People with opioid use disorders: A taxonomy of treatment entrants to support the development of a profile-based approach to care
Abstract
Introduction
People with opioid use disorders (OUD) present with high levels of medical and psychosocial vulnerabilities. In recent years, studies have highlighted a shift in demographic and biopsychosocial profiles of people with OUD. In order to support the development of a profile-based approach to care, this study aims to identify different profiles of people with OUD in a sample of patients admitted to a specialised opioid agonist treatment (OAT) facility.
Methods
Twenty-three categorical variables (demographic, clinical, indicators of health and social precariousness) were retrieved from a sample of 296 patient charts in a large Montréal-based OAT facility (2017–2019). Descriptive analyses were followed by a three-step latent class analysis (LCA) to identify different socio-clinical profiles and examine their association with demographic variables.
Results
The LCA revealed three socio-clinical profiles: (i) “polysubstance use with psychiatric, physical and social vulnerabilities” (37% of the sample); (ii) “heroin use with vulnerabilities to anxiety and depression” (33%); (iii) “pharmaceutical-type opioid use with vulnerabilities to anxiety, depression and chronic pain” (30%). Class 3 individuals were more likely to be aged 45 years and older.
Discussion and Conclusion
While current approaches (such as low- and regular-threshold services) may be suited for many OUD treatment entrants, there may be a need to improve the continuum of care between mental health, chronic pain, and addiction services for those characterised by the use of pharmaceutical-type opioids, chronic pain and older age. Overall, the results support further exploring profile-based approaches to care, tailored to subgroups of patients with differing needs or abilities.
Key Points
- Three classes of socio-clinical profiles emerged from the analysis.
- Profile 1 is characterised by polysubstance use and multiple vulnerabilities.
- Profile 2 is characterised by heroin use and fewer vulnerabilities.
- Profile 3 is characterised by older age, chronic pain and pharmaceutical-type opioid use.
- Results support further exploring profile-based approaches to care in opioid use disorder.
1 INTRODUCTION
Opioid use disorder (OUD) is a chronic condition characterised by persistent opioid use despite associated adverse consequences such as mental, physical and psychosocial difficulties [1]. The exact prevalence of OUD is unknown, but in 2013, it was estimated at 0.3% of Canadians aged 15 years and older [2]. In 2019, 1% of the population aged 15 years and older engaged in problematic opioid use [3], which includes OUD [4]. While the prevalence may appear low, OUD and OUD-related problems exert significant pressure in the fields of health care and social services, criminal justice, child protection services and the employment market [5]. People with OUD present high levels of comorbidities, such as chronic pain [6-10], mental health disorders and other substance use disorders [11]. Key psychosocial characteristics such as low employment levels and high rates of criminal charges have also been previously associated with OUD [12-15].
Effective treatment of OUD is based on opioid agonist treatment (OAT). OAT refers to ‘long-term treatment with an opioid agonist medication recognised for use in the treatment of opioid use disorder’ [16]. For instance, methadone and buprenorphine-naloxone have demonstrated substantial effectiveness in relieving opioid withdrawal symptoms, reducing opioid use, fostering individual uptake and retention in OUD treatment [17], reducing morbidity and mortality, and reducing risk of infections for people who use injection modes of administration [16]. Other treatment options such as slow-release oral morphine and injectable diacetylmorphine or hydromorphone (iOAT) are also available in some instances [17]. Clinical guidelines recommend combining pharmacological treatment with ancillary services (such as psychosocial treatment interventions, harm reduction strategies or general health-care services) to address other medical or psychosocial needs, as necessary [16-18].
Over the past 20 years, the rise in opioid use [19, 20], and opioid overdose-related deaths [19-22] has been increasingly documented in the United States and Canada. This phenomenon is explained, at least partly, by the rise of opioid prescribing that was observed during the first decade of the new millennia [23, 24]. To address this growing problem and prevent iatrogenic opioid addiction, strategies such as restricting supply and influencing prescribing habits have been implemented, through prescriber education, prescribing guidelines and prescription drug monitoring programs, for instance [25]. However, for patients who had already developed an iatrogenic opioid addiction or a high opioid tolerance, tightened opioid prescription protocols meant having to find an alternative access to the substance. In other words, restricting the supply of prescription opioids in the context of a high demand ‘contributed to a burgeoning use of illicit opioids’ [26]. In addition, highly potent opioids such as fentanyl have appeared on the illegal market, contributing to an increase in overdose deaths in Canada and the United States [27-30]. In sum, the successive expansion and tightening of opioid prescribing practices, the emergence of highly potent opioids on the illegal market as well as the rise in opioid use and opioid overdose-related deaths constitute shifts that need to be considered to address the negative consequences of opioid use.
