Benzodiazepine Use and Mortality Risk: A Nationwide Cohort Study on New Benzodiazepine Users With a 5‐Year Follow‐Up Authors: Hanna Särkilä, Heidi Taipale, Antti Tanskanen, Tero Taiminen, Reijo Sund, Terhi Kurko, Jarmo Hietala, Solja Niemelä Published: 2026 ABSTRACT ABSTRACTIntroductionWe aimed to investigate the risks of all‐cause and cause‐specific mortality associated with dose‐dependent benzodiazepine and Z‐drug (BZDR) use in the Finnish population initiating new BZDR use with five‐year follow‐up.MethodsStudy subjects were included if BZDR use had started in 2006 with no BZDR dispensing during the preceding years 2004–2005. The information drug dispensing was modeled with PRE2DUP method as time‐varying measure updated at every dispensing and then divided into three dose categories which were low dose (< 1.0 Defined Daily Doses [DDDs] per day), medium to high dose (1.0–< 3.0 DDDs/day) and very high‐dose (≥ 3.0 DDDs/day). Risk of mortality was studied with Cox regression models.ResultsThe study included 48,124 incident BZDR users aged 18–65 years (44.4% male). During the 5‐year follow‐up, 2294 (4.9%) individuals died. Compared to non‐use periods, BZDRs use was associated with increased mortality risk (adjusted Hazard Ratio HR 1.28, 95% Confidence Interval CI 1.16–1.41). Compared to the time periods when BZDRs were not used, both medium to high‐dose use (aHR 1.58, 1.36–1.74) and very‐high‐dose use (aHR 2.68, 95% CI 2.20–3.26) were associated with an increased risk of all‐cause mortality, whereas low dose use was not. The highest risk estimates were observed during very high‐dose use and for potentially preventable deaths such as overdoses (aHR 6.18, 95% CI 3.77–10.12), suicides (aHR 4.46, 95% CI 2.79–7.13), and accidents (aHR 3.77, 95% CI 2.66–5.35).ConclusionsPersons using BZDRs in very high doses are at a six‐fold increased risk of overdose deaths compared to non‐use, and those using BZDRs in medium to high doses are at a three‐fold increased risk. In low doses, however, the risk of all‐cause mortality was not elevated. The mortality risk strongly associates with higher doses and is often characterized by concomitant use of more than one BZDRs at the same time.Keywords: benzodiazepines, high‐dose use, mortality, substance use, Z‐drugs Introduction Benzodiazepines and Z‐drugs (BZDRs) have been widely used for many decades to treat symptoms such as anxiety, seizures and insomnia [1, 2]. Their function is based on potentiating the effects of neurotransmitter gamma‐aminobutyric acid (GABA) in GABA‐A receptors in the central nervous system (CNS) [3, 4]. Despite BZDRs undisputed usefulness in acute symptom relieving, they still tend to remain in use longer than many guidelines recommend, that is, few weeks [5, 6, 7]. The potential harms of long‐term and high‐dose use, such as tolerance and dependence, risky behaviors, increased exposure to trauma, and paradoxical increase of anxiety have been already vastly recognized [8, 9, 10, 11]. Increasing evidence is also shown on how BZDRs are associated with poor quality of life, increased risk of disability pension, long‐term unemployment and sickness absence [12, 13]. The knowledge of the association of BZDR use and mortality has, however, been rather limited [14]. Previous studies concerning the correlation between benzodiazepine use and all‐cause mortality have mostly been conducted in older or otherwise high‐risk classified groups and thus may not directly generalise to adult general population [15, 16, 17, 18]. One large US cohort conducted by Patorno et al. with a 6‐month follow‐up found only a minor increase in risk of all‐cause mortality associated with benzodiazepine initiation versus non‐initiation [19]. In another national cohort study from South Korea, it showed that the 5‐year all‐cause mortality risk was 10.0% in BZD users versus 9.4% in non‐users; time‐dependent HR = 1.15 (95% CI 1.09–1.22) [20]. Two large cohort studies conducted in French and UK populations also found a significant while moderate increase in all‐cause mortality in relation to benzodiazepines [21]. The study focused on the risk for all‐cause mortality for up to 12 months. The risk of mortality associated with different doses of BZDRs in the general adult population has not been explored. Previous studies have assessed dose‐dependent risks in specific sub‐populations, such as schizophrenia, or after myocardial infarction [22, 23]. In a study conducted by Brodeur et al., cumulative high‐dose benzodiazepine use was associated with increased mortality