Mortality during and after specialist alcohol and other drug treatment: Variation in rates according to principal drug of concern and treatment modality
Mortality during and after treatment
Havard et al.
National Drug and Alcohol Research Centre, UNSW Sydney Sydney Australia
Centre for Big Data Research in Health, UNSW Sydney Sydney Australia
* CorrespondenceDr Alys Havard, National Drug and Alcohol Research Centre, UNSW Sydney, Sydney, NSW 2052, Australia.
Email: alys.havard@unsw.edu.au
Abstract
Introduction
For people accessing treatment for problems with drugs other than opioids, little is known about the relationship between treatment and mortality risk, nor how mortality risk varies across treatment modalities. We addressed these evidence gaps by determining mortality rates during and after treatment for people accessing a range of treatment modalities for several drugs of concern.
Methods
We conducted a cohort study using linked data on publicly funded specialist alcohol or other drug treatment service use and mortality for people receiving treatment in New South Wales between January 2012 and December 2018. We calculated and compared during‐treatment and post‐treatment crude mortality rates and age‐ and sex‐standardised mortality rates, separately for each principal drug of concern and modality.
Results
Over the study period, 45,026 people accessed treatment for problems with alcohol, 26,407 for amphetamine‐type stimulants, 23,047 for cannabinoids and 21,556 for opioids. People treated for alcohol or opioid problems had higher crude mortality rates (1.48, 1.91, 1.09 per 100 person years, respectively) than those with problems with amphetamine‐type stimulants or cannabinoids (0.46, 0.30 per 100 person years, respectively). Mortality rates differed according to treatment status and modality only among people with alcohol or opioid problems.
Discussion and Conclusions
The observed variation in mortality rates indicates there is scope to reduce mortality among people accessing treatment with alcohol or opioid problems. Future research on mortality among people accessing drug and alcohol treatment should account for the variation in mortality by drug of concern and treatment modality.
Article notes
Havard A , Jones N , Bharat C , Gisev N , Pearson S‐A , Shakeshaft A , et al. Mortality during and after specialist alcohol and other drug treatment: Variation in rates according to principal drug of concern and treatment modality. Drug Alcohol Rev. 2023;42(6):1461–1471. 10.1111/dar.13669 37186492 PMC10946946
Boxed Text
- This is the first comprehensive study of mortality among people accessing different specialist alcohol and other drug treatment modalities for several drugs of concern.
- Mortality rates were higher among people with alcohol or opioid problems, relative to other drugs.
- Among people with alcohol or opioid problems, mortality rates were lower during treatment than after treatment for some treatment modalities.
- During‐treatment and post‐treatment mortality rates varied by treatment modality among people with alcohol or opioid problems.
- Findings indicate there is scope to reduce mortality among people accessing treatment with alcohol or opioid problems.
1INTRODUCTION
People with alcohol or other drugs problems are more likely to die prematurely than their peers in the general population. Compared to age and sex peers, mortality risk is 3.6 times higher among people with alcohol use disorder [1], 6.8 times higher among people with regular or dependent amphetamine use [2] and 14.7 times higher among people with regular or dependent opioid use [3]. Published data are insufficient to determine whether mortality is elevated among cannabis users [4].
There is evidence that engaging in drug and alcohol treatment is associated with reduced mortality, but this evidence primarily relates to treatment for opioid dependence. Systematic reviews of studies focused on people receiving opioid agonist treatment (OAT) found lower rates of all‐cause and cause‐specific mortality during treatment, compared to out‐of‐treatment [3, 5]. Individual cohort studies have found reduced in‐treatment mortality rates for people accessing other treatments for opioid dependence, including withdrawal management [6, 7], psychosocial treatments [6] and residential treatments [8]. There has been limited investigation of the relationship between treatment and mortality for drugs other than opioids. To our knowledge, this has been examined for cannabis only, with a single study finding no difference in mortality rates during and after treatment (various modalities combined) [9]. In a study among people accessing treatments for a range of drugs in Victoria, Australia, mortality during treatment was higher than in the 2 years after treatment [10]. The authors note the inconsistency between this finding and existing evidence, however, hypothesising it is due to the aggregation of a diverse range of treatment modalities and drugs. Whether engagement in treatment is associated with reduced mortality among people with problematic use of drugs other than opioids, and whether this varies across different treatment modalities, is largely unknown.
