Swedish Nationwide Time Series Analysis of Influenza and Suicide Deaths from 1910 to 1978
Total Page:16
File Type:pdf, Size:1020Kb
Open access Original research BMJ Open: first published as 10.1136/bmjopen-2021-049302 on 7 July 2021. Downloaded from Swedish nationwide time series analysis of influenza and suicide deaths from 1910 to 1978 Christian Rück , David Mataix- Cols, Kinda Malki, Mats Adler, Oskar Flygare , Bo Runeson, Anna Sidorchuk To cite: Rück C, Mataix- ABSTRACT Strengths and limitations of this study Cols D, Malki K, et al. Swedish Objectives There is concern that the COVID-19 pandemic nationwide time series analysis will be associated with an increase in suicides, but ► To our knowledge, this was the first study to inves- of influenza and suicide deaths evidence supporting a link between pandemics and suicide from 1910 to 1978. BMJ Open tigate the association between influenza deaths and is limited. Using data from the three influenza pandemics 2021;11:e049302. doi:10.1136/ suicide across several pandemics. of the 20th century, we aimed to investigate whether an bmjopen-2021-049302 ► The large amount of nationwide data on influenza association exists between influenza deaths and suicide and suicide death rates covering 1910–1978 is a ► Prepublication history and deaths. strength of the study. additional supplemental material Design Time series analysis. ► To guard against effects of changes in the recording for this paper are available Setting Sweden. online. To view these files, of causes of death, we created a series of dummy Participants Deaths from influenza and suicides please visit the journal online variables for each corresponding period, but no sig- extracted from the Statistical Yearbook of Sweden for (http:// dx. doi. org/ 10. 1136/ nificant effects were found. 1910–1978, covering three pandemics (the Spanish bmjopen- 2021- 049302). ► No historical data with higher temporal resolution influenza, the Asian influenza and the Hong Kong than yearly data could affect the results if the time Received 22 January 2021 influenza). sequence of association between changes in influ- Accepted 21 June 2021 Main outcome measures Annual suicide rates in Sweden enza death and suicide differ from the chosen time among the whole population, men and women. Non- linear interval. autoregressive distributed lag models was implemented to explore if there is a short- term and/or long- term relationship of increases and decreases in influenza death http://bmjopen.bmj.com/ rates with suicide rates during 1910–1978. Research have raised concerns of a possible Results Between 1910 and 1978, there was no evidence increase in suicide rates.5–7 Such concerns of either short- term or long- term significant associations originate from a combination of known risk between influenza death rates and changes in suicides factors for suicide, including the impacts of (β coefficients of 0.00002, p=0.931 and β=0.00103, p=0.764 for short- term relationship of increases and social distancing and disconnectedness, an decreases in influenza death rates, respectively, with economic downturn, the decreased access suicide rates, and β=−0.0002, p=0.998 and β=0.00211, to mental health services and increased p=0.962 for long- term relationship of increases and access to lethal means exemplified by an on September 24, 2021 by guest. Protected copyright. decreases in influenza death rates, respectively, with increase in gun purchases in the USA.8 For suicide rates). The same pattern emerged in separate example, a study of the economic recession analyses for men and women. in USA 2007–2009 found that for every Conclusions We found no evidence of short- term or percentage point increase in the unem- long-term association between influenza death rates and ployment rate, there was about a 1.6% © Author(s) (or their suicide death rates across three 20th century pandemics. 9 10 employer(s)) 2021. Re- use increase in the suicide rate. These find- permitted under CC BY-NC. No ings have been questioned. A study using commercial re- use. See rights INTRODUCTION an interrupted time- series analysis taking, and permissions. Published by Various international surveys have docu- for example, seasonality and long- term BMJ. mented a negative impact of the COVID-19 trends into account, found little evidence Centre for Psychiatry Research, pandemic on the population’s mental that the recession resulted in net excess Department of Clinical 11 Neuroscience, Karolinska health, with increased levels of psycholog- suicides across all age and gender groups. Institutet & Stockholm ical stress, psychiatric symptoms, insomnia However, there was some evidence of Health Care Services, Region and alcohol consumption.1–4 Whether excess suicides among men aged 65 years Stockholm, Stockholm, Sweden these acute impacts will persist long term and above and young women. Another Correspondence to is currently unknown. The Royal College of study failed to find an increase in suicide Dr Christian Rück; Psychiatrists in the UK and the WHO as well rates in Sweden during the two most recent