Study population and sampling

The study population comprises all Swedish residents living in all 21 regions, each with its regional government responsible for public healthcare and public transportation [26,27]. All age groups and genders were included, and no exclusion criteria were applied. While the overall number of Swedish residents was 10,378,483 in 2020 (with 50.32% males), it increased to 10,443,100 in 2021 (males – 50.33%), and further to 10,514,719 in 2022 (males – 50.35%). The demographic distribution of the Swedish population by 21 regions for the years 2020, 2021, and 2022 is available in the S1 Data [28]. A detailed description of Sweden’s geographical administrative division [29] can be found in the S2 Table. The data was included from individuals in Sweden who had been dispensed the selected psychotropic drugs from 2014 to 2022 to analyse changes before and during the COVID-19 pandemic. All data were collected at the regional level in all 21 Swedish counties [30].

Data sources – general information

The source data used in this study were obtained from publicly available Swedish Government databases, accessed on June 14, 2024 (medicines used and socioeconomic factors), and August 2, 2024 (additional information for regression analysis).

Data sources – measures of psychotropic medications (dependent variable)

The drug prescribing data were obtained from the National Prescribed Drug Register, maintained by the National Board of Health and Welfare (Socialstyrelsen) [31,32]. Hospital prescribing and OTC medications were not included in this database. The specific drug classes, according to the Anatomical Therapeutic Chemical (ATC) Classification system, included in this study are: N05B ‘Anxiolytics’; N05C ‘Hypnotics and sedatives’; N06A ‘Antidepressants’; N07B ‘Drugs used in addictive disorders’; N07BA ‘Drugs used in nicotine dependence’; N07BB ‘Drugs used in alcohol dependence’; N07BC ‘Drugs used in opioid dependence’.

To quantify drug prescribing patterns, ‘the number of patients per 100,000 inhabitants’ was used as a proxy measure for ‘the mental health drug prescribing rates’ for each ATC group at the county level for each year. Two periods were selected: the pre-COVID-19 (2015–2019) and the COVID-19 (2020–2022); see the S3 Text for detailed information. The range of years from 2014 to 2019 was included for the baseline trend, and 2020–2022 for pandemic year changes. Data on post-pandemic years were excluded.

Data sources – measures of socio-economic factors (independent variables)

All data on variables presented below were collected from the National Board of Health and Welfare [30]. The data on COVID-19 patients in hospitals was chosen as a proxy measure for the disease’s spread and labelled as ‘the number of COVID-19 patients in hospital per 1,000 population’ [30]. During the pandemic, it covered all COVID-19 patients in inpatient care, including intensive care units (ICU). The mean income and the post-school education levels were chosen to measure socio-economic factors. The data for the ‘mean income’ variable, presented in Swedish Krona (SEK) in thousands, was obtained from the Household Finances database of Statistics Sweden (Statistiska centralbyrån, SCB) [33]. The average yearly after-tax income (for both genders, aged 20 and older) for each year from 2020−2022, by region, was downloaded. The data for ‘the proportion of the population with post-school education’ variable was obtained from the Education and Research Database of Statistics Sweden [34], covering both genders, from 16 to 74 years old, who completed any post-secondary education. The distance from the capital of each county to Copenhagen, Denmark, was included as one covariate because personal trans-border importing of alcoholic beverages and tobacco products is quite popular in Sweden [35]. This is due to the Swedish monopoly system in alcohol retailing (Systembolaget) and high taxes on these products compared to neighbouring countries [36,37]. The primary sources of purchase remain countries south of Sweden, and the most convenient connection is often via Denmark. The variable, coded as ‘the distance to Copenhagen in kilometres (km)’, was chosen to investigate the association between cross-border purchasing and changes in drug prescribing with regional variation. The S4 Text provides a detailed explanation of the socio-economic data measures, while the source data for analyses performed in this study can be found in the S5 Data. From among the available data, any factors that were not reflective of changes in psychotropic medicine prescribing, according to the study’s aim and objectives (e.g., temperature or percentage of the population testing positive for COVID-19), were excluded.

