During a patient encounter, we observe only a snapshot of a patient’s clinical presentation.
Something that changes like a pendulum. Fluctuating upon the continuum of patient perceptions at different moments in time. But the ongoing, dynamic relationship between a patient’s perception and the patient’s ensuing behavior goes beyond the patient encounter.
The inability to understand this dynamic relationship explains the limitations in healthcare models that forecast healthcare outcomes. Most models use input from one or a few clinical studies, largely focusing on incidence or prevalence data, and compare the clinical data from the studies to financial data or quantified subjective metrics. The most common metric being quality adjusted life year (QALY), a measure of disease burden, derived in part by survey responses that incorporate data from specific populations.
The resulting inputs into the model are called expected values, a mathematical term that calculates the value of something discounted by the probability that it happens. If a patient needs a follow up mammogram costing $100, and the incidence of breast cancer is 10% for that patient, then the expected value is $10.
In reality, the expected value is not fixed at $10. It varies over time based upon the changing perceptions of the patient. For the woman in the example considering the follow up mammogram, this equation does not approximate her state of mind.
Her perception, and subsequently her decision-making, is based on the uncertainty of the initial result, making the likelihood that she seeks a follow up exam much greater than 10%. As the fear of an uncertain result shifts upward the perceived value of the exam, exposing a flaw in predicting behavior using fixed expected value as inputs.
Behavioral economists have observed that people perceive risk differently when faced with the risk of losing versus the risk of winning. Similarly, the true outcome for a patient is influenced by the interpretation of the outcome relative to how the patient currently perceives his or her health risks. Decisions made in this context are not always rational.
A model that more accurately predicts true patient behavior is called prospect theory, which integrates the impact of emotions into decision-making by adjusting the probability of outcomes with decision weights that account for different levels of emotions. Which produces more realistic perceptions of outcomes.
While the modeling can be quite complex, and many permutations have emerged over the years, each model derives from the core premise that we overvalue small risks and undervalue significant risks.
During the COVID-19 pandemic, many models projecting the mortality rate needed to be adjusted periodically every few months. Mostly because they failed to account for unforeseen behavioral responses by people represented in the model. Behavioral responses that are emotionally derived, based upon their interpretations of the events transpiring around them, and not rational.
Healthcare models should similarly integrate decision weights to approximate realistic patient behavior. Which changes per encounter, per healthcare engagement, in a continuous manner that is also influenced by discrete events.
Take an example of a patient who initially perceives her hypertension to be inconsequential to her overall health. After experiencing a hypertensive crisis requiring an emergency visit to the hospital, she perceives her condition to be serious and changes her lifestyle. That behavioral change is a reaction to an existing frame of reference, with both known and unpredictable consequences. If we model the expected value of her health, we must account for her lackadaisical approach to her health before her hypertensive crisis, her behavioral changes after her crisis, and the impact the crisis has on her long-term health. These are three discrete variables, but they are not fully independent and have varying impact on the patient’s health.
Yet most healthcare care models focus on a few variables, fixed in time, and hold all other factors constant. We cannot amortize a patient’s health over time any more than we can assume this patient maintains a constant outlook on her health regardless of what transpires.
Healthcare is dynamic and complex, in state of perpetual interpretive flux. Models that study decisions without the proper context of the underlying perceptions, or by holding most variables constant, frame patient decisions in an unrealistically rational manner. And are largely wrong.
We should explore models that are derived from decisions made in the context through which decisions are naturally made. For patients that means including the dominant perceptions that affect clinical decisions and behavior.
Rather than shy away from incorporating all the variables involved in a decision, we should elucidate the relationships among those variables and observe the effects of those relationships. Healthcare is filled with unique, unforeseen relationships that are hidden drivers behind patient behavior. Relationships that go beyond traditional clinical care.
Some of the first healthcare models calculated the statistical value of a patient’s life, by estimating and then fixing the economic and social value of that life. Lumping all the nuanced, dynamic relationships between a person’s health and their livelihood into one term called indirect value. A notoriously vague term that defines the loss of value due to the cessation or reduction of economic productivity, attributed to the morbidity and mortality associated with a disease – assuming it to be fixed over time.
Recently, more sophisticated models have attempted to adjust the socioeconomic value based on trends in the labor market. But even with more nuanced inputs, hardly any model attempts to incorporate the changing patient perceptions over time. Ask any patient with a chronic disease and you immediately learn there is nothing fixed about the disease.
When different economic and social factors are accounted for, many patients will have different expected values because patients prioritize things in their lives differently. Yet these models calculate the inputs through aggregated feedback surveys. By assuming broad overarching trends in patient perceptions that may not be true at an individual level.
Another obvious oversight grandfathered into these models. Which we can account for by adjusting the value in these models to better reflect permutations in individual preference, or payoffs relative to the aggregated preferences normally used in the models.
Patient perceptions are a complex concept, and it is not immediately obvious how they impact patient care or larger healthcare outcomes. The changing trends within perceptions affect patient behavior and can prompt reactions that change how a patient makes decisions. These prompts and ensuing changes are usually subtle initially but accumulate to become clinically meaningful. How that takes place is difficult to quantify since it changes in both rational and irrational ways.
A study conducted by Pews Research Center found that 7/10 people thought trust was decreasing in the corporate world, but only 4/10 felt it was a problem – demonstrating an interesting discrepancy between trust and the perception of trust. Trust, in our hyperconnected, online world, has moved from a centralized to decentralized concept, as trust in major institutions is giving way to trust in people’s own judgments and experiences.
When people surveyed felt that trust was decreasing, fewer than expected felt it was a problem because they developed a different perception of trust. Which is different than a reaction to the original perception of trust towards a centralized financial institution. Both are changes in perceptions, but the way the change took place is unexpected.
Similarly, perceptions among patients change in unexpected ways, and makes the study of patient perceptions regarding healthcare outcomes difficult to characterize. As perception changes are unpredictable and irrational, yet critical in modeling meaningful patterns of patient behavior.
Healthcare models must examine the impact of biases and irrational decision-making on patient care. We will never be fully certain of any behavior or of any decision. The best we can do is optimize the uncertainty in the decisions we make by structuring the perceptions influencing these decisions.
But in the fast pace world of healthcare, moving ever faster, we reflexively shift our awareness towards our dominant perceptions. And that becomes the basis for our decisions.
Which is anything but rational.
Vaccination rates vary by county, determined by local factors
COVID-19 has disproportionately affected certain underserved and high-risk populations, including people of color, those with underlying health conditions, and those who are socioeconomically disadvantaged. Ensuring access to COVID-19 vaccines for these communities can help address the disparate health effects of the virus and achieve herd immunity.
The Biden administration has identified vaccine equity as a priority, but states and local jurisdictions vary in how and the extent to which they prioritize equity. Given that vaccine roll-out in the U.S. is inherently local, understanding how vaccination rates vary at the local level is important for informing outreach efforts and addressing equity.
Earlier CDC analysis found that, as of early March, counties with high social vulnerability had lower vaccination rates than counties with low social vulnerability.
Source: Kaisesr Permanente Foundation
Dr. Anandi Gopal Joshi, the first Indian physician trained in the United States
Anandibai travelled to New York from Kolkata (Calcutta) by ship, chaperoned by two female English missionary acquaintances of the Thorborns. In New York, Theodicia Carpenter received her in June 1883. Anandibai wrote to the Woman’s Medical College of Pennsylvania in Philadelphia, asking to be admitted to their medical program, which was the second women’s medical […]