Citations

Abstract Background A vast amount of potentially useful information such as description of patient symptoms, family, and social history is recorded as free-text notes in electronic health records (EHRs) but is difficult to reliably extract at scale, limiting their utility in research. This study aims to assess whether an “out of the box” implementation of open-source large language models (LLMs) without any fine-tuning can accurately extract social determinants of health (SDoH) data from free-text clinical notes. Methods We conducted a cross-sectional study using EHR data from the Mass General Brigham (MGB) system, analyzing free-text notes for SDoH information. We selected a random sample of 200 patients and manually labeled nine SDoH aspects. Eight advanced open-source LLMs were evaluated against a baseline pattern-matching model. Two human reviewers provided the manual labels, achieving 93% inter-annotator agreement. LLM performance was assessed using accuracy metrics for overall, mentioned, and non-mentioned SDoH, and macro F1 scores. Results LLMs outperformed the baseline pattern-matching approach, particularly for explicitly mentioned SDoH, achieving up to 40% higher Accuracymentioned. openchat_3.5 was the best-performing model, surpassing the baseline in overall accuracy across all nine SDoH aspects. The refined pipeline with prompt engineering reduced hallucinations and improved accuracy. Conclusions Open-source LLMs are effective and scalable tools for extracting SDoH from unstructured EHRs, surpassing traditional pattern-matching methods. Further refinement and domain-specific training could enhance their utility in clinical research and predictive analytics, improving healthcare outcomes and addressing health disparities.

Abstract Background: Pneumonitis caused by radiotherapy for lung cancer may be missed, since it often occurs only several months later. In a previous trial including patients of any age, a symptom-based scoring system was tested to facilitate the correct diagnosis of radiation pneumonitis. Since elderly lung cancer patients have a greater risk of developing this complication, a separate scoring system particularly for this age group appears reasonable. This trial investigates such a specific tool for elderly patients irradiated for lung cancer. Methods: Patients with lung cancer patients aged ≥65 years will be included in this prospective multi-center trial and complete paper-based questionnaires (symptom-based scoring system). Patients indicate and rate symptoms potentially caused by pneumonitis, namely cough, dyspnea, and fever, once a week during and up to 24 weeks following their radiotherapy course. If symptomatic (grade ≥2) pneumonitis is confirmed by corresponding diagnostic procedures, patients receive prednisolone whenever possible. The total score of the symptom-based scoring system, which ranges between 0 and 9 points, is correlated to pneumonitis (yes or no). The discriminative power of the scoring system is evaluated by calculating the area under the receiver operating characteristic curve (AUC). Optimality is defined as a cut-off score with sensitivity ≥90% and specificity ≥80%. In addition, the Youden index (sensitivity + specificity – 1) will be applied. Fifty-nine patients (18 with and 41 without pneumonitis) are required in the full analysis set. Assuming that 5% of the patients will not qualify for this set, 62 patients should be enrolled. In addition, patient satisfaction with the scoring system is evaluated. If the dissatisfaction rate is >20%, the score needs modifications; if the dissatisfaction rate is >40%, the scoring system is considered not useful. Discussion: If the optimal cut-off score facilitates the diagnosis of radiation pneumonitis and its discrimination from other lung diseases in elderly patients with lung cancer, it will contribute to the development of a mobile application for identification of radiation pneumonitis, which can be used by the patients at home. Trial registration: clinicaltrials.gov (NCT06480734; URL: https://clinicaltrials.gov/show/NCT06480734); protocol version 1.2, registered on 27th of June, 2024.

Abstract Alzheimer's disease and related dementias (ADRD) significantly impact patients and their caregivers, causing emotional stress, lack of training, and financial strain for the latter. Addressing the need for effective support, this research developed and validated ADQueryAid, a conversational AI system designed to empower ADRD caregivers. Built on a Large Language Model (LLM) and enriched with authoritative ADRD information through a knowledge graph, ADQueryAid uses Retrieval Augmented Generation (RAG) techniques to retrieve relevant information from both structured and unstructured data sources. Prompt engineering and conversation policies ensure the delivery of informative, empathetic, and personalized responses. Evaluated via a Blind Within-Subjects Design study with 20 ADRD caregivers interacting with both ADQueryAid and a baseline model (ChatGPT 3.5) in fictional caregiving scenarios, the system's usability was assessed using the Chatbot Usability Questionnaire (CUQ). Results showed ADQueryAid significantly outperformed the baseline model across all CUQ metrics, offering more contextually relevant information, accurate guidance, and emotional support, which enhanced the caregiving experience. This study highlights the potential of AI to support ADRD caregivers by providing personalized assistance, with future research focusing on expanding the knowledge base, refining conversational strategies, and exploring the long-term impact on caregiver well-being and patient outcomes.

