DLMM as a lossless one-shot algorithm for collaborative multi-site distributed linear mixed models

Chongliang Luo; Md Nazmul Islam; Natalie E. Sheils; John Buresh; Jenna Reps; Martijn J. Schuemie; Patrick B. Ryan; Mackenzie Edmondson; Rui Duan; Jiayi Tong; Arielle Marks-Anglin; Jiang Bian; Zhaoyi Chen; Talita Duarte-Salles; Sergio Fernández-Bertolín; Thomas Falconer; Chungsoo Kim; Rae Woong Park; Stephen R. Pfohl; Nigam H. Shah; Andrew E. Williams; Hua Xu; Yujia Zhou; Ebbing Lautenbach; Jalpa A. Doshi; Rachel M. Werner; David A. Asch; Yong Chen

doi:10.1038/s41467-022-29160-4

DLMM as a lossless one-shot algorithm for collaborative multi-site distributed linear mixed models

Chongliang Luo, Md Nazmul Islam, Natalie E. Sheils, John Buresh, Jenna Reps, Martijn J. Schuemie, Patrick B. Ryan, Mackenzie Edmondson, Rui Duan, Jiayi Tong, Arielle Marks-Anglin, Jiang Bian, Zhaoyi Chen, Talita Duarte-Salles, Sergio Fernández-Bertolín, Thomas Falconer, Chungsoo Kim, Rae Woong Park, Stephen R. Pfohl, Nigam H. ShahAndrew E. Williams, Hua Xu, Yujia Zhou, Ebbing Lautenbach, Jalpa A. Doshi, Rachel M. Werner, David A. Asch, Yong Chen

Research output: Contribution to journal › Article › peer-review

16 Scopus citations

Abstract

Linear mixed models are commonly used in healthcare-based association analyses for analyzing multi-site data with heterogeneous site-specific random effects. Due to regulations for protecting patients’ privacy, sensitive individual patient data (IPD) typically cannot be shared across sites. We propose an algorithm for fitting distributed linear mixed models (DLMMs) without sharing IPD across sites. This algorithm achieves results identical to those achieved using pooled IPD from multiple sites (i.e., the same effect size and standard error estimates), hence demonstrating the lossless property. The algorithm requires each site to contribute minimal aggregated data in only one round of communication. We demonstrate the lossless property of the proposed DLMM algorithm by investigating the associations between demographic and clinical characteristics and length of hospital stay in COVID-19 patients using administrative claims from the UnitedHealth Group Clinical Discovery Database. We extend this association study by incorporating 120,609 COVID-19 patients from 11 collaborative data sources worldwide.

Original language	English
Article number	1678
Journal	Nature communications
Volume	13
Issue number	1
DOIs	https://doi.org/10.1038/s41467-022-29160-4
State	Published - Dec 2022

Access to Document

10.1038/s41467-022-29160-4

Cite this

Luo, C., Islam, M. N., Sheils, N. E., Buresh, J., Reps, J., Schuemie, M. J., Ryan, P. B., Edmondson, M., Duan, R., Tong, J., Marks-Anglin, A., Bian, J., Chen, Z., Duarte-Salles, T., Fernández-Bertolín, S., Falconer, T., Kim, C., Park, R. W., Pfohl, S. R., ... Chen, Y. (2022). DLMM as a lossless one-shot algorithm for collaborative multi-site distributed linear mixed models. Nature communications, 13(1), Article 1678. https://doi.org/10.1038/s41467-022-29160-4

@article{f6c8409ac3ed4b1cb27df3dfa95ac12c,

title = "DLMM as a lossless one-shot algorithm for collaborative multi-site distributed linear mixed models",

abstract = "Linear mixed models are commonly used in healthcare-based association analyses for analyzing multi-site data with heterogeneous site-specific random effects. Due to regulations for protecting patients{\textquoteright} privacy, sensitive individual patient data (IPD) typically cannot be shared across sites. We propose an algorithm for fitting distributed linear mixed models (DLMMs) without sharing IPD across sites. This algorithm achieves results identical to those achieved using pooled IPD from multiple sites (i.e., the same effect size and standard error estimates), hence demonstrating the lossless property. The algorithm requires each site to contribute minimal aggregated data in only one round of communication. We demonstrate the lossless property of the proposed DLMM algorithm by investigating the associations between demographic and clinical characteristics and length of hospital stay in COVID-19 patients using administrative claims from the UnitedHealth Group Clinical Discovery Database. We extend this association study by incorporating 120,609 COVID-19 patients from 11 collaborative data sources worldwide.",

