Chain-of-thought prompting is a method to make large language models generate intermediate reasoning steps when solving a complex problem. OpenAI's o1 preview and GPT-5 have been trained to create such a chain of thought internally before giving a response and have been claimed to surpass various benchmarks requiring complex reasoning. The purpose of this study was to evaluate their performance in text mining in oncology. Six hundred trials from high-impact medical journals were classified depending on whether they allowed for the inclusion of patients with localized and/or metastatic disease. GPT-4o, o1 preview, and GPT-5 at different reasoning effort settings were instructed to do the same classification based on the publications' abstracts. For predicting whether patients with localized disease were enrolled, GPT-4o and o1 preview achieved F1 scores of 0.80 (0.76-0.83) and 0.91 (0.89-0.94), respectively. For predicting whether patients with metastatic disease were enrolled, GPT-4o and o1 preview achieved F1 scores of 0.97 (0.95-0.98) and 0.99 (0.99-1.00), respectively. For GPT-5, the F1 scores for predicting the eligibility of patients with localized disease increased from 0.84 to 0.93 and 0.94 with increased reasoning effort. F1 scores for metastatic disease were 0.97, 0.99, and 0.99. o1 preview outperformed GPT-4o in extracting if people with localized and/or metastatic disease were eligible for a trial from its abstract. GPT-5 at high reasoning effort settings outperformed both GPT-4o and o1 preview, supporting the notion that reasoning models could become the new standard for text mining in medicine.

Partial nephrectomy has been advocated as the preferred surgical approach for small kidney tumors over total nephrectomy. However, partial nephrectomy is associated with increased perioperative risk. Estimating renal function after nephrectomy can facilitate personalized patient counseling, guide surgical approach, and identify patients who could benefit from perioperative interventions. Existing prediction models have several limitations including the lack of external validation or a user-friendly tool or application, and most have used traditional statistical methods. We used data from two academic medical institutions and machine learning (ML) methods to develop and externally validate renal function after nephrectomy-machine learning (RFAN-ML), a model to estimate long-term renal function after partial or total nephrectomy. Boruta feature selection was used to select four routinely available clinical features, specifically age, BMI, preoperative renal function, and nephrectomy type. In the training set of 1,932 patients, we compared six ML regression models representing a set of both ensemble and nonensemble ML algorithms and optimized for root mean squared error (RMSE). This model was evaluated in a test set of 1,995 patients, and the best performing model was selected as RFAN-ML. We compared RFAN-ML with existing renal function prediction benchmarks and found that RFAN-ML outperformed or had competitive performance with benchmarks on RMSE (16.6 [95% CI, 15.6 to 17.5]), R2, and mean absolute error. We developed and externally validated RFAN-ML, a ML model to predict renal function after nephrectomy, and have deployed our model online. RFAN-ML has the potential to improve the care and outcomes in patients with kidney tumors by informing personalized patient counseling and guiding surgical planning.

Cyclin-dependent kinase 4 and 6 inhibitors (CDKIs) are effective breast cancer therapies but pose adherence challenges because of cost, side effects, and complexity of medication schedule. We assessed the feasibility and usability of a smart label-enabled remote therapeutic monitoring (RTM) mHealth intervention for women with breast cancer prescribed a CDKI. Exploratory adjusted analyses examined factors associated with usability and CDKI adherence. Participants were recruited from a comprehensive cancer center between April and August 2024. For 3 months, participants used Tappt smart labels and web app to record CDKI doses, receive missed dose reminders, report symptoms biweekly, and complete baseline and follow-up surveys. Alerts were sent to oncology teams for nonadherence (>20% missed doses) or moderate-to-severe symptoms. Feasibility was defined as ≥70% of participants using the smart label >30 days and completing the follow-up survey. Usability was assessed using the System Usability Scale, with a benchmark score of ≥68. Linear regression was used to examine factors associated with usability and CDKI adherence. Among 168 screened, 107 were eligible and reached; 75.7% (81/107) consented; 90.1% (73/81) completed the follow-up survey, and 88.9% (72/81) used the intervention >30 days. Most participants self-identified as White (69.9%), were privately insured (72.6%), and had early-stage breast cancer (58.9%) and depression or anxiety (58.9%). The mean usability score was 75.8; participants who self-identified as Black reported 12.0 points higher usability than those who self-identified as White (P = .03). Mean CDKI adherence was 92.8%. A history of anxiety or depression was associated with an 8.6 percentage-point lower CDKI adherence rate (P = .02). A smart label-enabled RTM mHealth intervention exceeded feasibility and usability benchmarks and showed promise for supporting CDKI adherence and symptom management.

Lung cancer is a leading cause of death in Chile, where late-stage diagnoses and high mortality rates prevail. Here, we describe the development of OncovigIA, a novel digital tool powered by natural language processing that enhances the identification of potential lung cancer cases by surveilling computed tomography (CT) reports in a large public Hospital in Santiago, Chile. We combined natural language processing and large language models with state-of-the-art machine learning techniques and approaches to treat unbalanced data sets and determine the best solution to implement in OncovigIA. Focusing on key sections of the reports and using various machine learning models, including a balanced Random Forest, the tool achieved high performance with 0.90 accuracy and 0.84 F1-score on the test set. When applied to 13,326 CT chest reports from 2022, it successfully identified 377 CTs of patients with suspected lung cancer previously undetected and not managed by the multidisciplinary local lung cancer team. This study underscores the potential of artificial intelligence in early cancer detection and highlights the importance of its integration into local health care ecosystems. By promptly increasing the number of patients referred for specialized management, the tool OncovigIA offers a promising path toward improving lung cancer survival rates in Chile and beyond. Moreover, this article provides avenues for its broader implementation, extending it to other cancer types and/or health care-related texts for continuous surveillance, aiming at the early referral and treatment of cancer in low-resource settings.