In the midst of the shifting landscape, a few studies have shown changes in the characteristics associated with OUD [12, 31-33]. For instance, Cicero et al. [31] highlighted that opioid use was no longer confined solely to urban settings and marginalised populations who use heroin, but extended to middle and upper-class populations outside of urban centres [31]. Fischer et al. [34] found differences between people who use heroin and those who use prescription opioids in terms of age, other substance use, service utilisation, ethnicity, income source, modes of administration and physical health problems in people with OUD. Through a latent class analysis (LCA), Nielsen et al. [35] identified three subgroups of people with OUD in a convenience sample of Australian treatment seekers. The groups are described as ‘traditional opioid injectors in treatment’, ‘high-risk, disadvantaged injectors’ and ‘non-injectors and iatrogenic graduates’. The authors highlight the important differences between participants who inject heroin and those who use pharmaceutical opioid analgesics without injection. The statistical model suggests that the latter are more likely to be older, in poorer physical and psychological health, to be women, and to have initiated opioid use through a prescription. They are less likely to have a history of substance use disorder treatment and to report a lifetime exposure to heroin. The authors add that this profile is represented by a small proportion of the sample (12%), which suggests that they may not be reached by the addiction service offering [35]. These results are important in order to support the service organisation for people with OUD. Yet, no studies have used the LCA method in a Canadian setting to gain a better understanding of different groups of people with OUD.
In the medical field, tailored services for different patient profiles have existed for millennia to allow for adapting care to subpopulations with different needs or capacities [36]. In fact, adaptations in the service offering for people with OUD were implemented during the last decades to expand treatment reach to people who are disaffiliated from the health and social services network. More specifically, flexible ‘low-threshold’ programs were implemented to reduce barriers to access and retention for patients with low adherence or difficulty following program rules [37]. However, the current shifting landscape in the opioid market and opioid use calls for updating our understanding of different profiles of people with OUD, to ensure that no subgroup is left behind in the service offering.
Overall, it is important to understand how socio-clinical patterns may differ among different OUD patient profiles, suggesting distinct treatment and service needs [33, 35-38]. Profiles and associated needs may vary according to geographical differences and socio-political environment, which can also change over time [39]. Thus, this study aims to identify different profiles of people with OUD in a sample of patients admitted in a specialised OAT facility located in a Canadian urban setting, in order to support the development of a profile-based approach to care.
2 METHODS
2.1 Sample and ethics
A sample of patient charts in a large Montréal-based OAT facility was consulted. The decision to conduct the study in this facility was based on their high number of patients, their expanded OAT offer (i.e., methadone, buprenorphine-naloxone, slow-release oral morphine, safer supply and injectable hydromorphone) and their regular and low-threshold services. These characteristics permitted access to a representative sample of the study population. The selected charts were those of patients admitted electively between 2017 and 2019. Data were retrieved manually from a total of 296 paper charts. The study was conducted at the Centre de Recherche et d'Aide pour Narcomanes, a large specialised program for treatment of OUD based in downtown Montréal, Canada. Authorisation to conduct a chart review was obtained through the ethics board of the CIUSSS du Centre-Sud-de-l'île-de-Montréal. As is often the case with chart reviews, obtaining individual consent was not feasible. Additionally, the current study population can be especially hard to reach; many of them present residential instability or do not have a phone number. Measures were implemented to ensure full anonymity and confidentiality. Chart numbers and nominal information were removed from the database.
2.2 Measures and procedures for data collection and extraction
At the time of data collection, the treatment facility used paper charts. Medical notes were free text, while nursing and psychosocial notes were checkboxes, open-ended questions and free text. Nursing and psychosocial assessment tools were derived from the addiction severity index [40, 41].
