Digital epidemiology, deep phenotyping and the enduring fantasy of pathological omniscience

Abstract

Epidemiology is a field torn between practices of surveillance and methods of analysis. Since the onset of COVID-19, epidemiological expertise has been mostly identified with the first, as dashboards of case and mortality rates took centre stage. However, since its establishment as an academic field in the early 20th century, epidemiology’s methods have always impacted on how diseases are classified, how knowledge is collected, and what kind of knowledge was considered worth keeping and analysing. Recent advances in digital epidemiology, this article argues, are not just a quantitative expansion of epidemiology’s scope, but a qualitative extension of its analytical traditions. Digital epidemiology is enabled by deep and digital phenotyping, the large-scale re-purposing of any data scraped from the digital exhaust of human behaviour and social interaction. This technological innovation is in need of critical examination, as it poses a significant epistemic shift to the production of pathological knowledge. This article offers a critical revision of the key literature in this budding field to underline the extent to which digital epidemiology is envisioned to redefine the classification and understanding of disease from the ground up. Utilising analytical tools from science and technology studies, the article demonstrates the disruptive expectations built into this expansion of epidemiological surveillance. Given the sweeping claims and the radical visions articulated in the field, the article develops a tentative critique of what I call a fantasy of pathological omniscience; a vision of how data-driven engineering seeks to capture and resolve illness in the world, past, present and future.

Digital tools for the assessment of pharmacological treatment for depressive disorder: State of the art

“…This discussion is to be seen in te context of precision psychiatry, in the hope that with clear and deep phenotyping, integrating biological, behavioural, anamnestic (subjective) and objective information, stratification and profiling of patients can become more precise and help optimising treatment choices and the course of treatment. Some colleagues warn for an idealistic view and assessment of the possibilities of emerging digital tools and analysis techniques for healthcare ( Engelmann, 2022 ), on the other hand, there is still much to be gained from a better phenotypic understanding, more reliable stratification, and pharmacological options adjusted to the patients' unique pro-file, despite heterogeneity; towards a bottom-up reorientation of diagnostic classification ( Salagre and Vieta, 2021 ).…”

Short-circuiting biology: Digital phenotypes, digital biomarkers, and shifting gazes in psychiatry

“…Bycroft et al, 2018). However, Engelmann (2022) has suggested that the notion of "depth" in digital phenotyping is ambiguous and increasingly connected with the metaphors and language of artificial intelligence (e.g. DeepMind, DeepBlue, DeepFace, and DeepMood), where digital phenotyping is supposed to "provide causal depth" to illness phenotypes by going beyond the surface of "shallow" medicine by using techniques such as deep learning.…”

Digital phenotyping – Editorial

“…Baumgartner introduces the position of digital phenotyping in the project of precision, or personalised medicine, to clarify how the exploitation of data is envisioned to shift health care towards preventive and proactive interventions. Engelmann's contribution looks at the spurious claims that inform research built on the automated prediction of depression on Twitter (Engelmann, 2022), while Milne et al focus on the 'behavioural surplus' of surveillance capitalism (Milne et al, 2022;Zuboff, 2019), where the recognition of Alzheimer's disease is potentially entangled with voice-activated home assistants. Rowe's article introduces 'Opioid360', a system to determine risk factors for potential opioid addiction to guide and improve prescription practices (Rowe, 2021).…”

“…Within the second dimension, the authors map out the rapid uptake and analyse the outstanding popularity of digital phenotyping in mental health research and digital psychiatry (Pickersgill, 2019). Engelmann's article draws attention to the often-shallow underpinnings of the promises of deep medicine, when the assumptions and presumptions that inform the prediction of depression of Twitter users are scrutinized (Engelmann, 2022). Milne et al demonstrate in their paper how digital phenotyping extends and exaggerates a biological reductionism in the research environments of Alzheimer's disease (Milne et al, 2022), while Rowe shows that the analysis of social factors of addiction within a digital phenotyping context contributes to their individualised assessment (Rowe, 2021).…”

“…Baumgartner asks for increased analytical attention to the use and analysis of data in precision medicine and digital phenotyping, so that the often implied 'genetic determinism' and 'data fundamentalism' are acknowledged (Baumgartner, 2021). Engelmann investigates the depth of 'deep medicine' and asks how the emergence of digital phenotyping has been coupled to the reinvigoration of the dream of total archives, risking the introduction of totalising and normative conceptions of health and disease (Engelmann, 2022). Wackers rereads Dawkins' Extended Phenotype and uses the notion of the genotype to explore the problematic of probabilistic inference of causality; and the fallacy of the reification of generative, causal infrastructures (genotypes) underlying phenotypes derived from mathematical abstractions purported to express the strength of association (correlation) between variables defined as phenotypic that were assumed to share a common internal cause (Wackers 2022).…”