ABSTRACT
OBJECTIVE: This study compared the survival outcomes of non-traumatic intracerebral hemorrhage (ICH) patients with different famotidine administration routes and explored related risk factors. METHODS: Data from ICH patients between 2008-2019 were extracted from the MIMIC-IV database. Survival differences between patients with intravenous (IV) and non-intravenous (Non-IV) famotidine administration were analyzed using Cox analysis and Kaplan-Meier survival curves. RESULTS: The study included 351 patients, with 109 in the IV group and 84 in the Non-IV group after PSM. Cox analysis revealed that survival was significantly associated with age (HR = 1.031, 95%CI:1.011-1.050, p = 0.002), chloride ions (HR = 1.061, 95%CI:1.027-1.096, p < 0.001), BUN (HR = 1.034, 95%CI:1.007-1.062, p = 0.012), ICP (HR = 1.059, 95%CI:1.027-1.092, p < 0.001), RDW (HR = 1.156, 95%CI:1.030-1.299, p = 0.014), mechanical ventilation (HR = 2.526, 95%CI:1.341-4.760, p = 0.004), antibiotic use (HR = 0.331, 95%CI:0.144-0.759, p = 0.009), and Non-IV route (HR = 0.518, 95%CI:0.283-0.948, p = 0.033). Kaplan-Meier curves showed a significantly higher 30-day survival rate in the Non-IV group (p = 0.011), particularly in patients with normal ICP (HR = 0.518, 95%CI:0.283-0.948, p = 0.033). CONCLUSION: Non-IV famotidine administration significantly improves 30-day survival of ICH patients, especially for those with normal ICP, compared to IV administration.
ABSTRACT
BACKGROUND: Artificial intelligence (AI) chatbots, such as ChatGPT, have made significant progress. These chatbots, particularly popular among health care professionals and patients, are transforming patient education and disease experience with personalized information. Accurate, timely patient education is crucial for informed decision-making, especially regarding prostate-specific antigen screening and treatment options. However, the accuracy and reliability of AI chatbots' medical information must be rigorously evaluated. Studies testing ChatGPT's knowledge of prostate cancer are emerging, but there is a need for ongoing evaluation to ensure the quality and safety of information provided to patients. OBJECTIVE: This study aims to evaluate the quality, accuracy, and readability of ChatGPT-4's responses to common prostate cancer questions posed by patients. METHODS: Overall, 8 questions were formulated with an inductive approach based on information topics in peer-reviewed literature and Google Trends data. Adapted versions of the Patient Education Materials Assessment Tool for AI (PEMAT-AI), Global Quality Score, and DISCERN-AI tools were used by 4 independent reviewers to assess the quality of the AI responses. The 8 AI outputs were judged by 7 expert urologists, using an assessment framework developed to assess accuracy, safety, appropriateness, actionability, and effectiveness. The AI responses' readability was assessed using established algorithms (Flesch Reading Ease score, Gunning Fog Index, Flesch-Kincaid Grade Level, The Coleman-Liau Index, and Simple Measure of Gobbledygook [SMOG] Index). A brief tool (Reference Assessment AI [REF-AI]) was developed to analyze the references provided by AI outputs, assessing for reference hallucination, relevance, and quality of references. RESULTS: The PEMAT-AI understandability score was very good (mean 79.44%, SD 10.44%), the DISCERN-AI rating was scored as "good" quality (mean 13.88, SD 0.93), and the Global Quality Score was high (mean 4.46/5, SD 0.50). Natural Language Assessment Tool for AI had pooled mean accuracy of 3.96 (SD 0.91), safety of 4.32 (SD 0.86), appropriateness of 4.45 (SD 0.81), actionability of 4.05 (SD 1.15), and effectiveness of 4.09 (SD 0.98). The readability algorithm consensus was "difficult to read" (Flesch Reading Ease score mean 45.97, SD 8.69; Gunning Fog Index mean 14.55, SD 4.79), averaging an 11th-grade reading level, equivalent to 15- to 17-year-olds (Flesch-Kincaid Grade Level mean 12.12, SD 4.34; The Coleman-Liau Index mean 12.75, SD 1.98; SMOG Index mean 11.06, SD 3.20). REF-AI identified 2 reference hallucinations, while the majority (28/30, 93%) of references appropriately supplemented the text. Most references (26/30, 86%) were from reputable government organizations, while a handful were direct citations from scientific literature. CONCLUSIONS: Our analysis found that ChatGPT-4 provides generally good responses to common prostate cancer queries, making it a potentially valuable tool for patient education in prostate cancer care. Objective quality assessment tools indicated that the natural language processing outputs were generally reliable and appropriate, but there is room for improvement.
