Medical Decision Research Articles

Comparison of Faster R-CNN, YOLO, and SSD for Third Molar Angle Detection in Dental Panoramic X-rays.

The use of artificial intelligence algorithms (AI) has gained importance for dental applications in recent years. Analyzing AI information from different sensor data such as images or panoramic radiographs (panoramic X-rays) can help to improve medical decisions and achieve early diagnosis of different dental pathologies. In particular, the use of deep learning (DL) techniques based on convolutional neural networks (CNNs) has obtained promising results in dental applications based on images, in which approaches based on classification, detection, and segmentation are being studied with growing interest. However, there are still several challenges to be tackled, such as the data quality and quantity, the variability among categories, and the analysis of the possible bias and variance associated with each dataset distribution. This study aims to compare the performance of three deep learning object detection models-Faster R-CNN, YOLO V2, and SSD-using different ResNet architectures (ResNet-18, ResNet-50, and ResNet-101) as feature extractors for detecting and classifying third molar angles in panoramic X-rays according to Winter's classification criterion. Each object detection architecture was trained, calibrated, validated, and tested with three different feature extraction CNNs which are ResNet-18, ResNet-50, and ResNet-101, which were the networks that best fit our dataset distribution. Based on such detection networks, we detect four different categories of angles in third molars using panoramic X-rays by using Winter's classification criterion. This criterion characterizes the third molar's position relative to the second molar's longitudinal axis. The detected categories for the third molars are distoangular, vertical, mesioangular, and horizontal. For training, we used a total of 644 panoramic X-rays. The results obtained in the testing dataset reached up to 99% mean average accuracy performance, demonstrating the YOLOV2 obtained higher effectiveness in solving the third molar angle detection problem. These results demonstrate that the use of CNNs for object detection in panoramic radiographs represents a promising solution in dental applications.

General practitioners’ risk literacy and real-world prescribing of potentially hazardous drugs: a cross-sectional study

BackgroundOveruse of medical care is a pervasive problem. Studies using hypothetical scenarios suggest that physicians’ risk literacy influences medical decisions; real-world correlations, however, are lacking. We sought to determine the...

Top Ten Tips Palliative Care Clinicians Should Know About Caring for Patients Under Guardianships.

Palliative care clinicians often help facilitate coordination of care, complex serious illness, and end-of-life medical decision-making. However, the clinical and legal issues related to guardianship can complicate the decision-making process, care delivery, outcomes, and the role of the palliative care clinician. Adult patients who have a guardian have been found by a court to be unable to make some or all decisions for themselves. Providing care for patients under guardianship is where medicine overlaps with legal rights. It is crucial to be familiar with the patients' rights and the guardians' responsibilities to clarify medical decision-making processes and identify necessary authorities. This article uses an interprofessional approach to leverage the expertise of physicians, nurses, lawyers, and guardians and to guide palliative care clinicians to optimally support patients under guardianship.

How Much Information Is Too Much? An Experimental Examination of How Information Disclosures May Unintentionally Encourage the Withholding of Health Information.

Information disclosures are used in medicine to provide patients with relevant information. This research examines whether patients are less likely to discuss medical conditions with their physicians after seeing an insurance information disclosure. Three experimental studies with nonprobability online samples (ntotal = 875 US adult participants) examined the impact of information disclosures on patients' likelihood of disclosing symptoms to providers, using new symptoms and preexisting chronic conditions. The effects of insurance disclosures were also compared to those of pharmaceutical discount disclosures. These studies demonstrate that information disclosures can result in unintended consequences for patients and providers. Results showed that information disclosures about insurance claims significantly negatively affected willingness to discuss health information with providers. This effect was consistent for both new health concerns, b = -0.661, P < 0.001 (study 1, n = 250) and b = -0.893, P < 0.001 (study 3, n = 375), as well as chronic conditions, b = -1.175, P < .001 (study 2, n = 250); all studies were conducted in January 2023. Information provided to patients about pharmaceutical savings did not similarly affect willingness to discuss symptoms with providers. These were experimental studies with hypothetical scenarios. Future research is needed to understand how patients react to information disclosures in a physician's office. Future research is also needed to examine the role of specific wording and tone used in information disclosures. Prior research has shown that patients prefer more information and to be involved in their medical decisions; however, these studies demonstrate that some information disclosures can discourage full communication between patients and physicians. This research has important implications for the potential consequences of information disclosures in health care settings. Information disclosures should be presented in a way that will not discourage candid discussions of patient symptoms. This research found that information disclosures about insurance claims can negatively affect patient willingness to discuss health information with providers.Information disclosures may sometimes fall short of their intended purpose of aiding patient decisions with the goal of improved well-being.When information disclosures are focused on warning about potential new costs, patients may feel uncomfortable discussing new symptoms with their providers.Findings suggest patients may often be more concerned with costs than with addressing their ongoing health problems.

