From Ariadne’s thread to the labyrinth itself: Nosology and the infrastructure of modern medicine

This May the World Health Assembly quietly adopted the 11^th version of the International Classification of Disease (ICD-11), the first of its kind to be fully electronic. The digital-only format is needed in part because the compendium has itself become too big to print: weighing in at a hefty 55,000 pathological entities, the ICD-11 represents a nearly fourfold increase from the roughly 14,400 codes in its predecessor, the ICD-10.¹ Like a 21^st-century Pandora’s Box, the ICD-11 suddenly introduced 40,000 new plagues into the world, a move that will have major impacts on the organization of medical knowledge and the logistics of medical practice. But the Geneva event received relatively little fanfare, as the classification of disease is typically understood to be a boring, mundane affair—a far cry from more newsworthy stories of pandemics on the move or new cures in the making.

As mundane as the shifting codes of the ICD-11 may seem, the systems we use to classify disease shape the nature of medicine and public health in substantial and powerful ways. Seventy percent of the world’s health resources are now allocated based on ICD data.² As clinicians, we interpret a patient’s symptoms in light of a differential diagnosis comprised of different disease categories and then plot the final diagnoses using ICD codes to get reimbursement from insurers. As epidemiologists we use statistical data compiled on the basis of these agreed-upon categories and then publish our findings in journals that tag by disease. As patients, too, we increasingly understand ourselves in regard to disease names that are applied to our symptoms,³ and then use these names to seek more information via search engines and social media. As the internist Knud Faber concluded almost 100 years ago, the modern physician “cannot live, speak, or act without the concept of morbid categories.”⁴ Yet for a larger period in the history of medicine, this was not the case. Obsession with diagnostic classification is a hallmark of modern medicine, and tracing the history of disease classification in medical practice since 1700 can reveal substantive changes in the infrastructures that undergird medical knowledge and medical practice in the present day.

Ordering the world of disease

While Western medicine has always employed categories of disease to guide diagnosis and therapy (for a patient with a cough, for example, phthisis was not the same thing as catarrh), there was little interest in formal classification systems as we now know them until the 18^th century.⁵ Most premodern Western medical practice was rooted in the axiom that each diseased body was ill in its own particular way. The ability to interpret signs and individually tailor treatment distinguished the learned medical practitioner from lesser practitioners who offered the same treatment to every patient independent of their individual constitution.^6–8 Historian Owsei Temkin has contrasted this individualized “physiological” mode of practice from more abstract “ontological” approaches to diseases that have the same form across all bodies.⁹

When the 17^th century London physician Thomas Sydenham suggested that morbid phenomena should be grouped together as botanists grouped plants in a taxonomy based on aggregate observations of similarity and difference, he was arguing for empiricism in a profession that valued more theoretical approaches.^10,11 Not until the Montpellier-based physician François Boissier de Sauvages returned to the topic in the 18^th century, in correspondence with his friend and colleague taxonomist Carolus Linnaeus, did a universal nosology of disease become part of Western medical practice.¹² After 20 years collecting, reading, and systematizing descriptions from thousands of medical authors from different times and places, Sauvages published the five-volume Nosologia Methodica Sistens Morborum, representing the world of illnesses as a system of 10 classes, 295 genera and 2400 species of disease.^13,14 Sauvages intended the Nosologia as a guide to the individual physician: he compared it to the mythological thread of Ariadne, and argued that just as Ariadne’s thread helped Theseus find his way through the Labyrinth in Crete to defeat the monstrous Minotaur, his nosology would safely guide the practicing clinician through the vast labyrinth of human suffering to prevail against disease.¹⁵

After Sauvages, the 18^th century saw a proliferation of nosological systems.^16–18 Like Linneaus’ Systerma Naturae and Diderot’s Encyclopédie, such systems suited the Enlightenment ideal of the rational physician. But over the course of the 19^th century these classification systems also became useful in the rationalizing project of medical statistics, a growing site for of medicine’s increasing immersion within state bureaucracies and other large institutions.¹⁹ The most prominent physician of the Scottish Enlightenment, William Cullen, originally published his Synopsis Nosologiae Methodicae as a guide to individual physicians in 1769.²⁰ By the middle of the 19^th century, however, the British physician and epidemiologist William Farr commented that its continued use by the public service was too vague for the modern statistical purposes of the state.²¹ In 1839, Farr developed one of the first state-commissioned nosologies of morbidity and mortality to rationalize the collection of medical statistics in Great Britain. Other cities, states, and countries soon followed suit.

