Sonish Sivarajkumar

Sonish Sivarajkumar, Mark Kelley, Alyssa Samolyk-Mazzanti et al.

Large language models (LLMs) have shown remarkable capabilities in Natural Language Processing (NLP), especially in domains where labeled data is scarce or expensive, such as clinical domain. However, to unlock the clinical knowledge hidden in these LLMs, we need to design effective prompts that can guide them to perform specific clinical NLP tasks without any task-specific training data. This is known as in-context learning, which is an art and science that requires understanding the strengths and weaknesses of different LLMs and prompt engineering approaches. In this paper, we present a comprehensive and systematic experimental study on prompt engineering for five clinical NLP tasks: Clinical Sense Disambiguation, Biomedical Evidence Extraction, Coreference Resolution, Medication Status Extraction, and Medication Attribute Extraction. We assessed the prompts proposed in recent literature, including simple prefix, simple cloze, chain of thought, and anticipatory prompts, and introduced two new types of prompts, namely heuristic prompting and ensemble prompting. We evaluated the performance of these prompts on three state-of-the-art LLMs: GPT-3.5, BARD, and LLAMA2. We also contrasted zero-shot prompting with few-shot prompting, and provide novel insights and guidelines for prompt engineering for LLMs in clinical NLP. To the best of our knowledge, this is one of the first works on the empirical evaluation of different prompt engineering approaches for clinical NLP in this era of generative AI, and we hope that it will inspire and inform future research in this area.

Sonish Sivarajkumar, Yanshan Wang

Deep learning algorithms are dependent on the availability of large-scale annotated clinical text datasets. The lack of such publicly available datasets is the biggest bottleneck for the development of clinical Natural Language Processing(NLP) systems. Zero-Shot Learning(ZSL) refers to the use of deep learning models to classify instances from new classes of which no training data have been seen before. Prompt-based learning is an emerging ZSL technique where we define task-based templates for NLP tasks. We developed a novel prompt-based clinical NLP framework called HealthPrompt and applied the paradigm of prompt-based learning on clinical texts. In this technique, rather than fine-tuning a Pre-trained Language Model(PLM), the task definitions are tuned by defining a prompt template. We performed an in-depth analysis of HealthPrompt on six different PLMs in a no-data setting. Our experiments prove that prompts effectively capture the context of clinical texts and perform remarkably well without any training data.

Sonish Sivarajkumar, Fengyi Gao, Parker E. Denny et al.

Post-stroke patient rehabilitation requires precise, personalized treatment plans. Natural Language Processing (NLP) offers potential to extract valuable exercise information from clinical notes, aiding in the development of more effective rehabilitation strategies. Objective: This study aims to develop and evaluate a variety of NLP algorithms to extract and categorize physical rehabilitation exercise information from the clinical notes of post-stroke patients treated at the University of Pittsburgh Medical Center. A cohort of 13,605 patients diagnosed with stroke was identified, and their clinical notes containing rehabilitation therapy notes were retrieved. A comprehensive clinical ontology was created to represent various aspects of physical rehabilitation exercises. State-of-the-art NLP algorithms were then developed and compared, including rule-based, machine learning-based algorithms, and large language model (LLM)-based algorithms (ChatGPT). Analysis was conducted on a dataset comprising 23,724 notes with detailed demographic and clinical characteristics. The rule-based NLP algorithm demonstrated superior performance in most areas, particularly in detecting the 'Right Side' location with an F1 score of 0.975, outperforming Gradient Boosting by 0.063. Gradient Boosting excelled in 'Lower Extremity' location detection (F1 score: 0.978), surpassing rule-based NLP by 0.023. It also showed notable performance in 'Passive Range of Motion' with an F1 score of 0.970, a 0.032 improvement over rule-based NLP. The rule-based algorithm efficiently handled 'Duration', 'Sets', and 'Reps' with F1 scores up to 0.65. LLM-based NLP, particularly ChatGPT with few-shot prompts, achieved high recall but generally lower precision and F1 scores. However, it notably excelled in 'Backward Plane' motion detection, achieving an F1 score of 0.846, surpassing the rule-based algorithm's 0.720.

