Epidemiology provides information about the incidence of a condition, meaning the number of new cases, which arise during a certain period of time, and its prevalence, meaning how many individuals have the condition during a certain period. The goal of prevention is to decrease a condition’s incidence, i.e. prevent new cases from occurring, while successful treatment results in the decrease of the prevalence. Mental health disorders have presented challenges to epidemiology on several levels. In order to determine how many cases of a certain condition exist within the population, it is necessary to know how to define a case, but this is far from straightforward in the realm of mental health. Classification of medical disorders tends to evolve from symptom-based to mechanismbased as the links between a specific pathophysiology and the observed signs and symptoms are established. The lack of knowledge about the mechanisms producing cognitive and behavioural symptoms means that classification of mental disorders still relies upon descriptions of constellations of symptoms. The International Classification of Disease version 10(ICD-10),⁽¹¹⁾ and its United States counterpart the Diagnostic and Statistical Manual TR-IV (DSM-TR-IV),⁽¹²⁾ are the results of iterative attempts by experts in the field to create meaningful classifications of psychiatric disorders based upon such observations in conjunction with applicable considerations of length and severity of illness, age of onset, and risk factors. This work has provided the standardized terminology that made possible the first large-scale epidemiologic descriptions of mental disorders. However, problematic issues pertinent to epidemiology remain, including questions regarding the relative merits of categorical versus dimensional classification systems; how to interpret the high rate of comorbidities for several disorders; and how best to account for individuals who have subthreshold symptoms, including how to determine the starting point of a disorder. It is not uncommon for individuals who have come to meet criteria for a mental health disorder such as schizophrenia or depression to have had a preceding period of subthreshold ‘prodromal’ symptoms, but healthy individuals also have occasional subthreshold symptoms that resolve without intervention. Unfortunately this differentiation often cannot be determined except retrospectively, despite the fact that there may be different implications for epidemiologic and preventive efforts.

The question of how symptoms change over time gains additional relevance when attempting to describe the epidemiology of mental health disorders in children and adolescents. As described in more detail elsewhere in this volume,⁽¹³⁾ the science of developmental epidemiology has arisen as a response to the recognition that mental disorders may manifest in different ways over the lifespan, and that certain types of symptoms at one age may indicate that an individual is at high risk for developing a different disorder at a later stage of maturation. Risk factors may also have differing impact depending on an individual’s developmental stage. Function may appear impaired if children are developing slowly in comparison with their peers, and it must be decided when this is normal variation and when it should be considered pathological. An additional layer of complexity in epidemiology in paediatric populations is the incorporation of information from additional informants such as parents and/or teachers, and determining how to evaluate the relationship of symptoms to particular contexts.

Epidemiology is also used to assess for the presence of risk factors. Fixed risk factors are those that cannot be altered, such as genotype. Malleable risk factors are susceptible to intervention, such as exposure to lead-based paint or domestic violence. Causative risk factors are those with known relationships to a particular outcome, and are of particular interest to prevention because they represent potential points of intervention. Protective factors instead decrease the risk of an adverse outcome. Resilience is a term used to describe an individual’s ability to do well despite exposure to a typically high-risk situation.

Effective intervention to decrease risk factors or increase protective factors requires determining how these factors relate to each other and to the targeted health issues. The ultimate goal is a chain of causative steps leading from risk factor to outcome, but epidemiological data itself may provide sufficient guidance for action. One of the most famous examples of this was John Snow’s identification of tainted drinking water from a particular well as the root
of a cholera epidemic in London, which he did based solely on epidemiological observations. Removal of the pump handle stopped the epidemic and proved that exposure was a causative risk, decades before the bacteria itself was identified. We are currently in a similar situation to Snow in regards to connecting risk factors to mechanisms for many mental health disorders, with the additional complication that mental health disorders are typically associated with combinations of a large number of individually modest potential risk factors.

Risk factors can be classified in terms of how they relate to each other and to the specified outcome,⁽¹⁴⁾ and thus what type of intervention if any is appropriate. Mediating risk factors are those which explain how or why another factor affects the outcome; for example, the phenylketonuria enzyme mediates the effects of the phenylketonuria gene on IQ.⁽¹⁴⁾ Although all causal factors are mediators, the reverse is not true, and experimental conditions are generally necessary to demonstrate that a particular mediator plays a causal role. A moderating risk factor instead specifies under what conditions or for whom another risk factor will affect outcome. Moderating risk factors describe populations that have differing responses to a given exposure, and may also represent potential sites of intervention to prevent an adverse outcome by reducing vulnerability or increasing resilience. A proxy risk factor, also called a pseudocorrelation, is one that itself does not strongly predict outcome but is highly correlated to a risk factor that does. Overlapping risk factors are those that arise from the same underlying construct and are observed to equally predict outcome, be highly correlated with each other and not stand in a specific temporal relationship; these can often be combined into a single factor. Independent risk factors conversely are unrelated to each other; they both predict outcome but without correlation or temporal precedence.