https://orcid.org/0000-0001-9113-0334
Figures
Abstract
Background and objectives
Substance use disorders (SUDs) usually involve a complex natural trajectory of recovery alternating with symptom reoccurrence. This study examined treatment course patterns over time in a community SUD clinic. We examined depressive symptoms level, primary SUD assigned at each admission, and lifetime misuse of multiple substances as potential risk factors for premature treatment termination and subsequent treatment readmission.
Methods
De-identified longitudinal data were extracted from charts of 542 patients from an SUD treatment center. Survival analysis methods were applied to predict two time-to-event outcomes: premature treatment termination and treatment readmission.
Results
Primary opioid (vs alcohol) use disorder diagnosis at admission was associated with higher hazard of premature termination (HR = 1.91, p<0.001). The interaction between depressive symptoms level and substance use status (multiple vs single use) on treatment readmission was significant (p = 0.024), such that higher depressive symptoms level was predictive of readmission only among those with a history of single substance use (marginally significant effect). Lifetime use of multiple (vs single) substances (HR = 1.55, p = 0.002) and age (HR = 1.01, p = 0.019) predicted increased hazard of readmission.
Conclusions
Findings did not support a universal role for depressive symptoms level in treatment course patterns. Primary SUD diagnosis, age, and history of substance misuse can be easily assessed and incorporated into treatment planning to support SUD patients and families. This study is the first to our knowledge that afforded a stringent test of these relationships and their interactions in a time-dependent, recurrent event, competing risks survival analysis examining both termination and readmission patterns utilizing a real-world clinic-based sample.
Citation: Argyriou E, Bakoyannis G, Wu W, Rattermann MJ, Cyders MA (2023) Individual factors predict substance use treatment course patterns among patients in community-based substance use disorder treatment. PLoS ONE 18(1): e0280407. https://doi.org/10.1371/journal.pone.0280407
Editor: Keith Leverett Warren, The Ohio State University, UNITED STATES
Received: August 17, 2022; Accepted: December 28, 2022; Published: January 12, 2023
Copyright: © 2023 Argyriou et al. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Data Availability: Data cannot be shared publicly because the authors on this manuscript do not own the data and the data use agreement between the authors and the treatment center (who are the data owners) specified that we are not allowed to share it with any other entities. The data underlying the results presented in the study are available through contacting the Fairbanks Addiction Treatment Center (fairbankscd.org).
Funding: Financial support was provided by the National Institute on Alcohol Abuse and Alcoholism F31 fellowship grant (F31 AA028998) to E.A. and the Indiana University Responding to the Addictions Crisis grant to M.A.C and M.J.R. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.
Competing interests: The authors have declared that no competing interests exist.
Introduction
Substance use disorders (SUDs) involve a complex trajectory of recovery alternating with symptom reoccurrences [1, 2]. This is partly reflected in patterns of repeated treatment readmissions, which are associated with greater disruption to patients and their families and higher healthcare costs [3]. Premature treatment termination further contributes to these patterns by accelerating subsequent symptom relapse [4, 5]. It is essential to evaluate risk factors of premature SUD treatment discharge and readmission patterns to design more effective treatment and improve outcomes. A variety of risk factors for premature treatment termination and readmission have been identified in the literature (although not consistently) including systematic and social factors often driven by external variables, such as type of treatment modality [3], homelessness [6], health insurance [6], unemployment [7], as well as individual factors which could be more effectively used in treatment selection and planning, including demographics (primarily age), cognitive deficits, treatment alliance, and psychiatric comorbidity [8]. Depressive disorders are often comorbid with or secondary to SUD, predict a more severe and complicated SUD presentation [9], and may interfere with SUD treatment, reducing treatment completion and increasing the risk for relapse [10–12]. Higher depressive symptoms level could complicate clinical diagnosis and treatment [13] by impeding rapport and engagement in the treatment process, resulting in premature treatment attrition [14, 15] and accelerated treatment readmission through accelerating relapse [2, 10]. Depressive symptoms predict more subsequent treatment readmissions [11, 12] but may also decelerate readmission by impeding motivation to seek treatment [16]. The relationship between depressive symptoms level and SUD treatment outcomes has varied across studies [15, 17–23], suggesting the need to examine moderators of this relationship.
