A 30-year follow-up of substance misusers in Sweden – differences in predictors of mortality between women and men

Abstract

Background: Differing results on gender specific factors related to mortality risks among substance misusers highlights the need for further research. The present article is based on a 30-year follow-up study on substance misusers treated in residential care for drug problems in Sweden in 1982-1983 aiming to identify and compare gender differences in predictors of mortality.

Method: Original data consists of personal interviews with 1163 substance misusers treated in inpatient units in Sweden during 1982-1983. The outcome variable is death retrieved from the National Cause of Death Register held by the National Board of Health and Welfare. Gender differences and similarities regarding predictors of mortality was estimated in univariate and multivariate models, using Cox proportional hazards models.

Results: School failure, imprisonment and being a parent without custody of the child seem to constitute risk factors for mortality among women, but not among men. A social network of friends seemed to be more important for men. Treatment-dropout was a significant risk factor for premature death among men, but not among women. Both gender reporting alcohol as their self-reported most dominant substance misuse showed higher mortality risks compared with those with stimulants as dominant substance misuse.

Conclusions: Imprisonment was highly predictive of mortality for the women, suggesting that this group is important to pay particular attention to. Suggested differences in the importance of social factors need to be investigated more thoroughly. The substantial hazard revealed for women with polydrug misuse including alcohol calls for attention to this in treatment for substance misuse.

Keywords:

• Alcohol; gender differences; substance misuse; mortality; follow-up; polydrug use

Introduction

Higher standardized mortality rates for women compared to men have been reported in several follow-up studies of drug abusers (Ghodse et al. 1998; Bartu et al. 2003; Bird et al. 2003; Degenhardt et al. 2010; Stenbacka et al. 2010; Mathers et al. 2013; Evans et al. 2015) as well as of alcohol abusers (Roerecke and Rhem 2013). However, other studies report no gender differences in mortality rates among drug abusers (Oppenheimer et al. 1994; Jimenez-Treviño et al. 2011) or alcohol abusers (Rogers et al. 2015). These differing results on gender specific factors related to mortality risks among substance abusers highlights the need for further research.

The present article is based on a 30-year follow-up study on the majority (estimated at 50–75% of substance abusers in Sweden, Olsson 1988) of substance misusers treated in residential care for drug problems in Sweden in 1982–1983. The aim was to identify predictors of mortality and to compare differences in predictors of mortality between women and men in the group studied.

The SWEDATE project

Data used for this study were originally collected by the Swedish Drug Addict Treatment Evaluation (SWEDATE) research project investigating Swedish care for substance misusers in 31 inpatient treatment units (Olsson 1988; Berglund et al. 1991; Bergmark et al. 1994). In the 1980s, when the data were collected, inpatient treatment units in Sweden were not specialized according to substance misuse, but instead admitted and mixed all types of clients. A 30-year follow-up investigating standardized mortality rates (SMR) in the SWEDATE group and identifying the cause of death in different groups based on type of misuse (von Greiff et al. 2018) showed gender differences. The largest differences were found in the groups whose self-reported dominant substance misuse was cannabis (4.3 for women versus 9.9 for men) and opiates (13.5 for women versus 18.3 for men). Standardized mortality rates were higher among men in all the groups studied except for those mainly misusing alcohol (14.4 for women versus 12.8 for men). Drug and alcohol related demise was the most common cause of mortality in the group as a whole, with the highest rates of drug related deaths in the opiate group. There was also a large increase in the proportion of deaths related to drug and alcohol misuse when contributing causes were considered, compared with when only the underlying (primary) cause of death was accounted for.

Previous research on predictors of mortality in substance users

Follow-up studies on mortality have identified both protective and risk factors for early demise. A study from Austria on predictors of mortality showed, for example, that after release from opioid maintenance therapy, patients with more lifetime hospitalization and a poor financial situation were more likely to die, while social relationships were described as protective (Bauer et al. 2008). Criminality (Scott et al. 2011; Ravndal et al. 2015), psychiatric disorders (Bartu et al. 2003; Ravndal et al. 2015) and homelessness (Gossop et al. 2002; Bird et al. 2003) were also stressed as risk factors for mortality. However, a Danish registry study did not show excess mortality in people with psychiatric comorbidity, except for cocaine/amphetamine users. According to the authors, this indicates that these drugs are either particularly harmful to individuals with psychiatric disorders or that the use of stimulants can lead to psychiatric disorders resulting in excess mortality (Arendt et al. 2011).

