Patient-Driven Privacy Control through Generalized Distillation

Z. Berkay Celik, David Lopez-Paz, Patrick Drew McDaniel

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Abstract

The introduction of data analytics into medicine has changed the nature of patient treatment. In this, patients are asked to disclose personal information such as genetic markers, lifestyle habits, and clinical history. This data is then used by statistical models to predict personalized treatments. However, due to privacy concerns, patients often desire to withhold sensitive information. This self-censorship can impede proper diagnosis and treatment, which may lead to serious health complications and even death over time. In this paper, we present privacy distillation, a mechanism which allows patients to control the type and amount of information they wish to disclose to the healthcare providers for use in statistical models. Meanwhile, it retains the accuracy of models that have access to all patient data under a sufficient but not full set of privacy-relevant information. We validate privacy distillation using a corpus of patients prescribed to warfarin for a personalized dosage. We use a deep neural network to implement privacy distillation for training and making dose predictions. We find that privacy distillation withsufficient privacy-relevant information i) retains accuracy almost as good as having all patient data (only 3% worse), and ii) is effective at preventing errors that introduce health-related risks (only 3.9% worse under-or over-prescriptions).

Original languageEnglish (US)
Title of host publicationProceedings - 2017 IEEE Symposium on Privacy-Aware Computing, PAC 2017
PublisherInstitute of Electrical and Electronics Engineers Inc.
Pages1-12
Number of pages12
ISBN (Electronic)9781538610275
DOIs
StatePublished - Dec 4 2017
Event1st IEEE Symposium on Privacy-Aware Computing, PAC 2017 - Washington, United States
Duration: Aug 1 2017Aug 3 2017

Publication series

NameProceedings - 2017 IEEE Symposium on Privacy-Aware Computing, PAC 2017
Volume2017-January

Other

Other1st IEEE Symposium on Privacy-Aware Computing, PAC 2017
CountryUnited States
CityWashington
Period8/1/178/3/17

Fingerprint

Distillation
Health
Patient treatment
Medicine
Statistical Models

All Science Journal Classification (ASJC) codes

  • Computer Networks and Communications
  • Safety, Risk, Reliability and Quality

Cite this

Celik, Z. B., Lopez-Paz, D., & McDaniel, P. D. (2017). Patient-Driven Privacy Control through Generalized Distillation. In Proceedings - 2017 IEEE Symposium on Privacy-Aware Computing, PAC 2017 (pp. 1-12). (Proceedings - 2017 IEEE Symposium on Privacy-Aware Computing, PAC 2017; Vol. 2017-January). Institute of Electrical and Electronics Engineers Inc.. https://doi.org/10.1109/PAC.2017.13
Celik, Z. Berkay ; Lopez-Paz, David ; McDaniel, Patrick Drew. / Patient-Driven Privacy Control through Generalized Distillation. Proceedings - 2017 IEEE Symposium on Privacy-Aware Computing, PAC 2017. Institute of Electrical and Electronics Engineers Inc., 2017. pp. 1-12 (Proceedings - 2017 IEEE Symposium on Privacy-Aware Computing, PAC 2017).
@inproceedings{a293255fa6b244ac8a8c45d5225cbd47,
title = "Patient-Driven Privacy Control through Generalized Distillation",
abstract = "The introduction of data analytics into medicine has changed the nature of patient treatment. In this, patients are asked to disclose personal information such as genetic markers, lifestyle habits, and clinical history. This data is then used by statistical models to predict personalized treatments. However, due to privacy concerns, patients often desire to withhold sensitive information. This self-censorship can impede proper diagnosis and treatment, which may lead to serious health complications and even death over time. In this paper, we present privacy distillation, a mechanism which allows patients to control the type and amount of information they wish to disclose to the healthcare providers for use in statistical models. Meanwhile, it retains the accuracy of models that have access to all patient data under a sufficient but not full set of privacy-relevant information. We validate privacy distillation using a corpus of patients prescribed to warfarin for a personalized dosage. We use a deep neural network to implement privacy distillation for training and making dose predictions. We find that privacy distillation withsufficient privacy-relevant information i) retains accuracy almost as good as having all patient data (only 3{\%} worse), and ii) is effective at preventing errors that introduce health-related risks (only 3.9{\%} worse under-or over-prescriptions).",
author = "Celik, {Z. Berkay} and David Lopez-Paz and McDaniel, {Patrick Drew}",
year = "2017",
month = "12",
day = "4",
doi = "10.1109/PAC.2017.13",
language = "English (US)",
series = "Proceedings - 2017 IEEE Symposium on Privacy-Aware Computing, PAC 2017",
publisher = "Institute of Electrical and Electronics Engineers Inc.",
pages = "1--12",
booktitle = "Proceedings - 2017 IEEE Symposium on Privacy-Aware Computing, PAC 2017",
address = "United States",