Sonish Sivarajkumar, Yufei Huang, Yanshan Wang

Objective: To pre-train fair and unbiased patient representations from Electronic Health Records (EHRs) using a novel weighted loss function that reduces bias and improves fairness in deep representation learning models. Methods: We defined a new loss function, called weighted loss function, in the deep representation learning model to balance the importance of different groups of patients and features. We applied the proposed model, called Fair Patient Model (FPM), to a sample of 34,739 patients from the MIMIC-III dataset and learned patient representations for four clinical outcome prediction tasks. Results: FPM outperformed the baseline models in terms of three fairness metrics: demographic parity, equality of opportunity difference, and equalized odds ratio. FPM also achieved comparable predictive performance with the baselines, with an average accuracy of 0.7912. Feature analysis revealed that FPM captured more information from clinical features than the baselines. Conclusion: FPM is a novel method to pre-train fair and unbiased patient representations from EHR data using a weighted loss function. The learned representations can be used for various downstream tasks in healthcare and can be extended to other domains where bias and fairness are important.

David Oniani, Sonish Sivarajkumar, Yanshan Wang

Clinical Natural Language Processing (NLP) has become an emerging technology in healthcare that leverages a large amount of free-text data in electronic health records (EHRs) to improve patient care, support clinical decisions, and facilitate clinical and translational science research. Recently, deep learning has achieved state-of-the-art performance in many clinical NLP tasks. However, training deep learning models usually requires large annotated datasets, which are normally not publicly available and can be time-consuming to build in clinical domains. Working with smaller annotated datasets is typical in clinical NLP and therefore, ensuring that deep learning models perform well is crucial for the models to be used in real-world applications. A widely adopted approach is fine-tuning existing Pre-trained Language Models (PLMs), but these attempts fall short when the training dataset contains only a few annotated samples. Few-Shot Learning (FSL) has recently been investigated to tackle this problem. Siamese Neural Network (SNN) has been widely utilized as an FSL approach in computer vision, but has not been studied well in NLP. Furthermore, the literature on its applications in clinical domains is scarce. In this paper, we propose two SNN-based FSL approaches for clinical NLP, including Pre-Trained SNN (PT-SNN) and SNN with Second-Order Embeddings (SOE-SNN). We evaluated the proposed approaches on two clinical tasks, namely clinical text classification and clinical named entity recognition. We tested three few-shot settings including 4-shot, 8-shot, and 16-shot learning. Both clinical NLP tasks were benchmarked using three PLMs, including BERT,BioBERT, and BioClinicalBERT. The experimental results verified the effectiveness of the proposed SNN-based FSL approaches in both NLP tasks.

Sonish Sivarajkumar, Pratyush Tandale, Ankit Bhardwaj et al.

Transcription factors (TFs) play a vital role in the regulation of gene expression thereby making them critical to many cellular processes. In this study, we used graph machine learning methods to create a compendium of TF cascades using data extracted from the STRING database. A TF cascade is a sequence of TFs that regulate each other, forming a directed path in the TF network. We constructed a knowledge graph of 81,488 unique TF cascades, with the longest cascade consisting of 62 TFs. Our results highlight the complex and intricate nature of TF interactions, where multiple TFs work together to regulate gene expression. We also identified 10 TFs with the highest regulatory influence based on centrality measurements, providing valuable information for researchers interested in studying specific TFs. Furthermore, our pathway enrichment analysis revealed significant enrichment of various pathways and functional categories, including those involved in cancer and other diseases, as well as those involved in development, differentiation, and cell signaling. The enriched pathways identified in this study may have potential as targets for therapeutic intervention in diseases associated with dysregulation of transcription factors. We have released the dataset, knowledge graph, and graphML methods for the TF cascades, and created a website to display the results, which can be accessed by researchers interested in using this dataset. Our study provides a valuable resource for understanding the complex network of interactions between TFs and their regulatory roles in cellular processes.

