Artificial Intelligence and Telemedicine in a World of Value-Based Healthcare: A Deep Dive

This article focuses on the American healthcare system, and uses diabetes management — and diabetic retinopathy in particular — as an example to explain the value-based care landscape.

Value-based healthcare incentivizes quality and resourcefulness in care. Both AI and Telemedicine present an opportunity to swing the pendulum towards higher quality, more affordable, and more accessible care. However, several stakeholders need to be properly aligned for progress at pace. On December 2nd 2020, the Center for Medicare & medicaid Services (CMS) released its final rule for CY 2021 Physician Fee Schedule. The document contained its final decision on CPT code 92229 (“AI detection of diabetic retinopathy”) as well as revisions to teleretinal codes 92227 and 92228.

In this article I do a deep dive into the value-based healthcare landscape, using diabetic retinopathy screening as a case in point.

Diabetes affects approximately 35 million Americans, each of whom needs at least one retinal exam per year. However, majority do not get their eye exam (Benoit et al.) due to multiple prohibitive factors such as cost, transportation, difficulty of taking time off from work, and inconvenience, amongst others. The standard of care in diabetes requires at least an annual eye exam to detect onset of diabetic retinopathy and to treat when indicated. This is important as diabetes is the most common cause of visual impairment and blindness in working age adults in the United States.

NCHQA and the HEDIS measure

The National Committee for Healthcare Quality Assurance (NCHQA) is an independent 501(c)(3) non-profit organization that seeks to improve healthcare quality by nationally administering evidence-based standards, metrics, and accreditations. The NCHQA administers a set performance metrics called the Healthcare Effectiveness Data and Information Set (HEDIS) which is used to assess the performance of Health plans such as Healthcare Maintenance Organizations (HMOs), Preferred Provider Organizations (PPOs), Medicare, and Medicaid. The HEDIS metrics assess the quality of care provided for particular disease conditions. And the resulting score, the HEDIS score, is a quality grade of the Provider, HMO, or Health plan.

The NCHQA derives its power from the HEDIS performance metric which has now become a standard for comparing the quality of care provided by various HMOs, Health plans, or individual providers. For instance, CMS requires that HMOs and Private Health Plans submit HEDIS scores so as to provide care for Medicare enrollees according the Medicare Advantage Plan (Medicare Part C). Medicare Advantage is a program in which Medicare pays the private plan a capitated amount per enrollee so as to provide the Medicare A and B services. Health plans collect the HEDIS data annually and report it to the NCHQA. And to do so they utilize a variety of methods including insurance claims data, surveys, sampling, and electronic medical records queries. Once a given Health Plan’s HEDIS data has been audited by an NCHQA-approved audit 3rd party firm, that HEDIS data becomes publicly available to stakeholders such as insurance brokers, consumer groups, employers, government agencies etc who can then make decisions about various Plans based on their quality as determined by HEDIS scores. For instance the U.S. News & World Report publishes a “Best Health Plans” ranking based on the set of HEDIS scores. Additionally, NCHQA rates health plans on a scale from 0 to 5, with 5 being best. It does so by combining HEDIS scores with patient satisfaction survey called the Consumer Assessment of Healthcare Providers and Systems (CAHPS) score.

The HEDIS Comprehensive Diabetes Care measure includes the Hemoglobin A1c testing, Retinal eye exam, Medical attention to nephropathy, and blood pressure control. And the eligible population are patients aged 18–75 with either Type I or II diabetes. As an example, Cigna’s phrasing to care providers states that per the 2020 HEDIS retinal exam measure requirements, one of the following must be documented in the medical record:

  • A retinal or dilated eye exam by an eye care professional (optometrist or ophthalmologist) in measurement year
  • A negative (for retinopathy) retinal or dilated eye exam by an eye care professional in prior measurement year
  • Bilateral eye enucleation any time during the member’s history

The other private health insurance companies participating in Medicare Advantage have similar phrasing to the above. As the Health plans are externally incentivized to achieve high HEDIS scores, they in turn pass down incentives to the primary care providers participating in their plan. The incentives can include financial bonuses, or in the case of poor performers, expulsion from participation.