Abstract Background: The relationships between exercise timing and health outcomes including anthropometric indices, metabolic markers, and sleep health have not been clearly established. To gain better insights, we performed the meta-analysis to determine the effects of morning versus evening exercise on various health outcomes. Methods: We collected studies from MEDLINE, EMBASE, and Cochrane databases from inception to March 2023 to perform a meta-analysis. Exercise is defined as performing activity at least moderate intensity. For the outcomes of interest, the mean differences (MD) were employed using a random-effects model, the generic inverse variance method of DerSimonian and Laird. Results: A total of 716 patients were analyzed from 14 studies including randomized controlled trial and observational studies. We found no significant associations between exercise timing and several health-related parameters. Specifically, the data showed no notable differences in body weight (MD: -0.12 kg [95% CI −0.81 to 0.58]) and BMI (MD: -0.07 kg/m2 [95% CI −0.46 to 0.32]) when comparing morning and evening exercise sessions. Similarly, sleep latency was not significantly affected by exercise timing (MD: -2.24 minutes [95% CI −5.83 to 1.35]). Furthermore, changes in metabolic biomarkers, including LDL (MD: -6.64 mg/dL [95% CI −20.08 to 6.79]) and HDL (MD: 0.09 mg/dL [95% CI 0 to 0.17]), were not significantly associated with the timing of exercise. Conclusion: This meta-analysis found no significant relationship between exercise timing and health outcomes including anthropometric indices, metabolic markers, and sleep health. While considered as hypothesis generating, timing of exercise should equally prune patients’ health outcomes.

OBJECTIVE Surgical decompression is often indicated for symptomatic cases of cervical radiculopathy. In the cervical spine, minimally invasive posterior cervical foraminotomy (MIS-PCF) and the anterior transcorporeal approach (ATCA) are modern techniques available to surgeons. This systematic review and single-arm meta-analysis aimed to assess surgical and patient-reported outcomes of MIS-PCF and ATCA for cervical radiculopathy. METHODS A systematic review of the literature was conducted using 1) Ovid; 2) Epub Ahead of Print and In-Process, In-Data-Review & Other Non-Indexed Citations; and 3) Scopus databases, which reported outcomes following cervical decompression using MIS-PCF or the ATCA. Specifically, baseline characteristics, operative outcomes, and changes in visual analog scale (VAS) neck pain score were assessed. The quality of the studies was graded using the modified Newcastle-Ottawa Scale for observational studies. RESULTS Forty studies with 1661 patients were identified. The comparative analysis of both techniques revealed no significant differences in complication (7%, 95% CI 5%–10%, p = 0.75) or reoperation rates (5%, 95% CI 3%–7%, p = 0.41). Additionally, there were no significant differences in estimated blood loss (55.39, 95% CI 44.62–66.16 ml, p = 0.55) or operative time (85.15, 95% CI 65.38–104.92 minutes, p = 0.05). The ATCA showed significantly greater improvement (p < 0.01) in VAS neck pain scores following surgery (ATCA point reduction 6.7, 95% CI 6.0–7.5 points vs MIS-PCF 3.0, 95% CI 1.0–5.0 points). CONCLUSIONS The ATCA and MIS-PCF are effective modern techniques for the surgical treatment of radiculopathy. Both approaches showed comparable postoperative outcomes, including complication and reoperation rates. However, the ATCA was shown to provide significantly greater improvement in VAS neck pain scores.

Objective: We evaluated the impact of a methicillin-resistant Staphylococcus aureus (MRSA) nasal polymerase chain reaction (PCR) protocol on the vancomycin length of therapy (LOT) for skin and soft tissue infections (SSTIs). Design: Retrospective quasi-experimental pre- and post- MRSA nasal PCR protocol implementation study. Setting: Tertiary-care academic medical center in Jacksonville, Florida. Patients: Eligible patients received empiric vancomycin for SSTIs from January 1st to September 30th 2020 (pre-implementation group) and from January 1st to September 30th 2022 (post-implementation group). Intervention: The electronic health system software was modified to provide a best-practice advisory (BPA) prompt to the pharmacist upon order verification of vancomycin for patients with SSTIs. Methods: We reviewed patient records to determine the time from vancomycin prescription to de-escalation. The secondary outcomes were incidence of acute kidney injury (AKI), number of vancomycin levels collected, and hospital length of stay (LOS). Results: The study included 131 patients (pre-implementation, n = 86 and post-implementation, n = 45). There was no significant difference in vancomycin length of therapy (LOT) between implementation groups: mean LOT in days and standard deviation (SD) were 2.7 (1.9) and 2.6 (1.3), respectively, p-value 0.493. Of significance, in the post-implementation group, vancomycin LOT between patients with a negative and positive MRSA PCR were 2.3 (1.1) and 3.9 (1.6), p-value 0.006. There was no difference in secondary outcomes. Conclusion: The utilization of the MRSA nasal PCR to guide vancomycin de-escalation did not significantly change the vancomycin LOT, however in the post-implementation group there was a significant difference in vancomycin LOT between negative and positive MRSA PCRs.

