text
stringlengths 1
1.75k
|
|---|
Text |
Learn to Ask: Patient-Generated Health Data, Artificial Intelligence, and the Synthesis of Care |
Abstract |
[...] |
Healthcare is both an intersubjective experience and an increasingly automated system for making meaningful decisions on the basis of captured data. Two evolutionary tendencies in healthcare technology indicate a trend towards drawing these disparate characteristics into closer alignment, first: the arc initiated more than forty years ago by the shift from paper to electronic health records (EHR), which continues apace with the introduction of artificial intelligence (AI); and second: the use of patient-reported outcome measures (PROMs) as instruments for subjective, embodied, and context-rich measurement of patient experience. While the consequences of this convergence have only begun to be studied from a clinical standpoint, this dissertation seeks to understand the cultural and social ramifications of computational approaches to understanding patient experience, by placing large scale data analytics into a shared context with the medical humanities and disability studies. To this end, three separate approaches are interleaved: First, an exploration of the possibilities for analysis as a precursor to imagination, centered on the use of patient-generated health data and AI. Second, a narrative review of technical and scientific literature that addresses how PROMs and AI may be used together. The aim of this review is to collect the various narratives that produce, and are produced by, this convergence. In sum, these narratives are seen to constitute a sociotechnical imaginary, or collective sense of what is possible. The third method applies creative techniques—poetry, lyric essay, and image-making—to highlight the need for artifacts to understand and navigate intersubjective experience of both illness and technology. |
Keywords: |
Medical Humanities, Disability Studies, Critical AI Studies, Patient-Reported Outcomes, Medical Technology: Social and Cultural Effects |
Introduction |
Overview: |
I am most struck by studies that show people survive more when they fill out surveys about how they are feeling. |
A visit to the clinic is filled with pointed questions, asked of the patient by no one in particular—a portal, a prompt, automated and oriented by compliance. Questions about status, satisfaction, sadness, and work; about getting out of bed, lifting heavy objects, and how much of our social life has been interrupted. Some questions seem to recognize the specificity of our experience, while others seem wholly misdirected. What pattern could capture us completely? |
In my effort to explain the sensations of having a bodymind inside of automated systems of capture, I will refer to experiences which you, reader, may not share. This is ok: these gaps between us—intervals—are the basis for my efforts, where shared experience resides as allegory, recognition, and room to learn. |
There are three main axes for this dissertation: |
1. Analysis and Imagination |
2. Artificial Intelligence (AI) and Patient-Generated Health Data (PGHD) |
3. Informed Consent and Shared Decision Making |
Each of these axes represents a possible assemblage, or synthesis, of different approaches to care. Analysis is a precondition for imagination: what we learn, combined with what we know, helps us to confront what we think of as possible. The integration of automated analysis and prediction (AI) with first-person testimony (PGHD) is enthusiastically promoted as a means of easing the burden of disease by better understanding a patient’s experience in context. At the same time, this integration carries enormous risk if not implemented ethically, and with respect for the whole patient. As a cautionary example at the center of this integration, informed consent practices fail to keep pace with changes in both technology and moral imagination. Attuned to compliance requirements, these practices are insensitive to patient needs. This failure motivates an embrace of shared decision making as an ethical approach to building trust and comprehension. However, both of these models are supported by a set of assumptions about autonomy and dependence that continue to shape what is possible in terms of law, technology, and imagination. |
Aim: |
This dissertation seeks to locate possibilities for incorporating patients’ points of view into systematized models of care, particularly those premised on the use of predictive analytics, automation, and artificial intelligence. This process of incorporation, or synthesis, is held up as a model for producing better representations of, and insight into, the experience and effects of illness, with the overall goal of improving patient experience. |
The research as it is presented here is intended to be plainly legible to a general audience. In particular, it is written for those who have, as a patient or caregiver contending with serious illness or disability, become entangled with medicalized technologies. It is presented with an interdisciplinary curiosity and experimental spirit, in the hope of opening numerous points of entry to the concepts which it explores. |
