LOADING

Type to search

Generational Effects: What Public Health Administrative Data Tells Us About Heroin Use Disorder

Share

Illicit use of heroin has been a major public health crisis in the United States for decades, and the more recent availability of other opioid-based substances has caused the number of heroin users to increase further. More than seventy thousand Americans died from drug overdoses in 2017, from both illicit drugs and prescription opioids—a two-fold increase in a decade (NIDA, 2020). An analysis on data retrieved from the National Survey on Drug Use and Health (NSDUH) reported that from 2002 to 2013 the overall rate of heroin addiction increased from 1.0 per one thousand to 1.9 per one thousand, a 90 percent increase overall. They also found that annual average rates of past-year heroin use increased from 1.6 per one thousand from 2002 to 2004 to 2.6 per one thousand from 2011 to 2013 (Jones, Logan, Gladden, & Bohm, 2015).

The impact of increasing heroin use is felt all across the United States, including in the state of Maryland. In Maryland alone, 1,078 patients died in 2018 from heroin overdoses, representing more than half of opioid-related deaths (Maryland Department of Heath, 2017). There is some evidence that the death rate is decreasing in Maryland (CDC, 2020), suggesting that efforts to effectively address this crisis may be showing progress. In order to progress further, we need to understand more about how to effectively treat substance use disorder (SUD) clients when those clients come from different segments of the population. In this study, we analyze the Statewide Maryland Automated Record Tracking (SMART) system, which gathers administrative data on SUD treatment clients and treatment histories, to assess whether or not meaningful differences exist in how generational subgroups respond to different approaches to treatment.

Generational Effects and Substance Use Disorder

Substance use patterns go hand-in-hand with societal changes, as well as the availability of particular drugs and other illicit substances at different points in time (Durrant & Thakker, 2003). For example, baby boomers may have higher rates of SUDs relative to other generations due to the social movements of the 1960s and 1970s, which included a dramatic increase in experimentation with illicit drugs (Cross & Kleinhesselink, 1985). Other research has found that the baby boomers are becoming a larger proportion of clients seeking treatment (Chhatre, Cook, Mallik, & Jayadevappa, 2017). On the other hand, the tendency to experiment with heroin has dropped dramatically in individuals between the ages of eighteen and twenty-five, though the drop may be offset by higher rates of cigarette, alcohol, cocaine, methamphetamine, and LSD use (SAMHSA, 2018). A deeper understanding of generational differences can be expected to help government officials, medical professionals, researchers, and caregivers better understand the contours of substance use among different generations, and perhaps inform different approaches to treatment. In this study we analyze generational effects to give more insight into the interaction between treatment options and treatment effectiveness. We build a longitudinal data set that enables us to track subjects’ substance use and treatment trajectory as long as they are in treatment in Maryland. The goal of this study is to better understand treatment decisions and outcomes over the course of treatment for different generational populations; insights into this relationship are expected to play a useful role in directing heroin use disorder treatment efforts and improving models of care.

Data: Maryland’s SMART System

Data were obtained from administrative data collected by the SMART system, a case management database for individuals receiving SUD treatment within the state of Maryland (Rezai-Zadeh et al., 2019). The SMART system captures statewide administrative client data from both publicly funded and private treatment providers from client admission to discharge. Data collected by the SMART system includes client demographic data and event-level treatment information. Another important feature of this system is that we are able to link individual-level treatment admission records within and across years via unique client IDs, enabling us to track individual clients over time. Though large-scale, multiyear studies of those receiving SUD treatment are rare, research analyzing large, treatment-related administrative data sets have shown interesting results (Evans, Grella, Murphy, & Hser, 2010; Luchansky, Brown, Longhi, Stark, & Krupski, 2000).

For this study, a total of 5,390 treatment-naïve clients indicating heroin as a primary drug of use at their time of initial treatment enrollment had a total of 13,009 treatment events recorded in the SMART system from 2007 through 2013. Treatment-naïve clients in this study were defined as those with a first treatment record in the SMART system between 2007 and 2010 who also self-reported no prior treatment for opioid or heroin use disorder. Clients who had heroin use disorder treatment records before 2007, who resided outside of Maryland, or who were discharged from treatment due to incarceration or death were excluded from the study population. A small number of clients who had missing race/ethnicity data were also excluded from the study population. For all cases studied, data from clients were collected from first treatment admission to four years after first admission. Each time clients reappeared in the SMART system after their initial treatment episode, a treatment readmission was recorded for them. The principal outcome measure in this study was an indicator of whether clients completed or were readmitted to treatment within four years of initial treatment.

