This week, Clinical Informatics News is exploring those intangible, sometimes invisible, but altogether indispensable pieces of the health care system – the patients themselves. How can data-driven methods be brought to bear on patients’ lives outside the clinic? And if we ask the right questions, can they tell us how to provide better care?
By Aaron Krol
October 17, 2013 | The management team of Eliza Corporation was having a troublesome meeting. It was 2009, a decade after Eliza was founded with the goal of translating one-on-one interactions with patients into novel outreach programs that improved health behaviors and the quality and cost of care. Many of their programs had done just that, said Alexandra Drane, the Founder and Chief Visionary Officer of Eliza, in an interview with Clinical Informatics News: “With a specific population, with a goal in mind, we [could] absolutely bend trends.” Yet despite the efforts of Eliza and similar companies, the national trends in health outcomes were moving relentlessly downward.
Poring over a list of the major health behaviors Eliza and its partners recommended – basic things like healthy diet, exercise, and getting preventive screenings – the members of Eliza’s management realized that even they didn’t follow through with every item on a regular basis. “We’re in the healthcare space,” Drane remembers thinking. “We’ve dedicated our lives to this mission, we’re over-informed, over-resourced – and we don’t do it. How could we ever expect that anyone else would?”
A major premise of health and wellness outreach in America has been that better education, more information about leading a healthy lifestyle, will inevitably lead to clinically recommended behaviors. But what if a lack of good information isn’t the only, or even the biggest, obstacle to people taking charge of their health? Discussing their own barriers to changing their lifestyles, the Eliza team discovered that none of them had to do with poor information. It may seem obvious to say that everyday life challenges – problems like struggling with work, or living through a fraying relationship – are much more likely to get in the way of our health than not understanding the food pyramid, but it was Eliza’s insight that if these factors play such a profound role in our health outcomes, they belong in the purview of the health care system.
As a result of that meeting, Drane and her colleagues came up with a list of eleven life-context variables that might affect health outcomes, which they labeled the “unmentionables” after the reluctance of medical researchers to examine them. The biggest five were financial stress, workplace stress, relationship problems, sex life, and the demands of being a caregiver; other unmentionables might be “magnifiers” that intensified the effects of these obstacles, like lack of sleep, or “buffers” that relaxed the pressure, like having a strong social support network. Eliza's object was to see whether their own anecdotal observations would be borne out in large-scale patient studies, comparing patients' private life contexts against their health outcomes. To hone this heterogeneous list of factors into a useable measurement tool, Eliza partnered with the Altarum Institute, a nonprofit in the field of health system research.
“This Might Be What’s Going on in Health”
Eliza wasn’t the first organization to draw a connection between the “unmentionables” and poor health. A review of the literature, like the one Eliza and Altarum undertook in 2009, will reveal a wealth of studies that examined, say, a cohort of heart disease patients, and found the results confounded by these messy variables. Wendy Lynch, the Co-Director of Altarum’s Center for Consumer Choice in Healthcare and a member of the project with Eliza, recalls reading phrases like, “‘and it seems to be the worst for those who also have trouble with sleep,’ or, ‘it seems to be better for those people who have a strong peer network.’... We know that these things matter, but what we do is control for them and take them out of the equation to get at what’s going on in health. Rather than deciding that this might, in fact, be what is going on in health.”
Gradually, these variables took shape as a measurement tool, in the form of an automated survey where respondents would self-report their experience with each unmentionable on a sliding scale. Two unmentionables, sex life and workplace stress, had to be dropped because they didn’t yet meet the triple criteria of accurate self-reporting, actionability and being reliable predictors of health outcomes. The other nine became the Vulnerability Index, a name that was chosen with some care. “Really what we’re talking about is resiliency,” says Drane. “But what we heard from a lot of folks was, ‘I’m not interested in being rated negatively on a positive… but I’m really interested in being validated in something that is hard.’” High scores on the index therefore signified high levels of vulnerability – an indication that everyday life challenges might put an individual at uncommon risk for adverse health events.
Eliza had long been a leader in automated patient interviews, and the company was quickly able to begin measuring vulnerability in the real world. With information on the vulnerability scores of a large volume of patients, Eliza and Altarum strongly suspected they had uncovered a new perspective on health risk that could have enormous value to the health care system. However, at the outset they could only map vulnerability to outcomes retroactively, leaving open the possibility that the Vulnerability Index’s predictive power was just an artifact of the data its designers fed into it. What Eliza needed was partners in the insurance industry who were willing to measure their patients’ vulnerability, and see what correlations existed between scores on the index and future insurance claims.