Statistical analysis of data

We measured the drug prescribing rates using time-series plots and map plots for each drug group. The different trends of the actual and expected numbers of patients prescribed the chosen drugs during the COVID-19 pandemic were described using time-series plots. The expected number of patients for 2020, 2021, and 2022 was calculated based on the average patient growth rates from 2015 to 2019. Moreover, the percentage differences between actual and expected patient numbers from 2020 to 2022 across the 21 regions were described using spatial visualization, with map plots to examine regional variation.

The relationships between drug prescribing patterns and four independent variables were presented using scatter plots for 2020, 2021, and 2022. The N07BA ‘Drugs used in nicotine dependence’ was chosen to test the correlation, as the analysis had shown a reduction in the consumption of this type of medication throughout the pandemic. For the scatter plots, the ‘percentage difference in the actual number of patients and the expected number of patients’ (previously presented) has been tested against four key variables.

A panel regression model was applied after examining the scatter plots to investigate correlations across the entire period from 2020 to 2022. Using a panel regression, the unobserved heterogeneity was controlled, and individual effects and time effects were captured to explain the relationships between the variables across different regions that change over time. Additionally, for each year during the pandemic, a multiple linear regression model was used to examine the correlations between the primary dependent variable and the three key independent variables. The fourth independent variable (distance to Copenhagen in km) was excluded from the fixed effect model due to its time invariance. All models were weighted by each county’s population (i.e., the number of inhabitants).

To mitigate the selection bias regarding patients using psychotropic medications, all age and gender groups from all 21 regions of Sweden were included. The information bias was mitigated by using data from standardized national registries. The measurement bias was mitigated by using the same measurement approach (‘the number of patients per 100,000 inhabitants’ to quantify drug prescribing patterns) across all regions and periods. The data were weighted by population in the regression analysis to mitigate the regional bias.

We used the statistical computing R software version 4.4.0 (2024-04-24) [38]. The maps were created using swemaps2 in R [39].

Ethical considerations

The authors had no access to information that could identify any individuals during or after data collection. The study received ethics approval from the University Research Ethics Committee of the University of Sheffield (registration number 220198354, reference number 058312).

Regression analysis

The whole pandemic period was analysed using the panel regression. In panel regression, the Hausman test was used to determine whether a fixed-effects or random-effects panel regression model was more appropriate for the analysis (S6 Table). As the Hausman test results appear statistically significant (χ² = 24.90, p < 0.001), the fixed-effects model was chosen for panel regression for the whole pandemic period of 2020 to 2022 (Table 1). Its results show a high R² value of 0.97 and an even higher adjusted R² value of 0.95, indicating that the model explains a large proportion of variance over time in the dependent variable (N07BA prescribing). Among the independent variables, only the mean income variable showed a statistically significant relationship with nicotine dependence drug prescribing, with a negative coefficient (β = -0.02, 95% CI: [-0.03, -0.01], p < 0.01).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Fixed-effects panel regression results for nicotine dependence drug prescribing.

https://doi.org/10.1371/journal.pone.0330081.t001

Additionally, the multiple linear regression models were used to perform regression analyses for the separate years 2020, 2021, and 2022 (S7 Table). All of these models show low R² values. No significant correlations were found between any of the investigated covariates and the differences in patients with N07BA in these regressions.

Prescribing trends of psychotropic medicines in Sweden

The anxiolytic (N05B) prescribing in Sweden demonstrated a decreasing trend with a slight recovery in 2022. These findings align with the decrease in prescribing among newly diagnosed patients in 2020 in Germany [15]. It can be assumed that a less restrictive approach may reduce social restriction-related anxiety. Moreover, the Swedish society’s high trust in the government’s pandemic response [40] may also have contributed to the reduction of pandemic-induced anxiety and the need for anxiolytics. Another possible explanation is that the healthcare system in Sweden may have encouraged the use of non-pharmacological interventions for anxiety, such as stress management, rather than medication, while maintaining healthcare access.

