SurvMaximin: Robust federated approach to transporting survival risk prediction models
Xuan Wang
(1)
,
Harrison Zhang
(2)
,
Xin Xiong
(1)
,
Chuan Hong
(2)
,
Griffin Weber
(2)
,
Gabriel Brat
(2)
,
Clara-Lea Bonzel
(2)
,
Yuan Luo
(3)
,
Rui Duan
(4)
,
Nathan Palmer
(2)
,
Meghan Hutch
(3)
,
Alba Gutiérrez-Sacristán
(2)
,
Riccardo Bellazzi
(5)
,
Luca Chiovato
(6)
,
Kelly Cho
(7)
,
Arianna Dagliati
(5)
,
Hossein Estiri
(8)
,
Noelia García-Barrio
(9)
,
Romain Griffier
(10, 11)
,
David Hanauer
(12)
,
Yuk-Lam Ho
(7)
,
John Holmes
(13)
,
Mark Keller
(2)
,
Jeffrey Klann Meng
(8)
,
Sehi L'Yi
(2)
,
Sara Lozano-Zahonero
(14)
,
Sarah Maidlow
(12)
,
Adeline Makoudjou
(14)
,
Alberto Malovini
(6)
,
Bertrand Moal
(11)
,
Jason Moore
(15)
,
Michele Morris
(16)
,
Danielle Mowery
(13)
,
Shawn Murphy
(8)
,
Antoine Neuraz
(17, 18)
,
Kee Yuan Ngiam
(19)
,
Gilbert Omenn
(12)
,
Lav Patel
(20)
,
Miguel Pedrera-Jiménez
(9)
,
Andrea Prunotto
(14)
,
Malarkodi Jebathilagam Samayamuthu
(16)
,
Fernando Sanz Vidorreta
(21)
,
Emily Schriver
(13)
,
Petra Schubert
(7)
,
Pablo Serrano-Balazote
(9)
,
Andrew South
(22)
,
Amelia L.M. Tan
(2)
,
Byorn W.L. Tan
(23)
,
Valentina Tibollo
(14)
,
Patric Tippmann
(14)
,
Shyam Visweswaran
(16)
,
Zongqi Xia
(16)
,
William Yuan
(2)
,
Daniela Zöller
(14)
,
Isaac Kohane
(2)
,
Paul Avillach
(2)
,
Zijian Guo
(24)
,
Tianxi Cai
(2)
1
Harvard T.H. Chan School of Public Health
2 HMS - Harvard Medical School [Boston]
3 Northwestern University [Chicago, Ill. USA]
4 Harvard University
5 UNIPV - Università degli Studi di Pavia = University of Pavia
6 IRCCS Pavia - ICS Maugeri - Istituti Clinici Scientifici Maugeri [Pavia]
7 VA Boston Healthcare System
8 Massachusetts General Hospital [Boston]
9 Hospital Universitario 12 de Octubre [Madrid]
10 BPH - Bordeaux population health
11 CHU Bordeaux
12 University of Michigan [Ann Arbor]
13 University of Pennsylvania
14 University of Freiburg [Freiburg]
15 Cedars-Sinai Medical Center
16 PITT - University of Pittsburgh
17 UPCité - Université Paris Cité
18 Service d'informatique médicale et biostatistiques [CHU Necker]
19 NUHS - National University Health System [Singapore]
20 University of Kansas Medical Center [Kansas City, KS, USA]
21 David Geffen School of Medicine [Los Angeles]
22 Wake Forest School of Medicine [Winston-Salem]
23 NUH - National University Hospital [Singapore]
24 RU - Rutgers, The State University of New Jersey [New Brunswick]
2 HMS - Harvard Medical School [Boston]
3 Northwestern University [Chicago, Ill. USA]
4 Harvard University
5 UNIPV - Università degli Studi di Pavia = University of Pavia
6 IRCCS Pavia - ICS Maugeri - Istituti Clinici Scientifici Maugeri [Pavia]
7 VA Boston Healthcare System
8 Massachusetts General Hospital [Boston]
9 Hospital Universitario 12 de Octubre [Madrid]
10 BPH - Bordeaux population health
11 CHU Bordeaux
12 University of Michigan [Ann Arbor]
13 University of Pennsylvania
14 University of Freiburg [Freiburg]
15 Cedars-Sinai Medical Center
16 PITT - University of Pittsburgh
17 UPCité - Université Paris Cité
18 Service d'informatique médicale et biostatistiques [CHU Necker]
19 NUHS - National University Health System [Singapore]
20 University of Kansas Medical Center [Kansas City, KS, USA]
21 David Geffen School of Medicine [Los Angeles]
22 Wake Forest School of Medicine [Winston-Salem]
23 NUH - National University Hospital [Singapore]
24 RU - Rutgers, The State University of New Jersey [New Brunswick]
Xuan Wang
- Fonction : Auteur
- PersonId : 1198789
- ORCID : 0000-0002-0011-5663
Xin Xiong
- Fonction : Auteur
- PersonId : 1197157
- ORCID : 0000-0002-1162-5220
Griffin Weber
- Fonction : Auteur
- PersonId : 818086
- ORCID : 0000-0002-2597-881X
Gabriel Brat
- Fonction : Auteur
- PersonId : 818068
- ORCID : 0000-0003-3928-5931
Clara-Lea Bonzel
- Fonction : Auteur
- PersonId : 818067
- ORCID : 0000-0001-5238-1413
Nathan Palmer
- Fonction : Auteur
- PersonId : 818081
- ORCID : 0000-0002-4361-207X
Alba Gutiérrez-Sacristán
- Fonction : Auteur
- PersonId : 818061
- ORCID : 0000-0002-1245-198X
