Several years ago we highlighted the analytics work behind the Camden Healthcare Coalition’s hotspotting methodology used to identify super utilizers in the local health system so that appropriate interventions could be provided to decrease unnecessary utilization. The approach was highlighted in a widely read article by Atul Gawande in the New Yorker. To refresh memories, the Camden Core Model evolved out of insights from the Camden Coalition Health Information Exchange that developed criteria used to identify those at risk of becoming high utilizers of health services. These criteria included the presence of chronic diseases, co-morbidities, past utilization patterns, housing stability and other social complexities that complicate care.
Once patients at risk were identified they were connected to interdisciplinary care management teams that included nurses, social workers and community health workers who would work with these patients after discharge. Behavioral interventions and social services figure highly in the ‘treatments’ that super-utilizers often require. In many ways, the hotspotting approach has become a mainstream approach to risk stratification and dealing with low-income, high-risk patients who fall between the cracks. The founder of the Camden Coalition of Healthcare Providers, Dr. Jeffrey Brenner, went on to work with UnitedHealth in the following years and has recently received attention for providing housing for high utilizers rather than homeless shelters that only made patients sicker.
An Evaluation Demonstrates the Limitations of Hotspotting
In the fall of 2019, the Nobel Prize in Economics was awarded to Ester Duflo, Abhijit Banerjee and Michael Kremer for their work in development economics and evaluation of anti-poverty strategies utilizing randomized control trials (RCTs). Over the past decade, RCTs have become very popular in the international development discourse for evaluating anti-poverty programs and policies and have uncovered insights such as the well-known microloan/micro-credit programs may not have been as effective at addressing poverty as we had thought.
Recently researchers utilizing the RCT approach conducted an evaluation of the Camden Coalition’s work and were startled to discover that over time the impact or ability of the program to curtail excess utilization decreased. They focused their research on whether pairing high utilizers of hospital services with nurses and social workers could lower readmissions. They found that those receiving extra assistance from the program were just as likely to return to the hospital within 180 days as those not receiving assistance.
The Deeper Insight: Trauma and Social Violence are Difficult to Treat
When interviewed about the findings Brenner noted that the core issue boiled down to the quality of care and that average services fail patients with more complex social contexts and histories of abuse or trauma, particularly in childhood. These patients received a lot of unnecessary testing, inappropriate treatment, and incorrect diagnoses. High rates of addiction and mental illnesses are often due to the traumas we experience earlier in life that re-wire the brain. Addressing these issues requires longer-term and difficult structural changes and interventions that go well beyond typical care coordination and care management programs.
High rates of addiction and mental illnesses are often due to the traumas we experience earlier in life that re-wire the brain.
Statistically, what they observed was also a case of “regression to the mean” where high utilizers in a given year may actually be much lower utilizers in following years. The phenomenon of regression to the mean can confound researchers evaluating programs and lead to over-estimating the impact of programs if they rely only on pre-post data.
Which Data Count?
Perhaps the most interesting lesson from this story is Brenner’s call for addressing the context of structural violence and trauma. Yes, in the health IT space this is a difficult ask because we have a difficult time getting well-off folks to use a mobile app for more than six months or actually become engaged patients when they have a chronic disease. But if we take a look at global health where we are often dealing with the aftermath of wars and extreme trauma there have been successful interventions.
Paul Farmer, an anthropologist-physician and founder of Partners in Health (PIH) pioneered a lot of the work in social medicine that enabled the scaling up of anti-retroviral medications in low-resource contexts such as Haiti and Rwanda. PIH has been collaborating with the Rwandan government for years now in building the health IT system, a new hospital and one of the most innovative health systems in Africa. In many ways, they have accomplished things that Americans can only dream about. Yes. Far better interoperability and drones that can deliver blood supplies and medications throughout the country.
I worked in Rwanda in the 1990s after the genocide and the performance of the health system today was unimaginable back then. One of the keys to Farmer’s grasp of low-resource contexts has always been his anthropological training in ethnography and having a deep understanding of the contexts and politics of poor people’s lives. Quantitative data often only gets us so far. The mere categories we use to frame data may not be the categories patients themselves use. Taking a hard look at power structures has also been central to his work. The other poverty that Farmer addresses is the poverty of imagination that is pervasive among healthcare experts and how we think about the choices of the poor and why they are super-utilizers.
It is encouraging to see UnitedHealth now taking a page from the social medicine handbook and provide housing rather than homeless shelters in order to reduce Medicaid spending and improve the health of the poorest who are most often the high utilizers. Another important feature of this work is that the provision of housing is de-linked from adherence to treatment for mental health or substance abuse. The “housing first” notion recognizes that shelter can provide the foundation for adherence and enabling other positive changes. Something another Nobel laureate in economics, Amartya Sen, recognized decades ago that improving the fall-back position for low-income women could improve their bargaining in the household where they often faced abuse. The result was more investment of household resources in education and healthcare.
In the future it would be interesting to see the marriage of social medicine, ethnography and machine learning to innovate in the area of social determinants. While it may sound farfetched, we are seeing examples beginning to emerge in global health and it may be wise to learn from these efforts as they evolve. Much of the qualitative research also points to the notion that there is no such thing as “raw data” and how we need to interrogate categories used in our analytics occasionally to open up the black box to see what is going on beneath the surface. These analytics count too.