Informing antimicrobial stewardship with explainable AI

Massimo Cavallaro; Ed Moran; Benjamin Collyer; Noel D. McCarthy; Christopher Green; Matt J. Keeling

doi:10.1371/journal.pdig.0000162

Informing antimicrobial stewardship with explainable AI

Massimo Cavallaro, Ed Moran, Benjamin Collyer, Noel D. McCarthy, Christopher Green, Matt J. Keeling

Microbiology and Infection

Research output: Contribution to journal › Article › peer-review

Abstract

The accuracy and flexibility of artificial intelligence (AI) systems often comes at the cost of a decreased ability to offer an intuitive explanation of their predictions. This hinders trust and discourage adoption of AI in healthcare, exacerbated by concerns over liabilities and risks to patients’ health in case of misdiagnosis. Providing an explanation for a model’s prediction is possible due to recent advances in the field of interpretable machine learning. We considered a data set of hospital admissions linked to records of antibiotic prescriptions and susceptibilities of bacterial isolates. An appropriately trained gradient boosted decision tree algorithm, supplemented by a Shapley explanation model, predicts the likely antimicrobial drug resistance, with the odds of resistance informed by characteristics of the patient, admission data, and historical drug treatments and culture test results. Applying this AI-based system, we found that it substantially reduces the risk of mismatched treatment compared with the observed prescriptions. The Shapley values provide an intuitive association between observations/data and outcomes; the associations identified are broadly consistent with expectations based on prior knowledge from health specialists. The results, and the ability to attribute confidence and explanations, support the wider adoption of AI in healthcare.

Original language	English
Article number	e0000162
Journal	PLOS digital health
Volume	2
Issue number	1
DOIs	https://doi.org/10.1371/journal.pdig.0000162
Publication status	Published - 5 Jan 2023

Bibliographical note

Funding: This work was supported by Health Data Research UK, which is funded by the UK Medical Research Council, EPSRC, Economic and Social Research Council, Department of Health and Social Care (England), Chief Scientist Office of the Scottish Government Health and Social Care Directorates, Health and Social Care Research and Development Division (Welsh Government), Public Health Agency (Northern Ireland), British Heart Foundation and the Wellcome Trust supported MC, MJK, and NDM. MJK and NDM are affiliated to the National Institute for Health Research Health Protection Research Units (NIHR HPRUs) in Gastrointestinal Infections and in Genomics and Enabling Data. MJK is funded by UK Research and Innovation through the JUNIPER modelling consortium (MR/V038613/1). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.

Access to Document

10.1371/journal.pdig.0000162Licence: Creative Commons: Attribution (CC BY)

CavallaroM2023InformingFinal published version, 2.03 MBLicence: Creative Commons: Attribution (CC BY)

Cite this

@article{34454aa453d64dd5a802747e9fd008ca,

title = "Informing antimicrobial stewardship with explainable AI",

abstract = "The accuracy and flexibility of artificial intelligence (AI) systems often comes at the cost of a decreased ability to offer an intuitive explanation of their predictions. This hinders trust and discourage adoption of AI in healthcare, exacerbated by concerns over liabilities and risks to patients{\textquoteright} health in case of misdiagnosis. Providing an explanation for a model{\textquoteright}s prediction is possible due to recent advances in the field of interpretable machine learning. We considered a data set of hospital admissions linked to records of antibiotic prescriptions and susceptibilities of bacterial isolates. An appropriately trained gradient boosted decision tree algorithm, supplemented by a Shapley explanation model, predicts the likely antimicrobial drug resistance, with the odds of resistance informed by characteristics of the patient, admission data, and historical drug treatments and culture test results. Applying this AI-based system, we found that it substantially reduces the risk of mismatched treatment compared with the observed prescriptions. The Shapley values provide an intuitive association between observations/data and outcomes; the associations identified are broadly consistent with expectations based on prior knowledge from health specialists. The results, and the ability to attribute confidence and explanations, support the wider adoption of AI in healthcare.",

author = "Massimo Cavallaro and Ed Moran and Benjamin Collyer and McCarthy, {Noel D.} and Christopher Green and Keeling, {Matt J.}",