A protective factor highlighted in several studies is substance abuse treatment (Bartu et al. 2003; Degenhardt et al. 2010; Ravndal et al. 2015). However, results also imply that drug users who receive treatment later on in their addiction career or who spend a greater part of their life in treatment do not benefit from treatment to any great extent (Scott et al. 2011).

When comparing gender differences and mortality among drug abusers, Jimenez-Treviño et al. (2011) found in their 25-year follow-up of patients admitted to methadone treatment that women who survived were more likely to have stopped using heroin than men. Later studies comparing gender differences have found that in patients treated for opioid dependence, employment decreased the mortality risk in men, but not women (Evans et al. 2015). A study of street-recruited injecting drug users reported that sex work increased the mortality risk in women, while an increased mortality risk in men was related to experience of incarceration (Gjersing and Bretteville-Jensen 2014). Also, injury-related mortality has been found to be higher among women in patients with alcohol use disorders (Guitart et al. 2015).

A study that is highly relevant for the present investigation is a Swedish follow-up by Storbjörk and Ullman (2012) looking into mortality, eight years after baseline, in substance misusers starting treatment in Stockholm County between 2000 and 2002 (both alcohol and illicit/prescription drug misuse). No gender differences were found in SMR. Logistic regression showed that being older, male, and having reported living with a substance misuser were identified as risk factors for mortality. The mortality risk did not differ between alcohol and drug-dependent cases, and neither homelessness nor education were predictive of mortality.

General gender differences in research on substance misuse

In a literature review on women and addiction, Tuchman (2010) pointed out significant gender differences in several areas of research on substance abuse, for example, in the epidemiology of substance abuse, social factors and characteristics. Tuchman also concluded that there were gender differences in treatment entry, retention and completion. However, another literature review found evidence that women with substance use disorders were less likely, over a lifetime, to enter treatment compared with men, but once in treatment, gender was not a predictor of either retention or completion (Greenfield 2007). An Italian study of gender differences among heroin addicts in treatment showed that when women first gained access to treatment they were often younger, more frequently unemployed (but better educated), lived more frequently with a partner and/or children and had more severe psychiatric symptoms than men (Vigna-Talianti et al. 2016). Gender differences have been reported regarding help-seeking and reasons for entering treatment (Tucker 2001; Blomqvist and Cristophs 2005; Ravndal et al. 2015). The fact that women have more serious psychosocial problems than men when they decide to enter treatment has been a consistent result in previous research (Chou and Dawson 1994; Hesselbrook and Hesselbrook 1997; Tucker 2001). For women with children who were placed in substitute care, entering treatment quickly and completing treatment were factors that increased the likelihood of being reunified with their children (Green et al. 2006). The importance of having a comprehensive focus in the treatment of parenting women with substance use problems has also been highlighted (Jeong et al. 2015).

Overall, previous research indicates that substance abusers are very often a marginalized group and that there are differences between men and women regarding predictors of mortality.

Methods

This study is based on SWEDATE data that have been linked to the National Cause of Death Register (CDR) managed by the National Board of Health and Welfare. The SWEDATE data come from personal interviews with 1163 of the 1656 substance misusers treated in the period 1982–1983. All treatment units focused on substance misuse, including polydrug misuse, often combined with alcohol misuse.

The interviews were mainly carried out in 1982 and 1983, but eight interviews were conducted in 1981 and 90 in 1984 and 1985. An overall evaluation of the quality of the SWEDATE data, based on both validity and reliability testing, showed a high level of quality (Olsson 1988).