Even for treatment modalities with a demonstrated association with reduced mortality relative to periods out of treatment, there remains excess risk of mortality relative to the general population. Specifically, mortality rates during and after OAT exceed mortality rates among the general population [11]. Excess mortality risks have also been found among people accessing other treatment modalities for other drugs of concern [1, 10, 12, 13]. This suggests that there are further opportunities to reduce the risk of premature mortality among people accessing drug and alcohol treatment. Designing service enhancements that optimise client outcomes while making best use of limited resources requires an understanding of the treatment populations most at risk of premature mortality, including whether the risk during treatment and after treatment varies according to treatment modality. This question has been somewhat addressed for opioid dependence treatments, with OAT associated with lower during‐treatment mortality risk than community psychological support [8], and similar during‐treatment risk as residential treatments [8] and withdrawal management [7]. In contrast, mortality risk in the post‐treatment period is higher among those discharged from residential treatment and OAT than those discharged from psychological support [8].
To our knowledge, there has been no prior investigation of how mortality rates during treatment vary across treatment modalities for drugs other than opioids. There has, however, been some examination of how post‐treatment mortality rates vary. The above‐mentioned Victorian study found the highest rates of mortality in the 2 years following residential withdrawal management, and lowest mortality in the year following counselling [10]. A similar study in Texas found significantly higher drug‐related mortality rates in the year following discharge from residential treatments, withdrawal management and OAT compared to those discharged from outpatient treatment [14]. Neither of these studies, however, reported mortality rates separately by principal drug of concern (PDoC), which is associated with much variation in mortality risk.
Our objective was to address these evidence gaps for people accessing a range of specialist drug and alcohol treatment modalities across a range of PDoCs. Specifically, we compared mortality rates during and after treatment, separately for each PDoC and treatment modality. We also examined whether mortality rates during and after treatment vary according to treatment modality, separately for each PDoC.
2METHODS
This cohort study was based on linked data on health service use and mortality for people receiving publicly funded specialist alcohol or other drug treatment in New South Wales (NSW), Australia's most populous state.
2.1 Data sources
The NSW Minimum Data Set for Drug and Alcohol Treatment Services (MDS DATS) records all services provided by government and non‐government drug and alcohol agencies receiving NSW Ministry of Health funding for providing specialist drug and/or alcohol and/or gambling services. The PDoC for each treatment episode is coded according to the Australian Standard Classification of Drugs of Concern [15, 16]. The MDS DATS extract for this study included only closed treatment episodes.
The NSW Admitted Patient Data Collection records admissions to all NSW hospitals. Diagnoses are coded according to the International Statistical Classification of Diseases and Related Health Problems, Tenth Revision, Australian Modification [17], procedures are coded according to the Australian Classification of Health Interventions [18] and admissions as a whole are classified according to Australian‐Refined Diagnostic Related Groups [19].
Death records from the NSW Registry of Births, Deaths and Marriages records the date of every death registered in NSW.
Records from these collections were linked for all people with a MDS DATS record with a treatment exit date between 1 January 2012 and 31 March 2019 or a record in the non‐admitted patient (NAP) data collection with a service start date between 1 July 2015 and 30 September 2018, where the service provided was coded as ‘20.52: Addiction Medicine’ or ‘40.30: Alcohol and Other Drugs’ [20]. The NAP data collection includes data on outpatient services provided by the NSW public hospital system. Although people receiving NAP services were included in our cohort, because the PDoC is not specified, we did not use NAP treatment records in our analyses.
The linkage was performed by the NSW Centre for Health Record Linkage using personal identifiers. As MDS DATS records from before July 2015 did not contain full personal identifiers, linkage of these early records used a statistical linkage key based on partial identifiers. The rate of false positive links, based on the Centre for Health Record Linkage's review of a random 1000 person IDs, was 0.5%.
2.2 Study cohort
We identified people who, in the period 1 January 2012 and 31 December 2018, had at least one MDS DATS record or a hospital admission for withdrawal management for alcohol use disorder, as recorded in the Admitted Patient Data Collection. We excluded: people who had no treatment episodes with one of the four nominated PDoCs of interest (see Section 2.5); those seeking treatment for gambling disorder or for another person's substance use; people with inconsistent information across different data sources; and people for whom age and sex could not be determined.
