The DOI matches its Crossref record; 59 of the 59 references it cites were found in a registry or an index; 24 of the 53 sources its claims cite were checked for support.
Systematic review of cost effectiveness and budget impact of artificial intelligence in healthcare
Open accessAbstract
This systematic review examines 19 studies of clinical AI interventions across multiple specialties and finds economic promise, while warning that cost components are often underreported.
Overview
CoChat Summary
## TL;DR
This systematic review of 19 studies found that clinical AI interventions were generally associated with improved diagnostic accuracy, gains in QALYs, and frequent cost savings or favorable ICERs across multiple specialties. The authors caution that many studies used static models and omitted important cost components, so the reported economic value of AI may be overstated.
## Key findings
- The review included 19 studies published from 2020 to 2024 across oncology, atrial fibrillation screening, gastrointestinal endoscopy, diabetic retinopathy screening, dental caries detection, ICU management, medication management, tuberculosis screening/monitoring, glaucoma screening, lung cancer screening, and breast cancer screening.
- Economic evaluation types were CEA (n = 14), CUA (n = 4), CMA (n = 2), and BIA (n = 1); 13 studies used a healthcare system perspective, 4 a societal perspective, and 1 a payer perspective.
- Example results reported in the paper include AF screening ICERs of £4847–£5544 per QALY gained, diabetic retinopathy screening cost reductions of 14–19.5% with ICERs as low as $1107.63 per QALY, and AI-assisted colonoscopy savings of $149.2 million in Japan and $85.2 million in the United States annually.
- Reporting quality was high: CHEERS 2022 scores ranged from 21/28 to 27/28, and Drummond scores ranged from 8/10 to 9/10.
## Methods
- The authors conducted a PRISMA 2020 systematic review of clinical AI economic evaluations, searching PubMed/MEDLINE, Embase, Web of Science, and the Cochrane Library from database inception to January 2025.
- They screened 312 unique records, assessed 50 full-text articles, and included 19 studies after excluding studies for unclear/insufficient economic reporting, ineligible design, or no economic outcomes.
- They extracted study characteristics and economic outcomes using a standardized form, then performed a thematic synthesis and quality appraisal with CHEERS 2022 and the Drummond checklist.
## Limitations
- The paper states that many studies omitted indirect costs, upfront infrastructure costs, integration costs, and ongoing maintenance costs, which may overstate economic benefits.
- The paper states that many evaluations used static decision-analytic models and did not capture adaptive learning or time-dependent performance improvements in AI; this may overestimate cost-effectiveness.
- The paper states that equity was underexplored, with limited subgroup analyses and limited disaggregation by demographic variables.
- Evident from the text: the review synthesized heterogeneous studies across settings, specialties, and methods, which the authors say may limit generalizability.
Model-written summary. Check it against the paper before relying on it.
Integrity signals
Registry
Cites
Statistics
Figures

Verification
Claims the paper makes
21 Valid · 2 Not valid · 17 Not checked · 76 Own results and interpretations
Every source the paper cites for this claim was checked and supports it.
In the paper
“Mori et al.20 reported annual national savings of $149.2 million in Japan and $85.2 million in the United States from an AI-assisted colonoscopy strategy”
§ Systematic review of cost effectiveness and budget impact of artificial intelligence in healthcare > Results > Comparative economic outcomes
What the sources say
- Supportedcites Mori et al.20 Mori, Y. et al. Cost savings in colonoscopy with artificial intelligence-aided polyp diagnosis: an add-on analysis of a clinical trial (with video). Gastrointest. Endosc. 92, 905–911.e1 (2020). source
The Abstract reports estimated reductions in gross annual colonoscopy reimbursement of $149.2 million in Japan and $85.2 million in the United States. The Discussion identifies these as savings from an AI-supported diagnose-and-leave strategy compared with resecting all polyps; they are estimates, not observed national savings. (LLM, conf=0.93)
plainstates a cause- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: unclear — It is unclear whether the paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Supported
The cited sources support only part of this claim — read each source’s verdict below.
In the paper
“Ziegelmayer et al.25 used up to 30,000 Monte Carlo simulations to demonstrate that AI-assisted LDCT remained cost-saving under various scenarios.”
