I had the pleasure of attending HIMSS Asia-Pacific this year, where I talked about the importance of value in the delivery of care. Indeed, the continuously escalating healthcare cost added to a growing and aging population, resulting in a rise in chronic and multiple complex conditions to treat, have led to efforts to model and incentivize care based on value.
Value-based care programs follow a common pattern: identify, target and engage with patients and assess risk to prevent progression to costly illness. Should illnesses take place, then proactive care at lower settings of care with less costly interventions can slow disease progression and reduce the utilization of costly treatment or costly settings of care, such as the emergency room or hospital.
These programs are designed for populations as a whole, based on population needs, opportunities and desired outcomes. But the reality is that a typical population will be a mix of individuals with varying levels of clinical status, severity and needs. The inherent heterogeneity of any population, no matter how specifically defined the cohort is, means that the program design and interventions have issues with relevance and specificity: care will be neither evenly provided nor evenly needed across all individuals within that population.
A value-based program expects to do more good for more individuals in the group and to produce a net cost benefit on average, even after spreading cost and benefit across the entire population. This is fine for financial modelling and actuarial purposes; but what about for the purpose of providing care and a human-centred care experience?
As advancements are made with the approaches of precision medicine, can the approach to population management and value-based care model be refined with precision medicine capabilities?
Precision medicine seeks to use as many individual attributes and other relevant variables as possible to apply artificial intelligence (AI) and machine learning (ML) to personalize treatments at the individual level rather than for everyone with the same condition.
This goes against the basic model of a population management and public health model, where the expectation is to apply an intervention to all individuals meeting the population criteria. The interventions are not discriminating, and the desire is to achieve a high rate of the intervention: those who would benefit from it would be treated; those who wouldn’t benefit from the intervention would just receive unnecessary but not harmful care. The net result would ideally be value for the population.
With the availability of longitudinal clinical history in digital format, plus the computing power of AI and ML, the opportunity now is to target at-risk populations with precise interventions. This approach does produce a challenge with the reimbursement model and the financial incentives for caregivers: now the goal isn’t to apply the desired intervention on as many as possible; instead, the opportunity is to apply less interventions but to target interventions more precisely.
The focus shifts towards outcomes as the only way to know if the care pathway is actually suited for the individual: if done right, the precision-based intervention improves the right individual’s outcome without missing anyone, while also not causing unfavorable outcomes for those identified as not needing or benefiting from that intervention. This means caregivers will need the analytics insights from AI and ML to identify the most relevant at-risk individuals who will actually benefit from the interventions.
The need for analytics insights – not as retrospective static reports but as an integral part of patient care – is paramount to precision medicine. It is more than just the medical history or quantitative structured data of lab measurements and other biometrics data; it is also the inferred and derived data that isn’t inputted by the patient or caregiver. This juxtaposition of traditional clinical data plus inferred or derived insights is crucial to precision medicine.
Longitudinal patient data, plus other available and useful associated data, including relevant social determinants of health, may be determined through AI and ML to be predictive of risk and responsiveness to intervention towards the desired outcomes. This person-specific precision insight can thus guide clinician decision-making through enhancing clinical decision support based on broadly defined clinical guidelines and pathways.
Now, value is directly attributable and precise; we move beyond hoping for overall value based on an average applied across a population, and the public health approach of recommended best practice fades into the background as a fall-back approach.
Value must be redefined and re-scaled across a set of smaller but more similar cohorts, dynamically evolving treatment recommendations as patient needs evolve: provide care where it is needed, account for and optimize the care of some patients requiring costly intervention due to clinical reasons beyond anyone’s control, and seek to have more people healthy or achieving the desired outcomes than not.
This valued-based model can only be done by making precision medicine an integral part of the approach to population health management.