Individual and social determinants of treatment prognosis in people with substance use disorders

What predicts therapeutic outcome and recurrence is one of the most critical and less known aspects of addiction treatment. Prognostic studies have mostly examined individual based characteristics typically used in the biomedical literature such as age, principal drug of choice, personality traits or brain features (1, 2). However, there is growing recognition of the relevance of social determinants such as socioeconomic background, job and housing security or wealth (3). To date, very few studies have jointly examined the individual and social determinants of treatment prognosis, as illustrated by a limited number of systematic reviews on the topic over the last decade (4-7). Two of these reviews included participants with miscellaneous substance use disorders (4, 5), one focused on alcohol (7) and one on smoking (6). The Lewer et al. review (4) encompassed all treatment approaches and used health care utilisation as the primary outcome. Their findings showed that unstable housing, use of injected drugs and mental health problems were negatively associated with health care utilisation, whereas engagement with opioid substitution treatments was positively associated with the outcome. The Lappan et al review (5) focused on face-to-face psychosocial treatments, mostly in the United States, and used treatment retention as the primary outcome. Their findings showed that higher percentage of African Americans, lower income, inclusion of participants who used stimulant drugs, heavier cocaine use and treatments with more treatment sessions and longer session lengths were negatively associated with retention, whereas smoking more cigarettes per day and higher frequency of heroin use were linked to better retention. Among people with tobacco use disorder, greater dependence and having made a previous quit attempt were associated with lower quitting success, whereas higher affluence and social grade predicted better outcomes although they were comparatively less studied (6). In people with alcohol use disorder, severity of dependence, mental health problems, lower self-efficacy and motivation, and shallower treatment goals were prospectively associated with higher recurrence (selected studies had a minimum 3-month follow up) (7). Cognitive and socioeconomic variables were also negatively associated with recurrence although not as consistently. Overall, findings show that both individual (e.g. drug type, patterns of use), treatment (e.g. number and length of sessions) and social (e.g. affluence, housing) significantly predict clinical outcome and risk of recurrence, with the latter (social) variables consistently appearing across reviews but lacking sufficient evidence. This gap resonates with emerging evidence on and increased awareness of the importance of social determinants of health for addiction mechanisms (3). To our knowledge, there are no existing adequately sampled studies that have systematically assessed the cumulative, differential, and interactive contribution of both individual and social determinants in predicting addiction treatment outcome and risk of recurrence, nor attempts to generate predictive models that can optimise individualised risk prediction. The latter is critical for translation of research into clinical practice, as it may enable clinicians to use algorithms to estimate risk and personalise treatment for each client. In this project, we will leverage an existing dataset of circa 100,000 participants with substance use disorders who were registered and monitored through the digital health information system of a state-wide addiction treatment network between 2015 and 2023, to examine the prognostic contribution of both individual and social factors to clinical outcomes (treatment outcome and recurrence) over two years since treatment onset. Given the large sample size we will randomly split the sample in two subsamples (i.e., henceforth the training and testing subsamples) to build a predictive model in a training subsample and test the ability of the predictive model to forecast individualised risk using unseen data in a testing subsample. As the influence of social factors was likely intensified during the COVID-19 pandemic (8-10), we will use time-lagged data (pre- versus post-pandemic onset) to test this potential intensification as a secondary aim. Aims 1. To examine the prognostic contribution of individual and social determinants of addiction to treatment outcome and risk of recurrence over a period of two years in people with substance use disorders monitored through the health information system of a state-wide treatment network. 1a) To quantify both the cumulative and differential contribution of individual versus social determinants to therapeutic outcome and risk of recurrence. 1b) To explore meaningful interactions between individual and social determinants in prognostically predicting treatment outcome and risk of recurrence. 1c) To study if the contribution of individual vs social determinants to treatment outcomes and risk of recurrence shifted during the COVID-19 pandemic in Spain (2020-2022) relative to the immediately preceding period (2017-2019). 2. To test the accuracy of a data-driven model combining individual and social determinants (generated in a training subsample) to predict individual treatment outcome and recurrence in a testing subsample.