Outcome variable

The outcome variable was death retrieved from the CDR, from the date of intake to the end of 2013. The quality of the CDR was regarded as good and covered all registered deaths in the population of Sweden since 1961, whether occurring within or outside the country (National Board of Health and Welfare 2010). The registers were linked through the use of the individually unique ten-digit personal identity number (PIN) that all Swedish residents have from birth (or date of immigration) to death.

Background variables

Background information was retrieved from the SWEDATE data. In several previous studies, societal Out-of-home care and school failure (Not finished primary school) have proven to be strong risk factors for adverse outcomes later in life (e.g. Berlin et al. 2011) and were both included in this study. The variables Foreign background and Own housing last 12 months before intake were also included at first (shown in Table 1), but were in the end excluded from the analysis due to their weak association with mortality. This exclusion had no (or only a slight) effect on the results. In addition, the interpretation of these variables was uncertain due to changes in the immigrant group, and in the housing situation, in Sweden since the 1980s. Although there are no robust statistics, available studies imply that the proportion of clients with a foreign background has increased, especially with regard to opiate abusers (Armelius and Armelius 2011) and that the proportion of Finnish immigrants that dominate the group with a foreign background in the present study has decreased (Finne 2003). According to the data available, homelessness is a more widespread social problem in Sweden today compared with the situation in the 1980s (Knutagard and Swärd 2006).

Table 1. Distribution of background factors. Numbers (N) and percent (%).

	Men		Women		Total
	N	%	N	%	N	%
Foreign background^a	104	13.0	41	11.4	145	12.5
Out-of-home care before teen^a	286	35.7	99	27.4	385	33.1
Not finished primary school^a	156	19.4	96	26.6	252	21.7
Age at exit from treatment^b (baseline age)
−19 yrs	47	5.9	54	15.0	101	8.7
20–29 yrs	465	58.0	231	64.0	696	59.8
30+ yrs	290	36.2	76	21.0	366	31.5
In life^a
Care for drug and/or alcohol abuse	566	70.6	262	72.6	828	71.2
Psychiatric care and/or suicide attempts	317	39.5	188	52.1	505	43.4
Have had a nonfatal overdose	373	46.5	171	47.4	544	46.8
Previous imprisonment^b
Missing information	40	5.0	31	8.6	71	6.1
Yes	419	52.2	59	16.3	478	41.1
No	343	42.8	271	75.1	614	52.8
Children^b
Missing information	110	13.7	56	15.5	166	14.3
No children	445	55.5	172	47.6	617	53.1
Children without custody	206	25.7	83	23.0	289	24.8
Children with custody	41	5.1	50	13.9	91	7.8
Abuse last 12 months before intake^b
Alcohol	188	23.4	67	18.6	255	21.9
Cannabis	187	23.3	52	14.4	239	20.6
Stimulants	277	34.5	137	38.0	414	35.6
Opiates	105	13.1	77	21.3	182	15.6
Other	45	5.6	28	7.8	73	6.3
Source of income last 12 months before intake^b
Missing information	17	2.1	11	3.0	28	2.4
Disability pension/sick leave/welfare	244	30.4	135	37.4	379	32.6
Crime	369	46.0	98	27.1	467	40.2
Prostitution	4	0.5	52	14.4	56	4.8
Other	168	20.9	65	18.0	233	20.0
Daily contact with friends last 12 months^b
Missing information	35	4.4	24	6.6	59	5.1
No	132	16.5	70	19.4	202	17.4
With non-addicts	67	8.4	22	6.1	89	7.7
With addicts	469	58.5	215	59.6	684	58.8
Both with addicts & non-addicts	99	12.3	30	8.3	129	11.1
Own housing last 12 months before intake^a	308	38.4	164	45.4	472	40.6
Partner at the time of intake^b
Missing information	8	1.0	5	1.4	13	1.1
No	568	70.8	203	56.2	771	66.3
Yes, with addiction	85	10.6	117	32.4	202	17.4
Yes, without addiction	141	17.6	36	10.0	177	15.2
Treatment factors^a
Drop-out incl. ‘thrown out’	367	45.8	177	49.0	544	46.8
Acute/detox intake	160	20.0	82	22.7	242	20.8
Total number	802		361		1163

^a Percent with that specific attribute.

