Socio‐economic differences in the effectiveness of pharmacotherapy use in alcohol use disorder: A cohort study of 148 626 individuals in Sweden

Abstract

Aim

To examine the socioeconomic differences in the effectiveness of alcohol use disorders (AUD) pharmacotherapy and risk of AUD hospitalisation.

Design

A prospective register‐based cohort study.

Setting

Sweden.

Participants

Individuals who were registered as living in Sweden in 2005 (16–64 years) with a first‐time AUD diagnosis and complete information on their socioeconomic position (SEP) between 2005 and 2019 (n = 148 626).

Measurement

The outcome was AUD hospitalisation. The use of AUD pharmacotherapy was treated as a time‐varying exposure. SEP was the moderator. The association between the joint‐exposure (pharmacotherapy use and SEP) and AUD hospitalisation was assessed using a competing‐risk regression model, adjusted for sociodemographic factors, previous mental health diagnoses and use of other psychiatric medications.

Findings

Pharmacotherapy use was associated with a lower risk of AUD hospitalisation among high SEP individuals [subdistribution hazard ratio (SHR) = 0.83, 95% confidence interval (CI) = 0.77–0.90], but not among those with low SEP (SHR = 1.02, 95% CI = 0.94–1.10) and middle SEP (SHR = 1.01, 95% CI = 0.94–1.09), compared with low SEP individuals when not using pharmacotherapy.

Conclusions

In Sweden, alcohol use disorder (AUD) pharmacotherapy appears to be effective to reduce the risk of AUD hospitalisation only among individuals of high socioeconomic position.

INTRODUCTION

Alcohol use disorder (AUD) is a major public health problem contributing to approximately 5% of the global mortality rate [1]. The lifetime prevalence of AUD that requires an in‐patient stay was estimated at around 2% in 2018 in a register‐based Swedish study [2]. From a public health perspective, AUD is unequally distributed among the population [3, 4, 5]. Individuals in low socio‐economic position (SEP) generally experience worse health outcomes (e.g. with co‐occurring mental health conditions [6], higher risk of hospitalisation [7] and higher mortality rate [3, 8]) than those with higher SEP, despite similar levels of alcohol consumption [3, 4, 5, 9, 10]. Alcohol‐related costs were valued at over 100 billion Swedish kronor annually in 2020 [11], including injuries [12], sickness presenteeism [13] and sickness absence [14]. Despite the burden and costs of AUD, it remains greatly undertreated [15], especially among individuals of low SEP [5].

In the Swedish context, four AUD pharmacotherapies (acamprosate, disulfiram, nalmefene and naltrexone) have been approved for use [16, 17]. These AUD pharmacotherapies have been shown to be efficacious [18, 19] and moderately effective in treating AUD [20]. However, AUD pharmacotherapies remain underutilised compared with psychosocial treatment alternatives, such as cognitive behavioural therapy, motivational enhancement therapy and the Twelve Steps programme [18, 21]. AUD pharmacotherapies have been shown to reduce total alcohol consumption, prevent relapse to heavy drinking and/or maintain abstinence, which are associated with a lower risk of hospitalisation [19, 22]. For instance, the use of naltrexone as monotherapy or in combination with disulfiram or acamprosate was shown to be effective in reducing the risk of hospitalisation in a Swedish study [20]. This study did not, however, explore socio‐economic differences, which may be influenced by a variety of treatment barriers. These barriers include stigma surrounding the diagnosis, a lack of confidence in treatment centres, a lack of access and insufficient knowledge about treatment alternatives, and the low availability of non‐abstinence‐oriented treatments [23, 24, 25, 26, 27].

Differences in prescription rates and medication purchase have also been observed in earlier studies [9, 28]. Pharmacotherapies are prescribed to individuals in high SEP to a greater extent than to those in lower SEP [9, 28]. One explanation may be the hypothetical higher levels of health literacy among individuals in high SEP, compared with individuals in low SEP, and access to various treatment alternatives [28, 29, 30, 31]. As a result, positive health outcomes, such as the reduction in the risk of hospitalisation, may be observed to a greater extent among individuals in high SEP, compared with those in lower SEP. Estimating the use of AUD pharmacotherapy across SEP levels may provide additional insights regarding patterns of pharmacotherapy use, and potentially explains the variations in vulnerability toward AUD‐related consequences, such as differences in AUD hospitalisation. Moreover, the possible joint effects of the use of AUD pharmacotherapy and SEP when examining the association with AUD hospitalisation need to be explored. We hypothesise that, among individuals with an AUD diagnosis, the combined effects of AUD pharmacotherapy and SEP levels on the risk of hospitalisation for AUD is greater than the sum of the effects of AUD pharmacotherapy use and SEP separately. Thus, we aim to investigate the socio‐economic differences on the effectiveness of pharmacotherapy use on the risk of AUD hospitalisation.