§ Systematic review of cost effectiveness and budget impact of artificial intelligence in healthcare > Results > Comparative economic outcomes
What the sources say
- Partly supportedcites Ziegelmayer et al.25 Ziegelmayer, S., Graf, M., Makowski, M., Gawlitza, J. & Gassert, F. Cost-effectiveness of artificial intelligence support in computed tomography-based lung cancer screening. Cancers 14, 1729 (2022). sourceExperimental — accuracy not yet measured
Methods §2.7 reports a Monte Carlo probabilistic sensitivity analysis with 30,000 iterations, rather than 30,000 separate simulations. Results §3.1 finds lower costs for CT + AI in the base case, but Results §3.2 says the ICER turned positive when specificity varied. Those cases remained cost-effective at the stated willingness-to-pay threshold, not cost-saving. (LLM, conf=0.90)
boostedstates a cause- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: unclear — It is unclear whether the paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Partly supported
The cited sources support only part of this claim — read each source’s verdict below.
In the paper
“the performance differences between semi-automated and fully automated AI models—as shown by Xie et al.23 and Huang et al.36 —directly impacted cost per patient and ICERs”
§ Systematic review of cost effectiveness and budget impact of artificial intelligence in healthcare > Results > Comparative economic outcomes
What the sources say
- Not checked — no full text of the sourcecites Xie et al.23 Xie, Y. et al. Artificial intelligence for teleophthalmology-based diabetic retinopathy screening in a national programme: an economic analysis modelling study. Lancet Digit Health 2, e240–e249 (2020). source
- Partly supportedcites Huang et al.36 Huang, X. M. et al. Cost-effectiveness of artificial intelligence screening for diabetic retinopathy in rural China. BMC Health Serv. Res 22, 1–12 (2022). sourceExperimental — accuracy not yet measured
The Discussion (page unavailable) reports lower per-patient grading costs for AI than for ophthalmologist screening, and the Results > Cost-effectiveness analysis reports an ICER for AI screening. The paper does not compare semi-automated with fully automated AI models or show that performance differences between those models directly affected costs or ICERs; it mentions those model types separately when discussing other studies. (LLM, conf=0.86)
plainstates a cause- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: unclear — It is unclear whether the paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
The cited sources support only part of this claim — read each source’s verdict below.
In the paper
“Salcedo et al.30 found that AiCure was cost-saving in 96.4% of simulations at a $150,000/QALY threshold.”
§ Systematic review of cost effectiveness and budget impact of artificial intelligence in healthcare > Results > Comparative economic outcomes
What the sources say
- Partly supportedcites Salcedo et al.30 Salcedo, J. et al. Cost-effectiveness of artificial intelligence monitoring for active tuberculosis treatment: a modeling study. PLoS One 16, e0254950 (2021). sourceExperimental — accuracy not yet measured
The Results report that AiCure was cost-effective in 96.4% of simulations at a $150,000/QALY threshold. They separately report that it produced more QALYs at a lower cost than DOT in 93.5% of simulations. The claim conflates cost-effectiveness with cost savings. (LLM, conf=0.94)
plainstates a cause- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: unclear — It is unclear whether the paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Partly supported
Every source the paper cites for this claim was checked and supports it.
In the paper
“Szymanski et al.28 further estimated that ML-based AF screening in the UK could reduce NHS and Personal Social Services (PSS) costs by up to –£80.4 million over three years28.”
§ Systematic review of cost effectiveness and budget impact of artificial intelligence in healthcare > Results > Comparative economic outcomes
What the sources say
- Supportedcites Szymanski et al.28 Szymanski, T. et al. Budget impact analysis of a machine learning algorithm to predict high risk of atrial fibrillation among primary care patients. Europace 24, 1240–1247 (2022). source
The Estimated budget impact section reports £80,441,386 in savings over three years for Scenario 2 when Personal Social Services costs are included alongside NHS costs. This is the largest saving reported across the three screening scenarios, matching the claim’s ‘up to £80.4 million’ estimate. (LLM, conf=0.93)
hedgedstates a cause- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: unclear — It is unclear whether the paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Supported
64 more claims · 3 with their cited sources checked
Create a free account to see all 116 claims and to verify the 10 references still open.
- It may state two separate claims as one. 29
- Probably not a substantial claim. 18
Claims other papers make about this one
Each is checked against this paper’s own text: does it say what the citing paper says it does?
The source backs part of the claim, not all of it — or backs it for a different population, condition or comparison.
In the citing paper
“Dynamic modeling suggests that long-term value is maximized when adaptive learning and workflow integration are considered, but indirect costs and equity issues are often underreported in current analyses [59].”
From Artificial Intelligence in Oncologic Thoracic Surgery: Clinical Decision Support and Emerging Applications (2026) · cites this paper as [59]
Experimental — accuracy not yet measuredThe source backs part of the claim, not all of it — or backs it for a different population, condition or comparison.
In the citing paper
“Artificial intelligence in thoracic surgery demonstrates potential for cost-effectiveness primarily by improving diagnostic accuracy, reducing unnecessary procedures, and optimizing resource utilization.”