The Data Pours In
Getting patients to volunteer the private sources of distress that stood between them and a healthy lifestyle was, perhaps surprisingly, not an obstacle. Eliza has found that around 90% of people who start taking the Vulnerability Index complete the entire test. In one vulnerability survey of the Medicare dual-eligible population, 94% of people contacted expressed satisfaction with the encounter – an unheard-of figure at a company that has engaged in over a billion patient interactions. “We’re hearing people saying, ‘thank you for finally asking about the things that matter to me,’” says Drane.
On the other hand, persuading health providers to take the first step into these seemingly intractable problems has taken some coaxing. Almost three years passed from the creation of the Vulnerability Index before a few of Eliza’s clients started adopting it. Today, a handful of large providers like Cigna have been using the Index for around a year, and claims data is finally being released for patients with vulnerability scores.
The early results are striking. If you compare those patients who scored highest on the vulnerability scale against those who scored lowest, the claims data shows an average difference of $4,000 worth of insurance claims over the course of a year. For conditions that showed the closest correlation with vulnerability, like mental health problems and back pain, the highest-scoring patients generated between three and five times as many insurance claims as their low-scoring counterparts. High-vulnerability patients also required three times as many prescription medications and were more likely to log a hospital visit. Perhaps most dramatic is the role of vulnerability in people’s own perceptions of their health: knowing a person’s vulnerability score predicts between twice and three times the variance in self-reported health as knowing the full list of conditions that person suffers from.
Drilling into the demographics of this data provided further validation for the core concept of vulnerability. Lynch had initial concerns that vulnerability might be acting as a proxy for income, so she broke down patients into Medicaid, Medicare and commercial populations. “Medicaid populations did report the highest rates of vulnerability,” she said, “but within the ranges of income, more money did not guarantee you lower vulnerability.” In fact, vulnerability mapped to worse outcomes at every income level, diverging sharply from income as a predictive factor in patient health. What Eliza had developed was not a new perspective on a known trend, but a novel and strongly explanatory set of variables driving costs in the health system.
Alexandra Drane (L) was a panelist at the NEHI Innovation Conference on Patient Engagement this October. Image credit: NEHI
“Hungry for New Ideas”
Individual findings from Eliza’s work with the Vulnerability Index may be startling: the fact that high-vulnerability patients are two and a half times more likely to have diabetes, for instance, or that men are more likely to report caregiver stress in the Medicaid population than women. But when you back away from the specifics, the core conclusion is unlikely to come as a surprise to anyone. After all, we almost all live with the effects of vulnerability. 95% of people report problems with at least one unmentionable, and 40% are struggling with four or more. While Eliza’s work is allowing us to quantify the scope of the issue, and develop a language for discussing it, these problems could have been inferred long ago. Actually addressing the toll of vulnerability on our collective health will require a new mindset from clinics, insurance companies, and employers.
Still, certain measures to address the separate components of vulnerability already exist. Drane cites the example of Employee Assistance Programs, the employer-provided benefits that offer paid counseling or referral to specialists for life problems outside the workplace. These are widespread in the employer-provided insurance matrix, but little utilized or promoted. There are also free social networks available for caregivers, like Caregiver SOS – but few people know about them, despite caregiver stress being the single most powerful item on the Vulnerability Index correlating to poor health outcomes. While brand-new services to help overcome the unmentionables could yield substantial benefits for patients, and savings for providers, in the short term there are opportunities to be seized simply directing people with high vulnerability to qualified resources that are already out there.
Cigna, an early adopter of the Vulnerability Index, has launched a pilot program to reach out to its members with high vulnerability scores, and offer a live health coach or relevant online resources. Other insurers may soon follow suit. “The health care space is hungry for new ideas,” says Drane. “And when you have evidence that something’s working… they’re very quick to jump on board.”
Just as hungry for new ideas are the patients themselves. Eight-five percent of Vulnerability Index respondents said they would like help dealing with their vulnerability factors, rising to 95% for those with caregiver stress – with no statistically significant difference emerging when asked whether they would like help from a physician, employer, or insurance provider. Despite the sometimes deeply personal nature of the factors that make up vulnerability, patients are eager to share their information if it means someone will reach out to them about the problems that matter most to their lives and health.
This is “how the average human thinks about when they feel well and when they don’t feel well,” says Drane. “It’s really the healthcare system that has arbitrarily defined health in a more limited way.” It may be that insurance companies and clinics offering resources to patients once they develop specific conditions is akin to doctors waiting to intervene until an infection becomes bad enough to amputate. By asking patients about their life-context obstacles to health, and reaching out to the most affected before their vulnerability becomes symptomatic, the health care system may be able to save a great deal of cost and hardship.
Read Part 1: Judy Hibbard and the Patient Activation Measure
Read Part 3: NQF and the Standards for Patient-Reported Outcomes