A Model of More Effective Chronic Condition Management

At ProjectVision, we talk a lot about two elements that are essential to building condition management programs and interventions that not only retain patients, but keep them actively engaged in behavior change:

  • Personalization
  • Quantifying environmental barriers to behavior change

Because of the difficulty in gathering complete behavioral data, many chronic condition management programs focus “personalization” based on clinical risk – A1C level, BMI, blood pressure, etc. Using historic demographic data, they can predict with reasonable accuracy when a patient will shift from Pre-Diabetes to full Type 2 Diabetes.

However, clinical risk doesn’t tell us the “why” in the same way that behavior risk does. Focusing on clinical risk also pushes providers into a mental trap of treating patients with the same clinical risk and demographic characteristics as monolithic. Studies have shown that not more than 20% of patients who successfully lose weight in a structured program are able to maintain that weight loss in the long term.

The unsatisfactory long term outcomes of current condition management programs points to three key issues with the current models of chronic condition management:

  • The program is disconnected from the social, cultural, and environmental reality for most of the patients
  • Behavior risk stratification is not used to further tailor the program to each patient’s unique psychological needs
  • The environmental barriers faced by patients are beyond the scope of clinicians and care providers

So let’s dive deeper into each of these three issues:

Programs Disconnected From Patient Reality

In early November, I presented a webinar centered on more effectively engaging low-income and minority patients in health behavior change. I started that presentation by analyzing the rapidly growing amount of money retailers were spending on digital ads, making the point that in an industry where failing to understand your customers deeply means going out of business, companies make a big investment in tools to help them understand and speak directly to their customers.

A big problem in healthcare is that biases about certain patient groups – the overweight, the poor, the non-white – are allowed to drive program development without being acknowledged and confronted. Failures to recruit patients in adequate numbers or keeping patients engaged in programs often comes down to incongruent messaging on the part of providers. For example, using pictures of thin, white women doing yoga when advertising a condition management program to mothers of color in lower-income communities with high obesity prevalence will not create the necessary connection with those we are trying to reach.


Programs that don’t acknowledge and build behavior change around existing deep structures within a community can expect to continue to get unimpressive results. Making this change requires letting go of assumptions about how people in certain communities think and feel, and instead, relying on observations from actual data gathering.

Behavior stratification

Simply put, a good program is responsive to psychological variance within the participating patient population. In an inpatient setting, the frequency and intensity of care is based on the clinical risk faced by the patient. Good chronic condition programs should be able to similarly moderate the level of care provided based on the behavioral risk faced by each patient. We need to move away from one-size-fits-all models of intervention. There are many different paths a person can take from being healthy to becoming diabetic.



Effective Interventions Are More Than Just Clinical

We are shaped by our environment, and we generally take the path of least resistance when making day to day decisions. If a patient lives in a healthy environment – where they have walkable access to healthy food options that they can afford, with clean air, low crime, and easy access to recreation – they are more likely to be healthy. Obesity rate has a very strong negative correlation with income.

By the same token, it’s not within a hospital’s scope to provide adequate healthy food options in a food desert, or job resources for unemployed males in high crime neighborhoods, or for addressing a myriad of other social determinants of health that are not strictly clinical.

In order to be more effective, chronic condition programs run by clinicians need to engage with non-clinical resources in direct partnership. They need to work in concert with food banks, legal services, community health programs like the YMCA, and others to properly address the environmental issues that are causing their patients to fail in more traditional programs.

This more networked approach helps address the two key issues above around being disconnected from the patients’ reality and lacking true behavior stratification. If we understand what each patient’s personal goals are as well as their unique barriers, we can better understand the norms that shape their life and the specific type of support they need to get the most out of a clinical behavior change program. By partnering with non-clinical resources, we can now start actually tailoring the services around those unique needs in a scalable fashion.

By building networks of resources to support patients, we can start leveraging network effects to exponentially increase the value of condition management programs to patients as more resources and more patients become part of the network. You start to have opportunities to leverage past participants as peer advocates as well as your non-clinical partners to bring in billable, reimbursable patient cases.