From Artificial Intelligence in Oncologic Thoracic Surgery: Clinical Decision Support and Emerging Applications (2026) · cites this paper as [59]
Experimental — accuracy not yet measured
Self-check
Percentages, sums and reported differences re-computed from the paper’s own numbers.
Self-check results appear here once an independent check has confirmed them.
Cited works that were retracted
Works this paper cites that were later retracted, from the retraction notices publishers deposit. Major when the notice came out before this paper was published.
A note, not an issue: for example, a cited work that was retracted only after this paper was published, which its authors could not have known.
In the paper
“Whether an incremental analysis was performed, 8.”
Cited work 10.1016/j.heliyon.2024.e33186 · retracted 2025-11-12 · notice 10.1016/j.heliyon.2025.e44166 · this paper published 2025-08-26
Citations
Cites
1 retracted
Hussain, A., Umair, M., Khan, S., Alonazi, W. B., Almutairi, S. S., & Malik, A. (2024) · Heliyon · doi:10.1016/j.heliyon.2024.e33186 · Cited by 141 · cited 1×
“Whether an incremental analysis was performed, 8.”
RetractedUsing thematic analysis in psychology
Braun, V., & Clarke, V. (2006) · Qualitative Research in Psychology · doi:10.1191/1478088706qp063oa · Cited by 194,414
Rayyan-a web and mobile app for systematic reviews.
Ouzzani, M., Hammady, H., Fedorowicz, Z., & Elmagarmid, A. (2016) · Systematic Reviews · doi:10.1186/s13643-016-0384-4 · Cited by 26,092
Methods for the thematic synthesis of qualitative research in systematic reviews
Thomas, J., & Harden, A. (2008) · BMC Medical Research Methodology · doi:10.1186/1471-2288-8-45 · Cited by 9,312
Updating guidance for reporting systematic reviews: development of the PRISMA 2020 statement
Page, M. J., McKenzie, J. E., Bossuyt, P. M., Boutron, I., Hoffmann, T. C., Mulrow, C. D., Shamseer, L., Tetzlaff, J. M., & Moher, D. (2021) · Journal of Clinical Epidemiology · doi:10.1016/j.jclinepi.2021.02.003 · Cited by 3,490
Artificial intelligence in healthcare: transforming the practice of medicine
Bajwa, J., Munir, U., Nori, A., & Williams, B. (2021) · Future Healthcare Journal · doi:10.7861/fhj.2021-0095 · Cited by 1,856
Husereau, D., Drummond, M., Augustovski, F., de Bekker-Grob, E., Briggs, A. H., Carswell, C., Caulley, L., Chaiyakunapruk, N., Greenberg, D., Loder, E., et al. (2022) · Value in Health · doi:10.1016/j.jval.2021.10.008 · Cited by 847 · cited 2×
“Strategies to mitigate these obstacles include encouraging global adherence to standards such as CHEERS 2022 49 and CHEERS-AI 45 , facilitating international consensus-building forums, and standardizing reporting practices to enhance trans-parency and comparability across healthcare systems.”
Bias in artificial intelligence algorithms and recommendations for mitigation
Nazer, L. H., Zatarah, R., Waldrip, S., Ke, J. X. C., Moukheiber, M., Khanna, A. K., Hicklen, R. S., Moukheiber, L., Moukheiber, D., Ma, H., et al. (2023) · PLOS Digital Health · doi:10.1371/journal.pdig.0000278 · Cited by 621
Sources of bias in artificial intelligence that perpetuate healthcare disparities—A global review
Celi, L. A., Cellini, J., Charpignon, M.-L., Dee, E. C., Dernoncourt, F., Eber, R., Mitchell, W. G., Moukheiber, L., Schirmer, J., Situ, J., et al. (2022) · PLOS Digital Health · doi:10.1371/journal.pdig.0000022 · Cited by 576
Medical image analysis using deep learning algorithms
Li, M., Jiang, Y., Zhang, Y., & Zhu, H. (2023) · Frontiers in Public Health · doi:10.3389/fpubh.2023.1273253 · Cited by 505 · cited 1×
“For instance, deep learning algorithms applied to medical imaging have demonstrated diagnostic accuracies that rival — and in some cases surpass — that of expert clinicians 11 .”
Research works printed but not linked: 1
- Drummond, M. F., Sculpher, M. J., Claxton, K., Stoddart, G. L. & Torrance, G. W. Methods for the Economic Evaluation of Health Care Programmes. Consumption Benefits of Healthcare (Oxford University Press, London, 2015).
