Excess mortality following discharge from substance use disorder treatment in Chile

Abstract

Background and aims

People with substance‐use disorders (SUDs) have increased mortality risk, yet Chilean estimates of SUD‐based mortality are scarce. This study aimed to quantify all‐cause and cause‐specific mortality following SUD treatment in Chile compared with the general population and assess variation across key clinical and demographic subgroups.

Design

National‐level registry‐based retrospective data linkage cohort study.

Setting

Publicly funded SUD psychosocial treatments offered by the Chilean National Drug Agency, linked with official national mortality records from 2010 to 2020.

Participants

70064 adults aged 18–64 years (24% women, median age 35 at treatment entry) were followed after their first treatment episode until death or 31 December 2020.

Measurements

Primary outcome was all‐cause mortality. Secondary outcomes were cause‐specific mortality by the International Classification of Diseases, 10th revision, underlying and external causes. We estimated age–sex–calendar year directly standardized rates (DSR), and standardized mortality ratios (SMR) compared with the expected rate for the (sub)population. We also stratified rates and ratios by sex (men/women), attained age (18–29, 30–44, 45–59, 60+), setting (ambulatory/residential), primary substance (alcohol; illicit: predominantly cocaine paste base, marijuana and cocaine hydrochloride) and treatment compliance (not completed/completed). Additionally, we estimated rates and SMRs for underlying and external causes of mortality.

Findings

Over a median 4.9‐year follow‐up (353 826 person‐years), 2996 deaths occurred [DSR = 10.6, 95% confidence interval (CI) = 8.6–13.1]. Overall SMR was 3.65 (95% CI = 3.52–3.79). Excess risk was particularly pronounced for women (SMR = 5.57, 95% CI = 5.14–6.03), patients admitted due to alcohol use disorder (SMR = 4.59, 95% CI = 4.33–4.86), in residential care (SMR = 4.91, 95% CI = 4.45–5.42) and treatment noncompletion (SMR = 4.04, 95% CI = 3.85–4.24). Cause‐specific mortality revealed elevated external‐cause excess risk for SUD patients, including intentional self‐harm (SMR = 6.67, 95% CI = 6.05–7.36), unintentional injuries (SMR = 5.37, 95% CI = 4.79–6.02) and assaults (SMR = 4.98, 95% CI = 4.16–5.96). Notable excess risk was also observed for non‐external mortality causes: digestive system (SMR = 8.20, 95% CI = 7.62–8.83), symptoms and signs (SMR = 5.18, 95% CI = 4.29–6.26) and respiratory diseases (SMR = 5.18, 95% CI = 4.47–5.99) were greater than expected.

Conclusions

In Chile, patients with a history of publicly funded substance‐use disorder treatment appear to have an all‐cause mortality up to 3.7 times higher than the general population, driven predominantly by digestive and respiratory causes, as well as self‐harm, unintentional injuries and assaults.

INTRODUCTION

People with alcohol and substance use disorders (SUDs) have higher premature mortality than the general population [1, 2]. Globally, SUDs account for approximately 0.52% of deaths [3, 4] and are highly prevalent in Latin America [5]. Moreover, SUD mortality rose markedly in this region from 2000 to 2019, with Chile showing one of the largest increases (170%), highlighting the need to understand local epidemiological trends [6]. Yet, estimating directly attributable deaths is difficult for two main reasons. First, opioid and injected substance use is low compared to the Global North, therefore, deaths classified as poisoning or overdose are infrequent. Even in developed countries, the substances involved are uncertain [7]. Second, Chile and most Latin American countries report only underlying, not contributory, causes of death, limiting the ability to find drug‐related deaths in official statistics [8].

In response to these regional trends, the Chilean National Health Strategy has identified SUDs as a primary challenge for the decade. The burden is driven primarily by alcohol and cocaine (and their derivatives), which constitute the majority of publicly funded treatment episodes [8]. The Chilean government, primarily through the National Service for the Prevention and Rehabilitation of Drug and Alcohol Use (SENDA), funds and oversees a national network of residential and outpatient SUD treatment that covers approximately 80% of the publicly insured population. SENDA captures most specialized public admissions (~17 000/year) but excludes lower‐intensity Ministry of Health programs (~10 000/year) and private treatment (unknown) [8]. Consequently, this cohort represents a high‐complexity subgroup, because SENDA‐funded programs are designed for patients requiring specialized, higher‐intensity care [9]. However, treatment coverage remains low, reaching only 16% to 26% (if including Ministry of Health services) of the estimated population in need [8].

Describing mortality within cohorts with a history of SUD treatment is crucial. Many definitions of treatment highlight mortality as a critical consequence that interventions aim to minimize [10] and an essential patient‐centered outcome [11]. However, patients accessing public treatment often represent a population with higher SUD severity, comorbidities and complex social needs, inherently placing them at elevated mortality risk [12, 13]. Furthermore, factors such as the type of treatment received (e.g. residential vs. outpatient), the primary substance used, treatment completion or dropout impact subsequent mortality risk [2, 14, 15, 16, 17, 18, 19].