Bringing Things Together

Current condition management programs have long-term success rates that are anything but impressive. Given the cost concerns of CMS, it is imperative that we aggressively identify new models that can more reliably support behavior change.

Successful new models will address implicit bias and involving high-risk patients more directly in care planning, focus data collection on behavioral risk stratification, and build close partnerships with non-clinical resources.

Why We Need Health-Centered Public Policy

Social determinants of health such as crime, quality of education, and transportation mobility play a major role in the health expectations of communities, particularly around chronic conditions. In order for medical interventions to be most effective, public policy must support an overall reduction in the barriers certain communities face to health improvement.  We know that in the United States, weight-related conditions such as Type 2 Diabetes, Hyptertension, or Obesity are not medical problems, per se.  We know, medically, how to mitigate the effect of such conditions on a patient’s quality of life.  The problem of a rising obesity-rate and its subsequent impact on comorbid conditions is largely behavioral.  What makes the behavioral component so difficult to address is that even though there have been historically clear relationships between obesity and poverty, the overall rate of obesity has risen such that it is becoming a characteristic of the overall population rather than just a product of economic disparity.

As a result, the tired argument of blaming the poor for their own issues is a non-starter.  Even simply blaming the existence of food deserts or the unaffordability of healthy foods misses more foundational issues.  The whole of the population is susceptible to obesity and obesity-related chronic conditions, with apparently different factors driving obesity for different segments.  Recent studies have shown that by the early 2000’s, high income earners were the demograpic that experienced the fastest growth in obesity, which should encourage a more holistic look at the behavioral factors that might affect such a trend.

Using an urban planning simulator like Sim City, we can examine the civic factors of urban design such as transportation choices, work preferences, zoning, and planned development to better understand the aspects of public policy themselves that increase behavioral risk for obesity.

Modeling reality

While Sim City doesn’t really tackle microeconomics like car ownership or household consumption choices, it does provide a good model for conceptualizing how public planning decisions drive macro behaviors.

Take zoning – which in Sim City is limited to Residential, Commercial, and Industrial.  Placement of zones when first building a city has long term consequences on traffic patterns, prevalence of air and water pollution, land value/income demographics, quality of public services such as education, and of course citizen health.



In the example above, the green zones represent Residential zoning, the yellow Industrial zoning, and the purple Commercial zoning.  This city has multiple avenues surrounding the commercial district in the center, allowing a high volume of traffic to funnel in to and from the surrounding Residential and Industrial Zones.  Much like real life, high density development without appropriate scaling of the capacity of roads will result in traffic congestion.



In the picture above, just like in Google Maps, red roads mean heavy traffic, while green roads mean clear traffic.  And because cars emit carbon monoxide exhaust, heavy congestion predictably results in air pollution



As you can see, the heavy traffic roads also have higher pollution.  The heavy pollution at the bottom right of the screen also comes from the industrial zone.

Based on the preferences of the citizens in this Sim City town, we can see that the car is the favored form of transportation.  Wealthier citizens are willing to put up with long commute times before switching to alternatives if their place of work is beyond a short walk away.


The net result is that in spite of high quality hospitals nearby (the bright green buildings), the overall population health in the low-income residential areas nearby the commercial district is disappointing – where green is healthy and red is unhealthy.


These same concepts apply in real life as well, where the way citizens respond to urban planning and public policy leads to health hindering behaviors.  Long commutes by car to jobs that involve sitting for most of the day result in worse health for the commuters, while the pollution caused by emissions from heavy traffic negatively impact the health of citizens throughout the region.  Sprawl of the city population out to suburbs that causes the long commutes also increases the expense of providing public services like police, education, and utilities, straining the city’s ability to equitably serve the whole population.

Without even touching issues of food insecurity or food deserts we can see the complicated relationship between public policy and health.

Applying concepts to reality

Sim City is just a simulator, but it still provides us with a window for the interconnected nature of different public policies.  Our understanding of policy is often dominated by “single issue” special interests that frequently leverage emotional appeals to gain political legitimacy.  However, what we can learn from applying concepts from games like Sim city onto our empirical experience is that no matter how singular the intended effect, all policies have ramifications.

