Abstract
BACKGROUND: Food insecurity affects 1 in 8 households in Canada, with serious health consequences. We investigated the association between household food insecurity and all-cause and cause-specific mortality.
METHODS: We assessed the food insecurity status of Canadian adults using the Canadian Community Health Survey 2005–2017 and identified premature deaths among the survey respondents using the Canadian Vital Statistics Database 2005–2017. Applying Cox survival analyses to the linked data sets, we compared adults’ all-cause and cause-specific mortality hazard by their household food insecurity status.
RESULTS: Of the 510 010 adults sampled (3 390 500 person-years), 25 460 died prematurely by 2017. Death rates of food-secure adults and their counterparts experiencing marginal, moderate and severe food insecurity were 736, 752, 834 and 1124 per 100 000 person-years, respectively. The adjusted hazard ratios (HRs) of all-cause premature mortality for marginal, moderate and severe food insecurity were 1.10 (95% confidence interval [CI] 1.03–1.18), 1.11 (95% CI 1.05–1.18) and 1.37 (95% CI 1.27–1.47), respectively. Among adults who died prematurely, those experiencing severe food insecurity died on average 9 years earlier than their food-secure counterparts (age 59.5 v. 68.9 yr). Severe food insecurity was consistently associated with higher mortality across all causes of death except cancers; the association was particularly pronounced for infectious-parasitic diseases (adjusted HR 2.24, 95% CI 1.42–3.55), unintentional injuries (adjusted HR 2.69, 95% CI 2.04–3.56) and suicides (adjusted HR 2.21, 95% CI 1.50–3.24).
INTERPRETATION: Canadian adults from food-insecure households were more likely to die prematurely than their food-secure counterparts. Efforts to reduce premature mortality should consider food insecurity as a relevant social determinant.
Food insecurity — inadequate access to food because of financial constraints — is a public health concern in Canada. 1,2 In 2007/08, 11.3% of households experienced food insecurity; the figure rose to 12.4% by 2011/12.1,2 Food insecurity is associated with negative health outcomes and higher health care spending in a graded fashion, with more severe food insecurity corresponding to worse health and greater health care utilization. 3–5 Food insecurity predicts higher incidence of chronic conditions and poorer management of them.6–14 Food insecurity is also associated with mental disorders,15–18 suicidal thoughts19,20 and suicide attempts.21 Despite much evidence linking food insecurity to poor health, less is known about the association between food insecurity and mortality.
A graded association between severity of food insecurity and all-cause mortality risk was found among working-age adults in Ontario, Canada.22 A study of the US adult population found very low food insecurity — equivalent to severe food insecurity in Canada — was associated with higher all-cause mortality.23 In addition, 2 small studies found food insecurity to be associated with higher mortality odds among HIV-infected adults.24,25 However, none of these studies examined causes of death. More comprehensive, population-based research is needed to advance our understanding of the health burden associated with food insecurity.
Linking vital statistics to multiple cycles of a national health survey, we conducted a population-based retrospective cohort study to assess the association between household food insecurity status and Canadian adults’ all-cause and cause-specific premature mortality.
Methods
Study sample
We linked the 2005–2017 Canadian Community Health Survey (CCHS) to the 2005–2017 Canadian Vital Statistics Database and 2003–2017 Discharge Abstract Database.
