Page 115 - Contributed Paper Session (CPS) - Volume 3

CPS1956 Areti B. et al.

Detecting life expectancy anomalies in England
using a Bayesian hierarchical model
Areti Boulieri, Marta Blangiardo
MRC-PHE Centre for Environment and Health, Department of Epidemiology and Biostatistics,
Imperial College London, UK

Abstract
In England, life expectancy has shown a steady increase over many years,
however these improvements have recently started to slow down
considerably. This work aims to investigate the changes in life expectancy in
England over time and across its local authorities, and to identify local
authorities with unusual time trends that might help with hypothesis
generation and point to emerging risk factors. We analyse mortality count
data in England for females at the local authority level (324 areas), from 2001
to 2016 (17 years), and by age group, assuming 19 age groups of 5 year bands.
We develop a statistical model within the Bayesian hierarchical framework that
accounts for spatial, temporal, and age effects, as well as for pairwise
interactions. The space-time interaction parameter is used to detect areas
whose time trends deviate from the national one. The detection rule that we
specify focuses on areas that are detected as unusual over the last 5 years of
the time period (2013 – 2017). The model is implemented in Integrated Nested
Laplace Approximations (INLA). We found roughly 40 areas to be highlighted
as unusual, following a different time trend in the mortality rates compared to
the national trend.

Keywords
Bayesian statistics; spatio-temporal modelling; anomaly detection; life
expectancy surveillance

1. Introduction
The study of life expectancy is of primary interest in public health practice,
where it is needed to plan for health and social services. Recently, it has
attracted a lot of attention (Hill 2018; Therrien 2018), due to the stalling effect
that has been observed in several countries, including USA and UK (Olshansky
et al. 2005; Hiam et al. 2018). That is, while life expectancy has been improving
steadily since the early 80s when records began, mainly due to better lifestyles
and healthcare, this phenomenon has started to slow down since 2012
onwards. According to Office for National Statistcs (ONS), between 2014 and
2015, life expectancy fell by 0.2 years for both sexes. A larger decrease was
noticed for females, and for larger age groups. In addition, there have been
fluctuations in life expectancy among local authorities in England (Office for
104 | I S I W S C 2 0 1 9