Sonish Sivarajkumar, Thomas Yu CHow Tam, Haneef Ahamed Mohammad et al.

Alzheimer's Disease (AD) is the most common form of dementia in the United States. Sleep is one of the lifestyle-related factors that has been shown critical for optimal cognitive function in old age. However, there is a lack of research studying the association between sleep and AD incidence. A major bottleneck for conducting such research is that the traditional way to acquire sleep information is time-consuming, inefficient, non-scalable, and limited to patients' subjective experience. A gold standard dataset is created from manual annotation of 570 randomly sampled clinical note documents from the adSLEEP, a corpus of 192,000 de-identified clinical notes of 7,266 AD patients retrieved from the University of Pittsburgh Medical Center (UPMC). We developed a rule-based Natural Language Processing (NLP) algorithm, machine learning models, and Large Language Model(LLM)-based NLP algorithms to automate the extraction of sleep-related concepts, including snoring, napping, sleep problem, bad sleep quality, daytime sleepiness, night wakings, and sleep duration, from the gold standard dataset. Rule-based NLP algorithm achieved the best performance of F1 across all sleep-related concepts. In terms of Positive Predictive Value (PPV), rule-based NLP algorithm achieved 1.00 for daytime sleepiness and sleep duration, machine learning models: 0.95 and for napping, 0.86 for bad sleep quality and 0.90 for snoring; and LLAMA2 with finetuning achieved PPV of 0.93 for Night Wakings, 0.89 for sleep problem, and 1.00 for sleep duration. The results show that the rule-based NLP algorithm consistently achieved the best performance for all sleep concepts. This study focused on the clinical notes of patients with AD, but could be extended to general sleep information extraction for other diseases.

Thomas Yu Chow Tam, Sonish Sivarajkumar, Sumit Kapoor et al.

With generative artificial intelligence (AI), particularly large language models (LLMs), continuing to make inroads in healthcare, it is critical to supplement traditional automated evaluations with human evaluations. Understanding and evaluating the output of LLMs is essential to assuring safety, reliability, and effectiveness. However, human evaluation's cumbersome, time-consuming, and non-standardized nature presents significant obstacles to comprehensive evaluation and widespread adoption of LLMs in practice. This study reviews existing literature on human evaluation methodologies for LLMs in healthcare. We highlight a notable need for a standardized and consistent human evaluation approach. Our extensive literature search, adhering to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines, includes publications from January 2018 to February 2024. The review examines the human evaluation of LLMs across various medical specialties, addressing factors such as evaluation dimensions, sample types and sizes, selection, and recruitment of evaluators, frameworks and metrics, evaluation process, and statistical analysis type. Drawing on the diverse evaluation strategies employed in these studies, we propose a comprehensive and practical framework for human evaluation of LLMs: QUEST: Quality of Information, Understanding and Reasoning, Expression Style and Persona, Safety and Harm, and Trust and Confidence. This framework aims to improve the reliability, generalizability, and applicability of human evaluation of LLMs in different healthcare applications by defining clear evaluation dimensions and offering detailed guidelines.

Yuelyu Ji, Zhuochun Li, Rui Meng et al.

This paper introduces the RAG-RLRC-LaySum framework, designed to make complex biomedical research understandable to laymen through advanced Natural Language Processing (NLP) techniques. Our Retrieval Augmented Generation (RAG) solution, enhanced by a reranking method, utilizes multiple knowledge sources to ensure the precision and pertinence of lay summaries. Additionally, our Reinforcement Learning for Readability Control (RLRC) strategy improves readability, making scientific content comprehensible to non-specialists. Evaluations using the publicly accessible PLOS and eLife datasets show that our methods surpass Plain Gemini model, demonstrating a 20% increase in readability scores, a 15% improvement in ROUGE-2 relevance scores, and a 10% enhancement in factual accuracy. The RAG-RLRC-LaySum framework effectively democratizes scientific knowledge, enhancing public engagement with biomedical discoveries.