Subject(s)
Patient Education as Topic , Prostatic Neoplasms , Humans , Male , Patient Education as Topic/methods , Artificial IntelligenceABSTRACT
PURPOSE: Previous studies showed that long-term use of proton pump inhibitors (PPIs) was associated with cardiovascular events. However, the impact of short-term PPI exposure on intensive care unit (ICU) patients with myocardial infarction (MI) remains largely unknown. This study aims to determine the precise correlation between short-term PPI usage during hospitalization and prognostic outcomes of ICU-admitted MI patients using Medical Information Mart for Intensive Care IV database (MIMIC-IV). METHODS: Propensity score matching (PSM) was applied to adjust confounding factors. The primary study outcome was rehospitalization with mortality and length of stay as secondary outcomes. Binary logistic, multivariable Cox, and linear regression analyses were employed to estimate the impact of short-term PPI exposure on ICU-admitted MI patients. RESULTS: A total of 7249 patients were included, involving 3628 PPI users and 3621 non-PPI users. After PSM, 2687 pairs of patients were matched. The results demonstrated a significant association between PPI exposure and increased risk of rehospitalization for MI in both univariate and multivariate [odds ratio (OR) = 1.157, 95% confidence interval (CI) 1.020-1.313] analyses through logistic regression after PSM. Furthermore, this risk was also observed in patients using PPIs > 7 days, despite decreased risk of all-cause mortality among these patients. It was also found that pantoprazole increased the risk of rehospitalization, whereas omeprazole did not. CONCLUSION: Short-term PPI usage during hospitalization was still associated with higher risk of rehospitalization for MI in ICU-admitted MI patients. Furthermore, omeprazole might be superior to pantoprazole regarding the risk of rehospitalization in ICU-admitted MI patients.
ABSTRACT
BACKGROUND: Halitosis, characterized by an undesirable mouth odor, represents a common concern. OBJECTIVE: This study aims to assess the quality and readability of web-based Arabic health information on halitosis as the internet is becoming a prominent global source of medical information. METHODS: A total of 300 Arabic websites were retrieved from Google using 3 commonly used phrases for halitosis in Arabic. The quality of the websites was assessed using benchmark criteria established by the Journal of the American Medical Association, the DISCERN tool, and the presence of the Health on the Net Foundation Code of Conduct (HONcode). The assessment of readability (Flesch Reading Ease [FRE], Simple Measure of Gobbledygook, and Flesch-Kincaid Grade Level [FKGL]) was conducted using web-based readability indexes. RESULTS: A total of 127 websites were examined. Regarding quality assessment, 87.4% (n=111) of websites failed to fulfill any Journal of the American Medical Association requirements, highlighting a lack of authorship (authors' contributions), attribution (references), disclosure (sponsorship), and currency (publication date). The DISCERN tool had a mean score of 34.55 (SD 7.46), with the majority (n=72, 56.6%) rated as moderate quality, 43.3% (n=55) as having a low score, and none receiving a high DISCERN score, indicating a general inadequacy in providing quality health information to make decisions and treatment choices. No website had HONcode certification, emphasizing the concern over the credibility and trustworthiness of these resources. Regarding readability assessment, Arabic halitosis websites had high readability scores, with 90.5% (n=115) receiving an FRE score ≥80, 98.4% (n=125) receiving a Simple Measure of Gobbledygook score <7, and 67.7% (n=86) receiving an FKGL score <7. There were significant correlations between the DISCERN scores and the quantity of words (P<.001) and sentences (P<.001) on the websites. Additionally, there was a significant relationship (P<.001) between the number of sentences and FKGL and FRE scores. CONCLUSIONS: While readability was found to be very good, indicating that the information is accessible to the public, the quality of Arabic halitosis websites was poor, reflecting a significant gap in providing reliable and comprehensive health information. This highlights the need for improving the availability of high-quality materials to ensure Arabic-speaking populations have access to reliable information about halitosis and its treatment options, tying quality and availability together as critical for effective health communication.
