RESUMO
Background: Predicting fall injuries can mitigate the sequelae of falls and potentially utilize medical resources effectively. This study aimed to externally validate the accuracy of the Saga Fall Injury Risk Model (SFIRM), consisting of six factors including age, sex, emergency transport, medical referral letter, Bedriddenness Rank, and history of falls, assessed upon admission. Methods: This was a two-center, prospective, observational study. We included inpatients aged 20 years or older in two hospitals, an acute and a chronic care hospital, from October 2018 to September 2019. The predictive performance of the model was evaluated by calculating the area under the curve (AUC), 95% confidence interval (CI), and shrinkage coefficient of the entire study population. The minimum sample size of this study was 2,235 cases. Results: A total of 3,549 patients, with a median age of 78 years, were included in the analysis, and men accounted for 47.9% of all the patients. Among these, 35 (0.99%) had fall injuries. The performance of the SFIRM, as measured by the AUC, was 0.721 (95% CI: 0.662-0.781). The observed fall incidence closely aligned with the predicted incidence calculated using the SFIRM, with a shrinkage coefficient of 0.867. Conclusions: The external validation of the SFIRM in this two-center, prospective study showed good discrimination and calibration. This model can be easily applied upon admission and is valuable for fall injury prediction.
Assuntos
Acidentes por Quedas , Humanos , Acidentes por Quedas/estatística & dados numéricos , Masculino , Feminino , Idoso , Estudos Prospectivos , Pessoa de Meia-Idade , Medição de Risco/estatística & dados numéricos , Medição de Risco/métodos , Idoso de 80 Anos ou mais , Adulto , Fatores de Risco , Ferimentos e Lesões/epidemiologia , Incidência , Adulto JovemRESUMO
BACKGROUND Many hospitalized aged patients in Japan, the most super-aged society, are unable to be discharged home. This study was performed to clarify the factors associated with home discharge, not to a long-term care (LTC) facility or another hospital, among inpatients aged ≥75 years. MATERIAL AND METHODS A single-center prospective cohort study was performed for inpatients aged ≥75 years in a rural acute-care hospital in Japan, from November 2017 to October 2019. We divided the patients into 2 groups: those who resided at home or had died at home by 30 days after discharge, and others. We obtained data from medical charts and questionnaires given to patients and their caregivers. For each factor shown to be statistically significant by the univariable analysis, a multivariable analysis with adjustment was conducted. RESULTS In all, 285 patients agreed to participate. With adjustment by where the patient was admitted from, residing with other family members, cognitive function scores, and Barthel index, multivariable analysis using each factor identified as relevant by univariable analysis identified the following as associated with home discharge: being less informed about LTC insurance; cost of home-visit medical, nursing, or LTC services; shorter hospital stays; close proximity between patient and caregiver; main caregiver being female; and life expectancy of over 6 months (P<0.05). CONCLUSIONS Male gender and a long distance between the caregiver and patient's home significantly hindered home discharge in patients aged ≥75 years, suggesting the need to provide information regarding home-visit services under Japan's LTC insurance system for such caregivers.
Assuntos
Cuidadores , Alta do Paciente , Humanos , Masculino , Feminino , Cuidadores/psicologia , Japão , Estudos Prospectivos , HospitaisRESUMO
BACKGROUND While several predictive models for falls have been reported such as we reported in 2020, those for fall "injury" have been unreported. This study was designed to develop a model to predict fall injuries in adult inpatients using simple predictors available immediately after hospitalization. MATERIAL AND METHODS This was a single-center, retrospective cohort study. We enrolled inpatients aged ≥20 years admitted to an acute care hospital from April 2012 to March 2018. The variables routinely obtained in clinical practice were compared between the patients with fall injury and the patients without fall itself or fall injury. Multivariable analysis was performed using covariables available on admission. A predictive model was constructed using only variables showing significant association in prior multivariable analysis. RESULTS During hospitalization of 17 062 patients, 646 (3.8%) had falls and 113 (0.7%) had fall injuries. Multivariable analysis showed 6 variables that were significantly associated with fall injuries during hospitalization: age (P=0.001), sex (P=0.001), emergency transport (P<0.001), medical referral letter (P=0.041), history of falls (P=0.012), and abnormal bedriddenness ranks (all P≤0.001). The area under the curve of this predictive model was 0.794 and the shrinkage coefficient was 0.955 using the same data set given above. CONCLUSIONS We developed a predictive model for fall injuries during hospitalization using 6 predictors, including bedriddenness ranks from official Activities of Daily Living indicators in Japan, which were all easily available on admission. The model showed good discrimination by internal validation and promises to be a useful tool to assess the risk of fall injuries.
