Can AI Reduce Unnecessary Medical Testing?

Author: Rasit Dinc

Artificial intelligence (AI) is rapidly transforming the healthcare landscape, offering unprecedented opportunities to enhance diagnostic accuracy, streamline workflows, and improve patient outcomes. One of the most promising applications of AI in medicine is its potential to reduce unnecessary medical testing. Over-testing is a significant concern in modern healthcare, leading to increased costs, patient anxiety, and the risk of false positives. This article explores how AI can be leveraged to address this challenge, drawing on recent research to highlight the benefits and limitations of this technology.

The Core Problem: Unnecessary Medical Testing

Unnecessary medical tests contribute to a significant portion of healthcare spending and can lead to a cascade of negative consequences. They expose patients to potential harm, cause undue anxiety, and strain limited healthcare resources. The reasons for over-testing are multifaceted, ranging from defensive medicine and patient demand to diagnostic uncertainty. Addressing this issue requires a shift towards more precise and evidence-based diagnostic strategies, a domain where AI is poised to make a substantial impact.

AI-Powered Diagnostic Stewardship

Diagnostic stewardship refers to the coordinated effort to use diagnostic tests judiciously to improve patient outcomes. AI can serve as a powerful tool for diagnostic stewardship by providing data-driven insights at the point of care. For instance, machine learning models can analyze vast datasets of electronic health records (EHRs) to identify patients who are at low risk for certain conditions, thereby helping clinicians avoid unnecessary tests. A 2022 study on the use of machine learning to predict blood culture outcomes demonstrated that an AI model could have prevented 20% of negative cultures from being ordered, showcasing the potential for significant resource savings and improved patient care [4].

Enhancing Accuracy and Reducing Workload

One of the primary ways AI can reduce unnecessary testing is by improving the accuracy of initial diagnoses. In fields like radiology and pathology, which involve the interpretation of complex images and large volumes of data, AI has demonstrated remarkable success. A comprehensive review published in 2025 found that in these specialties, AI not only improved diagnostic accuracy but also reduced diagnostic time by as much as 90% [1].

Furthermore, a multicenter study on hepatocellular carcinoma (HCC) screening highlighted a human-AI collaboration strategy that significantly enhanced performance. By using AI for initial lesion detection and classification, with radiologists reviewing only the negative or ambiguous cases, the system improved specificity, reduced the radiologist workload by 54.5%, and decreased both recall and false-positive rates [2]. This reduction in false positives directly translates to fewer unnecessary follow-up tests, mitigating patient anxiety and reducing the burden on the healthcare system.

Cost-Effectiveness and Future Directions

The integration of AI into diagnostic workflows is not only about improving clinical outcomes but also about enhancing cost-effectiveness. The Microsoft AI Diagnostic Orchestrator (MAI-DxO), for example, has been shown to deliver both higher diagnostic accuracy and lower overall testing costs compared to physicians or individual AI models [3]. By configuring AI systems to operate within specific cost constraints, healthcare providers can explicitly manage the trade-offs between cost and diagnostic value, preventing the system from defaulting to ordering every possible test.

Conclusion

The evidence strongly suggests that artificial intelligence can play a pivotal role in reducing unnecessary medical testing. By enhancing diagnostic accuracy, supporting clinical decision-making, and optimizing resource allocation, AI-powered tools offer a pathway to a more efficient, cost-effective, and patient-centered healthcare system. While challenges related to implementation, standardization, and ethical considerations remain, the continued development and thoughtful integration of AI in diagnostic stewardship hold the promise of transforming medical practice for the better.

References

[1] Jeong, J., et al. (2025). Reducing the workload of medical diagnosis through artificial intelligence: A narrative review. Medicine, 104(6), e41470. https://pmc.ncbi.nlm.nih.gov/articles/PMC11813001/

[2] Lu, R. F., et al. (2025). AI enhanced diagnostic accuracy and workload reduction in hepatocellular carcinoma screening. npj Digital Medicine, 8(1), 1-11. https://www.nature.com/articles/s41746-025-01892-9

[3] King, D., & Nori, H. (2025). The Path to Medical Superintelligence. Microsoft AI News. https://microsoft.ai/news/the-path-to-medical-superintelligence/

[4] Schinkel, M., et al. (2022). Diagnostic stewardship for blood cultures in the emergency department: a multicenter validation and prospective evaluation of a machine learning prediction model. The Lancet Digital Health, 4(10), e739-e748. https://www.thelancet.com/journals/ebiom/article/PIIS2352-3964(22)00357-7/fulltext