Predictive care: a protocol for a computational ethnographic approach to building fair models of inpatient violence in emergency psychiatry
Bibliographic record
Abstract
INTRODUCTION: Managing violence or aggression is an ongoing challenge in emergency psychiatry. Many patients identified as being at risk do not go on to become violent or aggressive. Efforts to automate the assessment of risk involve training machine learning (ML) models on data from electronic health records (EHRs) to predict these behaviours. However, no studies to date have examined which patient groups may be over-represented in false positive predictions, despite evidence of social and clinical biases that may lead to higher perceptions of risk in patients defined by intersecting features (eg, race, gender). Because risk assessment can impact psychiatric care (eg, via coercive measures, such as restraints), it is unclear which patients might be underserved or harmed by the application of ML. METHODS AND ANALYSIS: We pilot a computational ethnography to study how the integration of ML into risk assessment might impact acute psychiatric care, with a focus on how EHR data is compiled and used to predict a risk of violence or aggression. Our objectives include: (1) evaluating an ML model trained on psychiatric EHRs to predict violent or aggressive incidents for intersectional bias; and (2) completing participant observation and qualitative interviews in an emergency psychiatric setting to explore how social, clinical and structural biases are encoded in the training data. Our overall aim is to study the impact of ML applications in acute psychiatry on marginalised and underserved patient groups. ETHICS AND DISSEMINATION: The project was approved by the research ethics board at The Centre for Addiction and Mental Health (053/2021). Study findings will be presented in peer-reviewed journals, conferences and shared with service users and providers.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame distilled prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. Learned from the 10,348 direct Codex labels and 10,348 direct Gemma labels. Candidate is the union of thresholded teacher heads; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels or direct frontier model labels.
Codex and Gemma teacher scores by category
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.001 | 0.000 |
| Meta-epidemiology (narrow) | 0.000 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.000 |
| Bibliometrics | 0.000 | 0.001 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.000 | 0.000 |
| Open science | 0.001 | 0.000 |
| Research integrity | 0.000 | 0.000 |
| Insufficient payload (model declined to judge) | 0.000 | 0.000 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one teacher head, not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".