Cited by
El Arab, R. A., Alshakihs, A. H., Alabdulwahab, S. H., Almubarak, Y. S., Alkhalifah, S. S., Abdrbo, A., Hassanein, S., & Sagbakken, M. (2025) · Frontiers in Digital Health · doi:10.3389/fdgth.2025.1666005 · Cited by 113
Młynarska, E., Bojdo, K., Bulicz, A., Frankenstein, H., Gąsior, M., Kustosik, N., Rysz, J., & Franczyk, B. (2025) · Current Issues in Molecular Biology · doi:10.3390/cimb47100787 · Cited by 105
“Consequently, their accessibility could be severely limited, exacerbating inequities in healthcare resource distribution, as only those systems with considerable budgets can afford the initial investment and ongoing maintenance of such sophisticated AI [178,180].”
Boussi Rahmouni, H., Hassine, N. B. E. H., Chouchen, M., Ceylan, H. İ., Muntean, R. I., Bragazzi, N. L., & Dergaa, I. (2025) · Healthcare · doi:10.3390/healthcare13202553 · Cited by 25
“Operationally, these investments may be considerable, emphasizing that the “predicted financial outcome” might be overstated if such costs are overlooked [189,190].”
Al-Ewaidat, O. A., & Naffaa, M. M. (2025) · Rheumato · doi:10.3390/rheumato5040017 · Cited by 28
“Interoperability remains a persistent barrier, as variability across hospital information systems often impedes the standardized exchange of multimodal data essential for model recalibration and longitudinal monitoring [245–247].”
Artificial Intelligence for Noninvasive Health Diagnostics
Wankhede, P., Bhuyar, D. L., Zanwar, S. R., Pawar, R., Jadhav, M. R., Gandhewar, N., Kulkarni, M. B., Bhaiyya, M. L., & Haick, H. (2025) · ACS Sensors · doi:10.1021/acssensors.5c03171 · Cited by 31
Metadata
- Authors
- El Arab, R. A., & Al Moosa, O. A.
- Venue
- npj Digital Medicine
- Published
- 2025-08-26
- Volume
- 8
- Issue
- 1
- Pages
- 548-548
- ISSN
- 2398-6352 (ISSN-L)
- PMC
- PMC12381244
- OpenAlex
- W4413645249
- Semantic Scholar
- 477e6d6bf842a0e02d123b0e917a98e340b31cff
- Cited by
- 149
- References
- 58
- Citations by year
- 2026: 123 · 2025: 25
- FWCI
- 8.74
- Type
- journal article
- Language
- en
- Open access
- gold
- doi.org
- Licence
- cc-by
- creativecommons.org/licenses/by/4.0 (tdm)
- creativecommons.org/licenses/by/4.0 (vor)
- Locations
- npj Digital Medicine (publishedVersion, cc-by)
- Europe PMC (PubMed Central) (publishedVersion, cc-by)
- PubMed Central (publishedVersion, cc-by)
- PubMed (publishedVersion)
- DOAJ (DOAJ: Directory of Open Access Journals) (submittedVersion, cc-by-sa)
- Funders
- Almoosa College of Health Sciences
- Topics
- Health Sciences › Medicine › Health Informatics › Artificial Intelligence in Healthcare and Education
- Social Sciences › Economics, Econometrics and Finance › Economics and Econometrics › Health Systems, Economic Evaluations, Quality of Life
- Health Sciences › Health Professions › Health Information Management › Artificial Intelligence in Healthcare
- Keywords
- clinical AI, cost-effectiveness, budget impact, incremental cost-effectiveness ratio, diagnostic accuracy, quality-adjusted life years, health equity, ophthalmology, infectious diseases
About these checks
Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.
The claim verdicts this report counts are experimental: their accuracy is not yet measured.
A language model drew the claims from the paper on 2026-10-10, under version 3 of the claim definition (model deepseek-v4.1-flash, prompt extract-v3).
A second model labelled each claim on 2026-10-10 (model jev-1.13-20260917, definition revision gold-guideline-v0.2).
Cited sources are checked with a language model’s help (gpt-6-sol).
Statistics and arithmetic are recomputed, not judged.
Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.
What counts as a claim ›How we measure confidence ›How we check this ›
Are you an author, or do you see a problem with this page? Request a review
How to cite
Cite CoChat for the checks on this page, and the paper for its own findings.
- This page
- CoChat. (2026). Citation check: Systematic review of cost effectiveness and budget impact of artificial intelligence in healthcare. https://app.cochat.ai/papers/10.1038/s41746-025-01722-y
- The paper
- El Arab, R. A., & Al Moosa, O. A. (2025). Systematic review of cost effectiveness and budget impact of artificial intelligence in healthcare. npj Digital Medicine. https://doi.org/10.1038/s41746-025-01722-y
Data sources
Bibliographic data: Semantic Scholar, OpenAlex, Crossref.
Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).
Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.