Assuntos
Atividades Cotidianas , Hospitalização , Adulto , Humanos , Estudos Retrospectivos , Japão , Pacientes Internados , Fatores de RiscoRESUMO
BACKGROUND In our previous single-center study, we developed an infective endocarditis (IE) prediction model among patients with undiagnosed fever (UF) based on 5 factors that can be obtained on admission: ambulance transfer, presence of cardiac murmur or pleural effusion, blood neutrophil percentage, and platelet count. The present study aimed to retrospectively evaluate the prediction model for IE in 320 patients presenting with fever at 4 university hospitals in Japan from January 2018 to December 2020. MATERIAL AND METHODS Patients aged ≥20 years admitted to 4 hospitals with I-330 (IE) or R-50-9 (UF) according to the International Statistical Classification of Diseases and Related Health Problems-10 were enrolled. More than 2 physicians at each hospital reviewed the patient diagnoses using the modified Duke criteria, allocating "definite IE" to IE group (n=119) and "non-definite IE" to UF group (n=201). Five factors on admission were analyzed by multivariate logistic regression. The discriminative ability and calibration of the model were evaluated using the area under the curve (AUC) and the shrinkage coefficient, respectively. RESULTS A total of 320 patients were enrolled. The odds ratios (95% confidence intervals) were as follows: ambulance transfer 1.81 (0.91-3.55); cardiac murmur 13.13 (6.69-27.36); pleural effusion 2.34 (0.62-2.42); blood neutrophil percentage 1.09 (1.06-1.14); and platelet count 0.96 (0.93-0.99). The AUC was 0.783 (0.732-0.834) with a shrinkage coefficient of 0.961. CONCLUSIONS The IE prediction model is useful to estimate the probability of IE immediately after admission for fever in patients aged ≥20 years.
Assuntos
Endocardite Bacteriana , Endocardite , Humanos , Estudos Retrospectivos , Japão/epidemiologia , Endocardite Bacteriana/diagnóstico , Endocardite/complicações , Febre , Hospitais UniversitáriosRESUMO
BACKGROUND In 2019, we developed a predictive formula of in-hospital mortality for inpatients aged ≥65 years transported by ambulance for endogenous diseases. In this study, we aimed to validate this previously developed predictive formula. MATERIAL AND METHODS In this single-center prospective observational study, we enrolled all patients aged ≥65 years who were transported by ambulance and admitted to an acute care hospital in Japan for endogenous diseases. We calculated the score according to our developed formula using age, disturbance of consciousness, the shock index, and amount of oxygen administered, with each item scoring 1 point and then totaling them. We subsequently evaluated the in-hospital mortality rate, stratum-specific likelihood ratio, and area under the receiver operating characteristic curve (AUC) of the formula, using in-hospital mortality as the primary outcome. RESULTS In total, 325 patients were included in this study. Forty-two patients died during hospitalization. Multivariable logistic regression analysis (forced-entry method) revealed that disturbance of consciousness and oxygen administration 5 L/min or more were significantly associated with mortality during hospitalization. In contrast, aged ≥90 years and shock index of 1 or higher were not significant. The mortality and stratum-specific likelihood ratios for scores in the predictive formula of 3 and 4 were 40.9% and 4.66, and 66.7% and 13.48, respectively. The AUC of the predictive formula for in-hospital mortality was 0.670 (95% confidence interval: 0.574-0.767). CONCLUSIONS This study showed that our predictive formula, consisting of factors available immediately after ambulance transport in older patients, is feasible with sufficient discrimination ability and reliability.
