The Importance of Social Determinants of Health in Determining the Right Clinical Interventions White Paper

The environment a person lives in, their socioeconomic status, lifestyle, wealth and cultural/ethnic background all affect that person's health status in far more significant ways than was previously understood. Across many studies, authors consistently agree that social determinants of health (SDOH) affect up to 80% of an individual’s health outcomes.

To illustrate how social factors may be integrated with a patient's standard clinical care, consider the key step of choosing the right clinical interventions. Clinician knowledge of the patient’s SDOH factors can allow for earlier, more targeted interventions when a successful outcome is more likely. If a patient has one or more chronic conditions combined with some social problems, say unstable housing and poverty, the chances of them needing a prolonged expensive hospital stay are significantly increased. Early intervention guided by an understanding of all the relevant factors impacting our patients’ health—both clinical and non-clinical—is therefore essential.

Understanding Social Determinants of Health Can Help Predict Worsening Outcomes

It makes intuitive sense that an understanding of social factors gives valuable insight into possible negative health outcomes for patients in many domains, including, for instance, high-risk pregnancies, early hospital readmissions, low adherence to prescribed medication regimens and susceptibility to chronic disease, to name just a few. Further it makes intuitive sense that appropriate interventions applied as soon as the situation is recognized could be very effective in helping SDOH-related issues and in turn ameliorating the patient’s clinical condition. Physicians have learned to better support their patients faced with social challenges by asking about the social history, offering advice and referrals to local support services, facilitating access to these services and acting as a reliable resource-person throughout the process.

More recently, interest in care coordination has highlighted the importance of SDOH factors as an organizing concept in the coordination of patient care. For instance, patients with limited resources may well struggle to get to specialist appointments on the far side of town. Patients without the usual supports at home, especially caregivers, are more likely to require a stay in hospital before their diagnosis can be fully addressed.

The “Hotspotting” Study

A recent study challenged the current wisdom that patients with complex needs benefit from commonly used clinical interventions aimed at improving care coordination. The study found that patients with complex medical and social needs, who were given intensive case management, did not have lower hospital readmission rates than a control group that did not receive comparable interventions. The study has been used as a reason not to increase a health system’s focus on care coordination for high-need patients. That would almost certainly be the wrong lesson to take from the study.

Patients had multiple challenges across several dimensions including mental health, substance abuse, chronic disease, poor socioeconomic status and for many, homelessness. The control group received standard care after being discharged from the hospital. The other half of the study participants received intensive case management, including coordinated follow-up care after they were discharged from the hospital. The intervention consisted of some home visits and phone calls focused on care coordination. On reflection, it's unlikely that the intervention was capable of positively impacting patients with this level of need in any meaningful way.

Such patients typically need the services of a broad multidisciplinary care team with some team members devoted to social services, perhaps legal advocacy and so much more. A focus on understanding and addressing mental health issues is frequently also highly relevant to a successful outcome. In other words, the clinical intervention needs to be capable of delivering the change being sought in a complex situation where predicting efficacy may be very difficult. Clinicians, care managers and population-level decision makers need current data relating to the outcomes of similar interventions on similar patients in similar circumstances to determine whether a proposed socially oriented intervention will be likely to have the desired impact.

Predicting and Reducing Hospital Readmissions

The challenge in reducing hospital readmissions is to recognize and address the reality that each patient has unique needs across many dimensions including the critical clinical, social and behavioral dimensions. How can an organization or individual clinician best ensure their interventions are organized so they deliver care to the right patient, at the right place and time, at the right cost?

Many patients recently discharged from the hospital can manage perfectly well at home without frequent calls from a care coordinator. It makes little sense to invest significant care coordinator time and effort on this group. At the other end of the patient-need scale, regarding the patients with multiple issues such as the subjects of the Hotspotting research, it would be helpful to know exactly what clinical interventions deliver the most benefit, so we can focus our efforts where they are likely to be rewarded. This is in clear contradistinction to the approach taken in the Hotspotting paper which was largely a standardized “one size fits all.” It should be apparent at this point that complex patients are likely to require more complex interventions and choosing the right intervention for each patient is complicated.

Technology alone cannot address the multiple social, justice and health inequities very likely to be impacting many of the patients in the trial. However, it can be used to help with good effect in a variety of ways. For instance, one of the authors pointed out the prevalence of unnecessary testing, inappropriate treatment and incorrect diagnoses in this patient population. There are many downsides to unnecessary or repeated testing and inappropriate treatment. They include the possibility of deterring patients from engaging further with the health system, a random abnormal result being taken out of context can easily lead to yet more inappropriate testing or interventions, some of which may be harmful. If this wastage were to be reduced, the time and clinical resources saved could be redirected to patients who are more likely to benefit from medical and social care interventions.

Readily available health information and technology, notably health information exchanges (HIEs), can help address many of these issues. For instance, unnecessary repeat testing can be addressed by highlighting recent test results independent of where those tests were performed or who ordered them. HIEs supply a huge range of data that can help address the risk of incorrect diagnoses or a lack of awareness of important complicating comorbidities. They are a logical place to capture store and present data from multiple sources across the region, giving clinicians a holistic understanding of their patient.

How Artificial Intelligence and Machine Learning Can Help Reduce Inequities

More recently, advances in the capabilities of HIEs including the incorporation of AI and machine learning tools allow for a better understanding of a complex patient's issues and assisting in choosing the right clinical interventions customized for each patient.

AI and machine learning tools can alert clinicians when a patient’s social situation is potentially contributing to their clinical state by highlighting the relevant, often unique key factors in play for each patient. They can also recommend appropriate targeted interventions such as choosing specific social services based on knowledge of the outcomes of other patients previously given the same intervention.

Given the complexity involved in choosing a clinical intervention for a complex patient, it’s important the AI and machine learning tools do not introduce unanticipated bias. A careful ethics review, along with in depth assessment by a multidisciplinary, representative team including clinicians’ managers and technical experts is needed to ensure the technology delivers accurate, unbiased results. Once assured that the tool is neutral, especially with respect to minority rights and social status, clinicians and managers can be confident it is guiding care in an equitable and just way with equal respect for all patients, yet targeted toward those who stand to gain the most from specific interventions.

Conclusion

Numerous studies have demonstrated benefits to identifying and proactively offering care coordination interventions to patients with complex chronic conditions. The Hotspotting study was unsuccessful in achieving significant improvements as a result of limited care coordination activities, such as follow-up phone calls and home visits. Nevertheless, it serves as an excellent starting point from which to base our recommendations for improvements. For instance, the study highlights that clinical interventions need to be carefully tailored to each individual patient’s unique needs, and that today, clinicians are working in a vacuum when it comes to knowing the true efficacy of interventions such as care management or social services for each unique patient.

AI and machine learning can help clinicians know and understand the characteristics of patients who may benefit from intensive case management. In the case of complex patients with complex needs, clinicians are confronted with a range of possible interventions, each with uncertain efficacy. In this situation the use of AI and machine learning offers a tangible approach to supporting clinical decision making and ultimately achieving better outcomes for all patients. While unanticipated bias is a known potential issue for AI and machine learning, it can be addressed with careful review and attention to detail. Unbiased algorithms that include social data as well as clinical data can make a significant benefit to individual patient care and the equitable distribution of clinical interventions across a population.

The views and opinions expressed in this content or by commenters are those of the author and do not necessarily reflect the official policy or position of HIMSS or its affiliates.