Despite available national treatments, there is a significant gap in the literature regarding mortality among SUD patients in Chile. Studies from high‐income countries provide valuable insights [17, 20], but are hardly generalizable to the Latin American context because of differences in epidemiology, health system structures, treatment approaches and socio‐cultural factors. Chilean larger in‐person psychosocial treatment programs focus mainly on alcohol and cocaine derivatives (specifically cocaine paste base); however, they do not offer differentiated, substance‐specific treatment protocols. Instead, individual clinical assessment partially considers it (e.g. pharmacological support) [8, 21].

It is concerning that evidence on SUDs remains limited in Chile and in most Latin American countries [22, 23, 24]. A related study from a cohort in a Brazilian city tracking psychiatric patients hospitalized between 2002 and 2007 and followed up to 14 years, found higher mortality than the general population among those hospitalized for SUDs [25]. In Chile, evidence on the mortality of patients with a history of SUD treatment nation‐wide remains non‐existent. However, Chile is uniquely positioned to address this gap because of its robust substance use treatment registration system (SISTRAT). SISTRAT became operative in 2010 and is one of the first national systems implemented in Latin America. In addition and despite its limitation on contributory causes, Chile also has a high‐quality national mortality registry managed by the Chilean National Department of Statistics and Health Information (DEIS) of the Ministry of Health, which has high completeness and few ill‐defined causes [8, 26, 27, 28]. Therefore, leveraging data linkage across these sources provides a powerful method for examining long‐term outcomes in this population [2].

This study aimed to quantify excess all‐cause and cause‐specific mortality among adults after publicly funded SUD treatment in Chile compared with the general population. The primary aim of this study was to quantify excess all‐cause mortality in adults following SUD treatment in Chile relative to the general population. Secondary aims were to compare mortality rates across key demographic and clinical subgroups and to identify major underlying and external causes of death driving the excess risk in the overall cohort.

METHOD

Design

This is a nation‐wide, registry‐based, retrospective cohort study. It included all adult patients age 18 to 64 years who initiated publicly funded SUD treatment in Chile between 2010 and 2020. This age range concentrates SUD‐attributable years of life lost and allows stable mortality estimates across older age subgroups [29].

SISTRAT records all services from government‐funded public and private SUD treatment centers, which are available for the population affiliated with public health insurance (~80% of the Chilean population) [30]. We cleaned the data by removing duplicates and records with aberrant admission ages, incorrect dates or missing types. Overlapping treatments were merged, prioritizing longer episodes. Records for patients who were not discharged or discharged after the censoring date (31 December 2020) were excluded (Figure S1 summarizes this process).

Mortality records included date, demographic information and International Classification of Diseases, 10th revision (ICD‐10) causes. Two mortality fields were available in this database: the ‘underlying cause’ as the disease or injury initiating the chain of events leading to death, where natural deaths were classified using this primary field; and ‘external causes’ (e.g. injury or poisoning mechanisms, codes V–Y) were identified using the dataset's external‐cause mechanism field [31]. Individual patient records from SISTRAT were linked to national mortality data. Records were linked using encrypted national identification numbers (RUNs), while the decryption key was withheld from the research team and held only by DEIS and SENDA. Follow‐up was from the first treatment termination date until death or the end of follow‐up, whichever occurred first. ‘First treatment’ refers to the first SUD‐related treatment presentation within our time frame, which may not be the patient's first overall. This study was approved by the University of Chile Human Research Ethics Committee (project number: 013–2025; approval ref. no: 2025/011). The analysis for this study was not pre‐registered, therefore, the results should be considered exploratory. The study adhered to the Reporting of Studies Conducted using Observational Routinely Collected Health Data (RECORD) reporting guideline [32].

Variables

SISTRAT records included patient birth date, age at treatment entry (range: 15–64 years), age at treatment discharge, sex (men/women) and year of treatment initiation. Additionally, SISTRAT included key variables used for stratification in the analyses: (1) treatment setting, categorized as residential (5–7 days a week, lasting up to 1 year) and ambulatory (2–3 and up to 5 weekly interventions, lasting 6 and up to 8 months) care; (2) primary substance of use at treatment entry, classified into licit (alcohol) or illicit (dominated by cocaine paste base, marijuana and cocaine hydrochloride); and (3) initial treatment outcome, categorized as completion (recorded as ‘therapeutic discharge’ on achieving the goals of the patient's individual intervention plan) or non‐completion (including patient‐initiated dropout, loss of contact or administrative discharge for disciplinary reasons, and recorded after loss of contact for 60 days or more or despite at least two outreach attempts to encourage them to resume treatment) [33].