The effect of depressive symptoms level on treatment course may differ across primary SUD and lifetime substance misuse history. Evidence suggests that opioid use predicts increased dropout rates [24–26]; however, these findings have not been consistent and depend on the definition of substance use (e.g., diagnosis or general misuse) and samples studied [8]. Illicit substance use disorders are more strongly linked to depressive symptoms [27, 28] and are associated with distinct patterns of time to relapse and treatment retention [14, 29]. Misuse of multiple compared to single substances or presence of multiple SUDs is associated with increased likelihood of depression [27] and readmission to treatment after discharge [17].
Previous studies have primarily examined the number of readmissions in a specific time period, drop-out rates, or treatment length [11, 12]. This fails to appreciate the aspect of time on treatment course, the interaction between depressive symptoms level or substance use variables over time, and more complex aspects of treatment course (e.g., reasons for termination). Baseline measures of depressive symptoms level and substance use often represent the previous two weeks of symptoms and may not track very closely with subsequent changes more proximal to later treatment admissions. Treatment length should be considered in the context of discharge reason: Discharge after successfully completing treatment may be underpinned by different processes than discharge due to premature termination. Collapsing across successful and premature termination masks important differences and prevents examination of unique underlying mechanisms.
The current study examined discharge and readmission over time in a community-based SUD clinic. We evaluate three individual factors that are commonly assessed in community-based clinics and readily available for treatment planning without additional burden: depressive symptoms level at each admission, primary SUD at each admission, and lifetime misuse of multiple substances. The current study makes several unique contributions rarely seen in previous literature. First, we examined both primary SUD and lifetime misuse of multiple substances in the same model, controlling for the effect of each to the other, in order to determine which may be more important to consider in treatment planning. Second, we assessed the interactions between depressive symptoms level and substance use variables, controlling for demographic variables previously found to predict these SUD treatment outcomes [8]. Given the low variance of other SUD diagnoses outside AUD and OUD in our sample, we focused on these two for our hypotheses. Treatment for OUD is significantly different in nature than that for AUD and this may be one mechanism driving premature treatment termination among those with OUD and subsequent readmission [30]. Third, we examined SUD treatment outcomes using survival analysis, which enables assessing the instantaneous rate of treatment completion and readmission and adjusts for the varying follow-up lengths across patients. Fourth, we analyzed treatment discharge using advanced completing risks analysis, which jointly estimates the probability of discharge type of interest (i.e., premature termination) and time to discharge from any cause. Finally, we analyzed all observed admission and treatment events over time.
We hypothesized:
- Higher depressive symptoms level at admission, OUD (vs AUD) at admission, and history of misuse of multiple substances (vs single substance misuse) would predict increased hazard of premature termination.
- Higher depressive symptoms level at admission, OUD (vs AUD) at admission, and history of misuse of multiple substances (vs single substance misuse) would predict increased hazard of treatment readmission.
- Depressive symptoms level at admission would interact with the two substance use factors to predict treatment course, such that for those with OUD (vs AUD) at admission and history of misuse of multiple substances (vs single substance misuse), the association between depression level and premature termination and readmission would be stronger.
Method
Participants
De-identified longitudinal data were extracted from the charts of patients admitted to an SUD treatment center at least once between October 2014 and April 2018. This treatment center is one of the state’s most comprehensive systems for treating substance use and mental health disorders offering inpatient (detox, rehab, and residential treatment), outpatient (including intensive outpatient, partial hospitalization programs), long-term residential programs, and recovery housing. The final sample included charts from 542 patients (51.3% male, 88.2% Caucasian) between the ages of 15 and 72 (Mean = 36.5, standard deviation [SD] = 13.9) at their first intake to the clinic. Given the small percentage of patients with non-AUD or OUD diagnoses, these were collapsed into one category for analysis (i.e., ‘Other’).
Measures
Demographics.
Upon intake to the clinic, data were collected about patients’ sex assigned at birth, age, reported substances misused during one’s lifetime, and SUD diagnosis. SUD variables were assessed via a structured clinical interview performed by licensed mental health professionals.
