activity were also associated with treatment receipt.

Conclusions

There were considerable differences in OUD symptom patterns and substance use treatment among individuals who misused opioids. The findings indicate a substantial unmet need for OUD treatment and point to patterns of heterogeneity within OUD that can inform development of treatment programs.

Psychometric Performance of a Substance Use Symptom Checklist to Help Clinicians Assess Substance Use Disorder in Primary Care

Importance

Substance use disorders (SUDs) are underrecognized in primary care, where structured clinical interviews are often infeasible. A brief, standardized substance use symptom checklist could help clinicians assess SUD.

Objective

To evaluate the psychometric properties of the Substance Use Symptom Checklist (hereafter symptom checklist) used in primary care among patients reporting daily cannabis use and/or other drug use as part of population-based screening and assessment.

Design, Setting, and Participants

This cross-sectional study was conducted among adult primary care patients who completed the symptom checklist during routine care between March 1, 2015, and March 1, 2020, at an integrated health care system. Data analysis was conducted from June 1, 2021, to May 1, 2022.

Main Outcomes and Measures

The symptom checklist included 11 items corresponding to SUD criteria in the Diagnostic and Statistical Manual for Mental Disorders (Fifth Edition) (DSM-5). Item response theory (IRT) analyses tested whether the symptom checklist was unidimensional and reflected a continuum of SUD severity and evaluated item characteristics (discrimination and severity). Differential item functioning analyses examined whether the symptom checklist performed similarly across age, sex, race, and ethnicity. Analyses were stratified by cannabis and/or other drug use.

Results

A total of 23 304 screens were included (mean [SD] age, 38.2 [5.6] years; 12 554 [53.9%] male patients; 17 439 [78.8%] White patients; 20 393 [87.5%] non-Hispanic patients). Overall, 16 140 patients reported daily cannabis use only, 4791 patients reported other drug use only, and 2373 patients reported both daily cannabis and other drug use. Among patients with daily cannabis use only, other drug use only, or both daily cannabis and other drug use, 4242 (26.3%), 1446 (30.2%), and 1229 (51.8%), respectively, endorsed 2 or more items on the symptom checklist, consistent with DSM-5 SUD. For all cannabis and drug subsamples, IRT models supported the unidimensionality of the symptom checklist, and all items discriminated between higher and lower levels of SUD severity. Differential item functioning was observed for some items across sociodemographic subgroups but did not result in meaningful change (<1 point difference) in the overall score (0-11).

Conclusions and Relevance

In this cross-sectional study, a symptom checklist, administered to primary care patients who reported daily cannabis and/or other drug use during routine screening, discriminated SUD severity as expected and performed well across subgroups. Findings support the clinical utility of the symptom checklist for standardized and more complete SUD symptom assessment to help clinicians make diagnostic and treatment decisions in primary care.

A latent class analysis of drug and substance use patterns among patients treated in emergency departments for suspected drug overdose

INTRODUCTION

Polysubstance use and misuse can increase risks for nonfatal and fatal drug overdose. To categorize drugs used in combination in nonfatal overdoses, we analyzed data from emergency department (ED) overdose-related visits in 18 states funded by CDC's Enhanced State Opioid Overdose Surveillance (ESOOS) program.

METHODS

From 2017 to 2018, 120,706 ED visits included at least one hospital discharge code indicating acute drug poisoning for opioids, stimulants, hallucinogens, cannabis, anti-depressants, sedatives, alcohol, benzodiazepines, or other psychotropic drugs. Latent class analyses were conducted to determine the groupings of drug combinations in overdose visits.

RESULTS

Latent class analyses indicated a model of 5 classes - mostly heroin overdose (42.5% of visits); mostly non-heroin opioid overdose/use (27.3%); opioid, polysubstance (11.0%); female, younger (<25 years), other non-opioid drugs (10.5%); female, older (>55 years), benzodiazepine (8.0%). Findings indicated that heroin continues to be a large burden to EDs, yet EDs are also seeing overdose survivors with polydrug toxicity.