in patients with schizophrenia [23]. Liu et al. found that self‐reported BZDRs use was associated with an increased risk of all‐cause and cardiovascular mortality, with dose‐dependent increases in risk in older patients after myocardial infarction [22]. Aims of the Study Our aim in this nationwide population‐based 5‐year follow‐up study was to investigate the risks of all‐cause and cause‐specific mortality associated with dose‐dependent BZDR use in the Finnish population initiating new BZDR use starting from the year 2006. Materials and Methods Ethical Concerns This study was based on secondary register data, no ethics board approval was required (Finnish National Board on Research Integrity TENK) according to Finnish legislation. Social Insurance Institution (SII) approved the use of the data for the current study. All registers are linked via personal identity codes which are assigned to each resident at birth or immigration. The data used in the study were fully pseudonymized before the authors accessed the data. All data preparation and linkage in the study were performed with pseudo‐identifiers, and the authors did not have access to information that could identify individual participants at any stage of the study. Legal restrictions prevent the open sharing of the pseudonymised data supporting the current study, as individual‐level health data are considered highly sensitive, and access is strictly regulated by law in Finland (Act on Secondary Use of Health and Social Data (552/2019)). Interested parties may, however, apply for permission to access the data from Findata [24]. Ethics Board approval is not required in Finland for purely register‐based studies. In addition, according to the General Data Protection Regulation of the European Union and the Finnish Data Protection Act, the processing of personal data is permitted without informed consent of study subjects if the task is carried out in the public interest, such as scientific research. Data Source and Study Population The population of this study (N = 129,732 incident BZDR users in Finland) was collected from the National Health Insurance Scheme register, which is administered by the SII, as described in our previous study [25]. For further information on the population and inclusion criteria, see also Figure S1. The register contains information on all reimbursed prescription drug purchases in Finland. This study was restricted to working‐aged persons, aged 18–65 years (N = 95,721), as our previous study informed that use patterns among older people are different than in working‐aged persons [13]. Users, aged 18–65 years, were included if BZDRs use had started in 2006 with no BZDR dispensing during the preceding 2 years 2004–2005. Users purchasing only one BZDR package (N = 47,597) were excluded from the final study sample as their use status was uncertain. The final study cohort included thus 48,124 incident BZDR users and all study subjects had used BZDRs at some point, and the reference category was non‐use of BZDRs that is, time after discontinuation of use. The follow‐up started from BZDR initiation and was 5 years or until death, whichever occurred first. The Anatomical Therapeutic Chemical (ATC) classification system was utilized to define study population and BZDRs used [26]. Only orally administered dosage forms were included, except for oral suspensions (mainly used for epilepsy indication). BZDR was defined as benzodiazepines (N03AE01, N05BA excluding clobazam, N05CD, N05CF, and N06CA01 that is, amitriptyline combined with chlordiazepoxide) and Z‐drugs (N05CF). Those who initiated the use with clonazepam or clobazam for epilepsy indication were not included in this study. To calculate the time periods when BZDRs were used we used PRE2DUP method which is an approach based on mathematical modeling of personal drug purchasing behaviors [27]. Since the definition of high‐dose BZDR use is still somewhat undefined [28], we decided to partly adjust our previous study's terminology [25]. The drug use periods were divided into three dose categories as follows: (1) low dose that is, < 1.0 Defined Daily Doses (DDDs) per day, (2) medium to high‐dose that is, 1.0–< 3.0 DDDs/day (instead of medium‐high dose), and (3) very high‐dose that is, ≥ 3.0 DDDs/day. DDD is the assumed average maintenance dose per day for a drug used for its main indication in adults and it is only assigned for medicines given an ATC code [26]. One DDD equates to 10 mg of diazepam [29]. The dose was re‐evaluated at every