Subject(s)
Comprehension , Halitosis , Internet , Humans , Halitosis/therapy , Consumer Health Information/standardsABSTRACT
Pharmacy decision support systems (PDSS) help clinical pharmacists to prevent and detect adverse drug events. The coding of hospital stays by the department of medical information (DMI) requires expertise, as it determines hospital revenues and the epidemiological data transmitted via the French national hospital database. The aim was to study the interest and feasibility of using a PDSS, in collaboration with the DMI, to help with the coding of hospital stays. Over 5 months, three rules were implemented in the PDSS to detect gout, Parkinson's disease and oro-pharyngeal candidiasis. The PDSS alerts were analyzed by a pharmacy resident and then forwarded to the DMI, who analyzed the stays to see whether or not the coding for the disease corresponding to the alert was present. The absence of coding was evaluated and tracked, along with the resulting change in severity and valuation. Three hundred and ninety-nine alerts from the PDSS were analyzed and sent to the DMI, representing 211 stays and 309 uniform hospital standardized discharge abstract (UHSDA) in the fields of medicine, surgery and obstetrics. Two hundred and eight (67.3%) UHSDA did not have the coding corresponding to the alert. For the majority of these UHSDAs, apart from diagnostic precision, there was no impact on the valuation of stays. For 4 UHSDAs, the addition of the diagnosis code led to an increase in the value of the stay and the severity of the homogeneous patient groups. The total revaluation corresponding to this modification was 5416. The use of PDSS has helped in the precision of diagnosis coding and the valuation of stays. This result must be weighed against the time invested in analyzing alerts and associated coding. An improvement in disease detection and data processing is needed to be feasible in practice, given the more than 227,600 RSS performed per year at our facility.
ABSTRACT
BACKGROUND: Although uncertainties exist regarding implementation, artificial intelligence-driven generative language models (GLMs) have enormous potential in medicine. Deployment of GLMs could improve patient comprehension of clinical texts and improve low health literacy. OBJECTIVE: The goal of this study is to evaluate the potential of ChatGPT-3.5 and GPT-4 to tailor the complexity of medical information to patient-specific input education level, which is crucial if it is to serve as a tool in addressing low health literacy. METHODS: Input templates related to 2 prevalent chronic diseases-type II diabetes and hypertension-were designed. Each clinical vignette was adjusted for hypothetical patient education levels to evaluate output personalization. To assess the success of a GLM (GPT-3.5 and GPT-4) in tailoring output writing, the readability of pre- and posttransformation outputs were quantified using the Flesch reading ease score (FKRE) and the Flesch-Kincaid grade level (FKGL). RESULTS: Responses (n=80) were generated using GPT-3.5 and GPT-4 across 2 clinical vignettes. For GPT-3.5, FKRE means were 57.75 (SD 4.75), 51.28 (SD 5.14), 32.28 (SD 4.52), and 28.31 (SD 5.22) for 6th grade, 8th grade, high school, and bachelor's, respectively; FKGL mean scores were 9.08 (SD 0.90), 10.27 (SD 1.06), 13.4 (SD 0.80), and 13.74 (SD 1.18). GPT-3.5 only aligned with the prespecified education levels at the bachelor's degree. Conversely, GPT-4's FKRE mean scores were 74.54 (SD 2.6), 71.25 (SD 4.96), 47.61 (SD 6.13), and 13.71 (SD 5.77), with FKGL mean scores of 6.3 (SD 0.73), 6.7 (SD 1.11), 11.09 (SD 1.26), and 17.03 (SD 1.11) for the same respective education levels. GPT-4 met the target readability for all groups except the 6th-grade FKRE average. Both GLMs produced outputs with statistically significant differences (P<.001; 8th grade P<.001; high school P<.001; bachelors P=.003; FKGL: 6th grade P=.001; 8th grade P<.001; high school P<.001; bachelors P<.001) between mean FKRE and FKGL across input education levels. CONCLUSIONS: GLMs can change the structure and readability of medical text outputs according to input-specified education. However, GLMs categorize input education designation into 3 broad tiers of output readability: easy (6th and 8th grade), medium (high school), and difficult (bachelor's degree). This is the first result to suggest that there are broader boundaries in the success of GLMs in output text simplification. Future research must establish how GLMs can reliably personalize medical texts to prespecified education levels to enable a broader impact on health care literacy.
