Felice Burn

Background: Medical documentation imposes a significant administrative burden on physicians and reduces time for direct patient care. Artificial intelligence (AI)-assisted tools such as automatic speech recognition and large language models (LLMs) promise to reduce this burden, but their performance in multilingual environments has not been explored. Switzerland is highly multilingual, and non-native German-speaking physicians may find documentation particularly challenging. Objective: This study aimed to compare the efficiency and documentation quality of four clinical documentation workflows—including both AI-assisted and traditional methods—in a Swiss tertiary hospital setting characterized by linguistic diversity. Methods: In this proof-of-concept study at a Swiss tertiary hospital (Department of Plastic and Hand Surgery, Cantonal Hospital Aarau), two physicians—a native Swiss German speaker and a non-native German speaker—documented encounters with simulated patients having common hand disorders. Four documentation workflows were tested: (1) traditional dictation with transcription by a secretary; (2) real-time dictation using speech recognition software for voice to text transcription; (3) postencounter dictation transcribed by an AI (Whisper) and processed by a GPT-based agent; and (4) AI-assisted ambient dictation of entire appointments using audio recording and automatic transcription. Documentation efficiency was measured by recorded physician time, and note quality was assessed using a modified Physician Documentation Quality Instrument (PDQI-9) scored by three different LLMs. To protect patient privacy, only synthetic (simulated) patient data were used. Results: AI-assisted workflows—particularly workflow 4 (AI-assisted ambient dictation)—produced the shortest physician documentation times per report. In post-hoc comparisons, workflow 4 was significantly faster than solely the speech recognition software workflow (workflow 2) for both physicians (adjusted P <.001). For the non-native speaker, workflow 4 was not significantly faster than traditional dictation (workflow 1) after adjustment ( P =.08). LLM evaluators assigned high absolute scores (median PDQI-9 >47/50); however, inter-rater reliability was poor (Krippendorff’s alpha=−.433, 95% CI: −0.444 to −0.416), indicating systematic disagreement that precludes definitive conclusions about documentation quality from these scores alone. Conclusions: AI-assisted documentation demonstrated significant time savings for the native speaker, though the reduction for the non-native speaker did not reach statistical significance in this pilot ( P =.08). Such tools show potential to alleviate the linguistic challenges faced by non-native speakers, reduce administrative burdens, and enable physicians to spend more time with patients. However, the inconsistency of AI-based quality scoring suggests that LLMs cannot yet reliably replace human evaluation. Future studies should evaluate these workflows in real-world clinical implementation, address data privacy and security issues, and include human evaluators to validate the benefits observed in this study.

Purpose: Hyponatremia remains one of the most prevalent electrolyte disorders among geriatric inpatients but is frequently under-recognized in clinical coding. This cost-of-illness (COI) study aimed to (1) estimate the corrected prevalence of uncoded hyponatremia (ucHn) by integrating laboratory and administrative data, and (2) quantify its incremental social and economic burden on elderly patients. Patients and Methods: A retrospective COI analysis covered 72,730 inpatient cases ≥ 70 years (2016– 2024) in a Swiss hospital network. Hyponatremia was classified as coded (cHn; ICD-10 E87.1) or uncoded (ucHn; Na < 135 mmol/L). Incremental costs were estimated via generalized linear models and probabilistic sensitivity analysis. Results: Among 13,657 patients with hyponatremia (18.7%), only 2,070 (2.8%) were coded. Prevalences were as follows: coded 2.8%, hidden15.9%, true prevalence 18.7%. ucHn was more prevalent in men (p< 0.0001), presented with more mild cases (83.1/34.9%, p< 0.0001), more incident cases (14.3/4.7%, p< 0.0001), lower chronic hyponatremia (20.8/39.6% p< 0.0001), significantly more frequent heart failure and lung cancer (p< 0.0001), had higher asset cost (2,463 vs 1,654, p< 0.0001), lower contribution margin 1 and 2 (p< 0.0001) and a markedly higher 30-day mortality (46% vs 7%, p< 0.0001). Mean length of stay (LOS) was 9.9 days vs 7.4 days (p< 0.0001). ucHn generated incremental costs of CHF per case with a total systemic burden (2016– 2024) amounted to ≈ 355 to 473 MCHF. Losses of autonomy, mobility, cognitive control and life years underline the impressive social impact for nearly every ucHn case. Conclusion: For the first time, the detected ucHn is evaluated. It presents a substantial social and economic burden with a previously unquantified excess mortality associated with uncoded hyponatremia. ucHn is widely underestimated in administrative hospital data. This COI study supports policy measures to improve documentation and awareness of hyponatremia in elderly patients with the aim of reducing its social impact. Plain Language Summary: Why was this study done? Hyponatremia means that the concentration of sodium in the blood is too low, most often because of an excess of water rather than a true lack of sodium. This condition is common in older hospital patients and can cause serious problems like falls, confusion and even death. However, doctors often fail to record this condition in medical files. The researchers wanted to find out how many cases are missed and what this means for patients and hospitals. What did the researchers do and find? The team analyzed medical records from over 72,000 patients aged 70 and older at a Swiss hospital between 2016 and 2024. They combined blood test results with hospital records to identify both recorded and unrecorded cases of hyponatremia. Their findings were striking: while blood tests showed that 13,657 patients (18.7%) had hyponatremia, only 2,070 cases (2.8%) were recorded in their medical files. This means 85% of cases went undocumented. Patients with unrecorded hyponatremia fared worse than those whose condition was documented. They stayed in hospital longer (9.9 days versus 7.4 days) and their death rate within 30 days was dramatically higher (46% versus 7%). Many lost their independence, mobility, and mental sharpness. The economic impact was substantial, with unrecorded cases costing Swiss hospitals between 355 and 473 million francs over nine years. What do these results mean? This study reveals that unrecorded hyponatremia creates a hidden burden on older patients and healthcare systems. Better awareness and documentation could help doctors recognize and treat this condition earlier, potentially preventing serious complications and saving lives. The findings support the need for improved recording practices in hospitals to ensure patients receive appropriate care.

