Medical practices display power law behaviors similar to spoken languages

Abstract

BackgroundMedical care commonly involves the apprehension of complex patterns of patient derangements to which the practitioner responds with patterns of interventions, as opposed to single therapeutic maneuvers. This complexity renders the objective assessment of practice patterns using conventional statistical approaches difficult.MethodsCombinatorial approaches drawn from symbolic dynamics are used to encode the observed patterns of patient derangement and associated practitioner response patterns as sequences of symbols. Concatenating each patient derangement symbol with the contemporaneous practitioner response symbol creates “words” encoding the simultaneous patient derangement and provider response patterns and yields an observed vocabulary with quantifiable statistical characteristics.ResultsA fundamental observation in many natural languages is the existence of a power law relationship between the rank order of word usage and the absolute frequency with which particular words are uttered. We show that population level patterns of patient derangement: practitioner intervention word usage in two entirely unrelated domains of medical care display power law relationships similar to those of natural languages, and that–in one of these domains–power law behavior at the population level reflects power law behavior at the level of individual practitioners.ConclusionsOur results suggest that patterns of medical care can be approached using quantitative linguistic techniques, a finding that has implications for the assessment of expertise, machine learning identification of optimal practices, and construction of bedside decision support tools.

Empirical analysis of Zipf’s law, power law, and lognormal distributions in medical discharge reports

“…Our CCDF plots show a similar behaviour, with the power law fitting the tail of the distribution properly, and deviating from the head of the discharge report data and subsections. Two other studies [18,19] also showed that the terms with the highest frequency (rank 1-10) deviated from the power law, and the long-tail was fit well by the power law distribution.…”

“…"gastric cancer") and (2) name codes (six characters long and building on ICD10 codes) [18]. One other study explored whether patient physiologic derangements encoded as symbols followed a power law distribution, the data generated from virtual patients in a digital training simulator [19]. Epidemiology of GP clinical encounters has also been found to follow a power law distribution [20].…”

“…To date, no work has explored Zipf's law in medical and clinical language, such as analysis discharge report text. Zipf's law has been explored in clinical codes [17,18], clinical diagnoses [18], virtual patient physiologic derangements [19], and epidemiology [20], but not in unstructured medical text. We aim to test whether medical discharge reports in the MIMIC-III dataset [21] follow Zipf's law, a power law distribution, and assess the fit of alternative probability distributions.…”