ABSTRACT
The goal of this work is to develop a Machine Learning model to predict the need for both invasive and non-invasive mechanical ventilation in intensive care unit (ICU) patients. Using the Philips eICU Research Institute (ERI) database, 2.6 million ICU patient data from 2010 to 2019 were analyzed. This data was randomly split into training (63%), validation (27%), and test (10%) sets. Additionally, an external test set from a single hospital from the ERI database was employed to assess the model's generalizability. Model performance was determined by comparing the model probability predictions with the actual incidence of ventilation use, either invasive or non-invasive. The model demonstrated a prediction performance with an AUC of 0.921 for overall ventilation, 0.937 for invasive, and 0.827 for non-invasive. Factors such as high Glasgow Coma Scores, younger age, lower BMI, and lower PaCO2 were highlighted as indicators of a lower likelihood for the need for ventilation. The model can serve as a retrospective benchmarking tool for hospitals to assess ICU performance concerning mechanical ventilation necessity. It also enables analysis of ventilation strategy trends and risk-adjusted comparisons, with potential for future testing as a clinical decision tool for optimizing ICU ventilation management.
ABSTRACT
BACKGROUND: This study aimed to create a nomogram using the model for end-stage liver disease (MELD) that can better predict the risk of 28-day mortality in patients with bleeding esophageal varices. METHODS: Data on patients with bleeding esophageal varices were retrospectively collected from the Medical Information Mart for Intensive Care database. Variables were selected using the least absolute shrinkage and selection operator logistic regression model and were used to construct a prognostic nomogram. The nomogram was evaluated against the MELD model using various methods, including receiver operating characteristic (ROC) curve analysis, calibration plotting, net reclassification improvement (NRI), integrated discrimination improvement (IDI), and decision curve analysis (DCA). RESULTS: A total of 280 patients were included in the study. The patient's use of vasopressin and norepinephrine, respiratory rate, temperature, mean corpuscular volume, and MELD score were included in the nomogram. The area under the ROC curve, NRI, IDI, and DCA of the nomogram indicated that it performs better than the MELD alone. CONCLUSION: A nomogram was created that outperformed the MELD score in forecasting the risk of 28-day mortality in individuals with bleeding esophageal varices.
ABSTRACT
BACKGROUND: Health literacy (HL) refers to the ability to understand and process information provided by the healthcare system and depends on various factors, such as language comprehension, education, and social environment. Low HL was recently associated with increased readmission, morbidity, and mortality. Little is known about HL levels in physical trauma patients. The aim of this study was to determine general HL in physical trauma patients in an outpatient setting and to evaluate possible differences based on demographic characteristics. MATERIAL AND METHODS: A total of 100 physical trauma patients were recruited in the outpatient trauma facility of the Medical University of Vienna. All recruited patients completed the German Short Test of Functional Health Literacy (S-TOFHLA). RESULTS: The evaluated HL index ranged between 20 and 36 points (highest achievable score: 36 points), with the mean value calculated at 34.3 (adequate). Out of 100 participants, 97 patients (97%) showed adequate HL and 3 patients (3%) reached a score corresponding to a marginal understanding. No patient showed inadequate HL utilizing the STOFHLA tool. No significant differences were found between different demographic categories, including age, education level, native language, and injury location. CONCLUSION: In this study, included outpatient trauma patients demonstrated an overall adequate understanding of healthcare related information. Age, sociodemographic background, and/or educational status did not influence performance, which leads to the question as to whether the German version of the STOFHLA is valid to representatively measure HL in these patients. Furthermore, regarding the obvious shortcomings of the STOFHLA, the education standard of the respective population should be taken into consideration when choosing an appropriate testing tool.
ABSTRACT
ChatGPT is an artificial intelligence (AI) chatbot application. In this study, we explore the creation and use of a customized version of ChatGPT designed specifically for patient education, called "Lab Explainer." Lab Explainer aims to simplify and clarify the results of complex laboratory tests for patients, using the sophisticated capabilities of AI in natural language processing; it analyses various laboratory test data and provides clear explanations and contextual information. The approach involved adapting OpenAI's ChatGPT model specifically to analyze laboratory test data. The results suggest that Lab Explainer has the potential to improve understanding by providing an interpretation of laboratory tests to the patient. In conclusion, the Lab Explainer can assist patient education by providing intelligible interpretations of laboratory tests.
