Back to results

University of Cambridge

Probabilistic Programming for Postoperative Bleeding

Abstract

dc:description.abstract

This thesis investigates a novel approach to clinical decision support design: empowering clinicians to participate in creating their own decision support tools, rather than being recipients of pre-built systems. While existing research has focused on making medical AI systems interpretable to clinicians, we explore what it might take for clinicians to be authors of their own decision support tools. We focus specifically on the challenge of modelling uncertainty and study this in the context of postoperative bleeding management in a hospital intensive care unit. We explore the translation of a tool from machine learning and statistics, probabilistic programming, into the clinical context, examining the assumptions inherent in these languages and comparing them with insights elicited from clinicians in context. We explore three research questions: what can we learn about clinical decision making under uncertainty from the perspective of probabilistic programming languages (PPLs)? What, in turn, do we learn about the usability of these languages as we work to adapt them to a new set of users? And finally, what are the adaptations required to design a domain-specific PPL that clinicians can integrate into their decision making in this context? Through a mixed-methods approach combining ethnography, semi-structured interviews, graphic elicitation, and iterative prototyping, we study how clinicians conceptualize and reason about uncertainty in the high-stakes context of postoperative bleeding in the cardiothoracic ICU. Using Concept-based Analysis of Surface and Structural Misfits, we analyse the conceptual fit between clinicians' mental models of uncertainty and the ways uncertainty is represented in these computational modelling tools. This work moves beyond traditional goals of AI interpretability in clinical contexts to explore what it might take to give clinicians authorship over their decision support tools. By studying the translation between clinical and computational thinking, we contribute insights for designing systems that support clinician agency in an increasingly computationally-mediated healthcare environment.

Degree

thesis:*
Name dc:type.qualificationname
Doctor of Philosophy (PhD)
Level dc:type.qualificationlevel
Doctoral
Grantor dc:publisher.institution
University of Cambridge
Year dc:date.issued
2024

Author and committee

dc:creator, dc:contributor.*
Author dc:creator
  • Robinson, Diana
Advisors dc:contributor.advisor
  • Blackwell, Alan
  • Lawrence, Neil

Subjects

dc:subject × 3

Rights

dc:rights
Language dc:language
eng

Identifiers

dc:identifier.*
DOI dc:identifier.doi
https://doi.org/10.17863/CAM.117795
OAI identifier oai:identifier
oai:www.repository.cam.ac.uk:1810/383405

Chain of custody

source
Harvested from
Cambridge University
Base URL
api.repository.cam.ac.uk/server/oai/request
Last updated
2026-07-22
Source record
OAI-PMH GetRecord
citation

Robinson, Diana. Probabilistic Programming for Postoperative Bleeding. Doctoral thesis, University of Cambridge, 2024. https://doi.org/10.17863/CAM.117795