Wellcome Library Visualisations

Public Health Reports

April 2016

I have been doing some work with a series of historic public health reports digitised by the Wellcome Library, the Medical Officer of Health (MOH) reports. This collection consists of over 6,000 fully-OCRed volumes between 1848-1972.

MOH reports, image credit: Wellcome Library blog

Example pages from the Medical Officer of Health reports

Rather than visualising quantitative data, I am interested in tracing commentary and attitudes around particular themes through time, and in digging into the text content over exploring the shape of the collection as a whole. I also wanted to experiment with using the text itself as markers within the visualisation rather than abstract shapes.

'typhoid carrier' visualisation

Detail from 'typhoid carrier' visualisation

These visualisations, which I built in JavaScript and D3, show every instance in all 6,000 reports of a particular search term with a snippet of surrounding text. Each snippet is horizontally centred on its date.

In this way, it's possible to see every instance of a search term in these reports with the context of what is being said, and to compare those contexts over time.

'heroin' visualisation

Detail from 'heroin' visualisation

Stacking the snippets in order of ascending date results in a slope stretching from order to newer. Additional meaning can be gained from the overall resulting shape. For instance, the sharp turn at the bottom of this visualisation for the term ‘putrid’ indicates a sharp drop in frequency of use of the term 1920 onwards.

Detail from 'putrid' visualisation

This is a work in progress, and next steps I want to take are:

  • to explore ways of visually grouping/removing repeated snippets (the previous year’s report was often used as a template).
  • to explore ways of highlighting or drawing attention to terms connected with the original search term – for instance commentary about typhoid sometimes mentions ice cream or dirty milk.
  • to review the general structure of these reports to check whether connected terms are actually in the same text passage, rather than being close together either side of a chapter break.
  • to explore zooming strategies for when the number of snippets in large. It is currently not possible to 'see everything' at a readable resolution where this occurs. (see ‘putrid’ visualisation, below)
  • 'putrid' visualisation