Task development, living conditions and time to drop-out in patients of 18-29 years of age, under treatment for substance abuse disorders in Chile, 2010-2019

A. González-Santa Cruz1, 2, , J. Castellano Soto1, 3, 4,


1 Young Researcher, Millennium Nucleus for the evaluation and analysis of Drug Policies
2 Ph.D. student, School of Public Health, Universidad de Chile
3 HEOR-Consultant IQVIA, Chile
4 Universidad de los Andes, Chile

Background

Emerging adults (aged 18-29) are considered youngsters in Chile, with more risk factors and poorer treatment outcomes. For these patients, whether living With their family of origin exclusively, Alone, or With their couple and/or children, can be a proxy of what Erikson and Valliant called task development, where Identity phase implies living independently of family of origin, Intimacy means being capable of maintaining a relationship with a partner, and Generativity, raising children \(^{[1; 2]}\). Substance use treatment (SUT) outcomes may be associated with living conditions at admission. To our knowledge, there are no similar studies in Chile.

Objectives

To estimate the association between living conditions at admission to first SUT (Alone, with family of origin/parents and with couple & children) and the time to dropout from admission, of emerging adulthood patients (aged 18-29), admitted to public substance use treatments between 2010-2019 in Chile.

Methods

This research relies on a population-based record-linkage retrospective cohort of adults in publicly funded Chilean SUTs programs nationally between 2010-2019 (n= 23,979). Patients were weighted by the inverse probability of living conditions at admission (IPW) and, based on several predictors, calculated weight-adjusted restricted mean survival times (RMST) of avoiding dropouts (administrative discharges or without clinical advice) after three years of baseline admission \(^{[3]}\). RMSTs can be useful for non-proportional hazards \(^{[4]}\). We also did a log-rank test with IPWs as sampling weights. The following covariates were used to estimate IPWs:

  • Sex (sexo_2)= Indicated by patients at admission. Can adopt the following values: “woman” or “man”.

  • Educational Attainment (escolaridad_rec)= Categories were “Completed primary school or lower”, “Completed high school or lower”, and “More than high school”.

  • Starting substance (sus_ini_mod_mvv)= Recorded by the therapeutic team and indicated by patients as the drug use onset substance. We grouped them to get the main substances: Cocaine Paste Base, Cocaine (snort) hydrochloride, Marijuana, Alcohol, Others (less than 5%).

  • Consumption frequency of primary or main substance (freq_cons_sus_prin)= Categories considered: “Less than 1 day per week”, “1 day per week”, “2-3 days per week”, “4-6 days per week”, and “Daily”. Among the categories of the Frequency of Drug Consumption, we grouped Less than 1 day a week (3.16%) with Did not use in the last 30 days (1.80%)

  • Psychiatric comorbidity ICD-10 (with amount of different diagnosis) (cnt_mod_cie_10_or)= According to the ICD-10 (International Classification of Diseases 10 th Edition) classification. The system allowed for submitting up to three psychiatric conditions. We counted the amount of diagnoses patients had.

  • Admission age (edad_al_ing)= Patient age when admitted to treatment.

  • Substance Use Severity (dg_trs_cons_sus_or)= Secondary substance use disorder according to the ICD-10 classification. Available values were “Drug dependence” or “Hazardous consumption”.

  • Primary substance (sus_principal_mod)= The therapeutic team records the substance identified by patients as causing problems at admission or the main substance of concern. We grouped them to get the main substances: Cocaine Paste Base, Cocaine (snort) hydrochloride, Marijuana, Alcohol, Others (less than 5%).

Codes are available at bit.ly/3Ir6jsT. This research was approved by the Human Research Ethics Committee of the Universidad Mayor.

Preliminary Results

Of the 109,756 records of treatments (patients= 85,048) SENDA records of admissions, we selected cases of emerging adults (patients= 23,979).

Table 1: Glimpse of the survival analysis, Years to drop-out in the First Treatment by Living Conditions
rn2 tstop event rate lower upper
Alone 2182.98 926 0.42 0.40 0.45
Family of origin 27734.74 10209 0.37 0.36 0.38
With couple/children 11721.96 5076 0.43 0.42 0.45

We tested inverse probability weights that account for a multinomial response using Propensity score weighting using a multinomial Logistic, not stabilized, stabilized at 1% (smoothing 1% of outliers), 5% and 10%. We also tested models using Bayesian additive regression trees (BART), multinomial regression, and Generalized Boosted Regression Modeling with multinomial response on the average effect on the treated or ATE, weights with multiple treatments and weights to simultaneously compare three treatment groups.