Twenty-three self-reported or clinically documented categorical variables were extracted from the charts to draw profiles at the time of admission to treatment. Variable selection was based on typically reported correlates of OUD [42-44] in comparable profile-based studies [35], and contingent on availability of the data in the patient files. Demographic variables (gender identity* and age), clinical variables (substances used in the last 30 days, history of mental health,† chronic pain—over 3 months—at admission, time since first opioid use, past OAT episodes) and indicators of social precariousness and health risk at the time of admission (employment, injection in the last 30 days, history of overdoses, sexually transmitted and blood-borne infections (STBBI), sex work, judiciarisation, residential instability) were collected. Data were collected based on the information available in the charts, either self-reported by the patient as part of their medical history, or confirmed by formal assessments by clinicians, and logged through free text notes or checkboxes. Variable description, collection and operationalisation are described in Table 1.
| Variable | Description | Collection and operationalisation | |
|---|---|---|---|
| Sociodemographic variables | Gender identity at admission | Male or femalea | Patient-reported |
| Age at admission | Less than 45 or 45 and up | ||
| Background variables | Time since first opioid use (years) | 10 or less or more than 10 | Patient-reported |
| Past OAT episodes | Yes or no | Patient-reported | |
| Substances used | Heroin use at admission (last 30 days) | Yes or no | Patient-reported |
| Pharmaceutical-type opioidsb at admission (last 30 days) | |||
| Cannabis use at admission (last 30 days) | |||
| Cocaine or crack use at admission (last 30 days) | |||
| Alcohol use at admission (last 30 days) | |||
| Non-prescribed medication use at admission (last 30 days) | |||
| Other substance use (last 30 days) | |||
| Mental health diagnoses or symptoms | History of anxiety disorders diagnoses or symptoms | Yes or no | Patient-reported OR clinician's diagnostic impression OR confirmed diagnosis |
| History of mood disorders diagnoses or symptoms | Patient-reported OR clinician's diagnostic impression OR confirmed diagnosis | ||
| Personality disorders | Patient-reported OR clinician's diagnostic impression OR confirmed diagnosis | ||
| History of psychosis | Patient-reported OR clinician's diagnostic impression OR confirmed diagnosis | ||
| Chronic pain | Chronic pain at admission | Yes or no | Patient-reported OR clinician's diagnostic impression OR confirmed diagnosis |
| Indicators of social precariousness and health risk | Employment at admission | Yes or no | Patient-reported |
| Injection at admission | Patient-reported | ||
| Overdose history | Patient-reported (at least one overdose) | ||
| STBBIs at admission | Patients-reported OR confirmed diagnosis | ||
| Sex work at admission | Patient-reported | ||
| History of judiciarisation | Patient-reported | ||
| Residential instability at admission | Patient-reported |
- Abbreviations: OAT, opioid agonist treatment; STBBI, sexually transmitted and blood-borne infections.
- a In the sample, no patient was reported identifying as ‘other’.
- b Produced or obtained licitly or illicitly. The term ‘pharmaceutical-type opioids’ is used to describe any non-heroin opioids.
Three coders were involved in data collection. One of them had experience with chart reviews and was on site throughout the whole process to assist the other coders. The standardised extraction form was accompanied by a document describing all the inclusion and exclusion criteria for each variable. Patient charts were read from beginning to end in order to document the patient characteristics at admission. At treatment entry, patients with OUD often present opioid withdrawal symptoms, which can create difficult conditions for a comprehensive assessment within one session. For that reason, the initial assessment is often broken down into several appointments with different professionals. For the purpose of this study, coders reviewed all the notes to identify information on the study variables and worked in close collaboration to consult with each other when in doubt. Calibration exercises were regularly performed by coders. One out of every 15 charts was blind-coded by two coders and results were compared to prevent discrepancies.
