Protected Health Information Research Articles

Shareable artificial intelligence to extract cancer outcomes from electronic health records for precision oncology research

Abstract Databases that link molecular data to clinical outcomes can inform precision cancer research into novel prognostic and predictive biomarkers. However, outside of clinical trials, cancer outcomes are typically recorded only in text form within electronic health records (EHRs). Artificial intelligence (AI) models have been trained to extract outcomes from individual EHRs. However, patient privacy restrictions have historically precluded dissemination of these models beyond the centers at which they were trained. In this study, the vulnerability of text classification models trained directly on protected health information to membership inference attacks is confirmed. A teacher-student distillation approach is applied to develop shareable models for annotating outcomes from imaging reports and medical oncologist notes. ‘Teacher’ models trained on EHR data from Dana-Farber Cancer Institute (DFCI) are used to label imaging reports and discharge summaries from the Medical Information Mart for Intensive Care (MIMIC)-IV dataset. ‘Student’ models are trained to use these MIMIC documents to predict the labels assigned by teacher models and sent to Memorial Sloan Kettering (MSK) for evaluation. The student models exhibit high discrimination across outcomes in both the DFCI and MSK test sets. Leveraging private labeling of public datasets to distill publishable clinical AI models from academic centers could facilitate deployment of machine learning to accelerate precision oncology research.

Advancing immunisation information systems: a proposal for the “unvaccinated registry”

Abstract Through a systematic review of 253 studies, we identified significant advancements in IIS functionalities including real-time data access, interoperability, and life-course vaccination records, crucial for robust public health responses. Our review reveals that vaccinations are vital for public health and societal welfare; however, vaccine hesitancy is a significant barrier, driven by complex factors. Most IIS implementations predominantly utilise web-based platforms focused on children’s vaccinations with features that enhance data interoperability and public health surveillance. Despite these advancements, gaps in vaccine coverage and hesitancy mapping persist due to suboptimal data integration and system interoperability. Parallelly, we propose a comprehensive registry that includes unvaccinated individuals, aiming to address vaccine hesitancy by integrating demographic and socio-economic data. This initiative promises to enhance health equity by providing actionable insights into vaccination gaps and facilitating targeted interventions. Ethical considerations, particularly data privacy, are paramount, as well as adhering to regulations such as GDPR to ensure the protection of personal health information. The integration of IIS with a registry that includes unvaccinated populations represents a transformative approach to public health, potentially increasing vaccine coverage and effectively informing policy decisions. This dual strategy leverages not only technological advancements but also incorporates essential behavioural insights, offering a path forward in the ongoing battle against vaccine-preventable diseases. This model aligns with the Italian National Vaccination Prevention Plan and underscores proactive, evidence-based strategies to enhance vaccination uptake, serving as a cornerstone for broader application in other national contexts.

A-208 Price Transparency of Laboratory Tests Across all Licensed Hospitals in Tennessee

Abstract Background The Hospital Price Transparency Final Rule federally mandates hospitals to provide gross price, discounted cash, payer-specific negotiated charges, and de-identified minimum and maximum negotiated rates within a machine-readable file and user-friendly display of shoppable services. Since implementation, compliance has been low and price variability has continued. Methods To assess hospital compliance, barriers to pricing information, and price variability in Tennessee, all hospitals websites were queried for gross, cash, and Blue Cross Blue Shield (BCBS) prices for 8 high-frequency laboratory tests in both mandated pricing sources. Barriers including click counts, data availability, and intra-hospital price discrepancies were noted. Results Only 2.8% of the 145 hospitals were compliant with federal law. Sub-analyses of the available machine-readable files, price estimators, and shoppable services files, only 50.4%, 4.9%, and 21.4% were found to be compliant, respectively. Barriers to pricing information included requiring protected health information (55.9%), missing at least 1 pricing source (7.6%), having no pricing sources available (6.2%), and involving more than 3 clicks to access cash price in machine-readable files (54.1%) and price estimators (68.6%.) The median total number of clicks from homepage to cash price in the machine-readable file and price estimator was 4 (range: 1-7) and 10.5 (range: 6-15), respectively. Average intra-hospital discrepancy for Basic Metabolic Panel cash prices across pricing sources was $101.30 (range: $0-1012.40). Price variability across Tennessee is shown in Figure 1. Maximum to minimum price multiples ranged from 29 to 114 for gross price, 57 to 243 for cash price, and 25 to 115 for BCBS price indicating price variability. Gross and cash prices were affected by median household income of the hospital’s county. Conclusions Our study shows low compliance with price transparency mandates, inequitable pricing affected by median household income, inconsistent and inaccessible pricing, and continued price variability in Tennessee.