dispensing during the follow‐up, so the same person could be in different dose groups at different times of the study period. Discontinuation happened and non‐use of BZDRs started when the modeling showed that no drug supply was left. Study Endpoints The primary endpoint of the follow‐up was all‐cause mortality. In addition, mortality was further categorized as natural deaths (A00–N99), external causes (V01–Y89), accidental deaths (V01–X59), suicides (X60–X84, Y10–Y34), and overdoses (X41–X42, X61–X62, F11–F12, F14–F16, F19). In Finland, a forensic autopsy is performed at the request of police if the death was caused by a crime, accident, suicide, poisoning, occupational disease or medical treatment, or there is cause to suspect that the death was caused by such a reason, or if the death otherwise occurred unexpectedly [30]. A medical autopsy is performed at the request of a physician or relatives of the deceased if the cause of death is otherwise unknown. The detailed information of performed autopsies and other means to determine cause of death by age and sex of deceased in years 1975–2024 is found on Statistics Finland. Finland has exceedingly high medicolegal autopsy rates compared with other countries [31]. Covariates The following covariates were considered as confounders, that is, common causes of BZDR exposure and mortality, and adjusted for in the analysis [13, 19, 20, 25, 32, 33]. Sex, age, level of education, receiving social benefits, the use of gabapentinoids (pregabalin/gabapentin, proxy for chronic/neuropathic pain), the use of prescription opioids, and disability pension status (proxy for functioning and severity of morbidity) were measured as baseline. Receipt of social benefits was defined as receiving social assistance, labor market subsidy, basic unemployment allowance, national pension, study grants, maternity allowance, paternity allowance, parental allowance, and child home care allowance during the year 2005, that is, previous year before BZDR initiation. Receipt of social benefits was considered as additional measure for low socioeconomic position, in addition to education. Clinical covariates included suicide attempt (X60–X84, Y10–Y34), schizophrenia‐spectrum (F20–F29), bipolar disorder (F30–F31), depression (F32–F33), personality disorder (F60–F61), autism‐spectrum (F84), ADHD (F90), anxiety disorder (F40–F42), and stress‐related disorder (F43) which were combined as a sum measure, and Charlson's comorbidity score [34] was utilized to describe somatic comorbidities impacting on the risk of mortality. The sum measure was also constructed of the number of specific SUDs (F10, F11, F12, F13, F14, F15, F16, F18, F19). All these sum measures (sum of psychiatric comorbidities, Charlson's comorbidity score and sum of SUDs) were categorized as 0 versus 1 versus 2 versus 3 versus 4 or more. The comorbidity measures were updated in the model in time with newly diagnosed conditions. Statistical Analysis Risk of mortality was investigated with Cox regression models. BZDR use was modeled as a time‐varying exposure, and persons contribute person‐time to different dose categories and can thus vary between different dose categories and non‐use of BZDRs. The method is depicted more closely on Figure S2. Current and past users of BZDRs both had indications for BZDR initiation. Factors considered impacting on the likelihood of being a current versus past user (and risk of mortality) were adjusted for in the analysis as described in the covariates section. Data management and statistical analysis were conducted primarily with SAS 9.4 and IBM SPSS Statistics 29.0. The figures were drawn with R version 4.5.1. The results were reported as adjusted hazard ratios (aHR) with 95% confidence intervals (CI). Results Study Population Characteristics The study cohort included 48,124 people (44.4% male) who initiated BZDR use in 2006. Most of the subjects (54.6%) had completed a degree in high school or technical school (vocational school, i.e., had more schooling than 9 years), 34.0% received social benefits and 15.8% were on disability pension. For more detailed information of the sociodemographic factors of the study cohort, see Table 1. TABLE 1Baseline characteristics of BZDR users and stratified by mortality status in 5‐year follow‐up.All N = 48,124Survivor/did not die N = 45,830Non‐survivor/died a N = 2294 N % N % N %SexMale21,36744.419,78043.2158769.2Female26,75755.626,05056.870730.8Age categories b 