Cause of death and significant disease found at autopsy

Evaluation of ChatGPT-Generated Differential Diagnosis for Common Diseases With Atypical Presentation: Descriptive Research

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.

Videoautopsy-A Minimally Invasive Autopsy Method Using Endoscopic Techniques in Forensic Medicine: Clinical Features

In light of falling global autopsy rates, one of the causes of which is the resulting body disfigurement, it has become crucial to search for new, minimally invasive post-mortem diagnostic tools. One of these methods is videoautopsy, a minimally invasive autopsy technique using endoscopic methods. In the years 2020-2023, 15 videoautopsies were conducted at the Department of Forensic Medicine of the Poznan University of Medical Sciences in order to determine the usefulness of the method in forensic approaches. Each post-mortem examination included laparoscopy and thoracoscopy, followed by a classic autopsy to assess the effectiveness of the method. In total, the endoscopic examination allowed for determining the cause of death in 53.3% of cases, and when the cause of death was located in the abdominal cavity or chest, the percentage increased to 80%. Traumatic lesions had good recognition efficiency. In addition, it was also possible to collect material for histopathological and toxicological tests. Retroperitoneal organs were difficult to assess. The main limitation of the method is the inability to assess the inside of the skull and the structures of the central nervous system. Videoautopsy may become an important tool in post-mortem diagnostics and in forensic cases, especially when the alternative is to not perform an autopsy. Further research is necessary to standardise the examination protocol, optimise the instrumentation, and assess the potential synergistic effect with other methods of minimally and non-invasive post-mortem examination.

Importance of Patient History in Artificial Intelligence-Assisted Medical Diagnosis: Comparison Study

Background

Medical history contributes approximately 80% to a diagnosis, although physical examinations and laboratory investigations increase a physician's confidence in the medical diagnosis. The concept of artificial intelligence (AI) was first proposed more than 70 years ago. Recently, its role in various fields of medicine has grown remarkably. However, no studies have evaluated the importance of patient history in AI-assisted medical diagnosis.

Objective

This study explored the contribution of patient history to AI-assisted medical diagnoses and assessed the accuracy of ChatGPT in reaching a clinical diagnosis based on the medical history provided.

Methods

Using clinical vignettes of 30 cases identified in The BMJ, we evaluated the accuracy of diagnoses generated by ChatGPT. We compared the diagnoses made by ChatGPT based solely on medical history with the correct diagnoses. We also compared the diagnoses made by ChatGPT after incorporating additional physical examination findings and laboratory data alongside history with the correct diagnoses.

Results

ChatGPT accurately diagnosed 76.6% (23/30) of the cases with only the medical history, consistent with previous research targeting physicians. We also found that this rate was 93.3% (28/30) when additional information was included.

Conclusions

Although adding additional information improves diagnostic accuracy, patient history remains a significant factor in AI-assisted medical diagnosis. Thus, when using AI in medical diagnosis, it is crucial to include pertinent and correct patient histories for an accurate diagnosis. Our findings emphasize the continued significance of patient history in clinical diagnoses in this age and highlight the need for its integration into AI-assisted medical diagnosis systems.

