The Financial Case for AI in Medicine: A Step-by-Step ROI Guide to Unlocking Value
Rubin Pillay MD, PhD
Artificial intelligence (AI) stands at the forefront of innovation, promising to revolutionize patient care, streamline operations, and potentially reduce costs. However, for many healthcare leaders and practitioners, the question remains: Does AI truly deliver on its financial promises? As a physician, healthcare executive, and academic, I've witnessed firsthand the transformative potential of AI in medicine. Yet, I've also seen the hesitation and skepticism that often accompanies significant technological investments in our field.
This article aims to bridge the gap between AI's potential and its practical implementation by providing a comprehensive, step-by-step guide to calculating the return on investment (ROI) for AI solutions in healthcare. By leveraging the Time-Driven Activity-Based Costing (TDABC) model, a powerful tool for accurate cost measurement in healthcare settings, we'll demystify the process of evaluating AI's financial impact.
Throughout this guide, we'll explore two common scenarios: a radiologist using AI to read X-rays and a primary care physician using AI for note-taking. These examples will serve as practical illustrations of how to apply ROI calculations to real-world healthcare situations. By the end of this article, you'll have a clear framework for assessing the financial viability of AI implementations in your own healthcare setting.
Remember, while ROI is a crucial metric, it's just one piece of the puzzle. We'll also discuss other critical factors to consider when evaluating AI solutions, ensuring a holistic approach to technology adoption in healthcare. Whether you're a hospital administrator, a practicing physician, or a healthcare investor, this guide will equip you with the tools to make informed, data-driven decisions about AI implementation in medicine.
Let's consider two common scenarios: a radiologist using AI to read X-rays and a primary care physician using AI for note-taking. For both cases, we'll use the TDABC equation:
Radiologist using AI to read x-rays
Assumptions:
· Radiologist's annual salary: $300,000
· Annual overhead costs: $100,000 (incl supervision, space, technology)
· Working hours per year: 2,000
· Time to read one x-ray manually: 15 minutes
· Time to read one x-ray with AI assistance: 5 minutes
· Number of x-rays read per year: 8,000
1. Calculate capacity cost rate:
Total cost of capacity supplied = $300,000 + $100,000 = $400,000
Practical capacity of resources supplied = 2,000 hours = 120,000 minutes
Capacity cost rate = $400,000 / 120,000 = $3.33 per minute
2. Calculate cost per x-ray:
Without AI: $3.33 × 15 = $49.95 per x-ray
With AI: $3.33 × 5 = $16.65 per x-ray
Savings per x-ray = $49.95 - $16.65 = $33.30
3. Calculate annual cost:
Without AI: $49.95 × 8,000 = $399,600
With AI: $16.65 × 8,000 = $133,200
4. Calculate cost savings:
Annual cost savings = $399,600 - $133,200 = $266,400
5. Set up the breakeven equation:
Number of x-rays × Savings per x-ray = AI system cost
x × $33.30 = $100,000
Solve for x (number of x-rays):
x = $100,000 / $33.30
x ≈ 3,003 x-rays
6. Calculate ROI:
Assuming AI system cost: $100,000
ROI = (Cost savings - AI system cost) / AI system cost × 100%
ROI = ($266,400 - $100,000) / $100,000 × 100% = 166.4%
To put this in perspective:
· The hospital reads 8,000 x-rays per year
· Breakeven occurs at 3,003 x-rays
· This means the hospital will break even after approximately:
· (3,003 / 8,000) × 12 months ≈ 4.5 months
So, the hospital will recover its investment in the AI system in about 4.5 months, assuming a constant rate of x-ray readings throughout the year. After this point, the hospital will start realizing net savings from the AI implementation.
Now lets recalculate the ROI for a hospital with 5 radiologists. We will need to adjust some of our assumptions and calculations. Let's go through this step-by-step:
1. Assumptions:
5 radiologists
· Each radiologist works 2,000 hours per year
· Total annual salary cost: 5 × $300,000 = $1,500,000
· Annual overhead costs: $500,000 (increased due to more radiologists)
· AI system cost: $250,000 (increased due to larger scale implementation)
· Number of x-rays read per year: 40,000 (5 times the original amount)
2. Calculate capacity cost rate:
Total cost of capacity supplied = $1,500,000 + $500,000 = $2,000,000
Practical capacity of resources supplied = 5 × 2,000 hours = 10,000 hours = 600,000 minutes
Capacity cost rate = $2,000,000 / 600,000 = $3.33 per minute (same as before)
3. Calculate cost per x-ray:
Without AI: $3.33 × 15 = $49.95 per x-ray
With AI: $3.33 × 5 = $16.65 per x-ray
4. Calculate annual cost:
Without AI: $49.95 × 40,000 = $1,998,000
With AI: $16.65 × 40,000 = $666,000
5. Calculate cost savings:
Annual cost savings = $1,998,000 - $666,000 = $1,332,000
6. Calculate ROI:
ROI = (Cost savings - AI system cost) / AI system cost × 100%
ROI = ($1,332,000 - $250,000) / $250,000 × 100% = 432.8%
7. Calculate breakeven point:
Savings per x-ray = $49.95 - $16.65 = $33.30
Breakeven point = $250,000 / $33.30 ≈ 7,508 x-rays
8. Time to breakeven:
(7,508 / 40,000) × 12 months ≈ 2.25 months
In this scenario with 5 radiologists:
· The annual cost savings are $1,332,000
· The ROI over one year is 432.8%
· The hospital will break even after reading about 7,508 x-rays
· The breakeven point is reached in approximately 2.25 months
This larger-scale implementation shows an even more impressive ROI, primarily due to the increased volume of x-rays being processed and the economies of scale in the AI system cost. The hospital recoups its investment faster and sees greater annual savings, making the case for AI implementation even stronger in this scenario.