The relatively stable prescribing patterns of hypnotic and sedative medications (N05C) observed in this study stand in contrast to the dynamic and shifting trends reported in international research findings. Globally, a significant increase in insomnia and sleep disturbances was observed [5,41]. In Denmark, the increased prescribing of benzodiazepine-related hypnotics was reported [21]. This stability pattern, followed by a slight decrease in most regions of Sweden throughout the pandemic, suggests that the Swedish population may have experienced a gradual recovery from insomnia. The possible explanation is that maintaining daily routines, facilitated by Sweden’s less restrictive approach to the pandemic, may have played a crucial role in preserving sleep patterns. This interpretation is supported by research highlighting the maintenance of sleep quality, especially during stressful events [42].

The observed regional variations in N05B and N05C prescribing also highlight the importance of considering local cultural factors in mental health interventions. For instance, less populated regions may have less pandemic-related anxiety and insomnia. The regionally differentiated coping mechanisms and unique local healthcare approaches may also play an important role.

The national data on antidepressant (N06A) prescribing shows a continuous increase in the years 2021 and 2022, which aligns with the global pattern of increasing depression during the COVID-19 pandemic [2,4,43]. The findings also correspond with the increased antidepressant prescribing patterns in the Scandinavian countries [21], the UK [22], Canada [24], and the new use of antidepressants during the pandemic in Australia [23]. Nevertheless, Sweden experienced a slight decrease at the start of the pandemic, which aligns with the study indicating no differences in depression symptoms during the first four weeks of the COVID-19 pandemic compared to the pre-pandemic period [3]. The Swedish population’s higher threshold for seeking medical intervention may have led to delayed anxiolytic prescribing in the early pandemic. While almost all counties experienced increased prescribing levels in 2022, regional variations persisted, particularly in 2021. The highest prescribing rates were observed in Västerbottens and regions in East Sweden, possibly due to variations in age distribution and population characteristics, such as the at-risk population and individual health-seeking behaviours. The local economic context could be another factor to consider because increased depression rates are highly associated with economic recessions and unemployment [44]. We took a closer look at the decreasing anxiolytic and increasing antidepressant prescribing. However, the number of medicines found within particular subgroups of N05B and N06C (the fifth level of the ATC classification) proved to be very small at the regional level, so even small changes, expressed in absolute measures, could have impacted the findings and subsequently influenced the interpretation of the results. An in-depth analysis of medicines prescribing at lower levels of ATC classification was beyond the scope of the current study.

The dramatic decrease in prescribing of all kinds of N07B drugs used in addictive disorders seems contrary to the international findings on increased addiction and use of addictive substances during the pandemic [20,45]. It is interesting to note that distinct regional variations were not seen in this drug group, except for Västernorrland and Kalmar counties (increased drug prescribing in 2020). Among the subgroups of N07B, the nicotine dependence drugs group (N07BA) showed the most striking reduction in prescribing. The smoking behaviours and cessation efforts during the COVID-19 pandemic reported varying results across the world, including both increased and decreased trends depending on the underlying reasons. The fear of COVID-19 led to a reduction in smoking, whereas mental health conditions like anxiety exacerbated the risky behaviour [46–50]. However, the current study findings align with the decrease in varenicline and nicotine replacement therapy (NRT) in Australia [51] and not meeting the expected prescription rates in the USA [52]. Possibly, the decreasing trends of the utilization of nicotine-dependence drugs observed in this analysis may be due to a shift towards non-prescription cessation therapy preference during the pandemic. Alternatively, it could indicate an increase in smoking during the pandemic or untreated nicotine dependence, both of which would have important public health implications. Yet another explanation could be the death of smokers during the COVID-19 pandemic.

The slight increases in the prescribing of medicines used in alcohol dependence treatment (N07BB), particularly in females, are consistent with the global trends of increased alcohol consumption during the pandemic [20,53]. This could indicate that some factors, such as gender-specific stressors (e.g., females are more stressed than males) [14], coping mechanisms applied during the COVID-19 pandemic, or shifting to home-based drinking culture (due to limitations in social gatherings) could have induced variations of N07BB prescribing between regions and genders.