Ageing-Related Neurodegeneration and Cognitive Decline

Neuropathological assessment was conducted on 1630 subjects, representing 5% of all the deceased that had been sent to the morgue of Uppsala University Hospital during a 15-year-long period. Among the 1630 subjects, 1610 were ≥41 years of age (range 41 to 102 years). Overall, hyperphosphorylated (HP) τ was observed in the brains of 98% of the 1610 subjects, and amyloid β-protein (Aβ) in the brains of 64%. The most common alteration observed was Alzheimer disease neuropathologic change (ADNC) (56%), followed by primary age-related tauopathy (PART) in 26% of the subjects. In 16% of the subjects, HPτ was limited to the locus coeruleus. In 14 subjects (<1%), no altered proteins were observed. In 3 subjects, only Aβ was observed, and in 17, HPτ was observed in a distribution other than that seen in ADNC/PART. The transactive DNA-binding protein 43 (TDP43) associated with limbic-predominant age-related TDP encephalopathy (LATE) was observed in 565 (35%) subjects and α-synuclein (αS) pathology, i.e., Lewy body disease (LBD) or multi system atrophy (MSA) was observed in the brains of 21% of the subjects. A total of 39% of subjects with ADNC, 59% of subjects with PART, and 81% of subjects with HPτ limited to the locus coeruleus lacked concomitant pathologies, i.e., LATE-NC or LBD-NC. Of the 293 (18% of the 1610 subjects) subjects with dementia, 81% exhibited a high or intermediate level of ADNC. In 84% of all individuals with dementia, various degrees of concomitant alterations were observed; i.e., MIXED-NC was a common cause of dementia. A high or intermediate level of PART was observed in 10 subjects with dementia (3%), i.e., tangle-predominant dementia. No subjects exhibited only vascular NC (VNC), but in 17 subjects, severe VNC might have contributed to cognitive decline. Age-related tau astrogliopathy (ARTAG) was observed in 37% of the 1610 subjects and in 53% of those with dementia.

Prevalence of Multiple Causes of Death Within Young and Middle-Aged People Experiencing Sudden Cardiac Arrest

BACKGROUND

Multiple causes of death are increasingly reported, particularly in older populations. Rates of multiple causes of young sudden death have not been quantified.

METHOD

The End Unexplained Cardiac Death (EndUCD) registry was utilised to identify cases of young sudden death (aged 1-50 years) referred for forensic assessment from April 2019 to April 2022. Causes of death were coded according to whether one or more underlying causes of death were identified. Patients were compared according to the number of causes of death, with significant predictors assessed using logistic regression analysis.

RESULTS

1,085 cases of sudden death were identified. 263 (24.2%) cases had more than one competing cause of their sudden death. The most common multi-causal associations identified were dual non-cardiac causes of the sudden death (n=68), cardiomyopathy with non-cardiac event (n=64) and coronary artery disease with non-cardiac cause (n=63). Multi-causal death was more common in those undergoing comprehensive autopsy examination (95.8% vs 77.6%, p<0.0001), and in the setting of higher body mass index (median 31.3 kg/m2 vs 29.9 kg/m2, p=0.01), older age (44.3 years vs 41.4 years, p<0.0001), non-ventricular cardiac arrest rhythm (93.2% vs 87.3%, p=0.009), and smoking (22.8% vs 14.2%, p=0.001). The strongest predictor of multiple pathologies was comprehensive autopsy examination compared with external inspection, full-body post-mortem computed tomography and review of ancillary documentation and investigations (odds ratio 6.49, 95% confidence interval 3.47-12.14).

CONCLUSIONS

One-quarter of young sudden deaths have more than one underlying cause, highlighting the value of comprehensive investigations including autopsy. Awareness of the complexity of young sudden death is important, along with multidisciplinary involvement to ensure all contributors to death are identified.

Waiting more than 24 hours for hip fracture surgery is associated with increased risk of adverse outcomes for sicker patients: a nationwide cohort study of 63,998 patients using the Swedish Hip Fracture Register

BACKGROUND AND PURPOSE

Waiting time to surgery is a modifiable risk factor in hip fracture surgery. However, there is no consensus regarding the acceptable duration of waiting time. We used the Swedish Hip Fracture Register RIKSHÖFT and 3 administrative registers to explore the association between time to surgery and adverse outcomes after discharge.

PATIENTS AND METHODS

63,998 patients ≥ 65 years, admitted to a hospital between January 1, 2012, and August 31, 2017 were included. Time to surgery was divided into < 12, 12-24, and > 24 hours. Diagnoses investigated were atrial fibrillation/flutter (AF), congestive heart failure (CHF), pneumonia, and "acute ischemia" (a combination of stroke/intracranial bleeding, myocardial infarction, and acute kidney injury). Crude and adjusted survival analyses were performed. Time spent in hospital following the initial hospitalization was described for the 3 groups.