We conducted a sensitivity analysis by restricting our cohort to people entering treatment from 1 July 2015 onwards (sensitivity analysis 1) so as to examine the impact of potentially reduced accuracy of links involving early MDS DATS records.
2.3 In‐treatment and post‐treatment periods
For people already in treatment at the beginning of the study period, we began follow up on 1 January 2012. For people commencing their first treatment episode after 1 January 2012, we began follow‐up on their treatment start date. We followed people from this date through to 31 December 2018 (or until the date of death, if this occurred earlier). We used the treatment entry and exit dates recorded in the MDS DATS and the admission and discharge dates in the admitted patient data collection to group person‐time into in‐treatment and post‐treatment periods and subdivided these periods according to the treatment modality accessed and the PDoC.
We excluded post‐treatment days occurring after 31 March 2018. This is because we had data on closed treatment episodes only and people may have commenced a new treatment episode that was still open when the data were extracted (31 March 2019). We considered this risk to be highest for the last 12 months of data availability as the MDS DATS data collection rules specify (with some exceptions) that a treatment episode should be closed (and a new episode record created) when it reaches the maximum length of 12 months [21].
2.4 Treatment modality
We focused on nine modalities of specialist alcohol and other drug treatment: assessment only, counselling, involuntary drug and alcohol treatment (IDAT), OAT, outpatient consultation, rehabilitation, residential rehabilitation, support and case management, and withdrawal management. We identified these modalities by combining information from the MDS DATS field of ‘main service provided’ with information regarding the setting in which the service was delivered, as defined in Table S1, Supporting Information.
We also used hospital records to identify episodes of withdrawal management for alcohol use disorder based on an International Statistical Classification of Diseases and Related Health Problems, Tenth Revision, Australian Modification diagnosis code related to problematic use of alcohol, as well as a relevant Australian Classification of Health Interventions code in the primary procedure field or a relevant Australian‐Refined Diagnostic Related Groups code (Table S2).
We observed that 8.8% of treatment episodes overlapped with at least one other episode, suggesting that some people received multiple treatments simultaneously. For the purposes of the analyses, we considered each treatment modality independently. We also conducted a sensitivity analysis using data related to single service episodes only (sensitivity analysis 2).
2.5 Principal drug of concern
We defined four broad PDoC categories: alcohol, amphetamine‐type stimulants, cannabinoids and opioids. See Table S3, for the drugs comprising each of these categories.
2.6 Mortality rates
We attributed deaths to an in‐treatment period if they occurred during the episode or on the day after treatment exit. We included deaths on the day after treatment exit to account for the likely scenario in which an individual dies and stops presenting for treatment, hence the treatment provider closes the treatment episode and records the client was last seen as the treatment end date. While our main analysis assumed a one‐day grace period, sensitivity analysis 3 examined whether alternative grace periods (0 days and 7 days) returned different results.
For the treatment modality of assessment only, where the episode represents a single treatment session, we attributed all person‐time and deaths on the date of assessment or thereafter to the post‐treatment period. Accordingly, our analyses for assessment only included no in‐treatment person‐time.
For each in‐treatment and post‐treatment period of interest, we computed crude mortality rates (CMR) as the total number of deaths occurring during the period divided by the total person‐time in that period. We present this as a rate of death per 100 person‐years. As CMRs are assumed to follow the Poisson distribution, we calculated 95% confidence intervals (CI) for the CMRs based on a Poisson distribution [22]. We also computed direct age‐ and sex‐ standardised mortality rates (ASR) for each period of interest. To calculate ASRs, we calculated the CMRs for each 10‐year age and sex stratum in our cohorts, and applied the age‐, and sex‐ specific weights from 2015 NSW population estimates [23]. We calculated the 95% CIs for the ASRs using the normal approximation [24].
We considered CMRs/ASRs for different time periods or treatment modalities to be statistically significantly different if the 95% CIs around the rates did not overlap. In the case of overlapping confidence intervals, noting that this does not necessarily indicate the difference between the estimates is not significant [25], we based our interpretation on inspection of the CMR/ASR estimates and the uncertainty surrounding those estimates.