Considerations on the basis of medical reasoning for the use in AI applications.

This study discusses the integration of artificial intelligence (AI) and machine learning (ML) in medical reasoning and decision-making, with a focus on the challenges and opportunities associated with the massive consumption of data required for training AI systems, and contrasts this with the limited data typically available to medical practitioners. We advocate for a balanced approach that includes small data and emphasize the importance of maintaining the art of clinical reasoning amid technological advancements. Finally, we highlight the potential of multidisciplinary research in addressing the complexities of medical reasoning and suggest the necessity of careful abstraction and conceptual modeling in AI applications.

Acquisition of Prehospital Stroke Severity Scale is associated with shorter door-to-puncture times in patients with prehospital notifications transported directly to a thrombectomy center

BackgroundWe sought to identify systemic factors influencing door-to-puncture times (DTP) among patients with pre-arrival notifications presenting directly to a comprehensive stroke center (CSC) and undergoing emergent mechanical thrombectomy (MT).MethodsIn this...

Prediction model for poor short-term prognosis in patients with chronic subdural hematoma after burr hole drainage: a retrospective cohort study.

Chronic subdural hematoma (CSDH) is a common condition in neurosurgery. With an aging population, there is increasing attention on the prognosis of patients following surgical intervention. We developed a postoperative short-term prognostic prediction model using preoperative clinical indicators, aiming to assist in perioperative medical decision-making and management. The dataset was randomly divided into training and validation cohorts. An mRS score greater than 2 one month after discharge was considered indicative of a poor prognosis. In the training cohort, the least absolute shrinkage and selection operator (LASSO) regression analysis was used for multivariate analysis to identify independent risk factors and construct a prediction nomogram for poor prognosis one month after discharge. The performance of the nomogram was assessed using the Receiver Operating Characteristic (ROC) curve and calibration curve. A Decision Curve Analysis (DCA) was also conducted to determine the net benefit threshold of the prediction model. Among the 505 participants, 18.8% (95/505) had a poor prognosis one month after discharge. The baseline characteristics did not significantly differ between the training cohort and the validation cohort. LASSO regression analysis in the training cohort reduced the predictors to four potential factors. Further multivariate logistic analyses in the training cohort identified four independent predictors: age, admission Glasgow Coma Scale (GCS) score, hemiparesis, and hemoglobin count. These predictors were incorporated into the nomogram prediction model. Internal validation using ROC analysis, calibration curves, and other methods demonstrated a strong correlation between the observed and predicted likelihood of poor prognosis one month after discharge. The visualized nomogram prediction model we developed for short-term postoperative prognosis of chronic subdural hematoma after burr hole drainage aids in predicting short-term outcomes and guiding clinical treatment decisions. Further external validation is needed in the future to confirm its effectiveness.

Based on Medicine, The Now and Future of Large Language Models.

