---
title: "Input Complexity Trade-offs for Transformer Models in Longitudinal EHR Asthma Prediction"
id: "medrxiv-20-rethinking-input-complexity-in-transformer-based-clinical-prediction"
canonical_url: "https://medichelpline.com/clinical-feed/medrxiv-20-rethinking-input-complexity-in-transformer-based-clinical-prediction"
content_type: "clinical_feed_article"
specialty: "Pulmonology"
source_name: "medRxiv (Clinical Preprints)"
source_url: "https://www.medrxiv.org/content/10.64898/2026.09.03.26362165v1?rss=1"
published_at: "2026-09-07T12:00:00.000Z"
evidence_level: "Verified Feed"
license: "CC-BY-NC-4.0 / Informational Use"
---
# Input Complexity Trade-offs for Transformer Models in Longitudinal EHR Asthma Prediction
## Provenance & Clinical Metadata
- **Canonical URL:** https://medichelpline.com/clinical-feed/medrxiv-20-rethinking-input-complexity-in-transformer-based-clinical-prediction
- **Specialty:** [Pulmonology](https://medichelpline.com/clinical-feed/pulmonology.md)
- **Primary Source:** medRxiv (Clinical Preprints)
- **Source URL:** [Original Journal Publication](https://www.medrxiv.org/content/10.64898/2026.09.03.26362165v1?rss=1)
- **Published At:** 2026-09-07T12:00:00.000Z
- **Evidence Rating:** Verified Feed
## Executive GIST (TL;DR)
- Transformer-based clinical prediction models for longitudinal EHR data are commonly built using large feature sets and long patient histories, based on the assumption that more input data improves performance. The study evaluated how **feature dimensionality** and **sequence length** affect predictive discrimination, calibration, risk stratification, and computational efficiency. - The cohort consisted of adults with mild asthma from an integrated healthcare system; the supervised task was predicting **acute asthma exacerbation** using transformer-based architectures and longitudinal EHR inputs. - Feature reduction strategies tested included: attribution-based selection using Integrated Gradients, univariate performance-based selection, and clinician-guided selection. Sequence length was varied by truncating patient histories at different percentiles. - Models with substantially fewer features achieved discrimination close to an 80-feature reference model: AUROC range reported for reduced models was 0.843–0.864 versus 0.870 for the full model, indicating modest loss in discrimination with reduced dimensionality. - Moderate truncation of sequence length reduced training time by more than 70% while producing minimal loss in discrimination, demonstrating substantial computational gains from shorter input sequences. - Reduced-dimensionality models showed some attenuation of predicted risk at the upper tail (affecting calibration among highest predicted-risk patients), but they identified similar high-risk groups and captured comparable proportions of exacerbation events at clinically relevant thresholds. - Overall, transformer-based models maintained strong predictive performance across reduced feature sets, while moderate sequence-length reduction offered large computational savings with limited impact on discrimination; these findings illustrate trade-offs between input complexity, performance, calibration at extreme risk levels, and efficiency. - The study was approved by the Kaiser Permanente Southern California IRB. Competing interests include consulting or grant support for two authors; other authors declared no relevant conflicts. Data may be available under collaboration and data-use agreement conditions. Funding included an NHLBI R01 grant (R01 HL 163049).
## Clinical Analysis & Structured Key Points
Rethinking Input Complexity in Transformer-Based Clinical Prediction: Implications for Feature Dimensionality and Sequence Length in Longitudinal EHR Data | medRxiv Skip to main content Rethinking Input Complexity in Transformer-Based Clinical Prediction: Implications for Feature Dimensionality and Sequence Length in Longitudinal EHR Data Wansu Chen , Botao Zhou , Robert S Zeiger , William W Crawford , Michael Schatz , Eric J Puttock , Stanley Xu , Matthew T Slaughter , Fagen Xie doi: https://doi.org/10.64898/2026.09.03.26362165 Wansu Chen 1 Kaiser Permanente Southern California - Department of Research and Evaluation; Find this author on Google Scholar Find this author on PubMed Search for this author on this site For correspondence: wansu.chen{at}kp.org Botao Zhou 1 Kaiser Permanente Southern California - Department of Research and Evaluation; Find this author on Google Scholar Find this author on PubMed Search for this author on this site Robert S Zeiger 2 Kaiser Permanente Southern California - Department of Allergy; Find this author on Google Scholar Find this author on PubMed Search for this author on this site William W Crawford 2 Kaiser Permanente Southern California - Department of Allergy; Find this author on Google Scholar Find this author on PubMed Search for this author on this site Michael Schatz 2 Kaiser Permanente Southern California - Department of Allergy; Find this author on Google Scholar Find this author on PubMed Search for this author on this site Eric J Puttock 1 Kaiser Permanente Southern California - Department of Research and Evaluation; Find this author on Google Scholar Find this author on PubMed Search for this author on this site Stanley Xu 1 Kaiser Permanente Southern California - Department of Research and Evaluation; Find this author on Google Scholar Find this author on PubMed Search for this author on this site Matthew T Slaughter 3 Kaiser Permanente Center for Health Research Find this author on Google Scholar Find this author on PubMed Search for this author on this site Fagen Xie 1 Kaiser Permanente Southern California - Department of Research and Evaluation; Find this author on Google Scholar Find this author on PubMed Search for this author on this site Abstract Info/History Metrics Data/Code Preview PDF Abstract Objective: Transformer-based models for clinical prediction using longitudinal electronic health record (EHR) data are often developed with large feature sets and long patient histories under the assumption that more data improves performance. However, high-dimensional inputs and long sequences increase computational burden, potentially limiting scalability. We evaluated how feature dimensionality and sequence length affect predictive performance, calibration, risk stratification, and computational efficiency in EHR prediction. Methods: Using longitudinal EHR data from adults with mild asthma in an integrated healthcare system, we evaluated input representation design for predicting acute asthma exacerbation. Feature dimensionality was reduced using Integrated Gradients attribution scores, univariate performance-based selection, and clinically guided selection strategies. Sequence length was varied using percentile-based truncation of patient histories. Performance was assessed using discrimination, calibration, high-risk classification, threshold-based event capture, and computational efficiency across regions. Results: Models using fewer features achieved discrimination comparable to the 80-feature reference model, with AUROC values ranging from 0.843 to 0.864 versus 0.870 for the full model. Moderate sequence-length truncation reduced training time by more than 70% with minimal loss in discrimination. Although reduced-dimensional models showed attenuation of predicted risk at the upper tail, they identified similar high-risk populations and captured comparable proportions of asthma exacerbation events at clinically relevant thresholds. Conclusion: Transformer-based prediction models maintained strong performance across reduced feature sets. While dimensionality reduction modestly affected calibration at the highest risk levels, moderate sequence-length reduction substantially reduced computational burden with limited change in overall discrimination. These findings highlight trade-offs between input complexity, predictive performance, and computational efficiency. Competing Interest Statement Robert S. Zeiger reports consulting fees from AstraZeneca; grant support from Sanofi; and royalties from UpToDate. Michael Schatz reports grant support from Sanofi and royalties from UpToDate. The rest of the authors declare that they have no relevant conflicts of interest. Author Declarations I confirm all relevant ethical guidelines have been followed, and any necessary IRB and/or ethics committee approvals have been obtained. Yes The details of the IRB/oversight body that provided approval or exemption for the research described are given below: This study was approved by Kaiser Permanente Southern California IRB I confirm that all necessary patient/participant consent has been obtained and the appropriate institutional forms have been archived, and that any patient/participant/sample identifiers included were not known to anyone (e.g., hospital staff, patients or participants themselves) outside the research group so cannot be used to identify individuals. Yes I understand that all clinical trials and any other prospective interventional studies must be registered with an ICMJE-approved registry, such as ClinicalTrials.gov. I confirm that any such study reported in the manuscript has been registered and the trial registration ID is provided (note: if posting a prospective study registered retrospectively, please provide a statement in the trial ID field explaining why the study was not registered in advance). Yes I have followed all appropriate research reporting guidelines, such as any relevant EQUATOR Network research reporting checklist(s) and other pertinent material, if applicable. Yes Data Availability Anonymized data that support the findings of this study may be made available from the investigative team in the following conditions: 1) agreement to collaborate with the study team on all publications, 2) provision of external funding for administrative and investigator time necessary for this collaboration, 3) demonstration that the external investigative team is qualified and has documented evidence of training for human subjects protections, and 4) agreement to abide by the terms outlined in data use agreements between institutions. Funder Information Declared National Heart, Lung, and Blood Institute , R01 HL 163049 Copyright