The first Duke Margolis team that I started working with just had one of our big projects get published in the Journal of Pain and Symptom Management.** We examined a demonstration project in Western North Carolina that attempted to offer pre-hospice palliative care to individuals outside of the typical end of life spectrum. Our big finding is that control groups are awesome especially when the patient population is extremely morbid as propensity matching techniques don’t work well on a population that is inherently outliers. But that is not the point I want to expand on.
We need to talk about cost attribution and the need for clarity about savings as we stated in this paragraph:
Despite widely cited studies supporting the cost-saving potential of palliative care literature, the inability to consistently detect cost savings in the post period in this demonstration may be explained by both Medicare payment mechanisms and the inability to identify a credible comparison group.11, 12 Hospital internal cost information has demonstrated that some aggressive therapies and associated costs can be avoided via palliative care; however, for hospital care, Medicare often pays a prospective amount based on diagnosis. As a result, any savings in cost reduction accrue to the institution, not to Medicare. As alternative payment models proliferate, the studies intending to detect cost savings of palliative care will be less fruitful because reimbursement is increasingly disconnected from resource use in value based payment models.
Monetary savings can be seen from two perspectives; total payer (CMS) perspective for Medicare and clinician/provider perspective. When we operated on a pure fee for service model, these two things were mostly linked. Fewer services produced fewer billing opportunities which produced fewer total payer costs. A service provider would be profitable if they could offer an incremental service at below reimbursement. If the service provider had a great ability to negotiate a good deal on a saline bag, they made money by offering one more saline bag to the patient.
Hospital care for Medicare is mostly driven by the perspective payment system where CMS cuts a hospital a check that is determined by the diagnosis codes and a few other claims based attributes for that particular hospital. The hospital gets a big chunk of money to care care of a patient and if they can spend less money than the chunk, the hospital profits while the hospital loses money if they spend above the chunk of care. This is a long standing bundle. Hospice payments have always been per-diem rates with a 2016 change altering early care and the last week of life payment levels.
Studies can look at total costs and potential savings of early palliative care from multiple perspectives. If an individual enters the hospital for an inpatient stay, there is very little savings that CMS and broader society will accrue from palliative care. CMS saves money if palliative care prevents an inpatient admission but conditional on an admission, there is very little juice for CMS to squeeze. Conditional on an admission, the hospital can control costs and pocket the difference. A palliative care intervention that leads to the patient and/or their family deciding to not engage in last ditch/low hope/high cost interventions will save the hospital money without changing what CMS pays.
This is important as studies that claim large care savings from a palliative care intervention can come from either perspective. There is way more potential swingable costs from the provider perspective in any given episode than from the CMS perspective. Both perspectives are asking valid questions but they are asking and answering different questions.
As we move towards even more bundled payments, and population capitated payments and/or global budget models that disconnect the payment from service intensity, we just need to be even more aware of exactly what question and from which perspective cost savings or increases are being viewed from.
** https://doi.org/10.1016/j.jpainsymman.2019.06.019 Taylor et al. 2019