18–2538358.037108.11255.426–35686414.3668714.61177.736–4510,50121.810,15222.234915.246–5514,21229.513,47929.473332.056–6512,71226.411,80225.891039.7Mean age (Standard deviation)45.81 (12.23)45.60 (12.23)49.83 (11.42)EducationPrimary school or less15,03131.214,03930.699243.2High school/Technical school c 26,28654.625,17454.9111248.5University680714.1661714.41908.3OccupationManagerial/Professional11,16123.210,91123.825010.9Office worker26365.525775.6592.6Farming/Forestry6221.35981.3241.0Sales/industry/cleaning/other11,78224.511,41924.936315.8Unknown21,92345.620,32544.3159869.7Social benefits d 16,35634.015,17533.1118151.5Disability pension761315.8701015.360326.3Clinical characteristicsThe number of somatic comorbidities041,09685.439,58486.4151265.91583412.2527411.556024.4≥ 211942.59722.12229.6The number of SUD comorbidities045,23994.243,41294.7182779.6125845.421564.742818.7≥ 23010.62620.6391.7The number of psychiatric comorbidities (excluding SUDs)039,90082.938,09483.1180678.71611012.7574712.536315.8≥ 221144.415464.41255.4Use of gabapentinoids4911.04451.0462.0Use of opioids9442.08361.81084.7Abbreviation: BZDR, benzodiazepines and related drugs. a Within the total number of non‐survivors there were both BZDR users (N = 1314) and non‐users that is, reference group (N = 980). b Age measured at baseline. c Vocational school, that is, had more schooling than 9 years. d Receipt of social benefits, including basic social assistance, labor market subsidy, basic unemployment allowance, national pension, and study grants, maternity allowance, paternity allowance, parental allowance, and child home care allowance during year 2005. Baseline characteristics of BZDR users and stratified by mortality status in 5‐year follow‐up. Abbreviation: BZDR, benzodiazepines and related drugs. Within the total number of non‐survivors there were both BZDR users (N = 1314) and non‐users that is, reference group (N = 980). Age measured at baseline. Vocational school, that is, had more schooling than 9 years. Receipt of social benefits, including basic social assistance, labor market subsidy, basic unemployment allowance, national pension, and study grants, maternity allowance, paternity allowance, parental allowance, and child home care allowance during year 2005. During the follow‐up up to 5 years, 4381 (9.1%) of the subjects had periods of very high‐dose BZDR use, 24,469 (50.8%) had medium to high‐dose use, 43,751 (90.9%) had periods of low dose use, and 43,244 (89.9%) subjects had non‐use of BZDRs, that is, they discontinued use either for shorter or longer term. The dose re‐evaluation in every dispensing is depicted more closely in Figure S2. Most of the very high dose use periods were characterized as the use of more than one BZDRs concomitantly, that is, polyuse (4.14%, N = 1992), as seen in Table 2. The number of persons using clonazepam with very high dose (1.54%, N = 741) was remarkable, followed by alprazolam (1.08%, N = 519). Oxazepam, however, was quite rare in very‐high dose group (0.07%, N = 34). TABLE 2The number of users for each dose category and specific drugs used during the follow‐up.Low dose (< 1.0 DDDs/day) N = 43751 a Medium to high‐dose (1 to 3 DDDs/day) N = 24,469Very high‐dose (≥ 3.0 DDDs/day) N = 4381POLYUSE (> 2 BZDRs)1117391992Clonazepam (N03AE01)14072207741Alprazolam (N05BA12)43822773519Diazepam and temazepam (N05BA01, N05CD07)37738434Clonazepam and alprazolam (N03AE01, N05BA12)9119313Diazepam (N05BA01)45224081286Zopiclone (N05CF01)20,708457784Oxazepam (N05BA04)12,283190134Lorazepam (N05BA06)1117200< 5 a The number of users in each column is larger than the total number of users per dose category as the same person may have used multiple BZDRs during the follow‐up. The number of users for each dose category and specific drugs used during the follow‐up. The number of users in each column is larger than the total number of users per dose category as the same person may have used multiple BZDRs during the follow‐up. Mortality During the 5‐year follow‐up, altogether 2294 people died (4.77% of the cohort). Of them, 1314 died during BZDR use, 980 during non‐use. Risk of mortality associated with different combinations and amounts of BZDRs used is seen more precisely in Figure 1. Any BZDR use was associated with a 36% increased risk of death (aHR 1.28, 95% CI 1.16–1.41) compared to non‐use. The increased risk was observed for BZDRs (aHR 1.40, 95% CI 1.25–1.58) and the combination of BZD with Z‐Drug (aHR 1.86, 95% CI 1.56–2.22) but not with Z‐drugs alone. The non‐survivors had a higher prevalence of somatic and psychiatric comorbidities than