DTDaniel Cameron Thompson

Daniel Cameron Thompson

*This is an algorithmically generated author record based on publicly available information.

Current Organizations / Institutes

The organizations/institutes from the scholar's latest article


Daniel Cameron Thompson has a total of 34 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Open Access

    Natural Language Processing framework for identifying abdominal aortic aneurysm repairs using unstructured electronic health records

    by Daniel C. Thompson, Reza Mofidi

    Patient identification for national registries often relies upon clinician recognition of cases or retrospective searches using potentially inaccurate clinical codes, leading to incomplete data capture and inefficiencies. Natural Language Processing (NLP) offers a promising solution by automating analysis of electronic health records (EHRs). This study aimed to develop NLP models for identifying and classifying abdominal aortic aneurysm (AAA) repairs from unstructured EHRs, demonstrating a proof-of-concept for automated patient identification in registries like the National Vascular Registry. Using the MIMIC-IV-Note dataset, a multi-tiered approach was developed to identify vascular patients (Task 1), AAA repairs (Task 2), and classify repairs as primary or revision (Task 3). Four NLP models were trained and evaluated using 4870 annotated records: scispaCy, BERT-base, Bio-clinicalBERT, and a scispaCy/Bio-clinicalBERT ensemble. Models were compared using accuracy, precision, recall, F1-score, and area under the receiver operating characteristic curve (AUC). The scispaCy model demonstrated the fastest training (2 min/epoch) and inference times (2.87 samples/sec). For Task 1, scispaCy and ensemble models achieved the highest accuracy (0.97). In Task 2, all models performed exceptionally well, with ensemble, scispaCy, and Bio-clinicalBERT models achieving 0.99 accuracy and 1.00 AUC. For Task 3, Bio-clinicalBERT and the ensemble model achieved an AUC of 1.00, with Bio-clinicalBERT displaying the best overall accuracy (0.98). This study demonstrates that NLP models can accurately identify and classify AAA repair cases from unstructured EHRs, suggesting significant potential for automating patient identification in vascular surgery and other medical registries, reducing administra


  • Open Access

    Development and comparison of natural language processing models for abdominal aortic aneurysm repair identification and classification using unstructured electronic health records

    by Daniel Thompson, Reza Mofidi

    Background Patient identification for national registries often relies upon clinician recognition of cases or retrospective searches using potentially inaccurate clinical codes, potentially leading to incomplete data capture and inefficiencies. Natural Language Processing (NLP) offers a promising solution by automating analysis of electronic health records (EHRs). This study aimed to develop NLP models for identifying and classifying abdominal aortic aneurysm (AAA) repairs from unstructured EHRs, demonstrating proof-of-concept for automated patient identification in registries like the National Vascular Registry. Method Using the MIMIC-IV-Note dataset, a multi-tiered approach was developed to identify vascular patients (Task 1), AAA repairs (Task 2), and classify repairs as primary or revision (Task 3). Four NLP models were trained and evaluated using 4,870 annotated records: scispaCy, BERT-base, Bio-clinicalBERT, and a scispaCy/Bio-clinicalBERT ensemble. Models were compared using accuracy, precision, recall, F1-score, and area under the receiver operating characteristic curve. Results The scispaCy model demonstrated the fastest training (2 mins/epoch) and inference times (2.87 samples/sec). For Task 1, scispaCy and ensemble models achieved the highest accuracy (0.97). In Task 2, all models performed exceptionally well, with ensemble, scispaCy, and Bio-clinicalBERT models achieving 0.99 accuracy and 1.00 AUC. For Task 3, Bio-clinicalBERT and the ensemble model achieved an AUC of 1.00, with Bio-clinicalBERT displaying the best overall accuracy (0.98). Conclusion This study demonstrates that NLP models can accurately identify and classify AAA repair cases from unstructured EHRs, suggesting significant potential for automating patient identification in vascular surgery and other medical registries, reducing administrative burden and improving data capture for audit and research.

  • Prescribing in the prehospital environment: a review of the pharmaceutical Module 501 on UK Military Exercise SAIF SAREEA 3. Can such analysis assist with the scaling of healthcare assets?

    by L G Davies, D C Thompson, R Gillett, M B Smith

    Introduction Module 501 provides core medications which are fundamental to the capability of a prehospital treatment team (PHTT). The quantities of each medication in the module inventory undergo regular review, but these do not correspond to a population at risk (PAR) figure or deployment length for which they intend to be used. This article proposes how the quantities of Module 501 drugs can be scaled for a given deployment, in this example using statistics taken from static PHTTs on Exercise Saif Sareea 3 (SS3). Methods The statistics were gathered using a custom-built search of electronic records from the Deployed Defence Medical Information Capability Programme in addition to written record-keeping, which were aligned to the weekly PAR at each PHTT location throughout their full operational capability periods. A quotient was then derived for each module item using a formula. Results Among the 10 most commonly prescribed drugs were four analgesics and three antimicrobials. 42 of the 110 studied drugs were not prescribed during SS3. Discussion The data from SS3 reflect the typical scope of disease encountered in the deployed land setting. Employing these data, the use of a formula to estimate the drug quantities needed to sustain a Strike Armoured Infantry Brigade over a 28-day period is demonstrated. Recommendations Further study of Module 501 across varied deployment environments would be valuable in evolving this approach to medicinal scaling if proven effective for the warm desert climate. It could then be applied to other modules to further inform future Strike medical planning. Limitations Several considerations when drawing deductions from the data are mentioned, including the inaccuracy of predictor variables taken from the EpiNATO-2 reports. Conclusion The proposed formula provides an evidence-based framework for scaling drug quantities for a deployment planning. This may improve patient safety and confer logistical, storage and fiscal benefits.