The Canadian Community Health Survey is a biennial cross-sectional survey, representing 98% of the Canadian noninstitutionalized population aged 12 years and older. The survey randomly samples households nationwide, comprising roughly 130 000 households per 2-year cycle. The household food security module is a regular CCHS component, but certain jurisdictions chose not to administer the module when it was optional (Appendix 1, available at www.cmaj.ca/lookup/suppl/doi:10.1503/cmaj.190385/-/DC1). CCHS Sharelink is a CCHS sub-sample with about 85% of total respondents, who agreed to have their responses linked to administrative records.26 Statistics Canada created sample weights for CCHS Sharelink to retain population representativeness.26 CCHS Sharelink resembles CCHS in sociodemographic characteristics.26
The Canadian Vital Statistics Database is an administrative database containing the date and primary cause of all deaths registered by Canadian jurisdictions. The cause of death is coded using the Canadian version of the International Statistical Classification of Diseases and Related Health Problems, 10th Revision (ICD-10-CA; Appendix 2, available at www.cmaj.ca/lookup/suppl/doi:10.1503/cmaj.190385/-/DC1),27 which has been extensively applied in mortality research.28,29 Data on all individuals aged 12 years or older who died between 2000 and 2017 are linkable to CCHS data.26,30
The Discharge Abstract Database is an administrative database containing the clinical information of all patients admitted to hospital in Canada, excluding the province of Quebec. For each hospital admission from 1996 to 2018, the Discharge Abstract Database registers the admission and discharge dates and the most responsible diagnosis, coded into ICD-10-CA. To limit the focus to morbidity-related hospital admissions, we omitted admissions related to pregnancy, same-day surgery and prescheduled examinations.
We linked CCHS respondents to Canadian Vital Statistics Database death records; those without a link were presumed alive on Dec. 31, 2017. We also performed one-to-many linkage between CCHS respondents and the Discharge Abstract Database hospital admission records. The linkages of CCHS with the Canadian Vital Statistics Database and the Discharge Abstract Database are of high quality, with few false links.30,31 After linking the 3 data sets, we excluded respondents from jurisdiction cycles that opted out of the food security module (Figure 1). We also excluded children younger than 18 years, seniors older than 82, and individuals with missing food insecurity status or miscoded hospital records. The 2017 CCHS data on territories were not available for analyses. By applying weights from CCHS Sharelink, our sample is representative of the noninstitutionalized adult population aged 18–82 years with valid food security status in the selected jurisdictions and years included in our analyses. To protect confidentiality, we rounded numbers of observations according to Statistics Canada’s rules on analysis of linked data.
Outcome measures
The outcomes were all-cause and cause-specific premature mortality, defined as “death that occurs before the average age of death in a certain population.”32 Life expectancy for 18-year-old Canadians in 2008–2014 was approximately 82 years;33 thus, we regarded deaths as premature if they occurred before age 83 years between the CCHS interview date and Dec. 31, 2017.
Using ICD-10-CA, we separated deaths by “communicable diseases and injuries” from deaths by “noncommunicable diseases.” We categorized the former into “infectious-parasitic diseases,” “unintentional injuries,” “suicides” and “other miscellaneous communicable diseases and injuries.” We categorized the latter into “cancers,” “circulatory-respiratory diseases,” “metabolic-digestive diseases” and “other miscellaneous noncommunicable diseases.” To better gauge the effect of intervention, we identified causes of death considered to be potentially avoidable through adequate public health prevention and medical treatment (Appendix 3, available at www.cmaj.ca/lookup/suppl/doi:10.1503/cmaj.190385/-/DC1).34 The concept of avoidable mortality was introduced in 1976 as “unnecessary untimely deaths” to measure quality of medical care, and since then has been widely used to assess health system performance.35,36
Food insecurity measure
The exposure was 12-month household food insecurity status. This was based on a validated 18-item scale of severity developed by the US Department of Agriculture to assess the adequacy of households’ access to food over the past 12 months (Appendix 4, available at www.cmaj.ca/lookup/suppl/doi:10.1503/cmaj.190385/-/DC1).37 Based on the number of affirmative answers, households were classified as food-secure, marginally food-insecure, moderately food-insecure or severely food-insecure, following the classification scheme developed by Health Canada and applied in other studies (Appendix 5, available at www.cmaj.ca/lookup/suppl/doi:10.1503/cmaj.190385/-/DC1).4,5,22,38
Control variables
We controlled for individual demographic characteristics, baseline health and household socioeconomic characteristics to isolate the association between food insecurity and mortality. Demographics included respondents’ sex and age at interview. Baseline health included any acute care admission to hospital in the past 2 years, smoker status, past-year alcohol consumption and number of self-reported chronic conditions among cancer, hypertension, effects of stroke, diabetes and heart diseases. We further controlled for socioeconomic characteristics including household income, housing tenure, highest education in household and household type.