Writing discharge notes (DNs) is a time-intensive task in inpatient care that contributes substantially to physician workload and burnout. Large language models (LLMs) offer potential opportunities to automate clinical documentation. However, concerns regarding accuracy, reliability, data privacy protection laws and data governance, and compliance with healthcare privacy remain. This study evaluated the performance of locally deployed, domain-adapted LLMs for generating German-language DN in internal medicine. In this retrospective study, electronic health record data from a Swiss tertiary care hospital were used. Clinical information from 3 inpatient scenarios served as input for the 3 open source LLMs (Mixtral 8×7B, Mixtral 8×22B, and LLaMA 3.1 70B). Zero-shot prompting, 4-shot in-context learning, and supervised fine-tuning using QLoRA were compared. Generated DNs were evaluated using automated metrics, blinded physician assessment with direct comparison of all relevant clinical information using the modified Physician Documentation Quality Instrument, and additionally an LLM-as-a-judge framework and a Turing Test. Across all evaluation methods, the fine-tuned Mixtral 8×7B model achieved the best overall performance compared with prompting-based approaches and larger nonspecialized models. Simpler cases consistently received higher-quality ratings. Physician inter-rater agreement was low, highlighting the challenges of objective quality assessment for clinical documentation. In a Turing Test, physicians were unable to reliably distinguish artificial intelligence-generated from clinician-authored DNs, performing at chance level, demonstrating that high-quality DN generation is feasible using locally deployed LLMs within secure hospital infrastructures. Domain-specific fine-tuning substantially improves performance and may enable practical deployment. Future work should address further hallucination mitigation, robustness across medical specialties, and real-world integration.

BACKGROUND Medical documentation imposes a significant administrative burden on physicians, reducing time for patient interaction and increasing the risk of burnout.1 AI-assisted solutions have shown promise in easing this burden, but their effectiveness in multilingual healthcare settings remains underexplored. Switzerland’s linguistic diversity presents unique documentation challenges, particularly for non-native speakers who must navigate multiple languages and dialects. OBJECTIVE To compare the efficiency and documentation quality of four medical documentation workflows—including both AI-assisted and traditional methods—in a Swiss tertiary hospital setting characterized by linguistic diversity. METHODS In this proof-of-concept study at a Swiss tertiary hospital (Department of Plastic and Hand Surgery, Cantonal Hospital Aarau), two physicians—a native Swiss German speaker and a non-native German speaker—documented encounters with simulated patients having common hand disorders. Four documentation workflows were tested: (1) traditional dictation with transcription by a secretary; (2) real-time dictation using speech recognition software for voice to text transcription; (3) post-encounter dictation transcribed by an AI (Whisper) and processed by a GPT-based agent; and (4) AI-assisted ambient dictation of entire appointments using audio recording and automatic transcription. Documentation efficiency was measured by recorded physician time, and note quality was assessed using a modified Physician Documentation Quality Instrument (PDQI-9) scored by three different large language models (LLMs). To protect patient privacy, only synthetic (simulated) patient data were used. RESULTS AI-assisted workflows—particularly workflow 4 (AI-assisted ambient dictation)—produced the shortest physician documentation times per report. In post-hoc comparisons, workflow 4 was significantly faster than solely the speech recognition software workflow (workflow 2) for both physicians (adjusted P < 0.001). For the non-native speaker, workflow 4 was not significantly faster than traditional dictation (workflow 1) after adjustment (P = 0.077). All workflows achieved uniformly high documentation quality scores (median PDQI-9 total score 47.3/50). A Friedman test indicated a statistically significant difference in quality scores between workflows (χ² = 36, P < 0.001), but the absolute score differences were <1 point, which is not clinically meaningful. CONCLUSIONS AI-assisted documentation—especially ambient AI dictation—has significant potential to streamline medical documentation in multilingual healthcare settings without compromising quality. Such tools can help alleviate the linguistic challenges faced by non-native speakers, reduce administrative burdens, and enable physicians to spend more time with patients. Future studies should evaluate these workflows in real-world clinical implementation, address data privacy and security issues, and include human evaluators to validate the benefits observed in this study.