^b Categories sum up to total number (100%).

The ‘in life’ variable In care for drug and/or alcohol misuse was created out of four separate questions on the number of times a participant received hospital care for alcohol and/or drug misuse or institutional care for alcohol or drug misuse. The variable Psychiatric care and/or suicide attempts was created out of two questions: one on psychiatric care for other reasons than abuse, and one on performed suicide attempts. Have had a non-fatal overdose was created out of a question on the number of non-fatal overdoses.

A separate category for missing information was created for individuals with missing values on single variables (due to non-response at interview), which consisted of mutually exclusive categories that added up to 100%. That applied to the following variables (percentage with missing information in brackets): Previous imprisonment (5–6%), Children (14%), Source of income last 12 months before intake (2%), Daily contact with friends last 12 months (4–5%), and Partner at the time of intake (1%) (Table 1). This was done in order to retain as many study subjects as possible in the analysis, since different individuals had missing values on different variables, and thus a relatively large proportion would have been excluded from the study population if the analysis had been limited to complete cases only.

The variable Misuse in last 12 months before intake was created as five mutually exclusive categories: Alcohol, Cannabis, Stimulants, Opiates and Other, based on what the interviewees had told the interviewer was the most dominant misuse substance for that period (for more details on the categories, see von Greiff et al. 2018).

The variable Source of income in last 12 months before intake was created out of two questions on the source of income, the main and contributing source. The response categories were: Employment or unemployment benefit; Study subsidy; Parents or relatives; Sick pay; Disability pension; Social welfare; Crime; Prostitution; and Other.

The variable Daily contact with friends last 12 months before intake was measured using four mutually exclusive categories; no daily contact with friends; daily contact only with addicts; daily contact only with non-addicts; and daily contact with both addicts and non-addicts. The variable Partner at the time of intake was also separated depending on whether the partner had an addiction or not, and a separate category for non-response was added.

Two treatment factors were included in the analysis, derived from questions answered by the staff; Dropout (including being thrown out of treatment) and Acute/detox intake, both as dummy variables. Missing values are included in dropout.

The following variables were included in the multivariable analysis: Out-of-home care, Not finished primary school, Baseline age, In life care for drug and/or alcohol misuse, In life psychiatric care and/or suicide attempts, In life have had a nonfatal overdose, Previous imprisonment, Children, Dominant misuse last 12 months before intake, Source of income last 12 months before intake, Daily contact with friends last 12 months, Partner at time of intake, Drop-out from treatment and Acute/detox intake.

Statistical analysis

Analyses were made using SAS software package. The mortality risk (Table 2) was estimated in univariate and multivariate models, separate for women and for men, using Cox proportional hazards models (PROC PHREG). Left truncated age (person-days) was used as the timescale, starting at age at exit from treatment (baseline age), and ending at age at death or end of follow-up (December 2013). The reason behind the choice of age as the timescale was to avoid bias which might occur when age is associated with the covariates (Canchola et al. 2003; Thiébaut and Bénichou 2004), However, we also ran the analysis with time-on-study (person-days) as the timescale, adjusting for age as a covariate, which had no significant impact on the results (not shown in table).

Table 2. Mortality risks: Cox regression (HR)^a.