This review explores the potential applications of large language models (LLMs) such as ChatGPT, GPT-3.5, and GPT-4 in the medical field, aiming to encourage their prudent use, provide professional support, and develop accessible medical AI tools that adhere to healthcare standards. This paper examines the impact of technologies such as OpenAI's Generative Pre-trained Transformers (GPT) series, including GPT-3.5 and GPT-4, and other large language models (LLMs) in medical education, scientific research, clinical practice, and nursing. Specifically, it includes supporting curriculum design, acting as personalized learning assistants, creating standardized simulated patient scenarios in education; assisting with writing papers, data analysis, and optimizing experimental designs in scientific research; aiding in medical imaging analysis, decision-making, patient education, and communication in clinical practice; and reducing repetitive tasks, promoting personalized care and self-care, providing psychological support, and enhancing management efficiency in nursing. LLMs, including ChatGPT, have demonstrated significant potential and effectiveness in the aforementioned areas, yet their deployment in healthcare settings is fraught with ethical complexities, potential lack of empathy, and risks of biased responses. Despite these challenges, significant medical advancements can be expected through the proper use of LLMs and appropriate policy guidance. Future research should focus on overcoming these barriers to ensure the effective and ethical application of LLMs in the medical field.

Impact of regression modeling on the assessment and harmonization of a point-of-care analyzer and commercial laboratory analyzer for feline plasma biochemistry testing.

Regression describes the relationship of results from two analyzers, and the generated equation can be used to harmonize results. Point-of-care (POC) analyzers cannot be calibrated by the end user, so regression offers an opportunity for calculated harmonization. Harmonization (uniformity) of laboratory results facilitates the use of common reference intervals and medical decision thresholds. Our aims were to characterize the relationship of results for multiple biochemistry analytes on a POC and a commercial laboratory analyzer (CL) with three regression techniques and to use regression equations to harmonize the POC results with those of the CL. Harmonized results were assessed by recognized quality goals. We used harmonized results to assess the regression techniques. After analyzer imprecision assessments, paired clinical samples were assessed with one dataset to calculate regression parameters that were applied to a second dataset. Three regression techniques were performed, and each was used to harmonize the POC results with those from the CL. POC results were assessed for bias and the number of results reaching quality goals before and after harmonization. All regression techniques could be used to harmonize most analytes so that 95% of results were within ASVCP TEa guidelines. Harmonization could be further improved with alternate regression techniques or exclusions. Regression offers a means to harmonize POC and CL analyzers. Further work is needed to assess how few samples can reliably be used and to assess likely species differences. No regression technique reliably describes the relationship between methods when correlation is poor.

The critical need for expert oversight of ChatGPT: Prompt engineering for safeguarding child healthcare information.

ChatGPT and other large language models have the potential to transform the health information landscape online. However, lack of domain-specific expertise and known errors in large language models raise concerns about the widespread adoption of content generated by these tools for parents making healthcare decisions for their children. The aim of this study is to determine if health-related text generated by ChatGPT under the supervision of an expert is comparable to that generated by an expert regarding persuasiveness and credibility from the perspective of a parent. In a cross-sectional study 116 parents aged 18-65 years (M = 45.02, SD = 10.92) were asked to complete a baseline assessment of their behavioral intentions regarding pediatric healthcare topics. Subsequently, participants were asked to rate text generated by either an expert or by ChatGPT under supervision of an expert. Results indicate that prompt engineered ChatGPT is capable of impacting behavioral intentions for medication, sleep, and diet decision-making. Additionally, there was little distinction between prompt engineered ChatGPT and content experts on perceived morality, trustworthiness, expertise, accuracy, and reliance. Notably, when differences were present, prompt engineered ChatGPT was rated as higher in trustworthiness and accuracy, and participants indicated they would be more likely to rely on the information presented by prompt engineered ChatGPT compared to the expert. Given that parents will trust and rely on information generated by ChatGPT, it is critically important that human domain-specific expertise be applied to healthcare information that will ultimately be presented to consumers (e.g., parents).

Trial selection criteria should not be used for clinical decisions and recommendations: the thrombectomy trials example

BackgroundDespite multiple calls for more inclusive studies, most clinical trial eligibility criteria remain too restrictive. Thrombectomy trials have been no exception. MethodsWe review the landmark trials that have shown the benefits of thrombectomy, their eligibility criteria, and consequences on clinical practice. We discuss the rationale behind various reasons for exclusions. We also examine the logical problem involved in using eligibility criteria as indications for treatment. ResultsMost thrombectomy trials have been too restrictive. This has been shown by a plethora of follow-up studies that have refuted most of the previously recommended trial eligibility restrictions. Meanwhile, the effect of clinical recommendations based on restrictive eligibility criteria is that treatment has been denied to the majority of patients who could have benefitted. Trial eligibility criteria cannot be used to make clinical decisions or recommendations unless, like any other medical diagnosis, they have been shown capable of reliably differentiating patients into those that will, and those that will not benefit from treatment. This goal can only be achieved with all-inclusive pragmatic trials. ConclusionRestrictive eligibility criteria render clinical trials incapable of guiding medical decisions or recommendations.