  • Medical equipment trends during the UK military exercise SAIF SAREEA 3 in Oman

    by Daniel C Thompson, L G Davies, M B Smith

    Introduction Ex SAIF SAREEA 3 was a joint British-Omani military exercise involving approximately 5500 British personnel deployed to Oman over a 6-month period. Role 1 medical care was provided by medical treatment facilities (MTFs) deployed with medical equipment as per the UK 300 medical module. Method Retrospective analysis was undertaken of prospectively collected equipment usage data from two Role 1 MTFs in Duqm (MTF 1) and Muaskar Al Murtafa (MTF 2) camps over a period of 6–8 weeks. Data were analysed alongside routinely collected epidemiological data (EPINATO) during the deployment. Equipment used in addition to the module was also recorded. Results MTF 1 used 50 out of the 179 different items from the module over the 8-week period. MTF 2 used 45 out of the 179 different items from their module over the 6-week period. The most commonly used items across the sites were non-sterile examination gloves, plastic aprons, tympanic thermometer probe covers, disinfectant wipes and self-adhesive plasters. Extramodular items (blunt fill needle, water pump sprayer, Jelonet gauze and stool specimen pot) accounted for 5% of all equipment used in MTF 1. Conclusion The study showed that the 300 module accommodates 95% of Role 1 patients’ needs but highlights the requirement for dedicated equipment for the treatment of heat casualties if deemed likely and blunt fill/filter needles for the administration of parenteral medication. Commanders must perform a thorough medical estimate and risk assessment prior to deployment to ensure that the 300 medical module is likely to provide the necessary equipment and supplement the module if required.


  • Encephalitis on deployment in Kenya: think beyond the infections

    by Daniel Cameron Thompson, M S Bailey, D Bowley, S Jacob

    A 34-year-old female soldier presented with fever and behavioural changes while deployed in Kenya and was diagnosed with encephalitis. The patient underwent urgent aeromedical evacuation to the Queen Elizabeth Hospital, Birmingham for further management. Microbiology tests excluded common infectious causes that are endemic in the East Africa region. However, an autoantibody screen was positive for antibodies against the N-methyl-D-aspartate receptor (NMDAR). Full body imaging confirmed the presence of limbic encephalitis and an ovarian mass suggestive of a teratoma. The patient was diagnosed with ovarian teratoma-associated anti-NMDAR encephalitis, a potentially fatal disease. The patient underwent surgery to remove the teratoma and commenced immunotherapy with steroids, plasma exchange and rituximab. This case highlights the diagnostic challenges of fever with behavioural changes in military personnel deployed in a tropical environment.


  • The pattern of paediatric blast injury in Afghanistan

    by Daniel C Thompson, R J Crooks, J C Clasper, A Lupu, S A Stapley, D J Cloke

    Introduction Between 2009 and 2015, 3746 children died, and 7904 were injured as a result of armed conflict within Afghanistan. Improvised explosive devices (IEDs) and explosive remnants of war accounted for 29% of child casualties in 2015. The aim of this study was to review the burden of paediatric blast injuries admitted to Camp Bastion, Afghanistan, and to investigate the hypothesis that children suffer proportionally more head injuries than adults. Method A retrospective analysis was undertaken of prospectively collected data derived from the UK Joint Theatre Trauma Registry of ambulant paediatric (aged 2–15 years) admissions with blast injuries at the Role 3 Field Hospital, Camp Bastion from June 2006 to March 2013. The data set included demographic information, injury profile and severity (New Injury Severity Score) and operative findings. The pattern of injuries were investigated by looking at trends in the number and severity of injuries sustained by each body region. Results During this period, 295 admissions were identified, 76% of whom were male, with an overall mortality rate of 18.5%. The most common blast mechanism was an IED (68%) causing 80% of fatalities. The lower extremities were the most commonly injured body region, accounting for 31% of total injuries and occurring in 62% of cases. 24.3% of children between 2 and 7 years suffered severe head or neck injuries compared with 19.8% of children aged between 8 and 15 years. 34% of head injuries were rated unsurvivable and accounted for 88% of fatalities. 77% of cases required an operation with a mean operating time of 125 min. The most common first operations were debridement of soft tissues (50%), laparotomy (16%) and lower limb amputation (11%). Conclusion Although paediatric blast casualties represented a small percentage of the overall workload at Camp Bastion Role 3 Medical Facility, the pattern of injuries seen suggests that children are more likely to sustain severe head, face and neck injuries than adults.