Assuntos
Ambulâncias , Oxigênio , Humanos , Idoso , Mortalidade Hospitalar , Japão , Reprodutibilidade dos TestesRESUMO
BACKGROUND: Several reliable predictive models for falls have been reported, but are too complicated and time-consuming to evaluate. We recently developed a new predictive model using just eight easily-available parameters including the official Japanese activities of daily living scale, Bedriddenness ranks, from the Ministry of Health, Labour and Welfare. This model has not yet been prospectively validated. This study aims to prospectively validate our new predictive model for falls among inpatients admitted to two different hospitals. METHODS: A double-centered prospective cohort study was performed from October 1, 2018, to September 30, 2019 in an acute care hospital and a chronic care hospital. We analyzed data from all adult inpatients, for whom all data required by the predictive model were evaluated and recorded. The eight items required by the predictive model were age, gender, emergency admission, department of admission, use of hypnotic medications, previous falls, independence of eating, and Bedriddenness ranks. The main outcome is in-hospital falls among adult inpatients, and the model was assessed by area under the curve. RESULTS: A total of 3,551 adult participants were available, who experienced 125 falls (3.5%). The median age (interquartile range) was 78 (66-87) years, 1,701 (47.9%) were men, and the incidence of falls was 2.25 per 1,000 patient-days and 2.06 per 1,000 occupied bed days. The area under the curve of the model was 0.793 (95% confidence interval: 0.761-0.825). The cutoff value was set as - 2.18, making the specificity 90% with the positive predictive value and negative predictive value at 11.4% and 97%. CONCLUSIONS: This double-centered prospective cohort external validation study showed that the new predictive model had excellent validity for falls among inpatients. This reliable and easy-to-use model is therefore recommended for prediction of falls among inpatients, to improve preventive interventions. TRIAL REGISTRATION: UMIN000040103 (2020/04/08).
Assuntos
Acidentes por Quedas , Atividades Cotidianas , Acidentes por Quedas/prevenção & controle , Idoso , Feminino , Humanos , Pacientes Internados , Japão/epidemiologia , Masculino , Estudos Prospectivos , Medição de RiscoRESUMO
BACKGROUND: With no approved treatments in Japan for the prevention of hereditary angioedema (HAE) attacks, there is a significant unmet need for long-term prophylactic therapies for Japanese patients with HAE. Berotralstat (BCX7353) is an oral, once-daily, highly selective inhibitor of plasma kallikrein in development for prophylaxis of angioedema attacks in HAE patients. METHODS: APeX-J is a phase 3, randomized, double-blind, placebo-controlled, parallel-group, 3-part trial conducted in Japan (University Hospital Medical Information Network identifier, UMIN000034869; ClinicalTrials.gov identifier, NCT03873116). Patients with a clinical diagnosis of type 1 or 2 HAE underwent a prospective run-in period of 56 days to determine eligibility, allowing enrollment of those with ≥2 expert-confirmed angioedema attacks. Patients were randomly assigned (1:1:1) and stratified by baseline attack rate (≥2 vs. <2 expert-confirmed attacks/month between screening and randomization) to receive once-daily berotralstat 110 mg, berotralstat 150 mg, or placebo. The primary endpoint was the rate of expert-confirmed angioedema attacks during dosing in the 24-week treatment period. RESULTS: Nineteen patients were randomized to receive once-daily berotralstat 110 mg (n = 6), berotralstat 150 mg (n = 7), or placebo (n = 6). Treatment with berotralstat 150 mg significantly reduced HAE attacks relative to placebo (1.11 vs. 2.18 attacks/month, p = .003). The most frequently reported treatment-emergent adverse events (TEAEs) in berotralstat-treated patients (n = 13) were nasopharyngitis (n = 4, 31%), abdominal pain, cough, diarrhea, and pyrexia (n = 2 each, 15%). CONCLUSIONS: Orally administered, once-daily berotralstat 150 mg significantly reduced the frequency of HAE attacks and was safe and well tolerated, supporting its use as a prophylactic therapy in patients with type 1 or 2 HAE in Japan.