The primary outcome was all‐cause mortality. Death dates were obtained from the DEIS national mortality registry. Causes of death were coded to ICD‐10 and grouped as underlying: infectious and parasitic (A00–B99), malignant neoplasms (C00–C97), endocrine and metabolic (E00–E90), mental and behavioral (F00–F99), nervous system (G00–G99), circulatory (I00–I99), respiratory (J00–J99), digestive (K00–K93), symptoms and signs (R00–R99) and other chapters (remaining D, H, L, M, N, O, P, Q, S, T, U). External causes were grouped into transport accidents (V01–V99), other unintentional causes of injury (W00–X59), intentional self‐harm (X60–X84) and assault/aggression (X85–Y09).

Statistical analysis

The main analysis was restricted to cases with known and valid admission and discharge information dates, resulting in a complete dataset for which complete‐case analysis was conducted.

Descriptive characteristics are summarized as medians, interquartile ranges (IQR) and percentages. To summarize the differences between survivors and patients who died during follow‐up, absolute standardized mean differences (|SMD|) were calculated. An SMD threshold of less than 0.15 was used to flag clinically meaningful imbalances [34].

We split follow‐up for each patient into person‐time segments (Lexis format). This structure enables dynamic classification of exposure time based on attained age (computed from birth date and follow‐up time, starting at age at discharge) and the corresponding calendar year, allowing individuals to transition across age groups as follow‐up progresses. We calculated crude mortality rates (CMR) for the cohort. Person‐years (PY) at risk were computed from the date of treatment termination until death or censoring. In 2024 (the most recent available data), the CMR in Chile was 6.3 per 1000 population [35].

To estimate standardized mortality rates, we calculated the expected number of deaths using age (attained during follow‐up) stratified in 5‐year age bins, sex and calendar year‐specific death rates for Chile, based on the 2017 CENSUS external reference population with official population projections. We used standardized mortality ratios (SMRs) to compare observed and expected deaths.

We calculated directly year‐, age‐ and sex‐standardized mortality rates (DSR) per 1000 PYs using the general population as the standard, enabling complementary comparisons with our cohort. Analyses were implemented using the popEpi package [36]. To examine potential variations, we calculated SMRs and DSRs for subgroups defined by attained age (18–29, 30–44, 45–59 and 60+ years) and sex. To account for potential dispersion for SMRs and DSRs, we scaled the standard errors by the square root of the dispersion parameter √ɸ in the case of overdispersion. We estimated ɸ using the quasi‐Poisson method, as the Pearson χ² statistic divided by residual degrees of freedom. This approach ensures the precision of rates reflects the observed variance [37, 38].

We stratified mortality rates by treatment setting, primary substance of use and initial treatment outcome. We quantified heterogeneity among SMRs and DSRs at different levels using Cochran's Q statistic [39]. Given that tests for heterogeneity frequently have low statistical power and may favor evidence supporting homogeneity, we adopted a significance level of P < 0.10 to better detect variation across subgroups [39, 40]. Additionally, we calculated SMRs and DSRs for each major underlying and external cause of death.

Sensitivity analyses

We adopted a last‐observation‐carried‐forward approach, defining cohort entry at each final recorded treatment episode to assess whether indexing on the last treatment alters the results, thereby assessing whether ignoring later episodes underestimates cumulative treatment exposure among individuals who cycle in and out of care [22]. Although our analysis focused on mortality following SUD treatment, we also replicated the main analysis with unfinished treatments [patients referred to treatments outside SENDA's network, currently in treatment or discharged after death (usually deceased while in active treatment and administratively discharged)].

All analyses were conducted in R version 4.4.1 (R Core Team). Following recommendations for reporting on complete populations, we included 95% CIs to reflect the stability of estimates and account for random fluctuations over time [41]. Programming code is available at https://tinyurl.com/mpeu8c3h.

RESULTS

Of the initial 150 046 treatment records, 4007 (2.7%) were excluded because of missing or invalid discharge information (primarily ongoing treatments). Following exclusions for age and study window, we removed records based on administrative ineligibility: 13886 (9.3%) referrals to treatment outside the SENDA network and 418 (0.3%) truncated episodes that were not administratively closed. Finally, 4406 (2.9%) patients were excluded because they did not contribute valid person‐time to the post‐discharge analysis: 3888 (2.6%) were discharged after the censoring date (31 December 2020), 368 (0.2%) died before treatment completion and 150 (0.1%) had inconsistent dates (negative follow‐up).