Treatment completion and readmission.
Although included charts indicated at least one admission between October 2014 and April 2018, data extracted included all prior treatment admissions for each individual. Data included dates of intakes and discharges, program(s) of treatment attended, SUD diagnosis at each admission, and reasons for discharge for each treatment admission to the clinic. Reasons for discharge were collapsed into two categories: successful completion and premature termination (including leaving against medical advice, being expelled, and medical reasons). After data compilation, we calculated the length of each treatment admission and the length of time to the subsequent treatment admissions for each treatment admission and discharge. Length of treatment was defined as the time period, in number of days, that patients remained continuously enrolled in the treatment facility receiving a sequence of inpatient and/or outpatient treatment programs for a given primary diagnosis. Length of time to the subsequent treatment readmission was defined as the time period, in number of days, from previous discharge to subsequent treatment readmission.
Patient Health Questionnaire—9 (PHQ-9).
The PHQ-9 was used to measure depression level over the previous two weeks at each admission to the clinic before initiating treatment. The PHQ-9 is a self-report measure consisting of 9 items based on the DSM–IV diagnostic criteria for Major Depressive Disorder. Items are on a four-point Likert scale (0 = not at all, 1 = several days, 2 = more than half the days, 3 = every day). Depression level is calculated by summing up item scores (range 0–27), with higher scores indicating higher depression level. Scores of 0–4 indicate no depression symptoms, 5–9 mild depression, 10–14 moderate depression, 15–19 moderately severe depression, and 20–27 severe depression. Validity of the PHQ-9 has been established using both community [31] and clinical samples [32].
Procedures
During one admission between October 2014 and April 2018, all patients were offered the opportunity to allow their data to be utilized in a separate, larger study. Patients were told that their refusal to participate would not influence their treatment at the center. Adults provided consent and patients under 18 years old at admission provided assent, with consent provided by a parent/guardian. At each intake admission, patients completed the PHQ-9 as part of regular clinical care, which was also recorded in their charts. The data for the current analyses were compiled from these charts and were de-identified before being shared with the research team; thus, the project was deemed exempt from institutional board review by the local university. No previous papers have been published from these data.
Data analysis plan
All analyses were conducted using the R statistical software. To capture the treatment course as a function of time we applied survival and competing risks analyses using the survival package. These methods were used to predict the two time-to-event outcomes of interest: premature treatment termination and treatment readmission.
Competing risks analysis for premature treatment termination.
To estimate the effect of the time-dependent depression level (i.e., depression reported at each admission intake), primary SUD diagnosis (OUD vs AUD vs Other) at each admission, and lifetime misuse of multiple substances (vs single substance misuse), as well as the interactions between these variables, we implemented a competing risks survival analysis. The outcome in this analysis was the instantaneous rate (i.e., hazard) of premature termination. This rate reflects both the probability of the discharge type of interest (i.e., premature termination) and the time to discharge. Thus, this analysis incorporates both the likelihood of the discharge reason and the time to discharge, of each treatment admission for each individual, into one analysis. We also incorporated repeated measurements of depression level and primary SUD diagnosis at each treatment admission as predictor variables, allowing for the analysis of richer within-individual observations. The goal of our competing risks analysis was to estimate predictor and interaction effects on the instantaneous rate of transitioning from the initial “in treatment” state to the “premature termination” state, which was the main event of interest. A hazard ratio (HR)>1 for “premature termination” indicates that increase on a continuous predictor (e.g., PHQ-9) or a certain group in a categorical variable over the reference group (e.g., multiple vs single misuse) is associated with an increased rate (or hazard) of premature termination.
