Burden of Mental,Neurological,Substance Use Disorders and Self-Harm: Counting the Cards,or Shuffling the Deck?

In this issue’s paper entitled: “Burden of Mental, Neurological, Substance Use Disorders and Self-Harm in North America: A Comparative Epidemiology of Canada, Mexico, and the United States,” Vigo et al. provide an alternative perspective on disease-burden estimates from the Global Burden of Disease (GBD) Project. ¹ The authors criticize the GBD for what they consider to be an arbitrary reclassification of “mental health-related burden” into other categories and express concern that the GBD approach dilutes the apparent burden attributable to mental health issues. The authors base their reclassification on the creation of a new aggregate category that they call “mental, neurological, substance use disorders and self-harm” (MNSS). MNSS includes suicide mortality, some selected neurological conditions (neurocognitive disorders, epilepsy, and headaches), physical conditions attributable to alcohol, and a proportion of chronic pain issues (to capture the burden of somatic symptom disorders). Analysis of the reclassified data indicates that MNSS is the largest source of disease burden between the ages of 10 and 60. ²

The GBD covers 359 diseases and injuries for 195 countries, with the most recently published estimates covering from 1990 to 2017. The GBD aggregates disease-specific estimates using four hierarchical levels. The fourth level includes burden related to specific disorders (e.g., major depression, dysthymia) although some conditions remain grouped (e.g., anxiety disorders are a Level 4 grouping as of 2017). The third level includes additional aggregation (e.g., depressive disorders), the second level includes a category called “mental disorders,” and at the top of the hierarchy, the first level includes only three overarching categories: noncommunicable diseases; injuries; and communicable, maternal, neonatal, and nutritional diseases. It is these aggregate groupings, and their labeling, that Vigo et al. consider problematic. They point out that the “mental disorders” category does not capture suicide deaths, some relevant neurological morbidity, alcohol-related morbidity, and somatic symptom disorders. Their concern is that the GBD’s approach may reduce “the visibility of the burden related to mental health.” For example, in the 2017 Canadian ranking of disease burden, “mental disorders” rank fifth (although they rank first in the 15 to 24 age range).

The purpose of the GBD is to describe mortality and morbidity from health conditions (causes) and risk factors at global, national, and (in some countries, but not in Canada) subnational or regional levels. They make comparisons across populations that enable a better understanding of changing health across the world. While it is not the intention of the GBD to provide an arena in which health-care sectors compete for resources and prestige, it is possible that naive decision makers, especially those who are unaware of the GBD methodology, could misinterpret GBD estimates in ways that undervalue mental health. For example, some stakeholders may come to believe that the “mental disorders” Level 2 grouping describes, or is intended to describe, mental health-related disease burden in its entirety. Vigo et al. help to counter this risk by producing the MNSS classification—a category that ranks Number 1.

Despite some harsh criticisms of the GBD by Vigo et al. (e.g., labeling the aggregate categories as “biased” and lacking in “minimum face validity”), it would be unwise to think of their reanalysis as discrediting the GBD or as correcting errors made by the GBD. Prior to the original GBD report, ³ priority setting in health policy was largely based on cause-specific mortality statistics. With the popularization of disability-adjusted life years (DALYs) as a metric for disease burden, the GBD made use of a composite measure that included both years lived with disability (weighted by the associated health loss using a disability weight) and years of life lost due to premature mortality. In doing so, the GBD brought mental health into the spotlight as never before. ⁴ Vigo et al.’s main target seems to be the Level 2 category label “mental disorders” rather than the overall quantification of disease burden provided by the fourth level of the hierarchy. However, to the extent that they impute methodological bad judgment and bias to the GBD consortium, they risk undermining the credibility of a source of helpful information. They also risk transforming a debate about health burdens into a less substantive debate about the choice of words for hierarchical category labels.