METHODS

Study design and participants

This prospective cohort study was conducted in Sweden using the Swedish Work, Illness and Labour market Participation (SWIP) cohort data from 2005 to 2020. Briefly, the SWIP cohort consisted of individuals (16–64 years, born 1941–1989) who were registered as living in Sweden in 2005. The SWIP cohort is a linkage of nationwide registers, such as the National Patient Register (NPR, including in‐patient and specialised outpatient care), the National Prescribed Drug Register (PDR), the Cause of Death Register, and the Longitudinal Integrated Database for Health Insurance and Labour Market Statistics (LISA). More detailed information on the SWIP cohort has been described previously [32, 33].

Figure 1 illustrates the inclusion of study participants in the analytical sample. Individuals were included in the study sample if they received a first‐time AUD diagnosis between 2005 and 2019 and had complete SEP information. We were able to historically observe all alcohol‐related diagnoses from 1997 onwards in the NPR. As the PDR was established in July 2005 [34], we excluded individuals who received a first‐time AUD diagnosis prior to the year 2005 to minimise the risk of misclassification of pharmacotherapy use. The last baseline year was chosen to allow for a follow‐up of at least 1 year.

FIGURE 1.

Flow chart for the selection process of the analytical sample. AUD, alcohol use disorder; SEP, socio‐economic position; SWIP, Swedish Work, Illness and Labour market Participation cohort.

First‐time AUD diagnosis was defined using the Swedish index of alcohol‐related health problems, which was based on the Swedish version of the International Classification of Diseases and Related Health Problems, 10th revision (ICD‐10). The following diagnoses were included: mental and behavioural disorders due to use of alcohol (F10), alcoholic liver disease (K70), toxic effects of alcohol (T51) and other alcohol‐related diagnoses (E24.2, G31.2, G62.1, G72.1, I42.6, K29.2, K85.2, K86.0, O35.4, R78.0, Z04.0, Z71.4 and Z72.1) [2, 35]. The complete list of AUD diagnoses can be found in Table S1.

The use of the SWIP cohort has been approved by the Regional Ethics Board of Stockholm (ref.: 2017/1224–31; 2018/1675–32; 2022–02725‐02). The study protocol was not pre‐registered in an open access platform and should be considered as exploratory.

Measures

Outcome: AUD hospitalisation

The outcome of AUD hospitalisation was defined according to the same ICD‐10 codes mentioned above and were measured between 2005 and 2020. Using in‐patient data held in the NPR, we defined AUD hospitalisation as hospitalisation for a principal or contributing diagnosis related to alcohol with an in‐patient stay of at least 24 hours after a first‐time AUD diagnosis.

Exposure: pharmacotherapy use

The use of AUD pharmacotherapy was treated as a time‐varying exposure [36]. Use and non‐use periods of AUD pharmacotherapy were estimated based on the dispensation of medication extracted from the PDR, following the Anatomical Therapeutic Chemical (ATC) classifications: acamprosate (N07BB03), disulfiram (N07BB01), nalmefene (N07BB05) and naltrexone (N07BB04) [37]. Periods of pharmacotherapy use were computed based on treatment episodes using the AdhereR package in RStudio [38]. Treatment episodes were estimated using the ‘tablet‐per‐day’ method, which calculated the duration of use based on purchase dates, assumed dose per day (number of tablets per day) and the maximum allowed gap between medications (grace period) of 30 days. The grace period was defined a priori after discussions between the investigators (D.L.E. and H.T.). Dose per day for the non‐tablet medication was estimated based on the frequency of dose administration and time interval between two doses. The short grace period was defined to minimise the misclassification of individuals who discontinued treatment. Because of the low proportion of the population getting an AUD prescription, the time‐varying use of medication was dichotomised into two categories (no pharmacotherapy use; pharmacotherapy use) after the treatment episodes had been estimated. The use and non‐use periods of AUD pharmacotherapy are illustrated in Figure S1.