  • The effect of operating lights on laminar flow

    by R. Refaie, P. Rushton, P. McGovern, D. Thompson, I. Serrano-Pedraza, K. S. Rankin, M. Reed

    Aims: The interaction between surgical lighting and laminar airflow is poorly understood. We undertook an experiment to identify any effect contemporary surgical lights have on laminar flow and recommend practical strategies to limit any negative effects. Materials and Methods: Neutrally buoyant bubbles were introduced into the surgical field of a simulated setup for a routine total knee arthroplasty in a laminar flow theatre. Patterns of airflow were observed and the number of bubbles remaining above the surgical field over time identified. Five different lighting configurations were assessed. Data were analysed using simple linear regression after logarithmic transformation. Results: In the absence of surgical lights, laminar airflow was observed, bubbles were cleared rapidly and did not accumulate. If lights were placed above the surgical field laminar airflow was abolished and bubbles rose from the surgical field to the lights then circulated back to the surgical field. The value of the decay parameter (slope) of the two setups differed significantly; no light (b = -1.589) versus one light (b = -0.1273, p < 0.001). Two lights touching (b = -0.1191) above the surgical field had a similar effect to that of a single light (p = 0. 2719). Two lights positioned by arms outstretched had a similar effect (b = -0.1204) to two lights touching (p = 0.998) and one light (p = 0.444). When lights were separated widely (160 cm), laminar airflow was observed but the rate of clearance of the bubbles remained slower (b = -1.1165) than with no lights present (p = 0.004). Conclusion: Surgical lights have a significantly negative effect on laminar airflow. Lights should be positioned as far away as practicable from the surgical field to limit this effect. Cite this article: Bone Joint J 2017;99-B:1061–6.


  • Open Access

    7 Implantable cardiac device generator change pathways: an opportunity for optimising device prescription

    by Helena Bolam, Daniel Thompson, James Waller, Neil Andrews, Senthil Kirubakaran, Paul Kalra, Geraint Morton

    Introduction: Implantation of cardiac devices including permanent pacemakers (PPM), cardiac resynchronisation therapy with defibrillators (CRT-D) or without (CRT-P) is increasingly common. Follow up, including arranging generator change, is frequently carried out by cardiac physiologists, with minimal physician input. Many patients clinical status will change during the lifetime of the device. Generator change is an opportunity to review the original device prescription. We report the impact of introducing routine screening prior to generator change into an established comprehensive device programme. Methods: Two pathways were implemented. Patients with a PPM due generator change were screened by cardiac physiologists using a proforma assessing symptoms, QRS duration and right ventricular (RV) pacing burden. Those with heart failure symptoms, in particular breathlessness, and a prolonged native QRS or >40% RV pacing had an echocardiogram and multidisciplinary team (MDT) discussion within 2 weeks. If device upgrade was potentially indicated the patient was reviewed in clinic prior to generator change. The goal was to discuss all patients with an ICD/CRT-D due generator change at MDT. Those with changes in clinical status that may affect the appropriateness of ongoing defibrillator therapy were reviewed in clinic. Results: The first 11 months, October 2015 to August 2016, following implementation of these pathways were studied. 189 consecutive patients (145 PPM, 20 CRT-P and 24 CRT-D/ICD), mean age of 77, were included. The pathways resulted in a change in the type of device therapy in 8 patients; 6/145 (4%) of those due PPM and 2/24 (8%) due CRT-D/ICD generator change (table 1). 145 (100%) of PPM patients were proforma screened. 139/145 PPM patients (96%) went on to have PPM generator change. 17 (12%), identified as potential candidates for upgrade to CRT, had echocardiography. 5 (3%) of these had severe LV impairment and underwent physician and MDT review. 4 patients (3%) were ultimately upgraded to CRT devices. CRT was not appropriate in 1 patient. In 2 (1%) patients ongoing device therapy of any kind was felt to be inappropriate (1 potential upgrade and 1 PPM box change) and no procedure was undertaken. 10/24 (42%) patients due a CRT-D/ICD generator change were reviewed at MDT. Of these 2 (20% of those reviewed) went on to have their devices downgraded from CRT-D to CRT-P after informed discussion. Conclusions: Physiologist-led device follow up can be effectively extended to include robust screening of patients to ensure they get the most appropriate device at the time of generator change. Excellent rates of PPM screening were achieved whereas ICD screening needs to be improved. Relatively few patients had their destination device changed as a result of these pathways. However there is significant morbidity associated with receiving inappropriate device therapy, justifying implementation of a simple screening process to improve care. Abstract 7 Table 1 Outcomes for patients due implantable cardiac device generator change after the implementation of routine screening Outcomes Number of patients Type of device unchanged 181 Upgrade from PPM to CRT-P 2 Upgrade from PPM to CRT-D 2 Downgrade from CRT-D to CRT-P 2 Device therapy withdrawn 2