Assuntos
Angioedemas Hereditários , Angioedemas Hereditários/diagnóstico , Angioedemas Hereditários/tratamento farmacológico , Angioedemas Hereditários/epidemiologia , Proteína Inibidora do Complemento C1/efeitos adversos , Humanos , Japão/epidemiologia , Estudos Prospectivos , PirazóisRESUMO
BACKGROUND: The statistical validities of the official Japanese classifications of activities of daily living (ADLs), including bedriddenness ranks (BR) and cognitive function scores (CFS), have yet to be assessed. To this aim, we evaluated the ability of BR and CFS to assess ADLs using inter-rater reliability and criterion-related validity. METHODS: New inpatients aged ≥75 years were enrolled in this hospital-based prospective observational study. BR and CFS were assessed once by an attending nurse, and then by a social worker/medical clerk. We evaluated inter-rater reliability between different professions by calculating the concordance rate, kappa coefficient, Cronbach's α, and intraclass correlation coefficient. We also estimated the relationship of the Barthel Index and Katz Index with the BR and CFS using Spearman's correlation coefficients. RESULTS: For the 271 patients enrolled, BR at the first assessment revealed 66 normal, 10 of J1, 15 of J2, 18 of A1, 31 of A2, 37 of B1, 35 of B2, 22 of C1, and 32 of C2. The concordance rate between the two BR assessments was 68.6%, with a kappa coefficient of 0.61, Cronbach's α of 0.91, and an intraclass correlation coefficient of 0.83, thus showing good inter-rater reliability. BR was negatively correlated with the Barthel Index (r = - 0.848, p < 0.001) and Katz Index (r = - 0.820, p < 0.001), showing justifiable criterion-related validity. Meanwhile, CFS at the first assessment revealed 92 normal, 47 of 1, 19 of 2a, 30 of 2b, 60 of 3a, 8 of 3b, 8 of 4, and 0 of M. The concordance rate between the two CFS assessments was 70.1%, with a kappa coefficient of 0.62, Cronbach's α of 0.87, and an intraclass correlation coefficient 0.78, thus also showing good inter-rater reliability. CFS was negatively correlated with the Barthel Index (r = - 0.667, p < 0.001) and Katz Index (r = - 0.661, p < 0.001), showing justifiable criterion-related validity. CONCLUSIONS: BR and CFS could be reliable and easy-to-use grading scales of ADLs in acute clinical practice or large-scale screening, with high inter-rater reliabilities among different professions and significant correlations with well-established, though complicated to use, instruments to assess ADLs. TRIAL REGISTRATION: UMIN000041051 (2020/7/10).
Assuntos
Atividades Cotidianas , Cognição , Idoso , Hospitais , Humanos , Japão , Reprodutibilidade dos TestesRESUMO
Key Clinical Message: In patients with symptoms of viral infection and marked thickening of the gallbladder wall, it is important to suspect acalculous cholecystitis due to Epstein-Barr virus-induced infectious mononucleosis. Abstract: A 35-year-old Japanese man presented with fever, abdominal right upper quadrant pain, and liver dysfunction. Positive immunoglobulin M and -G antibodies and negative nuclear antigen for Epstein-Barr virus were observed. Abdominal ultrasonography revealed a markedly thickened gallbladder wall. Acalculous cholecystitis due to Epstein-Barr virus-induced infectious mononucleosis was diagnosed.
RESUMO
Key Clinical Message: When seeing patients who present with atypical lymphocytes and abdominal pain without accompanying symptoms of pharyngitis or lymphadenopathy, acalculous cholecystitis caused by CMV infection should be considered as a differential diagnosis. Abstract: A teenage man presented with a fever and epigastric pain. The patient tested positive for cytomegalovirus IgG and IgM. Abdominal ultrasonography and contrast-enhanced CT revealed hepatosplenomegaly and gallbladder wall thickening. MRI did not identify gallstones or tumorous lesions. He was diagnosed with infectious mononucleosis and acalculous cholecystitis caused by cytomegalovirus.
