Modelling the impact of climate change on health

The main objective of this thesis is to develop a robust statistical model

by accounting the non-linear relationships between hospital admissions due to

lower respiratory (LR) disease and factors of climate and pollution, and their

delayed effects on hospital admissions. This study also evaluates whether the

model fits can be improved by considering the non-linearity of the data, delayed

effect of the significant factors, and thus calculate threshold levels of the

significant climate and pollution factors for emergency LR hospital admissions.

For the first time three unique administrative datasets were merged: Hospital

Episode Statistics, Met office observational data for climate factors, and data from

London Air Quality Network.

The results of the final GLM, showed that daily temperature, rain, wind

speed, sun hours, relative humidity, and PM10 significantly affected the LR

emergency hospital admissions. Then, we developed a Distributed lag non-linear

model (DLNM) model considering the significant climate and pollution factors.

Time and ‘day of the week’ was incorporated as linear terms in the final model.

Higher temperatures around ≥270C a quicker effect of 0-2 days lag but

lower temperatures (≤00C) had delayed effects of 5-25 days lag. Humidity

showed a strong immediate effect (0-3 days) of the low relative humidity at

around ≤40% and a moderate effect for higher humidity (≥80%) with lag period

of 0-2 days. Higher PM10 around ≥70-μg/m3 has both shorter (0-3 days) and

longer lag effects (15-20 days) but the latter one is stronger comparatively. A strong effect of wind speed around ≥25 knots showed longer lag period of 8-15

days. There is a moderate effect for a shorter lag period of 0-3 days for lower

wind speed (approximately 2 knots). We also notice a stronger effect of sun hours

around ≥14 hours having a longer lag period of 15-20 days and moderate effect

between 1-2 hours of 5-12 days lag. Similarly, higher amount of rain (≥30mm)

has stronger effects, especially for the shorter lag of 0-2 days and longer lag of 7-

10 days.

So far, very little research has been carried out on DLNM model in such

research area and setting. This PhD research will contribute to the quantitative

assessment of delayed and non-linear lag effects of climate and pollutants for the

Greater London region. The methodology could easily be replicated on other

disease categories and regions and not limited to LR admissions. The findings

may provide useful information for the development and implementation of public

health policies to reduce and prevent the impact of climate change on health

problems.