christian. ruck@ ki. se as the International Academy of Suicide economic recessions.12 Rück C, et al. BMJ Open 2021;11:e049302. doi:10.1136/bmjopen-2021-049302 1 Open access BMJ Open: first published as 10.1136/bmjopen-2021-049302 on 7 July 2021. Downloaded from While the concern is widespread, there is currently SARS,21 we additionally report on associations sepa- little evidence to support a clear association between rately in men and women. the ongoing COVID-19 outbreak and an increased risk of suicide. Data from the US Centers for Disease Control and Prevention show that suicide decreased in 2020, METHODS compared with the preceding year.13 Data from April to Data acquisition October 2020 from the UK did not show an increase of Annual data on influenza death rates and suicide rates 14 suicides. The first months of the COVID-19 pandemic were estimated based on information from the Statis- 15 in 21 countries was studied by Pirkis and colleagues and tical Yearbook of Sweden from 1910 to 1978.24 Over this overall there was little support for an increase in suicides. period, Sweden experienced three influenza pandemics Other reports highlight that the outcomes differ and that that occurred during different sociopolitical contexts, 16 certain minorities may be at higher risk. namely, the Spanish influenza (1918–1920, with the Historic US mortality data from the largest pandemic first case appearing in Sweden in June 1918), the Asian in the 20th century, the Spanish influenza, showed that influenza (1957–1958, being first documented in August the Spanish influenza was associated with an increase in 1957) and the Hong Kong influenza (1968–1969, starting suicides but those effects may have been mitigated by a in the autumn of 1968).20 The yearbooks were produced 17 decline in alcohol consumption. Suicide during the by Statistics Sweden, a governmental agency respon- same pandemic in Taiwan, where the Japanese Colonial sible for the official statistics in Sweden, with a history Government implemented physical distancing, school of population statistics going back to the 18th century.24 closings and prohibited religious activities, was studied in For each year from 1910 to 1978, we retrieved informa- 18 a paper by Chang and colleagues. They reported a small tion on the total population of Sweden, the number of and short- lived increase in suicides in the second wave of deaths by influenza (if death cause was indicated as ‘influ- the pandemic. A study of the impact of social distancing enza’) and the number of suicides (if death cause was measures on suicide in 1918 in the USA showed that indicated as ‘suicide’), as well as the corresponding data increasing social distancing was associated with increased separately for men and women. Since yearbooks reported suicide rates independent of the influenza mortality data retrospectively for several years prior to the year 19 rate. In Sweden, there were measures in place during each book was published, we checked the correctness of the Spanish influenza to minimise the spread of the retrieved data by comparing yearbooks with overlapping disease, such as school closings, and public gatherings, reporting periods. We constructed influenza mortality cinemas and religious meetings were temporarily stopped rates and suicide rates per 100 000 inhabitants for each 20 in some places. year by dividing the number of deaths by influenza and, The outbreak of the severe acute respiratory syndrome separately, suicide, by the total number of individuals http://bmjopen.bmj.com/ (SARS) epidemic in 2003 in Hong Kong was according registered in Sweden in a corresponding year and then to two studies associated with an increase of suicide in multiplying by 100 000. The corresponding rates for men 21 22 the population above the age of 65 years. In the and women were constructed likewise. In addition, we first of these two reports, the association was only statis- collected information on the changes in registration of 21 tically significant in elderly women. Another study of deaths in Sweden that included the cause of death clas- suicides during SARS suggested that disconnectedness sification based on the Bertillon criteria (prior to 1931), and fear of contracting SARS were more prevalent in the new classification introduced in cooperation with older adult SARS- related suicide victims than non-SARS- statistical authorities from other Nordic countries (1931– on September 24, 2021 by guest. Protected copyright. related suicide victims. However, this was based on a 1950), the International Classification of Diseases (ICD) 23 small number of suicides. To summarise, at present, Sixth Revision (ICD-6; 1951–1957), ICD-7 (1958–1968) our understanding of the effects of pandemics on suicide and ICD-8 (1969–1978).25 To capture a potential effect rates is very limited. of changes in classification, we created a series of dummy The availability of high- quality historical data on variables for each corresponding period, but these were mortality due to influenza and suicide in Sweden only kept in the models if statistically significant.