Combining adult with pediatric patient data to develop a clinical decision support tool intended for children: leveraging machine learning to model heterogeneity.

Abstract

Background

Clinical decision support (CDS) tools built using adult data do not typically perform well for children. We explored how best to leverage adult data to improve the performance of such tools. This study assesses whether it is better to build CDS tools for children using data from children alone or to use combined data from both adults and children.

Methods

Retrospective cohort using data from 2017 to 2020. Participants include all individuals (adults and children) receiving an elective surgery at a large academic medical center that provides adult and pediatric services. We predicted need for mechanical ventilation or admission to the intensive care unit (ICU). Predictor variables included demographic, clinical, and service utilization factors known prior to surgery. We compared predictive models built using machine learning to regression-based methods that used a pediatric or combined adult-pediatric cohort. We compared model performance based on Area Under the Receiver Operator Characteristic.

Results

While we found that adults and children have different risk factors, machine learning methods are able to appropriately model the underlying heterogeneity of each population and produce equally accurate predictive models whether using data only from pediatric patients or combined data from both children and adults. Results from regression-based methods were improved by the use of pediatric-specific data.

Conclusions

CDS tools for children can successfully use combined data from adults and children if the model accounts for underlying heterogeneity, as in machine learning models.

Department

Description

Provenance

Citation

Published Version (Please cite this version)

10.1186/s12911-022-01827-4

Publication Info

Sabharwal, Paul, Jillian H Hurst, Rohit Tejwani, Kevin T Hobbs, Jonathan C Routh and Benjamin A Goldstein (2022). Combining adult with pediatric patient data to develop a clinical decision support tool intended for children: leveraging machine learning to model heterogeneity. BMC medical informatics and decision making, 22(1). p. 84. 10.1186/s12911-022-01827-4 Retrieved from https://hdl.handle.net/10161/29064.

This is constructed from limited available data and may be imprecise. To cite this article, please review & use the official citation provided by the journal.

Scholars@Duke

Hurst

Jillian Hurst

Assistant Professor in Pediatrics

Children's Health & Discovery Initiative:
The prenatal period, infancy, childhood, and adolescence, represent critical time periods of human development that include more developmental milestones than any other period of the lifespan. Conditions during these developmental windows – including biological, social, economic, health, and environmental factors – have a profound impact on lifelong health. The Children’s Health and Discovery Initiative (CHDI) was founded on the hypothesis that interventions early in life will improve population health across the lifespan. To this end, the overarching goal of the CHDI is to create a robust coalition of multidisciplinary investigators and a pipeline of infrastructure, data, and research projects focused on developing innovative approaches to identifying and modulating early life factors that impact lifelong health and well-being.

Intersections of the upper respiratory microbiome, environmental exposures, and childhood respiratory infections
Early life exposure to and colonization with microbes has a profound influence on the education of the immune system and susceptibility to viral and bacterial infections later in life. My research is focused on the influence of the upper respiratory microbiome on the development of recurrent respiratory infections, including acute otitis media (AOM), the leading cause of antibiotic prescriptions and healthcare consultations among children. Importantly, some children develop recurrent infections that are thought to be linked to dysbiosis of the nasopharyngeal microbiome. My overarching goals are to identify alterations in the upper respiratory microbiome associated with AOM and to elucidate host factors and exposures that predispose some children to the development of recurrent AOM episodes.

Routh

Jonathan Charles Routh

Paul H. Sherman, M.D. Distinguished Associate Professor of Surgery

I am a pediatric urologist and health services researcher who is interested in caring for children with urological problems, conducting research on how to improve that care, and mentoring young researchers to ensure that the next generation does both better than I currently can. 

My clinical interests include minimally-invasive surgery, neurogenic and non-neurogenic voiding dysfunction, complex urologic reconstruction (particularly in children with spina bifida), and pediatric urologic oncology (particularly Wilms tumor and rhabdomyosarcoma). My research has been funded by awards from the NIH, CDC, FDA, and multiple foundations and industry partners, and during my time on faculty at Duke I have had the pleasure of collaborating with many groups and individuals around the world on a number of projects. Over the past 15 years, I have formally mentored nearly 3 dozen undergraduates, medical students, urology residents, post-doctoral students, and junior faculty members across multiple disciplines (pediatrics, urogynecology, urology, and nursing).

Goldstein

Benjamin Alan Goldstein

Professor of Biostatistics & Bioinformatics

I study the meaningful use of Electronic Health Records data. My research interests sit at the intersection of biostatistics, biomedical informatics, machine learning and epidemiology. I collaborate with researchers both locally at Duke as well as nationally. I am interested in speaking with any students, methodologistis or collaborators interested in EHR data.

Please find more information at: https://sites.duke.edu/bgoldstein/


Unless otherwise indicated, scholarly articles published by Duke faculty members are made available here with a CC-BY-NC (Creative Commons Attribution Non-Commercial) license, as enabled by the Duke Open Access Policy. If you wish to use the materials in ways not already permitted under CC-BY-NC, please consult the copyright owner. Other materials are made available here through the author’s grant of a non-exclusive license to make their work openly accessible.