Before-tax household income was unavailable for 20.6% of the respondents, including all from 2017. Quebec respondents, representing 24.3% of the sample, all had missing hospital admission records. A total of 5.2% of the sample had missing values for 1 or more other covariates. We kept all the jurisdictions (including Quebec) and cycles (including 2017) with valid food security status for analyses. We adjusted for missing hospital admission records from Quebec, missing income from 2017 and previous years, and other missing values as additional model covariates.
Statistical analyses
We first compared sample characteristics between those who died prematurely and those who did not. The statistics were generalizable to the noninstitutionalized Canadian adults aged 18–82 years with valid food insecurity status in the sampled jurisdiction cycles. We then computed death counts and crude mortality rates by cause of death as well as age at interview and age at death by food insecurity status. We fitted Cox proportional hazard models to estimate mortality hazard by food insecurity status while adjusting for potential confounders. Based on results from the Schoenfeld residuals test, we set respondents’ sex, age at interview, recent hospital admission and number of chronic conditions as strata in Cox models to cope with their violation of the proportional hazard assumption (Appendix 6, available at www.cmaj.ca/lookup/suppl/doi:10.1503/cmaj.190385/-/DC1).
In the survival analyses, we first estimated all-cause mortality for the overall sample and then split the sample by sex and analyzed all-cause mortality for men and women separately. We finally examined the association of food insecurity with cause-specific mortality for the overall sample. We used Statistics Canada’s sample weights to compute sample characteristics by CCHS respondents’ vital status. We also applied weights to compute average age at interview and age at death by food insecurity status. We conducted sensitivity tests on all-cause mortality to ensure that findings were not driven by weights, outliers, missing values, sampling bias or choice of measurements; results resembled those from the main analyses. All analyses were done with 2-sided confidence intervals using Stata SE 15.1. Coefficients with p < 0.05 were considered statistically significant.
Ethics approval
We obtained approval from the Health Sciences Research Ethics Board at the University of Toronto.
Results
The analytic sample consisted of 3 390 500 person-years from 510 010 adults. A total of 25 460 adults died before age 83 years between CCHS interview and Dec. 31, 2017 (Table 1). The 484 550 respondents who did not die prematurely by 2017 and the 25 460 who died prematurely by 2017 represented, respectively, 267 331 000 and 8 488 000 noninstitutionalized Canadian adults aged 18–82 years with valid food insecurity status. Compared with adults who did not die prematurely, those who did were less likely to be food secure and more likely to be male, older, smokers, chronically ill and with low income and education (p < 0.05 for all).
The crude mortality rate — number of deaths per 100 000 person-years — was higher for adults experiencing more severe food insecurity: 736 for food-secure adults compared with 752, 834 and 1124 for their marginally, moderately and severely food-insecure counterparts, respectively (Table 2). About 68% of premature deaths were potentially avoidable among food-secure adults; the comparable figures were 72%, 72% and 75% among their marginally, moderately and severely food-insecure counterparts, respectively. Noncommunicable diseases accounted for 90% (22 830/25 460) of total deaths, including 91% of deaths among food-secure adults and 84% of deaths among food-insecure adults. However, the share of food-insecure adults was proportionally higher among those who died from communicable diseases and injuries (19.5%) versus noncommunicable diseases (11.8%). Food-insecure adults also died earlier. The mean age at death was 68.9 years for food-secure adults compared with 64.4, 62.7 and 59.5 years for marginally, moderately and severely food-insecure adults, respectively (Table 3; p < 0.0001 for all, compared with food-secure adults).
