Table 2 Suggested directions for future research

Anxiety disorders as risk factors for CHD and stroke

Population-based prospective studies are needed that clearly define both anxiety disorders and the outcome of interest according to internationally accepted diagnostic criteria.

Screening for undiagnosed or subthreshold CVD at baseline to determine the direction of temporal association is needed.

Possible confounders, including depression, should be identified, measured and controlled for in statistical models.

Determine whether the risk for CVD is reduced in people who receive an evidence-based intervention for anxiety.

Bipolar disorder and schizophrenia as risk factors for CHD, stroke and type 2 diabetes

Prospective studies of representative community samples are needed to reduce the risk of recall bias and temporal ambiguity.

The direction of association will be further clarified through baseline screening for undiagnosed or subthreshold chronic disease.

In data collection and analysis, the possible bidirectional effects of treatment should be considered in addition to other known risk factors.

Depression as a risk factor for stroke and type 2 diabetes

Longitudinal studies are required where consistent clinical definitions are used to identify patients with depression.

Screening for undiagnosed or subthreshold CVD and diabetes at baseline would strengthen the case for temporality.

Risk of confounding can be reduced by the collection of data on treatment modality, adherence to treatment and comorbid disorders at both baseline and follow-up.

The use of accepted diagnostic criteria should be considered when identifying the potential outcome.

Autistic spectrum disorders and injury

Prospective studies of birth cohorts with frequent follow-up to help determine the order of onset are needed.

Where the sample size is limited, multisite collaborations should improve the power of statistical analyses in outcomes of patients with relatively rare childhood disorders.

Data collection should encompass possible confounders including common comorbidities (for example, epilepsy, intellectual disability) and environmental factors (for example, level of supervision, safety of physical environment). Data on environmental factors are difficult to obtain, but collaboration between qualitative and quantitative researchers provides scope for novel pathways in data collection on complex issues.

Studies of non-Western populations would contribute to the generalizability of findings outside the current narrow social context.

Childhood behavioural disorders and injury

Prospective studies with community samples are needed to establish the order of onset for health outcomes. An approach that has been underutilized thus far is the long-term follow-up of cases from national childhood mental health surveys.

Adjustment should be made for possible confounders such as common comorbidities (for example, conduct disorder, substance use) and treatment.

Consideration should be given to extending the collection of data beyond vehicular accidents to capture the proportion of people who suffer an injury, as well as the type and severity of injury.