Gathering and Interpreting Data about Tuberculosis in the U.S.

In nineteenth century America, tuberculosis accounted for nearly one out of every ten deaths. Known most commonly as “consumption,” this disease was dreaded across society because it affected all age groups, cut across social and class lines, was incurable, and often resulted in long debilitating illnesses. To explore the social effects of tuberculosis, a group of Virginia Tech undergraduates spent the spring 2015 semester exploring “That Dread Disease,” using newspaper obituaries to document the lives lost to this disease. In this second of three posts, Circulating Now welcomes guest bloggers Veronica Kimmerly and E. Thomas Ewing, who look at techniques for measuring the impact of this disease.

Does knowing how many people die from a particular disease contribute to efforts to control the number of deaths from this cause? More than a century ago, statisticians employed their tools and techniques for gathering mortality data to support the widespread public health campaign against tuberculosis, the most common cause of death in the United States. In his “letter of transmittal” for a 1908 U.S. Census Bureau report on Tuberculosis in the United States, Oscar S. Straus, Secretary of Commerce and Labor, declared emphatically that collecting statistics was vital to public health: “No more important service can be rendered by mortality statistics than the compilation of data which will prove of practical assistance to those actively engaged in the prevention of disease.”

a paper card with rows and columns of numbers some of which have been punched out.

Punch Card from Tuberculosis in the United States, prepared for the International Congress on Tuberculosis, Washington, September 21 to October 12, 1908.

This moment in history was significant not only for launching the global campaign against tuberculosis, but also as a time when data collection was undergoing a dramatic transformation. Punch cards, which had been first used during the 1890 census, were increasingly used by health officials to document causes of death. As explained in the 1908 publication, the compilation of tuberculosis statistics required a sequence of contributions, starting with the attending physician or coroner who identified the cause of death, the undertaker who completed a death certificate, the local registrar who recorded each death in a register and compiled regular reports, the state registrar who indexed these returns and published them in monthly or annual reports, and finally the US Census office, which received all transcripts from the states and classified the deaths according to standard categories. At this point, the statistical information was “transferred to a card by means of a punch,” leading to the only automated step in the whole process: “The punched cards are then compiled by means of electric tabulating machines, which record on many dials the various items necessary for the statistical table.” After this compilation was completed by the machines, the resulting tables were arranged, edited, and published as the annual report on mortality statistics, “which is distributed to state and city sanitary authorities, physicians, foreign governments, etc., and is for the use of all persons interested in the condition of the public health of the United States.” As conceded by the document’s authors, “accurate mortality statistics and cooperation of many individuals,” was needed, including the participation of physicians, coroners, undertakers, registrars, and statisticians in the census office. Improving the collection, recording, and analysis of mortality figures was determined to be an essential tool, as shown in this sweeping statement: “The vital statistics of the future must be statistics of greater precision, for they are to be the working tools of an aggressive scientific public health administration.” Modernizing the process to include automation to reduce errors was therefore of great benefit because the punch cards, as “electrical tabulating machines” had reduced the “tabulation of punch cards” to “a mechanical basis and thus largely eliminated the ‘personal equation’.”

A pie chart showing rates of mortality for different causes for 1907; Tuberculosis ranks first at 21%.

Death Table from Tuberculosis in the United States, prepared for the International Congress on Tuberculosis, Washington, September 21 to October 12, 1908.

Counting the number of tuberculosis victims was considered an especially urgent public health measure, because it was the single leading cause of death for the American population. In an 1893 publication, Practical Measures for the Prevention of Tuberculosis, Lawrence F. Flick issued this imperative statement to the medical profession: “…we cannot, without failing in our duty, remain inactive while this disease carries off hundreds of thousands of human beings to a premature grave and inflicts indescribable suffering and loss upon hundreds of thousands of others.” The importance attached to documenting tuberculosis cases led Flick to declare that “registration” of tuberculosis victims was the first step in preventing further spread of disease. Responding to the perception that preparing reports was too time consuming, Flick declared confidently that “…with properly prepared blanks, the making out of reports can be termed neither a labor nor a burden, and could easily be attended by the busiest practitioner.”