2.3 Statistical analyses
Descriptive analyses (frequencies) were performed using the SPSS software. Fifty-nine percent of cases had no missing data, 23% of cases had one missing variable and the remaining 18% had more than one variable missing. To ensure that missing data was missing completely at random, chi-square analyses were performed to establish that they were fairly distributed across demographic variables (gender identity and age). Table 2 presents the correlation matrix of all the variables used for the LCA.
| 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 | 10 | 11 | 12 | 13 | 14 | 15 | 16 | 17 | 18 | 19 | 20 | 21 | |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
|
1 | ||||||||||||||||||||
|
−0.421** | 1 | |||||||||||||||||||
|
1 | ||||||||||||||||||||
|
0.136* | 0.269** | 1 | ||||||||||||||||||
|
1 | ||||||||||||||||||||
|
0.214** | 1 | |||||||||||||||||||
|
0.114* | 1 | |||||||||||||||||||
|
0.166** | 1 | |||||||||||||||||||
|
0.216** | 0.168** | 1 | ||||||||||||||||||
|
−0.161** | −0.195** | 0.141* | 1 | |||||||||||||||||
|
−0.161* | 0.361** | 1 | ||||||||||||||||||
|
0.140* | 0.171** | 1 | ||||||||||||||||||
|
0.335** | 1 | |||||||||||||||||||
|
−0.314** | 0.214** | −0.228** | −0.131* | 1 | ||||||||||||||||
|
0.430** | −0.130* | 0.235** | 0.231** | −0.212* | 1 | |||||||||||||||
|
0.166** | 0.131* | 0.153* | 0.124* | 0.138* | 0.268** | 1 | ||||||||||||||
|
0.154* | 0.242** | 0.191** | −0.185** | 0.226* | 0.358** | 0.304** | 1 | |||||||||||||
|
0.141* | 0.145* | 0.169* | 0.249** | 0.131* | 0.288** | 1 | ||||||||||||||
|
0.134* | 0.161** | 0.183** | 0.245** | 0.235** | 1 | |||||||||||||||
|
0.125* | 0.200** | 0.303** | 0.120* | −0.131* | 0.137* | 0.302** | 0.377** | 0.286** | 0.559** | 0.273** | 0.325** | 1 | ||||||||
|
−0.175** | −0.157** | −0.149* | −0.199** | −0.182** | −0.170** | −0.215** | −0.400** | 1 |
- Abbreviations: OAT, opioid agonist treatment; STBBI, sexually transmitted and blood-borne infections.
- * p < 0.05.
- ** p < 0.01.
A LCA was performed on all socio-clinical variables using the Mplus8 software [45] in order to identify subtypes of individuals displaying similar patterns of characteristics, that is, having a certain homogeneity [46]. This method allows one to determine the probability of belonging to one class or another, as well as the probability of an individual in a given class exhibiting a specific characteristics [46]. ‘LCA can be particularly useful for identifying subgroups of individuals who could benefit from a common intervention based on their shared characteristics’ [47]. LCA can be performed using categorical data [46], with a desirable sample size of 300 cases or more [48]. The ‘maximum likelihood estimation with robust standard errors’ method was used to handle missing data [46]. Three indicators were used to identify the optimal solution: the Bayesian information criterion, entropy and the Lo–Mendell–Rubin test.
The new three-step maximum likelihood method was used [49, 50]. This approach allows for assessing the relationships of the latent class memberships with distal variables, while accounting for the classification uncertainty rate, and without needing to re-calculate estimations when including distal variables [50]. The first step serves to build the latent class model based on socio-clinical characteristics. In the second step, individuals are assigned to classes, based on class membership probabilities. For the third step, logistic regressions are conducted to assess significant associations between the socio-clinical classes and both distal demographic variables (age and gender identity).
3 RESULTS
3.1 Descriptive analysis of variables of interest
The sample's characteristics are presented in Table 3. Notably, at the time of admission, the majority of the valid cases identified as men, were younger than 45 years old, tried OAT in the past, had a history of judiciarisation, used an injection method of opioid administration and reported lifetime symptoms or diagnosis of anxiety or mood disorder.
