Data and code for: Local infectious disease experience influences vaccine refusal rates: a natural experiment

Vaccination has been critical to the decline in infectious disease prevalence in recent centuries. Nonetheless, vaccine refusal has increased in recent years, with complacency associated with reductions in disease prevalence highlighted as an important contributor. We exploit a natural experiment in Glasgow at the beginning of the 20th century to investigate whether prior local experience of an infectious disease matters for vaccination decisions. Our study is based on smallpox surveillance data and administrative records of parental refusal to vaccinate their infants. We analyse variation between administrative units of Glasgow in cases and deaths from smallpox during two epidemics over the period 1900–1904, and vaccine refusal following its legalisation in Scotland in 1907 after a long period of compulsory vaccination. We find that lower local disease incidence and mortality during the epidemics were associated with higher rates of subsequent vaccine refusal. This finding indicates that complacency influenced vaccination decisions in periods of higher infectious disease risk, responding to local prior experience of the relevant disease, and has not emerged solely in the context of the generally low levels of infectious disease risk of recent decades. These results suggest that vaccine delivery strategies may benefit from information on local variation in incidence.

Methods
Overview of the record

This record provides the main analysis dataset for the following paper:

Angelopoulos K, Stewart G, Mancy R. 2022 Local infectious disease experience influences vaccine refusal rates: a natural experiment. Proc. R. Soc. B 20221986. https://doi.org/10.1098/rspb.2022.1986

It also serves as a portal for the remaining code and datasets used in the paper.

Specifically, the upload consists of:

The main analysis dataset, hosted directly in DataDryad as an Excel spreadsheet. This is the final dataset used to conduct the statistical analysis reported in the paper.
'Related Works: Software': a link to the R code used for the data transformations and analysis, and hosted on Zenodo. Note that this also includes a copy of the main dataset and the shapefiles in (3) below.
'Related Works: Dataset': a link to the shapefiles used to generate maps and other spatial manipulations required for the manuscript, and hosted on Zenodo

Each Related Work contains its own README files and/or appropriate documentation. The following is a description of the methods for constructing the main dataset (1).

Main dataset construction and processing

COV_Main_Dataset.xlsx contains data relating to smallpox and conscientious object to smallpox vaccination (COV), alongside socioeconomic variables, at municipal ward and registration district level for Glasgow 1900-1913. It contains the following worksheets:

 Data dictionary: provides descriptions of variables and explains abbreviations used.
 Ward_All: provides variables at municipal ward level, for all 25 wards that existed in Glasgow in 1900.
 W_Drop_Blyt_Exch: provides variables at municipal ward level, excluding two atypical central wards, Blythswood and Exchange.
 RD_All: provides variables at registration district level, for all 20 registration districts, or parts of these registration districts that cover the area of the 25 municipal wards that existed in 1900.
 RD_Drop_Blyt: provides variables at registration district level, excluding Blythswood where the maternity hospital was situated and where data are atypical.

To construct this dataset, data were first manually transcribed into Excel, primarily from Medical Officer of Health reports and census reports. They were then processed in R as described in the manuscript (code and further documentation are available in 'Related Works: Software', a repository that also includes the raw dataset as transcribed). The file provided here is the final dataset used for the statistical analysis reported in the paper.