Deconfounding age effects with fair representation learning when assessing dementia

TL;DR

This paper introduces a fair representation learning approach to remove age-related confounding effects from linguistic features used in dementia detection, improving model generalizability across clinical datasets.

Contribution

It proposes neural network classifiers that learn age-invariant representations for dementia detection, outperforming traditional statistical deconfounding methods.

Findings

Models achieve only 2.56% and 1.54% accuracy loss compared to DNNs.

Deconfounded models outperform statistical adjustment methods.

Proposed approach effectively reduces age bias in linguistic features.

Abstract

One of the most prevalent symptoms among the elderly population, dementia, can be detected by classifiers trained on linguistic features extracted from narrative transcripts. However, these linguistic features are impacted in a similar but different fashion by the normal aging process. Aging is therefore a confounding factor, whose effects have been hard for machine learning classifiers (especially deep neural network based models) to ignore. We show DNN models are capable of estimating ages based on linguistic features. Predicting dementia based on this aging bias could lead to potentially non-generalizable accuracies on clinical datasets, if not properly deconfounded. In this paper, we propose to address this deconfounding problem with fair representation learning. We build neural network classifiers that learn low-dimensional representations reflecting the impacts of dementia yet discarding the effects of age. To evaluate these classifiers, we specify a model-agnostic score $\Delta_{eo}^{(N)}$ measuring how classifier results are deconfounded from age. Our best models compromise accuracy by only 2.56\% and 1.54\% on two clinical datasets compared to DNNs, and their $\Delta_{eo}^{(2)}$ scores are better than statistical (residulization and inverse probability weight) adjustments.