It is further hoped that the questions addressed here will resonate closely with those whose expertise in patient-centered medicine—clinicians, researchers, patient advocates and policy makers—has led to similar lines of inquiry. |
Outline: |
There are three analysis sections, each takes a different path. There are two creative texts (“Self-Writing” and “Sentences about rivers and cancers”). There are many illustrations. There is a glossary. There is an exhibition text (“There is almost not an interval”). |
Analysis 1: Analysis Helps Us to Imagine Better |
* Format: Critical analysis |
* Methodologies: post-qualitative, non-representational, context-rich |
* Research questions: |
* What are the possibilities for using analysis to assist with imagination? How do we define these terms? |
* How do the shapes of analysis, and representation, restrict what can be imagined or expressed? |
Analysis 2: The validated instruments |
* Format: narrative review |
* Subject: The integration of artificial intelligence and patient-reported outcome measures |
* Research questions: |
* What would an integration of artificial intelligence and patient-reported outcome measure (AI-PROM) do? |
* What questions are researchers pursuing this integration asking, to know they are on a good path? |
Self-Writing |
* Format: lyric essay, illness narrative |
* Thesis: Life-writing is a way to address excess signification and the limits of representational systems |
Sentences About Rivers and Cancers |
* Format: poetry that arises from research practice, generative visual artwork |
* Thesis: by assembling a shared imaginary between rivers and cancers, bodies and environment, illness and landscape, we can consider the ecological and the personal together. |
Analysis 3: From Informed Consent to Shared Decision Making |
* Format: critical analysis |
* Research question: |
* Given that one hope for the integration of AI and PROMs is that it will provide patient-centered decision support, what are the underlying assumptions about autonomy and dependence that define how decisions are made? |
* What can a comparison of informed consent and shared decision making teach us about emergent forms of relational autonomy and expertise? |
There is almost not an interval |
* Format: exhibition notes |
* Thesis: An exhibition presents a novel and expansive way of approaching the research areas covered in this dissertation. In learning from artists whose work resonates with and problematizes the dissertation’s themes, rigid investigative frameworks are exchanged for loose coherence and multiplicity, with the hope of generating new lines of inquiry. |
* This short text contains a discussion of the exhibition theme of “intervals” in the context of this dissertation. Some context and description for artworks in the program is included. |
Glossary |
* A glossary of terms is included as a guide to the specific genealogies for the terms used throughout, as many have traveled across specialized disciplines. |
Discussion: |
This dissertation identifies the role of synthesis in patient-centered care as an active integration of multiple subjective views and mixed-method analyses, instruments, documents, records, artifacts and interactions. Synthesis, as I have used the term here, is a complement to analysis, where each gives rise to the other. The synthesis of care, practiced as shared analyses and decision-making, is a co-creative act that centers consent. |
Working with the synthetic methodology of research-oriented artistic practice (what is also called research-creation), the lines of inquiry opened here are intended to sustain an expansive understanding of how health-related data becomes insight, how it is reasoned and acted on, and how participation figures in the process. Throughout this study, imagination is a fundamental undercurrent: as sociological imagination—the capacity to link individual experience with broader social processes—and as cultural imaginaries—the collective capacity to imagine, expansively and speculatively, what could be otherwise. |
Central to this study are: |
* The analytical frameworks of feminist science and technology studies, in particular those that illuminate the work of maintaining categorical boundaries, such as Karen Barad, Donna Haraway, Susan Leigh Star, and Adele E. Clarke. |
* Critical investigations of AI infrastructures, computational logics, and machinic perception, as represented in the work of Beatriz Fazi, Nora N. Khan, and Meredith Whittaker, among others: computational representation of embodied, social experience and attendant techniques for analysis and resynthesis. |
* The work of scholars and activists engaging with disability studies and crip theory, particularly the work of Mia Mingus, Alison Kafer, Robert McRuer, and Eli Clare. provide critical tools for examining the myriad ways that intersubjective experience of living with non-normative bodyminds are produced, represented and mediated—in terms of time, duration, sequence, environment, behavior, cognition, imagination, analysis, and so on. |
* Life-writing as a core qualitative methodology for practices of narrative medicine and medical anthropology. |
