Monday, February 6, 2023

Chat GPT Summarizes White Paper; Litmus Test for Clarity of Human Writing?

On January 20, 2023, I published a new white paper on trends in genomic coding for tumors - here.

Summary of Executive Summary

The white paper has a page long executive summary, and about 10 pages of text.   

I gave the executive summary to ChatGPT to summarize, and it did a pretty good job.  (Clipped at bottom).

Summary of Half of Paper

Next, while the whole ten-page single-spaced paper is too long for ChatGPT to absorb, I did cut and paste  the second half of the paper, for a summary.   This also came out pretty well.   

What's the Point?  A Clarity Litmus Test?

So it occured to me, people often think of ChatGPT in forward-mode:  Can it write a paper?  Can it write a discussion section?   This is using it backwards as a sort of cross-check.  If you have written a paper, essay, op ed, or executive summary, can ChatGPT "understand it" and summarize it clearly?  If so, you've written clearly enough that even a computer (basically a pile of tin cans and some batteries) can understand it, then it's probably fairly clear.

And on the other hand, if you give your essay to ChatGPT to summarize, and it returns with a confusing, conflicted mess, maybe - just maybe - your writing isn't as clear as you thought it was.

#   #   #   #  

The Computer Summary of my own Executive Summary:

  • Comprehensive Genomic Profiling (CGP) in cancer is widely recognized as having clinical utility and is included in updated guidelines, Medicare coverage policies, and white papers. The newer plasma-based CGP technology is preferred by some cancer patients as it is available when solid tumor tissue is not and can be performed serially to determine therapy resistance. The AMA code set for genomics does not distinguish between specimen sources but the newer PLA code set does and there are already codes for plasma-based tumor testing. Reimbursement for plasma-based testing requires codes specifying it as the underlying technology and there are already codes for two FDA-approved tests. The cost of plasma-based testing is higher due to the depth of sequencing required, but it offers cost offsets compared to tissue-based testing. The AMA CPT has recently introduced new specificity codes for RNA sequencing in cancer and is working on updating the coding system for cancer genomics. 2023 is expected to be an exciting period for policy around plasma-based CGP.
Computer Summary of Second Half of Paper:
  • The coding system for genomics in the US, the AMA CPT codes, has seen a rise in codes related to genomics in recent years, including the creation of the "Proprietary Laboratory Analysis" (PLA) codes. There are currently only a few codes for tumor genomics and these codes have not yet been specific to sample source such as plasma or paraffin. The AMA CPT has recently revised codes for DNA and RNA analyses but more needs to be done to keep up with the latest advancements. The coding system has not yet recognized the coding for plasma-based comprehensive genomic profiling (CGP) testing, which is an important area now seeing FDA-approved tests and rapid innovation from laboratory-developed tests. The resources for plasma-based CGP tests are different and higher than for FFPE-based tests, but there are cost offsets obtained from using a plasma-based approach. There is still time to debate these topics and develop the best consensus strategies through the AMA CPT meeting in May 2023.

ChatGPT Discusses AI in Pathology; Pathology vs Radiology

Please write a summary and also a powerpoint deck discussing the growing impact of artificial intelligence on the medical specialty of Pathology.   You are an expert and you are speaking to an expert audience.


Artificial Intelligence (AI) is rapidly changing the field of Pathology. With the increasing ability to process large amounts of data, AI is providing pathologists with new tools to diagnose and treat diseases more accurately and efficiently. This is particularly relevant in the area of digital pathology, where large amounts of tissue images can be analyzed using AI algorithms. AI is also helping to streamline the workflow of pathologists, reducing the time taken for diagnosis and allowing them to focus on more complex cases. Overall, the impact of AI on pathology is expected to be significant, leading to improved patient outcomes and cost savings for healthcare systems.


Slide 1: Introduction

  • Explanation of Artificial Intelligence (AI) and its increasing impact on various fields
  • Importance of AI in Pathology and its potential to revolutionize the field

Slide 2: The impact of AI on Pathology

  • Increased accuracy of diagnoses through the use of AI algorithms to analyze large amounts of data and identify patterns
  • Improved workflow efficiency through the automation of routine tasks and the ability to process large amounts of data quickly
  • Better patient outcomes through earlier and more accurate diagnoses, leading to earlier and more effective treatment

Slide 3: Digital Pathology

  • Definition and explanation of digital pathology and how it relates to AI in pathology
  • Advantages of using AI in digital pathology, including the ability to process large amounts of tissue images quickly and accurately
  • Examples of AI applications in digital pathology, such as the analysis of whole slide images, tissue classification, and tissue segmentation

Slide 4: AI in diagnostics

  • Explanation of how AI is being used to support pathologists in diagnosing diseases by analyzing large amounts of data, including tissue images, and identifying patterns and anomalies
  • Discussion of the potential impact on patient outcomes and healthcare systems, including earlier and more accurate diagnoses, improved patient care, and reduced healthcare costs

