Henry Np Munby
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Henry Np Munby has a total of 8 co-authors. Here are the top co-authors by number of shared publications:
- A. Sarah Walker
Nuffield Department of Medicine, University of Oxford
1 shared publication - Aravind V. Ramesh
Bristol Veterinary School, University of Bristol, Bristol, UK
1 shared publication - Chang Ho Yoon
University of Oxford
1 shared publication - David William Eyre
Nuffield Department of Medicine, University of Oxford, Oxford, UK
1 shared publication - Kevin Yuan
Big Data Institute, Nuffield Department of Population Health, University of Oxford, Oxford, UK
1 shared publication - Matt Thomas
Intensive Care Unit, North Bristol NHS Trust, Bristol, UK
1 shared publication
Latest Publications
- Open Access
Leveraging transformers and large language models with antimicrobial prescribing data to predict sources of infection for electronic health record studies
Background Electronic health records frequently contain extensive unstructured free-text data, but extracting information accurately from these data and at scale is challenging. Using free-text from antibiotic prescribing data as an example, we investigate the performance of modern natural language processing methods (NLP) and large language models (LLMs) as tools for extracting features from medical records. Methods We used 938,150 hospital antibiotic prescriptions from Oxfordshire, UK. The 4000 most frequently used free-text indications justifying antibiotic use were labelled by clinical researchers into 11 categories describing the infection source/clinical syndrome being treated and used for model training. Traditional classification methods, fuzzy regex matching and n-grams with XGBoost, were compared against modern transformer models: we fine-tuned generic and domain-specific BERT models, fine-tuned GPT3.5, and investigated few-shot learning with GPT4. Models were evaluated on internal and external test datasets (2000 prescriptions each). Infection sources determined from ICD10 codes were also used for comparisons. Results In internal and external test datasets, the fine-tuned domain-specific Bio+Clinical BERT model averaged an F1 score of 0.97 and 0.98 respectively across the classes and outperformed the traditional regex (F1=0.71 and 0.74) and n-grams/XGBoost (F1=0.86 and 0.84). OpenAI’s GPT4 model achieved F1 scores of 0.71 and 0.86 without using labelled training data and a fine-tuned GPT3.5 model F1 scores of 0.95 and 0.97. Comparing infection sources extracted from ICD10 codes to those parsed from free-text indications, free-text indications revealed 31% more specific infection sources. Conclusion Modern transformer-based models can efficiently and accurately categorise semi-structured free-text in medical records, such as prescription free-text. Finetuned local transformer models outperform LLMs currently for structured tasks. Few shot LLMs match the performance of traditional NLP without the need for labelling. Transformer-based models have the potential to be used widely throughout medicine to analyse medical records more accurately, facilitating beter research and patient care.
Open AccessLeveraging transformers and large language models with antimicrobial prescribing data to predict sources of infection for electronic health record studies
Background Electronic health records frequently contain extensive unstructured free-text data, but extracting information accurately from these data and at scale is challenging. Using free-text from antibiotic prescribing data as an example, we investigate the performance of modern natural language processing methods (NLP) and large language models (LLMs) as tools for extracting features from medical records. Methods We used 938,150 hospital antibiotic prescriptions from Oxfordshire, UK. The 4000 most frequently used free-text indications justifying antibiotic use were labelled by clinical researchers into 11 categories describing the infection source/clinical syndrome being treated and used for model training. Traditional classification methods, fuzzy regex matching and n-grams with XGBoost, were compared against modern transformer models: we fine-tuned generic and domain-specific BERT models, fine-tuned GPT3.5, and investigated few-shot learning with GPT4. Models were evaluated on internal and external test datasets (2000 prescriptions each). Infection sources determined from ICD10 codes were also used for comparisons. Results In internal and external test datasets, the fine-tuned domain-specific Bio+Clinical BERT model averaged an F1 score of 0.97 and 0.98 respectively across the classes and outperformed the traditional regex (F1=0.71 and 0.74) and n-grams/XGBoost (F1=0.86 and 0.84). OpenAI’s GPT4 model achieved F1 scores of 0.71 and 0.86 without using labelled training data and a fine-tuned GPT3.5 model F1 scores of 0.95 and 0.97. Comparing infection sources extracted from ICD10 codes to those parsed from free-text indications, free-text indications revealed 31% more specific infection sources. Conclusion Modern transformer-based models can efficiently and accurately categorise semi-structured free-text in medical records, such as prescription free-text. Finetuned local transformer models outperform LLMs currently for structured tasks. Few shot LLMs match the performance of traditional NLP without the need for labelling. Transformer-based models have the potential to be used widely throughout medicine to analyse medical records more accurately, facilitating beter research and patient care.
- 28 December 2023
- Published by SAGE Publications in Journal of the Intensive Care Society
- Vol. 25 (2) , 164-170
The fragility index in randomised controlled trials of interventions for aneurysmal subarachnoid haemorrhage: A systematic review
Background: Fragility analysis supplements the p-value and risk of bias assessment in the interpretation of results of randomised controlled trials. In this systematic review we determine the fragility index (FI) and fragility quotient (FQ) of randomised trials in aneurysmal subarachnoid haemorrhage. Methods: This is a systematic review registered with PROSPERO (ID: CRD42020173604). Randomised controlled trials in adults with aneurysmal subarachnoid haemorrhage were analysed if they reported a statistically significant primary outcome of mortality, function (e.g. modified Rankin Scale), vasospasm or delayed neurological deterioration. Results: We identified 4825 records with 18 randomised trials selected for analysis. The median fragility index was 2.5 (inter-quartile range 0.25–5) and the median fragility quotient was 0.015 (IQR 0.02–0.039). Five of 20 trial outcomes (25%) had a fragility index of 0. In seven trials (39.0%), the number of participants lost to follow-up was greater than or equal to the fragility index. Only 16.7% of trials are at low risk of bias. Conclusion: Randomised controlled trial evidence supporting management of aneurysmal subarachnoid haemorrhage is weaker than indicated by conventional analysis using p-values alone. Increased use of fragility analysis by clinicians and researchers could improve the translation of evidence to practice.
- 28 December 2023
- Published by SAGE Publications in Journal of the Intensive Care Society
- Vol. 25 (2) , 164-170
The fragility index in randomised controlled trials of interventions for aneurysmal subarachnoid haemorrhage: A systematic review
Background: Fragility analysis supplements the p-value and risk of bias assessment in the interpretation of results of randomised controlled trials. In this systematic review we determine the fragility index (FI) and fragility quotient (FQ) of randomised trials in aneurysmal subarachnoid haemorrhage. Methods: This is a systematic review registered with PROSPERO (ID: CRD42020173604). Randomised controlled trials in adults with aneurysmal subarachnoid haemorrhage were analysed if they reported a statistically significant primary outcome of mortality, function (e.g. modified Rankin Scale), vasospasm or delayed neurological deterioration. Results: We identified 4825 records with 18 randomised trials selected for analysis. The median fragility index was 2.5 (inter-quartile range 0.25–5) and the median fragility quotient was 0.015 (IQR 0.02–0.039). Five of 20 trial outcomes (25%) had a fragility index of 0. In seven trials (39.0%), the number of participants lost to follow-up was greater than or equal to the fragility index. Only 16.7% of trials are at low risk of bias. Conclusion: Randomised controlled trial evidence supporting management of aneurysmal subarachnoid haemorrhage is weaker than indicated by conventional analysis using p-values alone. Increased use of fragility analysis by clinicians and researchers could improve the translation of evidence to practice.