* The role of patient experience and patient advocacy in the development of PROMs as medical instruments. |
Patient-Reported Outcomes Measures (PROMs), as defined by the Food and Drug Administration, capture timely aspects of a patient’s experience of illness that come directly from the patient, prior to any interpretation of the patient’s response by a clinician or anyone else. |
They comprise active forms of patient-generated health data (PGHD), along with unstructured text from clinical interviews or communication between patients and providers, along with insights collected passively from fitness trackers and personal devices. |
PROMs offer an object of analysis that exemplifies broader shifts in the methods—and meaning—of health records as they are acquired, aggregated, accessed, and interpreted. |
These shifts are acutely evident in the ways patients, caregivers, clinicians, insurers and researchers come together around shared constructions of illness, disability, and quality of life. |
Decades of systematic digitization and categorization of patient data, along with the emergence of standardized application programming interfaces (API’s), have facilitated a system-wide tilt towards so-called “value-based care,” in which health insurers seek to correlate costs with holistic, “patient-centered” health outcomes. |
This correlation, along with decision support for clinicians and patients alike, is facilitated by the advent of large-scale artificial intelligence (AI) models pre-trained for adaptability to novel tasks involving language and reasoning, fine-tuned to draw insights from PROM data. |
My research, from a critical disability studies standpoint, considers the theoretical implications of using AI as a mediating layer for a patient's embodied experience, as it is captured in health records. Additionally, I place emphasis on the way PROMs are put to use in a clinical setting to assist with decision making and patient-provider communication. |
At every step, this dissertation considers PROMs to be media objects for analysis, as mechanisms for mediating experience that inevitably produce their own imprint on social and cultural life, as literacies, behavioral patterns, anticipations, desires, and biases. |
Finally, this dissertation is committed to the effective use of speculative, creative work for putting critique into practice, by finding points of convergence across research disciplines, and refining too-broad questions through attention to specific artifacts of experience. |
Analysis helps us to imagine better |
Researchers provide an adult chimpanzee with four sheets of paper. Three of the sheets have regular shapes printed on them (counterclockwise from the top right: dots forming a disc inscribed in a triangle, dots forming two concentric triangles, dots forming a small square cross). The fourth sheet, at bottom right, is blank. The chimpanzee, equipped with chalk, immediately adds markings to the paper in what appears to be a coherent response to the printed shapes (e.g. filling them in, echoing or extending their pattern). For the blank sheet, they add different kinds of controlled marks to the top edge, corners and central area of the sheet, reinforcing the rectangle of the page. |
What follows is a statement about the shapes analysis may take, and how interacting with these shapes might provide the conditions for better imagination. Along the way, it is a critique of methods that rely on stability and optimization: fixing things. Facts, and points of view, are always situated but they move and change and contradict because they are also animated. This movement and change is sometimes narrative and it is sometimes playful. |
Latent images |
I want to begin by describing the pleasure of receiving a diagnosis. |
I am standing by the entrance of a big-box home improvement store looking at large metal enclosures. Boxes. The ear, nose and throat doctor who, a few days before, had used a hollow needle to take a sample from a lump on my neck, is on the phone. He describes the particular type of cancerous tumor the cells from the sample had been classified as, and recommends that I visit an oncologist as soon as possible to discuss next steps. At this moment, I am not deteriorating, moving from healthy to sick: I am getting information about lymphoma and how to live with it. I am pausing for a moment. |
“For those with the privileges of food, care, and physical support,” Mel Y. Chen writes, this pause becomes “a meditation (if forced) on the conditions that underlie both illness and wellness, that is, the biopoliticized animacies that foretell what may become of a changing body, human or not, living or nonliving.” |
Predictions of impairment, or changed quality of life, are always already there, latent forms to be actualized as a consequence of becoming recognizable to myself and others. Who are these others? And what is becoming recognizable? |
I will start with the second question: |