Is there still protected health information in openly available PowerPoint files?

IntroductionPowerPoint presentations containing radiographic images continue to be an important educational tool for healthcare professionals. When properly deidentified, these radiographic imaging files are valuable and do not pose any privacy concerns for patients. However, when these images are not properly managed, patient information can be uncovered; raising concerns about patient privacy and potential legal consequences for healthcare systems and academic institutions. This study revisited and expanded upon previous work published by Weadock et. al in an attempt to see how this issue has changed in the last 15 years. MethodsThis study consisted of 8 separate Google Web searches related to a medical imaging modality. Each PowerPoint file was manually inspected for imaging files including x rays, CT scans, MRI's, PET scans, CT and MR angiography and ultrasounds. If a PowerPoint contained an image file, it was then inspected for PHI. Full PHI included all of the following: patient's full name, medical registration number (MRN), date and a geographic indicator smaller than a state. ResultsThe first two Google searches “Magnetic Resonance Imaging filetype:ppt” and “Cardiac CT CAT Scan filetype:ppt” returned 146 search results with 7.6 % containing full PHI. “Radiology Chest X-ray filetype:ppt” resulted in accessible PHI in 40 % of the presentations with images and “Post-Operative CT Scan filetype:ppt” resulted in accessible PHI in 29 %. Magnetic Resonance Imaging filetype:pptx” returned 0 results with PHI. “Cardiac CT CAT scan filetype:pptx” returned 1 result with partial PHI.“Radiology Chest X-ray filetype:pptx” contained 4 results with partial PHI and 1 result with full PHI. “Post-Operative CT scan filetype:pptx” contained 2 presentations with partial PHI. DiscussionWhile a declining percentage of PowerPoint files containing PHI is reassuring, this follow up study demonstrated that PHI can still be easily accessed openly on the internet by anyone who has basic familiarity with Microsoft PowerPoint. Work remains to continue to educate users on how to properly protect patient information in PowerPoint files. Appropriately protecting private healthcare information is essential for patient safety and can prevent unintentional HIPAA violations. Radiologists are uniquely positioned to educate other healthcare providers on how to properly remove PHI from radiologic imaging files.

Cybersecurity in Healthcare: Securing Patient Health Information (PHI), HIPPA compliance framework and the responsibilities of healthcare providers.

Healthcare industry is major target for cyberattacks, making the protection of public health information (PHI) and Personal Identifiable Information (PII) a prime issue. In this digital era, with health organizations shifting to electronic health records and telemedicine, they are facing a major cybersecurity attack such as ransomware, phishing, and data breaches. These attacks compromise patient privacy, pose serious risks to patient safety, and hinder healthcare operations. The Health Insurance Portability and Accountability Act (HIPAA) provides a comprehensive compliance framework to protect PHI, wherein health providers shall undertake administrative, physical and technical safeguards. Despite these regulations, many healthcare providers struggle with achieving and maintaining HIPAA compliance due to limited resources, outdated technologies, and the rapidly evolving nature of cyber threats. This paper explores the HIPAA compliance framework, examining the specific responsibilities of healthcare providers to secure PHI. Key measures taken include periodic analysis of risks, establishment of encryption and access control systems, and comprehensive employee training to minimize risks of cyber-attacks. The study highlights that there is a growing need for healthcare providers to adopt proactive, adaptive cybersecurity strategies to deal with emerging threats. By following HIPAA regulations and updating security practices continuously, healthcare providers can protect the PHI, ensure regulatory compliance and safeguarding patient trust. The findings emphasize the role of adhering to regulation and innovation both in managing cyber risks for the healthcare sector.

Development of Secure Infrastructure for Advancing Generative AI Research in Healthcare at an Academic Medical Center.