RESUMO
The Model Core Curriculum for Medical Education in Japan was revised in 2022. It aimed to reflect changes in the nature of medical care, including the advancement of medical technology through the use of information science and technology and artificial intelligence in the Society 5.0 era. We summarize recommendations for good practice regarding learning strategies from the perspective of general medicine.
RESUMO
BACKGROUND Several factors have been reported as possible predictors of intestinal necrosis in patients with portal venous gas (PVG). We describe potential indicators of intestinal necrosis in PVG identified by contrasting 3 episodes of PVG in a patient on hemodialysis against previously verified factors. CASE REPORT An 82-year-old woman undergoing hemodialysis was admitted to our hospital thrice for acute abdominal pain. On first admission, she was alert, with a body temperature of 36.3°C, blood pressure (BP) of 125/53 mmHg, pulse rate of 60/min, respiratory rate of 18/min, and 100% oxygen saturation on room air. Computed tomography (CT) revealed PVG, intestinal distension, poor bowel wall enhancement, bubble-like pneumatosis in the intestinal wall, and minimal ascites. PVG caused by intestinal ischemia was diagnosed, and she recovered after bowel rest and hydration. Three months later, she had a second episode of abdominal pain. BP was 115/56 mmHg. CT revealed PVG and a slight accumulation of ascites, without pneumatosis in the intestinal wall. She again recovered after conservative measures. Ten months later, the patient experienced a third episode of abdominal pain, with BP of 107/52 mmHg. CT imaging indicated PVG, considerable ascites, and linear pneumatosis of the intestinal walls. Despite receiving conservative treatment, the patient died. CONCLUSIONS A large accumulation of ascites and linear pneumatosis in the intestinal walls could be potential indicators of intestinal necrosis in patients with PVG caused by intestinal ischemia. As previously reported, hypotension was further confirmed to be a reliable predictor of intestinal necrosis.
Assuntos
Enteropatias , Isquemia Mesentérica , Pneumatose Cistoide Intestinal , Lesões do Sistema Vascular , Feminino , Humanos , Idoso de 80 Anos ou mais , Ascite/complicações , Veia Porta , Pneumatose Cistoide Intestinal/diagnóstico por imagem , Pneumatose Cistoide Intestinal/terapia , Dor Abdominal/etiologia , Oxigênio , Lesões do Sistema Vascular/complicações , Diálise Renal/efeitos adversos , Isquemia/complicações , NecroseRESUMO
Purpose: There has been growing interest in generalists in Japan in recent years. However, due to the diverse use of the term "generalist", the specific roles of these physicians remain ambiguous. Consequently, the target population for research on generalists is unclear, making it challenging to conduct studies within the generalist practice framework. Therefore, a literature search was conducted to examine how generalists are defined and classified in research worldwide. Methods: We conducted a literature search that focused exclusively on articles written in English and used keywords related to generalists, general medicine (GM), primary care, and family medicine. Based on the results, six physicians working in GM reviewed the findings and discussed the identified issues and their potential solutions. Results: The definition of generalists in studies targeting GM, family medicine, and primary care conducted worldwide, including Japan, varies. Generalists exhibit diverse roles even within university hospitals in Japan. No studies provide a precise categorization or definition of generalists based on specific medical practices or roles, except for hospitalists, who are primarily involved in inpatient management in the United States. Conclusion: The definition of GM was unclear based on the results of the literature search, and the lack of uniformity in backgrounds has rendered the target population unclear. Consequently, in healthcare settings where medical systems vary by country or region, evidence from studies targeting generalists cannot readily apply to actual practice. Clarifying generalists through an explicit definition based on clinical practice will allow for a more precise target population for research on generalists and enable the accumulation of evidence related to well-defined groups of generalists, contributing to the advancement of GM. Therefore, future research is required to develop new indicators to precisely classify and define generalists.
