Who's your PCP?
Patient Attribution in VBC: Where Accountability Meets Confusion
This is the third article in my Value-Based Care series, which aims to lay the foundation for understanding healthcare's current landscape and how it affects primary care.
In the last two newsletters, we covered Population Health and Value-Based Contracts.
To quote the definition of Population Health from my article “From Doctors to Social Workers”:
“the health outcomes of a group of individuals, including the distribution of such outcomes within the group.”
From a Population Health perspective, the “group of individuals” in whom we need to measure health outcomes are called “attributed lives” or “covered lives.” Without knowing this population, we cannot determine how sick people are and how much money to set aside, aka budget, to take care of them.
To quote from my last article, “Value-Based Care: The Illusion of Improvement”:
At the most basic level, there are two components to this style of contract:
Fixed budget to deliver care to a population (aka attributed lives).
Performance on quality metrics based on targets set by the contract.
The population on which the budget is based is called “Attributed Lives” or “Covered Lives.” The simplest way to determine attributed lives is to:
List all primary care providers (PCPs) participating in the ACO.
The panel of patients these participating PCPs provide ongoing care.
While on the surface, this seems simple, it can get complicated very quickly.
In this newsletter, we will discuss why it is complicated to answer this seemingly simple question of “attributed lives.”
PCP and Attributed Lives
From VBC's perspective, health plans (insurance companies, Medicare, or Medicaid) use “attributed lives” to determine which provider should be held accountable for the entire cost & quality of care a person receives (yes, you read that correctly). And they have proclaimed that it will be the PCP for the following reasons:
Have a “panel of patients.”
PCPs are the entry point to the healthcare system, i.e., they are generally the first contact a person has when a medical issue arises.
PCPs generally refer to specialists (although some insurance plans allow self-referral).
It is easier to hold one “person accountable” than the entire medical system.
If a patient does not have a PCP, then the patient is attributed to a specialist who renders or is expected to render the most care.
To keep things simple, we will assume that all patients have a regular PCP for now and address specialists later on in the article.
Who is considered a Primary Care Physician (PCP)?
Pediatrics, Family Medicine, and Internal Medicine were traditionally considered primary care specialties. This designation has become more complicated with these specialties performing “non-traditional” work :
Hospitalists: doctors that only take care of people admitted to a hospital.
SNFist: doctors who only care for people admitted to a Skilled Nursing Facility (SNF).
Locum tenens: doctors filling in as “temporary PCP” when a practice is short-staffed.
Doctors who work in Urgent Care Centers or Retail Clinics such as CVS Minute Clinics to provide urgent care services.
Virtual Care only providers: (Hims/Hers, Weight management providers).
These “non-traditional” roles have made it challenging to use specialty designation to assign PCP. To solve this challenge, several methods have been developed, including:
Credentialing status with Health Plan: the provider chooses to be listed as a PCP or specialist with the insurance company.
Job Taxonomy: where a provider chooses their job classification.
Requiring patients to pick a “PCP” from the insurance website manually.
However, all these methods fail when doctors work part-time and/or in different roles. The best example is Gynecologists. Several young women (especially those without significant medical issues) use their gynecologist as their sole provider. This raises two problems:
Should we consider Gynecologists as “PCP” for women receiving their entire care from them?
Since gynecologists also care for other patients on a consulting basis (generally from a referral from a PCP), should they become PCP for these patients, too?
The same problem exists with doctors who moonlight in Urgent Care Centers, take care of people admitted to SNF (especially for short periods), or work part-time for Virtual Care companies.
Primary Care Designation for Mid-Level Providers (NPs and PAs)
Until now, we have been considering MDs and DOs with a “specialty designation” based on board certification.
For Mid-level providers, such as Nurse Practitioners (NPs) and Physician Assistants (PAs), there are very few specialty designations based on board certification. Most NPs and PAs working in subspecialty offices have “generalist board certification.”1
This makes it very hard to determine whether a mid-level provider works as a PCP or a specialist. Some ways that health plans have tried to solve this challenge include:
Determine PCP status based on supervising MD/DO (when providers bill the insurance companies, there is a box for “supervising provider”). In this case, if the MD/DO is a PCP, the patient is generally attributed to them.
If NPs don’t require a supervising provider, their credentialing status with the health plan may be used to determine if they are PCPs.
Requiring patients to pick a APRN/PA as a “PCP” from the insurance website manually
Retrospective vs Prospective Attribution
After defining who a PCP is, the next challenge is to determine which PCP the health plan will hold accountable if patients change their PCP:
PCP who provided care in the past, i.e., retrospective attribution.
PCP who is expected to provide care in the future, i.e., prospective attribution.
Retrospective Attribution
Retrospective attribution looks at which PCP provided care in the past within a defined time frame, aka a look-back period. Insurance companies may use different look-back periods, varying from 1 to 3 years.
If a patient has seen multiple PCPs in the look-back period, then the attributed PCP is based on the following:
PCP that provided the most care - based on the number of visits or total reimbursement received.
PCP that provided the most recent care.
Most attribution models in use today are retrospective. The biggest challenge with retrospective attribution is that PCPs do not know which patients they are being held accountable for during the time period their performance is being measured.2 For example, if the performance measurement year is Jan to Dec 2023, the final attribution (called attribution reconciliation) may not be completed till mid or late 2024.
One can reasonably argue that PCPs should treat all patients equally, and health outcomes should be similar across all patients irrespective of retrospective attribution.