The increasing interest in leveraging generative AI models in healthcare necessitates secure infrastructure at academic medical centers. Without an all-encompassing secure system, researchers may create their own insecure microprocesses, risking the exposure of protected health information (PHI) to the public internet or its inadvertent incorporation into AI model training. To address these challenges, our institution implemented a secure pathway to the Azure OpenAI Service using our own private OpenAI instance which we fully control to facilitate high-throughput, secure LLM queries. This pathway ensures data privacy while allowing researchers to harness the capabilities of LLMs for diverse healthcare applications. Our approach supports compliant, efficient, and innovative AI research in healthcare. This paper discusses the implementation, advantages, and use cases of this secure infrastructure, underscoring the critical need for centralized, secure AI solutions in academic medical environments.

CensorCheck: A Tool for Evaluating Protected Health Information Detection Systems.

Identifying protected health identifiers (PHIs) in free text is core to anonymising clinical notes for research. There is no standard metric for the quality of a PHI detection algorithm, resulting in papers reporting heterogeneous results. CensorCheck is presented, which computes specialised metrics for evaluating PHI detection taking into account the peculiarities of anonymisation tasks.

Securing Clinical Data in Honduras: Exploring Cybersecurity Regulations and the Role of Blockchain

The existence of cybersecurity regulations in Honduras aimed at safeguarding clinical data remains uncertain, leaving the populace’s health information vulnerable due to the absence of robust legal frameworks. This study sheds light on the current landscape concerning health information protection in the country, underscoring its profound impact on securing medical information systems and the realm of Internet of Things (IoT). It underscores the potential of blockchain technology to provide an innovative solution to address security challenges in this domain. Employing a systematic approach, the research delineates a timeline to elucidate the evolution of legal measures against cybercrime in Honduras, consolidating pertinent governmental documents associated with cybersecurity and the protection of medical data. Furthermore, it underscores the healthcare sector’s susceptibility to cyber threats and the primary privacy apprehensions pertaining to IoT applications in medicine. The study concludes that Honduras lacks a dedicated data protection statute, posing a significant impediment to cybersecurity regulatory efforts. This legal void impedes the effective implementation of measures to safeguard clinical information and shield it from potential cyber risks amidst the proliferation of IoT initiatives and healthcare-related ventures. Nonetheless, this circumstance may present an opportunity for the adoption of technologies such as blockchain. By enabling the establishment of tailored regulatory frameworks, blockchain can adapt to its unique attributes, thereby facilitating its integration in fortifying clinical data protection and enhancing overall cybersecurity protocols.

Clinical and biobehavioral phenotypic assessments and data harmonization for the RE-JOIN research consortium: Recommendations for common data element selection

BackgroundThe Restoring Joint Health and Function to Reduce Pain (RE-JOIN) Consortium is part of the Helping to End Addiction Long-term® (HEAL) Initiative. HEAL is an ambitious, NIH-wide initiative to speed scientific solutions to stem the national opioid public health crisis. The RE-JOIN consortium’s over-arching goal is to define how chronic joint pain-mediating neurons innervate different articular and peri-articular tissues, with a focus on the knee and temporomandibular joints (TMJ) across species employing the latest neuroscience approaches. The aim of this manuscript is to elucidate the human data gathered by the RE-JOIN consortium, as well as to expound upon its underlying rationale and the methodologies and protocols for harmonization and standardization that have been instituted by the RE-JOIN Consortium. MethodsThe consortium-wide human models working subgroup established the RE-JOIN minimal harmonized data elements that will be collected across all human studies and set the stage to develop parallel pre-clinical data collection standards. Data harmonization considerations included requirements from the HEAL program and recommendations from the consortium’s researchers and experts on informatics, knowledge management, and data curation. ResultsMultidisciplinary experts − including preclinical and clinical researchers, with both clinician-scientists- developed the RE-JOIN’s Minimal Human Data Standard with required domains and outcome measures to be collected across projects and institutions. The RE-JOIN minimal data standard will include HEAL Common Data Elements (CDEs) (e.g., standardized demographics, general pain, psychosocial and functional measures), and RE-JOIN common data elements (R-CDE) (i.e., both general and joint-specific standardized and clinically important self-reported pain and function measures, as well as pressure pain thresholds part of quantitative sensory testing). In addition, discretionary, site-specific measures will be collected by individual institutions (e.g., expanded quantitative sensory testing and gait biomechanical assessments), specific to the knee or TMJ. Research teams will submit datasets of standardized metadata to the RE-JOIN Data Coordinating Center (DCG) via a secure cloud-based central data repository and computing infrastructure for researchers to share and conduct analyses on data collected by or acquired for RE-JOIN. RE-JOIN datasets will have protected health information (PHI) removed and be publicly available on the SPARC portal and accessible through the HEAL Data Ecosystem. ConclusionData Harmonization efforts provide the multidisciplinary consortium with an opportunity to effectively collaborate across decentralized research teams, and data standardization sets the framework for efficient future analyses of RE-JOIN data collected by the consortium. The harmonized phenotypic information obtained will significantly enhance our understanding of the neurobiology of the pain-pathology relationships in humans, providing valuable insights for comparison with pre-clinical models.