The copyright holder for this preprint is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. All rights reserved. No reuse allowed without permission. Back to top Previous Next Posted September 07, 2026. Download PDF Data/Code Email Thank you for your interest in spreading the word about medRxiv. NOTE: Your email address is requested solely to identify you as the sender of this article. Your Email * Your Name * Send To * Enter multiple addresses on separate lines or separate them with commas. You are going to email the following Rethinking Input Complexity in Transformer-Based Clinical Prediction: Implications for Feature Dimensionality and Sequence Length in Longitudinal EHR Data Message Subject (Your Name) has forwarded a page to you from medRxiv Message Body (Your Name) thought you would like to see this page from the medRxiv website. Your Personal Message CAPTCHA This question is for testing whether or not you are a human visitor and to prevent automated spam submissions. Share Rethinking Input Complexity in Transformer-Based Clinical Prediction: Implications for Feature Dimensionality and Sequence Length in Longitudinal EHR Data Wansu Chen , Botao Zhou , Robert S Zeiger , William W Crawford , Michael Schatz , Eric J Puttock , Stanley Xu , Matthew T Slaughter , Fagen Xie medRxiv 2026.09.03.26362165; doi: https://doi.org/10.64898/2026.09.03.26362165 Share This Article: Copy Citation Tools Rethinking Input Complexity in Transformer-Based Clinical Prediction: Implications for Feature Dimensionality and Sequence Length in Longitudinal EHR Data Wansu Chen , Botao Zhou , Robert S Zeiger , William W Crawford , Michael Schatz , Eric J Puttock , Stanley Xu , Matthew T Slaughter , Fagen Xie medRxiv 2026.09.03.26362165; doi: https://doi.org/10.64898/2026.09.03.26362165 Citation Manager Formats BibTeX Bookends EasyBib EndNote (tagged) EndNote 8 (xml) Medlars Mendeley Papers RefWorks Tagged Ref Manager RIS Zotero Tweet Widget Facebook Like Google Plus One Subject Areas All Articles Addiction Medicine (616) Allergy and Immunology (897) Anesthesia (328) Cardiovascular Medicine (4781) Dentistry and Oral Medicine (476) Dermatology (411) Emergency Medicine (646) Endocrinology (including Diabetes Mellitus and Metabolic Disease) (1617) Epidemiology (15865) Forensic Medicine (32) Gastroenterology (1196) Genetic and Genomic Medicine (7006) Geriatric Medicine (728) Health Economics (1060) Health Informatics (5007) Health Policy (1429) Health Systems and Quality Improvement (1747) Hematology (583) HIV/AIDS (1338) Infectious Diseases (except HIV/AIDS) (16258) Intensive Care and Critical Care Medicine (1164) Medical Education (665) Medical Ethics (151) Nephrology (721) Neurology (7216) Nursing (366) Nutrition (1074) Obstetrics and Gynecology (1234) Occupational and Environmental Health (1002) Oncology (3584) Ophthalmology (1044) Orthopedics (395) Otolaryngology (448) Pain Medicine (466) Palliative Medicine (139) Pathology (704) Pediatrics (1784) Pharmacology and Therapeutics (733) Primary Care Research (759) Psychiatry and Clinical Psychology (5851) Public and Global Health (9716) Radiology and Imaging (2399) Rehabilitation Medicine and Physical Therapy (1438) Respiratory Medicine (1245) Rheumatology (636) Sexual and Reproductive Health (769) Sports Medicine (577) Surgery (772) Toxicology (107) Transplantation (303) Urology (288)
## Related Clinical Research

- [Multi-ancestry GWAS Reveals Shared and Distinct Genetic Architecture in Mild vs Moderate-to-Severe](https://medichelpline.com/clinical-feed/medrxiv-1-multi-ancestry-genome-wide-association-analyses-identify-shared-and-specific.md)
- [Extended Treatment Duration and Recurrence in Cavitary Drug-Susceptible Pulmonary Tuberculosis](https://medichelpline.com/clinical-feed/plos-one-11-extended-treatment-duration-and-recurrence-in-successfully-treated-drug.md)
- [Triglyceride-Glucose (TyG) Index and Mortality in Non-Diabetic Fibrotic Lung Disease on Antifibrot](https://medichelpline.com/clinical-feed/plos-one-1-association-between-triglyceride-glucose-index-and-all-cause-mortality-in-non.md)
- [Nasal transcriptomics identify five asthma endotypes including T2‑low profiles in Puerto Rican and](https://medichelpline.com/clinical-feed/medrxiv-18-nasal-transcriptomics-characterize-t-helper-t-2-low-asthma-endotypes-in-a-study.md)
- [AI-designed lung drug rentosertib shows shifts in proteomic aging clocks in phase 2a trial](https://medichelpline.com/clinical-feed/medical-news-today-0-ai-designed-lung-disease-drug-lowers-biological-age-markers-in-early-trial.md)

## Navigation
- [← Back to Pulmonology Feed](https://medichelpline.com/clinical-feed/pulmonology.md)
- [← All Clinical Specialties](https://medichelpline.com/clinical-feed.md)
## Medical & Regulatory Disclaimer

> [!CAUTION]
> MedicHelpline content is structured for research, educational, and professional discovery purposes. It does not constitute individual medical advice, clinical diagnosis, or treatment recommendations.
> Always verify dosing, contraindications, and regulatory alerts against official product labeling and primary regulatory sources before clinical decision-making.