Market participants apply different models (some sophisticated, some not) to ascertain the value of a security. Their collective input generates a market bid-ask spread, not always accurate but in most cases more accurate than any one entity could otherwise determine over an extended time period. (Of the 8,500 plus mutual funds almost none beats the market consistently.)
Seems like logical process, right? Well, we should be at least familiar with it since we engage in price discovery in just about every aspect of our lives. That is every aspect except for health care.
In health care, the actual delivery of care (going to the doctor for a checkup, for example), a transparent price discovery process does not exist. Health plans contract with provider groups to pay fees for services that physicians perform. The price setter is in many cases the U.S. government, specifically Medicare the single largest payer which covers health care's highest utilizers, folks aged 65 and older. (Picture Goldman Sachs times one hundred, and everyone else following its lead.)
Unlike the capital markets, a transparent marketplace does not exist. Commercial health plans don't compete across state lines, and within states, plans can exercise near-monopoly power.
Moreover, because of the first-dollar nature of health insurance, what plans charge beneficiaries in premiums includes considerable actuarial assumptions about the health of their population, and how this changes over time.
Health reform (whether the Patient Protection and Affordable Care Act, "PPACA", or competing proposals) targets fundamental change in the payment system.
PPACA requires plans to offer equal access and prohibits differentiated pricing based on health status. It also establishes the formation of accountable care organizations ("ACOs"), care coordination vehicles that require providers to assume degrees of risk.
As a result, risk selection—a result of not having full information in the marketplace—is shifting from plans to providers as a business strategy.
But before providers can consider selecting risk they need to model and quantify it first. Not an easy task when patient populations vary substantially, depending on demographics and location.
The process of measuring risk factors for the purpose of risk-adjusted reimbursement is called simply risk adjustment. (Something that Wall Street does quite efficiently in a marketplace setting.)
As with any type of modeling, risk adjustment is an imperfect science, especially as it pertains to a price discovery process that involves thousands of beneficiaries but only one payer and one provider.
In a recent white paper written for the Massachusetts Medical Society, Milliman, the actuarial consultancy, tackles the intricacies of risk adjustment and submits key principles that both the payer and provider need to heed when looking to form ACOs. [Read the Milliman paper here.]
Many people, the Milliman paper explains, view the ACO concept as a viable alternative to the existing fee-for-service payment system. Not waiting for the U.S. government, commercial entities have already conceived and deployed ACO-type models, which included both a risk-adjusted global payment and a performance-based payment.
The government model continues the fee-for-service system but introduces additional payments based on a set of benchmarks for health care costs, outcomes, and quality. HHS is expected to recognize risk adjustment tools that will determine how much it reimburses groups for exceeding these benchmarks.
Milliman lists the following five key risk adjustment design principles:
- The groupings of medical conditions in a risk adjustment model should be clinically meaningful and reasonably specific, in order to minimize opportunities for gaming or discretionary coding.
- Diagnoses within the same condition category should be reasonably homogeneous with respect to health care cost and utilization, in order to optimize predictive accuracy and robustness of the model.
- Condition categories should have adequate sample sizes, to permit accuracy and stability of model predictions.
- The risk adjustment model design should encourage specific coding and discourage vague coding. Vague codes and nonspecific diagnoses should be excluded from the risk adjustment model
- The risk adjustment model should not reward coding proliferation. Providers should not be penalized for recording additional diagnoses. In other words, coding more diagnoses should not reduce the risk scores.
The ACO environment requires additional principles to address the organization's core competency and patient assignments. Under an ACO structure, the care delivery entity (a hospital or physician group practice) bears responsibility for assigned patient outcomes regardless of who ultimately provides the care—for example, a patients seeking care outside his assigned ACO.
What's more, there isn't a single risk adjustment model. Selecting and engaging a risk adjuster involves a detailed understanding the model itself and how it compares with different vendors.
And there's the matter of coding, the record-keeping system physicians use to indicate diagnosis and treatment. How physicians enter codes affects risk scores, and the coding system itself (encompassing tens of thousands of codes) is changing to an even more extensive system—a sort-of Y2K for health care—with the introduction of ICD-10-CM.
Imagine now that you're a physician and that your practice or affiliation is considering an ACO structure featuring a highly sophisticated risk adjustment process.
For many, the obvious answer is 'why bother?' At least in the fee-for-service model you had a clearer understanding of what your reimbursement rates were. Your compensation now comes down to a pool of money that the ACO's officers will allocate based on a black-box model.
Some who see no benefits in either system are existing the insurance model altogether, or simply vacating the profession.
Over recent years, Wall Street hasn't exactly demonstrated perfection, but at least the marketplace model is transparent. The answer to risk adjustment's complexity as it's implemented might just be market competition and consumer choice.