The Book
Digitizing Diagnosis: Medicine, Minds, and Machines in Twentieth-Century America
The Author(s)
Andrew S. Lea
In a classic 1996 episode of Seinfeld, Elaine Benes finds herself in a doctor’s office and manages to take a peek at her chart. Her medical history is all there, but what she notices immediately is another physician’s comment that she’s a “difficult” patient – based on an encounter four years earlier. Typically, Elaine cannot let go of the enervating issue, even though she was there to see about a rash. Why was I marked down as difficult, she asks. The doctor shoos her along, but the stain on her record is unremoved.[1]
Most of us never think about what’s on our chart or electronic medical records (EMRs). But we all have a faint sense that these details and woes are being documented somewhere. Andrew S. Lea’s new book, Digitizing Diagnosis, gives a truly eye-opening view into the ways that the protean and rascally host of symptoms that patients present get turned into quantifiable information.
In the process, Lea lays out some fundamental questions that are both practical and epistemological: how do doctors most effectively turn a patient’s personal narrative into a diagnosis? Are there ways to make the process more efficient, especially given the overwhelming demand for care and the limited resources of doctors, nurses, and hospitals? Could a paper form or a computer program do a better job of nailing down what is really “wrong” with a patient than a human physician could do?
Think about going to a psychiatric appointment. Often, you will be asked to fill out a form that ranks how distracted you’ve been in the last week, how often you’ve had suicidal thoughts, or how often depression or anxiety has interfered in your work, on a numerical scale. The doctor can then determine, “Well, you were a 3.3 last month but you’re a 3.1 now, so you must be improving…” Lea’s Digitizing Diagnosis is a story about how this kind of methodical and standardized assessment came to overtake medicine in the United States, to a significant yet stubbornly limited extent.
Lea joins a variety of scholars who have considered how knowledge in the twentieth century got turned into “information.” Historian Craig Robertson’s The Filing Cabinet: A Vertical History of Information (2021) partners well with Digitizing Diagnosis, as it shows how the materiality and specificity of particular organizational tools shaped how we understand the world around us, often in ways not very apparent to the average person or even to their inventors. Lea’s book also echoes Sarah E. Igo’s influential The Averaged American: Surveys, Citizens, and the Making of a Mass Public (2008), which showed how other heuristics, such as public opinion polls, came to define how we think about the public as quantifiable units of analysis.[2]
Lea rightly chooses diagnosis as his subject, since the act of assessing a patient’s needs can depend on judgments both subjective and statistical, artisanal and automated. Diagnosis is the spear of the sword when it comes to professionals such as doctors (along with lawyers and other practitioners) asserting that their own unique insight and training should prevail, against the idea that a computer can do better, without the biases and mishaps of fallible humans. Hence, Digitizing Diagnosis looks at early efforts in the 1940s and 1950s to use paper questionnaires to catalog the problems of patients, intended to diminish the demands on a busy physician to take an extensive personal history themselves. This standardization almost inevitably led to efforts to computerize diagnosis, which sought to prove that a data program could correlate patients’ complaints into groups or clusters and yield a more effective judgment than the physician could do just by talking to a patient.
Predictably, this foray into electronic standardization raised the hackles of some medical professionals – much as the debate about automation pervaded manufacturing and other fields in the 1950s and 1960s. Doctors pushed back on the idea that a questionnaire or computer program could substitute for their learned judgment, as well as their intuitive connection with individual people. As Lea reveals, the quest for standardized questionnaires ran into its own difficulties. When such tools were deployed around the world, translation difficulties and cultural differences meant that patient responses to a survey about their health were nowhere near as uniform or telling as was hoped.
The structure of Digitizing Diagnosis does well to elucidate the core issues for readers. The first part, “Patients,” deals with the aforementioned efforts to document and analyze symptoms. The second part of the book, “Disease,” addresses the basic question of how illnesses are defined, and whether computers could correlate symptoms and categorize them into some amorphous constellation like “borderline personality disorder.” The final section, “Physicians,” looks at how actual practitioners on the ground attempted to implement the new electronic systems that would, allegedly, assist them.
Throughout, Lea centers the fundamental question behind all attempts to automate or digitize medicine: how do we use an algorithmic logic to improve patient care, without at the same time endangering it? Could automation be a “cure” that is actually worse than the disease? “How could human physicians hope to fully understand a computerized system that took many individuals, working across many years and many disciplines, to develop?” Lea asks toward the book’s conclusion. “How could physicians be expected to place matters of life and death in the hands of faceless algorithms? What kinds of biases might become calcified by an opaque, computerized algorithm?”[3]
In today’s world, agonized as it is by concerns about artificial intelligence, Digitizing Diagnosis is a richly rewarding read. Andrew S. Lea reminds us that we have already gone through many wrought debates about new technologies that might threaten jobs and human agency, especially in the post-1945 era. This book joins the work of scholars such as Amy Sue Bix, Howard Brick, and Jason Resnikoff, who have also traced controversies over automation into their unsettled, frequently unexpected fallouts.[4]
More than anything, Lea’s book evokes Paul N. Edward’s 1997 classic The Closed World: Computers and the Politics of Discourse in Cold War America.[5] That book followed the rise of computing and cognitive science amid the Cold War, when people began to think of their computers as minds and their minds as computers, modeling one after the other. Lea ably shows us how these mind-machine metaphors unfolded simultaneously in the field of medicine. But in healthcare, the metaphors don’t matter just in the sense of whether a piece of code or hardware works, but whether the human body works. And that is a place where the ways we think about evidence, reasoning, and judgment matter more than anything else.
[1] “The Package,” dir. Andy Ackerman, Seinfeld (October 17, 1996).
[2] Craig Robertson, The Filing Cabinet: A Vertical History of Information (Minneapolis: University of Minnesota Press, 2021); Sarah E. Igo, The Averaged American: Surveys, Citizens, and the Making of a Mass Public (Cambridge: Harvard University Press, 2008).
[3] Andrew S. Lea, Digitizing Diagnosis: Medicine, Minds, and Machines in Twentieth-Century America (Baltimore: Johns Hopkins University Press, 2023), 173.
[4] Amy Sue Bix, Inventing Ourselves Out of Jobs?: America’s Debate over Technological Unemployment, 1929–1981 (Baltmore: Johns Hopkins University Press, 2002); Howard Brick, Transcending Capitalism: Visions of a New Society in Modern American Thought (Ithaca: Cornell University Press, 2006); Jason Resnikoff, Labor’s End: How the Promise of Automation Degraded Work (Champaign: University of Illinois Press, 2022).
[5] Paul N. Edwards, The Closed World: Computers and the Politics of Discourse in Cold War America (Cambridge: MIT Press, 1997).
About the Reviewer
Alex Sayf Cummings is a professor of History at Georgia State University. She is the author of Democracy of Sound (Oxford, 2013) and Brain Magnet (Columbia, 2020).
0