Primary care physician using AI to take notes
Assumptions:
· Physician's annual salary: $200,000
· Annual overhead costs: $50,000
· Working hours per year: 2,000
· Time spent on note-taking per patient without AI: 10 minutes
· Time spent on note-taking per patient with AI: 3 minutes
· Number of patients seen per year: 4,000
1. Calculate capacity cost rate:
Total cost of capacity supplied = $200,000 + $50,000 = $250,000
Practical capacity of resources supplied = 2,000 hours = 120,000 minutes
Capacity cost rate = $250,000 / 120,000 = $2.08 per minute
2. Calculate cost per patient note:
Without AI: $2.08 × 10 = $20.80 per patient
With AI: $2.08 × 3 = $6.24 per patient
Savings per patient note = $20.80 - $6.24 = $14.56
3. Calculate annual cost:
Without AI: $20.80 × 4,000 = $83,200
With AI: $6.24 × 4,000 = $24,960
4. Calculate cost savings:
Annual cost savings = $83,200 - $24,960 = $58,240
5. Set up the breakeven equation:
Number of patient notes × Savings per note = AI system cost
x × $14.56 = $50,000
Solve for x (number of patient notes):
x = $50,000 / $14.56
x ≈ 3,434 patient notes
6. Calculate ROI:
Assuming AI system cost: $50,000
ROI = (Cost savings - AI system cost) / AI system cost × 100%
ROI = ($58,240 - $50,000) / $50,000 × 100% = 16.48%
Therefore, the breakeven point is approximately 3,434 patient notes. After this number of notes have been completed using the AI system, the physician or practice will have recouped the initial $50,000 investment in the AI technology.
To put this in perspective:
· The physician sees 4,000 patients per year
· Breakeven occurs at 3,434 patient notes
· This means the physician will break even after approximately:
· (3,434 / 4,000) × 12 months ≈ 10.3 months
So, the physician or practice will recover its investment in the AI system in about 10.3 months, assuming a constant rate of patient visits throughout the year. After this point, they will start realizing net savings from the AI implementation.
This breakeven analysis shows that while the investment in AI for note-taking does provide a positive return, it takes longer to recoup the initial investment compared to the radiologist example. However, it's important to note that there may be additional benefits not captured in this financial analysis, such as improved note quality, reduced physician burnout, and potentially increased patient satisfaction due to more focused physician attention during visits.
Let's recalculate the scenario for 5 primary care physicians using AI for note-taking. We'll adjust our assumptions and calculations accordingly:
Assumptions:
· 5 primary care physicians
· Each physician's annual salary: $200,000
· Annual overhead costs: $250,000 (increased due to more physicians)
· Working hours per year per physician: 2,000
· Time spent on note-taking per patient without AI: 10 minutes
· Time spent on note-taking per patient with AI: 3 minutes
· Number of patients seen per year: 20,000 (5 times the original amount)
· AI system cost: $150,000 (increased due to larger scale implementation)
Calculations:
1. Calculate capacity cost rate:
Total cost of capacity supplied = (5 × $200,000) + $250,000 = $1,250,000
Practical capacity of resources supplied = 5 × 2,000 hours = 10,000 hours = 600,000 minutes
Capacity cost rate = $1,250,000 / 600,000 = $2.08 per minute (same as before)
2. Calculate cost per patient note:
Without AI: $2.08 × 10 = $20.80 per patient
With AI: $2.08 × 3 = $6.24 per patient
3. Calculate annual cost:
Without AI: $20.80 × 20,000 = $416,000
With AI: $6.24 × 20,000 = $124,800
4. Calculate cost savings:
Annual cost savings = $416,000 - $124,800 = $291,200
5. Calculate ROI:
ROI = (Cost savings - AI system cost) / AI system cost × 100%
ROI = ($291,200 - $150,000) / $150,000 × 100% = 94.13%
6. Calculate breakeven point:
Savings per patient note = $20.80 - $6.24 = $14.56
Breakeven point = $150,000 / $14.56 ≈ 10,302 patient notes
7. Time to breakeven:
(10,302 / 20,000) × 12 months ≈ 6.18 months
Results for 5 primary care physicians:
· The annual cost savings are $291,200
· The ROI over one year is 94.13%
· The practice will break even after completing about 10,302 patient notes
· The breakeven point is reached in approximately 6.18 months
This larger-scale implementation shows a significantly improved ROI compared to the single physician scenario. The practice recoups its investment faster (6.18 months vs. 10.3 months) and sees greater annual savings. The ROI has increased from 16.48% to 94.13%, making the case for AI implementation much stronger in this scenario. The improved economics are due to the economies of scale: while the cost of the AI system increased, it didn't increase proportionally to the number of physicians. This allows the practice to spread the cost over a larger number of patient visits, leading to faster breakeven and higher ROI. Implementing AI for note-taking becomes more financially attractive as the scale of the practice increases, potentially making it a very appealing option for larger medical groups or hospital systems.