Between 2010 and 2020, 70 064 individuals in Chile were discharged (finished) from SUD treatment (with or without clinical advice). Most patients (72.6%) did not complete treatment. Over a median follow‐up of 4.9 years (range: 0.3–11.0; total PY = 353 826), 2996 patients (4.3%) died (Table 1). These patients reported alcohol as the primary substance more often (58% vs. 33%) and were older (5% age 60–65 years vs. 2%; 15% age 18–29 years vs. 34%), compared to survivors. Sex differences were modest, with slightly more women among survivors (24% vs. 20%). Geographic differences were minor, although a lower proportion of deceased patients came from central regions, while a slightly higher proportion came from metropolitan areas. For treatment compliance, non‐completion was more prevalent among the deceased patients (78% vs. 72%). A cohort effect is observable, such as older patients, who were more prevalent among the deceased, also entered treatment in earlier calendar years, leading to longer potential follow‐up times and justifying person‐time analyses.

TABLE 1.

Characteristics of patients admitted to substance use treatments, by demographics and survival status in Chile.

Variable	Total, n = 70 064	Survived, n = 67 068	Deceased, n = 2996	SMD b
Alcohol as primary substance, n (%)	24 016 (34.3)	22 274 (33.2)	1742 (58.1)	0.517
Age at discharge, years, n (%) a
18–29	24 341 (34.7)	23 896 (35.6)	445 (14.9)	0.657
30–44	31 485 (44.9)	30 287 (45.2)	1198 (40.0)
45–59	13 098 (18.7)	11 895 (17.7)	1203 (40.2)
60+	1140 (1.6)	990 (1.5)	150 (5.0)
Age at deceased or censored, years, Mdn [IQR]	39.0 [33.0–48.0]	39.0 [33.0–48.0]	47.0 [38.0–54.0]	0.521
Residential treatment (%)	8525 (12.2)	8017 (12.0)	508 (17.0)	0.143
Women (%)	16 733 (23.9)	16 133 (24.1)	600 (20.0)	0.097
Macrozone of treatment center, n (%)	7988 (11.4)	7695 (11.5)	293 (9.8)	0.139
North
Center	7310 (10.4)	7085 (10.6)	225 (7.5)
South‐center	11 943 (17.0)	11 433 (17.0)	510 (17.0)
South	3332 (4.8)	3167 (4.7)	165 (5.5)
Austral	782 (1.1)	730 (1.1)	52 (1.7)
Metropolitan	38 709 (55.2)	36 958 (55.1)	1751 (58.4)
Treatment not completed, n (%)	50 879 (72.6)	48 554 (72.4)	2325 (77.6)	0.120
Length of stay in treatment, Mdn [IQR]	165.0 [91.0–290.0]	165.0 [91.0–290.0]	155.5 [90.0–287.8]	0.020
Birth date, Mdn [IQR]	1981.0 [1972.2–1987.7]	1981.3 [1972.8–1987.9]	1970.7 [1963.0–1979.5]	0.826
Admission date, Mdn [IQR]	2015.3 [2013.1–2017.5]	2015.4 [2013.2–2017.6]	2013.8 [2012.0–2015.7]	0.507
Discharge date, Mdn [IQR]	2015.9 [2013.7–2018.1]	2016.0 [2013.7–2018.2]	2014.4 [2012.5–2016.3]	0.507

Note: Patients were followed from their first discharge between 2010 and 2020. If the outcome event (death) was not observed, the exposure was censored as of 31 December 2020.

Abbreviations: Mdn, median; IQR, interquartile range; SMD, absolute standardized mean difference.

^a Age of patient at first substance use disorder treatment discharge.

^b For continuous variables, SMD is the difference in means divided by the pooled SD; for categorical variables, SMD is the difference in proportions divided by the pooled SD. An absolute SMD >0.15 was used as a reference threshold to flag potentially meaningful differences.

The average age‐sex‐year‐adjusted all‐cause mortality rate was 10.6 (95% CI = 8.6–13.1) per 1000, and the overall all‐cause death rate was 3.7‐fold higher than expected in the general population (SMR = 3.65, 95% CI = 3.52–3.79). Although men had higher absolute mortality rates (DSR = 13.7, 95% CI = 9.5–19.9; Cochran's Q = 6.84, P = 0.009), women exhibited a higher excess risk relative to the general female population (SMR = 5.57, 95% CI = 5.14–6.03; Cochran's Q = 122.33, P < 0.001) (Table 2).

TABLE 2.

All‐cause mortality rate and standardized mortality ratio for patients who accessed SUD treatment by sex and age group.