The Merit-based Incentive Payment System (MIPS)

The Merit-based Incentive Payment program (MIPS) is a federally mandated program that aims to pay physicians for performance as opposed to service alone. It incentivizes quality and cost-effectiveness in care by collecting physician performance data and comparing it to that that of their peers to determine a reimbursement adjustment. The program was established as part of the Medicare Access and CHIP Reauthorization Act of 2015 (MACRA). More specifically, MACRA established the Quality Payment Program which consists of both MIPS and Alternative Payment Model (APM). Physicians can choose from either payment models. The MIPS assigns physicians a composite performance score which determines whether they will receive a payment bonus, a payment penalty, or neither. The MIPS score (0 to 100) is based on the following categories (and their respective weights for the 2020 performance year):

  • Quality (45% weight)
  • Cost (15% weight)
  • Improvement activities (15% weight)
  • Promoting interoperability (25% weight)

Of note, under MIPS physicians are still payed on a fee-for-service basis, but they collect and report performance data according to the above four categories and get a pay bonus, penalty, or nothing two years after the performance year. For example, data collected and submitted for work done in 2020 will receive medicare payment adjustment in the year 2022.

A Zero-Sum Game

The sum total of bonuses distributed under MIPS in any given year equals the sum total of penalties issued in that year. As such MIPS is referred to as a “budget neutral” program in that no specific allocation or subtraction is needed for it. The payment adjustments are applied to Medicare part B services two years after said performance occurred.

MIPS Quality Measure #117: Diabetic Eye Exam

One example of a MIPS quality measure is the diabetic eye exam. The physician reports a ratio. The numerator of that ratio is the number of diabetic patients aged 18–75 who have a diagnosis of diabetic retinopathy overlapping the measurement period and who received a diabetic eye exam during the measurement period; or who have no diagnosis of diabetic retinopathy overlapping the measurement period and who did receive a diabetic eye exam during the measurement period or prior year.

The denominator is the total number of diabetic patients aged 18–75 who were under that physician’s care in that measurement period. Specifically,

The diabetic eye exam quality measure is also identified as NQF #0055 and CMS #131.

MVPs: MIPS Value Pathways

They say the only constant in life is change. Just as physicians are starting to get used to MIPS born only in 2015 of MACRA and a consolidation of three programs which were themselves still then fledgling — Physician Quality Reporting System (PQRS), Value-based Payment Modifier (VBM) and Medicare Electronic Health Records (EHR) incentive program— MIPS is morphing onwards into MIPS Value Pathways (MVPs), and likely for good reason. In 2019 CMS announced that it will present a proposed rule for MVP with the purpose of decreasing the reporting burden on physicians and tailoring to physician’s scope of practice. Since MIPS is a peer comparison program, i.e. grading on a curve, proper alignment requires that everyone in a class is taking the same test as one’s peers, and that the test assesses relevant material. This increases standardization at the expense of flexibility. For example, endocrinologists and primary physicians caring for diabetes patients will all report on essentially the same set of measures across the four categories. And those measures will be standardized as illustrated in figures 3 and 4 below. Of note, due to COVID, as of CY 2021 Final Rule, CMS has moved the initial performance period of MVP from 2021 to 2022.

Illustrative example of MVP for Diabetes Prevention and Management. Source=CMS
CMS’ Proposed Timeline and Blueprint for MVP Implementation. Source=CMS.

We have now reviewed the quality and resourcefulness incentivization mechanisms in the care of diabetes patients. In particular, we’ve seen that through programs like NCQA’s HEDIS scoring and CMS’ MIPS, primary physicians and health plans are strongly incentivized to perform yearly retinal exams on their patients with diabetes. Let us know take a closer look at the CPT codes that enable reimbursement of those retinal exams when done in the primary care office with remote interpretation elsewhere.

Teleretina CPT Codes

Both of the Teleretina CPT codes for diabetic retinopathy have been in existence since 2011. They were introduced by CMS for the explicit purpose of helping with the burden of diabetic retinopathy screening.

Both Teleretinal CPT codes (92227 and 92228) were introduced in 2011 by CMS for the explicit purpose of helping alleviate the burden of diabetic retinopathy. However, there has been some confusion and incongruence in both codes since their inception. Hence as of CY 2021 Physician Fee Schedule final rule, the AMA CPT Editorial Panel revised both codes and changed their long form descriptors. One impetus for this change was a need to distinguish the teleretinal codes from the incoming AI code 92229.