FedGMMAT: Federated generalized linear mixed model association tests.

Increasing genetic and phenotypic data size is critical for understanding the genetic determinants of diseases. Evidently, establishing practical means for collaboration and data sharing among institutions is a fundamental methodological barrier for performing high-powered studies. As the sample sizes become more heterogeneous, complex statistical approaches, such as generalized linear mixed effects models, must be used to correct for the confounders that may bias results. On another front, due to the privacy concerns around Protected Health Information (PHI), genetic information is restrictively protected by sharing according to regulations such as Health Insurance Portability and Accountability Act (HIPAA). This limits data sharing among institutions and hampers efforts around executing high-powered collaborative studies. Federated approaches are promising to alleviate the issues around privacy and performance, since sensitive data never leaves the local sites. Motivated by these, we developed FedGMMAT, a federated genetic association testing tool that utilizes a federated statistical testing approach for efficient association tests that can correct for confounding fixed and additive polygenic random effects among different collaborating sites. Genetic data is never shared among collaborating sites, and the intermediate statistics are protected by encryption. Using simulated and real datasets, we demonstrate FedGMMAT can achieve the virtually same results as pooled analysis under a privacy-preserving framework with practical resource requirements.

Inductive thematic analysis of healthcare qualitative interviews using open-source large language models: How does it compare to traditional methods?

BackgroundLarge language models (LLMs) are generative artificial intelligence that have ignited much interest and discussion about their utility in clinical and research settings. Despite this interest there is sparse analysis of their use in qualitative thematic analysis comparing their current ability to that of human coding and analysis. In addition, there has been no published analysis of their use in real-world, protected health information. ObjectiveHere we fill that gap in the literature by comparing an LLM to standard human thematic analysis in real-world, semi-structured interviews of both patients and clinicians within a psychiatric setting. MethodsUsing a 70 billion parameter open-source LLM running on local hardware and advanced prompt engineering techniques, we produced themes that summarized a full corpus of interviews in minutes. Subsequently we used three different evaluation methods for quantifying similarity between themes produced by the LLM and those produced by humans. ResultsThese revealed similarities ranging from moderate to substantial (Jaccard similarity coefficients 0.44–0.69), which are promising preliminary results. ConclusionOur study demonstrates that open-source LLMs can effectively generate robust themes from qualitative data, achieving substantial similarity to human-generated themes. The validation of LLMs in thematic analysis, coupled with evaluation methodologies, highlights their potential to enhance and democratize qualitative research across diverse fields.

Study on Prosecutorial Public Interest Litigation in the Protection of Personal Health Information

As mentioned above, with the continuous development of big data in the world, there are numerous incidents of using this technology to infringe on personal health information, and it has brought personal, property and spiritual damages to the subject of the information and even his/her family members. China's Personal Information Protection Law explicitly incorporates personal information protection into the statutory scope of prosecutorial public interest litigation, the procuratorate shoulders the burden of safeguarding the public interests of the state and society, and is actively exploring the path of protection related to personal health information.

Combining Federated Machine Learning and Qualitative Methods to Investigate Novel Pediatric Asthma Subtypes: Protocol for a Mixed Methods Study.