Moderator: socio‐economic position

Education level at first‐time AUD diagnosis was chosen as the measure for SEP, as it captures long‐term SEP and is not sensitive to reverse causality [10, 39, 40]. We used the highest educational attainment obtained from LISA, which was categorised into three categories: low SEP (primary education, ≤9 years); middle SEP (secondary education, 10–12 years); and high SEP (tertiary education, >12 years).

Covariates

Socio‐demographic characteristics (sex, age, country of origin) at the individual level were collected at first‐time diagnosis from LISA. Information on previous mental health diagnoses was gathered from the NPR based on the ICD‐10 codes at baseline: schizophrenia‐spectrum disorders (F20–F29), affective disorders (F30–F39), stress‐related disorders (F40–F48), personality and behavioural disorders (F60–F69), hyperkinetic disorders (F90) and self‐harm (X60, X70, X80–X84). To capture symptoms of comorbid psychiatric disorders that did not require hospitalisation and/or specialised outpatient care, several medications were extracted from the PDR, including: mood stabilisers (ATC: N03AF01, N03AG01, N03AX09); antipsychotics, hypnotics and sedatives (ATC: N05); and antidepressants (ATC: N06A). Information regarding high‐cost protection of AUD pharmacotherapy was collected from the PDR, which refers to governmental subsidies when the annual cost of any type of prescribed medication has been reached, which subsequently increases the accessibility of AUD pharmacotherapies. Dispensation of these medications was continuously updated during the follow‐up period and dichotomised as using or not using other psychiatric medications. A detailed description of all covariates can be found in Table S2.

Statistical analyses

The effectiveness of AUD pharmacotherapy on the risk of AUD hospitalisation was fitted using a competing risk regression model to obtain the subdistribution hazard ratio (SHR) with 95% CIs. All individuals, regardless of variations in exposure, contributed to the model. Non‐AUD hospitalisation and all‐cause mortality were regarded as competing risks. The crude model was adjusted for socio‐demographic factors (sex, age) and region of birth (Model 1). The model was further adjusted for high‐cost protection of pharmacotherapy, previous mental health diagnoses, use of other psychiatric medications and time since first diagnosis (Model 2). Follow‐up time started on the date when the individuals received their first AUD diagnosis (1 January 2005 at the earliest) and ended at first hospitalisation, death, emigration or the end of data linkage (31 December 2020), whichever came first.

To examine socio‐economic differences in AUD hospitalisation, we tested the additive interactions between using AUD pharmacotherapy and SEP. Additive interactions were achieved by creating a joint‐exposure variable to obtain a single comparison group [41]. From a public health perspective, we focussed on the additive interaction to understand whether the use of any pharmacotherapies would be more effective among those in high SEP than among those in low SEP. Comparison between the highest and lowest SEP level was conducted to contrast the potential differences across socio‐economic strata. We used the joint exposure of ‘no pharmacotherapy use’ and ‘low SEP’ as the reference group. The relative risk owing to interactions (RERI = SHR₁₁ – SHR₁₀ – SHR₀₁ + 1) and the 95% CI were calculated using the Delta method [41]. The attributable proportion owing to the interaction (AP = RERI/SHR₁₁) was also calculated.

Sensitivity analyses

Several sensitivity analyses were performed. To examine whether the effects of AUD pharmacotherapies on AUD‐related hospitalisation would differ across SEP level, multiplicative interaction was performed by including an interaction term of AUD pharmacotherapy use and SEP. Within‐individual analysis stratified by SEP for AUD hospitalisation outcome was performed because hospitalisation may occur several times in a lifetime. Here, each person acts as their own control to remove any time‐constant confounding by those whose risk is constant over time, and to identify which SEP groups are more vulnerable to hospitealisation owing to the effects of AUD pharmacotherapy use (as visualised in Figure S2). Only those with variations in both the outcome and the exposure contributed to the model (n = 45 573), and the follow‐up time was reset to zero after each hospitalisation outcome [20].

All statistical analyses were performed using the AdhereR package in RStudio 4.3 [38] and Stata 17 (Stata Corporation, College Station, TX, USA).