Objectives: A substantial number of incidental pulmonary embolisms (iPEs) in computed tomography scans are missed by radiologists in their daily routine. This study analyzes the radiological reports of iPE cases before and after implementation of an artificial intelligence (AI) algorithm for iPE detection. Furthermore, we investigate the anatomic distribution patterns within missed iPE cases and mortality within a 90-day follow-up in patients before and after AI use. Materials and Methods: This institutional review board–approved observational single-center study included 5298 chest computed tomography scans performed for reasons other than suspected pulmonary embolism (PE). We compared 2 cohorts: cohort 1, consisting of 1964 patients whose original radiology reports were generated before the implementation of an AI algorithm, and cohort 2, consisting of 3334 patients whose scans were analyzed after the implementation of an Food and Drug Administration–approved and CE-certified AI algorithm for iPE detection (Aidoc Medical, Tel Aviv, Israel). For both cohorts, any discrepancies between the original radiology reports and the AI results were reviewed by 2 thoracic imaging subspecialized radiologists. In the original radiology report and in case of discrepancies with the AI algorithm, the expert review served as reference standard. Sensitivity, specificity, prevalence, negative predictive value (NPV), and positive predictive value (PPV) were calculated. The rates of missed iPEs in both cohorts were compared statistically using STATA (Version 17.1). Kaplan-Meier curves and Cox proportional hazards models were used for survival analysis. Results: In cohort 1 (mean age 70.6 years, 48% female [n = 944], 52% male [n = 1020]), the prevalence of confirmed iPE was 2.2% (n = 42), and the AI detected 61 suspicious iPEs, resulting in a sensitivity of 95%, a specificity of 99%, a PPV of 69%, and an NPV of 99%. Radiologists missed 50% of iPE cases in cohort 1. In cohort 2 (mean age 69 years, 47% female [n = 1567], 53% male [n = 1767]), the prevalence of confirmed iPEs was 1.7% (56/3334), with AI detecting 59 suspicious cases (sensitivity 90%, specificity 99%, PPV 95%, NPV 99%). The rate of missed iPEs by radiologists dropped to 7.1% after AI implementation, showing a significant improvement (P < 0.001). Most overlooked iPEs (61%) were in the right lower lobe. The survival analysis showed no significantly decreased 90-day mortality rate, with a hazards ratio of 0.95 (95% confidence interval, 0.45–1.96; P = 0.88). Conclusions: The implementation of an AI algorithm significantly reduced the rate of missed iPEs from 50% to 7.1%, thereby enhancing diagnostic accuracy. Despite this improvement, the 90-day mortality rate remained unchanged. These findings highlight the AI tool's potential to assist radiologists in accurately identifying iPEs, although its implementation does not significantly affect short-term survival. Notably, most missed iPEs were located in the right lower lobe, suggesting that radiologists should pay particular attention to this area during evaluations.

In this work, several machine learning (ML) algorithms, both classical ML and modern deep learning, were investigated for their ability to improve the performance of a pipeline for the segmentation and classification of prostate lesions using MRI data. The algorithms were used to perform a binary classification of benign and malignant tissue visible in MRI sequences. The model choices include support vector machines (SVMs), random decision forests (RDFs), and multi-layer perceptrons (MLPs), along with radiomic features that are reduced by applying PCA or mRMR feature selection. Modern CNN-based architectures, such as ConvNeXt, ConvNet, and ResNet, were also evaluated in various setups, including transfer learning. To optimize the performance, different approaches were compared and applied to whole images, as well as gland, peripheral zone (PZ), and lesion segmentations. The contribution of this study is an investigation of several ML approaches regarding their performance in prostate cancer (PCa) diagnosis algorithms. This work delivers insights into the applicability of different approaches for this context based on an exhaustive examination. The outcome is a recommendation or preference for which machine learning model or family of models is best suited to optimize an existing pipeline when the model is applied as an upstream filter.