Slide 5: AI in workflow efficiency

  • Explanation of how AI can streamline the diagnostic process by automating routine tasks, reducing the time taken for diagnosis, and allowing pathologists to focus on more complex cases
  • Discussion of the potential impact on productivity and cost savings, including reduced workload for pathologists, improved patient outcomes, and reduced healthcare costs

Slide 6: Challenges and limitations

  • Discussion of the challenges facing the implementation of AI in pathology, including the need for high-quality data, the complexity of the algorithms, and the need for validation and regulatory approval
  • Explanation of the limitations of AI in this field, including the need for human expertise and judgment, the potential for errors, and the need for continued research and development

Slide 7: Future developments

  • Discussion of the potential future developments in the field of AI and Pathology, including the integration of AI into clinical decision-making, the development of more sophisticated algorithms, and the potential for personalized medicine
  • Explanation of the impact these developments may have on the field, including improved patient outcomes, reduced healthcare costs, and the potential to revolutionize pathology

Slide 8: Conclusion

  • Summary of the impact of AI on Pathology and its potential to revolutionize the field
  • Emphasis on the importance of continued research and development in this field to fully realize the potential of AI in pathology.

Speaking as an expert, and speaking to an expert audience, summary the differences in the impact of artificial intelligence on Radiology and on Pathology.

As an expert speaking to an expert audience, I would summarize the differences in the impact of artificial intelligence (AI) on Radiology and Pathology as follows:

Radiology:

  • In Radiology, AI has had a significant impact on imaging analysis and interpretation, particularly in the areas of CT, MRI, and X-ray.
  • AI algorithms have been developed to assist radiologists in analyzing and interpreting images, reducing the time taken for diagnosis and improving accuracy.
  • AI is also being used to automate routine tasks, such as image quantification, and to support decision-making in complex cases.

Pathology:

  • In Pathology, AI is having a growing impact, particularly in the area of digital pathology, where large amounts of tissue images can be analyzed using AI algorithms.
  • AI is helping to streamline the workflow of pathologists, reducing the time taken for diagnosis and allowing them to focus on more complex cases.
  • AI is also providing pathologists with new tools to diagnose and treat diseases more accurately and efficiently.
Overall, the impact of AI on both Radiology and Pathology is expected to lead to improved patient outcomes and cost savings for healthcare systems. However, the specific impact and applications of AI differ between the two fields, reflecting the unique challenges and opportunities in each.


Wednesday, February 1, 2023

ChatGPT - Not Great on LDCT (Reasons for Low Adoption)

Low dose CT (LDCT) is a preventive service for patients with a smoking history.   The adoption is very low, possibly 5% for example.   Why is the adoption rate so low?


There are several reasons why the adoption rate of low dose CT (LDCT) as a preventive service for patients with a smoking history is low:

  • Lack of Awareness: Many people may not be aware of the benefits of LDCT and its role in early detection of lung cancer.
  • Cost: LDCT can be expensive and may not be covered by all insurance plans, which can be a barrier for some patients.
  • False Positives: LDCT can sometimes produce false positive results, which can lead to unnecessary follow-up tests and procedures. This can cause anxiety and fear in some patients, and may discourage them from getting an LDCT.
  • Lack of Recommendation from Healthcare Providers: Some healthcare providers may not be familiar with LDCT or may not recommend it to their patients due to concerns about false positive results or cost.
  • Stigma Associated with Smoking: Some patients may be reluctant to get an LDCT because they feel ashamed of their smoking history.

These are some of the reasons why the adoption rate of LDCT as a preventive service for patients with a smoking history is low. To increase adoption, it will be important to educate patients and healthcare providers about the benefits of LDCT and to address any barriers that may prevent people from getting this important screening.


Why is reimbursement an issue? Is it not free as long as recommended by USPSTF?


Reimbursement for low dose CT (LDCT) can be an issue because, although it is recommended by the US Preventive Services Task Force (USPSTF), not all insurance plans are required to cover it. Medicaid and Medicare, for example, generally follow USPSTF recommendations, but private insurance plans may not. Additionally, even if a private insurance plan covers LDCT, there may be cost-sharing requirements, such as deductibles or co-payments, that can make the test unaffordable for some patients.

It's also worth noting that the Affordable Care Act (ACA) requires most private insurance plans to cover preventive services that are recommended by the USPSTF without cost-sharing, but there are some exceptions and exemptions, and the requirements can be complex.

In summary, while the USPSTF recommendation may increase the likelihood that LDCT will be covered by insurance, it is not a guarantee, and reimbursement can still be an issue for some patients.


I believe part of your answer is misleading.  While it is true that Medicare usually follows USPSTF guidelines, but isn't required to, in the case of LDCT, which is the question, yes, Medicare has definite coverage of LDCT.   

I apologize for any confusion. You are correct. 

Medicare does cover low dose CT (LDCT) scans for lung cancer screening in eligible individuals, as long as they meet certain criteria. This is in accordance with the US Preventive Services Task Force (USPSTF) recommendation. Thank you for bringing this to my attention.