Sample size calculations assuming a CMR of 0.89 deaths per 100 person years (the CMR among people who accessed OAT in NSW [11]) indicated that estimating the true mortality rate within ±0.5 deaths per 100 person years would require 1360 person years of observation. We therefore defined a reporting threshold such that we reported CMRs only for treatment modalities with at least 1360 person years of observation available for the in‐treatment and/or post‐treatment period. We allowed an exception to this reporting threshold for IDAT where alcohol was nominated as the PDoC given the added importance of understanding mortality risk in this group and the lack of existing evidence.
Ethics approval to conduct this study was granted by the UNSW Human Research Ethics Committee (ref: HC210443).
3RESULTS
A total of 102,256 people met our cohort inclusion criteria (Figure S1 quantifies the exclusions). Over the study period, 45,026 people accessed treatment for alcohol problems, 26,407 people for amphetamine‐type stimulants, 23,047 for cannabinoids and 21,556 for opioids.
The CMR among people accessing treatment for alcohol problems was 1.48 per 100 person years (PY) [95% CI 1.41–1.55] (Table 1), while the ASR was 1.64 per 100 PY [95% CI 1.54–1.73]. The CMR among people accessing treatment for opioid problems was significantly lower (1.09 [1.01–1.17]), but after age‐ and sex‐ standardisation, the mortality rate (ASR 1.56 [1.31–1.81]) was similar to that for alcohol. People accessing treatment for problems with amphetamine‐type stimulants or cannabinoids had significantly lower CMRs than people accessing treatment for alcohol or opioid problems. They also had significantly lower ASRs than people treated for alcohol problems.
| Principal drug of concern | Person years | Deaths | CMR per 100 PY [95% CI] | ASR per 100 PY [95% CI] |
|---|---|---|---|---|
| Alcohol | 117,766 | 1747 | 1.48 [1.41–1.55] | 1.64 [1.54–1.73] |
| Amphetamine‐type stimulants | 53,242 | 246 | 0.46 [0.41–0.52] | 1.03 [0.55–1.51] |
| Cannabinoids | 51,223 | 153 | 0.30 [0.25–0.35] | 0.68 [0.31–1.05] |
| Opioids | 70,093 | 763 | 1.09 [1.01–1.17] | 1.56 [1.31–1.81] |
The number and characteristics of people accessing each treatment modality are reported separately by PDoC in Tables S4a–d. Inspection of these descriptive data suggests substantial variation between treatment modalities in stable accommodation, injecting drug use history and hospitalisation history. Mortality rates during and after treatment are reported in the sections that follow, separately by PDoC and treatment modality.
3.1 Alcohol
Among people with alcohol problems, mortality rates during and after treatment were computed for seven of the nine treatment modalities (see Table 2). In‐treatment and post‐treatment person‐time for OAT and rehabilitation fell below the reporting threshold.
| Person years | Deaths | CMR per 100 PY [95% CI] | ASR per 100 PY [95%CI] | |
|---|---|---|---|---|
| Assessment only | ||||
| In treatment | N/A | N/A | N/A | N/A |
| Post treatment | 8936 | 75 | 0.84 [0.66–1.05] | 0.94 [0.67–1.21] |
| Counselling | ||||
| In treatment | 7325 | 32 | 0.44 [0.30–0.62] | 0.39 [0.23–0.54] |
| Post treatment | 39,968 | 349 | 0.87 [0.78–0.97] | 0.88 [0.77–0.99] |
| Involuntary drug and alcohol treatment | ||||
| In treatment | 43 | 0 | 0.00 [0.00–8.54] | a |
| Post treatment | 82 | 7 | 8.56 [3.44–17.64] | a |
| Outpatient consultation | ||||
| In treatment | 693 | 6 | 0.87 [0.32–1.89] | a |
| Post treatment | 6311 | 137 | 2.16 [1.82–2.56] | 1.90 [1.52–2.27] |
| Residential rehabilitation | ||||
| In treatment | 1229 | 0 | 0.00 [0.00–0.30] | a |
| Post treatment | 7609 | 43 | 0.57 [0.41–0.76] | 0.65 [0.40–0.91] |
| Support and case management | ||||
| In treatment | 1494 | 11 | 0.74 [0.37–1.32] | a |
| Post treatment | 9270 | 89 | 0.96 [0.77–1.18] | 1.06 [0.80–1.32] |
| Withdrawal management | ||||
| In treatment | 1128 | 10 | 0.89 [0.42–1.63] | a |
| Post treatment | 27,101 | 372 | 1.37 [1.24–1.52] | 1.22 [1.06–1.38] |
In‐treatment mortality rates were significantly lower than post‐treatment mortality rates for counselling and residential rehabilitation. A pattern of lower in‐treatment CMR point estimates was also observed for IDAT and outpatient consultation, but as confidence intervals overlapped statistical significance is uncertain. In‐treatment CMR point estimates did not differ substantially from post‐treatment CMRs for support and case management and for withdrawal management, and confidence intervals overlapped substantially.
