By Rachel Cuell
Introduction
Recently, I have had family and friends struggle with suicidal thoughts. Suicide is the ninth leading cause of death in Canada with 4405 deaths in 2017 (Statistics Canada, 2018). I wondered what factors would lead to suicide but recognized it would be inappropriate to interview individuals who have struggled with suicidal thoughts as such conversations could be a trigger. However, statistical analyses have always fascinated me and I learned I could access health and mortality data from Statistics Canada, and used these to study suicide in Canada.
QUESTION
Do factors such as socioeconomic status, health, hours of daylight, and life satisfaction correlate with suicide rates across Canada?
METHODS AND ANALYSIS
I compiled raw data for eleven variables from Statistics Canada. I grouped them by province, territory and region. The regions were west: AB, BC, and SASK; central: MB, ON and QUE; east: NB, NF, NS, and PEI; and the territories.
I calculated mean suicide rate by province and territory (2000-2015), province, and gender. In order to evaluate the effect of latitude on suicide rates, I split the data into two groups. The territories were used as the “high latitude” group and the provinces were classified as the “low latitude” group. I used t-tests to compare the means of the two groups. I examined associations between suicide rate and demographic and health data by province using Pearson Correlations. I used principal components analyses (PCA) and non-metric multidimensional scaling (NMDS) plots to compare the variables found to significantly correlate with suicide rates. My statistics mentors helped me use R and PRIMER. A p-value of 0.05 or less was considered significant.
Some analyses excluded Nunavut. Because of its high suicide rates, it was a statistical outlier and I wanted to ensure that Nunavut was not overwhelming other possible trends.
STATISTICS
I used a variety of correlations and multivariate statistics to analyze my data. The bar graphs used means and standard error for error bars. Standard error is the standard deviation within the group divided by the square root of the sample size. Simply put, if a data set has a large sample set with little variance within each group, there will be very little error. Next, I did unpaired t-tests which compared means between two variables. Third, I used Pearson correlations which analyzed correlation between two variables without assuming causality. Fourth, I used a non-metric multidimensional scaling plot using Euclidian distance. An NMDS plot is a multivariate ordination statistic and the distance between points describes the similarity among points. In other words, points close together are very similar and points far apart are not similar. To see if the distances in the NMDS plot were statistically significant, I used an analysis of similarities. An ANOSIM is similar to the more commonly known analysis of variance (ANOVA) but instead of using a theoretical distribution, an ANOSIM runs 999 iterations of the data set and yields a p value from real data. Finally, I used a principal component analysis (PCA). A PCA finds underlying variables (known as principle components) that best explain points in the data (Daust, 2018).
RESULTS
Nunavut had the highest suicide rate from 2000-2015 (Figure 1A). Males had at least three times suicide rate than females (Figure 1B) except Manitoba which had a 2.6 male to female rate.
Figure 1. Suicide rate per province from 2000-2015 (A) and by gender (B).
Higher latitudes with less daylight (territories) had a significantly higher suicide rate (Figure 2A; p = 0.02). Even with Nunavut excluded, the difference remained significant (Figure 2B; p = 0.01). I could not access data on the monthly suicide rates by province and territory to study daylight effects more closely.
Figure 2. Suicide rate by daylight. (A) Mean and standard error (p=0.02, t-test). (B) Excluding Nunavut (p=0.01, t-test).
At first glance, high wage seems correlated with high suicide rate (Figures 3A and 3B). When Nunavut is excluded, this trend still exists (Figure 3B). However, if all territories are excluded, there is no longer a significant correlation.
Figure 3. Correlation of suicide rate with weekly wage (A). Nunavut Excluded (B), Rp = 0.67, p=0.02.
In terms of demographics variables, lack of a high school diploma was correlated with a higher suicide rate (Figure 4A). Nunavut’s proportion of people without a high school diploma was twice the rate of all other provinces and territories. It was therefore considered an outlier and excluded. Post-secondary education was not associated with suicide (Figure 4B).
Figure 4. Suicide rate vs. proportion of population lacking a high school diploma. (Excluding Nunavut, Rp = 0.58, p = 0.05; Pearson Correlation, A) and correlation of post-secondary education with suicide rate, Rp = -0.01, p = 1, Pearson Correlation, B).