Pediatric asthma is a heterogeneous disease; however, current characterizations of its subtypes are limited. Machine learning (ML) methods are well-suited for identifying subtypes. In particular, deep neural networks can learn patient representations by leveraging longitudinal information captured in electronic health records (EHRs) while considering future outcomes. However, the traditional approach for subtype analysis requires large amounts of EHR data, which may contain protected health information causing potential concerns regarding patient privacy. Federated learning is the key technology to address privacy concerns while preserving the accuracy and performance of ML algorithms. Federated learning could enable multisite development and implementation of ML algorithms to facilitate the translation of artificial intelligence into clinical practice. The aim of this study is to develop a research protocol for implementation of federated ML across a large clinical research network to identify and discover pediatric asthma subtypes and their progression over time. This mixed methods study uses data and clinicians from the OneFlorida+ clinical research network, which is a large regional network covering linked and longitudinal patient-level real-world data (RWD) of over 20 million patients from Florida, Georgia, and Alabama in the United States. To characterize the subtypes, we will use OneFlorida+ data from 2011 to 2023 and develop a research-grade pediatric asthma computable phenotype and clinical natural language processing pipeline to identify pediatric patients with asthma aged 2-18 years. We will then apply federated learning to characterize pediatric asthma subtypes and their temporal progression. Using the Promoting Action on Research Implementation in Health Services framework, we will conduct focus groups with practicing pediatric asthma clinicians within the OneFlorida+ network to investigate the clinical utility of the subtypes. With a user-centered design, we will create prototypes to visualize the subtypes in the EHR to best assist with the clinical management of children with asthma. OneFlorida+ data from 2011 to 2023 have been collected for 411,628 patients aged 2-18 years along with 11,156,148 clinical notes. We expect to complete the computable phenotyping within the first year of the project, followed by subtyping during the second and third years, and then will perform the focus groups and establish the user-centered design in the fourth and fifth years of the project. Pediatric asthma subtypes incorporating RWD from diverse populations could improve patient outcomes by moving the field closer to precision pediatric asthma care. Our privacy-preserving federated learning methodology and qualitative implementation work will address several challenges of applying ML to large, multicenter RWD data. DERR1-10.2196/57981.

Using publicly available, interactive epidemiological dashboards: an innovative approach to sharing data from the Rakai Community Cohort Study.

Public sharing of de-identified biomedical data promotes collaboration between researchers and accelerates the development of disease prevention and treatment strategies. However, open-access data sharing presents challenges to researchers who need to protect the privacy of study participants, ensure that data are used appropriately, and acknowledge the inputs of all involved researchers. This article presents an approach to data sharing which addresses the above challenges by using a publicly available dashboard with de-identified, aggregated participant data from a large HIV surveillance cohort. Data in this study originated from the Rakai Community Cohort Study (RCCS), which was integrated into a centralized data mart as part of a larger data management strategy for the Rakai Health Sciences Program in Uganda. These data were used to build a publicly available, protected health information (PHI)-secured visualization dashboard for general research use. Using two unique case studies, we demonstrate the capability of the dashboard to generate the following hypotheses: firstly, that HIV prevention strategies ART and circumcision have differing levels of impact depending on the marital status of investigated communities; secondly, that ART is very successful in comparison to circumcision as an interventional strategy in certain communities. The democratization of large-scale anonymized epidemiological data using public-facing dashboards has multiple benefits, including facilitated exploration of research data and increased reproducibility of research findings. By allowing the public to explore data in depth and form new hypotheses, public-facing dashboard platforms have significant potential to generate new relationships and collaborations and further scientific discovery and reproducibility.

Merlin: A Vision Language Foundation Model for 3D Computed Tomography.

Over 85 million computed tomography (CT) scans are performed annually in the US, of which approximately one quarter focus on the abdomen. Given the current shortage of both general and specialized radiologists, there is a large impetus to use artificial intelligence to alleviate the burden of interpreting these complex imaging studies while simultaneously using the images to extract novel physiological insights. Prior state-of-the-art approaches for automated medical image interpretation leverage vision language models (VLMs) that utilize both the image and the corresponding textual radiology reports. However, current medical VLMs are generally limited to 2D images and short reports. To overcome these shortcomings for abdominal CT interpretation, we introduce Merlin - a 3D VLM that leverages both structured electronic health records (EHR) and unstructured radiology reports for pretraining without requiring additional manual annotations. We train Merlin using a high-quality clinical dataset of paired CT scans (6+ million images from 15,331 CTs), EHR diagnosis codes (1.8+ million codes), and radiology reports (6+ million tokens) for training. We comprehensively evaluate Merlin on 6 task types and 752 individual tasks. The non-adapted (off-the-shelf) tasks include zero-shot findings classification (31 findings), phenotype classification (692 phenotypes), and zero-shot cross-modal retrieval (image to findings and image to impressions), while model adapted tasks include 5-year chronic disease prediction (6 diseases), radiology report generation, and 3D semantic segmentation (20 organs). We perform internal validation on a test set of 5,137 CTs, and external validation on 7,000 clinical CTs and on two public CT datasets (VerSe, TotalSegmentator). Beyond these clinically-relevant evaluations, we assess the efficacy of various network architectures and training strategies to depict that Merlin has favorable performance to existing task-specific baselines. We derive data scaling laws to empirically assess training data needs for requisite downstream task performance. Furthermore, unlike conventional VLMs that require hundreds of GPUs for training, we perform all training on a single GPU. This computationally efficient design can help democratize foundation model training, especially for health systems with compute constraints. We plan to release our trained models, code, and dataset, pending manual removal of all protected health information.