	Men				Women
	Univariate		Multivariate		Univariate		Multivariate
	HR	p-value	HR	p-value	HR	p-value	HR	p-value
Out-of -home care before teen^a	1.17	.14	1.09	.47	1.63	<.05	1.11	.71
Not finished primary school^a	1.14	.30	1.02	.90	2.09	<.01	1.58	.11
In life^a
Care for drug and/or alcohol abuse	1.37	<.05	1.16	.33	2.17	<.01	1.50	.27
Psychiatric care and/or suicide attempts	1.26	<.05	1.18	.15	1.44	.11	1.29	.35
Have had a nonfatal overdose	1.52	<.01	1.25	.06	1.60	<.05	1.19	.52
Previous imprisonment
Missing information	0.86	.58	0.76	.54	1.88	.06	2.22	.09
Yes	0.92	.45	0.98	.88	2.07	<.01	2.30	<.01
No	ref.		ref.		ref.		ref.
Children
Missing information	1.43	.19	1.43	.24	2.51	<.05	1.93	.25
No children	1.27	.33	1.24	.43	2.10	.07	1.88	.19
Children without custody	1.33	.27	1.13	.65	2.70	<.05	2.55	<.05
Children with custody	ref.		ref.		ref.		ref.
Abuse last 12 months before intake
Alcohol	1.40	<.05	1.45	<.05	1.77	<.05	2.14	<.05
Cannabis	0.86	.33	1.02	.93	0.48	.14	0.56	.30
Stimulants	ref.		ref.		ref.		ref.
Opiates	2.11	<.01	1.93	<.01	1.68	.06	1.07	.85
Other	1.14	.59	1.02	.95	1.23	.61	0.82	.73
Source of income last 12 months before intake
Disability pension, sick leave or welfare	1.29	.10	1.16	.39	1.25	.51	1.03	.94
Crime	1.38	<.05	1.20	.30	1.44	.30	1.23	.65
Prostitution	0.73	.76	0.41	.38	2.34	<.05	1.61	.34
Other	ref.		ref.		ref.		ref.
Daily contact w ith friends last 12 months
Nor w. addicts or non-addicts	1.56	<.05	1.66	<.05	1.05	.92	0.79	.67
W. n-a. but not w. addicts	1.08	.77	1.24	.47	1.72	.37	2.01	.29
W. addicts but not w. n-a	1.59	<.05	1.72	<.05	1.73	.24	1.20	.71
W. both addicts & non-addicts	ref.		ref.		ref		ref.
Partner at the time of intake
No	1.00	.97	0.73	<.05	1.51	.31	1.42	.44
Yes, with addiction	1.52	<.05	1.07	.75	1.41	.41	1.34	.53
Yes, without addiction	ref.		ref.		ref.		ref.
Treatment factors^a
Drop-out from treatment incl. ‘thrown out’	1.37	<.01	1.43	<.01	1.11	.61	1.08	.76
Acute/detox intake	1.32	<.05	1.54	<.01	1.68	<.05	1.82	.06
Total number in models			724				316

^a Dummy variables.

p-values on 0.05-level or lower are marked in bold.

Ethics

This research project was scrutinized and approved by the Ethical Review Board in Stockholm, Sweden (2015/329-31/5, 2015/1205-32, 2016/542-32/5).

Results

Summarized sample description

The distribution of background variables is provided in Table 1. Age at exit from treatment (baseline age) differed according to gender; a higher proportion of the women were 19 or younger (15 vs. 6%) while the opposite applied to the oldest age group with a higher proportion of men (36 vs. 21%). Just over a quarter of the women and almost one fifth of the men had not finished primary school (27 and 19%, respectively). Further, the majority did not have their own housing in the last 12 months before intake (55% for women and 62 for men). However, difficulties in defining and measuring homelessness combined with the young age of most of the clients limited the possibilities of interpreting this variable.

Overall, the group was characterized by its vulnerable and marginalized situation. One third had been in out-of-home care before their teens (36% men and 27% women). Furthermore, 71% had experienced hospital and/or institutional care for drug or alcohol misuse. Almost half had suffered a non-fatal overdose. While the proportion who had been in prison was higher for men (52 vs. 16%), the opposite applied for suicide attempts and/or psychiatric care for other diagnoses than substance misuse (52 vs. 40%). Just under half had children, but a majority of these did not have custody of their children (62% of the women and 83% of the men).

Stimulants were the most common misuse substance in the last 12 months before intake for both men and women (36%), and while alcohol and cannabis misuse was more frequent among men (23 vs. 19 and 14%, respectively), the abuse of opiates was more common among women (21 vs. 13%).

A third reported welfare assistance (disability pension, sick pay or social assistance/income support) as their main source of income during the 12 months before intake. Crime as main source of income was more frequent among men (46 vs. 27%), while the opposite applied as regards prostitution (14% of the women and 0.5% of the men).