I evaluated six health-related variables: perceived life satisfaction, mental and physical health, overall health, more than five alcoholic drinks at once, and feeling of belonging. Having five or more alcoholic drinks at a time correlated with a higher suicide rate (Rp = 0.65, p = 0.02;). Populations with perceived overall health of “good” or “excellent” had a lower suicide rate (Rp = -0.75, p = 0.005; Pearson Correlation). The importance of each health-related variable varied significantly between provinces (Figure 5). A PCA was done comparing the variables found to significantly correlate with suicide rates: a lack of a high school diploma, wage, alcoholism, and overall health (Figure 6).
Figure 5. Correlation between health variables; Euclidean distance. (Global R= 0.84, p=0.002, ANOSIM).
Figure 6. PCA plot for health variables that were found to correlate with suicide: a lack of high school diploma, wage, alcoholism, and overall health.
DISCUSSION
In nearly all Canadian regions, males had a suicide rate that was at least three times higher than females. This gender paradox is found globally: Lee et al. (2016) found that although females are more likely to have suicidal thoughts (World Health Organization, 2001), men have a higher suicide rate. A 2008 study suggested that the social construction of gender caused men to be less likely to ask for help when struggling with mental disorders (Payne et al., 2008). Furthermore, Denning et al. (2000) found that men had a greater intent to die than women and used more violent means. Where men were more likely to use firearms and hanging to commit suicide, women were more likely to use drugs and carbon monoxide poisoning.
At first glance, it appears that regions with higher wages have a higher suicide rate. However, the regions with the highest average wages are Nunavut and the Northwest Territories which are very remote. Professionals such as teachers, doctors, and nurses get a very high yearly salary to work in more remote areas (Statistics Canada, 2018). This could mean that there is a large wage discrepancy in those regions: a few individuals earning a very high income who are pushing the average wage up but many more individuals living in poverty.
Across Canada, drinking more than five alcoholic beverages at one time was also correlated with increased suicide rates. This finding is also echoed in other countries. Lee found about one third of reported suicides in the US in 2007 had alcohol involvement (Lee et al., 2016).
My analyses cannot determine whether socioeconomic factors can cause suicide; they can only test whether these factors are associated with suicide. Other variables linked to both suicide and socioeconomic factors may instead explain the correlations. For example, Nunavut has ten times the national average suicide rate and low rate of high school completion. However, studies suggest transitioning from a traditional to modern society can lead to increased suicide rates (Schwarzenthal et al., n.d.; Kumar et al., 2016). After WWII, the Canadian government pressured Inuit to leave their traditional way of life and forced southern values upon them (Government of Nunavut et al., 2010). Nunavut’s traumatic history since that time has increased the rates of domestic abuse, which can lead to higher suicide rates (Chachamovich et al., 2013).
Similarly, Nunavut and NWT have dark winters and higher suicide rates, but this effect may be linked with other factors. Yukon’s suicide rate is similar to provinces which suggests that other factors – not merely daylight – cause higher suicide rates. Further, Lee et al. (2016) determined that there was a higher suicide rate in most countries during sunnier months, contradicting the common belief that more suicides occur during darker seasons.
Next steps for this project would include correlating suicide rates with variables such as historical factors (e.g., residential schools), levels of abuse in and out of homes, age, race, and the daylight variable in finer detail by obtaining monthly data.
CONCLUSIONS
I found higher suicide rates were associated with male gender, higher wage, poor perceived health, fewer annual hours of daylight, alcoholism and a lack of high school education. However, suicide is a complex disease and the variables I analyzed suggest correlation, not causation. Most likely, suicide is linked to factors working at multiple scales including the surrounding society, opportunities and income disparity at the community scale; and mental and physical health, feelings of belonging and meaningful work at an individual scale.
ERRORS
The data I collected from Statistics Canada were not available in the same year, so comparisons were made using different time periods which could have affected the results.
ACKNOWLEDGEMENTS
Thank you LJ MacMillan, Kiri Daust, Karen Price, and Paula Bartemucci .
REFERENCES
Chachamovich, Eduardo, et al. Suicide Report Nunavut English. 1 June 2013, www.documentcloud.org/documents/708953-suicide-report-nunavut-english.html#document/p1. Accessed 12 Apr. 2018.