Love plot

Figure 1: Love plot

The best model was the not stabilized multinomial Logistic model. Despite the groups With couple/children and Alone suffered the second greater lost in terms of effective sample sizes (ESS) (from 7189 to 5970 and from 1335 to 1091, respectively). Also BART model show the second lowest standardized mean differences between groups. If BART model were selected, the group With couple/children and Alone suffered the greater lost in terms of effective sample sizes (from 7189 to 5757 and from 1335 to 1086, respectively). We estimated survival measures using both weights but here we show the results of the not stabilized multinomial Logistic model.

Adjusted Survival

Kaplan meier (survey weighted, multinomial logistic weights)

Figure 2: Kaplan meier (survey weighted, multinomial logistic weights)

Patients living With couple/children had lower probabilities of avoiding dropouts (Q1= 0.29; Mdn= 0.26) than patients living alone (Q1= 0.32; Mdn= 0.31) and With their family of origin (Q1= 0.5; Mdn= 0.3) combined (Log-rank Chi Square 41.45 , p= 0).


Adjusted RMSTs

Table 2: Restricted Mean Survival Time (in years)
Time Cat1 RMST Cat2 RMST diff ratio
Six months Family of origin 0.39 Alone 0.36 0.03 (0.02, 0.04) 1.07 (1.04, 1.1)
Six months With couple/children 0.38 Alone 0.36 0.01 (0, 0.03) 1.04 (1.01, 1.07)
Six months With couple/children 0.38 Family of origin 0.39 -0.01 (-0.01, -0.01) 0.97 (0.96, 0.98)
One year Family of origin 0.60 Alone 0.55 0.05 (0.03, 0.07) 1.09 (1.05, 1.14)
One year With couple/children 0.57 Alone 0.55 0.02 (-0.01, 0.04) 1.03 (0.99, 1.07)
One year With couple/children 0.57 Family of origin 0.60 -0.03 (-0.04, -0.02) 0.94 (0.93, 0.96)
Three years Family of origin 1.24 Alone 1.13 0.11 (0.03, 0.19) 1.1 (1.02, 1.17)
Three years With couple/children 1.11 Alone 1.13 -0.02 (-0.1, 0.06) 0.98 (0.91, 1.05)
Three years With couple/children 1.11 Family of origin 1.24 -0.13 (-0.17, -0.1) 0.89 (0.86, 0.92)

  • Who lives with his/her couple and children had lower mean time without experiencing dropout vs. Family of origin since 6 months, meaning that experience it more quickly. Differences between Groups With couple/children vs. Alone (less time avoiding dropout for Family of origin). However this difference was only observed at 6 months.

  • Those that were with couple/children had less average time avoiding readmission, followed by those Alone.

Discussion

  • The results of this study did not support the proposed hypotheses. Being with one’s family of origin generally having a protective effect compared to being alone. However, the effect of being with a partner/children varies depending on the time point and may not always have a significant impact compared to being alone.

  • Further research is needed to account for competing risks (e.g., treatment completion) of the association. Additionally, studies related to mental health and substance use disorders to develop cost-effective strategies to support current and projected demand for healthcare services characteristics of patients living alone, their health status and what is currently known about their substance use and misuse.

References

[1] G. E. Vaillant. Triumphs of Experience. Harvard University Press, oct.. 2012. DOI: 10.2307/j.ctt2jbxs1. URL: https://doi.org/10.2307/j.ctt2jbxs1.

[2] L. Friedman. Identity’s Architect: A Biography of Erik H. Erikson. Scribner, 1999. ISBN: 9780684195254. URL: https://books.google.cl/books?id=t_QMAQAAMAAJ.

[3] S. C. Conner, L. M. Sullivan, E. J. Benjamin, et al. “Adjusted restricted mean survival times in observational studies”. In: Statistics in Medicine 38.20 (2019), pp. 3832-3860. DOI: https://doi.org/10.1002/sim.8206. eprint: https://onlinelibrary.wiley.com/doi/pdf/10.1002/sim.8206. URL: https://onlinelibrary.wiley.com/doi/abs/10.1002/sim.8206.

[4] M. A. Hernán. “The Hazards of Hazard Ratios”. In: Epidemiology 21.1 (ene.. 2010), pp. 13-15. DOI: 10.1097/ede.0b013e3181c1ea43. URL: https://doi.org/10.1097/ede.0b013e3181c1ea43.

Funding sources

  • This study is funded by the National Agency for Research and Development (ANID-FONDECYT regular grant #1191282 to Dr. Álvaro Castillo-Carniglia) and the Millennium Nucleus for the Evaluation and Analysis of Drug Policies (nDP), Chile to Dr. Álvaro Castillo-Carniglia

  • The author have no conflict of interest to declare

Correspondence to: Andrés González-Santa Cruz,