Abstract Background This investigator-initiated phase II trial aimed to evaluate the efficacy of cabozantinib in combination with nivolumab and ipilimumab (CaboNivoIpi) in previously treated patients with radioactive iodine-refractory differentiated thyroid cancer. Methods Eligible patients with radioactive iodine-refractory differentiated thyroid cancer who progressed on 1 prior line of vascular endothelial growth factor receptor-targeted therapy received a 2-week run-in of cabozantinib monotherapy followed by CaboNivoIpi for 4 cycles (cycle length = 6 weeks), followed by cabozantinib plus nivolumab (cycle length = 4 weeks) until disease progression. The primary endpoint was objective response rate (ORR) within the first 6 months of treatment. A Simon optimal 2-stage design allowed for an interim analysis after accrual of 10 evaluable patients. At least 5 responses were needed to proceed to stage 2. Results Among 11 patients enrolled, the median age was 69 years. Prior vascular endothelial growth factor receptor-targeted therapies included lenvatinib, pazopanib, and sorafenib plus everolimus. Median follow-up was 13.4 months. Among 10 evaluable patients, ORR within the first 6 months of treatment was 10% (1 partial response). Median progression-free survival was 9 months (95% CI, 3.0–not reached) and median overall survival was 19.2 months (95% CI, 4.6–not reached). Grade 3/4 treatment-related adverse events (AEs) were noted in 55% (6/11) and grade 5 AEs in 18% (2/11) of patients. The most common treatment-related AE was hypertension. The study did not reach its prespecified efficacy threshold. Conclusion CaboNivoIpi had low ORRs and a high rate of grade ≥3 treatment-related AEs. Clinical Trial Registration NCT03914300

PURPOSE Although cure rates for childhood acute lymphoblastic leukemia (ALL) exceed 90%, ALL remains a leading cause of cancer death in children. Half of relapses arise in children initially classified with standard-risk (SR) disease. MATERIALS AND METHODS To identify genomic determinants of relapse in children with SR ALL, we performed genome and transcriptome sequencing of diagnostic and remission samples of children with SR (n = 1,381) or high-risk B-ALL with favorable cytogenetic features (n = 115) enrolled on Children's Oncology Group trials. We used a case-control study design analyzing 439 patients who relapsed and 1,057 who remained in complete remission for at least 5 years. RESULTS Genomic subtype was associated with relapse, which occurred in approximately 50% of cases of PAX5 -altered ALL (odds ratio [OR], 3.31 [95% CI, 2.17 to 5.03]; P = 3.18 × 10 –8 ). Within high-hyperdiploid ALL, gain of chromosome 10 with disomy of chromosome 7 was associated with favorable outcome (OR, 0.27 [95% CI, 0.17 to 0.42]; P = 8.02 × 10 –10 ; St Jude Children's Research Hospital validation cohort: OR, 0.22 [95% CI, 0.05 to 0.80]; P = .009), and disomy of chromosomes 10 and 17 with gain of chromosome 6 was associated with relapse (OR, 7.16 [95% CI, 2.63 to 21.51]; P = 2.19 × 10 –5 ; validation cohort: OR, 21.32 [95% CI, 3.62 to 119.30]; P = .0004). Genomic alterations were associated with relapse in a subtype-dependent manner, including alterations of INO80 in ETV6::RUNX1 ALL, IKZF1 , and CREBBP in high-hyperdiploid ALL and FHIT in BCR::ABL1 -like ALL. Genomic alterations were also associated with the presence of minimal residual disease, including NRAS and CREBBP in high-hyperdiploid ALL. CONCLUSION Genetic subtype, patterns of aneuploidy, and secondary genomic alterations determine risk of relapse in childhood ALL. Comprehensive genomic analysis is required for optimal risk stratification.