note = "Funding: This work was supported by Health Data Research UK, which is funded by the UK Medical Research Council, EPSRC, Economic and Social Research Council, Department of Health and Social Care (England), Chief Scientist Office of the Scottish Government Health and Social Care Directorates, Health and Social Care Research and Development Division (Welsh Government), Public Health Agency (Northern Ireland), British Heart Foundation and the Wellcome Trust supported MC, MJK, and NDM. MJK and NDM are affiliated to the National Institute for Health Research Health Protection Research Units (NIHR HPRUs) in Gastrointestinal Infections and in Genomics and Enabling Data. MJK is funded by UK Research and Innovation through the JUNIPER modelling consortium (MR/V038613/1). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.",

year = "2023",

month = jan,

day = "5",

doi = "10.1371/journal.pdig.0000162",

language = "English",

volume = "2",

journal = "PLOS digital health",

issn = "2767-3170",

publisher = "Public Library of Science (PLOS)",

number = "1",

}

TY - JOUR

T1 - Informing antimicrobial stewardship with explainable AI

AU - Cavallaro, Massimo

AU - Moran, Ed

AU - Collyer, Benjamin

AU - McCarthy, Noel D.

AU - Green, Christopher

AU - Keeling, Matt J.

N1 - Funding: This work was supported by Health Data Research UK, which is funded by the UK Medical Research Council, EPSRC, Economic and Social Research Council, Department of Health and Social Care (England), Chief Scientist Office of the Scottish Government Health and Social Care Directorates, Health and Social Care Research and Development Division (Welsh Government), Public Health Agency (Northern Ireland), British Heart Foundation and the Wellcome Trust supported MC, MJK, and NDM. MJK and NDM are affiliated to the National Institute for Health Research Health Protection Research Units (NIHR HPRUs) in Gastrointestinal Infections and in Genomics and Enabling Data. MJK is funded by UK Research and Innovation through the JUNIPER modelling consortium (MR/V038613/1). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.

PY - 2023/1/5

Y1 - 2023/1/5

N2 - The accuracy and flexibility of artificial intelligence (AI) systems often comes at the cost of a decreased ability to offer an intuitive explanation of their predictions. This hinders trust and discourage adoption of AI in healthcare, exacerbated by concerns over liabilities and risks to patients’ health in case of misdiagnosis. Providing an explanation for a model’s prediction is possible due to recent advances in the field of interpretable machine learning. We considered a data set of hospital admissions linked to records of antibiotic prescriptions and susceptibilities of bacterial isolates. An appropriately trained gradient boosted decision tree algorithm, supplemented by a Shapley explanation model, predicts the likely antimicrobial drug resistance, with the odds of resistance informed by characteristics of the patient, admission data, and historical drug treatments and culture test results. Applying this AI-based system, we found that it substantially reduces the risk of mismatched treatment compared with the observed prescriptions. The Shapley values provide an intuitive association between observations/data and outcomes; the associations identified are broadly consistent with expectations based on prior knowledge from health specialists. The results, and the ability to attribute confidence and explanations, support the wider adoption of AI in healthcare.

AB - The accuracy and flexibility of artificial intelligence (AI) systems often comes at the cost of a decreased ability to offer an intuitive explanation of their predictions. This hinders trust and discourage adoption of AI in healthcare, exacerbated by concerns over liabilities and risks to patients’ health in case of misdiagnosis. Providing an explanation for a model’s prediction is possible due to recent advances in the field of interpretable machine learning. We considered a data set of hospital admissions linked to records of antibiotic prescriptions and susceptibilities of bacterial isolates. An appropriately trained gradient boosted decision tree algorithm, supplemented by a Shapley explanation model, predicts the likely antimicrobial drug resistance, with the odds of resistance informed by characteristics of the patient, admission data, and historical drug treatments and culture test results. Applying this AI-based system, we found that it substantially reduces the risk of mismatched treatment compared with the observed prescriptions. The Shapley values provide an intuitive association between observations/data and outcomes; the associations identified are broadly consistent with expectations based on prior knowledge from health specialists. The results, and the ability to attribute confidence and explanations, support the wider adoption of AI in healthcare.

U2 - 10.1371/journal.pdig.0000162

DO - 10.1371/journal.pdig.0000162

M3 - Article

SN - 2767-3170

VL - 2

JO - PLOS digital health

JF - PLOS digital health

IS - 1

M1 - e0000162

ER -

Informing antimicrobial stewardship with explainable AI

Abstract

Bibliographical note

Access to Document

Fingerprint

Cite this