First, we estimated the cumulative incidence of premature termination (i.e., cumulative probability of premature termination over time) using a moment-based nonparametric estimator that takes into account the potential association across different treatment episodes of the same individual [33]. We created cumulative incidence plots to display the probability of premature termination after admission across depression and substance use subgroups over time. Two-sample nonparametric comparisons of the cumulative incidences were conducted using a Kolmogorov-Smirnov-type test [34]. To take the within-individual dependence into account, significance level was calculated using 1000 cluster bootstrap replications [33]. For the pairwise comparison of the cumulative incidences across the SUD groups (primary AUD vs primary OUD vs primary Other), we applied the Bonferroni correction. The multivariable analysis was based on the semiparametric proportional cause-specific hazards model. To take the within-individual dependence into account, standard errors were estimated using a robust sandwich variance estimator. All analyses were weighted using the inverse of the number of episodes for a given individual (Bakoyannis, 2021) to adjust for potential informative cluster size (e.g., that individuals with more episodes might be more prone to a premature discharge due to treatment burden).We tested potential nonlinearity of the depression effect by including a quadratic term of depression in the model. This allows us to model U or inverse U type associations, which are common in psychological phenomena, but rarely assessed. In this analysis, we first performed a global Wald test for the null hypotheses that no covariates are associated with premature treatment termination, which is not prone to Type I error inflation due to multiple comparisons. Hypothesis tests for individual covariate effects were interpreted only if this global test was statistically significant.
Survival analysis for treatment readmission.
To estimate the effect of the time-dependent depression level, primary SUD diagnosis at each admission, and lifetime misuse of multiple substances (vs single substance misuse), as well as their interactions, we conducted a survival analysis using the semiparametric Cox proportional hazards model. The goal is to estimate the rate of transitioning from the initial “treatment discharge” state to “treatment readmission” state. The outcome in this analysis was the instantaneous rate of treatment readmission (i.e., hazard of treatment readmission). A HR>1 indicates that increase on a continuous predictor (e.g., PHQ-9) or a certain group in a categorical variable over the reference group (e.g., AUD vs OUD) is associated with increased rate of treatment readmission.
First, we estimated the cumulative incidence of readmission using a moment-based nonparametric estimator that takes into account the potential association across different treatment episodes of the same individual [33]. We created cumulative probability plots to display the probability of readmission after a discharge across depression and substance use subgroups over time. To account for repeated readmissions within each patient, we used the gap time formulation of the conditional risk set Cox model. We evaluated potential nonlinearity of the depression effect by including a quadratic term of depression in the model. Similar to the analysis of premature treatment termination, we first performed a global Wald test for the null hypotheses that no covariates are associated with treatment readmission. Hypothesis tests for individual covariate effects were interpreted only if this global test was statistically significant.
Results
Descriptive statistics
The sample had a mean PHQ-9 of 14.13 (SD = 6.19) at admission, indicative of moderate depressive symptoms, and most commonly presented with AUD (42.25%) or OUD (38.56) (see Table 1 for full sample characteristics). Being admitted with a primary OUD was associated with increased cumulative probability of premature termination over time compared to being admitted with primary AUD (p = 0.002) (see Fig 1).
Competing risks analysis for premature treatment termination
The global Wald test for the null hypothesis that no covariates are associated with premature treatment termination was statistically significant (p-value<0.001). The interaction between time-dependent depressive symptoms level and primary SUD did not significantly predict the hazard of premature treatment termination, and thus was removed from the final analysis to aid interpretation of the results and for model parsimony (Table 2). Time-dependent depressive symptoms level did not significantly predict instantaneous rate of premature treatment termination (HR = 1.00, p = 0.814). Primary SUD diagnosis at admission significantly predicted the rate of premature treatment termination; primary OUD predicted a higher hazard of premature termination compared to primary AUD (HR = 1.91, p<0.001). No other covariates significantly predicted premature treatment termination.
Survival analysis for treatment readmission
The global Wald test for the null hypothesis that no covariates are associated with treatment readmission was statistically significant (p-value<0.001). The interaction between depressive symptoms level and lifetime substance misuse (multiple vs. single) significantly predicted treatment readmission (p = 0.024; Table 2); for those with single substance lifetime misuse, depressive symptoms level fell just short of significantly predicting the hazard of treatment readmission (HR = 1.03, p = 0.067), whereas the effect of depressive symptoms level did not significantly predict treatment readmission among those with lifetime misuse of multiple substances (HR = 0.98, p = 0.169). Age was positively associated with rate of readmission (HR = 1.01, p = 0.019).
