Comparison with previous studies

Our study was strengthened by the use of longitudinal data commencing prior to the onset of the earthquakes and the use of appropriate controls. For example, if antidepressant dispensing were evaluated only in Canterbury following the onset of the earthquakes, large increases of antidepressant dispensing would be reported. However, pre-earthquake data reveal that increases in dispensing started before the earthquake sequence. In addition, the use of controlled data highlights that increases in Canterbury were matched by steeper increases in the rest of New Zealand, suggesting that adverse effects of the Canterbury earthquakes do not account for the increase in antidepressant dispensing recorded.

We previously identified two studies evaluating the mental health of young people in Canterbury following the Canterbury earthquake.^{Reference Beaglehole, Mulder, Boden and Bell18} The first study examined parent-reported emotional and behavioural problems in 4-year-old children as part of a New Zealand-wide health and development check prior to school entry and did not identify any adverse effects in Canterbury children as a result of the earthquake sequence.^{Reference Thomson, Seers, Frampton, Hider and Moor19} The second study, by Liberty et al,^{Reference Liberty, Tarren-Sweeney, Macfarlane, Basu and Reid20} utilised teachers to rate the behaviour of 212 5-year-olds several years after the onset of the earthquake sequence and reported the opposite. Although each study used pre-disaster measures and appropriate controls, large population studies include many individuals with lesser degrees of earthquake exposure and are therefore less sensitive to detecting change than studies that only sample individuals with high earthquake exposure. The study by Liberty et al ^{Reference Liberty, Tarren-Sweeney, Macfarlane, Basu and Reid20} was situated in an area of greater earthquake damage and children in this area may therefore have greater earthquake-related difficulties than children from a wider sampling pool. The use of teachers who were also likely to have been exposed to the earthquakes and have a heightened awareness of the negative effects of trauma on behaviour may also have introduced reporting bias to the study design that is not present in studies such as ours, which used dispensing as a proxy for rates of significant mental illness.

Our study sits within the broader literature evaluating the impacts of disasters on young people. In general, these studies report adverse effects on mental health,^{Reference Norris, Friedman, Watson, Byrne, Diaz and Kaniasty3,Reference Bromet and Dew8–Reference Vogel and Vernberg10} although methodological problems have also been highlighted in reviews of the disaster literature (principally the failure to use longitudinal data, the absence of appropriate controls, and sampling techniques that do not evaluate the population of interest).^{Reference Goldmann and Galea4,Reference Rubonis and Bickman5,Reference Norris21} These methodological concerns were addressed by our study design and therefore place weight in support of the findings of our study.

What is it about the post-earthquake disaster setting in Canterbury that meant only minimal adverse effects of the earthquake were detected? Following the earthquakes, there was a large-scale community and public health effort (the All Right? campaign: https://allright.org.nz/about-us), designed to promote mental health in those affected by the earthquakes. This was coupled with extensive school interventions, free counselling and extended general practice consultations for all. It is likely that these and other community initiatives minimised any adverse effects that may have occurred as a result of the earthquakes. It is also possible that the widespread availability of support services in Canterbury following the earthquakes reduced the demand for medications in Canterbury in a way that did not occur nationally, where such services were not present.

Limitations

For the majority of common psychiatric conditions affecting children and adolescents, first-line treatments are usually psychosocial, and medication is used after a failed trial of a psychological therapy, for prolonged symptomatology or after specialist consultation.^{Reference Malhi, Bassett, Boyce, Bryant, Fitzgerald and Fritz16,Reference Andrews, Bell, Boyce, Gale, Lampe and Marwat17} The percentage of young people in Canterbury taking any of the study medications ranged from 0.05% for anxiolytics and sedatives/hypnotics to 1.7% for antidepressants. These figures are lower than for the corresponding adult population.^{Reference Exeter, Robinson and Wheeler22} They suggest that in New Zealand, psychiatric medication is only being used for young people with significant and persistent mental health concerns. This means that there is a gap between total community morbidity and that measured by our dispensing records. However, the data provided tracked morbidity of particular relevance to physicians and healthcare services, although other methods will identify a larger pool of potential patients.^{Reference Tannenbaum, Lexchin, Tamblyn and Romans23}

The use of dispensing as a proxy for morbidity is only valid if the threshold for prescribing psychiatric medication by clinicians remained constant over the earthquake period. It is conceivable that general practitioners are less likely to prescribe psychiatric medication if there is an obvious psychosocial context for the presenting problem. In the adult population, increasing dispensing of psychiatric medication has been reported following disasters (at least in the short term).^{Reference Rossi, Maggio, Riccardi, Allegrini and Stratta13–Reference Beaglehole, Bell, Frampton, Hamilton and McKean15} This suggests that reduced propensity to prescribe is unlikely to be a dominant explanation, although we are unable to quantify this possibility for younger populations. We also considered whether other factors, such as religious faith, may provide a helpful context to the dispensing patterns reported. Religious faith increased in the post-earthquake period but this was not associated with improved subjective well-being.^{Reference Sibley and Bulbulia24} Subjective well-being fell for the minority who lost religious faith in the post-earthquake period but this group constituted only a small percentage of the overall population, suggesting the absence of significant effects of religion on prescribing for mental illness.^{Reference Sibley and Bulbulia24}

In the post-disaster setting, adequate access to healthcare is essential if dispensing records are to be a useful measure of morbidity and service need. Following the Canterbury earthquakes, access to primary and secondary services was enhanced for children and adolescents through post-quake initiatives, suggesting that in Canterbury the use of dispensing data as a proxy for community morbidity of a severity to result in treatment with psychiatric medication is reasonable.