BACKGROUND: Large language models (LLMs) have shown remarkable capabilities in natural language processing (NLP), especially in domains where labeled data are scarce or expensive, such as the clinical domain. However, to unlock the clinical knowledge hidden in these LLMs, we need to design effective prompts that can guide them to perform specific clinical NLP tasks without any task-specific training data. This is known as in-context learning, which is an art and science that requires understanding the strengths and weaknesses of different LLMs and prompt engineering approaches.
OBJECTIVE: The objective of this study is to assess the effectiveness of various prompt engineering techniques, including 2 newly introduced types-heuristic and ensemble prompts, for zero-shot and few-shot clinical information extraction using pretrained language models.
METHODS: This comprehensive experimental study evaluated different prompt types (simple prefix, simple cloze, chain of thought, anticipatory, heuristic, and ensemble) across 5 clinical NLP tasks: clinical sense disambiguation, biomedical evidence extraction, coreference resolution, medication status extraction, and medication attribute extraction. The performance of these prompts was assessed using 3 state-of-the-art language models: GPT-3.5 (OpenAI), Gemini (Google), and LLaMA-2 (Meta). The study contrasted zero-shot with few-shot prompting and explored the effectiveness of ensemble approaches.
RESULTS: The study revealed that task-specific prompt tailoring is vital for the high performance of LLMs for zero-shot clinical NLP. In clinical sense disambiguation, GPT-3.5 achieved an accuracy of 0.96 with heuristic prompts and 0.94 in biomedical evidence extraction. Heuristic prompts, alongside chain of thought prompts, were highly effective across tasks. Few-shot prompting improved performance in complex scenarios, and ensemble approaches capitalized on multiple prompt strengths. GPT-3.5 consistently outperformed Gemini and LLaMA-2 across tasks and prompt types.
CONCLUSIONS: This study provides a rigorous evaluation of prompt engineering methodologies and introduces innovative techniques for clinical information extraction, demonstrating the potential of in-context learning in the clinical domain. These findings offer clear guidelines for future prompt-based clinical NLP research, facilitating engagement by non-NLP experts in clinical NLP advancements. To the best of our knowledge, this is one of the first works on the empirical evaluation of different prompt engineering approaches for clinical NLP in this era of generative artificial intelligence, and we hope that it will inspire and inform future research in this area.

BACKGROUND: This exploratory study compares self-reported COVID-19 vaccine side effects and breakthrough infections in people who described themselves as having diabetes with those who did not identify as having diabetes.
OBJECTIVE: The study uses person-reported data to evaluate differences in the perception of COVID-19 vaccine side effects between adults with diabetes and those who did not report having diabetes.
METHODS: This is a retrospective cohort study conducted using data provided online by adults aged 18 years and older residing in the United States. The participants who voluntarily self-enrolled between March 19, 2021, and July 16, 2022, in the IQVIA COVID-19 Active Research Experience project reported clinical and demographic information, COVID-19 vaccination, whether they had experienced any side effects, test-confirmed infections, and consented to linkage with prescription claims. No distinction was made for this study to differentiate prediabetes or type 1 and type 2 diabetes nor to verify reports of positive COVID-19 tests. Person-reported medication use was validated using pharmacy claims and a subset of the linked data was used for a sensitivity analysis of medication effects. Multivariate logistic regression was used to estimate the adjusted odds ratios of vaccine side effects or breakthrough infections by diabetic status, adjusting for age, gender, education, race, ethnicity (Hispanic or Latino), BMI, smoker, receipt of an influenza vaccine, vaccine manufacturer, and all medical conditions. Evaluations of diabetes medication-specific vaccine side effects are illustrated graphically to support the examination of the magnitude of side effect differences for various medications and combinations of medications used to manage diabetes.
RESULTS: People with diabetes (n=724) reported experiencing fewer side effects within 2 weeks of vaccination for COVID-19 than those without diabetes (n=6417; mean 2.7, SD 2.0 vs mean 3.1, SD 2.0). The adjusted risk of having a specific side effect or any side effect was lower among those with diabetes, with significant reductions in fatigue and headache but no differences in breakthrough infections over participants\' maximum follow-up time. Diabetes medication use did not consistently affect the risk of specific side effects, either using self-reported medication use or using only diabetes medications that were confirmed by pharmacy health insurance claims for people who also reported having diabetes.
CONCLUSIONS: People with diabetes reported fewer vaccine side effects than participants not reporting having diabetes, with a similar risk of breakthrough infection.
BACKGROUND: ClinicalTrials.gov NCT04368065; https://clinicaltrials.gov/study/NCT04368065.