RESUMO
[This corrects the article DOI: 10.1371/journal.pone.0236130.].
RESUMO
INTRODUCTION: The danger of diagnostic errors exists in daily medical practice, and doctors are required to avoid such errors as much as possible. Although various factors, including cognitive, system-related, and patient-related factors, are involved in the occurrence of diagnostic errors, the percentage of doctors with insufficient medical knowledge among those factors is extremely low. Therefore, lectures on diagnostic errors might also be useful for medical students without experience working as doctors. This study investigated whether a 60-minute lecture on diagnostic errors would enable Japanese medical students to consider the factors involved in diagnostic errors and how their perceptions of diagnostic errors change. METHODS AND MATERIALS: This single-center interventional study was conducted in October 2022 among fourth-year medical students at the Faculty of Medicine, Saga University. A questionnaire survey was conducted before and immediately after the lecture to investigate changes in the perceptions of medical students regarding diagnostic errors. One mock case question was given on an exam the day after the lecture, and the number of responses to cognitive biases and system-related and patient-related factors involved in diagnostic errors were calculated. RESULTS: A total of 83 students were analyzed. After the lecture, medical students were significantly more aware of the existence of the concept of diagnostic error, the importance of learning about it, their willingness to continue learning about it, and their perception that learning about diagnostic errors improves their clinical skills. They were also significantly less likely to feel blame or shame over diagnostic errors. The mean numbers of responses per student for cognitive bias, system-related factors, and patient-related factors were 1.9, 3.4, and 0.9, respectively. The mean number of responses per student for all factors was 5.6. CONCLUSION: A 60-minute lecture on diagnostic errors among medical students is beneficial because it significantly changes their perception of diagnostic errors. The results of the present study also suggest that lectures may enable Japanese medical students to consider the factors involved in diagnostic errors.
RESUMO
Hospital Medicine in the United States has achieved significant progress in the accumulation of evidence. This development has influenced the increasing societal demand for General Medicine in Japan. Generalists in Japan actively engage in a wide range of interdisciplinary clinical practices, education, and management. Furthermore, Generalists have also contributed to advances in research. However, there is limited evidence regarding the benefits of General Medicine in Japan in all these areas, with most of the evidence derived from single-center studies. In Japan, the roles of Generalists are diverse, and the comprehensive definition of General Medicine makes it difficult to clearly delineate its scope. This results in an inadequate accumulation of evidence regarding the benefits of General Medicine, potentially making it less attractive to the public and younger physicians. Therefore, it is necessary to categorize General Medicine and collect clear evidence regarding its benefits.
RESUMO
Purpose: There has been no large-scale investigation into the association between the use of lemborexant, suvorexant, and ramelteon and falls in a large population. This study, serving as a pilot investigation, was aimed at examining the relationship between inpatient falls and various prescribed hypnotic medications at admission. Patients and Methods: This study was a sub-analysis of a multicenter retrospective observational study conducted over a period of 3 years. The target population comprised patients aged 20 years or above admitted to eight hospitals, including chronic care, acute care, and tertiary hospitals. We extracted data on the types of hypnotic medications prescribed at admission, including lemborexant, suvorexant, ramelteon, benzodiazepines, Z-drugs, and other hypnotics; the occurrence of inpatient falls during the hospital stay; and patients' background information. To determine the outcome of inpatient falls, items with low collinearity were selected and included as covariates in a forced-entry binary logistic regression analysis. Results: Overall, 150,278 patients were included in the analysis, among whom 3,458 experienced falls. The median age of the entire cohort was 70 years, with men constituting 53.1%. Binary logistic regression analysis revealed that the prescription of lemborexant, suvorexant, and ramelteon at admission was not significantly associated with inpatient falls. Conclusion: The administration of lemborexant, suvorexant, and ramelteon at admission may not be associated with inpatient falls.