author = "Chongliang Luo and Islam, {Md Nazmul} and Sheils, {Natalie E.} and John Buresh and Jenna Reps and Schuemie, {Martijn J.} and Ryan, {Patrick B.} and Mackenzie Edmondson and Rui Duan and Jiayi Tong and Arielle Marks-Anglin and Jiang Bian and Zhaoyi Chen and Talita Duarte-Salles and Sergio Fern{\'a}ndez-Bertol{\'i}n and Thomas Falconer and Chungsoo Kim and Park, {Rae Woong} and Pfohl, {Stephen R.} and Shah, {Nigam H.} and Williams, {Andrew E.} and Hua Xu and Yujia Zhou and Ebbing Lautenbach and Doshi, {Jalpa A.} and Werner, {Rachel M.} and Asch, {David A.} and Yong Chen",

note = "Publisher Copyright: {\textcopyright} 2022, The Author(s).",

year = "2022",

month = dec,

doi = "10.1038/s41467-022-29160-4",

language = "English",

volume = "13",

journal = "Nature communications",

issn = "2041-1723",

number = "1",

}

Luo, C, Islam, MN, Sheils, NE, Buresh, J, Reps, J, Schuemie, MJ, Ryan, PB, Edmondson, M, Duan, R, Tong, J, Marks-Anglin, A, Bian, J, Chen, Z, Duarte-Salles, T, Fernández-Bertolín, S, Falconer, T, Kim, C, Park, RW, Pfohl, SR, Shah, NH, Williams, AE, Xu, H, Zhou, Y, Lautenbach, E, Doshi, JA, Werner, RM, Asch, DA & Chen, Y 2022, 'DLMM as a lossless one-shot algorithm for collaborative multi-site distributed linear mixed models', Nature communications, vol. 13, no. 1, 1678. https://doi.org/10.1038/s41467-022-29160-4

TY - JOUR

T1 - DLMM as a lossless one-shot algorithm for collaborative multi-site distributed linear mixed models

AU - Luo, Chongliang

AU - Islam, Md Nazmul

AU - Sheils, Natalie E.

AU - Buresh, John

AU - Reps, Jenna

AU - Schuemie, Martijn J.

AU - Ryan, Patrick B.

AU - Edmondson, Mackenzie

AU - Duan, Rui

AU - Tong, Jiayi

AU - Marks-Anglin, Arielle

AU - Bian, Jiang

AU - Chen, Zhaoyi

AU - Duarte-Salles, Talita

AU - Fernández-Bertolín, Sergio

AU - Falconer, Thomas

AU - Kim, Chungsoo

AU - Park, Rae Woong

AU - Pfohl, Stephen R.

AU - Shah, Nigam H.

AU - Williams, Andrew E.

AU - Xu, Hua

AU - Zhou, Yujia

AU - Lautenbach, Ebbing

AU - Doshi, Jalpa A.

AU - Werner, Rachel M.

AU - Asch, David A.

AU - Chen, Yong

PY - 2022/12

Y1 - 2022/12

N2 - Linear mixed models are commonly used in healthcare-based association analyses for analyzing multi-site data with heterogeneous site-specific random effects. Due to regulations for protecting patients’ privacy, sensitive individual patient data (IPD) typically cannot be shared across sites. We propose an algorithm for fitting distributed linear mixed models (DLMMs) without sharing IPD across sites. This algorithm achieves results identical to those achieved using pooled IPD from multiple sites (i.e., the same effect size and standard error estimates), hence demonstrating the lossless property. The algorithm requires each site to contribute minimal aggregated data in only one round of communication. We demonstrate the lossless property of the proposed DLMM algorithm by investigating the associations between demographic and clinical characteristics and length of hospital stay in COVID-19 patients using administrative claims from the UnitedHealth Group Clinical Discovery Database. We extend this association study by incorporating 120,609 COVID-19 patients from 11 collaborative data sources worldwide.

AB - Linear mixed models are commonly used in healthcare-based association analyses for analyzing multi-site data with heterogeneous site-specific random effects. Due to regulations for protecting patients’ privacy, sensitive individual patient data (IPD) typically cannot be shared across sites. We propose an algorithm for fitting distributed linear mixed models (DLMMs) without sharing IPD across sites. This algorithm achieves results identical to those achieved using pooled IPD from multiple sites (i.e., the same effect size and standard error estimates), hence demonstrating the lossless property. The algorithm requires each site to contribute minimal aggregated data in only one round of communication. We demonstrate the lossless property of the proposed DLMM algorithm by investigating the associations between demographic and clinical characteristics and length of hospital stay in COVID-19 patients using administrative claims from the UnitedHealth Group Clinical Discovery Database. We extend this association study by incorporating 120,609 COVID-19 patients from 11 collaborative data sources worldwide.

UR - http://www.scopus.com/inward/record.url?scp=85127245114&partnerID=8YFLogxK

U2 - 10.1038/s41467-022-29160-4

DO - 10.1038/s41467-022-29160-4

M3 - Article

C2 - 35354802

AN - SCOPUS:85127245114

SN - 2041-1723

VL - 13

JO - Nature communications

JF - Nature communications

IS - 1

M1 - 1678

ER -

DLMM as a lossless one-shot algorithm for collaborative multi-site distributed linear mixed models

Abstract

Access to Document

Other files and links

Fingerprint

Cite this