Evaluation of apparent diffusion coefficient (ADC) with regards to reproducibility and diagnostic accuracy as well as possible significance of pre - and post - contrast acquisition and employment of different b values

PurposeOngoing efforts are focusing on optimizing diffusion-weighted imaging (DWI) as an essential part of breast MRI protocol. Our study aimed to evaluate the effect of contrast media (CM) on the apparent diffusion coefficients (ADC) acquired following current recommendations. Patient and Methods: Patients who underwent 3 T breast MRI with a histologically verified suspicious lesion were included in this IRB-approved, single-center, cross-sectional retrospective study. Breast MRI protocol included a DWI sequence with multiple b-values, which was acquired before and after CM administration. ADC maps were calculated by in-line monoexponential fitting with b-values 0 /800 and 50/800. Two independent readers (R1, R2) reviewed the images in separate sessions for b values 0/800 and 50/800, pre- and post-CM. Bland Altmann plots as well as intraclass correlation coefficients (ICCs) for inter-reader agreement, different b-values, and pre- and post-CM were calculated. Diagnostic accuracy was evaluated and compared by calculating the area under the receiver operating characteristics curve (AUC). Results: 91 lesions in 89 patients were examined (mean age 50.7 years, standard deviation 13.9). ADC values were significantly (P<0.05) lower post-CM (mean ranging from 1.28 x10−3 mm2/s to 1.30 x10−3 mm2/s) compared to pre-CM (mean ranging from 1.32 x10−3 mm2/s to 1.37 x10−3 mm2/s) for both b-values combinations (0/800 and 50/800 s/mm2). We found an almost perfect inter-reader agreement pre-/post-CM with b values 0/800 and 50/800 (ICC ranging from 0.853 to 0.939). Bland Altman plot demonstrated no systematic difference between readers. ROC analysis revealed good diagnostic performance without significant differences (P>0.05) between b values 0/800 and 50/800 s/mm2 as well as pre- and post-CM with areas under the ROC curve between 0.834 and 0.877. Conclusion: ADC values are slightly lower when acquiring b values 0/800 and post-CM. This effect does not reduce the diagnostic performance but may be relevant in case of definite cut-offs in medical decision making.

The conceptualisation of cardiometabolic disease policy model in the UK

BackgroundDecision models are increasingly used to inform policy-making processes, and there is a need to improve their credibility. The estimation of health and economic outcomes generated from decision models is influenced by the development process itself. This paper aims to present the conceptual model development process of cardiometabolic disease (CMD) policy models in the UK setting.MethodsThis conceptual model followed the International Society of Pharmacoeconomics and Outcomes Research-Society of Medical Decision Making (ISPOR-SMDM) Modelling Good Research Practices Task Force-2.ResultsFirst, for the conceptualisation of the problem, the CMD disease staging, progression and current clinical guidelines were summarised, followed by a systematic review of published policy models. We critically appraised policy models such as cardiovascular disease and type 2 diabetes. Key messages from the review emphasised the importance of understanding various determinants influencing model development, including risk factors, model structure, models’ parameters, data utilisation, economic perspective, equality/equity consideration, transparency and validation process. Second, as a sequential process, is model conceptualisation, to determine which modelling types and their attributes best represent the defined problem. Expert opinions, including a clinician and experienced modellers, provided input on the state transition model to ensure the structure is clinically relevant. From this stage, the consideration and agreement to establish a disease state in a state transition model was discussed.ConclusionThis conceptual model serves as a basis for representing the systematic process for structuring a CMD policy model to enhance its transparency and credibility.

"Hospice Care Could Be a Compassionate Choice": ChatGPT Responses to Questions About Decision Making in Advanced Cancer.

