Abstract
We write to provide additional context for the findings reported in this volume of HAMB by Sabic et al. They report an association between statewide veteran suicide rates and mean state altitudes. In this editorial, we seek to bring forward important epidemiological and methodological factors to consider when interpreting the study's results.
As the 10th leading cause of death in the United States overall, the second leading cause of death among young adults aged 10–24 years, and the third leading cause of death for adults aged 25–44 years (Heron, 2018), suicide is a significant public health issue. It is also a complicated one: mental health issues, substance abuse disorders, and sociodemographic factors (including age, gender, race, ethnicity, level of education, chronic illness, and firearm ownership) all contribute to completed suicides. Recent studies have also demonstrated that the rate of suicide increases in the face of unemployment and poverty (Reeves et al., 2012; Chang et al., 2013; Case and Deaton, 2015, 2017). Across the United States, overall mortality rates in the United States are suddenly rising, and as Dow recently observed, “this decrease in life expectancy reflects a dramatic increase in deaths from suicide, alcohol and drug use—the so-called ‘deaths of despair’”(Dow et al., 2019). These phenomena have many intermingled determinants with complex interactions among different protective and risk factors (Blakely et al., 2003; Hawton and van Heeringen, 2009; Kim et al., 2011).
In addition to medical, demographic, and socioeconomic influences, suicide rates also vary significantly according to geographic locale, with high rates in the Western and Rocky Mountain states (CDC, 2014). These regional differences in suicide rates persist, even after adjusting for gender, age, and race (Betz et al., 2011). Mental health experts and epidemiologists have postulated that the higher suicide rates in Western states may be explained by relative isolation, higher rates of firearm ownership, or a higher prevalence of substance use (or possibly, exposure to more dangerous and addictive substances (Masters et al., 2018; Ruhm, 2018). In some Western states, especially in rural locales, a lack of mental health care or a disinclination to utilize mental health services may also raise the risk of suicide (Frankel and Taylor, 1992; Miller et al., 2007, 2009). Understanding the epidemiological patterns of suicide is important for both clinical care and public health interventions; early recognition and treatment of at-risk individuals and communities may prevent suicides and improve overall health and wellness (Brodsky et al., 2018).
Could living at high altitude be another risk factor for suicide? Western states have higher average altitudes and previous studies have reported a correlation between state or county elevation and the population-based suicide rate (Cheng et al., 2005; Haws et al., 2009; Brenner et al., 2011). A number of mechanisms could explain the relationship between suicide or depression and living at altitude (Haws et al., 2009), including the lower blood oxygen saturations that occur at higher altitudes (Katz, 1982; Hackett and Roach, 2012). Indeed, barometric pressure, drug pharmacokinetics, and metabolism—all of which change with altitude—may also affect human behavior, mental health, and suicide risk (Kamimori et al., 1995; Ritschel et al., 1996; Jurgens et al., 2002; Arancibia et al., 2003; Schory et al., 2003; Maldonado et al., 2009).
At the same time, existing epidemiological and physiological studies linking high altitude with suicide do not provide sufficient evidence of causality. This is also the case with Sabic's report in this issue of HAMB.
Sabic and colleagues add to the existing literature on the association between altitude and suicide risk by exploring suicide rates in a specific high-risk population: military veterans. They tested the hypothesis that there is a correlation between suicide rates and altitude of residence. They used data obtained from the U.S. Veterans Health Administration (VA) to conduct a state-by-state analysis of veteran suicides. Their findings showed an association between altitude and veteran suicide rates, after adjusting for several important potential confounding variables, including education, employment, gender, number of VA facilities in the state, use of VA services as well as firearm ownership, and smoking rates by state. After conducting bivariate and multivariate logistic regression analyses, they determined that altitude has the strongest correlation to veteran suicides.
With rising rates of suicides among veterans (Kuehn, 2009) it is important to understand whether this association is real and whether it is causal, so that health care providers can provide optimal mental health and other health care and social services to this vulnerable population. Helping our military veterans is very much in our national interest.
