# When More is Less: Incorporating Additional Datasets Can Hurt Performance By Introducing Spurious Correlations

arXiv: https://arxiv.org/abs/2308.04431 · HTML: https://mlsanity.com/paper/2308.04431/

## Key facts

- Cited by 18 papers in the map (data through 14 Oct 2026).
- Builds on 24 of its 29 references that are in the map.
- First posted on arXiv 8 Aug 2023.

## Builds on (most cited first, 24 of 24)

- [Densely Connected Convolutional Networks](https://mlsanity.com/paper/1608.06993/) (2016)
- [Shortcut Learning in Deep Neural Networks](https://mlsanity.com/paper/2004.07780/) (2020)
- [CheXpert: A Large Chest Radiograph Dataset with Uncertainty Labels and Expert Comparison](https://mlsanity.com/paper/1901.07031/) (2019)
- [ChestX-ray8: Hospital-scale Chest X-ray Database and Benchmarks on Weakly-Supervised Classification and Localization of Common Thorax Diseases](https://mlsanity.com/paper/1705.02315/) (2017)
- [Revisiting Unreasonable Effectiveness of Data in Deep Learning Era](https://mlsanity.com/paper/1707.02968/) (2017)
- [MIMIC-CXR-JPG, a large publicly available database of labeled chest radiographs](https://mlsanity.com/paper/1901.07042/) (2019)
- [Last Layer Re-Training is Sufficient for Robustness to Spurious Correlations](https://mlsanity.com/paper/2204.02937/) (2022)
- [Confounding variables can degrade generalization performance of radiological deep learning models](https://mlsanity.com/paper/1807.00431/) (2018)
- [Transfusion: Understanding Transfer Learning for Medical Imaging](https://mlsanity.com/paper/1902.07208/) (2019)
- [Unified Deep Supervised Domain Adaptation and Generalization](https://mlsanity.com/paper/1709.10190/) (2017)
- [An Investigation of Why Overparameterization Exacerbates Spurious Correlations](https://mlsanity.com/paper/2005.04345/) (2020)
- [PadChest: A large chest x-ray image dataset with multi-label annotated reports](https://mlsanity.com/paper/1901.07441/) (2019)
- [Simple data balancing achieves competitive worst-group-accuracy](https://mlsanity.com/paper/2110.14503/) (2021)
- [Domain Adaptation for Visual Applications: A Comprehensive Survey](https://mlsanity.com/paper/1702.05374/) (2017)
- [Few-Shot Adversarial Domain Adaptation](https://mlsanity.com/paper/1711.02536/) (2017)
- [On the limits of cross-domain generalization in automated X-ray prediction](https://mlsanity.com/paper/2002.02497/) (2020)
- [Can we trust deep learning models diagnosis? The impact of domain shift in chest radiograph classification](https://mlsanity.com/paper/1909.01940/) (2019)
- [Deep Learning Predicts Hip Fracture using Confounding Patient and Healthcare Variables](https://mlsanity.com/paper/1811.03695/) (2018)
- [Out-of-distribution Generalization in the Presence of Nuisance-Induced Spurious Correlations](https://mlsanity.com/paper/2107.00520/) (2021)
- [An Empirical Framework for Domain Generalization in Clinical Settings](https://mlsanity.com/paper/2103.11163/) (2021)
- [DeepSleepNet-Lite: A Simplified Automatic Sleep Stage Scoring Model with Uncertainty Estimates](https://mlsanity.com/paper/2108.10600/) (2021)
- [Comparing Different Deep Learning Architectures for Classification of Chest Radiographs](https://mlsanity.com/paper/2002.08991/) (2020)
- [Undersampling is a Minimax Optimal Robustness Intervention in Nonparametric Classification](https://mlsanity.com/paper/2205.13094/) (2022)
- [Representation Learning with Information Theory for COVID-19 Detection](https://mlsanity.com/paper/2207.01437/) (2022)

## Built on by (most cited first, 18 of 18)

- [On the Challenges and Opportunities in Generative AI](https://mlsanity.com/paper/2403.00025/) (2024)
- [The Data Addition Dilemma](https://mlsanity.com/paper/2408.04154/) (2024)
- [Mind the Graph When Balancing Data for Fairness or Robustness](https://mlsanity.com/paper/2406.17433/) (2024)
- [In the Picture: Medical Imaging Datasets, Artifacts, and their Living Review](https://mlsanity.com/paper/2501.10727/) (2025)
- [Copycats: the many lives of a publicly available medical imaging dataset](https://mlsanity.com/paper/2402.06353/) (2024)
- [DataS^3: Dataset Subset Selection for Specialization](https://mlsanity.com/paper/2504.16277/) (2025)
- [Benchmark Profiling: Mechanistic Diagnosis of LLM Benchmarks](https://mlsanity.com/paper/2510.01232/) (2025)
- [Data Debiasing with Datamodels (D3M): Improving Subgroup Robustness via Data Selection](https://mlsanity.com/paper/2406.16846/) (2024)
- [Competing Risks: Impact on Risk Estimation and Algorithmic Fairness](https://mlsanity.com/paper/2508.05435/) (2025)
- [Decoupling Clinical and Class-Agnostic Features for Reliable Few-Shot Adaptation under Shift](https://mlsanity.com/paper/2509.09397/) (2025)
- [Privacy-Preserving Dataset Combination](https://mlsanity.com/paper/2502.05765/) (2025)
- [Residual Feature Integration is Sufficient to Prevent Negative Transfer](https://mlsanity.com/paper/2505.11771/) (2025)
- [Dataset Diversity Metrics and Impact on Classification Models](https://mlsanity.com/paper/2603.15276/) (2026)
- [Learning to Harmonize Cross-vendor X-ray Images by Non-linear Image Dynamics Correction](https://mlsanity.com/paper/2504.10080/) (2025)
- [Benchmark Success, Clinical Failure: When Reinforcement Learning Optimizes for Benchmarks, Not Patients](https://mlsanity.com/paper/2512.23090/) (2025)
- [DeconDTN-Toolkit: A Library for Evaluation and Enhancement of Robustness to Provenance Shift](https://mlsanity.com/paper/2605.11237/) (2026)
- [Identifying Latent Concepts and Structures for Generalized Category Discovery](https://mlsanity.com/paper/2607.00620/) (2026)
- [Team DACTYL at PAN 2026: Bayesian Data Mixing and Empirical X-risk Minimization for AI-text Detection](https://mlsanity.com/paper/2607.17382/) (2026)

## Abstract (arXiv)

In machine learning, incorporating more data is often seen as a reliable strategy for improving model performance; this work challenges that notion by demonstrating that the addition of external datasets in many cases can hurt the resulting model's performance. In a large-scale empirical study across combinations of four different open-source chest x-ray datasets and 9 different labels, we demonstrate that in 43% of settings, a model trained on data from two hospitals has poorer worst group accuracy over both hospitals than a model trained on just a single hospital's data. This surprising result occurs even though the added hospital makes the training distribution more similar to the test distribution. We explain that this phenomenon arises from the spurious correlation that emerges between the disease and hospital, due to hospital-specific image artifacts. We highlight the trade-off one encounters when training on multiple datasets, between the obvious benefit of additional data and insidious cost of the introduced spurious correlation. In some cases, balancing the dataset can remove the spurious correlation and improve performance, but it is not always an effective strategy. We contextualize our results within the literature on spurious correlations to help explain these outcomes. Our experiments underscore the importance of exercising caution when selecting training data for machine learning models, especially in settings where there is a risk of spurious correlations such as with medical imaging. The risks outlined highlight the need for careful data selection and model evaluation in future research and practice.