A point of confusion or inconsistency in the teleretinal codes was that providers were needing to downgrade service to 92227 if the patient turned out to not have diabetic retinopathy. Hence the exact same amount of work was billed differently depending on the interpretation of the image. Furthermore, a service that clearly involved physician work was only billable as an exclusively practice expense RVU whenever the patient turned out to not have retinopathy. The CPT Editorial panel resolved this issue as described below by binding the CPT code to the interpreter (physician/qualified health professional vs not physician/qualified health professional) instead of to the interpretation (diabetic retinopathy vs no diabetic retinopathy).

Overall the changes achieved two goals: They made the codes better suited to the actual primary usage, which is the screening or monitoring of patients in the primary care setting. In addition, the CPT Editorial panel attempts to increase access to care by decreasing the level of specialization required of the interpreters.

CPT 92227

This code is now for screening or monitoring of diabetic retinopathy when the interpretation is performed by clinical staff only. For example, the image may be read remotely by ophthalmic staff at a retinal imaging reading center. This code therefore only consists of practice expense (PE) RVU weight, and its work RVU is zero.

  • Previous long descriptor: “Remote imaging for detection of retinal disease (eg, retinopathy in a patient with diabetes) with analysis and report under physician supervision, unilateral or bilateral.
  • New long descriptor (starting CY 2021): “Imaging of retina for detection or monitoring of disease; with remote clinical staff review and report, unilateral or bilateral

One important change to note is that the code is no longer only for “detection of .. disease” but now for “detection or monitoring of disease.” Hence it is no longer only for screening in a patient without known diabetic retinopathy. But can also be used for monitoring of patients with some known retinopathy. Another important change to note in the above long descriptors is that the need for “physician supervision” has been removed.

The medical appropriateness and safety of these changes may remain open points of discussion in the future, as more data becomes available. The rationale for the changes is likely to increase access to care, and may be potentially justifiable on a population level, pending more claims and outcomes data.

Notably, 92227 still does not include a -26/-TC code split. The (-26) is a modifier that encodes the interpretation component of a service, while (-TC) encodes the technical component, e.g operating the camera to take the retinal image. There is no -26/-TC code split in 92227 because it includes no physician work.

CPT 92228

This code is now for remote screening or monitoring of diabetic retinopathy by a physician or other qualified healthcare professional. It therefore consists of physician work RVU, practice expense RVU, and malpractice RVU.

  • Previous long descriptor: “Remote imaging for monitoring and management of active retinal disease (eg, diabetic retinopathy) with physician review, interpretation and report, unilateral or bilateral
  • New long descriptor (starting CY 2021): “Imaging of retina for detection or monitoring of disease; with remote physician or other qualified health care professional interpretation and report, unilateral or bilateral

Changes include “active retinal disease” has been changed simply to “disease,” i.e. the disease need not be active it can be stable or quiescent. “monitoring and management” has been changed simply to “detection and monitoring.” Hence this now better suited for its main use case, screening or monitoring in primary care setting. Finally, “physician” has been changed to “physician or other qualified health care professional.” Here the CPT Editorial panel broadens who is able to bill 92228. There is ambiguity in the term “other qualified health care professional,” and as such this will likely need to be further specified by CMS in the future.

CPT 92229: AI detection of diabetic retinopathy

There has been much anticipation around this code which will enable CMS reimbursement of autonomous AI point-of-care diabetic retinopathy detection and monitoring.

  • Long descriptor: “Imaging of retina for detection or monitoring of disease; point-of-care; automated analysis and report, unilateral or bilateral
  • Activation date: Jan 2021
  • Ambulatory Payment Category: 5733 (Level III Minor Procedure)
  • Status indicator: “S” (Significant procedure)
  • RVU: Practice Expense Only (No physician work)
  • Reimbursement: around $54

CPT 92229 Deliberation

CMS had initially proposed reimbursement rate of $33.16 for 92229 because it was lumped with codes 92227 and 92228 which both fall into Ambulatory Payment Category 5732 (Level 2 Minor Procedure). By contrast, CPT 92250 for fundus photography done in the Eye clinic is classified as APC 5734 (Level 4 Minor Procedure) and reimbursed at $113.23.

The American Medical Association/Specialty Society Relative Value update Committee (RUC) provides recommendations annually to CMS on the values of RVUs. The American Academy of Ophthalmology, a specialty society of the RUC, had recommended in May 2019 that the automated AI retinopathy detection be billed using 92250 till the CMS PFS CY2021 Final Rule. Furthermore, the initial name given to 92229 by the AMA CPT Editorial Panel was 9225X, anticipating the argument that it was more similar to 92250 than to the teleretinal codes 92227 and 92228. CMS ultimately disagreed with this and classified it alongside 92227 and 92228, and moreso 92227 by asserting that it involves no physician work.