Daust, K. (2018, April 15). Statistics [Personal interview].
Denning, D G, et al. Method Choice, Intent, and Gender in Completed Suicide. 2000, www.ncbi.nlm.nih.gov/pubmed/11079640. Accessed 12 Apr. 2018.
Government of Nunavut, et al. Nunavut Suicide Prevention Strategy. Oct. 2010. Accessed 12 Apr. 2018.
Kumar, Mohan B, and Amy Nahwegahbow. Aboriginal Peoples Survey, 2012 Past-Year Suicidal Thoughts among off-Reserve First Nations, Métis and Inuit Adults Aged 18 to 25: Prevalence and Associated Characteristics Aboriginal Peoples Survey, 2012 Past-Year Suicidal Thoughts among off-Reserve First Nations, Métis and Inuit Adults Aged 18 to 25: Prevalence and Associated Characteristics. 13 Oct. 2016, www.statcan.gc.ca/pub/89-653-x/89-653-x2016011- eng.htm. Accessed 13 Apr. 2018.
Lee, Lindsay, et al. Suicide. July 2016, ourworldindata.org/suicide. Accessed 13 Apr. 2018. Payne, Sarah, et al. “The Social Construction of Gender and Its Influence on Suicide: a Review of the Literature.” Journal of Mens Health, vol. 5, no. 1, 2008, pp. 23–35., doi:10.1016/j.jomh.2007.11.002. Accessed 12 Apr. 2018.
Schwarzenthal, Miriam J., and Taciano L. Milfont. Suicide and Culture: Exploring Country-Level Relations between Suicide Rates and Dimensions of Cultural Variability. www.researchgate.net/profile/Miriam_Schwarzenthal/publication/311652696_Suicide_and_Culture_Exploring_Country-Level_Relations_between_Suicide_Rates_and_Dimensions_of_Cultural_Variability/links/58529 26608aef7d030a4fda5/Suicide-and-Culture-Exploring-Country-Level-Relations-between- Suicide-Rates-and-Dimensions-of-Cultural-Variability.pdf. Accessed 12 Apr. 2018.
Statistics Canada. Leading Causes of Death, by Sex (Both Sexes). 23 Feb. 2018, www.statcan.gc.ca/tables-tableaux/sum-som/l01/cst01/hlth36a-eng.htm. Accessed 12 Apr. 2018. World Health Organization. Self-Directed Violence. 2001. Accessed 12 Apr. 2018.
Statistics Canada. (2018). Average weekly earnings by industry, monthly, unadjusted for seasonality. Retrieved August 17, 2018, from https://www150.statcan.gc.ca/t1/tbl1/en/cv.action?pid=1410020301
BIBLIOGRAPHY
GraphPad Software. GraphPad QuickCalcs: t Test Calculator. 2018, www.graphpad.com/quickcalcs/ttest1/. Accessed 11 Apr. 2018.
Statistics Canada. Deaths and Mortality Rate (Age Standardization Using 2011 Population), by Selected Grouped Causes and Sex, Canada, Provinces and Territories. 2015, www5.statcan.gc.ca/cansim/a47. Accessed 2 Apr. 2018.
Statistics Canada. Distribution of the Population Aged 25 to 64 (Total and with Aboriginal Identity), by Sex and Highest Certificate, Diploma or Degree, Canada, Provinces and Territories. 2016, www5.statcan.gc.ca/cansim/a47#customizeTab. Accessed 2 Apr. 2018.
Statistics Canada. Average Weekly Earnings (Including Overtime), by Province and Territory. 28 Mar. 2018, www.statcan.gc.ca/tables-tableaux/sum-som/l01/cst01/labr79-eng.htm. Accessed 2 Apr. 2018.
About the Author
My name is Rachel and I live in Smithers, a small town in northwest British Columbia. I’m sixteen years old and I’ve had a passion for science since my first science fair in grade six. I’ve attended CWSF four times in the past and they have all been incredibly inspiring experiences. Mostly, I’ve done projects related to psychology but I’ve also studied Biology and Physics. I love learning about new scientific studies and breakthroughs and I can’t wait to study science in my post-secondary education.