ABSTRACT
BACKGROUND: The integration of artificial intelligence (AI) in medicine, particularly through AI-based language models like ChatGPT, offers a promising avenue for enhancing patient education and healthcare delivery. This study aims to evaluate the quality of medical information provided by Chat Generative Pre-trained Transformer (ChatGPT) regarding common orthopedic and trauma surgical procedures, assess its limitations, and explore its potential as a supplementary source for patient education. METHODS: Using the GPT-3.5-Turbo version of ChatGPT, simulated patient information was generated for 20 orthopedic and trauma surgical procedures. The study utilized standardized information forms as a reference for evaluating ChatGPT's responses. The accuracy and quality of the provided information were assessed using a modified DISCERN instrument, and a global medical assessment was conducted to categorize the information's usefulness and reliability. RESULTS: ChatGPT mentioned an average of 47% of relevant keywords across procedures, with a variance in the mention rate between 30.5% and 68.6%. The average modified DISCERN (mDISCERN) score was 2.4 out of 5, indicating a moderate to low quality of information. None of the ChatGPT-generated fact sheets were rated as "very useful," with 45% deemed "somewhat useful," 35% "not useful," and 20% classified as "dangerous." A positive correlation was found between higher mDISCERN scores and better physician ratings, suggesting that information quality directly impacts perceived utility. CONCLUSION: While AI-based language models like ChatGPT hold significant promise for medical education and patient care, the current quality of information provided in the field of orthopedics and trauma surgery is suboptimal. Further development and refinement of AI sources and algorithms are necessary to improve the accuracy and reliability of medical information. This study underscores the need for ongoing research and development in AI applications in healthcare, emphasizing the critical role of accurate, high-quality information in patient education and informed consent processes.
ABSTRACT
Background: Serum anion gap (AG) can potentially be applied to the diagnosis of various metabolic acidosis, and a recent study has reported the association of AG with the mortality of patients with coronavirus disease 2019 (COVID-19). However, the relationship of AG with the short-term mortality of patients with ventilator-associated pneumonia (VAP) is still unclear. Herein, we aimed to investigate the association between AG and the 30-day mortality of VAP patients, and construct and assess a multivariate predictive model for the 30-day mortality risk of VAP. Methods: This retrospective cohort study extracted data of 477 patients with VAP from the Medical Information Mart for Intensive Care III (MIMIC-III) database. Data of patients were divided into a training set and a testing set with a ratio of 7:3. In the training set, variables significantly associated with the 30-day mortality of VAP patients were included in the multivariate predictive model through univariate Cox regression and stepwise regression analyses. Then, the predictive performance of the multivariate predictive model was assessed in both training set and testing set, and compared with the single AG and other scoring systems including the Sequential Organ Failure Assessment (SOFA) score, the confusion, urea, respiratory rate (RR), blood pressure, and age (≥65 years old) (CURB-65) score, and the blood urea nitrogen (BUN), altered mental status, pulse, and age (>65 years old) (BAP-65) score. In addition, the association of AG with the 30-day mortality of VAP patients was explored in subgroups of gender, age, and infection status. The evaluation indexes were hazard ratios (HRs), C-index, and 95% confidence intervals (CIs). Results: A total of 70 patients died within 30 days. The multivariate predictive model consisted of AG (HR =1.052, 95% CI: 1.008-1.098), age (HR =1.037, 95% CI: 1.019-1.055), duration of mechanical ventilation (HR =0.998, 95% CI: 0.996-0.999), and vasopressors use (HR =1.795, 95% CI: 1.066-3.023). In both training set (C-index =0.725, 95% CI: 0.670-0.780) and testing set (C-index =0.717, 95% CI: 0.637-0.797), the multivariate model had a relatively superior predictive performance to the single AG value. Moreover, the association of AG with the 30-day mortality was also found in patients who were male (HR =1.088, 95% CI: 1.029-1.150), and whatever the pathogens they infected (bacterial infection: HR =1.059, 95% CI: 1.011-1.109; fungal infection: HR =1.057, 95% CI: 1.002-1.115). Conclusions: The AG-related multivariate model had a potential predictive value for the 30-day mortality of patients with VAP. These findings may provide some references for further exploration on simple and robust predictors of the short-term mortality risk of VAP, which may further help clinicians to identify patients with high risk of mortality in an early stage in the intensive care units (ICUs).