However, dear reader, if you read my previous articles, you may realize that your humble PCP is being held accountable for socioeconomic and environmental health determinants of care under VBC. As part of an ACO, the PCP may have limited resources and will want to use them on patients attributed to them! (The government and health plans abdicated their responsibility to PCPs.)
The best example to illustrate this point is - transportation. If patients don’t have transportation to access healthcare, their PCP, via their ACO, will prioritize providing transportation to attributed patients due to limited resources. The government, on the other hand, can build transportation infrastructure for everyone. And yes, under VBC, PCPs (or the ACOs in which they participate) are put in the position of arranging/providing transportation, housing, or helping with any socioeconomic determinants that may increase the cost of care.
And yes, under VBC, PCPs (or the participating ACOs) are put in the position of arranging/providing transportation, housing, or helping with any socioeconomic determinants that may increase the cost of care.
Prospective Attribution
Prospective attribution attempts to predict which providers will provide care in the upcoming year and attributes patients in advance. This modeling can become very complex and is outside the scope of this article. One point to note is that even if patients are attributed in advance based on a prediction model, there is still an attribution reconciliation several months after the measurement year. Depending on the prediction model, there can be wide variances in attributed patients and the patients who actually were part of the PCP panel.
Attribution to Specialists
Suppose no PCP visit occurred in a given year. In that case, the “accountable provider” will be the specialist who rendered the most care in the look-back period or is expected to render the most care in the future (based on data modeling).
Attributing “Very Healthy” and “Very Sick”
Until now, we assumed the patient was seen by a medical provider. But how do we attribute a patient who is “very healthy” and has no contact with the medical system?
When a person has no contact with the medical system, the insurance company either forces the patient to choose a PCP from the list when signing up for a health plan or automatically assigns one (I have been automatically assigned a PCP!).
And voila, a PCP somewhere becomes responsible for all the quality metrics and health outcomes for a person they have never seen!
On the other hand, people who are very sick, generally with end-organ damage such as kidney, heart, liver, or lung, are predominantly managed by their specialists. Yet, under most attribution models, these people are still attributed to their PCP.
Proliferation of Attribution Algorithms
According to an NCQA report,3 171 unique attribution models have been proposed or are currently used in accountability programs.
For example, Medicare uses a 2-step algorithm (pdf link - page 11) to attribute patients, as shown below.
Since each health plan uses its own attribution model, PCPs are left wondering why some patients are assigned to their panel while others are not—and are made responsible for their cost & quality of care.
The Toll of Patient Attribution on PCPs
Now that we understand the intricacies of patient attribution—let me list some of the problems PCPs face.
Impact on New Primary Care Practice
Under a fully capitated model, providers are (or will be) reimbursed monthly for attributed patients. If, on average, it takes six months for patients to be attributed, these new practices will not generate any revenue during this timeframe despite providing patient care. Furthermore, on average, it takes two years for a new medical practice to become profitable. Under the VBC model, it will take several years for new PCP practices to become profitable due to delays in attribution. This will make it unsustainable to start a new primary care practice.
The confusion between PCP Panel vs. Health Plan Attribution
There is often a mismatch between the PCP’s “real panel” and patients attributed to PCPs by health plans. The delta between this mismatch depends on the following:
Attribution algorithm used
Patient turnover in the Practice, e.g., turnover in a PCP practice caring for young adults, will be higher due to job relocation or patients being discharged for no-shows.
Provider type and role: e.g., patients seen by PCP in an SNF or Urgent Care may be misattributed to them.
There are ways to compensate for these issues (e.g., CPT billing codes for SNF visits differ from office visits), but this makes attribution algorithms more convoluted and opaque. Furthermore, since health plans control these algorithms, they have the final word on who is attributed to a PCP!
This mismatch may penalize (or reward) one PCP at the expense of another. Even worse, health plans may attribute high-cost patients without a PCP to ACO doctors who saw the patient in an acute care setting (e.g., SNF), raising their total cost of care and decreasing their chance of shared savings in VBC contracts. The onus then falls on the PCP/ACO to investigate and challenge the health plan's decision, increasing clerical workloads.
Public Humiliation with Publication of Quality Metrics
I understand the logic that publicly posting quality data will force organizations to provide better care. However, the combination of misattributed patients and data collection/reporting issues can cause a larger swing in quality metric performance for a small practice due to a small sample size (some entities try to account for small sample sizes and will not publish performance). It can be disheartening to providers when their publicly posted quality performance is inaccurate due to misattributed patients, leading to provider burnout.
Role in Consolidation
Patient attribution by itself is generally not a reason for consolidation in healthcare, but it affects performance on VBC contracts. This adds weight (consciously or unconsciously) when providers consider quitting private practice or practicing medicine altogether.
In summary, confusion in patient attribution models leads to confusion and higher costs associated with execution in value-based care contracts.
Up Next
Now that we understand how Patient Attribution works, we will tackle the art and science of medical Risk Adjustment in VBC and the havoc it creates for PCPs.
If you liked this article, please consider sharing it.
You can unlock access to the paid membership tier for free using the link below to share this article via email, text message, or social media. Learn more about these perks and track your progress on the leaderboard.
There are certificate programs for some subspecialties, but board certification is limited to a handful of subspecialties.
The attribution reports distributed by health plans during the performance year are not very accurate. Patients are added and removed every month from these reports. Approximately 6 months after the measurement year, health plans do an “attribution reconciliation” based on pre-specified criteria to calculate health outcomes and cost.
Ryan, A., Linden, A., Maurer, K., Werner, R., & Nallamothu, B. (n.d.). Attribution methods and implications for measuring performance in health care.