The Impact of Artificial Intelligence on Health Equity in Oncology: A Scoping Review
Bibliographic record
Abstract
Abstract Introduction: The expanding use of Artificial Intelligence (AI) in hematology and oncology research and practice creates an urgent need to consider the potential impact of these technologies on health equity at both local and global levels. Fairness and equity are issues of growing concern in AI ethics, raising problems ranging from bias in datasets and algorithms to disparities in access to technology. The impact of AI on health equity in oncology, however, remains underexplored. We conducted a scoping review to characterize, evaluate, and identify gaps in the existing literature on AI applications in oncology and their implications for health equity in cancer care. Methodology: We performed a systematic literature search of MEDLINE (Ovid) and EMBASE from January 1, 2000 to March 28, 2021 using key terms for AI, health equity, and cancer. Our search was restricted to English language abstracts with no restrictions by publication type. Two reviewers screened a total of 9519 abstracts, and 321 met inclusion criteria for full-text review. 247 were included in the final analysis after assessment by three reviewers. Studies were analysed descriptively, by location, type of cancer and AI application, as well as thematically, based on issues pertaining to health equity in oncology. Results: Of the 247 studies included in our analysis, 150 (60.7%) were based in North America, 57 (23.0%) in Asia, 29 (11.7%) in Europe, 5 (2.1%) in Central/South America, 4 (1.6%) in Oceania, and 2 (0.9%) in Africa. 71 (28.6%) were reviews and commentaries, and 176 were (71.3%) clinical studies. 25 (10.1%) focused on AI applications in screening, 42 (17.0%) in diagnostics, 46 (18.6%) in prognostication, and 7 (2.9%) in treatment. 40 (16.3%) used AI as a tool for clinical/epidemiological research and 87 (35.2%) discussed multiple applications of AI. A diverse range of cancers were represented in the studies, including hematologic malignancies. Our scoping review identified three overarching themes in the literature, which largely focused on how AI might improve health equity in oncology. These included: (1) the potential for AI reduce disparities by improving access to health services in resource-limited settings through applications such as low-cost cancer screening technologies and decision support systems; (2) the ability of AI to mitigate bias in clinical decision-making through algorithms that alert clinicians to potential sources of bias thereby allowing for more equitable and individualized care; (3) the use of AI as a research tool to identify disparities in cancer outcomes based on factors such as race, gender and socioeconomic status, and thus inform health policy. While most of the literature emphasized the positive impact of AI in oncology, there was only limited discussion of AI's potential adverse effects on health equity . Despite engaging with the use of AI in resource-limited settings, ethical issues surrounding data extraction and AI trials in low-resource settings were infrequently raised. Similarly, AI's potential to reinforce bias and widen disparities in cancer care was under-examined despite engagement with related topics of bias in clinical decision-making. Conclusion: The overwhelming majority of the literature identified by our scoping review highlights the benefits of AI applications in oncology, including its potential to improve access to care in low-resource settings, mitigate bias in clinical decision-making, and identify disparities in cancer outcomes. However, AI's potential negative impacts on health equity in oncology remain underexplored: ethical issues arising from deploying AI technologies in low-resources settings, and issues of bias in datasets and algorithms were infrequently discussed in articles dealing with related themes. Disclosures No relevant conflicts of interest to declare.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.043 | 0.222 |
| Meta-epidemiology (narrow) | 0.002 | 0.001 |
| Meta-epidemiology (broad) | 0.006 | 0.006 |
| Bibliometrics | 0.027 | 0.025 |
| Science and technology studies | 0.002 | 0.003 |
| Scholarly communication | 0.008 | 0.007 |
| Open science | 0.003 | 0.004 |
| Research integrity | 0.005 | 0.003 |
| Insufficient payload (model declined to judge) | 0.005 | 0.000 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".