Vigo et al. assert that some of the mortality and morbidity related to mental health is “arbitrarily” classified by the GBD into other categories. This may be true, with the proviso that the Level 2 hierarchy represents “mental disorders” not “mental health-related burden.” It is also true that some degree of arbitrariness is inevitable in such classifications. Questions of how diseases should be grouped are inevitably tied to administrative, professional, and political agendas and cannot fully align with the realities of actual human health. Every health issue has a mental health dimension, and mental health encompasses biological, psychological, and social dimensions of human life. Vigo et al. respond to the division of neurological and psychiatric disorders by adding “specific neurological disorders—those with prominent behavioral and cognitive presentations, namely neurocognitive disorders, epilepsy, and headaches” to the MNSS category. Many other neurological disorders can have prominent behavioral and cognitive presentations. One suspects that the decision to single out epilepsy relates to its tendency to be managed in mental health settings in some low- and middle-income countries. From a Canadian perspective, this decision itself seems somewhat arbitrary. The mental health disturbances that occur in epilepsy (and in multiple sclerosis, Parkinson’s disease, etc.) are often managed in mental health settings whereas neurological aspects are often managed in neurological settings. Their “ownership” is shared and no classification can occur without some arbitrary decisions.

Vigo et al. attribute alcohol-related complications such as liver disease to alcohol use disorders, but they stop short of attributing a large proportion of lung cancers and heart disease to nicotine dependence. Indeed, smoking but not nicotine dependence is included as a risk factor in the GBD data. Mental disorders are also risk factors for multiple chronic conditions ⁵ and associated mortality, but the reverse may also occur. Mental and physical conditions may also share risk factors such as childhood adversities. ⁶ Etiologic pathways leading to disease states typically involve chains, or webs, of causal events, and in many situations, reciprocal effects occur. For example, smoking is believed to be a risk factor for depression, but depression itself may increase the risk of smoking initiation and reduce the rate of successful cessation. Such problems have vexed the most advanced methodologies of analytical epidemiology and the etiological issues cannot possibly be addressed through a descriptive framework such as that provided by the GBD. Consequently, reassigning DALYs from medical conditions to mental health-related morbidity based on an interpretation of the disorder as a cause of the outcome is an exercise in judgment and interpretation. This is another reason why the MNSS aggregation is best regarded as an alternative classification, not as a correction of bias or error in the GBD methodology.

While Vigo et al.’s critique and reanalysis focuses on the midrange of the GBD hierarchy, the most important way in which the GBD underestimates the burden of mental health is, in my opinion, a different issue. Mental health issues often have their onset early in life, and while the GBD seeks to quantify the extent of associated disability, it cannot take into account the negative trajectories that the mental disorders can create in a person’s life course. A negative trajectory may occur due to interference with the successful completion of education, initiation of careers, establishment of relationships, navigation of shifting social roles, and so on. Young adults who do not achieve their educational potential due to the early onset of, for example, a bipolar disorder, may experience many subsequent burdens that are not captured by the GBD disability weights because they are not direct manifestations of the disorder. The GBD methodology, sophisticated as it is, simply cannot address such issues.

Psychiatry and the broader mental health field has always struggled with issues of stigma, which partially manifest as a trivialization of mental disorders and a devaluing of those affected. This may contribute a lower prioritization of mental health issues in health policy. Consequently, it is understandable and commendable that attempts, such as that by Vigo et al., be made to drive home key messages about the importance of these conditions for public health. While I do not think it is possible to say that their estimates are more valid than those of the GBD, they do provide a different perspective. They “cut the pie” of disease burden in a different way.

Finally, even though the GBD is a one of the largest and most complex health research initiatives ever undertaken, it ultimately derives from a poorly developed knowledge base. The GBD represents disease epidemiology using Bayesian meta-regression strategies applied to data collected from hundreds of studies and data sets identified through its network of contributors and by conducting systematic reviews. The disability weights come from surveys collecting data about health state preferences using general population samples. Many Canadian epidemiologists, including myself, have collaborated with the GBD in identifying Canadian data sources with the hope of helping the GBD produce the best estimates of the epidemiology of mental disorders in Canada. However, we should not forget that the state of knowledge about these disorders is limited. The sophistication of the methods employed by the GBD and by Vigo et al. makes it quite easy to forget that the advanced modeling, processing, and weighting procedures of the GBD are required largely because of the need to fill in the gaps in an inadequate and incomplete knowledge base. In Canada, only two national surveys of mental health have ever been conducted, and each had substantial limitations in their coverage of various disorders. ^7,8 Ultimately, the question of how to address the many challenges of mental health will require a stronger understanding of mental health issues than currently exists.