“Whispers from the Wrist”: Wearable Health Monitoring Devices and Privacy Regulations in the U.S.: The Loopholes, the Challenges, and the Opportunities

The growth of wearable technology has enabled the collection of even more personalized information on individuals. New health-related devices marketed to consumers collect health information that might not fall under the traditional category of Protected Health Information, and thus, HIPAA protections do not fully apply. Meaning, commercial wearable health devices do not fall under FDA oversight, and data not paired with a doctor–patient relationship do not fall under HIPAA privacy protection; thus, much of the gathered health-related metrics are left without regulation and open to be sold to data brokers. As such, these data can be leveraged by health insurance, law enforcement, and employers, to name a few. This manuscript explores the loopholes in current regulations and suggests a framework that categorizes wearable data and addresses challenges in data transfer. Furthermore, taking a user perspective, the suggested framework offers solutions that aim to guide users and policymakers in navigating privacy issues in wearable technology.

Study Protocol for the Artificial Intelligence-Driven Evaluation of Structural Heart Diseases Using Wearable Electrocardiogram (ID-SHD).

Portable devices capable of electrocardiogram (ECG) acquisition have the potential to enhance structural heart disease (SHD) management by enabling early detection through artificial intelligence-ECG (AI-ECG) algorithms. However, the performance of these AI algorithms for identifying SHD in a real-world screening setting is unknown. To address this gap, we aim to evaluate the validity of our wearable-adapted AI algorithm, which has been previously developed and validated for detecting SHD from single-lead portable ECGs in patients undergoing routine echocardiograms in the Yale New Haven Hospital (YNHH). This is the protocol for a cross-sectional study in the echocardiographic laboratories of YNHH. The study will enroll 585 patients referred for outpatient transthoracic echocardiogram (TTE) as part of their routine clinical care. Patients expressing interest in participating in the study will undergo a screening interview, followed by enrollment upon meeting eligibility criteria and providing informed consent. During their routine visit, patients will undergo a 1-lead ECG with two devices - one with an Apple Watch and the second with another portable 1-lead ECG device. With participant consent, these 1-lead ECG data will be linked to participant demographic and clinical data recorded in the YNHH electronic health records (EHR). The study will assess the performance of the AI-ECG algorithm in identifying SHD, including left ventricular systolic dysfunction (LVSD), valvular disease and severe left ventricular hypertrophy (LVH), by comparing the algorithm's results with data obtained from TTE, which is the established gold standard for diagnosing SHD. All patient EHR data required for assessing eligibility and conducting the AI-ECG will be accessed through secure servers approved for protected health information. Data will be maintained on secure, encrypted servers for a minimum of five years after the publication of our findings in a peer-reviewed journal, and any unanticipated adverse events or risks will be reported by the principal investigator to the Yale Institutional Review Board, which has reviewed and approved this protocol (Protocol Number: 2000035532).

Shareable artificial intelligence to extract cancer outcomes from electronic health records.