Characteristic	Observed deaths	Risk time (PY)	CMR (95% CI) per 1000 PY	DSR (95% CI) a per 1000 PY	Heterogeneity b	Expected deaths	SMR (95% CI) a	Heterogeneity c
Total	2996	353 826	8.5 (8.2–8.8)	10.6 (8.6–13.1)	–	820	3.65 (3.52–3.79)	–
Sex
Men	2396	268 325	8.9 (8.6–9.3)	13.7 (9.5–19.9)	6.84(1); P = 0.009	713	3.36 (3.23–3.50)	122.33(1); P < 0.001
Women	600	85 501	7.0 (6.5–7.6)	7.9 (6.6–9.5)		108	5.57 (5.14–6.03)
Attained age, years
18–29	224	77 125	2.9 (2.5–3.3)	2.9 (2.4–3.5)	275.67(3); P < 0.001	69	3.22 (2.82–3.67)	58.65(3); P < 0.001
30–44	1070	183 191	5.8 (5.5–6.2)	5.7 (4.7–6.9)		270	3.95 (3.69–4.23)
45–59	1343	82 119	16.4 (15.5–17.3)	14.2 (12.9–15.7)		341	3.94 (3.73–4.15)
60+	359	11 391	31.5 (28.4–34.9)	46.8 (20.7–106.1)		138	2.59 (2.33–2.87)
Treatment setting
Ambulatory	2488	302 963	8.2 (7.9–8.5)	10.5 (8.2–13.5)	0.24(1); P = 0.626	717	3.47 (3.32–3.63)	39.67(1); P < 0.001
Residential	508	50 863	10.0 (9.2–10.9)	9.7 (7.9–12.0)		103	4.91 (4.45–5.42)
Primary substance
Illicit	1254	246 046	5.1 (4.8–5.4)	8.8 (5.7–13.4)	2.65(1); P = 0.104	441	2.84 (2.69–3.00)	138.28(1); P < 0.001
Licit	1742	107 780	16.2 (15.4–16.9)	12.8 (10.9–15.1)		379	4.59 (4.33–4.86)
Compliance status
Completed	671	83 248	8.1 (7.5–8.7)	7.7 (5.9–9.9)	5.62(1); P = 0.018	245	2.74 (2.49–3.02)	49.67(1); P < 0.001
Not completed	2325	270 578	8.6 (8.3–8.9)	11.9 (9.2–15.4)		575	4.04 (3.85–4.24)

Note: The Q statistic is reported with its degrees of freedom in parentheses and the corresponding P‐value.

Abbreviations: CMR, crude mortality rate; DSR, directly standardized rate; PY, person‐years; SMR, standardized mortality ratio; SUD, substance use disorder.

^a 95% CIs for rates and ratios were adjusted for overdispersion (phi).

^b Heterogeneity test for DSRs across levels of each variable (Cochrane Q statistic).

^c Heterogeneity test for SMRs across levels of each variable (Cochrane Q statistic).

Both sexes show early spikes and then decline. Men have higher absolute rates in the first 2 weeks, but women have higher SMRs in early follow‐up. Post‐discharge absolute mortality rates were higher for men in the first months and slightly higher in subsequent time points. In contrast, post‐discharge relative mortality rates were substantially higher for women in the first few months and remained slightly higher at subsequent time points (Figure 1).

FIGURE 1.

Adjusted all‐cause directly standardized rates (DSRs) and standardized mortality ratios (SMRs) by years since treatment discharge and sex among adults treated for substance‐use disorders in Chile, 2010–2020. (a) Depicts age‐, sex‐ and calendar year DSRs by sex and follow‐up time since treatment discharge in 1000 person years (PY); (b) depicts SMRs by sex and follow‐up time since treatment discharge. The shaded area depicts 95% CIs. The y‐axis in the adjusted SMR plot is presented on a logarithmic scale. Dots mark the midpoints of the analysis intervals: day 0, 14 days (~2 weeks), 3 months, 6 months and 1, 3, 5, 7 and 9 years after discharge. The horizontal dashed gray line indicates no excess deaths relative to the reference population.

An analysis of SMRs across all age groups shows that the mortality of SUD patients exceeded that of the general population, although varying with age (Cochran's Q = 58.65, P < 0.001), exhibiting an inverted‐U pattern peaking in mid‐adulthood, particularly with contrasts between 30 and 44 (Holm‐corrected P < 0.001) and 45 to 59 (Holm‐corrected P < 0.001) compared with 60+ groups, along with contrasts between 30 and 44 (Holm‐corrected P = 0.021) and 45 to 59 (Holm‐corrected P = 0.021) compared with 18 to 29 groups. DSRs differed significantly across age groups (Cochran's Q = 275.67, P < 0.001; all contrasts showed Holm‐corrected P < 0.01).

Stratification by treatment setting showed similar DSRs (Cochran's Q = 0.24, P = 0.626), but heterogeneity in SMRs (Cochran's Q = 39.67, P < 0.001). Patients in residential treatment face nearly five times the mortality risk of the general population (SMR = 4.91, 95% CI 4.45–5.42). Similarly, for primary substance, alcohol use was associated with nearly fivefold higher mortality (SMR = 4.59, 95% CI = 4.33–4.86; Cochran's Q = 138.28, P < 0.001), despite no evidence of differences between patients admitted for alcohol and illicit substances in age–sex–year standardized absolute mortality rates (Cochran's Q = 2.65, P = 0.104). DSRs differed markedly by treatment completion (Cochran's Q = 5.62, P = 0.018), with higher mortality rates in patients who did not complete their treatment (11.9 vs. 7.7), and four times the expected mortality for the general population (SMR = 4.04, 95% CI = 3.85–4.24; Cochran's Q = 49.67, P < 0.001).