I am reading Hannah Arendt looking for how she talks about life. Human life, she writes, can be considered apart from biological life in the way it unfolds through speech and action, “the two activities whose end result will always be a story with enough coherence to be told, no matter how accidental or haphazard the single events and their causation may appear to be.” Biography, as the record of vital facts, is not important. What is important is how we become recognizable to one another, through mediated processes of mutual disclosure, and the evidence these processes leave, their hopeful coherence. |
Arendt begins the “Action” section of The Human Condition with a quote, in Latin, from Dante’s Monarchia: |
“Nam in omni actione principaliter intenditur ab agente, sive necessitate nature sive voluntarie agat, propriam similitudinem explicare. Unde fit quod omne agens, in quantum huiusmodi, delectatur; quia, cum omne quod est appetat suum esse, ac in agendo agentis esse quodammodo amplietur, sequitur de necessitate delectatio [...] Nichil igitur agit nisi tale existens quale patiens fieri debet.” |
Here I give my own loose translation (with delight): |
In every action, the first intention of the person acting, whether out of necessity or by choice, is to be recognizable [to make their own image legible]. Everything that acts, as an actor, finds delight in being active; since everything that exists desires its own existence, and in acting this existence is somehow intensified, delight necessarily follows. Nothing acts unless—by acting—it makes its latent self easily understandable. |
Dante pre-echoes my point, that it is essential (delightful) to put inner experience into the world, to be exposed to diagnosis, to make oneself understood. In fact, he writes, it is the only reason for acting. |
The closing sentence of Dante’s text, resistant to a singular translation, as Arendt acknowledges in a footnote, is one piece of a larger story that explains our need for a “space of appearance,” that is, an active forum of subjective co-creation, where our ‘latent selves’ come out—not a place where objects are made, but a space for making (and maintaining) ourselves and our relations: |
“Without a space of appearance and without trusting in action and speech as a mode of being together, neither the reality of one's self, of one's own identity, nor the reality of the surrounding world can be established beyond doubt. The human sense of reality demands that [humans] actualize the sheer passive givenness of their being, not in order to change it but in order to make articulate and call into full existence what otherwise they would have to suffer passively anyhow.” |
I read Dante’s early-14th century text about sovereign identity, and Arendt’s mid-20th century continuation of the thought, in terms of very contemporary crises of representation: how do we now establish identity and reality beyond doubt? How do we make latent selves easily understandable? |
The focal point of this dissertation sits at the intersection of patient-reported outcome measures (PROMs) and artificial intelligence (AI). I have selected this intersection because it seems to indicate an inflection point in how testimony (speaking or authenticating oneself), and analysis (extracting features and patterns to produce representations) will either reinforce or cancel each other. |
While the consequences of this are far-reaching, I will concentrate here on applications to patient-centered healthcare. I will take many walks outside to clear my head and enjoy the surroundings, following a peripatetic method characterized by Rosi Braidotti as the figuration of a nomadic subject, “a politically informed account of [...] alternative subjectivity. [...] A style of thinking, occasionally autobiographical, which may at times strike the readers as [...] stream-of-consciousness.” |
Rough answers to the preceding questions (the crises of representation) may be found in patient-reported outcome measures as validated instruments, tools by which inner experience becomes empirical evidence, with minimal mediation. For AI, the questions may be considered together with the mechanics of latent space, compressed representations that provide insight into how data are related, their patterns and features, implicit or explicit. |
The explanation of life seems to be its melody |
Before the Dante quote and just after the chapter heading (“Action”) Arendt put a brief epigraph from the Danish author Isak Dinesen: |
All sorrows can be borne if you put them into a story or tell a story about them. |
In part because I want to know what “all sorrows” means exactly I trace the source of this quotation to a 1957 New York TImes interview where I find Dinesen continues: |
To me, the explanation of life seems to be its melody, its pattern. And I feel in life such an infinite, truly inconceivable fantasy. |
Melody is not a metaphor, it is a guide. “All melody is a series of attitudes,” Suzanne Langer wrote, fifteen years before Dinesen’s interview took place. |
Thoughts and emotions are not represented by melody, but are carried along, in what Langer names “a connotative relationship between music and subjective experience, a certain similarity of logical form.” One is not the model of the other, there is no possibility of direct translation. And yet the sensation is there, to be analyzed. |
End of preview. Expand
in Data Studio
README.md exists but content is empty.
- Downloads last month
- 53