Maneesh Bilalpur, Saurabh Hinduja, Sonish Sivarajkumar et al.

The classical approach to detecting depression from vision emphasizes interpretable features, such as facial expression, and classifiers such as the Support Vector Machine (SVM). With the advent of deep learning, there has been a shift in feature representations and classification approaches. Contemporary approaches use learnt features from general-purpose vision models such as VGGNet to train machine learning models. Little is known about how classical and deep approaches compare in depression detection with respect to accuracy, fairness, and generalizability, especially across contexts. To address these questions, we compared classical and deep approaches to the detection of depression in the visual modality in two different contexts: Mother-child interactions in the TPOT database and patient-clinician interviews in the Pitt database. In the former, depression was operationalized as a history of depression per the DSM and current or recent clinically significant symptoms. In the latter, all participants met initial criteria for depression per DSM, and depression was reassessed over the course of treatment. The classical approach included handcrafted features with SVM classifiers. Learnt features were turn-level embeddings from the FMAE-IAT that were combined with Multi-Layer Perceptron classifiers. The classical approach achieved higher accuracy in both contexts. It was also significantly fairer than the deep approach in the patient-clinician context. Cross-context generalizability was modest at best for both approaches, which suggests that depression may be context-specific.

Chenyu Li, Zohaib Akhtar, Mingu Kwak et al.

As large language models (LLMs) increasingly generate and process clinical text, scalable evaluation has become critical. LLM-as-a-Judge (LaaJ), which uses LLMs to evaluate model outputs, offers a scalable alternative to costly expert review, but its healthcare adoption raises safety and bias concerns. We conducted a PRISMA-ScR scoping review of six databases (January 2020-January 2026), screening 11,727 studies and including 49. The landscape was dominated by evaluation and benchmarking applications (n=37, 75.5%), pointwise scoring (n=42, 85.7%), and GPT-family judges (n=36, 73.5%). Despite growing adoption, validation rigor was limited: among 36 studies with human involvement, the median number of expert validators was 3, while 13 (26.5%) used none. Risk of bias testing was absent in 36 studies (73.5%), only 1 (2.0%) examined demographic fairness, and none assessed temporal stability or patient context. Deployment remained limited, with 1 study (2.0%) reaching production and four (8.2%) prototype stage. Importantly, these gaps may interact: when judges and evaluated systems share training data or architectures, they may inherit similar blind spots, and agreement metrics may fail to distinguish true validity from shared errors. Minimal human oversight, limited bias assessment, and model monoculture together represent a governance gap where current validation may miss clinically significant errors. To address this, we propose MedJUDGE (Medical Judge Utility, De-biasing, Governance and Evaluation), a risk-stratified three-pillar framework organized around validity, safety, and accountability across clinical risk tiers, providing deployment-oriented evaluation guidance for healthcare LaaJ systems.

Sonish Sivarajkumar, Kimia Ameri, Chuqin Li et al.

Cardiovascular events, such as heart attacks and strokes, remain a leading cause of mortality globally, necessitating meticulous monitoring and adjudication in clinical trials. This process, traditionally performed manually by clinical experts, is time-consuming, resource-intensive, and prone to inter-reviewer variability, potentially introducing bias and hindering trial progress. This study addresses these critical limitations by presenting a novel framework for automating the adjudication of cardiovascular events in clinical trials using Large Language Models (LLMs). We developed a two-stage approach: first, employing an LLM-based pipeline for event information extraction from unstructured clinical data and second, using an LLM-based adjudication process guided by a Tree of Thoughts approach and clinical endpoint committee (CEC) guidelines. Using cardiovascular event-specific clinical trial data, the framework achieved an F1-score of 0.82 for event extraction and an accuracy of 0.68 for adjudication. Furthermore, we introduce the CLEART score, a novel, automated metric specifically designed for evaluating the quality of AI-generated clinical reasoning in adjudicating cardiovascular events. This approach demonstrates significant potential for substantially reducing adjudication time and costs while maintaining high-quality, consistent, and auditable outcomes in clinical trials. The reduced variability and enhanced standardization also allow for faster identification and mitigation of risks associated with cardiovascular therapies.