In‐treatment mortality rates were significantly higher for counselling, outpatient consultation, support and case management and withdrawal management than for residential rehabilitation. IDAT could not be included in this comparison due to limited in‐treatment person‐time and deaths. In‐treatment ASRs could be computed only for counselling, preventing us from adjusting for potential age and sex differences in our comparison across treatment modalities.
We found significantly higher post treatment mortality rates among people who previously received IDAT, outpatient consultation and withdrawal management, relative to all other treatment modalities. Post‐treatment mortality was significantly higher for IDAT than outpatient consultation, which in turn was associated with higher post‐treatment mortality than withdrawal management. After age‐ and sex‐standardisation, post‐treatment mortality rates remained elevated for outpatient consultation relative to other modalities, while post‐treatment ASRs for withdrawal management were significantly higher than those for counselling and residential rehabilitation only. IDAT could not be included in the age‐ and sex‐standardised comparisons owing to small cell counts.
Our sensitivity analyses among people with alcohol problems yielded results that were consistent with the main analysis (see Table S5a), although in some cases, a previously statistically significant difference was no longer significant. This is not surprising given the sensitivity analyses included fewer people and/or less person time, and we do not consider this a meaningful change to the study results.
3.2 Amphetamine‐type stimulants and cannabinoids
Among people with problems with amphetamine‐type stimulants or cannabinoids, mortality rates during and after treatment were computed for six treatment modalities (Table 3), with IDAT, OAT and rehabilitation excluded. For both these PDoCs, there was no evidence to suggest a significant difference between in‐treatment and post‐treatment CMRs, with estimates not differing markedly, and confidence intervals overlapping substantially for all treatment modalities.
| Amphetamine‐type stimulants | Cannabinoids | |||||
|---|---|---|---|---|---|---|
| Person years | Deaths | CMR per 100 PY [95% CI] | Person years | Deaths | CMR per 100 PY [95% CI] | |
| Assessment only | ||||||
| In treatment | N/A | N/A | N/A | N/A | N/A | N/A |
| Post treatment | 6849 | 26 | 0.38 [0.25–0.56] | 4895 | 19 | 0.39 [0.23–0.61] |
| Counselling | ||||||
| In treatment | 3733 | 8 | 0.21 [0.09–0.42] | a | <5 | 0.12 [0.03–0.31] |
| Post treatment | 16,090 | 75 | 0.47 [0.37–0.58] | 20,424 | 60 | 0.29 [0.22–0.38] |
| Outpatient consultation | ||||||
| In treatment | a | <5 | 0.50 [0.01–2.77] | a | <5 | 0.62 [0.02–3.48] |
| Post treatment | 2205 | 14 | 0.64 [0.35–1.07] | 2143 | 11 | 0.51 [0.26–0.92] |
| Residential rehabilitation | ||||||
| In treatment | 1054 | 0 | 0.00 [0.00–0.35] | 348 | 0 | 0.00 [0.00–1.06] |
| Post treatment | 6476 | 17 | 0.26 [0.15–0.42] | 2586 | 7 | 0.27 [0.11–0.56] |
| Support and case management | ||||||
| In treatment | a | <5 | 0.26 [0.07–0.67] | a | <5 | 0.07 [0.00–0.40] |
| Post treatment | 7855 | 26 | 0.33 [0.22–0.48] | 8847 | 20 | 0.23 [0.14–0.35] |
| Withdrawal management | ||||||
| In treatment | 89 | 0 | 0.00 [0.00–1.28] | a | <5 | 0.36 [0.01–1.99] |
| Post treatment | 6074 | 32 | 0.53 [0.36–0.74] | 5981 | 17 | 0.28 [0.17–0.46] |
During‐treatment CMRs did not differ markedly across different treatment modalities among people with problems with amphetamine‐type stimulants or cannabinoids, and confidence intervals overlapped substantially. Small cell sizes meant we could not adjust for potential age and sex difference in this comparison (see Table S6).