All-cause mortality
Marginal, moderate and severe food insecurity was associated with 1.62 (95% confidence interval [CI] 1.51–1.73), 1.96 (95% CI 1.86–2.07) and 3.25 (95% CI 3.04–3.48) times higher mortality hazard after adjusting for respondents’ sex and age at interview (Table 4; Appendix 7, available at www.cmaj.ca/lookup/suppl/doi:10.1503/cmaj.190385/-/DC1). The magnitude of the adjusted hazard ratios (HRs) decreased after further controlling for baseline health and socioeconomic status, but the associations remained significant at p < 0.05 for all levels of food insecurity, with adjusted HR being 1.10, 1.11 and 1.37 for marginal, moderate and severe food insecurity, respectively. Sensitivity tests confirmed that findings on all-cause mortality were not driven by outliers, missing records, sampling bias, weights or choice of measurements (Appendix 8, available at www.cmaj.ca/lookup/suppl/doi:10.1503/cmaj.190385/-/DC1). The association between food insecurity and all-cause mortality hazard was similar for men and women (Appendix 9, available at www.cmaj.ca/lookup/suppl/doi:10.1503/cmaj.190385/-/DC1).
Cause-specific mortality
Severe food insecurity was consistently associated with higher mortality hazard across all causes of death except cancers (Table 5; Appendix 10, available at www.cmaj.ca/lookup/suppl/doi:10.1503/cmaj.190385/-/DC1). In comparison, marginal and moderate food insecurity was associated with fewer causes of death. Notable exceptions included infectious-parasitic diseases, which were associated with all levels of food insecurity; suicides, which were associated with marginal and severe food insecurity; and circulatoryrespiratory diseases, which were associated with moderate and severe food insecurity. In addition, higher mortality from potentially avoidable causes was observed across all levels of food insecurity.
Interpretation
Drawing on a population-based sample of Canadian adults, we found a graded positive association between household food insecurity status and hazard of premature mortality. Severe food insecurity predicted higher mortality across all causes of death except cancers. Among adults who died prematurely, those experiencing severe food insecurity died at an age 9 years earlier than their food-secure counterparts. These findings are important for extending understanding of the health burden associated with food insecurity in Canada.
The graded association between food insecurity status and premature mortality — in particular, the strong association for severe food insecurity — is consistent with previous studies in the United States and Ontario.22,23 The significant correlations of all levels of food insecurity with potentially avoidable deaths imply that food-insecure adults benefit less from public health efforts to prevent and treat diseases and injuries than their food-secure counterparts. That severe food insecurity was associated with mortality hazard across circulatory-respiratory and metabolic-digestive diseases is in line with studies documenting greater incidence of these chronic conditions among food-insecure adults.6–10 The material deprivation and psychological distress typified by food insecurity can result in chronic inflammation, 39 nutritional vulnerability40 and medication nonadherence, 41 all of which hamper chronic disease management and increase the probability of adverse outcomes.
Our finding of higher risk of death from infectious-parasitic diseases across all levels of food insecurity is important for extending our understanding of the health risks associated with food insecurity in Canada. Although the underlying mechanisms still await elucidation, our finding is consistent with the existing evidence associating food insecurity with mortality among HIV-positive adults.24,25 Inadequate nutrition, depression, inability to follow therapeutic regimens and co-infection of hepatitis C are common factors inhibiting treatment of HIV/AIDS and precipitating death among food-insecure adults.11–13,42–48 More research is needed to understand the link between food insecurity and deaths by communicable diseases unrelated to HIV.