Riccardo Bellazzi
- Fonction : Auteur
- PersonId : 763117
- ORCID : 0000-0002-6974-9808
- IdRef : 204354269
Kelly Cho
- Fonction : Auteur
- PersonId : 818089
- ORCID : 0000-0003-1727-7076
Arianna Dagliati
- Fonction : Auteur
- PersonId : 818069
- ORCID : 0000-0002-5041-0409
Hossein Estiri
- Fonction : Auteur
- PersonId : 806516
- ORCID : 0000-0002-0204-8978
Noelia García-Barrio
- Fonction : Auteur
- PersonId : 818071
- ORCID : 0000-0002-2789-8426
David Hanauer
- Fonction : Auteur
- PersonId : 818072
- ORCID : 0000-0001-6931-3791
Yuk-Lam Ho
- Fonction : Auteur
- PersonId : 818094
- ORCID : 0000-0003-3305-3830
John Holmes
- Fonction : Auteur
- PersonId : 794711
- ORCID : 0000-0003-2167-3602
Sara Lozano-Zahonero
- Fonction : Auteur
- PersonId : 818090
- ORCID : 0000-0002-9467-0018
Sarah Maidlow
- Fonction : Auteur
- PersonId : 818078
- ORCID : 0000-0003-4421-6834
Michele Morris
- Fonction : Auteur
- PersonId : 806517
- ORCID : 0000-0002-3255-5727
Antoine Neuraz
- Fonction : Auteur
- PersonId : 1122536
- IdHAL : antoine-neuraz
Gilbert Omenn
- Fonction : Auteur
- PersonId : 794562
- ORCID : 0000-0002-8976-6074
- IdRef : 077657802
Lav Patel
- Fonction : Auteur
- PersonId : 818063
- ORCID : 0000-0002-8626-137X
Miguel Pedrera-Jiménez
- Fonction : Auteur
- PersonId : 818097
- ORCID : 0000-0003-0187-3826
Emily Schriver
- Fonction : Auteur
- PersonId : 806518
- ORCID : 0000-0003-4522-1029
Petra Schubert
- Fonction : Auteur
- PersonId : 818091
- ORCID : 0000-0002-6135-0173
Pablo Serrano-Balazote
- Fonction : Auteur
- PersonId : 818100
- ORCID : 0000-0003-2958-6319
Andrew South
- Fonction : Auteur
- PersonId : 802114
- ORCID : 0000-0002-3204-4142
Valentina Tibollo
- Fonction : Auteur
- PersonId : 802107
- ORCID : 0000-0003-2687-3822
Patric Tippmann
- Fonction : Auteur
- PersonId : 1198790
- ORCID : 0000-0001-6521-8607
Daniela Zöller
- Fonction : Auteur
- PersonId : 818102
- ORCID : 0000-0001-9929-7403
Paul Avillach
- Fonction : Auteur
- PersonId : 803243
- ORCID : 0000-0002-0235-7543
Tianxi Cai
- Fonction : Auteur
- PersonId : 785101
- ORCID : 0000-0002-5379-2502
Résumé
Objective
For multi-center heterogeneous Real-World Data (RWD) with time-to-event outcomes and high-dimensional features, we propose the SurvMaximin algorithm to estimate Cox model feature coefficients for a target population by borrowing summary information from a set of health care centers without sharing patient-level information.
Materials and Methods
For each of the centers from which we want to borrow information to improve the prediction performance for the target population, a penalized Cox model is fitted to estimate feature coefficients for the center. Using estimated feature coefficients and the covariance matrix of the target population, we then obtain a SurvMaximin estimated set of feature coefficients for the target population. The target population can be an entire cohort comprised of all centers, corresponding to federated learning, or can be a single center, corresponding to transfer learning.
Results
Simulation studies and a real-world international electronic health records application study, with 15 participating health care centers across three countries (France, Germany, and the U.S.), show that the proposed SurvMaximin algorithm achieves comparable or higher accuracy compared with the estimator using only the information of the target site and other existing methods. The SurvMaximin estimator is robust to variations in sample sizes and estimated feature coefficients between centers, which amounts to significantly improved estimates for target sites with fewer observations.
Conclusions
The SurvMaximin method is well suited for both federated and transfer learning in the high-dimensional survival analysis setting. SurvMaximin only requires a one-time summary information exchange from participating centers. Estimated regression vectors can be very heterogeneous. SurvMaximin provides robust Cox feature coefficient estimates without outcome information in the target population and is privacy-preserving.