Erythroderma is a condition characterized by erythema affecting at least 90% of the skin surface area. It can be caused by various underlying conditions. Due to nonspecific clinical and laboratory findings, determining the cause may pose a challenge. In the retrospective study, we identified 212 patients hospitalized for erythroderma in the Department of Dermatology, Venereology, and Allergology at Wroclaw Medical University between January 2012 and March 2022. Clinical, laboratory, and histopathological features, as well as the management of patients, were studied. The median age of adults was 61 years (IQR = 47-68). The most common causes of erythroderma were psoriasis (n = 49, 24.01%), followed by atopic dermatitis (AD) (n = 27, 13.23%), and cutaneous T-cell lymphomas (CTCL) (n = 27, 13.23%). Despite laboratory tests and histopathological examination, the etiology of erythroderma remained undetermined in 39 cases (19.12%). In 70.59% of patients, it was the first episode of erythroderma, while 29.41% experienced a recurrent episode. Regardless of the etiology of erythroderma, patients were most frequently treated with systemic antihistamines (146 cases, 71.57%) and systemic steroids (132 cases, 64.71%). Patients with idiopathic erythroderma constitute the greatest diagnostic and therapeutic challenge, requiring particularly thorough evaluation.

OBJECTIVE: Identification of patients at high risk of aggressive prostate cancer is a major clinical challenge. With the view of developing artificial intelligence-based methods for identification of these patients, we are constructing a comprehensive clinical database including 7448 prostate cancer (PCa) Danish patients. In this paper we provide an epidemiological description and patients\' trajectories of this retrospective observational population, to contribute to the understanding of the characteristics and pathways of PCa patients in Denmark.
RESULTS: Individuals receiving a PCa diagnosis during 2008-2014 in Region Southern Denmark were identified, and all diagnoses, operations, investigations, and biochemistry analyses, from 4 years prior, to 5 years after PCa diagnosis were obtained. About 85.1% were not diagnosed with metastatic PCa during the study period (unaggressive PCa); 9.2% were simultaneously diagnosed with PCa and metastasis (aggressive-advanced PCa), while 5.7% were not diagnosed with metastatic PCa at first, but they were diagnosed with metastasis at some point during the 5 years follow-up (aggressive-not advanced PCa). Patients with unaggressive PCa had more clinical investigations directly related to PCa detection (prostate ultrasounds and biopsies) during the 4 years prior to PCa diagnosis, compared to patients with aggressive PCa, which may have contributed to the early detection of PCa.

BACKGROUND: Large data sets comprising routine clinical data are becoming increasingly available for use in health research. These data sets contain many clinical variables that might not lend themselves to use in research. Structural equation modelling (SEM) is a statistical technique that might allow for the creation of \"research-friendly\" clinical constructs from these routine clinical variables and therefore could be an appropriate analytic method to apply more widely to routine clinical data.
OBJECTIVE: SEM was applied to a large data set of routine clinical data developed in East London to model well-established clinical associations. Depression is common among patients with type 2 diabetes, and is associated with poor diabetic control, increased diabetic complications, increased health service utilization, and increased health care costs. Evidence from trial data suggests that integrating psychological treatment into diabetes care can improve health status and reduce costs. Attempting to model these known associations using SEM will test the utility of this technique in routine clinical data sets.
METHODS: Data were cleaned extensively prior to analysis. SEM was used to investigate associations between depression, diabetic control, diabetic care, mental health treatment, and Accident & Emergency (A&E) use in patients with type 2 diabetes. The creation of the latent variables and the direction of association between latent variables in the model was based upon established clinical knowledge.
RESULTS: The results provided partial support for the application of SEM to routine clinical data. Overall, 19% (3106/16,353) of patients with type 2 diabetes had received a diagnosis of depression. In line with known clinical associations, depression was associated with worse diabetic control (β=.034, P<.001) and increased A&E use (β=.071, P<.001). However, contrary to expectation, worse diabetic control was associated with lower A&E use (β=-.055, P<.001) and receipt of mental health treatment did not impact upon diabetic control (P=.39). Receipt of diabetes care was associated with better diabetic control (β=-.072, P<.001), having depression (β=.018, P=.007), and receiving mental health treatment (β=.046, P<.001), which might suggest that comprehensive integrated care packages are being delivered in East London.
CONCLUSIONS: Some established clinical associations were successfully modelled in a sample of patients with type 2 diabetes in a way that made clinical sense, providing partial evidence for the utility of SEM in routine clinical data. Several issues relating to data quality emerged. Data improvement would have likely enhanced the utility of SEM in this data set.

