Thomas Pausch
Has grant
Senior Physician and Head of Digital Transformation and Data Science in Surgery
Research Interests
Make your next move.
Prepare your application
Review drafts before you send them.
Explore related searches
Recent Grants
Grant: Close
Development of an Intraoperative Indicator of Pancreatic Leakage to Reduce Postoperative Pancreatic Fistula
Open Date: 2016-08-01
Close Date: 2018-08-01
Articles (19)
Year:
2026
Background Patient-reported outcome measures (PROMs) such as health-related quality of life (HRQoL) are usually assessed at greater time intervals such as diagnostic time points, after treatment, and during follow-up. Many PROMs require frequent data collection (weekly or daily). Electronic PROMs enable high-resolution tracking but face declining response rates. Integrating PROMs into electronic health records (EHRs) could improve response rates and personalize therapy. Objective This study aimed to evaluate the technical and clinical feasibility of high-frequency HRQoL assessments for routine care in EHRs. Methods Patients receive emails on their mobile devices with 1-time links to a web-based app called MyEDC. This app communicates with an electronic data capture proxy in the demilitarized zone of the hospital. With a polling mechanism, these patient data are transferred to the protected hospital network and uploaded to the EHR system. HRQoL on a visual analog scale is assessed over the course of treatment in 4 clinical use cases: psychosomatics, hematology, visceral surgery, and neurosurgery. Results Quality of life (QoL) trajectories were collected for 110 patients with daily or weekly data collection between 2 weeks and 3 months. The HRQoL analyses revealed clinically relevant findings across the 4 different medical domains. In the use case psychosomatics, 36 patients showed a significant increase in HRQoL following 4 weeks of therapy, rising from a median of 42% (IQR 32%-52%) to 60% (IQR 41%-67%; P=.01). An analysis of 25 patients in hematology demonstrated a significant correlation between HRQoL and 30-item QoL Questionnaire (EORTC QLQ-C30) global health status score (P=.02). For 26 patients in visceral surgery, a significant association was observed between HRQoL and the reported pain level (P<.001). The clinical feasibility was further highlighted in the neurosurgery use case, where 23 patients showed a median response time to the electronic PROM questionnaires of 5.3 (IQR 0.6-17.7) hours. HRQoL values were associated with disease-specific symptoms and scores, indicating clinical validity of this readout. Considerable variability of HRQoL was observed over time, both intraindividually and interindividually. Median area under the curve of HRQoL ranged from 0.46 to 0.79. Median time to answer ranged from 0.9 to 7.1 hours. No significant association between number of responses and age was observed. Conclusions High-resolution QoL trajectories with EHR integration are technically and clinically feasible. They offer a novel readout beyond survival analysis or PROM end point, enabling precise disease characterization and treatment comparison.
Year:
2026
Background Advancements in clinical medicine depend on state-of-the-art healthcare analytics, which require patient data from electronic health record (EHR) systems and electronic data capture (EDC). We developed and implemented an independent, yet integrated EDC system (designated IMI-EDC) in a tertiary hospital. This system ensures that clinical data are available for patient care and research while complying with regulatory standards. In this study, we compared the first cohort of IMI-EDC users against those using the predecessor paper-mediated system, measuring patient comfort and satisfaction, data quality, and facilitation of workflow facilitation. Methods Until July 2025, more than 5000 patients were able to submit their medical history and health-related quality of life questionnaires via the IMI-EDC. All patients using tablets were surveyed regarding their user experience. Additionally, a subset of 100 EDC submissions was compared to 100 submissions via the paper-based system with respect to data completeness, correctness, clarity, and readability. Finally, the time to transfer data to the Heidelberg Pancreas Registry (HPR) research database was measured. Results IMI-EDC was successfully implemented at our center. Approximately three-quarters (75%) of patients using tablets were comfortable with the system. Tablet users were significantly more likely to submit completed documents than those using paper (98/100 vs. 66/100), and among submitted forms, fully completed questionnaires were more frequent in the tablet group (70/98 vs. 24/66). Data transfers to the HPR were quicker for tablet submissions than for paper submissions in a small pragmatic subsample (4.2 minutes vs. 9.9 minutes; 7 tablet vs. 7 paper cases). Conclusions IMI-EDC is satisfactory for patients and staff in terms of user experience, data completeness, and workload reduction in the setting of a high-volume surgical center.