Although the link between food insecurity and mental health disorders is well documented in Canada,5,15,18–21 our finding that severe food insecurity was associated with unintentional injuries and suicides highlights the urgent need for more effective interventions. Given the prevalence of deaths related to drug poisoning in our sample (results not shown), our finding on unintentional injuries may partly reflect the link between food insecurity and substance use,49,50 which is often complicated by the co-existence of HIV infection and depression.51–53 The higher risk of death from suicides is consistent with Canadian studies documenting links between food insecurity and suicidal ideation and attempts.19–21 Suicide prevention needs to be prioritized for adults experiencing food insecurity.
Previous research has consistently documented an association between lower income and higher mortality in Canada,54,55 including greater mortality by unintentional injuries and suicides. 54,56 However, food insecurity and low income are inherently different measures of economic well-being, with the former capturing experience-based material hardship and the latter providing a relative measure of monetary resources. The markedly higher mortality hazard of severe food insecurity highlights the importance of policy interventions that protect households from extreme deprivation. In Canada, policies that improve the material resources of low-income households have been shown to strengthen food security and health.57–62 This body of evidence should inform future policy-making in Canada so that household food insecurity is more effectively addressed.
Limitations
We recognize the limitations of this study. First, multiple jurisdiction cycles were excluded from this study owing to missing food insecurity data, which limited the generalizability of our findings to adults with valid food security status. In addition, hospital admission records from Quebec and income data from 2017 were missing. However, sensitivity tests showed a limited effect from missing food insecurity, hospital admission and income. Second, our findings are correlational, not causal. The chronic conditions we controlled for were self-reported and limited in scope, with a reference period (previous 6 mo) that overlapped with the food insecurity measurement (previous 12 mo). Our adjustment for hospital admission provided a more objective indication of baseline health, but likely captured more severe health problems. Omitted variable bias remains possible: unobserved factors may have led to food insecurity and premature death simultaneously. Third, food insecurity was measured once per person in the CCHS, obscuring important information about change of food insecurity status over the follow-up period. Repeated measurement of food insecurity in a panel design would yield insight into hardship dynamics and help determine causality. Finally, the number of deaths by specific causes were relatively small, especially for communicable diseases and injuries. Replication studies using different population-based samples may help to validate our findings.
Conclusion
Our study marks an important step toward understanding the epidemiology of food insecurity and establishing this condition as an independent social determinant of health. Food insecurity status is linked to all-cause and cause-specific premature mortality among Canadian adults. Future studies need to clarify the mechanisms underlying the linkage between food insecurity and various causes of death. Adults from severely food-insecure households warrant greater attention from clinicians and policymakers due to their elevated risk of premature death.
Acknowledgements
The authors are grateful for the invaluable comments from the editor, the 2 anonymous referees, Judi Bartfeld and audience at the 2018 Canadian Research Data Centre Network conference. We thank analysts from the Statistics Canada Research Data Centre for the logistical support.
Footnotes
Competing interests: None declared.
This article has been peer reviewed.
Contributors: Valerie Tarasuk, Marcelo L. Urquia and Craig Gundersen contributed to the conception of the study. Valerie Tarasuk and Marcelo L. Urquia acquired the data. Fei Men and Valerie Tarasuk designed the study. Fei Men analyzed and interpreted the data and drafted the manuscript. Valerie Tarasuk, Craig Gundersen and Marcelo L. Urquia contributed to data interpretation and revised the manuscript for important intellectual content. All of the authors gave final approval of the version to be published and agreed to be accountable for all aspects of the work.
Funding: This study was supported by Canadian Institutes of Health Research grant PJT 153260, awarded to Valerie Tarasuk and Marcelo L. Urquia. The funder had no role in the design and conduct of the study; collection, management, analysis and interpretation of data; or preparation, review, decision to submit for publication or approval of the manuscript.
Data sharing: Data for this study cannot be shared publicly because of the confidentiality agreement between Statistics Canada and respondents of the Canadian Community Health Survey. Data are available at the Statistics Canada Research Data Centre (contact via phone +1 905-525-9140 ext. 23661) for researchers who meet the criteria for access to confidential data.
- Accepted November 27, 2019.