Other aspects of the social network were studied by questions on having a partner (at time of intake) and daily contact with friends (last 12 months before intake). More female clients had a partner with an addiction compared to the men in the group (32 vs. 11%). There were no real gender differences regarding contact with friends other than slightly more men had contact with both addicts and non-addicts (12 vs. 8%). The most common scenario was daily contact only with addicts (59%).

Just over one fifth of the interviews were conducted at an acute or detox unit. Almost half of the group dropped out or were thrown out from treatment.

Gender differences in background characteristics as predictors of mortality

In total, 373 men and 93 women had died by the end of the follow-up (December 2013), which constituted 47% of the men and 26% of the women. The analysis of mortality and its risk factors was carried out using separate models for men and women, both univariate and multivariate (Table 2). In the multivariate models, 45 women (12%) and 78 men (10%) were excluded due to missing information (non-response) on one or several variables.

Women who had not finished primary school had a higher mortality risk, compared with women who had (Hazard Ratio, HR = 2.09 univariate analysis), while no significant difference was found in men. Further, imprisonment was a strong predictor for women (HR = 2.30 multivariate analysis), but not for men. Although non-significant, the higher risk concerning this variable in the group of female non-responders indicates that the non-responder group was skewed. The variable ‘care for drug and/or alcohol abuse’ included both institutional treatment and hospital care. This was a significant predictor for both women and men on a univariate level although this factor was connected to a higher risk for women (HR = 2.17 univariate analysis) than for men (HR = 1.37 univariate analysis). However, no significant risk was found for either gender in the multivariate model. Mental health factors showed a slightly higher hazard ratio on a univariate level for men but did not have a significant effect on mortality in the multivariate model. Earlier non-fatal overdoses implied an excess risk for both genders in the univariate analysis, but this did not remain in the multivariate model.

Women who had children, but without custody of them, had a 2.7 times as high mortality risk as women who were the caregivers of their children while this made no significant difference as concerned the men. The higher risk in the group of female non-responders for this variable indicates that the non-responder group was skewed. For other social living conditions that were tested, the picture was mixed. Neither partner at the time of intake nor income from public support/crime/prostitution during the last twelve months before intake to treatment was significant in the multivariate model. However, the strongest factor in the above-mentioned social living conditions on a univariate level was a more than doubled risk for women who reported prostitution as their main source of income during the last twelve months. The social network had a greater impact on men, where daily contact solely with addicts or no daily contact at all entailed an equal increase in the mortality risk. No significant difference were found for women neither in the univariate nor in the multivariate model.

Men who declared alcohol as their most dominant form of substance misuse had only a slightly higher mortality risk (HR = 1.45 multivariate analysis) compared with those with stimulants as their most dominant form of substance misuse. For women with alcohol as their most dominant form of substance misuse, the risk was more than doubled (HR = 2.13 multivariate analysis). In the opiate group, men had almost twice as high a mortality risk compared with men in the stimulant group, while no significantly higher risk was shown in women.

Compared with men ending their treatment in a planned manner, men that dropped out of treatment had an excess risk of 1.42 HR (multivariate analysis).

In the group where individual interviews were conducted during intake for detoxification on a hospital ward, the excess risk was 1.82 HR for women and 1.54 HR (multivariate analysis) for men, compared with those who were interviewed during their treatment programme. However, this variable was difficult to interpret. Data collected by the SWEDATE project revealed that about two thirds of the 232 individuals on the detoxification wards were transferred to some other treatment facility (such as in or outpatient care, somatic or ‘other’ treatment). More than a quarter in the group that were interviewed during their inpatient treatment (n = 789) had been transferred from detoxification wards. Even if the statistical differences found between the group that was interviewed during treatment and the group that was interviewed on detoxification wards implied that there were certain differences between these two groups, they interacted with each other and only differed concerning where they were staying when the data were collected for the SWEDATE project.

Discussion

The main aim of the present study was to investigate factors influencing mortality risk by gender. Overall, the members of the group in focus were characterized as having a marginalized situation including a low socioeconomic status.