11000 Background: Clinical outcomes such as response, progression, and metastasis represent crucial data for observational cancer research, but outside of clinical trials, such outcomes are usually recorded only in unstructured notes in electronic health records (EHRs). Manual EHR annotation is too resource-intensive to scale to large datasets. Individual cancer centers have trained artificial intelligence natural language processing (AI/NLP) models to extract outcomes from their EHRs. However, due to concerns that models trained on protected health information (PHI) might encode private data, such models usually cannot be exported to other centers. Methods: EHR data from Dana-Farber Cancer Institute (DFCI) and Memorial-Sloan Kettering (MSK) collected through the AACR Project GENIE Biopharma Collaborative were used to train and evaluate Bidirectional Encoder Representations from Transformers (BERT)-based NLP models to extract outcomes from imaging reports and oncologist notes annotated with the Pathology, Radiology/Imaging, Signs/Symptoms, Medical oncologist, and bioMarkers (PRISSMM) framework. Document-level outcomes included the presence of active cancer; response; progression; and metastatic sites. ‘Teacher’ models trained on DFCI EHR data were used to label imaging reports and discharge summaries from the public MIMIC-IV dataset. ‘Student’ models trained to use MIMIC documents to predict teacher labels were transferred to MSK for evaluation. Results: Teacher models were trained at DFCI on 30,332 imaging reports for 2609 patients and 32,173 oncologist notes for 2917 patients with non-small cell lung, colorectal, breast, prostate, pancreatic, or urothelial cancer. The models were used to label 217,642 imaging reports and 141,377 discharge summaries from MIMIC for student model training. These DFCI-trained student models were evaluated at MSK, demonstrating high discrimination (AUROC > 0.90) across outcomes. Conclusions: This privacy-preserving “teacher-student” AI/NLP framework could expedite linkage of genomic data to clinical outcomes across institutions, accelerating precision cancer research. [Table: see text]

Evaluation of a summer program for minority high school students in the clinical cancer center: The Future Gen Cancer Scholars program.

e13703 Background: Individuals from minoritized communities receive less cancer screening, have higher stage at cancer diagnosis, and experience higher death rates. Helping minority high school students find a pathway to enter college and continue with advanced education and training and become leaders in cancer treatment, research and prevention may help reduce these inequities. Many public high schools in urban centers struggle in STEM education. The Rhode Island Hospital / Legorreta Cancer Center at Brown University partnered with the Papitto Opportunity Connection in Rhode Island to create the Future Gen Cancer Scholars program. Methods: High school students were invited to apply from Providence and surrounding communities. Accepted applicants became paid hospital employees and received training in professional conduct, protected health information and other skills necessary to shadow physicians in inpatient and outpatient settings. The summer curriculum included physician shadowing, radiology, pathology, student run tumor boards, seminars and community outreach. Pre/post-surveys were collected where students rated their experiences and beliefs regarding cancer careers on a 5-point Likert scale. Open-ended questions were added to evaluate interest in pursuing a career in oncology. Results: 20 high school students participated in the program’s first cohort. 70% self-identified as Hispanic/Latinx and 30% as African American, 85% were female, 90% attended a public school. Pre- and post-surveys were completed by 17 students. The 5-point Likert scale showed improvement in an understanding of oncology careers (pre, 3.9; post, 4.5), interest in pursuing a related career (pre, 3.1; post, 3.8), and recognition of available educational resources within their communities (pre, 3.9; post 4.3). Confidence in overcoming educational access barriers slightly decreased (pre-4.1; post-3.8) due to the increased awareness of the length, cost and process of medical training. Most students described the Future Gen Program as helping them identify career goals and a path forward. Conclusions: The Future Gen program helped solidify minority high school students’ desire to become leaders in cancer care and may provide a path forward to achieve this goal. We will track students’ educational and research achievements over time to measure the program’s impact. The establishment of a successful local model can provide a framework for a nationwide program.

Tennessee hospital noncompliance with price transparency legislation for 8 common laboratory tests.

The goal of this study was to assess hospital compliance with federal price transparency mandates and barriers to pricing information in Tennessee. All hospitals websites were queried for gross, cash, and BlueCross BlueShield of Tennessee prices for 8 high-frequency laboratory tests in 2 Centers for Medicare & Medicaid Services-mandated pricing sources: (1) a machine-readable file of all available services and (2) a consumer-friendly display of 300 shoppable services. Barriers, including click counts, data availability, and intrahospital price discrepancies, were noted. Of the 145 Tennessee hospitals assessed, 97.2% were noncompliant with the Centers for Medicare & Medicaid Services final rule. Subanalysis of available machine-readable files, price estimators, and shoppable services files demonstrated 49.6%, 95.1%, and 78.6% noncompliance, respectively. Barriers to pricing information included requiring protected health information (55.9%), missing at least 1 pricing source (7.6%), having no pricing sources available (6.2%), and involving more than 3 clicks to access the cash price in machine-readable files (54.1%) and price estimators (68.6%.) Average intrahospital discrepancy for basic metabolic panel cash prices across pricing sources was $101.30 (range, $0-1012.40). Our study showed high levels of noncompliance with price transparency laws, inconsistent and inaccessible pricing, and continued challenges facing patients in Tennessee.