More importantly, these ramifications drive behaviors of citizens at a micro and macro level.  The decision to build a new retail development in a certain part of town will affect traffic, strain water/power infrastructure, and may result in economic stress for certain portions of the existing residents.  All of these effects will have some measurable impact on population health and health care expenditures.  Given that the physical and mental health of citizens is central to maintaining productivity, reducing crime, and supporting an overall high quality of life, health should naturally be a central focus of public policy.  Questions such as “how will this impact health-related behaviors?” and “how might this affect health-related expenditures?” should be a major consideration by policy-makers when evaluating any new development.

The case for health-driven policy

There are three major arguments in favor of a “health-first” approach to public policy:

1) Our national and local healthcare expenditure is a huge economic problem

2) Government and Payers are shifting responsibility for patient outcomes directly onto care providers (value-based care)

3) Heart attacks and strokes are not partisan issues

1) Our healthcare expenditure is a huge economic problem

Not very many would argue the fact that within the US, from the patients’ perspective, health services are generally not cost effective.  When adjusted for Purchasing Power Parity, the US outspends every country in the world per patient.  Yet the expenditure is not reflected in health outcomes.


I discuss in another article why life expectancy may not be the best measure of health outcomes, and how we might want to look at measure such as obesity rate instead.  But end result is the same; a great many other countries spend far less per person and get much better health outcomes.

Beyond health outcomes, there are significant economic consequences to having an expensive healthcare system that is still ineffective at addressing population health issues such as obesity.  For example, obese employees file twice as many worker’s compensation claims as non-obese workers, and a roughly 10-fold increase in loss of workdays (183.63 versus 14.19 per 100 FTEs), medical claims costs ($51 091 versus $7503 per 100 FTEs) and indemnity claims ($59 178 versus $5396 per 100 FTEs).  Absenteeism due to obesity-related issues cost employers a total of $153B per year(1 cite) while total obesity related care delivery costs range anywhere from $80B to $120B.  Given total health expenditures are roughly $1.4Trillion, that is nearly $1 in $10 spent as a result of obesity.

Another major economic issue is how patients pay for care.  Unpaid medical bills are the number 1 cause of personal bankrupcy in the US.  Even among patients with health insurance, 10 million adults with year-round insurance will accumulate medical bills that require multiple years to pay off.  In today’s climate, a person who suffers a severe accident due to misfortune may potentially end up removed as a participant in the economy as a result of the cost of care.  To have patients who have preventable conditions that have gotten out of hand and whose families are paying medical bills for decades, is nothing but pure failure.  To what extent this health crisis has stifled our ability to recover from the Great Recession we may never know.

2) The rise of value-based care

In response to the extreme cost inefficiencies of the standard, fee-for-service healthcare model, both private and public payers have been undergoing a philosophical shift over the past several decades.  Even before the Accountable Care Act nudged health insurers into an open market, there have been several key drivers that have lead to the disintegration of fee-for-service

  • The unsustainability of insurance premiums that have risen much faster than income growth
  • The pending insolvency of Medicare/Medicaid at current cost trends
  • A chaotic care coordination process that relies on patients knowing how to navigate multiple health systems


Patient advocates and fiscal conservatives are aligned in their incentive to develop new financial models of care that push more risk onto care providers.  One such model – the Accountable Care Organization – attempts to establish key metrics on which to evaluate quality of care delivered, and shares cost savings with providers if certain quality and cost standards are met.  Many other HMO’s, integrated payer-provider systems, and physician networks are developing other innovations on the traditional reimbursement model that simultaneously seek to improve the quality of outcomes and reduce total cost per patient.

One of the main benefits of the philosophical shift towards value-based care will be the increased data collection across the continuum of care.  This will enable vastly improved behavioral and health risk modeling, and allow policy-makers to more accurately predict the health impacts of potential policies.