survivors. Risk of mortality associated with different combinations and amounts of BZDRs used. The reference category was non‐use of BZDRs, and the results were adjusted for sex, age, level of education, receiving social benefits, the use of gabapentinoids and opioids, disability pension status, number of diagnosed psychiatric disorders, number of diagnosed substance use disorders and Charlson's comorbidity score. aHR, adjusted hazard ratio; BZDRs, benzodiazepines and related drugs; CI, confidence interval; PY, person‐years. The distribution of the mortality causes among the deceased is described in Figure 2. As shown in Figure 2, deaths per 1000 PY were 6.69 for low dose use, 14.58 for medium to high dose use and 23.70 for very high dose use. For external causes, that is, suicides, accidents and overdoses, there were a total of 447 deaths and for natural causes 852 deaths (unspecified cause of death, N = 15). The most common reason for external death overall were accidents (N = 272). In low dose, the risk of all‐cause mortality was not elevated (aHR 1.00, CI 0.90–1.12) compared to the time periods when BZDRs were not used. Both medium to high‐dose use (aHR 1.58, CI 1.36–1.74) and very high‐dose use (aHR 2.68, 95% CI 2.20–3.26), in turn, were distinctly associated with an increased risk of all‐cause mortality compared to the time periods of no BZDR use. Medium to high dose use was associated with an increased risk of both external causes (aHR 2.29, CI 1.84–2.85) and natural causes (aHR 1.31, CI 1.13–1.53). Risk of mortality associated with low, medium to high and very high‐dose BZDR use compared to non‐use of BZDRs by cause of death. Adjusted for temporal order of treatments, sex, age, disability pension, education, receipt of social benefits, the use of gabapentinoids and opioids, number of diagnosed psychiatric disorders, number of diagnosed substance use disorders and Charlson's comorbidity score. aHR, adjusted hazard ratio; CI, confidence interval; PY, person‐years. The highest risk of death for medium to high dose use was observed for overdose deaths (aHR 3.34, CI 2.18–5.12), followed by suicides (aHR 2.43, CI 1.71–3.46) and accidents (aHR 2.24, CI 1.68–2.97). Very‐high dose BZDR use was associated with a higher risk of external‐cause mortality (aHR 4.04, CI 3.06–5.32) than with natural causes (aHR 1.67, CI 1.23–2.28). The highest risk estimates for very high‐dose BZDR use were observed for overdose deaths (aHR 6.18, CI 3.77–10.12), followed by suicides (aHR 4.46, CI 2.79–7.13), and accidents (aHR 3.77, CI 2.66–5.35). Discussion In this nationwide register‐based follow‐up study, use of benzodiazepines and Z‐drugs (BZDRs) above the recommended defined daily doses compared to time‐periods of non‐use was associated with increased risk of death in a dose‐related manner. Very high dose BZDR use, that is, ≥ 3.0 DDDs/day, was associated with a six‐fold increased risk of overdose death and over four‐fold increased risk of suicide. Our results emphasize the need to intensify prevention and support from health care system, as these external deaths might be prevented. Our study shows that the mortality risk strongly associates with higher doses and is often characterised by concomitant use of more than one BZDRs at the same time. Most of the previous studies of the subject have focused on the use of BZDRs with other medications, such as opioids [35, 36, 37, 38] or other psychotropic medications [15, 39, 40]. In our study, the association with mortality was highest when using three or more BZDRs at the same time or when using benzodiazepine in combination with a Z‐drug. This pattern of using multiple benzodiazepines at the same time is not in line with guidelines and should always be fully explained as off‐label use. Although BZDRs are considered rather safe drugs per se and for example, overdose deaths primarily occur in combination with other sedatives [41, 42], high‐dose BZDR use may by itself be a marker of risk behavior, poor functional level or active and/or poorly treated substance use disorder. This may partly explain the increased risks of BZDR use and overdose deaths. As our previous study in 2024 [25] already indicated, those patients who initiate their BZDRuse with more than two BZDRs or with clonazepam (used for other than epileptic indications) and alprazolam seem to be most likely to escalate to very high dose use [43, 44, 45]. Knowing the high potency and misuse potential of the two last‐mentioned, careful clinical consideration should be paid when initiating