Discussion
The goal of this longitudinal study was to examine time-dependent depressive symptoms level at each admission, primary SUD at each admission, lifetime misuse of multiple substances, and their interactions as risk factors of treatment course patterns over time in a community-based SUD clinic. Consistent with our hypothesis, primary OUD diagnosis at admission significantly predicted premature termination compared to primary AUD diagnosis and this effect was large. Our work replicates and extends previous research documenting this effect [24–26] using a sophisticated longitudinal analysis, which offers stringent statistical control, and a community-based clinic sample, which offers ecological validity. Taken together, this work increases confidence in the effect of OUD diagnosis on premature termination. Premature termination may be a sign of low engagement in intensive or step-down outpatient programs [30], or may be driven by the demanding nature of OUD treatments, which require a high level of patient commitment, including regular monitoring by multiple treatment professionals. Not remaining engaged in ongoing outpatient or supportive care is a risk factor for SUD relapse [35] and future work should seek to better understand and prevent what is driving this increased premature termination risk among those with OUD.
Contrary to our hypotheses, depressive symptoms level, history of misuse of multiple substances, and their interactions did not significantly predict premature termination. This is inconsistent with several previous studies [15, 22, 23]; however, this previous work used less diverse samples, included patients with only one SUD diagnosis, and/or only used baseline measures of depression without accounting for time-dependent depressive symptoms level over time. This result is consistent with previous studies that show no relationship between depressive symptoms or depression and treatment course across a variety of study designs (e.g., studies with retrospective designs based on electronic health records to longitudinal trials), samples (e.g., local clinics or community samples of patients diagnosed with a variety of SUDs or a specific SUD diagnosis), sample sizes (ranging between 70 and 557 patients), depression measures (e.g., categorical clinical diagnosis or symptom-based scores), and treatment settings (e.g., more specialized inpatient, detoxification, residential SUD settings, and outpatient clinics) [17–21]. However, based on our study results, it is impossible to determine whether the null effect of depressive symptoms level and its interaction with history of misuse of multiple substances on SUD treatment course reflects a lack of a relationship or is a result of type II error.
Consistent with our hypotheses and previous work [17], history of misuse of multiple substances predicted readmission to treatment, indicating that more complex substance misuse histories impart risk for more frequent and faster treatment readmission. Also, as expected, there was a significant interaction between depressive symptoms level and lifetime history of substance misuse for treatment readmission, but the direction of this interaction did not support our hypothesis. Depressive symptoms level trended towards positively predicting the rate of treatment readmission, but only among those with lifetime misuse of a single substance. This effect was small but potentially of theoretical interest that needs to be further understood. Mood disorders are strong predictors of substance use symptom reoccurrence [2], which is expected to translate to faster subsequent treatment readmission. This finding may reflect that the relationship between depressive symptoms level and subsequent readmission differs as a function of substance misuse complexity. Although depressive symptoms level showed a trend of faster engagement with treatment among those with simpler misuse histories, it appears unrelated to treatment access for people with a more complex misuse history. This might reflect depressive symptoms level as a more primary driver of treatment readmission for those with a simpler substance misuse history and/or that depressive symptoms level may play less of a role (e.g., complex misuse may be a stronger contributor of readmission which may outweigh the effect of depressive symptoms), or its role may be masked (e.g., those with high depressive symptoms level and complex substance misuse do not reach care fast enough after relapse), in those with more complex misuse histories. This interaction may, in part, explain inconsistencies in previous research concerning the role of depressive symptoms on treatment course [15, 17–23].
We propose understanding the mechanisms driving this interaction effect as a prime next step in this program of research, since more complex substance misuse patterns increase likelihood for unintentional or intentional overdose fatalities, the prevalence of which are greater in the context of depressive symptoms and suicidality [2]. This finding points to an important disparity in SUD treatment utilization in a high-risk and vulnerable group, especially since depressive symptoms level, which is often a warning sign for relapse risk [10] may fail to signal risk among this higher risk group. There may be a need for concerted efforts among health care professionals and other supports to closely monitor patients with more complicated substance misuse histories for changes in mood and signs of substance use reoccurrence. This is critical given that untreated depressive symptoms can result in further problems, including increasing the potential for suicide [10] before reaching clinical care. Interestingly, age predicted treatment readmission, which has been inconsistently found in previous work [17]; thus, more research is needed to understand this effect.