RESULTS

Waiting > 24 hours was associated with an increased risk of AF (HR 1.4, 95%CI 1.2-1.6), CHF (HR 1.3, CI 1.1-1.4) and "acute ischemia" (HR 1.2, CI 1.01-1.3). However, stratifying for ASA grade revealed that these associations were present only in patients with ASA 3-4. There was no association between waiting time and pneumonia after the initial hospitalization (HR 1.1, CI 0.97-1.2), but one was found with pneumonia during hospital stay OR 1.2 (CI 1.1-1.4). Time in hospital after the initial hospitalization was similar over the waiting time groups.

CONCLUSION

The associations between waiting > 24 hours for hip fracture surgery and AF, CHF, and acute ischemia suggest that shorter waiting time may reduce adverse outcomes for the sicker patients.

When is a postmortem examination carried out? A retrospective analysis of all Swedish deaths 1999-2018

The objective of this study was to assess who is merited an autopsy in Sweden. Data from the Swedish cause of death (COD) registry over a period of 20 years was retrieved and analysed. A multinominal logistic regression analysis was performed to identify the variables that were most strongly associated with the performance of a clinical or forensic autopsy (CA/FA).A definite COD, i.e. a COD based on autopsy findings, was registered in 12.6% of all deceased during the investigated period. In the remaining cases, the COD was presumed by the clinicians. Being male, born in the Nordic region, dying in a private residence, and unnatural death were most strongly associated with the performance of CA/FA. In contrast, being female, dying from dementia, dying at a nursing home, being born outside of Europe, or living in a small city or rural area seldom led to the performance of CA/FA.The above is certainly surprising as an autopsy provides an opportunity to investigate the cause of death, validate clinical diagnoses, detect unexpected aberrations, audit health care, and provide feedback to clinicians to facilitate their continuing education.

Death as a Complication of Urologic Surgery-2 Cases Identified at Autopsy

ABSTRACT

While routine medical procedures often impose some level of risk for the patient, death after routine urologic care is rare. In this series, we present 2 cases in which a relatively healthy patient ultimately died after complications, one from a prostate biopsy and one from a total nephrectomy. In case 1, a 58-year-old male died due to a 1500-mL to 2000-mL left retroperitoneal hemorrhage that occurred during insertion of a central line for sepsis treatment that resulted from an infection after a transrectal prostate biopsy. In case 2, a patient who underwent a total nephrectomy for renal cell carcinoma expired on postoperative day 7. Autopsy revealed a 1500-mL hemoperitoneum and an "unclasped" vascular surgical clip with the likely source of origin being the surgical resection site. Performance of an autopsy is essential to ascertain the cause and manner of death after medically related deaths because autopsy results can have implications on patient care, patient safety, and quality improvement.

Integrated case-based clinical approach in understanding pathways, complexities, pitfalls and challenges in neurodegenerative disorders

INTRODUCTION

This paper presents 5 cases of neurodegenerative disorders from our tertiary care rural hospital in south India. The purpose of this paper is to generate an emerging common theme by thematic analysis of clinical data from each of these patients. A theme emerged, we identified that there was a common clinical ground in patients with movement disorders and psychiatric symptoms. From this common theme, these patients eventually went on to develop different courses of illnesses.

METHODOLOGY

Clinical analysis of a case series of 5 patients with neurodegenerative disorders attending the Medicine or Psychiatry services of our hospital.

CONCLUSION

A clear & consistent association between movement disorders and psychiatric symptoms was found. Although our data is limited, we conclude that movement disorders can be early clinical markers of organic psychopathology. However, we are aware that this association can be confounded by substance abuse, stress, sleep disruption and even therapeutic interventions, and thus these factors were accounted for and yet we conclude that movement disorders can be early clinical indictors of organic psychopathology.