RESUMO
Purpose: We conducted a pilot study in an acute care hospital and developed the Saga Fall Risk Model 2 (SFRM2), a fall prediction model comprising eight items: Bedriddenness rank, age, sex, emergency admission, admission to the neurosurgery department, history of falls, independence of eating, and use of hypnotics. The external validation results from the two hospitals showed that the area under the curve (AUC) of SFRM2 may be lower in other facilities. This study aimed to validate the accuracy of SFRM2 using data from eight hospitals, including chronic care hospitals, and adjust the coefficients to improve the accuracy of SFRM2 and validate it. Patients and Methods: This study included all patients aged ≥20 years admitted to eight hospitals, including chronic care, acute care, and tertiary hospitals, from April 1, 2018, to March 31, 2021. In-hospital falls were used as the outcome, and the AUC and shrinkage coefficient of SFRM2 were calculated. Additionally, SFRM2.1, which was modified from the coefficients of SFRM2 using logistic regression with the eight items comprising SFRM2, was developed using two-thirds of the data randomly selected from the entire population, and its accuracy was validated using the remaining one-third portion of the data. Results: Of the 124,521 inpatients analyzed, 2,986 (2.4%) experienced falls during hospitalization. The median age of all inpatients was 71 years, and 53.2% were men. The AUC of SFRM2 was 0.687 (95% confidence interval [CI]:0.678-0.697), and the shrinkage coefficient was 0.996. SFRM2.1 was created using 81,790 patients, and its accuracy was validated using the remaining 42,731 patients. The AUC of SFRM2.1 was 0.745 (95% CI: 0.731-0.758). Conclusion: SFRM2 showed good accuracy in predicting falls even on validating in diverse populations with significantly different backgrounds. Furthermore, the accuracy can be improved by adjusting the coefficients while keeping the model's parameters fixed.
Assuntos
Hospitalização , Hospitais , Masculino , Humanos , Idoso , Feminino , Medição de Risco/métodos , Projetos Piloto , Estudos Retrospectivos , Fatores de RiscoRESUMO
Background: The persistence of diagnostic errors, despite advances in medical knowledge and diagnostics, highlights the importance of understanding atypical disease presentations and their contribution to mortality and morbidity. Artificial intelligence (AI), particularly generative pre-trained transformers like GPT-4, holds promise for improving diagnostic accuracy, but requires further exploration in handling atypical presentations. Objective: This study aimed to assess the diagnostic accuracy of ChatGPT in generating differential diagnoses for atypical presentations of common diseases, with a focus on the model's reliance on patient history during the diagnostic process. Methods: We used 25 clinical vignettes from the Journal of Generalist Medicine characterizing atypical manifestations of common diseases. Two general medicine physicians categorized the cases based on atypicality. ChatGPT was then used to generate differential diagnoses based on the clinical information provided. The concordance between AI-generated and final diagnoses was measured, with a focus on the top-ranked disease (top 1) and the top 5 differential diagnoses (top 5). Results: ChatGPT's diagnostic accuracy decreased with an increase in atypical presentation. For category 1 (C1) cases, the concordance rates were 17% (n=1) for the top 1 and 67% (n=4) for the top 5. Categories 3 (C3) and 4 (C4) showed a 0% concordance for top 1 and markedly lower rates for the top 5, indicating difficulties in handling highly atypical cases. The χ2 test revealed no significant difference in the top 1 differential diagnosis accuracy between less atypical (C1+C2) and more atypical (C3+C4) groups (χ²1=2.07; n=25; P=.13). However, a significant difference was found in the top 5 analyses, with less atypical cases showing higher accuracy (χ²1=4.01; n=25; P=.048). Conclusions: ChatGPT-4 demonstrates potential as an auxiliary tool for diagnosing typical and mildly atypical presentations of common diseases. However, its performance declines with greater atypicality. The study findings underscore the need for AI systems to encompass a broader range of linguistic capabilities, cultural understanding, and diverse clinical scenarios to improve diagnostic utility in real-world settings.