ABSTRACT
The use of mobile-based personal health record (m-PHR) applications at the hospital level has been minimally studied. This study aimed to investigate the relationship between m-PHR use and quality of care. A cross-sectional study design was employed, analyzing data from 99 hospitals. Two data sources were utilized: a previous m-PHR investigation conducted from 26 May to 30 June 2022 and a hospital evaluation dataset on quality of care. The use of m-PHR applications was measured by the number of m-PHR application downloads. Three independent variables were assessed: quality of care in the use of antibiotic drugs, injection drugs, and polypharmacy with ≥6 drugs. A generalized linear model was used for the analysis. The hospitals providing high-quality care, as evaluated based on the rate of antibiotic prescription (relative risk [RR], 3.328; 95% confidence interval [CI], 1.840 to 6.020; p < 0.001) and polypharmacy (RR, 2.092; 95% CI, 1.027 to 4.261; p = 0.042), showed an increased number of m-PHR downloads. Among the hospital covariates, public foundation status and being part of multi-hospital systems were associated with the number of m-PHR downloads (p < 0.05). This exploratory study found a positive relationship between quality of care and m-PHR use. Hospitals providing high-quality care may also excel in various activities, including m-PHR application use.
ABSTRACT
BACKGROUND: With the global increase of cesarean deliveries, breech presentation is the third indication for elective cesarean delivery. Implementation of external cephalic version (ECV), in which the position of the baby is manipulated externally to prevent breech presentation at term, remains suboptimal. Increasing knowledge for caretakers and patients is beneficial in the uptake of ECV implementation. In recent decades, the internet has become the most important source of information for both patients and health care professionals. However, the use and availability of the internet also bring about concerns since the information is often not regulated or reviewed. Information needs to be understandable, correct, and easily obtainable for the patient. Owing to its global reach, YouTube has great potential to both hinder and support spreading medical information and can therefore be used as a tool for shared decision-making. OBJECTIVE: The objective of this study was to investigate the available information on YouTube about ECV and assess the quality and usefulness of the information in the videos. METHODS: A YouTube search was performed with five search terms and the first 35 results were selected for analysis. A quality assessment scale was developed to quantify the accuracy of medical information of each video. The main outcome measure was the usefulness score, dividing the videos into useful, slightly useful, and not useful categories. The source of upload was divided into five subcategories and two broad categories of medical or nonmedical. Secondary outcomes included audience engagement, misinformation, and encouraging or discouraging ECV. RESULTS: Among the 70 videos, only 14% (n=10) were defined as useful. Every useful video was uploaded by educational channels or health care professionals and 80% (8/10) were derived from a medical source. Over half of the not useful videos were uploaded by birth attendants and vloggers. Videos uploaded by birth attendants scored the highest on audience engagement. The presence of misinformation was low across all groups. Two-thirds of the vloggers encouraged ECV to their viewers. CONCLUSIONS: A minor percentage of videos about ECV on YouTube are considered useful. Vloggers often encourage their audience to opt for ECV. Videos with higher audience engagement had a lower usefulness score compared to videos with lower audience engagement. Sources from medically accurate videos should cooperate with sources with high audience engagement to contribute to the uptake of ECV by creating more awareness and a positive attitude of the procedure, thereby lowering the chance for a cesarean delivery due to breech presentation at term.
ABSTRACT
Background: Currently, a scarcity of prognostic research exists that concentrates on patients with nephrotic syndrome (NS) who also have tuberculosis. The purpose of this study was to assess the in-hospital mortality status of NS patients with tuberculosis, identify crucial risk factors, and create a sturdy prognostic prediction model that can improve disease evaluation and guide clinical decision-making. Methods: We utilized the Medical Information Mart for Intensive Care IV version 2.2 (MIMIC-IV v2.2) database to include 1,063 patients with NS complicated by TB infection. Confounding factors included demographics, vital signs, laboratory indicators, and comorbidities. The Least Absolute Shrinkage and Selection Operator (LASSO) regression and the diagnostic experiment the receiver operating characteristic (ROC) curve analyses were used to select determinant variables. A nomogram was established by using a logistic regression model. The performance of the nomogram was tested and validated using the concordance index (C-index) of the ROC curve, calibration curves, internal cross-validation, and clinical decision curve analysis. Results: The cumulative in-hospital mortality rate for patients with NS and TB was 18.7%. A nomogram was created to predict in-hospital mortality, utilizing Alb, Bun, INR, HR, Abp, Resp., Glu, CVD, Sepsis-3, and AKI stage 7 days. The area under the curve of the receiver operating characteristic evaluation was 0.847 (0.812-0.881), with a calibration curve slope of 1.00 (0.83-1.17) and a mean absolute error of 0.013. The cross-validated C-index was 0.860. The decision curves indicated that the patients benefited from this model when the risk threshold was 0.1 and 0.81. Conclusion: Our clinical prediction model nomogram demonstrated a good predictive ability for in-hospital mortality among patients with NS combined with TB. Therefore, it can aid clinicians in assessing the condition, judging prognosis, and making clinical decisions for such patients.