RESULTS

Table 1 summarises the description of individuals with an AUD diagnosis between 2005 and 2019 in the SWIP cohort. A total of 148 626 individuals with a first‐time AUD diagnosis were identified during 2005–2019. Across SEP, the majority were males (low SEP, 71.8%; middle SEP, 68.2%; high SEP, 61.2%) and were Swedish born (low SEP, 86.0%; middle SEP, 88.0%; high SEP, 87.1%). A higher proportion of individuals in high SEP had AUD pharmacotherapy (52.5%) compared with individuals in low SEP (41.5%). A larger proportion of individuals in high SEP (13.1%) have received a previous mental health diagnosis compared with individuals in middle (12.2%) or low (10.2%) SEP. A similar proportion of individuals using other psychiatric medications after cohort entry was observed across SEP levels. Table S3 shows the number of hospitalisations, both AUD and non‐AUD related, by use of pharmacotherapy and SEP level.

TABLE 1.

Baseline characteristics of individuals with a first‐time diagnosis of alcohol use disorder (AUD) in the SWIP cohort (n = 148 626) during 2005–2019, stratified by socio‐economic position (SEP).

	Low SEP		Middle SEP		High SEP
	(n = 42 132)		(n = 76 192)		(n = 30 302)
	n	%	n	%	n	%
Sex
Male	30 234	71.8	51 959	68.2	18 550	61.2
Female	11 898	28.2	24 233	31.8	11 752	38.8
Age, mean (SD)	49.3	18.6	47.9	15.9	49.9	15.8
Region of origin
Sweden	36 236	86.0	67 060	88.0	26 379	87.1
Outside of Sweden	5896	14.0	9132	12.0	3923	12.9
Use of AUD pharmacotherapy at any time after cohort entry
No	50 367	58.5	93 804	53.1	37 836	47.5
Yes	35 764	41.5	82 968	46.9	41 812	52.5
Previous mental health diagnosis
No	37 840	89.8	66 897	87.8	26 330	86.9
Yes	4292	10.2	9295	12.2	3972	13.1
Use of other psychiatric medications at any time after cohort entry
No	58 507	67.9	115 783	65.5	49 808	62.5
Yes	27 728	32.2	60 991	34.5	29 852	37.5
Year of cohort entry
2005	4124	9.8	5036	6.6	1538	5.1
2006	4075	9.7	5619	7.4	1815	6.0
2007	3786	9.0	5914	7.8	1933	6.4
2008	3519	8.4	6176	8.1	2020	6.7
2009	3186	7.6	5875	7.7	2042	6.7
2010	3186	7.6	5684	7.5	2157	7.1
2011	3057	7.3	5710	7.5	2170	7.2
2012	2815	6.7	5481	7.2	2148	7.1
2013	2554	6.1	5180	6.8	2334	7.7
2014	2270	5.4	4627	6.1	2142	7.1
2015	2152	5.1	4519	5.9	2033	6.7
2016	2096	5.0	4406	5.8	2163	7.1
2017	1829	4.3	4186	5.5	2051	6.8
2018	1843	4.4	4023	5.3	1997	6.6
2019	1640	3.9	3756	4.9	1759	5.8

Table 2 shows socio‐economic differences of pharmacotherapy use on the SHR for AUD hospitalisation after accounting for competing risks of other causes of hospitalisation and mortality (median follow‐up = 7.7 years; interquartile range, IQR = 3.5–11.7 years). In the unadjusted model, the use of pharmacotherapy was associated with a higher SHR for AUD hospitalisation among individuals in low SEP (SHR = 1.35, 95% CI = 1.29–1.41) and in middle SEP (SHR = 1.26, 95% CI = 1.22–1.31), but a lower SHR for hospitalisation among individuals in high SEP (SHR = 0.81, 95% CI = 0.77–0.85), compared with individuals in low SEP who did not use any pharmacotherapy. In the fully adjusted model, the associations between the joint exposure and AUD hospitalisation were attenuated (low SEP, SHR = 1.02, 95% CI = 0.94–1.10; middle SEP, SHR = 1.01, 95% CI = 0.94–1.09). Among individuals in high SEP, pharmacotherapy use was associated with a lower SHR of AUD hospitalisation in the fully adjusted model (SHR = 0.83, 95% CI = 0.77–0.90), compared with individuals in low SEP who did not use pharmacotherapy. Our RERI calculation, comparing individuals in high and low SEP, suggested that there may be some additive interaction between SEP and pharmacotherapy use (RERI = –0.08, 95% CI = –0.15, –0.01), with an AP of −0.10. This implies that approximately 10% of the reduction in the risk of AUD hospitalisation may be attributed to the interaction itself.

