Formal Idiographic Inference in Medicine

Abstract

Fifty years ago, Alvin Feinstein published his seminal book Clinical Judgment 1 in which he argued that medicine needs reliable clinical data, a sound taxonomy of disease, and attention to the nature of clinical reasoning. The years following saw the establishment of modern evidence-based medicine (EBM), whereby the best medical decisions about a current patient are made by applying rigorously established outcomes of previous patients. 2 The key instrument in both of these perspectives is a taxonomy of disease in which similarities among patient conditions are reflected in patient assignment to similar disease classes. 3 Courses of disease in previous patients with a particular diagnosis then serve as predictions for the disease course of a current patient with the same diagnosis. Interpatient variability within a disease class is unaccounted for with this approach and is not formally exploited for treatment decisions.Precision medicine seeks to remedy the challenge of patient heterogeneity by subdividing diagnostic classes more finely. 4 With expanding ability to compile data from more patients, additional disease subclasses provide greater potential to stratify patients into smaller, more similar cohorts for bettertailored treatment decisions. A fundamental limit of t h i s p h i l o s o p hy m a n i fe st s w h e n d i ag n o st i c subdivisions become so fine as to preclude a sufficient number of similar previous patients for adequately informing decisions about the current patient. Even big data will not solve this inherent problem because with fine enough subdivisions, all diseases become rare diseases, accompanied by all the associated challenges in identifying appropriate evidence from enough similar previous patients. To summarize the formal EBM philosophy, data from previous patients are used to construct classes of disease whereas data from the current patient are used to place that patient into the proper class. Treatment proceeds for the class, which is sometimes referred to as "treating the diagnosis."In her book How Doctors Think: Clinical Judgment and the Practice of Medicine, 5 Kathryn Montgomery describes medicine not as a science but rather as an "interpretive practice." Physicians obtain a patient's history, physical examination, and the results of relevant diagnostic tests, then juxtapose this information with clinical experience and empirical studies to construct a tentative account of the illness. This description is reminiscent of Hippocrates of Kos's arguments against his archrivals practicing in Knidus, where "diseases were divided and subdivided arbitrarily into endless varieties." Hippocrates's view was that "to accidental varieties of diseased action there was no limit, and that what was indefinite could not be reduced to

Computational Audiology: New Approaches to Advance Hearing Health Care in the Digital Age

“…2001). Individual refinement based on learning systems could be a promising way forward but raises many challenges to perform in an evidence-based manner (Barbour 2018).…”

Computational Audiology: New Approaches to Advance Hearing Health Care in the Digital Age

“…Historically, it has been restricted mainly to cohort-level inference, for example, in understanding the incidence and degree of hearing loss in the general population (Mościcki et al, 1985), and the prescription of sound-amplification for different types and degrees of hearing loss (Byrne et al, 2001). Individual refinement based on learning systems could be a promising way forward but raises many challenges to perform in an evidence-based manner (Barbour, 2018).…”

Precision medicine and the cursed dimensions

“…Clinical observation models can combine theoretical knowledge, empirical findings from previous patients, and past observations from the current patient to deliver predictions regarding new observations from the current patient. 9 They capture the multidimensional interactions of latent clinical factors specific to each patient.…”