For both these PDoCs, inspection of post‐treatment CMRs and the associated confidence intervals did not suggest marked differences across different treatment modalities. We were able to compute post‐treatment ASRs only for counselling and support and case management, as well as withdrawal management for amphetamine‐type stimulants; there was no evidence to suggest a significant difference in post‐treatment ASRs across these modalities.
All three sensitivity analyses among people with problems with amphetamine‐type stimulants or cannabinoids (Table S5b, c, respectively) yielded results that were consistent with the main analysis.
3.3 Opioids
Among people with opioid problems, mortality rates during and after treatment were computed for seven treatment modalities (Table 4), with IDAT and rehabilitation excluded. Comparisons of in‐treatment and post‐treatment mortality rates indicate that, for counselling and OAT, mortality during treatment was significantly lower than after treatment. A similar pattern was indicated for outpatient consultation and residential rehabilitation, although confidence intervals around these CMRs overlapped. In‐treatment CMRs did not differ markedly from post‐treatment CMRs for support and case management and for withdrawal management, and confidence intervals overlapped substantially.
| Person years | Deaths | CMR per 100 PY [95% CI] | ASR per 100 PY [95%CI] | |
|---|---|---|---|---|
| Assessment only | ||||
| In treatment | N/A | N/A | N/A | N/A |
| Post treatment | 3460 | 25 | 0.72 [0.47–1.07] | 1.00 [0.49–1.52] |
| Counselling | ||||
| In treatment | 1862 | 8 | 0.43 [0.19–0.85] | b |
| Post treatment | 7564 | 81 | 1.07 [0.85–1.33] | 1.22 [0.86–1.58] |
| Opioid agonist treatment | ||||
| In treatment | 20,460 | 172 | 0.84 [0.72–0.98] | 0.85 [0.68–1.02] |
| Post treatment | 22,483 | 273 | 1.21 [1.07–1.37] | 1.36 [1.14–1.58] |
| Outpatient consultation | ||||
| In treatment | a | <5 | 1.09 [0.23–3.19] | b |
| Post treatment | 1871 | 30 | 1.60 [1.08–2.29] | 2.13 [1.28–2.98] |
| Residential rehabilitation | ||||
| In treatment | a | <5 | 0.19 [0.00–1.03] | b |
| Post treatment | 2770 | 22 | 0.79 [0.50–1.20] | 1.67 [0.00–3.35] |
| Support and case management | ||||
| In treatment | 1160 | 14 | 1.21 [0.66–2.03] | b |
| Post treatment | 4028 | 29 | 0.72 [0.48–1.03] | 0.82 [0.35–1.30] |
| Withdrawal management | ||||
| In treatment | 464 | 5 | 1.08 [0.35–2.51] | b |
| Post treatment | 5901 | 71 | 1.20 [0.94–1.52] | 1.36 [0.90–1.82] |
Among people with opioid problems, a pattern of higher in‐treatment CMRs was seen for all treatment modalities relative to counselling and residential rehabilitation, but confidence intervals around all in‐treatment CMRs overlapped. Small cell counts meant an in‐treatment age‐ and sex‐standardised mortality rate could be computed only for OAT.
Post‐treatment CMRs for assessment only and support and case management were significantly lower than those for OAT and outpatient consultation. After age‐ and sex‐standardisation, this pattern remained for outpatient consultation only, however, confidence intervals overlapped.
Our sensitivity analyses among people with opioid problems yielded results that were mostly consistent with the main analysis (Table S5d), although some differences observed in the main analysis were not always accompanied by non‐overlapping confidence intervals in sensitivity analyses. The analysis restricted to treatment received between July 2015 and December 2018, however, did yield a contrasting result with no difference found between the in‐treatment and post‐treatment CMR for OAT, both in terms of similar CMR point estimates and overlapping confidence intervals.