Survival, causes of death and recurrence up to 3 years after stroke: A population-based study

BACKGROUND AND PURPOSE

Up-to-date population-based information about long-term survival, causes of death and recurrence after stroke is needed.

METHODS

Four hundred consecutive individuals in a population-based cohort of first-ever stroke between 2015 and 2016 in Lund, Sweden, were followed up to 3 years regarding (i) survival (Swedish Population Register); (ii) causes of death (Swedish Causes of Death Register); and (iii) stroke recurrence (interview and medical chart review). Index and recurrent ischaemic stroke cases were classified using the Trial of ORG 10172 in Acute Stroke Treatment (TOAST) and Oxfordshire Community Stroke Project; and comorbidities were classified using the Charlson Comorbidity Index. Cox regression was used to determine predictors for 3-year mortality. Survival rates were compared with three local studies over a 30-year timespan.

RESULTS

Amongst 400 first-ever stroke patients, 265 (66%) survived 3 years post-stroke. Age (hazard ratio [HR] 1.09; 95% confidence interval [CI] 1.06-1.11), stroke severity (HR 1.11; 95% CI 1.08-1.13) and comorbidities (HR 1.36; 95% CI 1.22-1.53) were independently related to 3-year mortality. Amongst index ischaemic stroke patients, survival was lowest amongst those with cardio-aortic embolism (51/91; 56%). Cerebrovascular disease (54/135; 40%) and ischaemic heart disease (25/135; 19%) were the most common causes of death. Within 3 years, 30 (8%) had recurrent stroke. Amongst patients with index ischaemic stroke, 16/29 (55%) had a different TOAST pathogenetic mechanism or hemorrhagic stroke upon recurrence. Stroke survival improved between 1983-1985 and 2015-2016 (p = 0.002), but no significant change was observed between 2001-2002 and 2015-2016 (p = 0.48).

CONCLUSIONS

Stroke survival rates are relatively high, but their improvement over recent decades may be slowing down, possibly due to the composition of the first-ever stroke population. The common occurrence of changed pathogenetic mechanisms between first-ever and recurrent stroke highlights the value of reassessment in recurrent stroke.

Frequency and Significance of Pathologic Pulmonary Findings in Postmortem Examinations-A Single Center Experience before COVID-19

Coronavirus disease 2019 (COVID-19) has shown the importance of postmortem investigation of deceased patients. For a correct interpretation of the pulmonary findings in this new era, it is, however, crucial to be familiar with pathologic pulmonary conditions observed in postmortem investigations in general. Adequate postmortem histopathological evaluation of the lungs may be affected by suboptimal gross work up, autolysis or poor fixation. Using a standardized preparation approach which consisted in instillation of 4% buffered formaldehyde through the large bronchi for proper fixation and preparing large frontal tissue sections of 1-2 cm thickness after at least 24 h fixation, we comprehensively analyzed postmortem pulmonary findings from consecutive adult autopsies of a two-year period before the occurrence of COVID-19 (2016-2017). In total, significant pathological findings were observed in 97/189 patients (51%), with 28 patients showing more than one pathologic condition. Acute pneumonia was diagnosed 33/128 times (26%), embolism 24 times (19%), primary pulmonary neoplasms 18 times (14%), organizing pneumonia and other fibrosing conditions 14 times (11%), pulmonary metastases 13 times (10%), diffuse alveolar damage 12 times (9%), severe emphysema 9 times (7%) and other pathologies, e.g., amyloidosis 5/128 times (4%). Pulmonary/cardiopulmonary disease was the cause of death in 60 patients (32%). Clinical and pathological diagnoses regarding lung findings correlated completely in 75 patients (40%). Autopsy led to confirmation of a clinically suspected pulmonary diagnosis in 57 patients (39%) and clarification of an unclear clinical lung finding in 16 patients (8%). Major discrepant findings regarding the lungs (N = 31; 16%) comprised cases with clinical suspicions that could not be confirmed or new findings not diagnosed intra vitam. A significant proportion of acute pneumonias (N = 8; 24% of all cases with this diagnosis; p = 0.011) was not diagnosed clinically. We confirmed the frequent occurrence of pulmonary pathologies in autopsies, including inflammatory and neoplastic lesions as the most frequent pathological findings. Acute pneumonia was an important cause for discrepancy between clinical and postmortem diagnostics.

