A regular series of unexpected and large increases in total deaths (all-cause mortality) for male and female residents of mid super output areas (MSOA) in England and Wales: How high level analysis can miss the contribution from complex small-area spatial spread of a presumed infectious agent

Abstract

Background: The trends in death in the UK, Europe, Australia, New Zealand and America show evidence for extended periods of higher than expected deaths. This study investigates the detail of these trends using Mid Super Output Areas (MSOA) in England and Wales, containing 10 to 100 male or female deaths per annum. Objective: To study the trend in deaths for both male and female residents of MSOA, and seek to identify a recurring series of curious and unexpected step-like increases in death which endure for around 12-months before abating. Methods: A running 12-month sum of deaths was constructed for each MSOA over the period Jan-01 to Dec-14. Blocks of 12-month deaths were compared each side of a given date to detect step-like increases in deaths which endure for a period of around 12 months before reverting back to the baseline level. Results: Deaths in each MSOA are marked by a regular series of step-like increases which initiate at roughly two year intervals, however across England and Wales around 1% of MSOA initiate a maximum magnitude event at any point in time. When clustered to larger areas, depending on the timing (synchrony) and magnitude in the smaller areas, these events can cause either a cancelling-out effect or re-inforce each other to create larger local, regional or national events. Female deaths show a tendency for larger step-like increases than males. The maximum increase in deaths declines with increasing size of the MSOA (average deaths per annum) and follows a Power Law function. For MSOA containing around 10 deaths per annum, the largest step-like increase in deaths range from 58% to 300%. Extrapolation back to around 0.1 deaths (roughly 8 households or a small social network) gives a maximum potential increase around 1,000%. Males and females appear to behave as separate compartments and each can lag behind the other due to random events surrounding initiation of each ‘outbreak’. Other studies indicate that these events instigate periods of higher emergency department attendance, emergency medical admissions, GP referral and sickness absence, and even a wobble in the gender ratio at birth. Conclusions: These events provide evidence for the existence of a new type of disease entity. There are implications to the monitoring of deaths and mortality rates between areas and countries, most importantly that the unique spatiotemporal spread can lead to these events being obscured in larger aggregates such as a country or region. The larger increase in death in females is also reflected in larger increases in female medical admissions during these events, and seeming selective loss of the female fetus. It remains unknown why there is an approximate two-year periodicity, and why the magnitude of each event in each MSOA shows such high variation over time, although chance transmission and a moderate degree of seasonal forcing (winter) appear involved.