them in the first place. The low dose BZDR use, that is, < 1.0 Defined Daily Doses (DDDs) per day, was not associated with increased risk of mortality when compared to the time periods when BZDRs were not used. This finding gives an interesting addition to the discussions on whether BZDRs are always purely harmful. Still, while the fatal outcomes of the low dose use might be somewhat uncommon, other harmful risks undeniably still exist as benzodiazepines are linked for example, with dependence and protracted withdrawal syndromes, even at therapeutic doses [13, 46]. The harms and side‐effects are emphasized especially in the elderly [47, 48]. The focus of future studies should therefore perhaps concentrate more on this specific aspect, and more information about the dose‐dependent use is clearly needed. In summary, high‐dose benzodiazepine use poses significant mortality risks, and the risk seems to be also dose‐related. Addressing these characteristics necessitates concerted efforts to re‐evaluate prescribing habits, enhance patient education, and undertake research into safer management practices for individuals on these medications. Strengths and Limitations The information used in this study is based on the Social Insurance Institution's register data and the advanced Finnish healthcare system. The cohort is large and nationwide. For constructing drug use periods of BZDR use and further categorizing it according to time‐varying dose we used PRE2DUP method and the classification system of BZDRs used included the whole range of benzodiazepine‐related substances and correlating factors. However, our data does not capture short‐term changes in clinical status and the actual indications for BZDR use except epilepsy were lacking, which may lead to bias in attributing events to benzodiazepine exposure. Evidence also suggests that the existing classification systems have their own limitations and there are some overlaps between different categories for example, in mortality causes. Although we adjusted results for sex, age, education, social benefits, gabapentinoid and opioid use, disability pension status, number of psychiatric and substance use disorders, and Charlson comorbidity score, residual confounding may remain. Our data was based specifically on prescription medications. As medications can also be bought from illegal markets without a doctor's prescription and some smaller BZDRs packages are not reimbursed, the data coverage may be adversely affected. The correlation of a medication's actual use and its purchase can affect the numbers, but as some studies show, this seems to be a minor concern [49]. Data for this study was gathered in 2006–2011. BZDR use overall has decreased in Nordic countries during 2000–2020 but there are considerable differences between countries and also trends in use vary by age [7]. To our knowledge, there are no studies reporting whether clinical characteristics of BZDR users have changed over these years, for example whether prescription patterns have changed toward lower or higher risk groups, for example, in terms of overdoses for which we found the highest relative risk. Due to reimbursement system cost savings, reimbursement of hypnotic benzodiazepines ended in December 2012 and consequently, extending follow‐up time beyond that would have resulted in exposure misclassification. The mandatory electronic prescription (e‐prescription) was introduced in 2017, and since then it has significantly improved prescribers' access to information about patients' overall medication use and other prescriptions (e.g., to avoid duplicate prescriptions). In addition, certain diagnoses, especially those related to substance use are still underreported and underdiagnosed among clinicians. In any case, the problems related to long‐term and high dose use of BZDRs are still clinically relevant. In this study, we used the Cox proportional hazards model for analysis and reported hazard ratios as the effect measure. Changes in HRs over time could be assessed by dividing follow‐up time to consequent years. However, instead of stable exposure groups, we had time‐varying exposure, meaning that people could change from use to non‐use and between different dose categories. This reduces the possibility that risk changes over time as also exposure categories can change. The analysis for temporal order of exposures (dose categories vs. non‐use) was adjusted. Median order of exposures was first low‐dose (median 1, IQR 1–1), followed by either non‐use (median 2, IQR 