Pulmonary embolism at autopsy in cancer patients

BACKGROUND

Pulmonary embolism (PE) is a potentially fatal disease, but data on the incidence of fatal PE in cancer patients are scant.

OBJECTIVE

We sought to estimate the proportion of cancer patients with PE at autopsy.

METHODS

For this retrospective cohort study, all autopsy reports of cancer patients were retrieved from PALGA: Dutch Pathology Registry and used for data extraction. The primary outcome was PE at time of autopsy, defined as any clot obstructing a pulmonary artery. The secondary outcome was venous thromboembolism, defined as the composite of thrombotic PE, deep vein thrombosis, splanchnic vein thrombosis, or internal jugular vein thrombosis.

RESULTS

A total of 9571 cancer patients were included. In 1191 (12.4%; 95% confidence interval [CI], 11.8-13.1) patients, one or more PE events were observed at autopsy, of whom 1074 (90.2%) had a thrombotic embolism, 168 (14.1%) a tumor embolism, 9 (0.8%) a septic embolism, 7 (0.6%) a fat tissue embolism, and 3 (0.3%) a bone marrow embolism. Among patients with PE for whom the cause of death was specified in the autopsy report, death was considered PE-related in 642 patients (66.7%), which was 6.7% of the total study population. Venous thromboembolism was observed in 1223 (12.8%; 95% CI, 12.1-13.5) patients.

CONCLUSION

The proportion of PE in cancer patients at autopsy is substantial. Although the study population is not representative for the total cancer population, it suggests that PE is an important disease complication in cancer patients.

Re-visioning clinical reasoning, or stepping out from the skull

THE PROBLEM

Progress in teaching and learning clinical reasoning depends upon more sophisticated modelling of the reasoning process itself. Current accounts of clinical reasoning, grounded in experimental psychology, show a bias towards situating reasoning inside the skull, further reduced to neural processes signified by imaging. Such a model is necessary but not sufficient to explain the clinical reasoning process where it fails to embrace cognition extended to the environment and social contexts.

A SOLUTION

Sufficiency for a model of clinical reasoning must include dialogues between doctor, patient, and colleagues, including the complex influences of history and culture, where artefacts and semiotics such as computers, testing, and narrative structures augment cognition. Here, 'extended' cognition is configured as an outside-in process of 'sensemaking' or 'adaptive expertise'.

THE FUTURE

Current 'predictive processing' cognition models place emphasis on anticipatory cognition, where memory is reconfigured as active reconstruction rather than recall and recognition. Such an 'ecological perception' or 'externalistic' model provides a counter to the current dominant paradigm of 'ego-logical' cognitive reasoning - the latter, again, abstracted from context and located inside the skull. New models of clinical reasoning as an open, dynamic, nonlinear, complex system are called for.

Comparison of antemortem clinical diagnosis and post-mortem findings in intensive care unit patients