TABLE 2.

Competing risk regression model of the joint exposure between socio‐economic position (SEP) and use of pharmacotherapy for alcohol use disorder (AUD) on the outcome of alcohol‐related hospitalisation; SHR, subdistribution hazard ratio.

Crude	Low SEP			Middle SEP			High SEP			Test of interaction
	SHR	95% CI		SHR	95% CI		SHR	95% CI		RERI a	95% CI		AP b
No pharmacotherapy use	1.00	–	–	1.10	1.08	1.13	0.98	0.95	1.01	–	–	–	–
Pharmacotherapy use	1.35	1.29	1.41	1.26	1.22	1.31	0.81	0.77	0.85	−0.52	−0.60	−0.44	−0.64
Model 1 c
No pharmacotherapy use	1.00	–	–	1.10	1.07	1.12	0.99	0.96	1.02	–	–	–	–
Pharmacotherapy use	1.34	1.28	1.40	1.26	1.22	1.30	0.83	0.79	0.87	−0.51	−0.58	−0.43	−0.61
Model 2 d
No pharmacotherapy use	1.00	–	–	0.99	0.97	1.02	0.89	0.86	0.92	–	–	–	–
Pharmacotherapy use	1.02	0.94	1.10	1.01	0.94	1.09	0.83	0.77	0.90	−0.08	−0.15	−0.01	−0.10

^a Relative risk owing to interaction (RERI) = SHR₁₁ – SHR₁₀ – SHR₀₁ + 1. RERI calculated between high SEP and low SEP.

^b Attributable proportion (AP) = RERI/SHR₁₁.

^c Model 1 was adjusted for sex, age and region of birth.

^d Model 2 was adjusted for sex, age, region of birth, high‐cost protection of AUD pharmacotherapy, previous mental health diagnoses and use of other psychiatric medications.

Sensitivity analyses

The results from all sensitivity analyses can be found in the supporting information. In the multiplicative interaction analysis, using AUD pharmacotherapy was associated with an overall lower SHR for AUD hospitalisation but not among individuals in low SEP (Table S4). The within‐individual analysis stratified by SEP showed that using AUD pharmacotherapy was associated with an overall lower SHR for AUD hospitalisation, but that these differences were minimal (Table S5).

DISCUSSION

This study aimed to investigate socio‐economic differences in the effectiveness of pharmacotherapy use on the risk of AUD hospitalisation. We found that the use of pharmacotherapy was associated with a lower risk of hospitalisation for AUD only among individuals in high SEP during the follow‐up period (2005–2020).

In the current study, even though AUD pharmacotherapy was used by approximately half of our study population, it was higher among individuals in high SEP compared with those in low and middle SEP, a result in concordance with previous findings [9, 28]. Individuals in high SEP may be likely to demonstrate more frequent treatment‐seeking behaviours compared with those in lower SEP [28], for example through private clinics for AUD. This may increase the likelihood of individuals in high SEP having better access to a variety of treatment alternatives, such as combining psychosocial treatments and pharmacotherapy for AUD. Access to various treatment alternatives could also reflect levels of integration into the healthcare system across SEP groups [9], subsequently contributing to the overall health status [28]. Another explanation is that AUD is highly stigmatised [27], perhaps more so among individuals in low SEP. This could lead to delayed treatment, which may further exacerbate alcohol‐attributable consequences among individuals in low SEP, who are already at a higher risk of alcohol‐related consequences [4, 7, 8, 39]. Differences in the receipt and subsequent dispensation of AUD pharmacotherapy could explain the differential risk for AUD hospitalisation across SEP levels [9, 28].