Background: After breast conserving surgery (BCS), surgical clips indicate the tumor bed and, thereby, the most probable area for tumor relapse. The aim of this study was to investigate whether a U-Net-based deep convolutional neural network (dCNN) may be used to detect surgical clips in follow-up mammograms after BCS. Methods: 884 mammograms and 517 tomosynthetic images depicting surgical clips and calcifications were manually segmented and classified. A U-Net-based segmentation network was trained with 922 images and validated with 394 images. An external test dataset consisting of 39 images was annotated by two radiologists with up to 7 years of experience in breast imaging. The network’s performance was compared to that of human readers using accuracy and interrater agreement (Cohen’s Kappa). Results: The overall classification accuracy on the validation set after 45 epochs ranged between 88.2% and 92.6%, indicating that the model’s performance is comparable to the decisions of a human reader. In 17.4% of cases, calcifications have been misclassified as post-operative clips. The interrater reliability of the model compared to the radiologists showed substantial agreement (κ_reader1 = 0.72, κ_reader2 = 0.78) while the readers compared to each other revealed a Cohen’s Kappa of 0.84, thus showing near-perfect agreement. Conclusions: With this study, we show that surgery clips can adequately be identified by an AI technique. A potential application of the proposed technique is patient triage as well as the automatic exclusion of post-operative cases from PGMI (Perfect, Good, Moderate, Inadequate) evaluation, thus improving the quality management workflow.

Background: Measurement of skeletal muscle index (SMI) in computed tomography has been suggested to improve the objective assessment of muscle mass. While most studies have focused on lumbar vertebrae, we examine the association of SMI at the thoracic level with nutritional and clinical outcomes and response to nutritional intervention. Methods: We conducted a secondary analysis of EFFORT, a Swiss-wide, multicenter, randomized trial. We investigated the association of low SMI at the 12th thoracic vertebra (T12) with adverse outcome within 30 days after hospital admission (primary endpoint). Results: 663 of 2028 patients from the EFFORT trial had available CT scans for T12, and 519 among them also had available L3 scans. Mean SMI at T12 was 22.4 ± 5.8 cm²/m² and 19.6 ± 5.5 cm²/m² in male and female patients, respectively, and correlated well with nutritional parameters, including nutritional risk based on NRS 2002 (adjusted coefficient −0.63, 95%CI −1.25 to −0.01, p = 0.047), BMI (adjusted coefficient 0.74, 95%CI 0.66 to 0.82, p < 0.001) and handgrip strength (adjusted coefficient 0.15, 95%CI 0.11 to 0.2, p < 0.001). In multivariate regression analyses, low SMI was not a significant predictor for either clinical outcome or for treatment response. Results for SMI measured at L3 were similar, with only little prognostic value. Conclusions: Within medical patients at risk for malnutrition, SMI at thoracic vertebra provided low prognostic information regarding clinical outcomes and nutritional treatment response.

Introduction Smoking prevalence is twice as high among patients admitted to hospital because of the acute condition of aneurysmal subarachnoid hemorrhage (aSAH) as in the general population. Smoking cessation may improve the prognosis of aSAH, but nicotine replacement therapy (NRT) administered at the time of aSAH remains controversial because of potential adverse effects such as cerebral vasospasm. We investigated the international practice of NRT use for aSAH among neurosurgeons. Methods The online SurveyMonkey software was used to administer a 15-question, 5-min online questionnaire. An invitation link was sent to those 1425 of 1988 members of the European Association of Neurosurgical Societies (EANS) who agreed to participate in surveys to assess treatment strategies for withdrawal of tobacco smoking during aSAH. Factors contributing to physicians’ posture towards NRT were assessed. Results A total of 158 physicians from 50 nations participated in the survey (response rate 11.1%); 68.4% (108) were affiliated with university hospitals and 67.7% (107) practiced at high-volume neurovascular centers with at least 30 treated aSAH cases per year. Overall, 55.7% (88) of physicians offered NRT to smokers with aSAH, 22.1% (35) offered non-NRT support including non-nicotine medication and counselling, while the remaining 22.1% (35) did not actively support smoking cessation. When smoking was not possible, 42.4% (67) of physicians expected better clinical outcomes when prescribing NRT instead of nicotine deprivation, 36.1% (57) were uncertain, 13.9% (22) assumed unaffected outcomes, and 7.6% (12) assumed worse outcomes. Only 22.8% (36) physicians had access to a local smoking cessation team in their practice, of whom half expected better outcomes with NRT as compared to deprivation. Conclusions A small majority of the surveyed physicians of the EANS offered NRT to support smoking cessation in hospitalized patients with aSAH. However, less than half believed that NRT could positively impact clinical outcome as compared to deprivation. This survey demonstrated the lack of consensus regarding use of NRT for hospitalized smokers with aSAH.