| n (%) | Valid N | |
|---|---|---|
| Identify as men | 210 (70.9%) | 296 |
| 45 years and older | 108 (36.5%) | 296 |
| Heroin use (last 30 days) | 181 (61.6%) | 294 |
| Pharmaceutical-type opioid use (last 30 days) | 226 (76.6%) | 295 |
| Cannabis use (last 30 days) | 103 (34.9%) | 295 |
| Cocaine or crack use (last 30 days) | 137 (46.3%) | 296 |
| Other substance use (last 30 days) | 37 (12.5%) | 295 |
| Alcohol use (last 30 days) | 120 (40.7%) | 295 |
| Non-prescribed medication use (last 30 days) | 82 (27.8%) | 295 |
| Opioid use for over 10 years | 127 (44.7%) | 284 |
| Past OAT episodes | 148 (51.4%) | 288 |
| History of anxiety disorders diagnoses or symptoms | 141 (53.4%) | 264 |
| History of mood disorders diagnoses or symptoms | 137 (51.5%) | 266 |
| Personality disorders | 36 (13.5%) | 267 |
| History of psychosis | 30 (11.3%) | 266 |
| Chronic pain | 84 (28.9%) | 291 |
| Employment | 82 (29.6%) | 277 |
| Injection (last 30 days) | 171 (57.8%) | 296 |
| Overdose history (at least one) | 120 (43.8%) | 274 |
| STBBI | 81 (30.3%) | 267 |
| History of judiciarisation | 181 (71.8%) | 252 |
| Sex work | 20 (7%) | 287 |
| Residential instability | 104 (35.4%) | 294 |
- Abbreviations: OAT, opioid agonist treatment; STBBI, sexually transmitted and blood-borne infections.
3.2 Latent class analysis
One- through six-class models were computed [49]. In reviewing the assessment indicators for the models, the three-class solution showed the best model fit, with the lowest Bayesian information criterion [47]. The entropy above 0.8 indicates that this model accurately defines classes. The Lo–Mendell–Rubin's p-values indicate that the third model is statistically better than the second one, but that the fourth is not statistically better than the third [51]. A full overview of all six models and their assessment indicators is presented in Table 4.
| 1 class | 2 classes | 3 classes | 4 classes | 5 classes | 6 classes | |
|---|---|---|---|---|---|---|
| BIC | 7133.688 | 6883.291 | 6839.470 | 6894.818 | 6950.955 | 7022.750 |
| Entropy | — | 0.851 | 0.861 | 0.847 | 0.885 | 0.869 |
| LMR test (p) | — | 0.0000 | 0.0000 | 0.2358 | 0.2193 | 0.5761 |
- Abbreviations: BIC, Bayesian information criterion; LMR, Lo–Mendell–Rubin.
The probabilities for an individual in a given class to present a specific characteristic are presented in Table 5. Compared to individuals from classes 2 and 3, individuals belonging to class 1 (37% of the sample) present a higher probability of using multiple types of substances, presenting a confirmed or suspected personality disorder diagnosis, reporting a history of psychosis, using a method of administration by injection, reporting a least one overdose, presenting a sexually transmitted blood-borne infections, having a criminal record, practising sex work and being residentially unstable. According to these characteristics, individuals belonging to class 1 are differentiated by ‘polysubstance use with psychiatric, physical and social vulnerabilities’. Compared to individuals from classes 1 and 3, individuals belonging to class 2 (33% of the sample) have a higher probability of using heroin and alcohol and of being employed. They have a greater probability than class 1 of reporting a diagnosis or symptoms of an anxiety disorder (present or past) or a mood disorder (present or past). Based on these characteristics, individuals belonging to class 2 are differentiated by ‘heroin use with vulnerabilities to anxiety and depression’. Finally, compared to individuals from classes 1 and 2, individuals belonging to class 3 (30% of the sample) have a higher probability (100%) of using pharmaceutical-type opioids and cannabis and of reporting chronic pain. They have a greater probability than class 1 of reporting a diagnosis or symptoms of an anxiety disorder (present or past) or a mood disorder (present or past). Based on these characteristics, individuals belonging to class 3 are differentiated by ‘pharmaceutical-type opioid use with vulnerabilities to anxiety, depression and chronic pain’.