This study had several limitations. Most of the patients were Caucasian with sufficient socioeconomic resources to either pay for or have insurance for treatment, which may limit generalizability. Due to low racial diversity, we were not able study racial differences. Patients received a variety of different treatment modalities, which may increase generalizability of findings, but could potentially confound results. We only had data from one treatment facility and could not ascertain whether patients were admitted to other clinics during the follow-up period. Rates of lifetime history of misuse of multiple substances were slightly lower (at 38%) than may be expected by previous estimates (e.g., 55–64% [36, 37]), which may reflect an underreporting self-report bias. Although the PHQ-9 is a well-supported measure of depressive symptoms [31, 32], other measures may contribute complementary information to better understand effects on treatment course. Depressive symptoms level was measured at the beginning of treatment; including measurements at the end of treatment would have helped us understand depressive symptom changes secondary to treatment and would provide a more proximal measure to characterize if unresolved depressive symptoms contributed to premature termination and/or readmission.
Overall, this work extends upon previous findings of this nature, by providing evidence utilizing a rigorous time-dependent, recurrent event and competing risks survival analysis approach and a real-world clinic-based sample, controlling for covariates that may have confounded this past literature. Thus, the impact of this study is that it provides a more stringent test and support for the finding that OUD diagnosis places one at risk for premature treatment termination, which is particularly important, as those with OUD who terminate prematurely are at higher risk for substance use relapse, which is associated with higher risk of overdose and death [2]. Since primary diagnosis is readily available in clinics, this is a low-cost/high-reward strategy without additional staff burden. Lifetime misuse of multiples substances, age, and premature termination in the previous treatment admission may be important to predict return to treatment; these variables can be used to focus close monitoring on patients who may most benefit from such efforts. This work also provides one explanation for the mixed results concerning how depressive symptoms level influences SUD treatment course and patterns by identifying substance misuse history complexity as a moderator of this relationship. Depressive symptoms level, which is easily assessed and tracked over time, may be an important early warning sign for those who have a history of only misusing one substance that returning to treatment may be warranted or necessary, but may be a less useful warning sign for those with more complex misuse histories. The current study utilized a real-world clinic-based sample, providing complementary findings to better controlled, but less ecologically valid, clinical trial approaches. We believe that assessing and testing these complex relationships in real-world clinic samples has potential to directly influence real-world clinics and treatment settings. With advancements in statistical modeling, we propose that this study can serve as an example of how future research should continue to utilize more complex models to attempt to replicate previously documented effects, with the long-term goal of better characterizing risks, designing and planning treatment, and supporting SUD patients and their families.
Acknowledgments
We wish to thank the Fairbanks Addiction Treatment Center for collecting the data, collaborating with us, and providing their valuable feedback.
References
- 1. Cunningham JA, McCambridge J. Is alcohol dependence best viewed as a chronic relapsing disorder? Addiction (Abingdon, England). 2012;107(1):6–12. Epub 2011/10/07. pmid:21981681.
- 2. Volkow ND, Jones EB, Einstein EB, Wargo EM. Prevention and Treatment of Opioid Misuse and Addiction: A Review. JAMA Psychiatry. 2019;76(2):208–16. Epub 2018/12/06. pmid:30516809.
- 3. Reif S, Acevedo A, Garnick DW, Fullerton CA. Reducing Behavioral Health Inpatient Readmissions for People With Substance Use Disorders: Do Follow-Up Services Matter? Psychiatric Services. 2017;68(8):810–8. pmid:28412900
- 4. Brewer DD, Catalano RF, Haggerty K, Gainey RR, Fleming CB. RESEARCH REPORT A meta-analysis of predictors of continued drug use during and after treatment for opiate addiction. Addiction. 1998;93(1):73–92.