Year:
2026
Background Recently, surgical data recorders that are comparable to flight data recorders, also known as black boxes in the aviation industry, have been developed to improve patient safety and performance in surgery. These devices allow for unique insights in the operating room by providing new data capture capabilities. No systematic review has been carried out to evaluate the areas of application of surgical data recorders to date. Objective This systematic review and single-arm meta-analysis aims to assess the aspects of the operating theater environment for which surgical data recorders are used and to make a preliminary assessment of the quantifiable data that can be collected, compared to traditional collection methods. Methods This systematic review followed the PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) guidelines. Medline, Embase, and Web of Science databases were lastly systematically searched for papers that focused on a clinical use case for surgical data recorders on February 10, 2025. In particular, not relevant papers focusing on implementation of surgical data recorders were excluded. Title, abstract, and full-text screening were completed to identify relevant articles. The included studies were analyzed descriptively using data extraction forms. Where possible, quantifiable data was also analyzed. Risk of bias was assessed using the Risk Of Bias In Non-Randomized Studies of Exposure (ROBINS-E) tool. Results In total, 70 studies were screened, and a total of 17 studies were included. A total of 10 of the 17 studies had a low overall risk of bias; however, confounding, selection bias, small sample sizes, short study periods, and potential Hawthorne effects were the notable limitations. Only 2 studies were assessed to have publication bias. Use cases could be grouped into 4 categories: economic, safety, behavior in the operating room, and technical skill assessment. A single-arm meta-analysis focusing on adverse events and distractions in the operating theater could be conducted, demonstrating accurate reporting of distractions in line with the existing literature. Conclusions Surgical data recorders provide an unobstructed view of various aspects of the operating theatre. Most published papers present preliminary studies on surgical data recorders, indicating the potential for further, larger-scale studies with enhanced methodological quality.
Year:
2025
The aim of this systematic review was to summarize and analyze the available literature on the application of artificial intelligence systems in esophageal surgery, focusing on anatomy recognition, instrument detection, and surgical phase recognition. Esophageal cancer poses a significant global health challenge, ranking as the seventh most common cancer worldwide. Esophagectomy is the only curative treatment for non-metastatic esophageal cancer. While the introduction of minimally invasive esophagectomy and later robot-assisted minimally invasive esophagectomy significantly improved surgical precision and patient outcome, this development promoted a transition to increasing digitalization and video processing. Subsequently facilitating the integration of artificial intelligence is a promising tool in the enhancement of esophageal surgery. A systematic search was conducted following the PRISMA guidelines in the Medline and Web of Science databases. Studies published between January 2019 and June 2025 published in English and without restrictions to study type were included. Inclusion criteria focused on artificial intelligence-based anatomy recognition, instrument recognition, and phase recognition in esophageal surgery. Studies addressing preoperative and postoperative risk prediction or artificial intelligence applications not directly related to the surgical procedure were excluded. The systematic literature search yielded 7063 results. After screening, we identified six studies examining artificial intelligence applications in esophagectomy focusing on anatomy, instrument, and phase recognition. Artificial intelligence can be a useful tool—especially for intraoperative anatomy recognition—reaching detection rates comparable to trained surgeons in real time as seen in one study, reaching a Dice coefficient of 0.58, which was close to that of an expert esophageal surgeon (0.62) and significantly higher than the general surgeon (0.47, p = 0.0019). Due to the heterogeneity of study aims, utilized algorithms and outcome measures direct comparison between studies was not feasible. Artificial intelligence has demonstrated significant potential in enhancing esophageal surgery by improving anatomical recognition and optimizing surgical workflow. Despite these advancements, challenges remain in standardizing datasets, refinement of annotation methodologies, and seamless integration into real-time surgical navigation systems. To ensure clinical applicability, future research should focus on large-scale validation and prospective clinical trials to establish artificial intelligence’s clinical utility and safety in minimally invasive esophagectomy.
Year:
2025
Year:
2025
This study explores the use of open-source large language models (LLMs) to automate generation of German discharge summaries from structured clinical data. The structured data used to produce AI-generated summaries were manually extracted from electronic health records (EHRs) by a trained medical professional. By leveraging structured documentation collected for research and quality management, the goal is to assist physicians with editable draft summaries. After de-identifying 25 patient datasets, we optimized the output of the LLaMA3 model through prompt engineering and evaluated it using error analysis, as well as quantitative and qualitative metrics. The LLM-generated summaries were rated by physicians on comprehensiveness, conciseness, correctness, and fluency. Key results include an error rate of 2.84 mistakes per summary, and low-to-moderate alignment between generated and physician-written summaries (ROUGE-1: 0.25, BERTScore: 0.64). Medical professionals rated the summaries 3.72 ± 0.89 for comprehensiveness and 3.88 ± 0.97 for factual correctness on a 5-point Likert-scale; however, only 60% rated the comprehensiveness as good (4 or 5 out of 5). Despite overall informativeness, essential details—such as patient history, lifestyle factors, and intraoperative findings—were frequently omitted, reflecting gaps in summary completeness. While the LLaMA3 model captured much of the clinical information, complex cases and temporal reasoning presented challenges, leading to factual inaccuracies, such as incorrect age calculations. Limitations include a small dataset size, missing structured data elements, and the model’s limited proficiency with German medical terminology, highlighting the need for large, more complete datasets and potential model fine-tuning. In conclusion, this work provides a set of real-world methods, findings, experiences, insights, and descriptive results for a focused use case that may be useful to guide future work in the LLM generation of discharge summaries, perhaps especially for those working with German and possibly other non-English content.