2–2) or medium‐high dose (median 2, IQR 1–5), and then very‐high dose (median 4, IQR 2–7). For more specific information about the survival curves of dose categories see Figure S3. Conclusions The association between high‐dose benzodiazepine use and mortality risk is a significant concern within the medical community, particularly as research indicates that both dosage and concurrent use with other substances can exacerbate this risk. Persons using BZDRs in very high doses, that is, ≥ 3.0 DDDs/day, are in six‐fold increased risk of overdose deaths four‐fold increased risk of suicide compared to non‐use. The use in low dose, however, does not seem to elevate the risk of mortality. The prescribing practices, for example, comprehensive approach to patient management, including careful monitoring, consideration of discontinuation, and adherence to clinical guidelines, should be standardized to minimize the risks of BZDR use. Author Contributions Concept and design: All authors. Acquisition, analysis, or interpretation of data: Hanna Särkilä and Heidi Taipale. Drafting of the manuscript: Hanna Särkilä and Heidi Taipale. Critical revision of the manuscript for important intellectual content: All authors. Statistical analysis: Hanna Särkilä, Heidi Taipale, and Antti Tanskanen. Obtained funding: Jarmo Hietala. Administrative, technical, or material support: Heidi Taipale, Solja Niemelä, Jarmo Hietala, and Tero Taiminen. Supervision: Heidi Taipale, Solja Niemelä, and Jarmo Hietala. Funding This work was supported by the Turku Psychiatric services (603459) and the Turku University Central Hospital (P3704). Conflicts of Interest H.S. reported receiving personal fees from Duodecim, The Finnish Institute of Occupational Health, and the Finnish Society of Addiction Medicine. H.T. and A.T. reported participating in research projects funded by grants from Janssen‐Cilag to their employment institution outside the submitted work. H.T. reported receiving grants and personal fees from Gedeon Richter, Janssen, Lundbeck and Otsuka. T.T. reported receiving grants and personal fees from Duodecim Publishing Company, Finnish Psychiatric Association, Janssen‐Cilag, Kuopio University Hospital, Law firm Jarkko Männistö, Lundbeck and University of Turku. S.N. reported personal fees from DNE Pharma/Navamedic, Lundbeck, Recordati, Otsuka and Takeda outside the submitted work. J.H. reported receiving research grants from the Social Insurance Institution of Finland and personal fees from Lundbeck and Orion during the conduct of the study. No other disclosures were reported. Supporting information Figure S1: Flow chart of the population and inclusion criteria. Figure S2: Description of study design. Figure S3: Time to death in years as Kaplan–Meier curves. BZDR use was modeled as time‐varying exposure. For survival curves, time was zeroed if and when exposure status changed. As exposure was modeled in time‐varying manner and most people changed between exposure groups, reaching maximum follow‐up of 5 years within one exposure category was relatively rare and the curves do not extend to the maximum of 5 years. Acknowledgments Open access publishing facilitated by Turun yliopisto, as part of the Wiley ‐ FinELib agreement. Data Availability Statement Hanna Särkilä and Heidi Taipale had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis. The data that support the findings of this study are available from the Finnish government agencies, but restrictions apply to the availability of these data, which were used under license for the current study and so are not publicly available. References Associated Data Supplementary Materials Figure S1: Flow chart of the population and inclusion criteria. Figure S2: Description of study design. Figure S3: Time to death in years as Kaplan–Meier curves. BZDR use was modeled as time‐varying exposure. For survival curves, time was zeroed if and when exposure status changed. As exposure was modeled in time‐varying manner and most people changed between exposure groups, reaching maximum follow‐up of 5 years within one exposure category was relatively rare and the curves do not extend to the maximum of 5 years. Data Availability Statement Hanna Särkilä and Heidi Taipale had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis. The data that support the findings of this study are available from the Finnish government agencies, but restrictions apply to the availability of these data, which were used under license for the current study and so are not publicly available.