Do hotter temperatures increase the incidence of self-harm hospitalisations?

Abstract

A relationship between air temperature and the incidence of suicide has been established in a number of previous studies. Interestingly, the relationship between geographical variation in temperature and suicide incidence has generally been found to be negative, while the relationship between temporal variation in temperature and suicide incidence has generally been found to be positive. It is less clear, however, how temperature relates to the incidence of self-harm. This topic is of particular importance given the presence of ongoing global warming. This study investigated the relationship between temperature and the incidence of self-harm resulting in hospitalisation in New Zealand. Self-harm hospitalisations by date and district for 1993–2009 were obtained from the Ministry of Health. Meteorological data was obtained from NIWA. Generalised linear mixed models were used to estimate the effects of three different components of variation in temperature: geographical, seasonal and irregular. Irregular (random) daily variation in temperature had a modest positive relationship with the incidence of acts of self-harm resulting in hospitalisation, with about 0.7% extra incidents for every 1 °C increase in temperature. However, there was no strong evidence for a positive effect of either seasonal or geographical variation in temperature. We conclude that temperature does appear to bear some relation to the incidence of self-harm, with irregular daily variation in temperature having a positive effect. However, inconsistencies in the effects of different components of variation in temperature make it challenging to accurately predict how global warming will influence the incidence of self-harm.

Keywords:

• self-harm
• self-injury
• temperature
• climate change
• global warming

Disclosure statement

No potential conflict of interest was reported by the authors.

Notes

1. Given the use of a logarithmic link function as part of the negative binomial model, the log transformation allowed population size to have an additive effect. Specifically, one would expect the logarithm of population size to have a coefficient near 1, if self-harm incidence was directly proportional to population.