Year:
2025
Background: The invisible fluid leaking from a partially resected pancreas is associated with complications including postoperative pancreatic fistula (POPF), calling for strategies to visualize intraoperative leakage. This single-arm, monocentric trial aims to evaluate the usefulness and safety of SmartPAN, a hydrogel that reacts to alkali pancreatic fluids by changing color and thus enables the surgeon to take immediate action to close leakage. Methods: Patients awaiting partial pancreatic resection for any indication were recruited to receive intraoperative SmartPAN application. Trial endpoints covered SmartPAN usability and safety according to reports completed by surgeons after each operation, laboratory measurements of nonbiodegradable compounds in body fluids, and clinical evaluations over 30 days of follow-up. Results: In total 42 patients were recruited to the trial and 29 received partial pancreatic resection with SmartPAN application according to protocol. All 16 attending surgeons rated SmartPAN as easy to learn and use, mostly agreeing that it was useful and that they intended to use it frequently. No adverse effects or complications were associated with SmartPAN, nor were its compounds detected in blood or abdominal fluids. Positive leakage response was detected in 10/29 surgeries. POPF developed in 7 patients, including 2 intraoperatively detected leakages, thereof 1 with targeted closure as well as 5 with no leakage detected. Conclusion: This study represents the first-in-human clinical trial of SmartPAN and the precursor to randomized controlled trials. The outcomes support SmartPAN’s clinical usability and safety and showcase the device’s potential to intraoperatively visualize precursors of POPF.
Year:
2025
Year:
2025
BACKGROUND Patient-reported outcome measures (PROMs) such as health-related quality of life (HRQoL) are usually assessed at greater time intervals such as diagnostic time points, after treatment, and during follow-up. Many PROMs require frequent data collection (weekly or daily). Electronic PROMs enable high-resolution tracking but face declining response rates. Integrating PROMs into electronic health records (EHRs) could improve response rates and personalize therapy. OBJECTIVE This study aimed to evaluate the technical and clinical feasibility of high-frequency HRQoL assessments for routine care in EHRs. METHODS Patients receive emails on their mobile devices with 1-time links to a web-based app called MyEDC. This app communicates with an electronic data capture proxy in the demilitarized zone of the hospital. With a polling mechanism, these patient data are transferred to the protected hospital network and uploaded to the EHR system. HRQoL on a visual analog scale is assessed over the course of treatment in 4 clinical use cases: psychosomatics, hematology, visceral surgery, and neurosurgery. RESULTS Quality of life (QoL) trajectories were collected for 110 patients with daily or weekly data collection between 2 weeks and 3 months. The HRQoL analyses revealed clinically relevant findings across the 4 different medical domains. In the use case psychosomatics, 36 patients showed a significant increase in HRQoL following 4 weeks of therapy, rising from a median of 42% (IQR 32%-52%) to 60% (IQR 41%-67%; P =.01). An analysis of 25 patients in hematology demonstrated a significant correlation between HRQoL and 30-item QoL Questionnaire (EORTC QLQ-C30) global health status score ( P =.02). For 26 patients in visceral surgery, a significant association was observed between HRQoL and the reported pain level ( P <.001). The clinical feasibility was further highlighted in the neurosurgery use case, where 23 patients showed a median response time to the electronic PROM questionnaires of 5.3 (IQR 0.6-17.7) hours. HRQoL values were associated with disease-specific symptoms and scores, indicating clinical validity of this readout. Considerable variability of HRQoL was observed over time, both intraindividually and interindividually. Median area under the curve of HRQoL ranged from 0.46 to 0.79. Median time to answer ranged from 0.9 to 7.1 hours. No significant association between number of responses and age was observed. CONCLUSIONS High-resolution QoL trajectories with EHR integration are technically and clinically feasible. They offer a novel readout beyond survival analysis or PROM end point, enabling precise disease characterization and treatment comparison.
Year:
2025
Year:
2025
Year:
2024
Year:
2024
Year:
2024
Year:
2023
Year:
2023
Year:
2023
Year:
2023
Year:
2023
Collaborators (8)
Rosa Klotz
University Hospital Heidelberg
Sandro M. Krieg
Chair
University Hospital Heidelberg
Lucy Joanne Kessler
University Hospital Heidelberg
Hans‐Christoph Friederich
-
Christoph Lehmann
Director, Clinical Informatics Center; Professor
University of Texas Southwestern Medical Center
Petra Knaup
University Hospital Heidelberg
Pascal Probst
Heidelberg University
Hannes G Kenngott
University Hospital Heidelberg
Make your next move.
Prepare your application
Review drafts before you send them.