Background: Patients with cancer use the internet to inform medical decision making. Objective: To examine the content of ChatGPT responses to a hypothetical patient question about decision making in advanced cancer. Design: We developed a medical advice-seeking vignette in English about a patient with metastatic melanoma. When inputting this vignette, we varied five characteristics (patient age, race, ethnicity, insurance status, and preexisting recommendation of hospice/the opinion of an adult daughter regarding the recommendation). ChatGPT responses (N = 96) were coded for mentions of: hospice care, palliative care, financial implications of treatment, second opinions, clinical trials, discussing the decision with loved ones, and discussing the decision with care providers. We conducted additional analyses to understand how ChatGPT described hospice and referenced the adult daughter. Data were analyzed using descriptive statistics and chi-square analysis. Results: Responses more frequently mentioned clinical trials for vignettes describing 45-year-old patients compared with 65- and 85-year-old patients. When vignettes mentioned a preexisting recommendation for hospice, responses more frequently mentioned seeking a second opinion and hospice care. ChatGPT's descriptions of hospice focused primarily on its ability to provide comfort and support. When vignettes referenced the daughter's opinion on the hospice recommendation, approximately one third of responses also referenced this, stating the importance of talking to her about treatment preferences and values. Conclusion: ChatGPT responses to questions about advanced cancer decision making can be heterogeneous based on demographic and clinical characteristics. Findings underscore the possible impact of this heterogeneity on treatment decision making in patients with cancer.

Conflict between parents, physicians, and healthcare professionals in medical decision-making: How to address it-A systematic review from the ESPGHAN Ethics Committee.

Medical decisions about pediatric gastroenterology pathologies often involve collaboration between the medical team and the family. On occasions, conflict may arise between the individuals involved in decision making (team-family conflict) causing delays in managing a child's health condition. Little is known on the strategies that can be implemented to address such conflicts. Using the systematic review model by McCullough et al., an electronic literature search was conducted using PUBMED databases and SCOPUS. Studies published between 2001 and 2022 were analyzed to identify high-risk families, the barriers and facilitators involved in the team-family conflict and the circumstances in which healthcare professionals can be ethically justified to override parents' medical decisions and to trigger the state intervention. The present review provides recommendations on the more suitable ways to manage team-family conflict and gives a practical approach using a case vignette.

James Cameron, Critically Ill Children and the Law: Medical Decision-Making and the Best Interests Principle

James Cameron, <i>Critically Ill Children and the Law: Medical Decision-Making and the Best Interests Principle</i>

Impact of Diagnosis-Related Groups (DRG) reform on cost homogeneity of treatment for patients with malignant tumours

The cost fluctuations associated with chemotherapy, radiotherapy, and immunotherapy, as primary modalities for treating malignant tumors, are closely related to medical decision-making and impose financial burdens on patients. In response to these challenges, China has implemented the Diagnosis-Related Group (DRG) payment system to standardize costs and control expenditures. This study collected hospitalization data from patients with malignant tumors who received chemotherapy, radiotherapy, and immunotherapy at Hospital H from 2018 to 2022. The dataset was segmented into two groups: the intervention group, treated with traditional Chinese medicine (TCM) alongside standard therapies, and the control group, treated with standard therapies alone. Changes and trends in hospitalization costs under the DRG policy were analyzed using propensity-score matching (PSM), standard deviation (SD), interquartile range (IQR), and concentration index (CI). Findings showed a decreasing trend in the standard deviation of hospitalization costs across all treatment modalities. Radiotherapy exhibited the most significant decrease, with costs reducing by 2547.37 CNY in the control group and 7387.35 CNY in the intervention group. Following the DRG implementation, the concentration indexes for chemotherapy and radiotherapy increased, while those for immunotherapy did not exhibit this pattern. Costs were more concentrated in patients who did not receive TCM treatment. In summary, DRG reform positively impacted the cost homogeneity of inpatient treatments for malignant tumors, particularly in the control group not receiving TCM treatment. The effects of DRG reform varied across different treatment modalities. Although short-term fluctuations in hospitalization costs may occur, initial evidence during the study period shows the positive impact of DRG reform on cost homogeneity.

