Men and women differ in factors precipitating opioid use disorder relapse

Rates of opioid use and opioid use disorder remain high, with life-altering personal and economic consequences for some. Efficacious treatments exist for patients with opioid use disorder, yet relapse after treatment remains common. Research on patient characteristics that predict post-treatment relapse risk can inform which types of individuals would benefit from additional supports. The authors of this study applied advanced statistical approaches to identify novel predictors of post-treatment opioid use, with a particular focus on factors associated with opioid use between men vs. women.

WHAT PROBLEM DOES THIS STUDY ADDRESS?

Opioid use and opioid use disorder remain significant public health concerns. Despite the availability of evidence-based treatments for opioid use disorder, maintaining opioid abstinence post-treatment is a challenge for many, especially within the first year following treatment. There is a growing body of research examining factors associated with opioid use and relapse, but relatively less research examining factors associated with post-treatment opioid use among men vs. women. This is important given research showing greater consequences of substance use disorder in women as well as evidence that women and men may benefit from services in different ways despite similar response to care. In addition, novel approaches to analyzing data, such as machine learning methods, offer innovative ways to identify the highest impact and most important predictors of post-treatment opioid use. Research that identifies these factors in novel and comprehensive ways can inform treatment for individuals with opioid use disorder. For example, individuals with factors that predict greater risk for post-treatment opioid use may warrant additional interventions and recovery support. In this study, researchers used a large, longitudinal dataset of individuals with opioid use disorder who received outpatient addiction treatment to examine the most critical characteristics related to opioid use during the first post-treatment year and whether these characteristics were similar or different for men and women.

HOW WAS THIS STUDY CONDUCTED?

This study was an exploratory analysis that leveraged existing, longitudinal data from the Global Appraisal of Individual Need (GAIN) dataset. The data for this study includes 1,126 individuals 12 years of age and older who received outpatient substance use disorder treatment between 2002 and 2013 and were diagnosed with opioid use disorder. These data were gathered from across 137 treatment centers across the United States who received funding from the Substance Abuse and Mental Health Services Administration (SAMHSA) as part of an initiative to improve and expand evidence-based treatments to patients with substance use disorder.

Individuals in the study first completed a baseline assessment when they entered outpatient treatment. Treatment types were not specified by the authors but included an array of “behavioral” interventions that varied by treatment site. Participants then completed follow up assessments 3, 6, and 12 months later.

Opioid use after treatment was assessed by asking participants to report whether they had used opioids since their last follow up appointment and, if “yes,” to indicate the number of days between their last assessment and first use of opioids. The researchers used this information to index the duration of opioid abstinence for each participant following initiation of treatment. In addition, the team had access to demographic information, as well as detailed psychiatric data that included co-occurring psychiatric diagnoses (e.g., major depressive disorder) and severity, co-occurring substance use disorders (e.g., alcohol use disorder) and severity, withdrawal symptoms, conduct problems/disorder, suicide risk, criminal justice involvement, mutual-help meeting involvement, and other stressors such as homelessness.

The team used two statistical approaches to identify factors associated with post-treatment opioid use. Using a more traditional approach, they tested which factors were associated, statistically, with opioid use. Using a more novel “machine learning” approach, they also examined the relative importance of each factor in the prediction of post-treatment opioid use. Together, both approaches provide a comprehensive understanding of the characteristics that place individuals with opioid use disorder at greatest risk for opioid use after receiving treatment.

The highest proportion of the sample was between 18 and 25 (41%), followed by 26+ (32%), and lastly 12-17 (27%). The average age of the sample was 24.5 years. Slightly over half of the sample was male (55%; 45% female), and identified as White (66%), followed by Hispanic (20%), multi-racial (10%), Black/African American (3%), and “other”/unspecified (1%). Just over half of men (52%) and women (55%) used opioids in the 12 months following treatment.

WHAT DID THIS STUDY FIND?

Importantly, opioid use outcomes during the year after receiving treatment were similar for men and women. Median time to opioid use for women was 64 days and for men was 60 days. There were some key differences by gender, however, on the factors that most strongly predict post-treatment opioid use.

Risk factors identified for women.

Using the more traditional approach, the authors found that more severe physical withdrawal symptoms, younger age, having a previous substance use disorder, or being more resistant to treatment (e.g., having greater difficulty in treatment or resisting substance use treatment) were the biggest predictors of post-treatment opioid use. Using the newer machine learning method, the authors found that having more substance use-related problems or negative consequences, followed by criminal justice involvement, younger age, and higher withdrawal severity, ordered from most to least important, were the biggest risk factors for opioid use. Although identified as lower magnitude and of lesser importance, depressive symptoms and posttraumatic stress disorder were also associated with greater opioid use risk for women but not men.

Risk factors identified for men.

For men, the traditional approach showed that more severe conduct disorder, younger age, and having multiple substance use disorder diagnoses were the highest-impact risk factors for post-treatment opioid use. With respect to relative importance in the prediction of opioid use, the machine learning results showed that younger age was first, followed by more severe conduct disorder symptoms, multiples substance use disorders, and finally being involved in the criminal justice system.