The collection of medical statistics in the late 19^th century took on increased focus on the public health threat of infectious diseases, as a world made smaller by steamboats, trains, and canals fell under the thrall of the new medical taxonomy of microbiology and a series of International Sanitary Conferences sought to coordinate responses to epidemic disease. Beginning in 1853, the International Statistical Congress, and its successor the International Statistical Institute, held a series of meetings that culminated in the first ICD, the International Classification of Causes of Death in 1893, based on an adapted and expanded version of Farr’s system developed by Jacques Bertillon, the Chief of Statistical Services of the City of Paris. By August 1900, a detailed classification of causes of death was published containing a mere 179 fatal illnesses. At the same meeting, a decision was made to develop a further list of diseases where the outcome was not fatal, though like many other international projects this was interrupted by the First World War.²²

In the aftermath of the first World War, the collection of medical statistics became a plank for new international congresses seeking to build order and stability through the sharing of information between states. Joining forces with the International Statistical Institute, the health division of the League of Nations drafted the proposals for the fourth and fifth International Classification of Death while simultaneously working on a broader International Classification of Disease. Beyond the member states of the League of Nations, many nonstate actors were also interested in the development of a universal taxonomy of disease, including the many private hospitals and health insurance companies now playing an increased role in the delivery of early 20^th century medical care. Yet it was not until 1948, with the sixth revision of the ICD, that a comprehensive list of non-fatal diseases was finally incorporated in the newly-renamed International Classification of Diseases, Injuries and Causes of Death, announced at the inauguration of the World Health Organization in Geneva that year.

The ICD was meant to be revised every decade: a committee would meet in Geneva in 1955 to form the ICD-7, in 1965 to form the ICD-8, in 1975 to form the ICD-9, and so forth. Each subsequent meeting saw an expansion not only in the number of diagnostic categories but also in the number of private actors in the room. ICD-8 contained entirely new functions dedicated to hospital indexing, while ICD-9 expanded the functions of classification to include the health management role, supporting the indexing of medical records, medical care review and ambulatory and other medical care programs.²³ As the ICD shifted from a guidebook to a system requiring the agreement of a wide variety of stakeholders, the lag between publication and implementation grew apace. ICD-8, published in 1965, was not implemented in the US until 1968. ICD-9, published in 1975, would not be implemented in the US until 1979. ICD-10, published in 1990, would not be implemented in the US until 2015. The earliest possible date that the ICD-11 will be implemented in the US is 2022—and if history is any guide the actual implementation is likely to take much longer than that.²⁴

Pathological negotiations

In contrast to 18^th and 19^th century taxonomies, the ICD has become a vast information system: no longer an Ariadne’s thread but a labyrinth unto itself. The ICD is also the case study for what sociologists Geoffrey Bowker and Susan Leigh Star call an information infrastructure – embedded in myriad databases, supporting work across a series of medical, financial, public health, and governmental institutions, with international reach, and linked with a series of standards and conventions.²⁵ As the ICD has become more central to day-to-day medical practice, medical economics, and medical research, it has also become a flashpoint for the politics of health. Much of the delay in the ratification of ICD-11 (which according to the original calendar should have been finalized in 1995) is due to the fact that developing new taxonomies of disease has become increasingly contentious among physician, patient advocacy groups, and other civil society groups.

Among the most controversial changes to the ICD-11 has been the contest over the pathologization or depathologization of transgender. Diagnostic categories in ICD-10 like “transsexualism” and “gender identity disorder of children” belonged to the chapter of mental health and behavioral disorder. During the ICD-11 revision, several parties argued that gender incongruence should no longer be regarded as a disease, and should be taken out of the classification system altogether in order to remove the stigma of mental disorder, much as homosexuality was removed from the ICD-10 in 1990 (three years after it was removed from the Diagnostic Standard Manual of the American Psychiatric Association, whose use has also spread internationally in recent decades). On the other hand, demedicalizing transgender risked the loss of reimbursements for medical and surgical interventions. After a long process involving debates among researchers, clinicians and activists,²⁶ the WHO settled on including the new category of “gender incongruence” in a new chapter of sexual health in order to ensure access to and coverage of gender-affirming health care.²⁷ As member states work to implement this change, economic and legal consequences loom for psychiatrists in the field, since trans-related and gender diverse identities are no longer considered mental health diagnoses.²⁸