The broad scope and large scale of our dataset allow for an analysis of the treatment and outcome patterns exhibited by heroin use disorder clients.

To control for baseline demographic measures, we included gender, race/ethnicity, age at time of initial treatment enrollment, employment status, and geographical region as control variables. Some measures, like employment status, varied over time or changed through readmission. For the purposes of this study, we coded variables as their values at initial treatment enrollment. We also included measures of initial treatment types and intensities, which corresponds to the American Society of Addiction Medicine (ASAM) levels of care (Mee-Lee et al., 2013). If clients were discharged from their initial treatment and did not require additional treatment, or completed their initial treatment plan and were only referred to additional treatment services, a successful treatment completion was recorded for them by the treatment facility. Finally, all clients were assigned an indicator for generation based on their age at first treatment: the millennial cohort (born between 1982 and the 2000s) contained 2,769 clients, the generation X cohort (born between 1961 and 1981) contained 2,380 clients, and the baby boomer cohort (born between 1940 and 1960) contained 241 clients.

Analytical Plan

First, summary statistics were analyzed for the client population as a whole as well as across generations. Pearson Chi-square statistics (Pearson, 1900) were calculated to evaluate significant associations between our measure of readmission and the covariates. Following that, a multivariate logistic model was fitted and tested with readmission status within four years after first discharge as the outcome measure, assessing the statistical significance level of each covariate. To help interpret the logistic model’s results, we calculated adjusted odds ratios for readmission and predicted probabilities across generational cohorts. All statistical analyses were performed using SAS 9.4 (SAS Institute, n.d.).

Results and Discussion

Using this rich data set, we were able to examine generational effects, and other potential significant covariates, for treatment success versus readmission. First we noted demographics of our clients clustered by generation. For the 2,769 client records that comprised the millennial cohort, the majority were male (59 percent, n=1,634), white (85 percent, n=2,367), unemployed, never married, and resided in an urban setting. As for the 2,380 client records that comprised the gen X cohort, the vast majority were male (77 percent, n=1,834), black (71 percent, n=1,687), unemployed, never married, and resided in an urban setting. The 241 baby boomer client records showed that the majority were even more male (80 percent, n=192), black (83 percent, n=199), unemployed, and resided in an urban setting. It is clear that notable client demographic characteristics differed across generations. Different drug availability across years and shifting reasons for consumption may help to explain this variation (“Drug and alcohol,” 2020). Some characteristics remained universal across generations, like a majority of our clients in all cohorts were unemployed and never married. With the assistance of Pearson Chi-square statistics, associations between those baseline characteristics and readmission were tested. For the millennial cohort, test results found gender (p < 0.01), race (p < 0.01), and employment status (p < 0.01) to be correlated to readmission status. For the gen X cohort, resident region (p < 0.01) was significant, indicating that clients who reside in suburban areas have a higher chance of completing treatment. Interestingly, the baby boomer cohort was the only group for which education (p= 0.04) was a significant covariate to substance use treatment completion.

Generational effects can also be seen initially when we examine the different treatments assigned to different clients; as we move from millennial, to generation X, to baby boomers we find that subjects are less likely be assigned to opioid maintenance therapy (OMT), even though this treatment seems to be a very effective option for everyone with respect to successfully completing treatment. Indeed, OMT was the most common treatment approach overall (n = 1468, 27 percent). Broken down by generation, however, about one-third of millennial clients got OMT as their first treatment while this proportion fell to 21 percent for gen X, and fell further to 17 percent for baby boomers. The gen X and the baby boomer cohorts are more likely to be assigned outpatient therapy treatment than OMT. From the perspective of treatment intensity, the millennial cohort was more likely to be assigned intensive inpatient services and their treatment plans have a higher likelihood of involving detox, while outpatient treatment (a lower intensity treatment) was more likely to be assigned to older generations.

Our results indicate generational effects are related to treatment approach, which is then related to the likelihood of successfully completing treatment. In the millennial cohort, of the 145 cases assigned clinically managed, high-intensity, residential treatment (5 percent of the sample), a large majority (76 percent, n=110) did not have a readmission within four years, making this treatment program the most effective one among all treatments we investigated. Outpatient treatment and OMT both have around a 47 percent no-readmission rate. However, medically monitored intensive inpatient treatment and outpatient ambulatory detox treatment had the lowest percentage with no readmission within four years (27 percent, n=55 and 29 percent, n=26, respectively).

From the gen X client records, of the three hundred cases in clinically managed, high-intensity, residential treatment (13 percent of the sample), a large majority (72 percent, n=215) did not have a readmission within four years, marking this as the most effective treatment. The second promising treatment was OMT, with a no-readmission rate within four years of 63 percent. However, medically monitored intensive inpatient detox services and intensive outpatient detox services had the highest readmission rates.