4DISCUSSION
To our knowledge, this is the first comprehensive study examining mortality among people accessing different specialist alcohol and other drug treatment modalities for several drugs of concern. We found significantly higher mortality rates among people accessing treatment for problems with alcohol or opioids, relative those with problems with amphetamine‐type stimulants or cannabinoids. This pattern persisted after age‐ and sex‐standardisation. We also found differences in mortality rates according to treatment status and treatment modality among people with alcohol or opioid problems. We found no such differences among people with problems with amphetamine‐type stimulants or cannabinoids, probably due in part to the low overall mortality rate in these groups and limited statistical power to detect differences, but potentially also due to the limited effectiveness of existing treatments for stimulant use disorders [26]. These findings highlight the importance of examining mortality rates separately by PDoC when examining variation in mortality rates. They also suggest there is scope to reduce mortality among people accessing treatment with alcohol or opioid problems. Larger scale studies are needed to determine whether this is also the case for people accessing treatment with amphetamine‐type stimulants or cannabinoids as their drug of concern.
Our comparisons of mortality during and after treatment point to service delivery and policy changes that might contribute to reduced mortality risk among people accessing treatment for problems with alcohol or opioids. Consistent evidence that engagement in OAT is associated with reduced mortality has led to calls for expanded access to OAT, strategies to maximise retention, and/or greater post‐treatment follow‐up [5, 6, 27]. Similar strategies should be considered for treatment modalities where mortality rates were reduced during treatment, relative to after treatment: counselling, residential rehabilitation, IDAT and outpatient consultation for people with problematic use of alcohol; and counselling, OAT, outpatient consultation and residential rehabilitation for people with opioid problems. It is important to note, however, that these are associations and we have not formally investigated causality. Establishing causality is extremely challenging in this setting where randomised studies are impractical and unethical [5, 6], and non‐randomised studies, including the current study, are subject to confounding by factors that are related to both the decision to participate in treatment (and the treatment modality chosen) and the risk of mortality [8]. Likely confounding factors include severity of the person's substance dependence, the extent to which they have problems with other drugs, their physical and mental health, their financial, housing and employment circumstances, and the strength of their support networks. It should be the goal of future research to obtain good quality data on likely confounding factors and to conduct robust causal analyses. While there is a dearth of evidence from causal studies, it is important to carefully evaluate the impact of any implemented policy reform or service changes so that they can be reversed, modified or extended based on the findings of the evaluation.
In implementing strategies to optimise retention and/or post‐treatment follow‐up, there could be merit in prioritising outpatient consultation for people with opioid problems given they had higher post‐treatment mortality relative to people accessing other modalities even after adjusting for age and sex. The same applies to outpatient consultation and withdrawal management for people with alcohol problems. Among people accessing treatment for alcohol problems, there was also variation between modalities in during‐treatment mortality, suggesting there may be value in adjunct interventions for people accessing certain treatment modalities for alcohol problems (counselling, outpatient consultation, support and case management, and withdrawal management). However, as we could not adjust for age and sex in this comparison, this result should be interpreted with caution. Although these between‐modality differences in mortality likely reflect variation in the underlying mortality risk among people accessing these different modalities (the distribution of risk factors in our cohort varied by treatment modality), this variation does suggest the mortality risk in this population is to some extent modifiable. Research building an understanding of how these and other risk factors relate to mortality risk in people accessing different treatment modalities may inform the design of interventions. Similarly, appropriately powered research investigating the causes of death occurring in these groups with elevated mortality risk would provide insight into the common pathways to mortality. Overdose is the most common cause of death among people with opioid dependence [3], while liver cirrhosis, mental health disorders and injuries account for a large proportion of the excess mortality among people with alcohol use disorders [28]. Whether this holds across all treatment modalities and in‐treatment and post‐treatment periods remains to be elucidated.