| C1 (n = 110) polysubstance use with psychiatric, physical and social vulnerabilities | C2 (n = 96) heroin use with vulnerabilities to anxiety and depression | C3 (n = 90) pharmaceutical-type opioid use with vulnerabilities to anxiety, depression and chronic pain | |
|---|---|---|---|
| Heroin use | 0.700 | 0.990 | 0.124 |
| Pharmaceutical-type opioid use | 0.812 | 0.491 | 1.000 |
| Opioid use (any type) for over 10 years | 0.560 | 0.375 | 0.396 |
| Past OAT episodes | 0.689 | 0.461 | 0.358 |
| Cannabis use | 0.333 | 0.315 | 0.405 |
| Cocaine or crack use | 0.670 | 0.422 | 0.255 |
| Other substance use | 0.158 | 0.125 | 0.087 |
| Alcohol use | 0.354 | 0.465 | 0.410 |
| Injection | 0.864 | 0.671 | 0.133 |
| Non-prescribed medication use | 0.340 | 0.220 | 0.264 |
| Overdose | 0.672 | 0.328 | 0.274 |
| Employment | 0.023 | 0.570 | 0.294 |
| History of anxiety disorders diagnoses or symptoms | 0.403 | 0.590 | 0.603 |
| History of mood disorders diagnoses or symptoms | 0.446 | 0.511 | 0.587 |
| Personality disorders | 0.188 | 0.076 | 0.143 |
| History of psychosis | 0.256 | 0.029 | 0.059 |
| Chronic pain | 0.276 | 0.092 | 0.510 |
| STBBIs | 0.706 | 0.157 | 0.015 |
| History of judiciarisation | 0.883 | 0.583 | 0.613 |
| Sex work | 0.194 | 0.000 | 0.000 |
| Residential instability | 0.885 | 0.050 | 0.034 |
- Abbreviations: OAT, opioid agonist treatment; STBBI, sexually transmitted and blood-borne infections.
- a The highest probability for each variable is highlighted in bold.
3.3 Logistic regressions
Bivariate analyses revealed that only age contributed to predicting belonging to the identified classes. In fact, the analysis demonstrated that individuals in class 3 (‘pharmaceutical-type opioid use with vulnerabilities to anxiety, depression and chronic pain’), are more likely to be 45 years of age and older. Results are presented in Table 6.
| Heroin use with vulnerabilities to anxiety and depression | Pharmaceutical-type opioid use with vulnerabilities to anxiety, depression and chronic pain | |||||||
|---|---|---|---|---|---|---|---|---|
| B | SE | p | OR | B | SE | p | OR | |
| Gender identity | 0.116 | 0.348 | 0.739 | - | −0.100 | 0.330 | 0.761 | - |
| Age 45 and over | −0.157 | 0.354 | 0.658 | - | 1.181 | 0.321 | 0.000 | 3.256 |
- Note: Reference category, ‘Polysubstance use with psychiatric, physical and social vulnerabilities’.
- Abbreviations: B, regression coefficient; OR, odds ratio; p, p-value; SE, standard error.
4 DISCUSSION
The purpose of this study was to identify different profiles (i.e., classes) of people dealing with OUD, in order to support the development of a profile-based approach to care. Significant associations between the socio-clinical classes and two distal demographic variables (age and gender identity) were assessed. In line with other studies conducted in recent years [34, 35, 52, 53], three classes of people with OUD emerge from the LCA.
The first profile identified in this study is characterised by multiple biopsychosocial vulnerabilities. Several studies highlight a similar subgroup of people who use multiple substances [34, 52], present high risk for overdose [34, 35], homelessness [35], mental health disorders [35], and higher addiction severity [53]. High overdose odds and health risks for these individuals calls for rapid and flexible service provision, such as low-threshold services [54] and one-stop shops [55]. The removal of structural and process barriers to OAT [56] appears essential for reaching people who may be homeless, phoneless, and have atypical schedules organised around sex work and panhandling. Moreover, people fitting this highly vulnerable profile present high odds of homelessness, sex work, STBBIs and injection, implying the need for services such as foot care [57, 58], treatment of skin and soft tissue infections [58], and sexual and reproductive health care [59]. For these reasons, they would benefit from easy access to global physical health care. In addition, high odds of psychotic disorders for this profile calls for easily available mental health services. In this regard, integrated and co-located comorbidity treatment (in the form of a one-stop shop model) may foster treatment adherence for this hard-to-reach patient profile [60]. Importantly, Chalabianloo et al. [61] insist that low-threshold services should not require abstinence. Incidentally, the choice of an appropriate opioid agonist medication should be coherent with the patients' treatment objectives. For instance, methadone may be preferred to buprenorphine-naloxone by patients who are not ready to stop opioid use altogether [62]. Moreover, because individuals in this profile exhibit high odds of injection, previous OAT attempts, overdose risk as well as biopsychosocial comorbidities, they might be eligible for iOAT [63]. Overall, highly flexible and comprehensive structures appear helpful to provide encompassing and easily accessible services to this vulnerable patient profile.
