skip to main content

Title: DRG-LLaMA : tuning LLaMA model to predict diagnosis-related group for hospitalized patients

In the U.S. inpatient payment system, the Diagnosis-Related Group (DRG) is pivotal, but its assignment process is inefficient. The study introduces , an advanced large language model (LLM) fine-tuned on clinical notes to enhance DRGs assignment. Utilizing LLaMA as the foundational model and optimizing it through Low-Rank Adaptation (LoRA) on 236,192 MIMIC-IV discharge summaries, our -7B model exhibited a noteworthy macro-averaged F1 score of 0.327, a top-1 prediction accuracy of 52.0%, and a macro-averaged Area Under the Curve (AUC) of 0.986, with a maximum input token length of 512. This model surpassed the performance of prior leading models in DRG prediction, showing a relative improvement of 40.3% and 35.7% in macro-averaged F1 score compared to ClinicalBERT and CAML, respectively. Applied to base DRG and complication or comorbidity (CC)/major complication or comorbidity (MCC) prediction, achieved a top-1 prediction accuracy of 67.8% and 67.5%, respectively. Additionally, our findings indicate that ’s performance correlates with increased model parameters and input context lengths.

more » « less
Author(s) / Creator(s):
; ; ; ;
Publisher / Repository:
Nature Publishing Group
Date Published:
Journal Name:
npj Digital Medicine
Medium: X
Sponsoring Org:
National Science Foundation
More Like this
  1. null (Ed.)
    Abstract The inter-residue contact prediction and deep learning showed the promise to improve the estimation of protein model accuracy (EMA) in the 13th Critical Assessment of Protein Structure Prediction (CASP13). To further leverage the improved inter-residue distance predictions to enhance EMA, during the 2020 CASP14 experiment, we integrated several new inter-residue distance features with the existing model quality assessment features in several deep learning methods to predict the quality of protein structural models. According to the evaluation of performance in selecting the best model from the models of CASP14 targets, our three multi-model predictors of estimating model accuracy (MULTICOM-CONSTRUCT, MULTICOM-AI, and MULTICOM-CLUSTER) achieve the averaged loss of 0.073, 0.079, and 0.081, respectively, in terms of the global distance test score (GDT-TS). The three methods are ranked first, second, and third out of all 68 CASP14 predictors. MULTICOM-DEEP, the single-model predictor of estimating model accuracy (EMA), is ranked within top 10 among all the single-model EMA methods according to GDT-TS score loss. The results demonstrate that inter-residue distance features are valuable inputs for deep learning to predict the quality of protein structural models. However, larger training datasets and better ways of leveraging inter-residue distance information are needed to fully explore its potentials. 
    more » « less
  2. Abstract Objective We develop natural language processing (NLP) methods capable of accurately classifying tumor attributes from pathology reports given minimal labeled examples. Our hierarchical cancer to cancer transfer (HCTC) and zero-shot string similarity (ZSS) methods are designed to exploit shared information between cancers and auxiliary class features, respectively, to boost performance using enriched annotations which give both location-based information and document level labels for each pathology report. Materials and Methods Our data consists of 250 pathology reports each for kidney, colon, and lung cancer from 2002 to 2019 from a single institution (UCSF). For each report, we classified 5 attributes: procedure, tumor location, histology, grade, and presence of lymphovascular invasion. We develop novel NLP techniques involving transfer learning and string similarity trained on enriched annotations. We compare HCTC and ZSS methods to the state-of-the-art including conventional machine learning methods as well as deep learning methods. Results For our HCTC method, we see an improvement of up to 0.1 micro-F1 score and 0.04 macro-F1 averaged across cancer and applicable attributes. For our ZSS method, we see an improvement of up to 0.26 micro-F1 and 0.23 macro-F1 averaged across cancer and applicable attributes. These comparisons are made after adjusting training data sizes to correct for the 20% increase in annotation time for enriched annotations compared to ordinary annotations. Conclusions Methods based on transfer learning across cancers and augmenting information methods with string similarity priors can significantly reduce the amount of labeled data needed for accurate information extraction from pathology reports. 
    more » « less
  3. The quality of parent–child interaction is critical for child cognitive development. The Dyadic Parent–Child Interaction Coding System (DPICS) is commonly used to assess parent and child behaviors. However, manual annotation of DPICS codes by parent–child interaction therapists is a time-consuming task. To assist therapists in the coding task, researchers have begun to explore the use of artificial intelligence in natural language processing to classify DPICS codes automatically. In this study, we utilized datasets from the DPICS book manual, five families, and an open-source PCIT dataset. To train DPICS code classifiers, we employed the pre-trained fine-tuned model RoBERTa as our learning algorithm. Our study shows that fine-tuning the pre-trained RoBERTa model achieves the highest results compared to other methods in sentence-based DPICS code classification assignments. For the DPICS manual dataset, the overall accuracy was 72.3% (72.2% macro-precision, 70.5% macro-recall, and 69.6% macro-F-score). Meanwhile, for the PCIT dataset, the overall accuracy was 79.8% (80.4% macro-precision, 79.7% macro-recall, and 79.8% macro-F-score), surpassing the previous highest results of 78.3% accuracy (79% precision, 77% recall) averaged over the eight DPICS classes. These results show that fine-tuning the pre-trained RoBERTa model could provide valuable assistance to experts in the labeling process.