Cause‐specific mortality analysis revealed elevated external‐cause mortality for SUD patients (Table 3). Intentional self‐harm showed nearly sevenfold excess risk (SMR = 6.67, 95% CI = 6.05–7.36). Unintentional injuries (SMR = 5.37, 95% CI = 4.79–6.02) and assaults (SMR = 4.98, 95% CI = 4.16–5.96) also showed higher risks. Non‐external underlying causes remained higher than the general population (SMR = 3.49, 95% CI = 3.34–3.65). Specifically, digestive diseases show eight times more deaths than expected in the general population (SMR = 8.20, 95% CI = 7.62–8.83). Extreme excess mortality was observed for mental/behavioral disorders (SMR = 8.71, 95% CI = 6.58–11.54), respiratory (SMR = 5.18, 95% CI = 4.47–5.99) and infectious diseases (SMR = 3.88, 95% CI = 3.27–4.62). Notably, deaths from symptoms, signs and ill‐defined conditions were over five times higher than expected (SMR = 5.18, 95% CI = 4.29–6.26). Code R99 (other ill‐defined and unspecified causes of mortality) accounted for 88% of these cases.

TABLE 3.

Deaths by cause of death group between individuals with and without a history of substance use disorder treatments in Chile.

Cause of death (ICD‐10 codes) a	SUD treatment group					General mortality
	%	CMR b , c	DSR b , c	Obs./exp.	SMR b , d	%	DSR c
Underlying causes	66.2	5.6 (5.4–5.9)	7.3 (6.4–8.3)	1982/567	3.49 (3.34–3.65)	89.4
Infectious and parasitic (A00–B99)	4.3	0.4 (0.3–0.4)	0.4 (0.2–0.7)	129/33	3.88 (3.27–4.62)	2.6	0.11
Malignant neoplasms (C00–C96)	7.3	0.6 (0.5–0.7)	1.2 (0.8–1.8)	220/158	1.38 (1.20–1.59)	29.9	1.30
Endocrine and metabolic (E00–E99)	1.4	0.1 (0.1–0.2)	0.1 (0.1–0.2)	42/21	1.97 (1.46–2.67)	4.8	0.21
Mental and behavioral (F01–F99)	1.6	0.1 (0.1–0.2)	0.1 (0.1–0.2)	49/5	8.71 (6.59–11.53)	0.9	0.04
Nervous system (G00–G99)	1.6	0.1 (0.1–0.2)	0.3 (0.1–1.0)	47/20	2.35 (1.76–3.13)	2.7	0.12
Circulatory (I00–I99)	13.6	1.2 (1.0–1.3)	1.6 (1.2–2.0)	407/131	3.11 (2.82–3.42)	23.3	1.01
Respiratory (J00–J99)	6.0	0.5 (0.4–0.6)	0.6 (0.4–0.9)	181/34	5.18 (4.47–5.99)	7.1	0.31
Digestive (K00–K93)	23.5	2.0 (1.8–2.1)	2.4 (2.0–3.0)	704/85	8.20 (7.62–8.83)	9.3	0.40
Symptoms and signs (R00–R99)	3.6	0.3 (0.3–0.4)	0.4 (0.3–0.7)	107/20	5.18 (4.29–6.26)	1.7	0.08
Other underlying causes (D,H,L,M,N,O,P,Q,S,T,U)	3.2	0.3 (0.2–0.3)	0.2 (0.1–0.2)	96/55	1.72 (1.41–2.11)	7.1	0.31
External causes	33.8	2.9 (2.7–3.0)	3.3 (1.8–6.1)	1014/203	4.99 (4.64–5.37)	10.6
Assaults/aggressions/homicide (X85–Y09)	4.0	0.3 (0.3–0.4)	0.2 (0.2–0.3)	120/24	4.98 (4.16–5.96)	1.0	0.04
Intentional self‐harm (X60–X84)	13.4	1.1 (1.0–1.3)	1.0 (0.7–1.4)	402/60	6.67 (6.05–7.36)	2.8	0.12
Transport accidents (V01–V99)	5.4	0.5 (0.4–0.5)	0.3 (0.2–0.4)	162/56	2.86 (2.45–3.34)	3.0	0.13
Other unintentional external causes (W00–X59)	10.8	0.9 (0.8–1.0)	1.8 (0.6–5.7)	323/60	5.37 (4.79–6.02)	3.5	0.15

Abbreviations: CMR, crude mortality rate; DSR, directly standardized rate; Exp, expected deaths; ICD‐10, International Classification of Diseases, 10th revision; Obs., observed deaths; PY, person‐years; SMR, standardized mortality ratio.