- 5. Moos RH, Pettit B, Gruber V. Longer episodes of community residential care reduce substance abuse patients’ readmission rates. Journal of Studies on Alcohol. 1995;56(4):433–43. pmid:7674679
- 6. Carrier E, McNeely J, Lobach I, Tay S, Gourevitch MN, Raven MC. Factors Associated with Frequent Utilization of Crisis Substance Use Detoxification Services. Journal of Addictive Diseases. 2011;30(2):116–22. pmid:21491293
- 7. Callaghan RC, Cunningham JA. Gender differences in detoxification: predictors of completion and re-admission. Journal of Substance Abuse Treatment. 2002;23(4):399–407. pmid:12495802
- 8. Brorson HH, Ajo Arnevik E, Rand-Hendriksen K, Duckert F. Drop-out from addiction treatment: A systematic review of risk factors. Clinical Psychology Review. 2013;33(8):1010–24. pmid:24029221
- 9. Pettinati HM, O’Brien CP, Dundon WD. Current status of co-occurring mood and substance use disorders: a new therapeutic target. Am J Psychiatry. 2013;170(1):23–30. Epub 2012/12/12. pmid:23223834; PubMed Central PMCID: PMC3595612.
- 10. Hasin D, Liu X, Nunes E, McCloud S, Samet S, Endicott J. Effects of major depression on remission and relapse of substance dependence. Arch Gen Psychiatry. 2002;59(4):375–80. Epub 2002/04/03. pmid:11926938.
- 11. Shanahan CW, Lincoln A, Horton NJ, Saitz R, Winter M, Samet JH. Relationship of depressive symptoms and mental health functioning to repeat detoxification. J Subst Abuse Treat. 2005;29(2):117–23. Epub 2005/09/02. pmid:16135340.
- 12. Yedlapati SH, Stewart SH. Predictors of Alcohol Withdrawal Readmissions. Alcohol and Alcoholism. 2018;53(4):448–52. pmid:29617711
- 13. Lubman DI, Allen NB, Rogers N, Cementon E, Bonomo Y. The impact of co-occurring mood and anxiety disorders among substance-abusing youth. J Affect Disord. 2007;103(1–3):105–12. Epub 2007/02/13. pmid:17291589.
- 14. Tate SR, Mrnak-Meyer J, Shriver CL, Atkinson JH, Robinson SK, Brown SA. Predictors of treatment retention for substance-dependent adults with co-occurring depression. Am J Addict. 2011;20(4):357–65. Epub 20110531. pmid:21679267; PubMed Central PMCID: PMC3821388.
- 15. Brown RA, Monti PM, Myers MG, Martin RA, Rivinus T, Dubreuil ME, et al. Depression Among Cocaine Abusers in Treatment: Relation to Cocaine and Alcohol Use and Treatment Outcome. American Journal of Psychiatry. 1998;155(2):220–5. pmid:9464201
- 16. Field CA, Duncan J, Washington K, Adinoff B. Association of baseline characteristics and motivation to change among patients seeking treatment for substance dependence. Drug Alcohol Depend. 2007;91(1):77–84. Epub 2007/07/04. pmid:17606335.
- 17. Böckmann V, Lay B, Seifritz E, Kawohl W, Roser P, Habermeyer B. Patient-Level Predictors of Psychiatric Readmission in Substance Use Disorders. Front Psychiatry. 2019;10. pmid:32038313
- 18. Daughters SB, Richards JM, Gorka SM, Sinha R. HPA axis response to psychological stress and treatment retention in residential substance abuse treatment: A prospective study. Drug and Alcohol Dependence. 2009;105(3):202–8. pmid:19713052
- 19. Davis A, McMaster P, Christie DC, Yang A, Kruk JS, Fisher KA. Psychiatric Comorbidities of Substance Use Disorders: Does Dual Diagnosis Predict Inpatient Detoxification Treatment Outcomes? International Journal of Mental Health and Addiction. 2022.
- 20. Marrero CA, Robles RR, Colón HM, Reyes JC, Matos TD, Sahai H, et al. Factors associated with drug treatment dropout among injection drug users in Puerto Rico. Addictive Behaviors. 2005;30(2):397–402. pmid:15621413
- 21. Santonja-Gómez F-J, Sánchez-Hervás E, Secades-Villa R, Zacarés-Romaguera F, García-Rodríguez O, García-Fernández G. Pretreatment Characteristics as Predictors of Retention in Cocaine-dependent Outpatients. Addictive Disorders & Their Treatment. 2010;9(2).