Sonish Sivarajkumar, Hang Zhang, Yuelyu Ji et al.

Electronic health records (EHRs) are rich clinical data sources but complex repositories of patient data, spanning structured elements (demographics, vitals, lab results, codes), unstructured clinical notes and other modalities of data. Harnessing this heterogeneity is critical for improving patient outcomes. Recent advances in large language models (LLMs) have enabled foundation models that can learn from multiple data modalities and support clinical tasks. However, most current approaches simply serialize numeric EHR data into text, which risks losing temporal and quantitative detail. We introduce Generative Deep Patient (GDP), a multimodal foundation model that natively encodes structured EHR time-series via a CNN-Transformer encoder and fuses it with unstructured EHRs through cross-modal attention into a LLaMA-based decoder. GDP is trained in two stages: (1) generative pretraining, where it learns to produce clinical narratives from raw patient timelines while also performing masked feature prediction (MFP) and next time-step prediction (NTP) to capture temporal dynamics; and (2) multi-task fine-tuning for clinically meaningful predictions (e.g., heart failure, type 2 diabetes, 30-day readmission). In clinical prediction, GDP demonstrated superior performance on MIMIC-IV: heart failure AUROC = 0.923, type 2 diabetes AUROC = 0.817, and 30-day readmission AUROC = 0.627. For narrative generation, GDP achieved ROUGE-L = 0.135 and BERTScore-F1 = 0.545. In a blinded human evaluation, GDP-Instruct scored highest on faithfulness, fluency, and overall clinical utility, suggesting reduced hospital documentation workload without sacrificing accuracy. Our results demonstrate that a single multimodal foundation model can both predict clinically actionable events and generate high-quality clinical narratives. Furthermore, GDP's flexible architecture can be extended to additional modalities.

Fengyi Gao, Xingyu Zhang, Sonish Sivarajkumar et al.

In this study, we utilized statistical analysis and machine learning methods to examine whether rehabilitation exercises can improve patients post-stroke functional abilities, as well as forecast the improvement in functional abilities. Our dataset is patients' rehabilitation exercises and demographic information recorded in the unstructured electronic health records (EHRs) data and free-text rehabilitation procedure notes. We collected data for 265 stroke patients from the University of Pittsburgh Medical Center. We employed a pre-existing natural language processing (NLP) algorithm to extract data on rehabilitation exercises and developed a rule-based NLP algorithm to extract Activity Measure for Post-Acute Care (AM-PAC) scores, covering basic mobility (BM) and applied cognitive (AC) domains, from procedure notes. Changes in AM-PAC scores were classified based on the minimal clinically important difference (MCID), and significance was assessed using Friedman and Wilcoxon tests. To identify impactful exercises, we used Chi-square tests, Fisher's exact tests, and logistic regression for odds ratios. Additionally, we developed five machine learning models-logistic regression (LR), Adaboost (ADB), support vector machine (SVM), gradient boosting (GB), and random forest (RF)-to predict outcomes in functional ability. Statistical analyses revealed significant associations between functional improvements and specific exercises. The RF model achieved the best performance in predicting functional outcomes. In this study, we identified three rehabilitation exercises that significantly contributed to patient post-stroke functional ability improvement in the first two months. Additionally, the successful application of a machine learning model to predict patient-specific functional outcomes underscores the potential for precision rehabilitation.