Contrary to increased opioid use during the pandemic in some countries [20,45], a gradual decline in N07BC prescribing (drugs used in opioid dependence) has been observed in this analysis. It can be assumed that the Swedish healthcare system may have already implemented successful prevention and treatment in harm reduction programs. Hence, Sweden’s unique pandemic and healthcare approaches may offer essential insights into opioid management during crises.

In summary, apart from antidepressants, the prescribing of psychotropics showed a decreasing trend during the pandemic, particularly in the N07BA group of drugs used for nicotine dependence. The diverse trends across different drug classes and regional variations reflect the complexity of factors surrounding the psychotropic medication prescribing patterns in Sweden during the COVID-19 pandemic. Since healthcare accessibility in Sweden was maintained, these changes could be driven mainly by other factors, such as cultural, local, socio-economic, and demographic factors, rather than by barriers to accessing healthcare.

Relationships between drug prescribing for nicotine dependence and socio-economic characteristics in Swedish regions

We found weak positive associations between the COVID-19 patients at hospital per 1,000 population and the NO7BA prescribing differences in the scatter plots for all three years. Some regions with higher COVID-19 patient numbers at hospitals tended to have N07BA prescribing patterns closer to expected values; however, the relationship was not statistically significant. The findings are consistent with reports from the UK, Italy, Spain, and Germany, where disease severity did not significantly impact the smoking cessation rate during the pandemic [54,55]. They also align with the decrease in attempts and success rate of quitting smoking during the pandemic, compared to the pre-pandemic period [54].

The analysis showed a weak positive relationship between the mean income and the N07BA prescribing differences in all observed years. Although some higher-income counties, such as Västra Götaland, showed less negative prescribing differences, the results varied across other counties and all three years. These weak relationships are contrary to the previously published studies, which reported higher smoking cessation success rates in individuals with higher education and full-time employment in Spain [54], and lower chances of lifestyle changes during the pandemic in unemployed groups in Sweden [9]. It is also essential to consider the social status and disparities of different Swedish community groups, which could be affected by Sweden’s unique pandemic approach [56].

Although some counties with higher post-school education levels, such as Uppsala and Västra Götaland showed a more positive prescribing pattern, the overall correlation is not significant. This partially aligns with the reports from other studies on the increased psychological distress in lower education level groups [43] and a higher rate of quitting smoking in individuals with higher education backgrounds during the pandemic [54]. The relationships between the distance from the capital to Copenhagen and N07BA prescribing remained weak throughout the pandemic years. There may be an impact of travel restrictions on access to cross-border purchasing of nicotine products and alcohol, as Sweden has higher taxes on these products than Denmark or Germany [35,57].

None of the variables showed a significant correlation with the N07BA drug prescribing in the initial testing with scatter plots. A persistent weak relationship and extensive regional variations in all variables across all three years highlight the complex nature of factors affecting the changes in medication prescribing patterns. Other factors, such as regional policies, healthcare literacy, and localised socio-economic factors in different community groups, may also have combined effects on prescribing trends throughout the pandemic.

Interpreting the relationships tested in the regression analysis

A fixed-effects panel regression over the entire period (2020–2022) was employed. As far as the education variable was concerned, the regression model showed an insignificant correlation. Some other studies reported that lower education levels are associated with higher levels of mental health problems [2,43]. In contrast, one study in China found the opposite finding that individuals with higher education are more prone to depression during the pandemic [58]. These inconsistent results highlight the complex role of education levels in health behaviour during crises. In addition to education levels, there could be several mediating factors to the changes in nicotine dependence drug prescribing, which are not captured in the current model.