- 22. Alterman AI, McKay JR, Mulvaney FD, McLellan AT. Prediction of attrition from day hospital treatment in lower socioeconomic cocaine-dependent men. Drug Alcohol Depend. 1996;40(3):227–33. Epub 1996/03/01. pmid:8861401.
- 23. Curran GM, Kirchner JE, Worley M, Rookey C, Booth BM. Depressive symptomatology and early attrition from intensive outpatient substance use treatment. J Behav Health Serv Res. 2002;29(2):138–43. Epub 2002/05/30. pmid:12032971.
- 24. Callaghan RC. Risk factors associated with dropout and readmission among First Nations individuals admitted to an inpatient alcohol and drug detoxification program. Cmaj. 2003;169(1):23–7. pmid:12847035
- 25. Li X, Sun H, Puri A, Marsh DC, Anis AH. Factors Associated with Pretreatment and Treatment Dropouts Among Clients Admitted to Medical Withdrawal Management. Journal of Addictive Diseases. 2007;26(3):77–85. pmid:18018810
- 26. Gilchrist G, Langohr K, Fonseca F, Muga R, Torrens M. Factors associated with discharge against medical advice from an alcohol and drug inpatient detoxifcation unit in Barcelona between 1993 and 2006. Heroin Addiction and Related Clinical Problems. 2012;14(1):35–44.
- 27. Kandel DB, Huang FY, Davies M. Comorbidity between patterns of substance use dependence and psychiatric syndromes. Drug Alcohol Depend. 2001;64(2):233–41. Epub 2001/09/07. pmid:11543993.
- 28. Lai HM, Cleary M, Sitharthan T, Hunt GE. Prevalence of comorbid substance use, anxiety and mood disorders in epidemiological surveys, 1990–2014: A systematic review and meta-analysis. Drug Alcohol Depend. 2015;154:1–13. Epub 2015/06/15. pmid:26072219.
- 29. Chen T, Zhong N, Du J, Li Z, Zhao Y, Sun H, et al. Polydrug use patterns and their impact on relapse among heroin-dependent patients in Shanghai, China. Addiction. 2019;114(2):259–67. Epub 2018/10/03. pmid:30276902.
- 30. Timko C, Below M, Schultz NR, Brief D, Cucciare MA. Patient and program factors that bridge the detoxification-treatment gap: a structured evidence review. J Subst Abuse Treat. 2015;52:31–9. Epub 2014/12/23. pmid:25530425.
- 31. Martin A, Rief W, Klaiberg A, Braehler E. Validity of the Brief Patient Health Questionnaire Mood Scale (PHQ-9) in the general population. Gen Hosp Psychiatry. 2006;28(1):71–7. Epub 2005/12/27. pmid:16377369.
- 32. Dum M, Pickren J, Sobell LC, Sobell MB. Comparing the BDI-II and the PHQ-9 with outpatient substance abusers. Addict Behav. 2008;33(2):381–7. Epub 2007/10/30. pmid:17964079.
- 33. Bakoyannis G. Nonparametric analysis of nonhomogeneous multistate processes with clustered observations. Biometrics. 2021;77(2):533–46. pmid:32640037
- 34. Bakoyannis G, Bandyopadhyay D. Nonparametric tests for multistate processes with clustered data. Annals of the Institute of Statistical Mathematics. 2022. pmid:36090245
- 35. Blodgett JC, Maisel NC, Fuh IL, Wilbourne PL, Finney JW. How effective is continuing care for substance use disorders? A meta-analytic review. Journal of substance abuse treatment. 2014;46(2):87–97. Epub 2013/09/26. pmid:24075796.
- 36. Hjemsæter AJ, Bramness JG, Drake R, Skeie I, Monsbakken B, Benth JŠ, et al. Mortality, cause of death and risk factors in patients with alcohol use disorder alone or poly-substance use disorders: a 19-year prospective cohort study. BMC Psychiatry. 2019;19(1):101. pmid:30922325
- 37. Staines GL, Magura S, Foote J, Deluca A, Kosanke N. Polysubstance Use Among Alcoholics. Journal of Addictive Diseases. 2001;20(4):57–73. pmid:11760926