Chasing Your Long Tails: Differentially Private Prediction in Health Care Settings

Abstract

Machine learning models in health care are often deployed in settings where it is important to protect patient privacy. In such settings, methods for differentially private (DP) learning provide a general-purpose approach to learn models with privacy guarantees. Modern methods for DP learning ensure privacy through mechanisms that censor information judged as too unique. The resulting privacy-preserving models, therefore, neglect information from the tails of a data distribution, resulting in a loss of accuracy that can disproportionately affect small groups. In this paper, we study the effects of DP learning in health care. We use state-of-the-art methods for DP learning to train privacy-preserving models in clinical prediction tasks, including x-ray classification of images and mortality prediction in time series data. We use these models to perform a comprehensive empirical investigation of the tradeoffs between privacy, utility, robustness to dataset shift, and fairness. Our results highlight lesser-known limitations of methods for DP learning in health care, models that exhibit steep tradeoffs between privacy and utility, and models whose predictions are disproportionately influenced by large demographic groups in the training data. We discuss the costs and benefits of differentially private learning in health care.

References in corpus (28)

• Scikit-learn: Machine Learning in Python
• Deep Learning with Differential Privacy
• Fashion-MNIST: a Novel Image Dataset for Benchmarking Machine Learning Algorithms
• Equality of Opportunity in Supervised Learning
• Understanding Black-box Predictions via Influence Functions
• Understanding deep learning requires rethinking generalization
• COVID-19 Image Data Collection: Prospective Predictions Are the Future
• Preventing Fairness Gerrymandering: Auditing and Learning for Subgroup Fairness
• A Review of Challenges and Opportunities in Machine Learning for Health
• Failing Loudly: An Empirical Study of Methods for Detecting Dataset Shift
• Privacy Loss in Apple's Implementation of Differential Privacy on MacOS 10.12
• An Empirical Characterization of Fair Machine Learning For Clinical Risk Prediction
• Collecting and Analyzing Data from Smart Device Users with Local Differential Privacy
• On the Accuracy of Influence Functions for Measuring Group Effects
• Feature Robustness in Non-stationary Health Records: Caveats to Deployable Model Performance in Common Clinical Machine Learning Tasks
• Auditing Differentially Private Machine Learning: How Private is Private SGD?
• Understanding Gradient Clipping in Private SGD: A Geometric Perspective
• Federated and Differentially Private Learning for Electronic Health Records
• Privacy-Preserving Distributed Deep Learning for Clinical Data
• On the Generalization Properties of Differential Privacy
• Individual Privacy Accounting via a Renyi Filter
• Preventing Failures Due to Dataset Shift: Learning Predictive Models That Transport
• Evading Curse of Dimensionality in Unconstrained Private GLMs via Private Gradient Descent
• Private Reinforcement Learning with PAC and Regret Guarantees
• Neither Private Nor Fair: Impact of Data Imbalance on Utility and Fairness in Differential Privacy
• Privacy for the Protected (Only)
• Hide-and-Seek Privacy Challenge
• Modeling the Biological Pathology Continuum with HSIC-regularized Wasserstein Auto-encoders

Cited by in corpus (2)

• Differential Privacy and Fairness in Decisions and Learning Tasks: A Survey
• Toward Operationalizing Pipeline-aware ML Fairness: A Research Agenda for Developing Practical Guidelines and Tools