Assessing the clinical reasoning of ChatGPT for mechanical thrombectomy in patients with stroke.

BACKGROUND: Artificial intelligence (AI) has become a promising tool in medicine. ChatGPT, a large language model AI Chatbot, shows promise in supporting clinical practice. We assess the potential of ChatGPT as a clinical reasoning tool for mechanical thrombectomy in patients with stroke. METHODS: An internal validation of the abilities of ChatGPT was first performed using artificially created patient scenarios before assessment of real patient scenarios from the medical center's stroke database. All patients with large vessel occlusions who underwent mechanical thrombectomy at Tulane Medical Center between January 1, 2022 and December 31, 2022 were included in the study. The performance of ChatGPT in evaluating which patients should undergo mechanical thrombectomy was compared with the decisions made by board-certified stroke neurologists and neurointerventionalists. The interpretation skills, clinical reasoning, and accuracy of ChatGPT were analyzed. RESULTS: 102 patients with large vessel occlusions underwent mechanical thrombectomy. ChatGPT agreed with the physician's decision whether or not to pursue thrombectomy in 54.3% of the cases. ChatGPT had mistakes in 8.8% of the cases, consisting of mathematics, logic, and misinterpretation errors. In the internal validation phase, ChatGPT was able to provide nuanced clinical reasoning and was able to perform multi-step thinking, although with an increased rate of making mistakes. CONCLUSION: ChatGPT shows promise in clinical reasoning, including the ability to factor a patient's underlying comorbidities when considering mechanical thrombectomy. However, ChatGPT is prone to errors as well and should not be relied on as a sole decision-making tool in its present form, but it has potential to assist clinicians with more efficient work flow.

Chat GPT as a Neuro-Score Calculator: Analysis of a Large Language Model's Performance on Various Neurological Exam Grading Scales.

BACKGROUND: ChatGPT is a large language model artificial intelligence chatbot that has been applied to different aspects of the medical field. Our study aims to assess the quality of chatGPT to evaluate patients based on their exams for different scores including Glasgow Coma Scale (GCS), intracranial hemorrhage score (ICH), and Hunt & Hess (H&H) classification. METHODS: We created batches of patient test cases with detailed neurological exams, totaling 20 cases and created variants of increasing complex phrasing of the test cases. Using ChatGPT, we assessed repeatability and quantified the errors, including the average error rate (AER) and magnitude of errors (AME). We repeated this process for the H&H and the ICH score using base cases. Specific prompts were created for each calculator. RESULTS: The GCS calculator on 10 base test cases had an AER/AME of 10%/0.150. The accuracy of ChatGPT decreased with increasing complexity; for example, in a variation where crucial information was missing, the AER was 45% for 20 cases. For H&H, AER/AME was 13%/0.13 and for ICH, AER/AME was 27.5%/0.325. Using a simple prompt resulted in a significantly higher error rate of 70%. CONCLUSIONS: ChatGPT demonstrates ability in this proof-of-concept experiment in evaluating neuroexams using established assessment scales including GCS, ICH, and H&H. However, it has limitations in accuracy and may "hallucinate" with complex or vague descriptions. Nonetheless, ChatGPT, has promising potential in medicine.

Surveying the Cost-Effectiveness of the 20 Procedures with the Largest Public Health Services Waiting Lists in Ireland: Implications for Ireland's Cost-Effectiveness Threshold.

OBJECTIVES: To survey the cost effectiveness of procedures with the largest waiting lists in the Irish public health system to inform a reconsideration of Ireland's current cost-effectiveness threshold of €45,000/quality-adjusted life-year (QALY). METHODS: Waiting list data for inpatient and day case procedures in the Irish public health system were obtained from the National Treatment Purchase Fund. The 20 interventions with the largest number of individuals waiting for inpatient and day case care were identified. The academic literature was searched to obtain cost-effectiveness estimates from Ireland and other high-income countries. Cost-effectiveness estimates from foreign studies were adjusted for differences in currency, purchasing power parity, and inflation. RESULTS: Of the top 20 waiting list procedures, 17 had incremental cost-effectiveness ratios (ICERs) lower than €45,000/QALY, 14 fell below €20,000/QALY, and 10 fell below €10,000/QALY. Only one procedure had an ICER higher than the current threshold. Two procedures had ICERs reported for different patient and indication groups that lay on either side of the threshold. CONCLUSIONS: Some cost-effective interventions that have large waiting lists may indicate resource misallocation and the threshold may be too high. An evidence-informed revision of the threshold may require a reduction to ensure it is consistent with its theoretical basis in the opportunity cost of other interventions foregone. A limitation of this study was the difficulty in matching specific procedures from waiting lists with ICER estimates from the literature. Nevertheless, our study represents a useful demonstration of a novel concept of using waiting list data to inform cost-effectiveness thresholds.

Dendritic distributions of I h channels in experimentally-derived multi-compartment models of oriens-lacunosum/moleculare (O-LM) hippocampal interneurons.

The O-LM cell type mediates feedback inhibition onto hippocampal pyramidal cells and gates information flow in the CA1. Its functions depend on the presence of voltage-gated channels (VGCs), which affect its integrative properties and response to synaptic input. Given the challenges associated with determining densities and distributions of VGCs on interneuron dendrites, we take advantage of computational modeling to consider different possibilities. In this work, we focus on hyperpolarization-activated channels (h-channels) in O-LM cells. While h-channels are known to be present in O-LM cells, it is unknown whether they are present on their dendrites. In previous work, we used ensemble modeling techniques with experimental data to obtain insights into potentially important conductance balances. We found that the best O-LM models that included uniformly distributed h-channels in the dendrites could not fully capture the "sag" response. This led us to examine activation kinetics and non-uniform distributions of h-channels in the present work. In tuning our models, we found that different kinetics and non-uniform distributions could better reproduce experimental O-LM cell responses. In contrast to CA1 pyramidal cells where higher conductance densities of h-channels occur in more distal dendrites, decreasing conductance densities of h-channels away from the soma were observed in O-LM models. Via an illustrative scenario, we showed that having dendritic h-channels clearly speeds up back-propagating action potentials in O-LM cells, unlike when h-channels are present only in the soma. Although the present results were morphology-dependent, our work shows that it should be possible to determine the distributions and characteristics of O-LM cells with recordings and morphologies from the same cell. We hypothesize that h-channels are distributed in O-LM cell dendrites and endow them with particular synaptic integration properties that shape information flow in hippocampus.