For baby boomers, of the thirty-five cases in outpatient treatment (15 percent of the sample), a large majority (83 percent, n=29) did not have a readmission within four years. Following that, a large percentage of medically monitored intensive inpatient detoxification services clients did not have a readmission within four years (78 percent, n = 25). In contrast, clinically managed medium or low-intensity residential treatment had the lowest percentage of no-readmission within four years (43 percent, n=6).

In conclusion, clinically managed high-intensity residential treatment services, on average, yield the best outcomes across younger generations, yet are not widely assigned. OMT was a popular approach to treating the millennial cohort and also came with exceptionally low readmission rates. It may be less effective for aging generations such as baby boomers, who also failed to show as positive a response to that form of treatment. In contrast, outpatient treatment outperforms other treatments in the older generation, but failed to show effective results for the younger generation.
To further explore our results, a multivariate logistic regression was conducted for the overall population with the inclusion of generational effects as covariates. In fact, generational cohort was the most significant covariate to readmission rate (p < 0.01). Roughly speaking, the odds of readmission for the baby boomer cohort was only 0.34 times of that in the millennial cohort, with a 95 percent confidence interval of 0.25, 0.46. A similar comparison between gen X and millennials indicates that the odds of readmission in gen X is 0.60 times the same measure in the millennial cohort, with a 95 percent confidence interval of 0.52, 0.69.

Finally, interaction terms between gender, race, region, and generation were investigated. These three interaction terms were found to be significant with p < 0.001, p < 0.001, and p = 0.005 respectively. Some interesting findings from the model with interaction terms include: female clients have a lower predicted probability of readmission among baby boomers, while the probability of readmission is roughly even in the gen X cohort, and then the probability of female readmission outpaces males among millennials. Regarding race, black and white clients did not show a significant difference in baby boomer and gen X cohorts. However, in the millennial cohort white clients were found to have a significantly higher chance of readmission within four years.

Our results suggest that health care providers can benefit from an understanding of generational cohorts and their relationship to both treatment and successful treatment completion. Different generations respond in systematic ways to different treatments. Paying attention to generational differences can be a tool to promote treatment effectiveness.

Study Limitations

The broad scope and large scale of our dataset allow for an analysis of the treatment and outcome patterns exhibited by heroin use disorder clients. Notwithstanding the clear generational trends presented here, there are a few limitations we would like to discuss. A very important limitation is that without clinical context it may be hard to rely too much on treatment readmission as an indicator of client outcomes. Readmission by itself is not a perfect indicator of program performance and outcome (Humphreys & Weingardt, 2000). Readmission may indicate that clients are seeking further help and support for their ongoing substance abuse, which could be a positive sign of treatment engagement and retention. Future studies, in collaboration with clinicians working in this area, are currently underway to build on these results and address the questions surrounding treatment readmission and its relation to completion, retention, and outcome across generations.

Additionally, our data source only captures treatment provided within Maryland. If clients move out of state, our data will not contain future treatment information. Using Maryland as a source of data limits the generalizability of our findings to the extent that the state is not necessarily representative of the nation or other areas. Our analysis also depends on self-reported data and accurate recording of treatment and outcome characteristics.

Another potential limitation is that we only examine the influence of clients’ first treatments. Clients can be admitted to multiple modalities in succession during a treatment episode. It is possible that clients’ first treatments are brief and therefore have little impact on readmission relative to subsequent treatment admissions during a single episode. At the same time, over 50 percent of our study population that experienced a readmission return to the same treatment modality, and about 60 percent return to the same level of care, suggesting that initial treatment has a profound influence on overall treatment plans.

In future studies more complicated treatment combinations will be measured to help us understand the effect of multiple treatment approaches. Despite these limitations, we do believe that these data are unusually detailed and provide us with the best available information on individual-level client characteristics and treatment decisions, over time, for those undergoing treatment for heroin use over the course of several years.

Conclusion

Results from this study indicate that there are generational effects when it comes to the assignment of treatments for, and outcomes associated with, a heroin-based SUD. Clinicians may be better able to serve their clients when they understand that generations respond in different ways to different treatment options. It is troubling to observe in our data that, while treatment decisions do systematically vary with clients’ generations, those variations do not conform to the findings about which treatments tend to be most effective for different generational groups.