We only found a reduction in the risk of AUD hospitalisation among individuals in high SEP who used pharmacotherapy. The attenuation of AUD hospitalisation risk among individuals in low and middle SEP in the fully adjusted model appears to be confounded by other comorbidities. Individuals in lower socio‐economic strata who had multiple health issues may be more likely to seek treatment for their other health issues that coincide with their AUD diagnosis. For example, as individuals in lower SEP have worse health compared with individuals in high SEP, delaying treatments for AUD may exacerbate their overall health status or lead to situations where AUD pharmacotherapies are contraindicated [42]. Subsequently, pharmacotherapy may be used as a first‐line treatment and thus result in worse outcomes. Another explanation is that psychosocial treatments are one of the mainstays of AUD treatment [18]. The differences in AUD pharmacotherapy outcomes among individuals in low, middle and high SEP may hypothetically be explained by the level of health literacy in these populations [29, 30]. Individuals in high SEP may have better access to information regarding available treatments or are able to advocate for their own needs [31]. Relative to individuals in low SEP, those in high SEP are also more likely to receive treatment from specialists [31]. The healthcare system is publicly funded in Sweden, but variations may still exist across socio‐economic strata, as suggested by the negative RERI and AP estimates found in this study. This may be explained by variations in the prescribing practices or dispensation of prescribed medications [9, 28]. As a result, this could lead to differences in health outcomes. Therefore, receiving an AUD pharmacotherapy prescription may indicate a more integrative treatment strategy, such as in combination with other types of treatment [18], rather than pharmacotherapies being more effective in individuals of high SEP, compared with individuals of low SEP.

Strengths and limitations

The large cohort study design along with the long follow‐up time and the use of register‐based information were considered major strengths of this study. The nationwide coverage of the registers allowed us to include all registered individuals in Sweden, which lowered the risk of selection bias. We were also able to follow all the included individuals over time with a well‐defined outcome based on the ICD‐10 codes. Our results could therefore be generalisable to other countries with similar healthcare systems to that in Sweden.

This study used education at first‐time AUD diagnosis as the main SEP measure, because it is not sensitive to reverse causality [39, 40]. Other measures of SEP, such as income and occupation were not considered as a good measure in the present study, as individuals are likely to transition in and out of work, influencing their level of income [39]. Using occupation as a measure of SEP in this study may also have biased the estimate owing to the high risk of missing information, given its complex relationship with AUD [39, 40]. However, because we only included those with complete SEP information, there is a risk of selection bias, as missing SEP information was more common among non‐Swedish‐born individuals. Our within‐individual analysis stratified by SEP, however, showed that pharmacotherapy use was associated with a lower risk of hospitalisation across all SEP levels. This suggests that the differences observed in our main analyses were likely influenced by structural‐level factors, such as the influence of geographical barriers on access to treatment facilities or differences in treatment provision that may affect our outcomes of interest differently [23, 26, 28].

The use of the PDR could also introduce a risk of misclassification of exposure. We were only able to assume that those who at some point during the follow‐up period redeemed their prescription continued their treatments as prescribed, rather than measuring the actual adherence to treatment. In the current study, we only limited AUD pharmacotherapy to acamprosate, disulfiram, nalmefene and naltrexone. As a result, individuals who used other types of pharmacotherapies to treat AUD, such as gabapentin and topiramate, may have been misclassified as never using any AUD pharmacotherapies during the follow‐up period [17]. Furthermore, our exposure (pharmacotherapy use) was a proxy that was estimated based on dispensed prescriptions, which assumed that individuals adhered to their prescribed medication regimen. We could not ascertain adherence to pharmacotherapy use; however, the tablet‐per‐day method has been shown to provide a good estimation of exposure period for AUD pharmacotherapy [43]. Information on AUD diagnosis was only available from the in‐patient and specialised outpatient registers, where the majority of the sample constitute severe cases of AUD. Finally, the PDR only captured those who received and dispensed their pharmacotherapy prescription. This may lead to missing information on individuals who received direct administration of AUD pharmacotherapy from healthcare centres for substance use disorders, the emergency room, during a hospital stay, or when prescriptions were not picked up from the pharmacy [34].

Other information that may have influenced our findings, such as the use of psychosocial treatments [18] and severity of AUD [44], was not available. The additional information, such as different types of AUD treatments, and intensity and frequency of treatments, could provide some insights into potential explanations for SEP differences regarding the risk of AUD hospitalisation, especially among those in low and middle SEP. As AUD poses a considerable burden to individuals and society, the additional information is unlikely to alter our findings significantly. Nevertheless, future research should seek to incorporate information on other treatment alternatives and examine the effectiveness of AUD pharmacotherapies in combination with these treatment alternatives to gain a comprehensive overview of treatment regimes.

CONCLUSION

Generally, using any type of AUD pharmacotherapy was associated with a lower risk of hospitalisation only among individuals in high SEP. This underscores the importance of variations in treatment access across social strata that lead to disparities in health outcomes. The prescription and use of pharmacotherapy are indicative of a more integrative treatment regimen in Sweden. Potential differences in the use of psychosocial treatments for AUD by itself or in combination with pharmacotherapy across socio‐economic strata should be explored further.