The RUC had proposed a $25 analysis fee (akin to a physician work reimbursement for interpretation). CMS sort of rejected this proposal arguing instead that the analysis component done by the AI in the cloud constitutes an indirect practice expense. The RUC rebutted this assertion, pointing out that the analysis is in fact done on a per patient basis, hence is more appropriately a direct expense. This is absolutely correct. For each patient, the retinal images are transmitted into the cloud where the AI analysis engine interprets them and sends back a PDF report.

In addition, CMS proposed a 1 minute reduction in the clinical labor task CA009 “Greet patient, provide gowning, ensure appropriate medical records are available.” They wanted it to match the allotted times of CA009 for 92227 and 92228. The RUC rebutted this stating that this task takes longer in 92229 because the data needs to be entered on both the camera device and the client computer.

During the comment-response period following the proposed CY 2021 rule, there was a petition to classify 92229 as APC 5734 (Level 4 Minor Procedure). CMS did not agree but did compromise and bumped it up to APC 5733 (Level 3 Minor Procedures). What this means is that the final reimbursement rate in Addendum B to the final Rule in CY 2021 will settle around $55 — based on what APC 5733 has reimbursed in recent years; historically in the range $54 to $56. The exact reimbursement will be published in Addendum B of the Physician Fee Schedule — as at the time of this writing the field for 92229 is not yet populated. Per the American Academy of Ophthalmology, there may ultimately be some regional variations in 92229 reimbursement to be determined by Medicare Administrative Contractors (MACs) [5].

Another event in the comment-response period for classification of 92229 was a reclassification from Status indicator “Q1” to status indicator “S” (“Significant procedure”). This implies the payment is not bundled with the rest of the visit, but is separate and therefore guaranteed to not be absorbed by other services in a capitated scenario.

CMS Final Decision on 92229: APC 5733, Status indicator “S”


Value-based healthcare incentivizes quality and resourcefulness in care. Technological approaches such as artificial intelligence and telemedicine can improve patient outcomes while decreasing overall healthcare spending. The Center for Medicare and Medicaid Services (CMS) and the Relative Value Update Committee (RUC) play a critical role in facilitating the implementation and adoption of novel technologies into our healthcare system. CPT codes 92227, 92228, and 92229 are encouraging signals of progress in the integration of telemedicine and AI screening into primary care.

BIO: Dr. Stephen G. Odaibo is CEO & Founder of RETINA-AI Health, Inc. He is a Physician, Retina Specialist, Mathematician, Computer Scientist, and Full Stack AI Engineer. In 2021 he was issued a U.S. Patent for inventing an AI system that automatically detects diseases from ophthalmic images. In 2017 he received UAB College of Arts & Sciences’ highest honor, the Distinguished Alumni Achievement Award. And in 2005 he won the Barrie Hurwitz Award for Excellence in Neurology at Duke Univ School of Medicine where he topped the class in Neurology and in Pediatrics. He is author of the books “Quantum Mechanics & The MRI Machine” and “The Form of Finite Groups: A Course on Finite Group Theory.” Dr. Odaibo Chaired the “Artificial Intelligence & Tech in Medicine Symposium” at the 2019 National Medical Association Meeting. Through RETINA-AI, he and his exceptionally talented team are building AI solutions to address the world’s most pressing healthcare problems. He resides in Houston Texas with his family.

1. CMS Quality Payment Program: MIPS Value Pathways (MVPs)
2. CMS OPPS Final Rule Summary
3. CMS CY 2021 Proposed Rule
4. Benoit et. al. Eye Care Utilization Among Insured People With Diabetes, U.S. 2010–2014
5. Glasser et. al. Medicare Update Webinar: The Financial Impact of the CMS Final Fee Schedule Rule on Ophthalmic Practice



Get the Medium app

A button that says 'Download on the App Store', and if clicked it will lead you to the iOS App store
A button that says 'Get it on, Google Play', and if clicked it will lead you to the Google Play store
Dr Stephen Odaibo

Physician. Retina Specialist. Computer Scientist. Mathematician. Full Stack AI Engineer. Christian. Husband. Dad. CEO/Founder RETINA-AI Health, Inc.