CONCLUSIONS

Medication-assisted treatment could be initiated in the emergency department following overdose for patients with opioid use disorder, and post-overdose protocols, such as naloxone provision and linkage to treatment and harm reduction services, have the potential to prevent future overdose for those at risk.

An Item Response Theory Analysis of DSM-5 Heroin Use Disorder in a Clinical Sample of Chinese Adolescents

The study examined the dimensionality and psychometric properties of Diagnostic and Statistical Manual of Mental Disorders, 5th Edition (DSM-5) criteria for heroin use disorder in a clinical sample of Chinese adolescents using item response theory approach. We examined 168 adolescents aged 14.8–17.9 years, who were in treatment for heroin use disorder. A two-parameter logistic item response theory model was conducted to examine the severity and discrimination of DSM-5 criteria for heroin use disorder. Differential item functioning across age and ethnicity was assessed. Results supported the hypothesis that the DSM-5 criteria for heroin use disorder were arrayed an underlying unitary dimension of severity in clinical adolescents. All the items exhibited good discriminatory power in distinguishing between clinical adolescent heroin users. Although three criteria exhibited measurement non-invariance at the item level, the overall DSM-5 heroin use disorder diagnostic criteria set appears to achieve measurement invariance at the scale level.

Opioid-related hospitalizations in Pennsylvania: A latent class analysis

BACKGROUND

Opioid abuse is associated with substantial morbidity and often results in hospitalization. Despite this, patient-level factors associated with opioid-related hospitalizations are not well understood.

METHODS

We used the Pennsylvania Health Care Cost Containment Council dataset (2000-2014) to identify opioid-related hospitalizations using primary and/or secondary ICD-9-CM hospital discharge codes for opioid use disorder (OUD), opioid poisoning, and heroin poisoning. Latent class analyses (LCA) of patient-level factors including sociodemographic characteristics, pregnancy, alcohol, tobacco, other substance use, and psychiatric disorders were used to identify common patterns within hospitalizations.

RESULTS

Among 28,538,499 hospitalizations, 430,569 (1.5%) were opioid-related. LCA identified five latent class (LC) patient groups associated with opioid-related hospitalizations: pregnant women with OUD (LC1); women over 65 with opioid overdose (LC2); OUD, polysubstance use and co-occurring psychiatric disorders (LC3); patients with opioid overdose without co-occurring polysubstance use (LC4); and African American patients with OUD and co-occurring cocaine use (LC5). LC3 was the largest latent class (58.2%) with annual hospitalizations doubling over time.

DISCUSSION

Among patients with opioid-related discharges, we identified five subpopulations among this sample. These findings suggest increased outpatient OUD treatment, mental health service support for patients with co-occurring psychiatric disorders and polysubstance use to prevent overdose and hospitalization.

The latent trait of ICD-11 nicotine dependence criteria: Dimensional and categorical phenotypes

We aimed to identify phenotypes of DSM-ICD nicotine dependence among a representative sample of lifetime weekly smokers in the largest metropolitan area in South America. Data came from 1,387 lifetime weekly smokers in the São Paulo Megacity Mental Health Survey. We used exploratory factor analysis (EFA) and latent class analysis (LCA) on ICD-11 nicotine dependence proposed criteria to explore dimensionality and phenotypes profiles, followed by logistic regression models to examine the association between latent classes and socio-demographic, psychiatric and chronic medical conditions. Analyses were performed using Mplus taking into account the complex survey design features. An unidimensional model had the best EFA fit with high loadings on all criteria. Response patterns detected by LCA indicated class differences based on severity continuum: a "non-symptomatic class" (32.0%), a "low-moderate symptomatic class" (34.9%)-with high probability of the criterion "use in larger amounts", and a "high-moderate symptomatic class" (33.1%). We found an association between high-income and the intermediate class that differs from findings in high-income countries, and high likelihood of psychiatric comorbidity among the most symptomatic smokers. The best dimensional model that pulled together nicotine dependence criteria supported a single factor, in concordance with the changes proposed for ICD-11.