Thursday, November 13, 2014

How Does Hospice Enrollment Impact Overall Health Care Utilization?

Ziad Obermeyer et. al published a large, rigorous study in JAMA this month (Open Access!) that provides cost and aggressiveness data to support the use of hospice for poor-prognosis cancer patients who stop cancer related treatments. The bottom line of the study is:

The choice to enroll in hospice definitely decreases cost and aggressiveness of care in patients with poor-prognosis cancer who are no longer receiving disease directed therapies.

This finding is consistent with Kelley et al’s 2013 study of hospice length of stay showing that hospice enrollment, for all diagnoses, saved Medicare dollars during all length of stay (LOS) periods measured, with the greatest benefit in patients enrolled 8-30 days($5,040-6,430). All LOS groups also had associated decrease in hospitalizations, ICU days, hospital deaths and hospital readmissions.

The Obermeyer study significantly confirms that this benefit also exists for hospice utilization in advanced cancer. So let’s break it down so that we can be sure to use/apply it appropriately:

Association Between the Medicare Hospice Benefit and Health Care Utilization and Costs for Patients With Poor-Prognosis Cancer

Background: People with cancer historically have been the largest recipients of the Medicare Hospice Benefit (per NHPCO, 37% of all beneficiaries in 2013) but their LOS is shorter than the general population, having the highest rates of hospice stay less than 3 days. Short hospice LOS and increased intensity of care outside of hospice spurred this research group to explore health care utilization and costs in the poor-prognosis cancer population.

Population:
  • Medicare fee-for-service beneficiaries with poor-prognosis malignancies who died in 2011 after a full year of Medicare coverage were included.
  • ICD codes for pancreatic, lung, brain, metastatic or ill-defined malignancies, or hematologic malignancies that are relapsed or not in remission.
  • Large study: 36,330 patients in the final matched cohort from an initial sample 86,851 patients.
Matching: (see the study for the complex details- this is significantly simplified)
  • Poor-prognosis cancer patients who enrolled with hospice were matched with those who did not
  • Coarsened exact matching variables: Place of residence, age, sex, time from first poor-prognosis cancer diagnosis to death
  • Compared utilization before and after hospice to try to determine what may have happened if the patient had chosen hospice.
  • Matching did not include: Objective performance status, caregiver status, availability of hospice or other palliative care services
  • Pairs were not included if either member of the pair utilized cancer directed treatment (defined as chemotherapy or cancer related surgery) during the exposure period (so, only people who stopped all cancer directed treatments were included in the study!)
Outcome Measures:

Primary: Health care utilization during hospice enrollment or the equivalent period of time in non-hospice patients, including hospitalization and ICU frequency, number of inpatient procedures and deaths in hospital or skilled nursing facility. Did not include death in assisted living or inpatient hospice.

Secondary: Total cost, including beneficiary payments, Medicare payments and third party payments for inpatient care, hospice costs, physician costs, non-institutional provider payments. Did not include medications, personal care or non-Medicare paid services.

Results:
  • Median time until death after diagnosis of poor-prognosis cancer was 13 months.
  • 60% of patients received hospice care.
  • Median hospice LOS 11 days, less than 6% had a LOS of more than 6 months.
  • Patients who chose hospice
    • Spent less time in the hospital
    • Spent less time in the ICU
    • Had less inpatient procedures billed.
  • Death in a facility (hospital or skilled nursing facility) decreased: 14% versus 74%
  • Cost of care in the last week of life decreased: $8,421 versus $3,892, - more than 50% difference
  • Costs were decreased in all hospice LOS periods except for the 2% of beneficiaries enrolled more than 1 year
  • The greatest cost savings: LOS 5-8 weeks which decreased cumulative costs by 24%
  • Cumulative cost for the last year of life* decreased: $71,517 versus $62,819, a 12% difference
    • *This is difficult to interpret since patients had varied hospice lengths of stay, with many patients enrolled in hospice only 3 days prior to death, this number hardly seems reflective of a hospice effect. And some of them were not even diagnosed with their cancer for the full year, resulting in spuriously lower costs.
Author's Conclusion:

In this sample of Medicare fee-for-service beneficiaries with poor-prognosis cancer, those receiving hospice care, compared with matched control patients not receiving hospice care, had significantly lower rates of hospitalization, intensive care unit admission and invasive procedures at the end of life, along with significantly lower health care expenditures during the last year of life.

Concerns/Thoughts:
Factors that influence patients’ decision to enroll in hospice services were not addressed. They may include:
  • Caregiver availability - Prior to the exposure period, hospice patients in the sample used more home health and non-hospice patients used more skilled nursing facilities. I would argue the increased use of skilled nursing may reflect inadequate resources for in home care. Patients in this study, like many patients I’ve care for, may have been faced with the decision to enroll in hospice OR get their skilled nursing home stay paid through Medicare. If care giving is an issue, they may have chosen ECF instead of hospice. This increases costs and moves site of death to the ECF. If there were better payment options for in-home custodial care, patients may choose to stay at home, with hospice, instead.
  • Hospice/palliative care availability - we don’t know why patients did not choose hospice. They were all clearly eligible. Since available services/facilities were not described, we do not know if the availability of hospice & palliative care influenced the propensity to enroll in hospice. This may be particularly relevant to patients living in a rural community where there is greater variability in services despite the study attempt to match place of residence.
  • Data from personal care or inpatient hospice facilities was not available - Therefore this data may miss significant caregiving costs and frequency of deaths in facilities other than hospital or ECF.
Other interesting findings:

Patients who were selected for the study, who did not received cancer directed therapies during the exposure period, lived in zip codes with higher incomes and had a shorter median time from diagnosis to death. This could be further investigated- maybe these factors influence patients’ decisions to stop cancer directed treatments?

My Conclusion:

Care for people who chose hospice after stopping cancer treatments cost 12-50% less than care for people who didn’t and less of them died in facilities, despite median hospice LOS of less than 2 weeks.

Other Takeaway Points:
  • Although cancer comprises the largest group of hospice diagnoses (37%), people with poor-prognosis cancers have shorter hospice length of stays than overall population (11 days versus 19 days)
  • When sharing this study with others, don’t forget that this cost-avoidance occurs in the setting of care that is proven to increase patient and family satisfaction, decreased symptom distress and improved outcomes for caregivers. (Even the most curmudgeonly administrator likes that combination!)
  • These data cannot be applied to:
    • Young patients who have private insurance
    • Patients who are receiving any chemotherapy or cancer related surgeries, including those who receive oral targeted agents or who (presumably) have palliative surgeries for bowel obstruction, cord compression etc.
  • This study cannot definitively separate whether the hospice services, or the propensity to choose hospice services has the greater effect, so we don’t know for sure if hospice in a “Concurrent Care” model would have similar results.
  • Although consistent with other studies, these findings cannot be applied to non-cancer patients
Studies like this are very difficult to do and cannot control all variables, so kudos and thank you to all of those involved in this project for contributing such great data to the field!

Image Credit: US Postal Service
Image Credit: Unknown infographic from Pinterest

Pallimed | Blogger Template adapted from Mash2 by Bloggermint