To investigate the variables important for men and women, gender specific analyses were performed. Thus, the gender differences in mortality per se were not in focus. However, the lower mortality rate in the female group in the study compared with the men might imply that the women who died were more marginalized, since death was a more uncommon outcome for the women. The way marginalization was expressed differed between women and men. While it was more common for men to have experienced out-of-home care before their teens as well as prison and not having their own housing, women were much more prone to mental illness and had not completed their primary school education to a greater extent. The gender differences found were that school failure (unfinished primary school), imprisonment, and being a parent without custody of the child seemed to constitute risk factors for mortality among women, but not men. A social network of friends seemed to be of greater importance for men, where having a poor social network or a social network of addicts alone was worse than having daily contact with a network of both addicts and non-addict friends. Dropout from treatment was a significant risk factor for premature death for men, but not for women. Further, male opiate abusers had a higher mortality risk compared with male stimulant abusers, while this comparatively higher risk was not indicated in female opiate abusers. Both men and women who reported alcohol as their self-reported most dominant form of substance misuse showed higher mortality risks compared with those with stimulants as their main substance. Before relating and discussing the results in relation to previous research, it should be noted that comparisons with other studies require caution because of the use of different study groups, different sample sizes, and different lengths of follow-up periods, as well as differences in methods and measurements. However, this study contributes with comparisons of gender and misuse differences, and the discussion will mainly focus on this.

Gender differences and similarities regarding predictors of mortality

In previous research on mortality, finishing primary school has been identified as a protective factor (Jackson 1994; Berlin et al. 2011; Storbjörk and Ullman 2012). This was true for the women on a univariate level but not at all for the men in the present study, which implies a gender difference.

The influence of social networks and the importance of having a supportive network has been highlighted in previous research on recovery from alcohol and drug problems (Orford et al. 2006; Best et al. 2008), and several studies have described a stable social network as a protective factor (Bauer et al. 2008; Becker and Hu 2008). Previous research has also stressed that men and women tend to use their social networks in different ways during recovery from alcohol and drug abuse (Timko et al. 2005; von Greiff and Skogens 2017). In a Swedish study investigating reasons for seeking and entering treatment, men more often described pressure from a spouse or from friends, while women more often described parental responsibility putting pressure on them (Blomqvist and Cristophs 2005). The present study implies that women and men differ in terms of which part of the social network is most important. Not having custody of children seemed to be a risk factor for women. However, with current data it was not possible to establish the order of events, i.e. whether losing custody of your children was what made the situation deteriorate or whether the situation in itself was a reason for losing custody. Having no friends or friends with addiction seemed to be a negative factor for men, but was not significant for women. In line with previous studies (Tuchman 2010), the current investigation suggests that social relationships are important, but function differently for women and for men to a certain extent. Moreover, since this seems to be quite a vital area, investigation using other kinds of data would be relevant in order to understand the influence of social relationships.

Previous research has shown that mental illness is a risk factor and that mental illness is more common among women compared with men (Chou and Dawson 1994; Hesselbrook and Hesselbrook 1997; Hser et al. 2005; Timko et al. 2005; Greenfield et al. 2007; Degenhardt et al. 2010). In the present study, this was confirmed with a higher proportion of mental illness among women. However, the factors studied – ‘have made suicide attempts’ and/or ‘previous psychiatric care for drug or alcohol misuse’ – had no significant effect in the multivariate analysis. One possible interpretation of this is that mental illness was not a risk factor in the group studied, either as concerns women or men; however, it can also be explained by differences in measuring mental illness or that other variables conceal the correlation.

In line with previous research, the number of spells of incarceration were fewer among women (see, e.g. Bird et al. 2003). Earlier studies that have not made separated gender analyses have found criminality to be a risk factor for mortality (Scott et al. 2011; Ravndal et al. 2015). In the present study, prison spells increased the mortality risk in women but not men, which contradicts the results from a study by Gjersing and Bretteville-Jensen (2014) on injecting drug users.