In our fixed-effects panel regression for 2020–2022, the mean income showed a significant negative correlation. It appears that the income factor is correlated with the temporal variation between regions during that period. Although the results align with the published reports [2,43], mentioning the association between lower socio-economic status and higher risk of mental health symptoms, the mean income variable in this study could not explain the continuous decline in nicotine dependence drug prescribing throughout the pandemic. Potentially, the statistically significant correlation between the mean income and nicotine dependence drug prescribing could be related to several socio-economic mechanisms, including, e.g., reduced access to non-pharmacological interventions among less affluent patients and their financial instability.

Additionally, linear regression was applied for each of the individual years (2020, 2021, 2022). These analyses found no statistically significant relationship between the COVID-19 patients at the hospital and N07BA prescribing patterns. Surprisingly, the tests of the possibility of influence on N07BA prescribing by the disruption in cross-border purchasing factor (using the distance to Copenhagen variable as a covariate in multiple regression analysis for each year) did not show any significant correlation either.

None of these chosen variables could fully explain the decline of N07BA drugs. This also suggests that other variables, not measured in this study, should be considered in future research. Nevertheless, a temporal effect is present throughout the entire period, which is strongly correlated with the income factor.

Public policy implications and recommendations

Most of the psychotropic medications, except antidepressants and hypnotics/sedatives, experienced decreasing prescribing trends during the pandemic, which was the most prominent in drugs used in nicotine dependence (N07BA). None of the studied factors could explain the initial reduced levels in 2020. However, socioeconomic factors, particularly income-related factors, play a significant role in changes in N07BA prescribing during the pandemic period. The negative correlations in the panel regression highlight that health authorities should consider increasing targeted mental health and smoking cessation awareness campaigns tailored to educational backgrounds in regions with higher income. Although the coefficient for the education factor is insignificant, the positive effect highlights that education influences the long-term management of nicotine dependence rather than abrupt changes in prescribing behavior during the pandemic. Implementing economic support could help to reduce the income-related disparities during the pandemic, as well as indirectly support coping with mental health problems arising from socio-economic inequalities. Moreover, policymakers should emphasize developing region-specific mental health or addiction policies, providing autonomy to regional health authorities, and adapting to local circumstances, especially in times of crisis.

The lack of clear correlations between COVID-19 hospitalization rates and N07BA prescribing levels suggests that the disease spread factor also could not explain the changes in drug prescribing trends. Nevertheless, health authorities should incorporate mental health and addiction treatment strategies into health plans as well as develop a flexible regulatory framework for the adaptation of mental health services during emergencies. In the event of future health crises, integrated public health strategies should be developed in conjunction with emergency health mitigation interventions. A real-time and long-term monitoring system implemented during and after the pandemic or other public health emergencies could support addressing mental healthcare needs.

Limitations and future research directions

The study has several limitations. First, the data available in this study were limited and subject to external generalizability constraints. Second, the data exclude OTC and hospital prescribing; therefore, to mitigate potential bias, the study results, including those on the use of N07BA drugs, were interpreted carefully. The lack of including OTC medicines in this study may affect the generalizability of the findings, particularly for vulnerable populations or other patient groups that do not regularly seek and/or obtain professional medical services. Third, the study results reflect the Swedish health system, which may not be generalizable to other countries with their distinct health delivery systems, comprising unique legislative and organizational characteristics, and within diverse clinical and political settings.

Future research, including OTC and non-pharmacological interventions, is recommended to characterize prescribing trends across Sweden. Subsequent studies should investigate other related and additional socioeconomic factors that may influence changes in medication prescribing during health emergencies. A further longitudinal study on the long-term relationships between income and education factors and psychotropic medication prescribing patterns would be helpful. To understand the localized factors and disparities, a detailed analysis at the municipality level could also be performed. Moreover, qualitative studies, including interviews and focus groups, would provide deeper insights into exploring the causal factors. Subgroup analyses are also recommended to identify prescribing patterns in specific groups, such as those based on age, gender, and ethnicity. An in-depth analysis of decreasing anxiolytic and increasing antidepressant prescribing, going down to the lowest levels of the ATC classification, could help to explain both of these intriguing phenomena.