4.1 Study limitations
Our identification of in‐treatment and post‐treatment periods is imperfect. Although the MDS DATS covers NSW Ministry of Health funded specialist alcohol and other treatment services, treatment for alcohol and other drug problems is also provided in other settings including private treatment agencies, inpatient and outpatient services without a specialisation in alcohol and other drug treatment and services provided in correctional settings. Additionally, while we had data on non‐specialist outpatient services through the NAP data collection, we did not include these episodes in the analysis because the PDoC could not be identified. These data coverage issues mean that, during person‐time we classified as in‐treatment and/or post‐treatment, our cohort may have received services we did not identify. OAT is the modality most impacted by these data coverage issues, with only one‐third of OAT episodes in NSW managed by services receiving NSW Ministry of Health funding [29]. Additionally, OAT episodes can be long in duration and were therefore more likely to still be open at the time of data extraction. The lack of data on open treatment episodes likely led to a greater risk of misclassification later in the study period, with more person‐time incorrectly classified as post‐treatment instead of in‐treatment. This may account for the inconsistent finding between the main analysis and the 2015–2018 sensitivity analyses in relation to OAT. As the 2015–2018 analysis did not yield altered conclusions for other treatment modalities or other PDoCs, the inconsistent finding is unlikely due to improved linkage accuracy in recent years. Given the high risk of misclassified person‐time arising from incomplete identification of OAT episodes, the results relating to mortality during and after treatment for people with opioid problems should be interpreted with caution.
As we used the non‐overlap of confidence intervals as an indicator of statistical significance, uncertainty remains as to whether those estimates with overlapping confidence intervals were statistically significantly different from each other. In these cases, we based our interpretation on inspection of the CMR/ASR estimates and the width of the confidence intervals. While this method can be subjective, we considered it reasonable given the growing acceptance in the scientific community that statistical interpretation and reporting would be improved by focusing on estimation (the size of effects and the uncertainty surrounding those estimates) instead of statistical testing [25]. It is also important to note that mortality rates are one measure of the harms experienced by people with alcohol or other drug problems. Future research measuring the relationship between treatment and alcohol and other drug‐related harm, and how the risk of harm varies across treatment modalities, might consider a more comprehensive range of outcomes including substance use, physical and mental health, and quality of life, as well as social outcomes.
4.2 Conclusions
Mortality rates differed according to treatment status and treatment modality among people with alcohol or opioid problems, indicating there is scope to reduce mortality among people accessing treatment for these drugs of concern. Larger‐scale studies are needed to determine whether this is also the case for people accessing treatment for problems with amphetamine‐type stimulants or cannabinoids. Research to inform the design of appropriate responses to the elevated mortality among people accessing drug and alcohol treatment should account for the variation in mortality by PDoC and treatment modality.
CONFLICT OF INTEREST STATEMENT
Louisa Degenhardt and Michael Farrell have received investigator‐initiated untied educational grants from Reckitt Benckiser/Indivior, Mundipharma and Seqirus Pty Ltd. Michael Farrell is also on the Board of Directors for a specialist alcohol and drug treatment service in NSW. The Centre for Big Data Research in Health at UNSW Sydney received funding from AbbVie Australia in 2020. These grants were unrelated to the current study, and none of these companies had any knowledge or involvement in this study. All other authors declare no competing interests.
Supporting information
ACKNOWLEDGEMENTS
The authors would like to thank the custodians of the data custodians of the NSW Minimum Data Set for Drug and Alcohol Treatment Services, the NSW Admitted Patient Data Collection and the NSW Register of Births, Deaths and Marriages for providing data for this study. The authors thank the NSW Centre for Health Record Linkage and conducting the linkage of records for this study. The authors would also like to thank the Alcohol and Other Drug‐related Mortality Working Group (convened by the NSW Ministry of Health) and representatives from the NSW Ministry of Health, for their contribution to setting priorities for this project, providing advice regarding the coding of the data and data limitations, and providing suggestions for the clinical and policy implications of the study findings. The NSW Ministry of Health commissioned this project and provided the linked data under a Co‐creating Evidence initiative. Although the funder played a role in setting priorities for this project, providing advice regarding the coding of the data and data limitations, and providing suggestions for the clinical and policy implications of the study findings, they had no role in the data analysis, interpretation or drafting of the final manuscript. The opinions expressed in this paper are those of the authors, and do not necessarily represent the views of NSW Health. Chrianna BharatB is supported by National Drug and Alcohol Research Centre and UNSW Scientia PhD scholarships. Natasa Gisev is supported by a UNSW Scientia Fellowship. Louisa Degenhardt is supported by an Australian National Health and Medical Research Council Research Fellowship. The National Drug and Alcohol Research Centre, UNSW Sydney, is supported by funding from the Australian Government Department of Health under the Drug and Alcohol Program. Open access publishing facilitated by University of New South Wales, as part of the Wiley ‐ University of New South Wales agreement via the Council of Australian University Librarians.