Ending treatment in a planned manner seemed in this study to be a protective factor for men compared with those who dropped out of treatment. This result confirms earlier research highlighting retention in treatment as a protective factor (Bartu et al. 2003; Degenhardt et al. 2010; Ravndal et al. 2015). Although the data in Bartus’ study on opiate and amphetamine users and the meta-analysis of studies on opioid abusers by Degenhardt et al. consisted of both male and female misusers, neither compared treatment effects according to gender.

Both male and female clients interviewed on detoxification wards had a significantly higher mortality risk; but as mentioned above, this was difficult to interpret, since the main difference compared to other groups was the time at which they were interviewed; many clients in this group moved on to other forms of treatment and many of those interviewed in other types of treatment had been transferred from detoxification wards. However, part of the explanation for the significant result could be that the acute/detox group also included those that did not continue to other forms of treatment. Thus, this result might implicitly confirm treatment as the protective factor highlighted in previous studies (Bartu et al. 2003; Degenhardt et al. 2010).

The majority of the clients in the present study who gave alcohol as their most dominant form of substance misuse were polydrug users (von Greiff et al. 2018). Both gender groups that reported alcohol as their most dominant form of substance misuse showed a higher mortality risk compared with those with stimulants as their main form of substance misuse. This is in line with the study by Ravndal et al. (2015) on mortality among drug abusers after seeking treatment, revealing that reported alcohol abuse before intake to treatment in the sample was a significant mortality predictor. Stenbacka et al. (2010) also found higher SMR among clients who were both alcohol and drug abusers compared with those were only drug abusers (opiates and/or central stimulants or cannabis). Another Swedish study investigating mortality after treatment for alcohol and drug problems did not find differences between alcohol and drug-dependent cases (Storbjörk and Ullman 2012). However, one reason for the different results in the latter study might be that they grouped drug and/or alcohol users by dependence (through ICD-10), while the other studies discriminated groups through interviews or data from the EuropAddiction Severity Index (Kokkevi and Hartgers 1995).

Alcohol as the most dominant form of substance misuse was the only variable that showed significant results for both genders in the present study, supporting the findings of earlier studies on the importance of including alcohol consumption in studies on the mortality of drug misusers. Moreover, the results suggest that polydrug use, including alcohol misuse, is a greater hazard for women, since the mortality risk was considerably higher for the female participants.

Limitations and strengths

The study investigated and compared predictors of mortality in men and women 30 years after they had been treated in residential care for substance misuse, and was limited to the factors that were available from the interviews conducted at intake into residential care. The dominant reason for not being interviewed was dropping out of treatment before the interview, which might indicate misuse that had progressed further and could imply higher mortality rates in this group, in which case the mortality rate in this study might be somewhat underestimated since it is limited to those who were interviewed.

The impact of the available variables deemed relevant were analysed, knowing that other factors, both during the time of the interview and later on in life, may be associated with premature death. Also, since the variables were based on self-reports during the interview or on reports from staff (information on dropout), bias is possible. A large number of variables were included in the multivariate analysis, which constitutes both a strength and a limitation. The strength is that this enables an analysis of the relative impact of different factors. However, this increases the statistical probability of false associations; but what is perhaps more significant for this study is that important factors may be concealed behind correlating variables. Still, there was no univariate correlation between the independent variables that exceeded R = 0.30, with the exception of Age at intake, which showed a high correlation with Children (R = 0.40 and 0.50 for men and for women, respectively) and Previous imprisonment (R = 0.37 and 0.34 for men and for women, respectively). The strengths of the study lie foremost in the length of the follow-up, the possibilities to compare groups with different substances as main form of misuse and the comparison of gender differences.

Conclusions

Three main gender differences emerged in the results. Imprisonment was highly predictive of mortality for the women in the study, which suggests that it is particularly important to pay attention to this group, for example, with interventions during and after sentences. Also, the substantial hazard revealed in the study for women with polydrug misuse issues including alcohol shows that there is a need to look into this during treatment for substance misuse. The suggested gender differences regarding the importance of social factors, such as family and friends, were in line with previous research but need to be investigated more thoroughly using different data.

Disclosure statement

No conflict of interest has been reported by the authors.

Additional information

Funding

This research was funded by the Swedish Research Council for Health, Working Life and Welfare (Grant # 2015-00980).