A second profile is characterised by heroin and alcohol use, lifetime anxiety and depressive symptoms, and higher odds of employment compared to other profiles. This higher socioeconomic status profile is similar to the subgroup referred to by Nielsen et al. [35] as the ‘traditional profile’. Other studies highlight a similar subgroup of heroin users [34, 53] with fewer comorbidities [34, 53]. Individuals with this profile may present higher functional levels, allowing for treatment uptake in many different settings, including primary care. People in this profile may even be good candidates for what Samso Jofra et al. [64] refer to as interim OAT, which involves initiating an OAT medication alone to bridge waitlist, thus avoiding risks of overdose or psychosocial deterioration while people wait for treatment. Nonetheless, high odds of anxiety or depressive symptoms history in this profile calls for eventually offering ancillary services such as psychosocial interventions. For those who are employed, it should be noted that the stigma around OUD can impede treatment uptake, due to fear of professional consequences [65]. To mitigate these concerns among those whose work schedule is incompatible with regular supervised dosing at the pharmacy or clinical appointments, flexibility of treatment rules and take-home doses should be considered [66-68]. Furthermore, buprenorphine implants or injections [69] could reduce obstacles to treatment. Additionally, higher odds of professional integration for this profile suggest that vocational counselling services should also be prioritised to support employment.
Lastly, a third profile is characterised by pharmaceutical-type opioid and cannabis use, lifetime anxiety and depressive symptoms, and chronic pain at admission to OAT. Individuals with this profile are more likely to be 45 years of age or older. This profile is similar to the subgroup of individuals who have developed a problem with opioid use in the context of pain prescriptions, as identified by Nielsen et al [35]. Several studies highlight a similar subgroup of pharmaceutical-type opioid users [32, 34, 53, 56], who present lower risk for using an injection mode of administration [34, 56], lower risk for other illicit substance use [52, 56] and higher risk for mental health or chronic pain comorbidities [32]. With low odds of injection and heroin use, it is likely that individuals in this profile self-medicate for psychological or physical pain. It can also be hypothesised that for many of them, prescribed opioid medications contributed to the development of an OUD. The results pertaining to the third profile highlight the importance of the continuum of care between mental health, chronic pain and addiction services. While integrated treatments that combine psychosocial, educational and psychiatric components have become the ‘gold standard’ [70] in the last decades, the continuum of care between pain and addiction is still fragmented [71]. In fact, reports of undertreated OUD in chronic pain patients [72] and undertreated pain issues in OUD patients [6, 73-75] can be found in the literature. To address this gap, some current guidelines touch upon the management of co-occurring chronic pain and OUD. They recommend the use of an approach that combines pharmacological and non-pharmacological strategies within a multidisciplinary management framework that may include a primary care physician, an addiction physician and a pain physician [76]. Additionally, for patients in OAT who have chronic pain, split doses [18, 76, 77] as well as non-opioid analgesics [77] should be considered. When it comes to organising care settings to address co-occurring pain and OUD, several authors favour primary care management to meet the needs of this specific clientele [78] and simultaneously address comorbidities [78, 79]. However, addressing this co-occurring condition remains difficult for many clinicians who mention that it is time- and energy-consuming [71], that there is a lack of evidence-based treatment options [71, 80] and that non-pharmacological resources are not available [71]. From the perspective of people who present co-occurring chronic pain and OUD, a qualitative study revealed that some patients are afraid that their pain will not be adequately managed and that they will be stigmatised if they receive treatments in facilities designed for people with heroin use disorders [81]. This example highlights the importance of destigmatising opioid use for all people regardless of their profile, as well as the relevance of a tailored service delivery model for this third profile of patients who exhibit low risk in terms of overdoses and injection complications, but high odds of psychological and physical pain. Moreover, older age in this profile calls for reflecting on treatment obstacles for those with reduced mobility, cognitive impairment and lowered tolerance to opioid agonists. In fact, the Canadian Guidelines on Opioid Use Disorder Among Older Adults recommend, for example, careful monitoring, reduced initial doses, slow dose escalation and easier access to OUD residential or hospital care for older adults with social, psychological or physical comorbidities [82].