}

Celik, ZB, Lopez-Paz, D & McDaniel, PD 2017, Patient-Driven Privacy Control through Generalized Distillation. in Proceedings - 2017 IEEE Symposium on Privacy-Aware Computing, PAC 2017. Proceedings - 2017 IEEE Symposium on Privacy-Aware Computing, PAC 2017, vol. 2017-January, Institute of Electrical and Electronics Engineers Inc., pp. 1-12, 1st IEEE Symposium on Privacy-Aware Computing, PAC 2017, Washington, United States, 8/1/17. https://doi.org/10.1109/PAC.2017.13

Patient-Driven Privacy Control through Generalized Distillation. / Celik, Z. Berkay; Lopez-Paz, David; McDaniel, Patrick Drew.

Proceedings - 2017 IEEE Symposium on Privacy-Aware Computing, PAC 2017. Institute of Electrical and Electronics Engineers Inc., 2017. p. 1-12 (Proceedings - 2017 IEEE Symposium on Privacy-Aware Computing, PAC 2017; Vol. 2017-January).

Research output: Chapter in Book/Report/Conference proceedingConference contribution

TY - GEN

T1 - Patient-Driven Privacy Control through Generalized Distillation

AU - Celik, Z. Berkay

AU - Lopez-Paz, David

AU - McDaniel, Patrick Drew

PY - 2017/12/4

Y1 - 2017/12/4

N2 - The introduction of data analytics into medicine has changed the nature of patient treatment. In this, patients are asked to disclose personal information such as genetic markers, lifestyle habits, and clinical history. This data is then used by statistical models to predict personalized treatments. However, due to privacy concerns, patients often desire to withhold sensitive information. This self-censorship can impede proper diagnosis and treatment, which may lead to serious health complications and even death over time. In this paper, we present privacy distillation, a mechanism which allows patients to control the type and amount of information they wish to disclose to the healthcare providers for use in statistical models. Meanwhile, it retains the accuracy of models that have access to all patient data under a sufficient but not full set of privacy-relevant information. We validate privacy distillation using a corpus of patients prescribed to warfarin for a personalized dosage. We use a deep neural network to implement privacy distillation for training and making dose predictions. We find that privacy distillation withsufficient privacy-relevant information i) retains accuracy almost as good as having all patient data (only 3% worse), and ii) is effective at preventing errors that introduce health-related risks (only 3.9% worse under-or over-prescriptions).

AB - The introduction of data analytics into medicine has changed the nature of patient treatment. In this, patients are asked to disclose personal information such as genetic markers, lifestyle habits, and clinical history. This data is then used by statistical models to predict personalized treatments. However, due to privacy concerns, patients often desire to withhold sensitive information. This self-censorship can impede proper diagnosis and treatment, which may lead to serious health complications and even death over time. In this paper, we present privacy distillation, a mechanism which allows patients to control the type and amount of information they wish to disclose to the healthcare providers for use in statistical models. Meanwhile, it retains the accuracy of models that have access to all patient data under a sufficient but not full set of privacy-relevant information. We validate privacy distillation using a corpus of patients prescribed to warfarin for a personalized dosage. We use a deep neural network to implement privacy distillation for training and making dose predictions. We find that privacy distillation withsufficient privacy-relevant information i) retains accuracy almost as good as having all patient data (only 3% worse), and ii) is effective at preventing errors that introduce health-related risks (only 3.9% worse under-or over-prescriptions).

UR - http://www.scopus.com/inward/record.url?scp=85046550197&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85046550197&partnerID=8YFLogxK

U2 - 10.1109/PAC.2017.13

DO - 10.1109/PAC.2017.13

M3 - Conference contribution

T3 - Proceedings - 2017 IEEE Symposium on Privacy-Aware Computing, PAC 2017

SP - 1

EP - 12

BT - Proceedings - 2017 IEEE Symposium on Privacy-Aware Computing, PAC 2017

PB - Institute of Electrical and Electronics Engineers Inc.

ER -

Celik ZB, Lopez-Paz D, McDaniel PD. Patient-Driven Privacy Control through Generalized Distillation. In Proceedings - 2017 IEEE Symposium on Privacy-Aware Computing, PAC 2017. Institute of Electrical and Electronics Engineers Inc. 2017. p. 1-12. (Proceedings - 2017 IEEE Symposium on Privacy-Aware Computing, PAC 2017). https://doi.org/10.1109/PAC.2017.13