Like its predecessors, the ICD-11 is the product of longstanding negotiation, compromise, and inertia. Classification systems are by nature slow and laborious to change, as they are bound up with infrastructures that shape the daily habits of clinicians, policymakers, economists and researchers. The result is a seemingly inevitable process of diagnostic accumulation: it is easier to add several more diseases than to subtract a single category. Moreover, since radical changes to any classificatory system cause trouble for the continuous collection of statistics, these changes tend to be both gradual and path-dependent. Past precedents remain engrained in medical classifications long after they would have been considered outmoded in daily practice; as the sociologist Mildred Blaxter has suggested these systems are best viewed as a “museum of past and present concepts of the nature of disease.”²⁹

Evolving infrastructures of medical practice

Reading the series of classificatory systems for disease from the 18^th century to the present, one can trace both a history of disease concepts over time and a history of the changing media of medical information. Where Sauvages and Cullen produced books to help the medical student and individual practitioner, Farr and Bertillon created standardized paper forms to render the tabulation of medical statistics possible. The ICD committees of the late 20^th and early 21st century, in turn, created an digital infrastructure that potentiated a dream of universal biomedical knowledge—but also served to further extend the reach of hospital administrators and insurance claim adjusters into the governance of daily clinical practice. If the physician in 1959 needed a diagnosis in order to choose the proper course of treatment, the physician of 1989 needed a diagnosis in order to bill for services, whereas the physician of 2019 needs a diagnosis in order to even write a prescription or sign a chart note from their electronic medical record.

The framers of the ICD at the turn of the 20^th century wanted the system to have no more than 200 categories, to fit the maximum length of a paper census form.²⁵ Freed from the constaints of the 8½ x 11 sheet of paper, the all-digital ICD-11 will be online, multipurpose, and multilingual. The newest version is designed as a database encompassing a foundation of disease classifications, linked directly with medical nomenclature systems, so that coders will not need to learn lists of codes anymore. Its electronic structure enables onscreen use in the clinic directly integrated into electronic health records, with the hope that disease classification will be a seamless aspect of clinical work in the e-health environment. Advocates for the new system claim ICD-11 will allow epidemiologic data to draw directly into clinical practice in a way that ICD-10 did not, and that the new platform will serve as a point-of-care information aid. In turn, the new electronic platform is meant to “crowdsource” diagnostic expertise, as users directly contribute to improvement of ICD by writing comments, improving translations and so forth. The success or failure of these attributes in practice ICD-11 will have everything to do with how useful the different stakeholders involved—clinician, researcher, insurer, or possibly event patient—find these categories to be in engaging with their own piece of the information structure of modern medicine.

The nosologies of the 18^th century aspired to provide practical knowledge (logos) about disease (noso-) for the individual clinician. If the expansion of the ICD system over the late 20^th century has had more practical uses outside of the clinic, a major aim of the new ICD-11 is to demonstrate its usefulness to the practicing clinician once more. Once implemented, the new system will affect the way we tabulate data about the prevalence and prognosis of conditions, and the ways in which we decide which forms of treatment are reimbursed. And even more importantly, these picayune details of how to name and order the world of disease will have an impact on how we constitute the normal and the pathological in the contemporary world.

As a central informational infrastructure of medical practice, the ICD-11 will have powerful effects not only on how the task of medicine is carried forward for future generations, but also on how health and disease are understood by both practitioners and patients. It is easy for the practicing physician, goggling at the ballooning number of ICD codes, to feel ensnared in an increasingly Kafkaesque architecture that constrains rather than aids their ability to practice humanistic medicine. Yet at the same time, this new classification system, linked to component health information systems and electronic health records, now promises a return to Sauvages’ original intent for nosology to lend a helping hand to the clinician, by creating information that can be used in the clinical setting where it is produced. Whether the newly digitized ICD-11 will serve as a string or as labyrinth will have much to do with its implementation—when, not if, that arrives.