Just over one hundred years later, a group of Virginia Tech undergraduate researchers has been trying to answer a similar question about the loss of life due to tuberculosis. Rather than relying on census reports derived from death certificates, the students searched historical newspapers for cases of individuals who died of consumption. For each individual, the students entered available information about key categories such as age, gender, and race of the deceased and also the length of their illness. If the obituary provided additional information about individuals’ family history, particularly in case where family members also died of consumption, this information was entered into the form as well. This research project, which studied approximately fifteen newspapers from Virginia, located just over one thousand individuals who died of consumption in the decades from the 1866 to 1922.

The data available from the Virginia newspapers can be usefully compared to similar results from the published documents of the U.S. Census in the table below giving the age of victims relative to death rates. The results of the students’ research demonstrates that obituaries reported deaths in patterns similar to census and other statistical reports, but with greater detail and particularly more attention to context, relations, and connections. Comparing age at death is one way to compare data from obituaries with census reports. More than one-half of obituaries (n=576) allowed researchers to identify a specific age of death. The accumulated data confirms that consumption caused the most deaths in the age group, 20-40 years, with 52.2% of deaths in newspapers and 44.5% of census deaths reported from this age group. This table also shows some discrepancies: the census reported that 6.5% of consumption deaths were under the age of 10, whereas the newspapers reported less than 1% from this age group. This gap can be explained by the fact that obituaries tended not to report infant deaths, especially from families of low status. Newspapers reported a higher proportion of deaths in the age group 40-60 years, suggesting that status in the community meant that victims were over-reported in newspapers.

Age at Death
Newspaper Reports, 1880-1920 Census Report on Virginia deaths, 1880
Number Percent Number Percent
1 – 10 years 5 0.9 39 6.5
11 – 20 years 67 11.6 59 9.8
21 – 30 years 181 31.4 160 26.5
31 – 40 years 120 20.8 109 18.0
41 – 50 years 81 14.1 82 13.6
51 – 60 years 73 12.7 63 10.4
61 – 70 years 30 5.2 55 9.1
71 + years 19 3.3 37 6.1


Health officials continue to seek answers to the question: How many people die each from a particular disease? Mortality statistics in the United States are compiled from death certificates, through a process by which the single underlying cause of death listed on each certificate is encoded with a numeric entity reference number, thus generating a number of deaths from each cause. Every year, an estimated 610,000 people in the United States die from heart disease, making this health condition the leading cause of death annually. The second most common cause, cancer, resulted in nearly 590,000 deaths in the United States in 2015. Government agencies such as the Centers for Disease Control and Prevention or the National Institutes of Health and private organizations such as the American Lung Association or the American Cancer Society use mortality statistics to measure the effectiveness of public health campaigns, medical procedures, and pharmaceutical interventions. More than one hundred years after the U.S. Census began using punch cards to tally the causes of death, health experts continue to rely on this data to make decisions about how to invest in public health improvements.

In these posts Virginia Tech students, working under the direction of Professor Tom Ewing, explore themes related to tuberculosis, including reporting on medical discoveries, the cultural implications of advertising cures, and techniques for measuring the impact of this disease. By studying these patterns in the past, historians of medicine can contribute to contemporary and future responses to infectious diseases. This research made extensive use of the digitized collections of the National Library of Medicine as well as Medical Heritage Library in the form of national and Virginia medical journals, photographs and other images, and published books. More information about the project, including an online presentation documenting Virginia lives lost to tuberculosis, is available from the project site:

Veronica Kimmerly is a Chemical Engineering & Mathematics major at Virginia Tech.

E. Thomas Ewing is Professor of History and Associate Dean in the College of Liberal Arts and Human Sciences at Virginia Polytechnic Institute and State University.