The U.S. health care system is increasingly testing programs to address patients’ social needs. Examples include large integrated delivery systems such as Kaiser Permanente and pilots funded through the Centers for Medicare and Medicaid Services (CMS) such as the Accountable Health Community (AHC) model and Comprehensive Primary Care Plus (CPC+) initiative. Within the next year, the number of patients and beneficiaries screened for social needs and referred to community services will jump from thousands to millions.
As we move toward a more person-centered and equitable health care system, the ongoing analysis of these and other related programs will shape our understanding of how such approaches affect the total cost of care, patterns of utilization, and health outcomes. Yet with all this experimentation and innovation, we lack a shared definitional framework of what success looks like in addressing social needs.
No Common Definition of Success
The lack of common definitions for interventions addressing social needs makes comparison across populations and interventions difficult, and it jeopardizes efforts to understand how such interventions impact overall value, i.e. cost, experience, and health outcomes. These challenges are similar to those in mental health, where non-standard measurements of severity (largely qualitative and subjective), unclear comparative impact of various interventions (e.g. group therapy vs. individual therapy vs. drug therapy), and strong vested interests make consensus-driven standards harder to achieve.
For example, for a patient unable to afford healthy food, various interventions have defined “successful resolution” of food insecurity as:
- A case manager provided an information sheet listing nearby food pantries
- The patient told her case manager she intends to visit the food pantry
- The patient told her case manager she received food at the pantry, and
- The food pantry notified the case manager that the patient came to visit and received food.
All of these are legitimate ways a social need might be resolved, but four definitions of success is effectively no definition of success. A range of information can be collected and is helpful, but success resolving a need is most powerfully defined as the need being met. The range of outcomes looks something like this:
- Failure: (a) Patient hit a roadblock and was unable to resolve his or her need, (b) Patient did not meet eligibility requirements (e.g., income too high), or (c) No resources exist to meet patient’s need.
- Wait list: Patient placed on a waiting list for two months or longer.
- Equipped: Patient’s food need not yet met, but the patient is actively working on solving the need and feels equipped to proceed without further assistance.
- Successful: Patient confirms successful resolution of need (food is secured).
There are numerous risks associated with both the failure to clarify which definition of success is being used and picking a definition of success that does not, at a minimum, establish whether or not a patient secured the needed resource. The risks include:
- Picking a definition, or definitions, only weakly associated with desirable outcomes such as improved health and reduced cost;
- Not learning which aspects or types of interventions work and which do not (e.g., more or less intensive follow up);
- Not understanding which interventions have sustained impact over time (e.g., which interventions achieve long term food security);
- Referring patients to the same non-viable resource multiple times; and,
- Measuring health outcomes or cost savings without understanding if each element of the intervention was successfully implemented. (Did the patient get healthy food? Hence, it is reasonable to expect diabetes control will improve and costs will be reduced?)
Each one of these issues is a threat to both internal and external validity — for the purpose of learning and quality improvement, it is necessary to record whether successful resolution was due to the intervention provided by the health care system or to unrelated factors. Unfortunately, no agreed upon framework currently exists, which only increases the risk of unintended consequences of screening social needs and confusion around which interventions work in different settings and patient populations.
Defining Success for Accountable Health Communities
The CMS AHC model is illustrative of this challenge. The goal of the AHC model is to “test whether systematically identifying and addressing the health-related social needs of community-dwelling beneficiaries, including those who are dually eligible [for both Medicare and Medicaid]…impacts total health care costs and inpatient and outpatient health care utilization.” As CMS officials recently noted, the Innovation Center models are extremely complex to evaluate; it is critical to determine how and why model participants succeed or fail, to provide an accurate sense of what will happen if models are expanded, and to inform how best to structure potential expansion.
CMS has committed to use information provided through the screening process, as well as retrospective methods such as follow-up surveys and calls, to attempt to determine if identified health-related social needs were resolved for participating beneficiaries. However, without clearly defining criteria in advance for what such resolution means and committing to securing that information in a reliable way, the risk of misattribution and spurious findings is extremely high. By operationally defining what resolution of a need means, CMS can increase the chances of generating more meaningful results that can be linked to cost and utilization, in addition to learning which implementation strategies are most successful at resolving beneficiaries’ needs.
Recommendations for Defining Success
To ensure the successful evaluation of social needs programs in the AHC model and beyond, we advance three key principles:
Define success from the patient’s perspective.
The Patient-Centered Outcome Research Institute provides valuable guidance through their Methodology Standards Framework. They recommend measuring outcomes important to patients and based on input directly elicited from patient informants and people representative of the population of interest. In the food insecurity scenarios above, we cannot know if the lighter-touch approach (just handing a patient an information sheet listing nearby food pantries) is less or more effective than the higher-touch approach (navigation and follow up) without confirming whether a hungry patient in fact secured food. Moreover, we know that many food resources in communities will change in a given year (location, phone number, services, funding), and without knowing if patients actually received food we cannot know which community resources are viable.
Tailor success by social need domain
Social need interventions often cover a wide variety of social needs/resource domains. For example, the AHC model includes core needs ranging from food insecurity to transportation to interpersonal violence. With such different content areas being addressed, it is not possible to create one definition of success that is relevant for every domain. Furthermore, it is important that definitions align with those that are used and understood by social service providers assisting patients with needs within that domain.
A unique but standard definition of success for each domain will make it clear which activities must be completed and which specific data must be collected to document successful results within that domain. For example, a measure of success for food insecurity may be “75 percent of patients secured sufficient healthy food,” while transportation may require “patient reduced no-shows for appointments by 50 percent.”
However, while it is necessary to have clear measures of success for each social need, there needs to be some synthesis across needs for each patient, in order for the whole patient to be the final frame of reference. For example, if a patient or family screens positive for food insecurity and a fuel need in the winter, success cannot be defined solely by effectively addressing the food need without addressing the fuel need.
Define a range of success
As the scenarios above demonstrate, a range of responses to social need interventions may be meaningful for the patient. Defining levels of success to capture that range will enable more nuanced understanding and improvements of interventions over time, akin to how classifications of response to treatment for solid tumors or depression have supported and accelerated research in those fields. For example, understanding what it means to respond, partially respond, or not respond to treatment of depressive disorders requires definitions of success/progress across both biological and behavioral outcomes.
The range of outcomes described above for patients’ needs (failure, wait listed, equipped, and successful) is analogous to this clinical response orientation. In the food need example above a patient may successfully secure food (response), report having sufficient information to secure food independently without further follow-up (partial response), or refuse to discuss the need for help securing food (non-response). And even when patients do secure food, we can break that down further into whether it was an emergency food assistance or a longer-term solution such as securing SNAP benefits.
Before health systems across the country begin systematically screening and navigating millions of Americans toward securing basic resources necessary for improved health, we need operational definitions of what it means to successfully address such social needs. By establishing clear, patient-centered definitions we improve our ability to understand whether outcomes relate to a successful intervention and enable comparisons of interventions across health systems to support learning, improvement, and research.
The authors thank Rebecca Onie, Sara Standish, and Zach Goldstein for reviewing previous versions of this post.