^a U00–U85 Codes for special purposes were not considered. Other external causes (Y10–Y98) were discarded because they represented 0.2% of the external causes of death (n = 7).

^b Values in parentheses represent the 95% CI.

^c Total PY = 353 826. All rates are per 1000 PY.

^d Standard errors were multiplied by an estimated dispersion parameter.

Sensitivity analysis revealed similar, but slightly larger adjusted rates and ratios, and the overall pattern of mortality risk remained consistent, suggesting that the main results are largely robust and reliable, despite different methodological approaches (See Tables S1 and S2).

DISCUSSION

This study described and explored mortality patterns among adults entering SUD treatments from 2010 to 2020. We found marked differences in all‐cause mortality, revealing that patients leaving publicly funded SUD treatment in Chile face a threefold higher mortality risk than expected in the general population. Among external causes, this excess mortality is driven by roughly sevenfold more intentional self‐harm and fivefold more assaults, which is consistent with the literature [20, 42, 43, 44]. Additionally, chronic SUD drives lethal liver/gut damage and immunosuppression, resulting in disproportionately high digestive, respiratory and infectious mortality (8, 5 and 4 times that of the general population, respectively) [45]. This finding raises critical questions about care fragmentation between SUD and mental health services (exacerbated by high out‐of‐pocket costs) as a preventable driver of premature mortality [9, 30].

Women showed disproportionately high relative mortality (SMR), which contrasts with the higher absolute risks seen in men [25, 46, 47], who often delay seeking medical care until their health problems worsen [48]. Although partly driven by the lower background mortality of women in the general population, this pronounced disparity is also consistent with the ‘telescoping effect’, where women experience an accelerated disease trajectory [49, 50]. This rapid progression may result from the intersection of biological factors (e.g. diminished gastric alcohol metabolism) and socio‐structural stressors, such as the intense stigmatization and isolation often faced by women with SUDs [50, 51, 52, 53]. Follow‐up patterns show this excess risk for women persists after 5 years post‐discharge.

The primary substance of use was also an important source of heterogeneity (Havard et al.) [17]. Patients admitted because of licit substances (i.e. alcohol) exhibited nearly fivefold higher mortality risk than the general population. This warrants attention, as people using alcohol constitute a significant portion of publicly funded treatment episodes in Chile (36%) [8]. A cited study from Brazil reported similar patterns, with alcohol‐related disorders showing an SMR of 4.91, higher than that for non‐alcohol psychoactive substance use (SMR = 2.02) [25].

Treatment setting was associated with mortality, with patients in residential care showing nearly five times the number of deaths expected based on general‐population mortality rates (SMR = 4.91). This may reflect residential patients often having a higher SUD severity profile (i.e. comorbidity) than peers in outpatient treatments [17, 33]. Hence, mortality comparisons between settings are descriptive and are consistent with differences in group composition.

Our 72.6% non‐completion rate aligns with national evaluations of Chile's public network, where completion often falls below 20% [54]. Although exceeding global dropout estimates (~30%) [55], this attrition highlights systemic barriers linked to readmission, substance use reductions and criminal justice contact [21, 56, 57]. Patients who did not complete treatment had higher crude rates (11.9 vs. 7.7 for patients with complete treatments) and four times higher expected mortality than the general population. Previous studies have found greater mortality risks in patients who dropped out [19, 58]. Because non‐completion is also linked to higher readmission rates [57], future research should assess post‐treatment events like readmission and hospitalization as mortality risk factors.

We focused on adults 18 to 64 years because they bear the heaviest burden of substance‐related harm [29]. Additionally, this group aligns with recent increases in SUD‐related mortality observed globally among middle‐age adults [4]. Similarly, this study found that mortality exceeded that of the general population across all age groups, but varied, showing an inverted‐U pattern that peaked in mid‐adulthood. Specifically, the 30 to 44 and 45 to 59 age groups showed greater SMRs, consistent with those reported by other studies [59], although patients age 60+ showed greater mortality rates. We suspect this pattern is explained by the mortality dynamics of alcohol use, as the cumulative consequences of chronic consumption often lead to fatal outcomes at a later age [60]. We hypothesize that elevated alcohol‐related mortality among older adults may partly mask prior illicit substance use histories. To illustrate this, the proportion of patients admitted primarily for alcohol use increased with age, from a range of 17% to 32% in younger to 60% to 86% in the 45+ age groups, which is in line with other studies [61]. However, this potential masking effect, and whether mid‐life substance substitution systematically occurs, warrant further investigation. Additionally, studies should put their focus on polysubstance use, because it may show high mortality risks similar to alcohol [62].