Access to clinical notes enhances patient engagement and trust, and the 21st Century Cures Act enabled immediate electronic patient access in April 2021. Yet, technological advances may perpetuate disparities, which remain understudied. Understanding whether inequities in note access exist in oncology would highlight challenges around making this foundational health information available to all patients receiving ongoing, complex medical care. This study at a high-volume specialty cancer center explored disparities around clinical notes posted to patients' portal accounts from September 1, 2021, to August 31, 2022, and accessed by March 19, 2024. Logistic and Poisson regression were used to identify patient characteristics associated with note access and note opening rates. The study included 124,554 patients and 815,104 clinical notes, 43.7% of which (356,290) were accessed. Although modest differences in access rates emerged around sex, age, and marital status, larger disparities appeared for ethnicity, race, and language: Black patients (odds ratio [OR], 0.63 [95% CI, 0.60 to 0.66]; P < .001; incidence rate ratio [IRR], 0.74 [95% CI, 0.73 to 0.76]; P < .001), Hispanic patients (OR, 0.85 [95% CI, 0.80 to 0.90]; P < .001; IRR, 0.90 [95% CI, 0.89 to 0.92]; P < .001), and non-English-preferred language speakers (OR, 0.78 [95% CI, 0.72 to 0.84]; P < .001; IRR, 0.82 [95% CI, 0.80 to 0.84]; P < .001) were, respectively, 37%, 15%, and 22% less likely to open at least one note, and opened 26%, 10%, and 18% percent fewer notes, compared with white, non-Hispanic, and English-preferred patients, respectively. This analysis highlighted disparities, by race, ethnicity, and language, in cancer patients' accessing clinical notes. Tailored interventions are crucial to ensure that diverse groups benefit from digital health care resources.

Health insurance claims comprising diagnosis and treatment information offer insights into clinical practice and medical care costs. However, inaccurate diagnosis codes listed in claims and the absence of staging information limit the understanding of colorectal cancer (CRC)-related clinical practice. We developed and validated an algorithm to accurately identify incident CRC cases and their progression phases using claims data. We conducted a retrospective study using claims data from three Japanese institutions, including two designated cancer care hospitals (DCCHs), between April 2016 and August 2022. An algorithm that uses CRC-associated diagnostic codes and claim codes for CRC-specific treatments was developed to identify incident CRC cases and classify patients into three progression phases (treatment-sequenced groups: endoscopic, surgical, and noncurative). The algorithm was refined using cohorts from two DCCHs in April-September 2017 and April-September 2019 to enhance performance metrics, with validity tested at these hospitals during different periods and at another hospital. The performance metrics of the algorithm included positive predictive value (PPV), sensitivity in identifying incident CRC, and accuracy in determining progression phases. The performance metrics of the algorithm were enhanced by filtering prevalent cases, selecting CRC-specific treatments, and targeting invasive CRC cases. The algorithm for identifying incident invasive CRC achieved high PPVs (91.2% [95% CI, 89.5 to 92.7] and 94.4% [95% CI, 87.6 to 97.6]), sensitivities (94.6% [95% CI, 93.1 to 95.7] and 100.0% [95% CI, 95.7 to 100.0]), and progression phase accuracies (91.5% [95% CI, 89.7 to 93.0] and 97.6% [95% CI, 91.8 to 99.4]) in two validation cohorts. The developed algorithm accurately identified incident invasive CRC cases and determined their progression phases using claims data. Application of this algorithm could contribute to research on real-world practices and medical care costs associated with CRC.

PURPOSECancer and its treatment cause symptoms. In this study, we aimed to develop machine learning (ML) systems that predict future symptom deterioration among people receiving treatment for cancer and then validate the systems in a simulated deployment across an entire health care system.METHODSWe trained and tested ML systems that predict a deterioration in nine patient-reported symptoms within 30 days after treatments for aerodigestive cancers, using internal electronic health record (EHR) data at Princess Margaret Cancer Centre (3,229 patients; 20,267 treatments). The primary performance metric was the area under the receiver operating characteristic curve (AUROC). The best-performing systems in the held-out internal test set were then externally validated across 82 cancer centers in Ontario (12,079 patients; 77,003 treatments) by adapting techniques from meta-analysis.RESULTSThe best ML systems predicted symptom deterioration with AUROCs ranging from 0.66 (95% CI, 0.63 to 0.69) for dyspnea to 0.73 (95% CI, 0.71 to 0.75) for drowsiness in the internal test cohort. Treatments flagged as high-risk were significantly associated with future symptom deterioration (odds ratios [ORs], 2.53-6.56; all P < .001) and emergency department visits for dyspnea (OR, 1.85; P = .008), depression (OR, 1.84; P = .04), and anxiety (OR, 2.66; P < .001). In the external validation cohort, the AUROCs for different symptoms meta-analyzed across centers ranged from 0.67 (95% CI, 0.66 to 0.68) to 0.73 (95% CI, 0.72 to 0.74). Performance across centers displayed significant heterogeneity for six of nine symptoms (I2, 46.4%-66.9%; P = .004 for dyspnea, P < .001 for the rest).CONCLUSIONML can predict future symptoms among people with cancer from routine EHR data, which could guide personalized interventions. Heterogeneous performance across centers must be considered when systems are deployed across a health care system.