The analysis presented in this article demonstrates the significant potential for AI to generate substantial cost savings and return on investment in healthcare settings, particularly when implemented at scale. Both scenarios examined - radiologists using AI for X-ray interpretation and primary care physicians using AI for note-taking - showed positive ROI, with the benefits amplifying as the scale of implementation increased.
Key points from the analysis:
· Scale matters: The ROI for both scenarios improved dramatically when scaled from a single practitioner to a group of five. This suggests that larger healthcare organizations may be better positioned to benefit from AI implementations.
· Faster breakeven for radiology: The radiology use case showed a quicker path to breakeven and higher ROI compared to the primary care scenario. This could be due to the higher volume of discrete tasks (X-ray readings) and the more significant time savings per task.
· Improved efficiency: In both cases, AI significantly reduced the time required for key tasks, potentially allowing healthcare providers to see more patients or focus on more complex cases.
· Non-financial benefits: While the analysis focused on financial metrics, it's important to note the potential for improved patient care, reduced burnout, and increased job satisfaction that may result from AI implementation.
· Variability in implementation costs: The assumed costs for AI systems were estimates and may vary significantly based on the specific solution and scale of implementation. Healthcare organizations should carefully assess these costs in their own contexts.
· Time savings translate to cost savings: The TDABC model clearly illustrates how time savings directly impact costs, providing a clear rationale for AI adoption in time-intensive tasks.
The financial case for AI in medicine, as demonstrated through these ROI calculations, is compelling, particularly for larger healthcare organizations. The ability of AI to significantly reduce the time required for routine tasks while maintaining or improving accuracy can lead to substantial cost savings and efficiency gains.
While ROI is a vital metric, it's not the only factor to consider when evaluating AI solutions in healthcare. Other important considerations include:
Quality improvement: AI may enhance diagnostic accuracy or patient outcomes, which can have long-term financial benefits.
Increased capacity: Time savings could allow for seeing more patients or conducting more procedures.
Physician satisfaction: Reducing administrative burden can improve job satisfaction and reduce burnout.
Long-term benefits: Some advantages of AI may not be immediately quantifiable but could provide significant value over time.
Address concerns about patient data protection and regulatory compliance.
Integration with existing systems: Consider the challenges of integrating AI solutions with current healthcare IT infrastructure.
Training and change management: Factor in the time and resources required to train staff and manage the transition to AI-assisted workflows.
Ongoing maintenance and updates: Account for the long-term costs of maintaining and updating AI systems.
While the financial benefits of AI in healthcare are clear, successful implementation requires a holistic approach that considers all these factors. Healthcare leaders should use this ROI framework as a starting point for evaluating AI solutions, while also considering the broader implications for their organizations and patients.
As AI technology continues to evolve, its potential to transform healthcare delivery and economics will likely grow. Healthcare organizations that can effectively leverage these technologies may find themselves at a significant advantage in an increasingly competitive and cost-conscious healthcare landscape.
@rubin - thanks for writing this blog post. I wish you had published this as a peer reviewed opinion paper so that you would have gotten feedback from reviewers and would have improved this piece.
I absolutely agree that AI can significantly impact Radiologists' interpretation times, patient outcomes and also ROI.
My issue with this blog is that you are using interpretation times that are not realistic. As someone who has done dozens of reader studies in his life, average xray interpretation time is actually close to 90 seconds and not 15 mins. If you look at just Chest Xrays, it is much lower. Here is a recent Nature peer reviewed paper (https://www.nature.com/articles/s41746-023-00829-4) showing Chest Xray interpretation time of 15 seconds without AI and 13 seconds with AI. I think you should redo the calculations using more realistic interpretation times.
Here is another peer-reviewed study showing reduction of reading times with AI from 10–65 seconds to 6–27 seconds (https://pubs.rsna.org/doi/10.1148/radiol.2021202818).
Here is a paper that actually excluded cases that took longer than 120s. (https://journals.plos.org/plosone/article?id=10.1371/journal.pone.0264383) Here is the reason why: To exclude the cases that remained open for long durations due to unexpected interruptions, we considered more than 120s as an unreliable reading time because readers may have been interrupted and excluded from the analysis.). 15 min interpretation time is totally unrealistic. In this study, the interpretation time actually went up from 14s without AI to 19s with AI.
I hope you will consider this as a positive criticism of your post.