Funded Grants

Lessons in death: disease dynamics in London since 1592 - current and future significance

An extraordinary series of manuscripts sits in the Guildhall Library in London England, manuscripts detailing the deaths in London on a weekly basis since 1592 by cause of death and by parish. The Bills of Mortality, meticulously kept and preserved for over four hundred years, are an unparalleled resource for understanding the ways in which infectious diseases invade and spread through populations in space and time. Yet this wealth of data and the vital patterns it contains remain largely opaque to theoretical biologists, because the handwritten Bills have never been converted into a digital form. Until now.

The purpose of my proposal is to study and model the spread of infectious diseases over the last four centuries in London, one of the most important centers of modern human history. I am strongly motivated to do this, because recent research has shown that extraordinary discoveries can be made from much shorter and less detailed data series. Indeed, great strides have been made in our understanding of infectious disease dynamics through analysis of published records of cases and death covering less than 100 years (almost exclusively since 1900), and through mathematical and computational modeling that has attempted to explain many of these patterns. In particular, 20th century data have revealed that while many infectious diseases recur frequently, the time between epidemics (and their overall pattern) changes dramatically over long time scales. For example, in a given city the observed pattern of incidence of an infectious disease can display simple annual cycles for many years and then suddenly change to a cycle of two or more years, or to apparently irregular behaviour, and then return to an annual cycle. Some of my own work has shown that these complex transitions in the character of infectious disease dynamics result from changes in birth rates and changes in infection control strategies, and that these transitions can be predicted using mathematical models.

The discoveries that have been made to date represent merely the tip of the iceberg in terms of what can be discovered from historical records of epidemics. In principle, exploring historical data is straightforward, but in practice a serious impediment is that many of the relevant sources are not easily accessible and exist only in unique archives (such as the Manuscripts Section of the Guildhall Library). The Guildhall’s weekly record of deaths in London, classified by cause, and separately classified by the parish in which burial occurred, is the empirical foundation of my proposal.

While annual counts of deaths in London have been studied extensively by historians, demog- raphers and epidemiologists, the complete weekly records have never been digitized and converted into a format suitable for analysis. Yet these weekly records can reveal the detailed temporal and spatial structure of all serious epidemics in the city over the centuries as London grew from a population of about 125,000 in 1592 to more than 7 million in the present day. There is extraordinary potential to make new and important discoveries through analysis of these data and through modeling the complex epidemiological dynamics that the data reveal.

The first step will be to digitize the mortality records and create a spreadsheet suitable for epidemiological analysis. You might hope that digitization would be a straightforward task that we could let a computer accomplish on its own using optical character recognition software. But because most of London’s mortality records exist only in handwritten form, we have no choice but to enter each data point by hand.

Epidemics of certain diseases are known to have occurred in London, and will form the starting point of my analysis. Major epidemics of plague occurred in London in 1593, 1603, 1625, 1636 and culminated with the “Great Plague of London” in 1665. Minor plague epidemics occurred in 15 other years during this period. I will begin by describing the spatial dynamics of these epidemics using standard statistical approaches and by creating movies that show spatial spread over a map of 17th century London.

It will then be possible to address a variety of questions that could not be approached previ- ously, and which have important implications for modern day infectious threats (including plague itself, for which there is still a reservoir in rodents and which has the potential to explode in hu- mans again). For example, how did the transmission rate of plague change between 1592 and 1665, during which the population density of London increased four-fold? Are the observed spatial and temporal mortality patterns consistent with plague being spread from rats to fleas to humans? Or must direct human-to-human transmission (e.g., pneumonic plague or a different pathogen altogether) have been critical during times of major plague epidemics?

These questions can be addressed using mathematical and computational modeling techniques. Such methods make it possible to discover which mechanisms give rise to which observed patterns. If we start with a simple model that can describe an epidemic, but the pattern of epidemic it predicts is not consistent with the data, then we can include more mechanisms in the model, step by step, isolating how each mechanism alters the patterns we would expect to observe. For example, when studying epidemics of plague in London, I will begin with a model that divides the whole population into a few compartments according to disease status (e.g., susceptible, infectious, recovered). Such a model can’t tell us anything about spatial spread, but it will allow us to estimate the transmissibility of plague in the city as a whole. I will then need to look at the effects of interactions among rats, fleas and humans. Next, spatial structure can be included in steps, for example by first considering the effects of the population being divided into two groups (within the walls of the city and outside the walls) and then considering the population broken down into the 130 parishes for which the Bills of Mortality provide weekly plague deaths. Finally, I will carry out computer simulations of transmission dynamics among the half million people living in London at the time (since we don’t have individual level data, these simulations will mainly test the robustness of simpler models).

Another important disease about which we can learn a great deal from the weekly London mortality records is smallpox. London experienced periodic smallpox epidemics throughout the 17th, 18th and 19th centuries, with declining disease-induced death during the 19th century. While the 19th century reduction in smallpox mortality is thought to be related to control practices, especially vaccination, detailed modeling and comparison with the weekly records will allow me to estimate the role that vaccination played, and whether changes in virulence or transmissibility were likely involved. Influenza did not appear as a consistent cause of death in London’s mortality records until the 19th century, yet influenza viruses are believed to have circulated in human populations for more than 400 years. We know that major changes in human influenza, known as antigenic shifts, occurred three times in the 20th century (1918, 1957 and 1968), but specific evidence for shifts before 1918 is very limited. With the detailed spatio-temporal record of mortality from all causes in London, and comparison of those data with expected patterns of mortality from newly invading influenza viruses, we have the potential to discover when previous shifts probably occurred. Such information is important not only for understanding the ecology and evolution of influenza in the past, but for predicting the likely frequency of influenza pandemics in the future.

The development of the field of modeling infectious disease dynamics has been greatly influ- enced by a few time series of incidence of measles and other childhood diseases covering a few decades of the 20th century. The complex patterns in these data immediately presented deep puz- zles and have generated thousands of research papers. The mortality patterns that I propose to study in this project cover a time period that is much more extensive than anything that has been analyzed to date with this level of spatio-temporal detail. It is difficult to foretell the discoveries that will emerge from the newly digitized data — and the modeling and analyses that they will inspire — beyond the likelihood that they will be profound.