Reconstructing A Century of Data: Transcription of Annual Reports From Alabama State Hospitals

The University of Alabama at Birmingham Libraries historical collections has archived annual activity reports from state-run hospitals. The collection includes 73 digitized reports from 1869 to 1964 which provide detailed recordings of the yearly activities at these institutions. The content has potential value for academic pursuits in sociology, history, anthropology, and psychology, especially those focusing on healthcare disparities in the U.S. Deep South during this era. However, the data within the reports are preserved in a format incompatible with modern quantitative tools. We launched a collections-as-data project to transcribe data from these reports to facilitate their use in research and instruction. We first created an electronic index of all tables and their location within each report. The index captures 164 types of data tables, 34 of which were only present in a single year. On average, a table was present in 16 (22%) of the reports, and the median was considerably lower at 7 reports. Items with consistent entries over time would presumably better reflect longitudinal trends in the operations of these facilities, thus we created a priority list of the most regularly reported data tables. We chose three tables from the list for transcription: annual hospital census (present in 99% of reports), patient cause of death (95%), and patient county of residence (75%). Transcription challenges included errors in the annual reports exposed by quality control measures, and variability in terminology. Lastly, using the DataOne framework, we developed lesson plans that incorporate the transcribed data. Future work includes transcribing additional tables, piloting the lessons in a sociology course, and publishing the data corpus. This session will discuss lessons learned during the process illustrated by examples from the datasets. Attendees will gain practical insights into building a workflow to transcribe data from historical collections using a FAIR and quality-controlled approach.