A colleague who runs a community mental health agency recently told me that the worst outcomes in her organization are seen in the patients who receive only medication management services from prescribing providers but who are not plugged into the organization’s broader counseling and other services. These are the patients who end up repeatedly destabilized and in the emergency department or those who attempt, or worse, succeed in committing suicide.
When they look back into the clinical notes as part of their morbidity and mortality reviews, the team sees that these patients have told their providers about rising social stresses—impending eviction, interpersonal violence, struggles to find a good job without a GED. Unfortunately, it is not within the narrow scope of the medication management practice to address these issues, other than through some brief counseling. At present there is also no means of extracting this information prospectively from the clinical notes to ensure these patients get outreach and encouragement to take advantage of the organization’s broader services.
These scenarios evoke the capacity of Augintel’s child welfare software, which looks cumulatively and over time at risk and protective factors discussed in the social workers’ notes. Using a variety of tools, the software helps both individual case managers and their supervisors know which families need more attention and which may be ready to move on from close supervision or towards family reunification. These same tools can be equally helpful in clinical settings. Augintel has done this with a behavioral health organization focused on moving patients out of their suicide prevention program to open up enrollment for others, but the potential applications are much broader.
Clearly capturing rising social stressors is critical for those at risk of suicide due to underlying mood or psychotic disorders as demonstrated by the post-hoc case reviews conducted by my colleague’s organization. However, it would likely be equally helpful for those with chronic illness of any type, particularly complex or co-morbid illnesses, where the patient’s ability to adhere to a treatment plan is constantly under threat by social circumstances.
During my work in one of New Jersey’s largest urban areas, a common challenge we found for keeping patients with complex medical illness well and out of the hospital was the inability of persons who became unhoused to take their injectable medications into a shelter with them. We found similar challenges with temperature sensitive medicines being prescribed to patients who were living in their cars. With no sign of the housing crisis easing, more and more patients are just one lower-than-expected paycheck or fight with a friend or family member away from this reality. So addressing the medication issue linked to housing is critical. While housing (like many other social issues) is a tough problem to solve, we do have the opportunity to listen to our patients and follow their journeys even when we can’t fully solve their problems. Adaption of the medical plan to fit a patient’s reality is a critical, if imperfect, step in improving health disparities. Highlighting a patient’s increasing risk of housing instability by systematically pulling this information from the clinical notes and presenting it back to the clinical team as an indicator of rising risk, can trigger both short- and long-term interventions. While team members make connections to housing (or job or counseling) services to eventually enable the person to obtain stable housing, in the near term the clinician can adapt and prescribe alternative medications that can be brought into a shelter or be safely stored in a locked glove compartment. While long term stable housing is typically the goal, the pay-off for clinical care plan adjustments is rapid and quantifiable in both biometrics and healthcare utilization.
Not only can NLP based software find and represent rising risk, but as with the suicide prevention program mentioned above, it can help to redirect scarce resources to those who most need them, by demonstrating risk mitigation. Many aspects of a patient’s progress toward treatment goals are not easily captured as structured data. Unlike laboratory data, even when quantified using validated approaches, clinical outcomes such as decreased frequency of thoughts of self-harm or improvements in pain or other physical function are more commonly captured in narrative, rather than structured, data. This use of NLP creates the opportunity for progress tracking and new outcome measures demonstrating effectiveness, not just intermediate markers, which can form the basis of more meaningful value-based payment schemes.
This capacity for outcomes capture is critical for the social factors on which health systems are increasingly focused. For example, a patient who has had a hard time following their medication regimen may report that now they are taking their medications daily since getting to the pharmacy every month is no longer a problem now that the community health worker helped them get a subsidized bus pass or small grant to fix their car. They may also report that the farmer’s market vouchers help them to eat healthy throughout the month, even after the SNAP EBT card is empty.
Augintel’s technology, as applied to child welfare, has shown the ability to provide this information about both risk and protective factors—allowing the child welfare worker to step up, step down, or modify the plan for a child and their caregivers without having to rely on chart audits or capacious memory to recognize the trajectory. Great opportunity awaits in applying the same approach to both physical and behavior clinical care.
.png)
.jpg)