    more » « less
  4. Abstract

    The timeliness of monitoring is essential to algal bloom management. However, acquiring algal bio-indicators can be time-consuming and laborious, and bloom biomass data often contain a large proportion of extreme values limiting the predictive models. Therefore, to predict algal blooms from readily water quality parameters (i.e. dissolved oxygen, pH, etc), and to provide a novel solution to the modeling challenges raised by the extremely distributed biomass data, a Bayesian scale-mixture of skew-normal (SMSN) model was proposed. In this study, our SMSN model accurately predicted over-dispersed biomass variations with skewed distributions in both rivers and lakes (in-sample and out-of-sample predictionR2ranged from 0.533 to 0.706 and 0.412 to 0.742, respectively). Moreover, we successfully achieve a probabilistic assessment of algal blooms with the Bayesian framework (accuracy >0.77 and macro-F1score >0.72), which robustly decreased the classic point-prediction-based inaccuracy by up to 34%. This work presented a promising Bayesian SMSN modeling technique, allowing for real-time prediction of algal biomass variations andin-situprobabilistic assessment of algal bloom.

    more » « less
  5. Abstract

    Solar energetic particles (SEPs) are associated with extreme solar events that can cause major damage to space- and ground-based life and infrastructure. High-intensity SEP events, particularly ∼100 MeV SEP events, can pose severe health risks for astronauts owing to radiation exposure and affect Earth’s orbiting satellites (e.g., Landsat and the International Space Station). A major challenge in the SEP event prediction task is the lack of adequate SEP data because of the rarity of these events. In this work, we aim to improve the prediction of ∼30, ∼60, and ∼100 MeV SEP events by synthetically increasing the number of SEP samples. We explore the use of a univariate and multivariate time series of proton flux data as input to machine-learning-based prediction methods, such as time series forest (TSF). Our study covers solar cycles 22, 23, and 24. Our findings show that using data augmentation methods, such as the synthetic minority oversampling technique, remarkably increases the accuracy and F1-score of the classifiers used in this research, especially for TSF, where the average accuracy increased by 20%, reaching around 90% accuracy in the ∼100 MeV SEP prediction task. We also achieved higher prediction accuracy when using the multivariate time series data of the proton flux. Finally, we build a pipeline framework for our best-performing model, TSF, and provide a comprehensive hierarchical classification of the ∼100, ∼60, and ∼30 MeV and non-SEP prediction scenarios.

    more » « less