The report by Sabic has methodological strengths as well as significant limitations. Among the strengths are a large sample size and the use of population-based data. At the same time, similar to the preceding studies, the authors relied upon aggregated region-level data in place of person-specific data, raising the likelihood of ecological fallacies. Suicide rates vary across different states; altitude, employment rates, wages, firearm ownership, and the availability of VA clinical facilities also vary across states. However, without knowing employment, smoking, firearm ownership status, and use of health care resources in individuals who did and did not commit suicide, an association cannot be inferred. Ecological fallacies play havoc with epidemiological studies in numerous areas of investigation, from studies of diet and chronic diseases to analyses of injury prevention laws and injury-related deaths; suicide is no exception. The concerns surrounding ecological fallacies and causal associations prompted 47 editors of 35 respiratory, sleep, and critical care journals to offer guidance and principles on the design and reporting of observational causal associations and inference studies to increase the rigor in observational research methods (Lederer et al., 2019)
One of the accepted criteria for inferring “causality” in epidemiological studies is the existence of a plausible biological mechanism for an observed association. Sabic and colleagues suggested hypoxia as a possible biological explanation for why altitude may be causative, referencing other reports with similar findings (Kim et al., 2011; Young, 2013). However, many physiological changes indicating hypoxic responsiveness, such as increased ventilation (Donoghue et al., 2005), raised hematocrits (Weil et al., 1968), and detectable erythropoietin release (Ge et al., 2002) occur only at much higher elevations (Swenson, 2011), casting doubt on this hypothesis.
How high altitude is measured and defined could also have a significant impact on the findings. The authors utilize state mean altitudes, which, as the authors admit, is a “relatively coarse grained measure of the altitude to which veterans in a state might be exposed.” Since their hypothesis is that the hypoxia associated with altitude may provide a biological reason for suicidal behavior, the fact that a state such as Colorado with a high overall mean altitude can have many veterans living in the eastern part of the state with altitudes closer to 2–3000 ft raises the question of whether altitude plays any part of the suicides that occur there. In addition, the duration of exposure to a given altitude is unknown in these veterans. It is possible that brief exposure from training at high altitude would provide different risks than exposure for years while stationed at high altitude.
Another important limitation of veteran suicide is the limited access to person-specific information about disability and combat exposure. For example, prior research has shown that suicide risk is heightened in the early years after exposure to military combat (Institute of Medicine, 2007). Other risk factors, including medical illnesses, psychiatric illness, substance abuse, traumatic brain injury, and chronic pain, are also important risk factors for suicide among veterans (Department of Veterans Affairs, 2018). Without the ability to utilize individual veteran data, the conclusions of Sabic's study may be misleading and divert attention away from more actionable risk factors.
Similarly, gun ownership is an important issue in understanding all suicides, including those by veterans. Many other studies (although without consideration of altitude) have identified firearms ownership rates as likely determinants of geographic variation in suicide mortality in the United States, largely because firearms have the highest case fatality rate of any suicide method (85%–90%) (Spicer and Miller, 2000; Miller et al., 2013). Gun ownership further illustrates the challenges of ecological fallacies. Sabic and colleagues did not have person-specific data about individual veterans' gun ownership. Rather, firearm ownership was based on a self-reported survey and only reflected ownership at a state level; every county in a given state was assumed to have the same rate of firearms ownership, despite differences in county elevations and rural and urban settings. In addition, the firearm ownership data did not include information about firearm storage patterns or efforts to restrict access in case of mental health crises.
Apart from the challenges of collecting person-specific data, Sabic's study, such as many similar studies, does not include information about recent social crises, such as death of a family member or friend, a sudden health challenge, or a financial, school, intimate partner, or legal problem. These acute triggering events may also contribute to veteran suicides (Betz et al., 2011). If so, they may be “actionable risk factors,” possibly responsive to increased crisis response systems and other mental health and social support services focused on crisis management and coping.
It appears that high- and low-altitude residents, including veterans, do have different suicide risks. Although physiological hypoxia might play a role, other social factors and demographic characteristics, such as race, Hispanic ethnicity, substance use, firearm ownership, report of depressed mood before suicide, and recent financial, job, legal, and interpersonal problems, appear more potent in elevating the risk of suicide in the general population and probably in veterans as well. Therefore, we must be cautious in concluding that high altitude is a cause of suicides, absent person-specific data.
However, even with limited data and an imperfect understanding of the causal connection between high altitudes and suicides, health care professionals at high altitudes should be especially attentive to this clinical and public health issue. Along with their counterparts at lower altitudes, they should follow best practices in assessing and responding to suicidality (Brodsky et al., 2018) when serving and treating veterans.