Associations Between Ambient Outdoor Temperature and Patterns of Morbidity and Mortality in Scotland

This report investigates the association between ambient air temperature and health impacts, measured as morbidity (hospital admissions) and mortality, in Scotland to determine what evidence there is to reject the (null) hypothesis that there is no relationship.

The mortality and hospital admission data came from the ACADME data mart and were available at a weekly level only, by age group and gender. All-cause mortality and cause specific mortality rates were obtained. The specific causes considered were: influenza, pneumonia, Chronic Obstructive Pulmonary Disease (COPD), cardiovascular and trauma. Morbidity was captured through emergency admissions to hospital with total weekly admissions for the same five specific causes. The morbidity and mortality data are available from 1981 onwards to the end of 2016.

Daily ambient temperature data were extracted from the Met Office integrated data archive system (MIDAS) land and marine surface stations data. A weekly ‘average’ temperature was calculated from the maximum and minimum air temperatures from all stations reporting in the week. A number of weekly summary temperatures were also calculated – average minimum, average maximum and a weighted average using Shepherd’s inverse distance weighting method.

For a restricted time period from 2009 to 2016 weekly GP consultation data for influenza like illness (ILI) and acute respiratory illness (ARI) were available from Health Protection Scotland (HPS).

The statistical analysis used generalised additive Poisson regression modelling with smooth spline terms for temperature and time and the logarithm of consultations and an autoregressive term to take into account serial correlation in the weekly data.

The analysis of the deaths data identified that there is an association between all-cause mortality and temperature. As average weekly temperature increases, the average weekly all-cause mortality rate decreases until about 10^oC, where it levels off and begins to increase again.

The association between all-cause mortality and temperature is present in all age groups but is less pronounced in those aged under 65. The increasing risk at higher temperatures is more evident in those aged 75-84.

The association between decreasing mortality and increasing temperature is also evident for specific causes of death - influenza deaths, deaths due to COPD and cardiovascular deaths. Data coding issues, associated with coding changes over time, prevented similar analyses for pneumonia and trauma deaths.

With respect to hospital admissions data, there are associations between temperature and the proportions of emergency hospitalisations associated with influenza, pneumonia, COPD, cardiovascular and trauma. With the exception of trauma the general pattern is for lower proportions of admissions for the selected causes with increasing temperature. For COPD, there is the same up turn at higher temperatures above an average of 10^oC for the week. Trauma has a V shaped association with higher admission risk at both low and high temperatures.

After adjustment for ARI consultations, the association between mortality and morbidity, and temperature for all-cause mortality, for cause specific mortality and morbidity is not as strong. This suggests that the main driver of the observed association with temperature is respiratory infection but there is still a residual impact of temperature.

While colder temperatures are associated with higher mortality and morbidity, there is no strong evidence of a step change in the association between mortality and morbidity, and temperature such that there is an even greater impact of very low temperatures on mortality. There is no evidence of the existence of a temperature below which there is a step change in the increased risk of mortality or an increased gradient.

Different measures of weekly average temperature produced similar interpretations. All-cause mortality decreases with increasing temperature and then there is a slight rise at higher temperatures. The only difference with different temperature measures is the location of the nadir. This means that this analysis should not be used to set temperature thresholds (for intervention actions) as they will depend on the weekly temperature summary statistic used.