Autopsy is an important quality assurance indicator and a tool to advance medical knowledge. This study aims to compare the premortem clinical and postmortem pathology findings in patients who died in the Intensive Care Unit (ICU), to analyze if there are any discrepancies between them, and to compare the results to two similar studies performed in our institution in 2004 and 2007. Between January 1, 2016, and December 31, 2018, 888 patients died in the ICU and 473 underwent post-mortem examination (PME) of whom 437 were included in the present study. Autopsies revealed discrepancies between clinical diagnosis and pathologic findings according to in 101 cases (23.1%) according to Goldman classification. Forty-eight major discrepancies (class I and class II) were identified in 44 cases and the most frequent identified discrepancies were pulmonary embolism (3/12) as class I and malignancies (13/35) as class II. They were more frequent in patients hospitalized for less than 10 days then in the group with more than 10 days of hospitalization (13.8% vs 4.5%; p = 0.002). No statistical difference has been noticed concerning age, gender, and ICU stay. We observed an increase of performed autopsies and a total discrepancy rate similar to the studies performed in the same institution in 2004 (22.5%) and 2007 (21%). In conclusion, discrepancies between clinical and PME diagnoses persist despite the medical progress. Secondly, the autopsy after a short hospital stay may reveal unexpected findings whose diagnosis is challenging even if it may be suspected by the intensivist.

Comparison of conventional autopsy with post-mortem magnetic resonance, computed tomography in determining the cause of unexplained death

Conventional autopsy is the gold standard for identifying unexplained death but due to declines in referrals, there is an emerging role for post-mortem imaging. We evaluated whether post-mortem magnetic resonance (PMMR) and computed tomography (PMCT) are inferior to conventional autopsy. Deceased individuals ≥ 2 years old with unexplained death referred for coronial investigation between October 2014 to December 2016 underwent PMCT and PMMR prior to conventional autopsy. Images were reported separately and then compared to the autopsy findings by independent and blinded investigators. Outcomes included the accuracy of imaging modalities to identify an organ system cause of death and other significant abnormalities. Sixty-nine individuals underwent post-mortem scanning and autopsy (50 males; 73%) with a median age of 61 years (IQR 50-73) and median time from death to imaging of 2 days (IQR 2-3). With autopsy, 48 (70%) had an organ system cause of death and were included in assessing primary outcome while the remaining 21 (30%) were only included in assessing secondary outcome; 12 (17%) had a non-structural cause and 9 (13%) had no identifiable cause. PMMR and PMCT identified the cause of death in 58% (28/48) of cases; 50% (24/48) for PMMR and 35% (17/48) for PMCT. The sensitivity and specificity were 57% and 57% for PMMR and 38% and 73% for PMCT. Both PMMR and PMCT identified 61% (57/94) of other significant abnormalities. Post-mortem imaging is inferior to autopsy but when reported by experienced clinicians, PMMR provides important information for cardiac and neurological deaths while PMCT is beneficial for neurological, traumatic and gastrointestinal deaths.

Can low autopsy rates be increased? Yes, we can! Should postmortem examinations in oncology be performed? Yes, we should! A postmortem analysis of oncological cases

Ever declining autopsy rates have been a concern of pathologists as well as clinicians for decades. Notably, in the field of oncology, data on autopsies and discrepancies between clinical and autoptic diagnoses are particularly scarce. In this retrospective study, we show the effect of a simple catalog of measures consisting of a different approach to obtain consent for autopsy, structured conferencing, and systematic teaching of residents, as well as a close collaboration between clinicians and pathologists on the numbers of autopsies, especially of oncological patients. Additionally, postmortem examination protocols from the years 2015 until 2019 were analyzed, regarding rates of discrepancies between clinical and autoptic causes of death in this category of patients. Autopsy numbers could be significantly increased from a minimum in 2014 (60 autopsies) to a maximum in 2018 (142 autopsies) (p < 0.0001). In the 67 autopsies of oncological cases, a high rate of 51% of major discrepancy between clinical and autoptic causes of death could be detected. In contrast to the general reported decline of autopsy rates, we present rising autopsy numbers over the past 5 years with an increasing number of oncological cases who underwent a postmortem examination. The high percentage of major discrepancies between clinical and autopsy diagnosis is in contrast to an expected decrease of major discrepancies in times of precise diagnostic methods and underlines the importance of autopsies to ensure high quality in diagnostics and therapy not only in the field of oncology.