Risk of returning to opioid use post-treatment. The researchers used hazard ratios (HR) to report the risk of returning to opioid use. There are 2 different ways to read HRs here, depending on the type of variable being looked at. For “yes or no” variables, any HR above 1 indicates an increased risk of returning to opioid use, while any HR below 1 indicates decreased risk. For example, in the female category, does someone have a PTSD diagnosis? If yes, the HR of 1.03 indicates they have increased risk of opioid use. For continuous variables, something like age or varied scores on the treatment resistance index where values exist on a scale, an HR above 1 would indicate that someone with more of that variable is at increased risk (a higher score on the treatment resistance index, for example), while an HR below 1 indicates that someone with more of that variable is at decreased risk, like higher age (alternatively, this can be read as someone with less of that variable being at increased risk). Essentially, for age (HR = 0.88), the lower HR means that individuals are less likely to return to opioid use as they get older, but younger individuals would be at increased risk comparatively.

WHAT ARE THE IMPLICATIONS OF THE STUDY FINDINGS?

Using different statistical approaches, a more traditional survival analysis approach and a newer machine learning approach, the authors identified the highest magnitude (survival analysis) and rank order (machine learning) of risk factors for opioid use following substance use disorder treatment. The authors spotlighted gender differences with respect to post-treatment opioid use as well and found shared and differential risk factors for men vs. women. Research of this kind is important for identifying risks for opioid use and could aid in further tailoring of assessment, prevention, and interventions to reduce risk of post-treatment opioid use for men and women.

Across both statistical approaches, younger age, more withdrawal symptoms, having previously attended substance use disorder treatment, more severe substance use-related problems, treatment resistance, and more days of criminal justice involvement were the identified risk factors for opioid use among women. Similarly, risk factors identified for men also included younger age and more days of criminal justice involvement. Additional risk factors for men that were not identified for women included more severe conduct disorder symptoms and having multiple substance use disorders. Conversely, risk factors identified for women and not men included withdrawal symptoms, having a previous substance use disorder, more severe substance use-related problems, treatment resistance, a posttraumatic stress disorder diagnosis, and depressive symptoms.

These findings suggest the presence of shared risk factors for post-treatment opioid use for both men and women but also distinct risk factors that could become targets of enhanced screening, prevention, and intervention. Specifically, younger women with more severe withdrawal and fewer days of criminal justice involvement, and younger men with higher conduct disorder related symptoms, were the highest risk groups identified, and these risk profiles could help to identify for whom enhanced screening and relapse prevention support may be needed. Specifically, regarding criminal justice involvement, it is possible that there were high stakes contingencies for these individuals after treatment, whereby they were required to provide random drug tests and testing positive for opioid use may have resulted in consequences. There are many risk factors introduced by criminal justice involvement among individuals with opioid use disorder, including additional barriers to building recovery capital (e.g., employment barriers based on criminal record) and heightened risk for overdose among those recently released from jail/prison. Thus, these data showing a protective aspect of criminal justice involvement likely based on contingency management principles suggesting research into what is beneficial about this involvement so that it can be leveraged in context of a public health rather than criminal justice approach to substance use disorder treatment. Lastly, posttraumatic stress disorder and depressive symptoms also increased risk for post-treatment opioid use among women, suggestive of a higher need for psychiatric care that is inclusive of these concerns to promote post-treatment recovery.

LIMITATIONS

1. The authors used retrospective recall of past 90-day opioid use as their outcome variable. This has been used extensively in prior research but has limitations to the extent that participants’ reports reflect imperfect recall of if/when they used opioids. Response bias, or the tendency to respond in ways that seem socially desirable, can also influence participant reports.
2. While the outcome focused on here was any opioid use, whether individuals identified their opioid use disorder as their primary disorder, and the nature and severity of their opioid use could not be determined from the data presented here.
3. The researchers included many relevant predictors in their models but were unable to account for several key factors, such as location of the individuals in the study, site-specific information, medication use (including medication for opioid use disorder), actual treatment received, length of treatment received, post-treatment recovery and relapse prevention services, or other relevant services and supports that could aid in recovery for patients with opioid use disorder.
4. Gender group analyses were limited to male and female identified individuals. Whether individuals identified as transgender or non-binary could not be determined from the data presented here.
5. The data from this study were gathered between 2002 and 2013 and replication using more current data may be of added value.

BOTTOM LINE

Patients with opioid use disorder are at significant risk for opioid use following treatment, and it is imperative to continue to identify risks for this occurrence. It is equally important to identify who is at risk and when to develop increasingly targeted screening, prevention, and intervention campaigns. Using both traditional and novel machine learning approaches, the research team found greatest risk for post-treatment opioid use among younger women with more severe withdrawal and fewer days of criminal justice involvement, and younger men with greater conduct disorder related symptoms.

CITATIONS

Davis, J. P., Eddie, D., Prindle, J., Dworkin, E. R., Christie, N. C., Saba, S., . . . & Kelly, J. F. (2021). Sex differences in factors predicting post-treatment opioid use. Addiction, 116(8), 2116–2126. doi: 10.1111/add.15396*

*This study features RRI personnel but was reviewed independently for this bulletin