Time preference shifts in medical decision-making after serious illness

This study explores the impact of serious illnesses, such as cancer, on patients' time preferences in medical decision-making. Specifically, we assess how patients value extending their lifespan by one year under varying survival prognoses through three experimental studies. The findings reveal that patients exhibit a higher Subjective Discount Rates (SDR) in their medical decisions after a serious illness diagnosis. Notably, this difference in individual health also affects the time preferences of their family members. Additionally, the subjective contextual setting of the illness can also increase an individual's SDR levels. The research highlights a tendency for patients and families facing a potential short life expectancy to focus more on immediate concerns, leading to potentially shortsighted and irrational medical choices. This behavior often results in regret during the end-of-life stage. These insights are vital for healthcare professionals in optimizing treatment plans and for policymakers in understanding patient behaviors more comprehensively. The study emphasizes the need for considering psychological and behavioral changes in patients grappling with severe health challenges.

Parental math anxiety is associated with negative emotional activation during hypothetical health decision making

Dealing with numbers is an inherent aspect of interpreting health statistics, and negative emotions may interfere with medical decision making. One emotionally charged decision-making context is parents making medical decisions for their children. Knowing which factors–such as anxiety specific to math contexts–are associated with parents’ negative emotions during the decision-making process may inform ways to better support families as they make critical medical decisions. The current study involved secondary data analyses of an experiment with 249 parents. Participants were randomly assigned to make hypothetical health decisions for themselves, their child, or a stranger. We examined which domain-specific math (e.g., math anxiety), domain general (i.e., need for cognition), and demographic variables (e.g., parents’ health-care coverage) were associated with ratings of negative emotional activation immediately after making the decisions. Results indicated that two factors were significantly associated with parents’ ratings of negative emotional activation: (1) the person they were making decisions about (i.e., higher negative emotion activation if they were randomly assigned to make hypothetical health decisions about their child versus themselves or a stranger), and (2) parents’ ratings of their own math anxiety (i.e., parents with higher self-reported math anxiety also reported higher negative emotional activation). Future research may further consider the joint roles of emotional activation and math anxiety in how parents make health decisions for their children. Further, understanding how much math anxiety causally contributes to people’s overall negative emotional activation could lead to a more nuanced understanding of negative emotional activation in health decision making.

Heart failure in Europe: Guideline-directed medical therapy use and decision making in chronic and acute, pre-existing and de novo, heart failure with reduced, mildly reduced, and preserved ejection fraction-the ESC EORP Heart Failure III Registry.

We analysed baseline characteristics and guideline-directed medical therapy (GDMT) use and decisions in the European Society of Cardiology (ESC) Heart Failure (HF) III Registry. Between 1 November 2018 and 31 December 2020, 10 162 patients with acute HF (AHF, 39%, age 70 [62-79], 36% women) or outpatient visit for HF (61%, age 66 [58-75], 33% women), with HF with reduced (HFrEF, 57%), mildly reduced (HFmrEF, 17%) or preserved (HFpEF, 26%) ejection fraction were enrolled from 220 centres in 41 European or ESC-affiliated countries. With AHF, 97% were hospitalized, 2.2% received intravenous treatment in the emergency department, and 0.9% received intravenous treatment in an outpatient clinic. AHF was seen by most by a general cardiologist (51%) and outpatient HF most by a HF specialist (48%). A majority had been hospitalized for HF before, but 26% of AHF and 6.1% of outpatient HF had de novo HF. Baseline use, initiation and discontinuation of GDMT varied according to AHF versus outpatient HF, de novo versus pre-existing HF, and by ejection fraction. After the AHF event or outpatient HF visit, use of any renin-angiotensin system inhibitor, angiotensin receptor-neprilysin inhibitor, beta-blocker, mineralocorticoid receptor antagonist and loop diuretics was 89%, 29%, 92%, 78%, and 85% in HFrEF; 89%, 9.7%, 90%, 64%, and 81% in HFmrEF; and 77%, 3.1%, 80%, 48%, and 80% in HFpEF. Use and initiation of GDMT was high in cardiology centres in Europe, compared to previous reports from cohorts and registries including more primary care and general medicine and regions more local or outside of Europe and ESC-affiliated countries.