3) Heart attacks and strokes are not partisan issues (yet)

Unlike vaccines or abortions, no one within the political realm is arguing the science of treating obesity.  The near ubiquitous acceptance of unhealthy diet and lack of exercise as core contributors to obesity and obesity-related conditions means that clinicians have room to focus more on medical or behavioral therapy rather than defending science.

Further, because population health is such an integral component of value-based care, care providers and payers are in a unique position to influence dialogue around policy as it affects health.  The fact that obesity has not yet been politicized means that clinicians and health administrators can speak about the health impacts of weight-related chronic conditions from a relatively unquestioned position of expertise.  This legitimacy is a crucial component of a health-centered public policy.

Whether or not the rest of the political sphere, including lobbyists and other special interests, can be persuaded to subordinate their priorities to citizen health is another matter altogether.  They are unlikely to do so if the clinical data predicts negative health impacts from adopting their policy positions.  Nonetheless, it remains of utmost economic importance for citizens, care providers, and payers to work together to present compelling arguments for policy-makers to evaluate policy proposals on the short and long term affects on population health.



Building Robust Models for Predicting Health Behaviors

As the healthcare tech market matures, the focus of consumer oriented weight management products will shift from merely data gathering and generic weight loss plans to applied analytics for building successful personalized plans for change resistant users. To be most effective at addressing preventable health costs, efforts need to be focused on patients at the wrong end of disparities in health outcomes.  The key will be identifying predictive frameworks through which data can be transformed into insights.

One of the frameworks that we employ within the algorithms behind Yabbit, our behavioral risk management platform, is the Health Belief Model.  We talked earlier about the Health Belief Model and how it works. It states that an individual’s likelihood of engaging in a health related behavior is determined by his/her perception of the following six variables:

  1. Perceived susceptibility (perceived risk for contracting the health condition of concern);

  2. Perceived severity (perception of the consequence of contracting the health condition of concern);

  3. Perceived benefit (perception of the good things that could happen from undertaking specific behaviors);

  4. Perceived barrier (perception of the difficulties and cost of performing behaviors);

  5. Cues to action (exposure to factors that prompt action);

  6. Self-efficacy (confidence in one’s ability to perform the new health behavior).

These six health determinants identified by the Health Belief Model together provide a framework for designing both long and short-term health behavior interventions for conditions that extend further than tuberculosis screening programs: osteoporosis prevention, smoking cessation and healthy eating are all valid and potential uses of this model. The common denominator is that the Health Belief Model focuses on health determinants; therefore, it is most suitable for addressing problem behaviors that have health consequences.

So how predictive is the model in general?

The initial four HBM variables (susceptibility, severity, benefit, and barrier) have been shown to predict about 20% of variance in healthy behavior. In the pursuit of greater predictive power two additional variable, self-efficacy and cues to action were added by 1988. Domain-focused studies for areas such as HIV related mental health have also added domain specific variables to the HBM. For example, a 2003 HIV study extended the HBM to include HIV-related stigma as a variable. Such studies, when combined with the self-efficacy variable addition, have seen an average predicted variance around 40%. The HIV study saw predicted variance of 63%. The downside of the domain specific approach to extending the HBM is that the added variables are not broadly applicable to a wide range of health behaviors.

To further expand the broad predictive power of the HBM, we have also incorporated the variables used by Orji, Vassileva, and Mandryk of the University of Saskatechewan in their 2012 study:
Self-Identity, Perceived Importance, Consideration of Future Consequences, and Concern for Appearance.

Self identity refers to a patient’s self-perception in relation to a particular health behavior – ie “I am a health conscious person” or “I have an incurable sweet tooth.” Perceived importance refers to the value a person places on the outcomes of a particular behavior, both positive and negative. Consideration of future consequences refers to the extent to which a person’s decision-making is influenced by the potential outcomes of their current behavior. Concern for appearance refers to the importance a person places on behaviors that affect their physical appearance.

The HBM, when extended by the variables included by Orji et al, increases its predicted variance in healthy behavior to 71%. In conjunction with the biometric and GPS data we are also gathering with Yabbit, we provide a system with a high predictive threshold for patient behavior at a very granular level by various demographic characteristics. This presents the necessary insights for care providers and population health managers to work together with health plan administrators to build more cost effective models for supporting healthy behavior change.