Overall, the results of this study suggest the existence of three distinct profiles of people with OUD in a sample of patients with OUD entering a large Montréal-based OAT facility. Several authors have suggested rethinking services to meet the needs of distinct profiles [32, 35-38, 83, 84]. Indeed, in the field of health services, a profile-based approach allows health care to be tailored to subgroups of patients with differing needs or abilities [36]. With this in mind, it should be noted that in the field of OUD treatment, differential care has emerged during the last decades, mostly in the form of low- and regular-threshold services. While low- and regular-threshold services may be well suited for the first and second identified profiles respectively, there may be a need to adjust the service offering for the specific characteristics exhibited by class 3 patients.
5 LIMITATIONS OF THE STUDY
The present study has certain limitations inherent to the design of the chart review study, such as missing data and self-reported medical and psychosocial history. More specifically, some information is not systematically entered in the patient charts by the clinicians, with parts of the admission questionnaire sometimes left blank. Similarly, some information might have been intentionally or unintentionally left out by patients during assessments. Additionally, the chart review design inherently does not allow for the use of standardised or time-specific measures. However, it should be noted that the systematic use of an adapted version of the Addiction Severity Index by the clinical team counterbalances this limit, improving validity. As for the data extraction process, measures were implemented to improve reliability. Notably, a detailed standardised extraction form was used, the coders always worked in pairs, and the dyads always included one coder with experience in chart reviews. In terms of transferability [85], the study is focused on a sample of patients at the time of admission to treatment, which excludes profiles of people who are not seeking treatment or those who were unable to access it. Nonetheless, the study was conducted in a specialised OAT centre which serves patients from all over the city of Montréal and beyond, and which offers high and low threshold services. In this regard, the sample includes patients who are often left out of services in areas where low threshold treatment is not available. Additionally, the retrospective nature of the study impedes the acknowledgment of the ever-evolving context. For instance, heroin has become much less available since the beginning of the pandemic, which is a phenomenon that could not be taken into account in the present study [86]. Finally, while this study supports the development and assessment of a profile-based approach to care by fostering a better understanding of distinct patients' characteristics, it does not provide a qualitative understanding of the associated needs from the patients' perspective, an important aspect that future research should examine.
6 CONCLUSIONS
The results of this study add to a scarce body of literature on different profiles of people with OUD by highlighting three distinct profiles of people with OUD at the time of admission to OAT in a Montréal specialised facility. Results support further exploring of profile-based approaches to care, allowing health care to be tailored to subgroups of patients with differing needs or abilities [75]. Future work should focus on validating these results in other settings (e.g., primary care and non-urban areas). Additionally, in order to further contribute to the adaptation of services for these different profiles, the use of qualitative methodologies would be valuable in order to gain a better understanding of these profiles, their distinct needs, as well as individual trajectories of opioid use and service experiences.
AUTHOR CONTRIBUTIONS
Study design: Léonie Archambault, Didier Jutras-Aswad, Karine Bertrand, Michel Perreault. Data collection: Léonie Archambault. Data analysis: Léonie Archambault, El Hadj Touré. Lead in writing: Léonie Archambault. Critical feedback and contribution to the manuscript: Didier Jutras-Aswad, Eva Monson, El Hadj Touré, Karine Bertrand, Michel Perreault.
ACKNOWLEDGMENTS
The authors wish to thank Niamh Power and Salima Belhouari for their contribution to the data collection, as well as Adriana Gentile for editing and formatting the manuscript and Diana Milton for administrative support. They also thank the Programme Cran's staff and management for facilitating data collection. This work was supported by Health Canada and the Ministère de la Santé et des Services sociaux du Québec and the Douglas Mental Health University Institute. The views expressed do not necessarily represent those of Health Canada and the Ministère de la Santé et des Services sociaux du Québec. Each author certifies that their contribution to this work meets the standards of the International Committee of Medical Journal Editors.
CONFLICT OF INTEREST STATEMENT
No competing interests to declare.