AUTHOR CONTRIBUTIONS

Devy L. Elling: Conceptualization (equal); data curation (lead); formal analysis (lead); funding acquisition (equal); methodology (lead); project administration (lead); writing—original draft (lead); writing—review and editing (lead). Emelie Thern: Conceptualization (equal); data curation (supporting); formal analysis (supporting); funding acquisition (equal); methodology (supporting); project administration (supporting); writing—review and editing (supporting). Lluís Mangot‐Sala: Data curation (supporting); formal analysis (supporting); writing—review and editing (supporting). Jari Tiihonen: Conceptualization (supporting); funding acquisition (supporting); methodology (supporting); writing—review and editing (supporting). Anders Hammarberg: Funding acquisition (supporting); writing—review and editing (supporting). Daniel Falkstedt: Funding acquisition (supporting); writing—review and editing (supporting). Heidi Taipale: Conceptualization (supporting); formal analysis (supporting); funding acquisition (supporting); methodology (supporting); supervision (supporting); writing—review and editing (supporting).

DECLARATION OF INTERESTS

J.T. has participated in research projects funded by grants from Janssen‐Cilag to his employing institution; he has served as a consultant to HealthCare Global Village, HLS Therapeutics, Janssen, Lundbeck, Orion Pharma, Teva and WebMD Global; given an expert testimony to Janssen; has received honoraria from Janssen, Lundbeck, Orion Pharma, Otsuka and Teva; and has received support for attending meetings and/or travel from Teva, all outside this work. H.T. has received lecture fees from Gedeon Richter, Janssen, Lundbeck and Otsuka, outside this work and outside the topic of this work. All other authors declare no competing interests.

Supporting information

Figure S1. Illustration of the study design during the follow‐up period. The main exposure, the use of pharmacotherapy for alcohol use disorder (AUD), was treated as a time‐varying exposure. Individuals are followed from first‐time AUD diagnosis (indicated by a blue cross) and censored owing to the outcomes of interest, competing risks or emigration (indicated by a red cross).

Figure S2. Illustration of the within‐individual design, where individuals acted as their own strata owing to hospitalisation being a recurring event. Time is reset after each outcome event (AUD hospitalisation). The exposure is a time‐varying exposure, where ‘exposed’ is using AUD pharmacotherapy and ‘unexposed’ is not using AUD pharmacotherapy. When time resets, the time periods from the same individual (i.e. stratum) are used in comparison the same way as different individuals are used in a between‐individual model. More detail on time resetting is described in Allison, P.D. (2009). Fixed effects regression models, SAGE publications. AUD, alcohol use disorder; Exp, exposed; Unexp, unexposed.

Table S1. The Swedish index of alcohol‐related health problems based on the Swedish ICD‐10 codes. Translated from Swedish (Sweden’s National Board and Welfare, 2016).

Table S2. Description of covariates, including the Anatomical Therapeutic Chemical (ATC) classification codes and the International Classification of Diseases and Health‐related problems, 10th revision (ICD‐10) codes.

Table S3. The number of events for exposure and hospitalisation outcome by socio‐economic position (SEP).

Table S4. Competing risk regression model of the multiplicative interaction between socio‐economic position (SEP) and use of pharmacotherapy for alcohol use disorder (AUD) on the outcome of alcohol‐related hospitalisation. Subdistribution hazard ratios (SHRs) and 95% confidence intervals (95% CIs).

Table S5. Within‐individual competing risk regression model of pharmacotherapy use for alcohol use disorder (AUD) on the risk of AUD hospitalisation. Stratified by socio‐economic position (SEP). Subdistribution hazard ratios (SHRs) and 95% confidence interval (95% CIs) (n = 45 573).

Elling DL, Thern E, Mangot‐Sala L, Tiihonen J, Hammarberg A, Falkstedt D, et al. Socio‐economic differences in the effectiveness of pharmacotherapy use in alcohol use disorder: A cohort study of 148 626 individuals in Sweden. Addiction. 2026;121(3):597–605. 10.1111/add.70238

DATA AVAILABILITY STATEMENT

The data that support the findings of this study are available from Statistics Sweden. Restrictions apply to the availability of these data, which were used under license for this study. Data may be available from the corresponding author with the permission of Statistics Sweden.