Buy It vs Build It – Why Care Providers Should Stay Out of Tech Development

There are a number of seismic shifts underway within the US healthcare system. The two that we’ve had our eyes on are the rise of capitated provider networks and the subsequent focus on population health management and preventive care to drive operational cost savings. The reason we’ve followed these two so closely is that they so completely expose the lack of proper data infrastructure supporting chronic condition management – specifically weight-related chronic condition management.

A good measure of the overall success of the capitated reimbursement model is the performance data from the Medicare Shared Savings Program (MSSP). The recent Year 2 report released in late 2014 indicates that while overall, participating Accountable Care Organizations have done a respectable job increasing overall quality, they are struggling to achieve a similar level of cost savings across the board.

Figure 1 – Year 2 (2013) Reporting data from the MSSP


While the composite scores are moving in the right direction from Year 1, a closer inspection of some very specific quality measures tells a very different story.

Figure 2 – Year 2 (2013) Selected Average Quality Measure Scores


CMS Quality Measure*

Group with Both

Hospital Group

Physician Group

All ACOs

5 – Health Promotion and Education





6 – Shared Decision Making





16 – Adult Weight Screening and Follow-up





22 – 26 Diabetes Composite





28 – Percent of beneficiaries with hypertension





Some of the lowest average quality measure scores were in areas relevant to weight-related chronic condition management as well as patient engagement.

There are a few potential justifications for why these scores lag so far behind the average scores in across other measures.

  1. Legacy of event-driven, reactive practice means that there is plenty of infrastructure and evidence-based approaches for reducing prevalence and severity of highly acute diagnoses.
  2. Revenue traditionally has been driven by inpatient events (related to item 1), and so the bulk of a hospital’s operating budget has been dedicated to serving and supporting inpatient care activities. Long term outpatient preventive care brought in proportionally very little.
  3. Unhealthy weight gain happens very slowly, and takes years of mismanagement to result in the highly acute outcomes that traditionally get close attention

The MSSP and capitation in general turn the traditional finance model on its head. Under capitation, reimbursement risk is passed from the payer to the provider as the relationship between the payer and the patient shifts from event-based coverage to managed care. Providers are not just accountable for treating whatever a patient shows up with that day, but also for their long term health outcomes. In this context, extremely low quality scores across the board in items related to weight management and patient engagement/education are very problematic – especially in light of the fact that the nation is in the midst of a long-term obesity epidemic. It’s problematic because

  • effective preventive care is the key to long term cost savings
  • active patient participation is critical for preventive care to work
  • gradual gains in body fat drive a myriad of other chronic conditions that become difficult and expensive to treat if poor behaviors are entrenched

In short, ACO and managed care plan long term profitability depends on a highly motivated and engaged customer base, and robust preventive care services. Which is a major weakness of Medicare ACO’s across the board.

This is a problem that needs to be addressed within the time frame of 1 or 2 years, for MSSP or Pioneer ACO participants, in order to maintain shared savings eligibility. But given the historical rate of change within healthcare, it is not reasonable for robust, consumer-technology based programs that assist in preventive care services to be developed in-house. In comes the market and a sea of digital health technology vendors who can provide solutions off-the-shelf or customized to meet unique needs.

Healthcare technology firms that display a good understanding of the business dynamics in which provider networks operate will become the invaluable partner as these networks seek scalable preventive care frameworks that improve patient outcomes and boost the Customer Lifetime Value for patients within managed care plans. These technology firms have the flexibility of rapid development cycles and access to a level of technical talent that simply has not been attracted to work within traditional healthcare IT. Additional benefits include:

  • Driving innovation by exposing healthcare to tech infrastructure paradigms common in other spaces but unheard of in healthcare
  • A patient centered design focus for programs that leverage digital health technology to improve patient engagement and satisfaction

Look to see digital health technology firms become much more influential in driving new paradigms in how patients interact with providers and payers, and how providers can work together with patients to drive healthy outcomes and reduced cost of utilization.