References

  • Centers for Disease Control and Prevention (CDC). (2020). Heroin overdose data. Retrieved from https://www.cdc.gov/drugoverdose/data/heroin.html
  • Chhatre, S., Cook, R., Mallik, E., & Jayadevappa, R. (2017). Trends in substance use admissions among older adults. BMC Health Services Research, 17(1), 1–8.
  • Cross, H. J., & Kleinhesselink, R. R. (1985). The impact of the 1960s on adolescence. The Journal of Early Adolescence, 5(4), 517–31.
  • “Drug and alcohol abuse across generations.” (2020). Retrieved from https://drugabuse.com/featured/drug-and-alcohol-abuse-across-generations/
  • Durrant, R., & Thakker, J. (2003). Substance use and abuse: Cultural and historical perspectives. London, UK: Sage Publications.
  • Evans, E., Grella, C. E., Murphy, D. A., & Hser, Y.-I. (2010). Using administrative data for longitudinal substance abuse research. The Journal of Behavioral Health Services & Research, 37(2), 252–71.
  • Humphreys, K., & Weingardt, K. R. (2000). Assessing readmission to substance abuse treatment as an indicator of outcome and program performance. Psychiatric Services, 51(12), 1568–9.
  • Jones, C. M., Logan, J., Gladden, R. M., & Bohm, M. K. (2015). Vital signs: Demographic and substance use trends among heroin users — United States, 2002–2013. MMWR, 64(26), 719–25.
  • Luchansky, B., Brown, M., Longhi, D., Stark, K., & Krupski, A. (2000). Chemical dependency treatment and employment outcomes: Results from the ‘ADATSA’ program in Washington State. Drug and Alcohol Dependence, 60(2), 151–9.
  • Maryland Department of Health. (2017). Unintentional drug- and alcohol-related intoxication deaths in Maryland: Annual report 2017. Retrieved from https://bha.health.maryland.gov/OVERDOSE_PREVENTION/Documents/Drug_Intox_Report_2017.pdf
  • Mee-Lee, D., Shulman, G. D., Fishman, M. J., Gastfriend, D., Miller, M. M., Provence, S. M. (2013). The ASAM criteria: Treatment criteria for addictive, substance-related, and co-occurring conditions. Chevy Chase, MD: American Society of Addiction Medicine, Inc.
  • National Institute on Drug Abuse (NIDA). (2020). Trends & statistics: Overdose death rates. Retrieved from https://www.drugabuse.gov/drug-topics/trends-statistics/overdose-death-rates
  • Pearson, K. (1900). On the criterion that a given system of deviations from the probable in the case of a correlated system of variables is such that it can be reasonably supposed to have arisen from random sampling. Philosophical Magazine Series, 5(302), 157–75.
  • Rezai-Zadeh, K. P., Engstrom, R. N., Sharma, A., Chen, Y., Chu, J., Cox, R. P., & Lee, M.-L. T. (2019). Generational trends and patterns in readmission within a statewide cohort of clients receiving heroin use disorder treatment in Maryland, 2007–2013. Journal of Substance Abuse Treatment, 96, 82–91.
  • SAS Institute, Inc. (n.d.). SAS 9.4 – Today and tomorrow. Retrieved from https://support.sas.com/software/94/index.html
  • Substance Abuse and Mental Health Services Administration (SAMHSA). (2018). Key substance use and mental health indicators in the United States: Results from the 2017 National Survey on Drug Use and Health. Retrieved from https://www.samhsa.gov/data/sites/default/files/cbhsq-reports/NSDUHFFR2017/NSDUHFFR2017.pdf

Editor’s Note:

This article was adapted from an article by the same authors previously published in the Journal of Substance Abuse Treatment (JSAT). This article has been adapted as part of Counselor’s memorandum of agreement with JSAT. The following citation provides the original source of the article: Rezai-Zadeh, K. P., Engstrom, R. N., Sharma, A., Chen, Y., Chu, J., Cox, R. P., & Lee, M.-L. T. (2019). Generational trends and patterns in readmission within a statewide cohort of clients receiving heroin use disorder treatment in Maryland, 2007–2013. Journal of Substance Abuse Treatment, 96, 82–91.

Richard Engstrom
Richard N. Engstrom, PhD

Richard N. Engstrom, PhD, is associate director of applied research and outreach at the University of Maryland’s Institute for Governmental Service and Research. He has published research on the influence of generational change on public policies, election systems, and other factors related to political representation.

Yiming Chen, MS

Yiming Chen, MS, is currently a second year PhD student in the biostatistics and bioinformatics program at the University of Maryland, College Park. Her research interests include data analysis, survival analysis, and clinical trial design. She has worked for the University of Maryland’s Institute for Governmental Service and Research as a graduate assistant for the past two years.