ABSTRACT
OBJECTIVE: To explore the relationship between lactate-to-albumin ratio (LAR) at ICU admission and prognosis in critically ill patients with acute kidney injury (AKI). METHODS: A retrospective analysis was conducted. Patients were divided into low (<0.659) LAR and high LAR (≥0.659) groups. Least absolute shrinkage and selection operator regression analysis was conducted to select variables associated with the 30-day prognosis. Cox regression analyses were performed to assess the association between LAR and mortality. Kaplan-Meier curves were plotted to compare cumulative survival rates between high and low LAR groups. Subgroup analysis was employed to assess the stability of the results. ROC curve was used to determine the diagnostic efficacy of LAR on prognosis. RESULTS: A nonlinear relationship was observed between LAR and the risk of 30-day and 360-day all-cause mortality in AKI patients (p < 0.001). Cox regulation showed that high LAR (≥ 0.659) was an independent risk factor for 30-day and 360-day all-cause mortality in patients with AKI (p < 0.001). The Kaplan-Meier survival curves demonstrated a noteworthy decrease in cumulative survival rates at both 30 and 360 days for the high LAR group in comparison to the low LAR group (p < 0.001). Subgroup analyses demonstrated the stability of the results. ROC curves showed that LAR had a diagnostic advantage when compared with lactate or albumin alone (p < 0.001). CONCLUSION: High LAR (≥0.659) at ICU admission was an independent risk factor for both short-term (30-day) and long-term (360-day) all-cause mortality in patients with AKI.
Subject(s)
Acute Kidney Injury , Critical Illness , Intensive Care Units , Lactic Acid , ROC Curve , Humans , Acute Kidney Injury/blood , Acute Kidney Injury/diagnosis , Acute Kidney Injury/mortality , Acute Kidney Injury/etiology , Male , Female , Retrospective Studies , Middle Aged , Prognosis , Aged , Lactic Acid/blood , Intensive Care Units/statistics & numerical data , Serum Albumin/analysis , Kaplan-Meier Estimate , Risk Factors , Biomarkers/blood , Proportional Hazards Models , Survival Rate , Adult , Clinical RelevanceABSTRACT
BACKGROUND: Artificial intelligence is increasingly being applied to many workflows. Large language models (LLMs) are publicly accessible platforms trained to understand, interact with, and produce human-readable text; their ability to deliver relevant and reliable information is also of particular interest for the health care providers and the patients. Hematopoietic stem cell transplantation (HSCT) is a complex medical field requiring extensive knowledge, background, and training to practice successfully and can be challenging for the nonspecialist audience to comprehend. OBJECTIVE: We aimed to test the applicability of 3 prominent LLMs, namely ChatGPT-3.5 (OpenAI), ChatGPT-4 (OpenAI), and Bard (Google AI), in guiding nonspecialist health care professionals and advising patients seeking information regarding HSCT. METHODS: We submitted 72 open-ended HSCT-related questions of variable difficulty to the LLMs and rated their responses based on consistency-defined as replicability of the response-response veracity, language comprehensibility, specificity to the topic, and the presence of hallucinations. We then rechallenged the 2 best performing chatbots by resubmitting the most difficult questions and prompting to respond as if communicating with either a health care professional or a patient and to provide verifiable sources of information. Responses were then rerated with the additional criterion of language appropriateness, defined as language adaptation for the intended audience. RESULTS: ChatGPT-4 outperformed both ChatGPT-3.5 and Bard in terms of response consistency (66/72, 92%; 54/72, 75%; and 63/69, 91%, respectively; P=.007), response veracity (58/66, 88%; 40/54, 74%; and 16/63, 25%, respectively; P<.001), and specificity to the topic (60/66, 91%; 43/54, 80%; and 27/63, 43%, respectively; P<.001). Both ChatGPT-4 and ChatGPT-3.5 outperformed Bard in terms of language comprehensibility (64/66, 97%; 53/54, 98%; and 52/63, 83%, respectively; P=.002). All displayed episodes of hallucinations. ChatGPT-3.5 and ChatGPT-4 were then rechallenged with a prompt to adapt their language to the audience and to provide source of information, and responses were rated. ChatGPT-3.5 showed better ability to adapt its language to nonmedical audience than ChatGPT-4 (17/21, 81% and 10/22, 46%, respectively; P=.03); however, both failed to consistently provide correct and up-to-date information resources, reporting either out-of-date materials, incorrect URLs, or unfocused references, making their output not verifiable by the reader. CONCLUSIONS: In conclusion, despite LLMs' potential capability in confronting challenging medical topics such as HSCT, the presence of mistakes and lack of clear references make them not yet appropriate for routine, unsupervised clinical use, or patient counseling. Implementation of LLMs' ability to access and to reference current and updated websites and research papers, as well as development of LLMs trained in specialized domain knowledge data sets, may offer potential solutions for their future clinical application.