Radiomic diagnosis models generally consider only a single dimension of information, leading to limitations in their diagnostic accuracy and reliability. The integration of multiple dimensions of information into the deep learning model have the potential to improve its diagnostic capabilities. The purpose of study was to evaluate the performance of deep learning model in distinguishing tuberculosis (TB) nodules and lung cancer (LC) based on deep learning features, radiomic features, and clinical information.
Positron emission tomography (PET) and computed tomography (CT) image data from 97 patients with LC and 77 patients with TB nodules were collected. One hundred radiomic features were extracted from both PET and CT imaging using the pyradiomics platform, and 2048 deep learning features were obtained through a residual neural network approach. Four models included traditional machine learning model with radiomic features as input (traditional radiomics), a deep learning model with separate input of image features (deep convolutional neural networks [DCNN]), a deep learning model with two inputs of radiomic features and deep learning features (radiomics-DCNN) and a deep learning model with inputs of radiomic features and deep learning features and clinical information (integrated model). The models were evaluated using area under the curve (AUC), sensitivity, accuracy, specificity, and F1-score metrics.
The results of the classification of TB nodules and LC showed that the integrated model achieved an AUC of 0.84 (0.82-0.88), sensitivity of 0.85 (0.80-0.88), and specificity of 0.84 (0.83-0.87), performing better than the other models.
The integrated model was found to be the best classification model in the diagnosis of TB nodules and solid LC.

UNASSIGNED: The CFTR-modulating therapy Elexaftor - Tezacaftor - Ivacaftor (ETI) has been widely prescribed since its approval in 2020 in the European Union. The aim of this study was to methodically evaluate the effects of an ETI treatment on clinical, biochemical data and Pseudomonas colonization in order to demonstrate its efficacy.
UNASSIGNED: This prospective monocentric study comprised 69 patients diagnosed with cystic fibrosis aged at least 12 years and treated with ETI between September 2020 and November 2021. Clinical and laboratory data of each patient and study visit were collected before and after 24 weeks of ETI treatment. Follow-up status of Pseudomonas aeruginosa (PsA) colonization was assessed after one year of therapy by regularly determined sputum or throat swab samples.
UNASSIGNED: Marked improvements biochemical markers of systemic inflammation as white blood cell count, levels of immunoglobulins A, G and M and albumin within 24 weeks of therapy were observed. ETI treatment proved to be effective as seen by amelioration of lung function and sweat chloride concentration. Assessment of PsA colonization status revealed a conversion from a positive to negative detection in 36% of the cases after one year of therapy.
UNASSIGNED: ETI treatment effectively improves systemic inflammation parameters and shows promising results in PsA status conversion.