Follow‐up inspection allowed us to notice that 6.1% (n = 184) of deaths occurred within the first 2 weeks after treatment episodes. We suspected that deaths could be triggering administrative termination as a classification artifact. However, sensitivity analyses including patients with unfinished treatments yielded virtually identical SMRs, indicating that excess mortality may be robust to this source of bias.

This study benefits from a robust, nation‐wide data infrastructure, in contrast to contexts like Brazil, where data quality is compromised by parallel reporting systems, misclassified deaths and missing information [63].

This study has several limitations. First, insufficient data granularity precluded the inclusion of geographical and socio‐demographic characteristics in this analysis [22, 64, 65]. Second, interpreting corrections for multiple comparisons and heterogeneity tests should be done cautiously, because they might inflate false negative rates [66]. Third, variables such as socio‐economic status were unavailable and may inflate relative mortality estimates, despite standardization by age, sex and year.

Implications

The observed mortality differences are descriptive and subject to unobserved confounding from many sources, such as patient (e.g. substance use history) and treatment (e.g. provider quality) factors [67]. Furthermore, individuals seeking treatment have higher baseline mortality risks, precluding causal interpretation of risks as evidence of effectiveness [14, 68, 69].

Nevertheless, our results possess tangible applications for public health policy and clinical practice in Chile. The high mortality among patients with SUDs, especially the relative mortality of women and those with alcohol‐related disorders, represents a crisis of preventable deaths occurring well below the national life expectancy (76.7 for men, 81.4 for women in 2022). This premature mortality is a stark indicator of the systemic inequities faced by this often‐marginalized population. Interventions to improve treatment retention and post‐discharge continuity, particularly for individuals using alcohol, could effectively avert significant numbers of additional deaths over time [70]. As a principle of health equity, the greater relative excess mortality for women in treatment (compared to the general female population) warrants future assessment to reduce this differential disparity [71]. Because many contributing factors are modifiable, these findings emphasize the need for targeted interventions [20, 72]. For example, knowledge about patterns of patient visits and mortality risk, derived from large datasets, could contribute to a more adaptive patient safety system [62]. Another strategy is the systematic monitoring and timely treatment of psychiatric comorbidities in these populations to reduce their severity and, consequently, self‐harm risk [73], implementing programs to remove financial barriers and improve care coordination for these patients [74].

Preventable deaths for people with alcohol‐related disorders in our findings point toward major shortcomings in national prevention and control strategies. For example, Chile has one of the highest safe‐drinking thresholds, according to a review of 37 countries [75]. This is compounded by under prioritized alcohol control policies [76], despite recent health warnings [77, 78]. Adopting proven alcohol control policies, such as those embedded in the LIVE LIFE suicide prevention strategy led by the World Health Organization, could, therefore, be critical in reducing the high mortality from intentional self‐harm [79]. Although research in the Chilean general population has found no association between alcohol control strategies and suicide rates, these policies may have a protective effect on individuals with alcohol use disorders [80].

Using data on people with SUDs is vital for developing effective treatments and improving cooperation between service providers [81]. The persistent burden of SUDs in Chile has already been recognized as a key policy response in the Chilean National Health Strategy for 2021 to 2030, which highlights the need for robust public health initiatives.

CONCLUSION

Individuals in Chile with a history of publicly funded SUD treatment face a more than threefold excess risk of death. The highest relative mortality was observed in women, people who used alcohol as their primary substance and people who were in residential settings. This risk is attributable to higher rates of death from external causes, notably suicide and homicide‐related and specific physical illnesses like digestive, respiratory and infectious diseases. Findings underscore the need for systematic post‐treatment surveillance and sustained support after treatment.

AUTHOR CONTRIBUTIONS

Andrés González‐Santa Cruz: Conceptualization; data curation; software; formal analysis; visualization; project administration; writing—original draft; writing—review and editing; investigation; funding acquisition; methodology. Alvaro Castillo‐Carniglia: Supervision; writing—review and editing; project administration; conceptualization. Jay S. Kaufman: Supervision; writing—review and editing; conceptualization.

DECLARATION OF INTERESTS

None.

Supporting information

Figure S1. Flowchart of cohort inclusions and exclusions.

Table S1. All‐cause mortality rate and standardized mortality rate for patients who accessed SUD treatment by sex and age group, last treatment available.

Table S2. All‐cause mortality rate and standardized mortality ratio for patients who accessed SUD treatment by sex and age group, first treatment including unfinished treatments (referral outside SENDA network, currently in treatment and discharged after death).

González‐Santa Cruz A, Castillo‐Carniglia A, Kaufman JS. Excess mortality following discharge from substance use disorder treatment in Chile. Addiction. 2026;121(8):2060–2070. 10.1111/add.70420

DATA AVAILABILITY STATEMENT

The data are not available due to data use agreement restrictions, but may be available upon request to the Chilean National Drug Agency.