Subject(s)
Health Personnel , Hematopoietic Stem Cell Transplantation , Humans , Artificial Intelligence , LanguageABSTRACT
OBJECTIVES: Given the high incidence and mortality rate of sepsis, early identification of high-risk patients and timely intervention are crucial. However, existing mortality risk prediction models still have shortcomings in terms of operation, applicability, and evaluation on long-term prognosis. This study aims to investigate the risk factors for death in patients with sepsis, and to construct the prediction model of short-term and long-term mortality risk. METHODS: Patients meeting sepsis 3.0 diagnostic criteria were selected from the Medical Information Mart for Intensive Care-IV (MIMIC-IV) database and randomly divided into a modeling group and a validation group at a ratio of 7ê3. Baseline data of patients were analyzed. Univariate Cox regression analysis and full subset regression were used to determine the risk factors of death in patients with sepsis and to screen out the variables to construct the prediction model. The time-dependent area under the curve (AUC), calibration curve, and decision curve were used to evaluate the differentiation, calibration, and clinical practicability of the model. RESULTS: A total of 14 240 patients with sepsis were included in our study. The 28-day and 1-year mortality were 21.45% (3 054 cases) and 36.50% (5 198 cases), respectively. Advanced age, female, high sepsis-related organ failure assessment (SOFA) score, high simplified acute physiology score II (SAPS II), rapid heart rate, rapid respiratory rate, septic shock, congestive heart failure, chronic obstructive pulmonary disease, liver disease, kidney disease, diabetes, malignant tumor, high white blood cell count (WBC), long prothrombin time (PT), and high serum creatinine (SCr) levels were all risk factors for sepsis death (all P<0.05). Eight variables, including PT, respiratory rate, body temperature, malignant tumor, liver disease, septic shock, SAPS II, and age were used to construct the model. The AUCs for 28-day and 1-year survival were 0.717 (95% CI 0.710 to 0.724) and 0.716 (95% CI 0.707 to 0.725), respectively. The calibration curve and decision curve showed that the model had good calibration degree and clinical application value. CONCLUSIONS: The short-term and long-term mortality risk prediction models of patients with sepsis based on the MIMIC-IV database have good recognition ability and certain clinical reference significance for prognostic risk assessment and intervention treatment of patients.
Subject(s)
Sepsis , Humans , Sepsis/mortality , Sepsis/diagnosis , Female , Male , Risk Factors , Prognosis , Databases, Factual , Risk Assessment/methods , Intensive Care Units/statistics & numerical data , Middle Aged , Area Under Curve , Aged , Organ Dysfunction Scores , Proportional Hazards ModelsABSTRACT
Introduction: In the context of the deep coupling and synergistic development of digital villages and healthy villages, the development of China's rural society harbors a huge potential for medical and healthcare consumption. Methods: On the basis of theoretical research, a framework was constructed to analyze the influence mechanism of farmers' medical and healthcare consumption in the context of Internet medical information overflow, and empirically examines the research and analysis framework by using the 2020 China Household Tracking Survey data with the OLS model, mediation effect model, and instrumental variable method. Results: It is found that Internet medical information spillover has a "crowding-in effect" on farmers' healthcare consumption; Medical attendance behavior, economic capital utilize the intermediary effect between Internet medical information spillover and farmers' healthcare consumption. And there is age group heterogeneity in the effect of Internet medical information spillover on farmers' healthcare consumption, The ability of rural middle-aged and old-aged groups to recognize new things such as Internet medical information needs to be improved, so the overflow of Internet medical information will induce rural middle-aged and old-aged groups to generate a certain amount of medical and health care consumption. However, the impact on healthcare consumption is not sensitive to the youth cohort group. Discussion: The sinking of Internet medical resources should be accelerated in the future to promote the high-quality development of rural medical and health services, at the same time the "Internet + healthcare services" should be optimized to promote scientific and rational stratification of farmers' access to healthcare, and economic capital for farmers' access to health care should be improved in order to alleviate the burden of health care, etc.