BACKGROUND: Drug-induced prolongation of the corrected QT interval (QTc) increases the risk for Torsades de Pointes (TdP) and sudden cardiac death. Medication effects on the QTc have been studied in controlled settings but may not be well evaluated in real-world settings where medication effects may be modulated by patient demographics and comorbidities as well as the usage of other concomitant medications.
OBJECTIVE: We demonstrate a new, high-throughput method leveraging electronic health records (EHRs) and the Surescripts pharmacy database to monitor real-world QTc-prolonging medication and potential interacting effects from demographics and comorbidities.
METHODS: We included all outpatient electrocardiograms (ECGs) from September 2008 to December 2019 at a large academic medical system, which were in sinus rhythm with a heart rate of 40-100 beats per minute, QRS duration of <120 milliseconds, and QTc of 300-700 milliseconds, determined using the Bazett formula. We used prescription information from the Surescripts pharmacy database and EHR medication lists to classify whether a patient was on a medication during an ECG. Negative control ECGs were obtained from patients not currently on the medication but who had been or would be on that medication within 1 year. We calculated the difference in mean QTc between ECGs of patients who are on and those who are off a medication and made comparisons to known medication TdP risks per the CredibleMeds.org database. Using linear regression analysis, we studied the interaction of patient-level demographics or comorbidities on medication-related QTc prolongation.
RESULTS: We analyzed the effects of 272 medications on 310,335 ECGs from 159,397 individuals. Medications associated with the greatest QTc prolongation were dofetilide (mean QTc difference 21.52, 95% CI 10.58-32.70 milliseconds), mexiletine (mean QTc difference 18.56, 95% CI 7.70-29.27 milliseconds), amiodarone (mean QTc difference 14.96, 95% CI 13.52-16.33 milliseconds), rifaximin (mean QTc difference 14.50, 95% CI 12.12-17.13 milliseconds), and sotalol (mean QTc difference 10.73, 95% CI 7.09-14.37 milliseconds). Several top QT prolonging medications such as rifaximin, lactulose, cinacalcet, and lenalidomide were not previously known but have plausible mechanistic explanations. Significant interactions were observed between demographics or comorbidities and QTc prolongation with many medications, such as coronary disease and amiodarone.
CONCLUSIONS: We demonstrate a new, high-throughput technique for monitoring real-world effects of QTc-prolonging medications from readily accessible clinical data. Using this approach, we confirmed known medications for QTc prolongation and identified potential new associations and demographic or comorbidity interactions that could supplement findings in curated databases. Our single-center results would benefit from additional verification in future multisite studies that incorporate larger numbers of patients and ECGs along with more precise medication adherence and comorbidity data.

Information systems used by platform trials should handle changes that are not predefined. Unfortunately, the technical architecture of most existing clinical data management systems (CDMS) do not support changes to be incorporated into an ongoing trial. Adaptive clinical trials need advanced architectural solutions setup to enable biomarker stratification and enrichment strategy necessary for the adaptive clinical trial operation. This short paper presents the microservices-based architecture solution that is used to run and support the adaptive RECORDS-Trial.

The primary aim was to predict upper airway collapse sites found in drug-induced sleep endoscopy (DISE) from demographic, anthropometric, clinical examination, sleep study, and snoring sound parameters in patients with obstructive sleep apnea (OSA). The secondary aim was to identify the above-mentioned parameters that are associated with complete concentric collapse of the soft palate.
All patients with OSA who underwent DISE and simultaneous snoring sound recording were enrolled in this study. Demographic, anthropometric, clinical examination (viz., modified Mallampati classification and Friedman tonsil classification), and sleep study parameters were extracted from the polysomnography and DISE reports. Snoring sound parameters during DISE were calculated.
One hundred and nineteen patients with OSA (79.8% men; age = 48.1 ± 12.4 years) were included. Increased body mass index was found to be associated with higher probability of oropharyngeal collapse (P < .01; odds ratio = 1.29). Patients with a high Friedman tonsil score were less likely to have tongue base collapse (P < .01; odd ratio = 0.12) and epiglottic collapse (P = .01; odds ratio = 0.20) than those with a low score. A longer duration of snoring events (P = .05; odds ratio = 2.99) was associated with a higher probability of complete concentric collapse of the soft palate.
Within the current patient profile and approach, given that only a limited number of predictors were identified, it does not seem feasible to predict upper airway collapse sites found in DISE from demographic, anthropometric, clinical examination, sleep study, and snoring sound parameters in patients with OSA.
Huang Z, Bosschieter PFN, Aarab G, et al. Predicting upper airway collapse sites found in drug-induced sleep endoscopy from clinical data and snoring sounds in obstructive sleep apnea patients: a prospective clinical study. J Clin Sleep Med. 2022;18(9):2119-2131.