Preliminary Evidence of the Impact of Hospice Payment Reform on Social Service Visits in the Last Week of Life

Introduction

The Tax Equity and Fiscal Responsibility Act of 1982 created the Medicare Hospice Benefit (MHB), allowing the Medicare program to reimburse previously volunteer-driven hospice organizations for a comprehensive interdisciplinary set of services to address the needs of terminally ill patients and their family members. ¹ The payment structure instituted by the MHB reimburses hospices on a per diem basis based on one of four levels of care, and thus does not necessarily reflect the resource intensity required to care for terminally ill beneficiaries.

The supply of hospice organizations and utilization of hospice services have grown perennially, with accelerated growth over the past decade. In 2017, there were 4488 hospice organizations providing care to ∼1.5 million hospice users, with payment to hospice organizations approaching $18 billion. ² There is wide variation in hospice utilization, with much attention to the significant proportion of hospice stays that are deemed too short (<7 days) or too long (>180 days).^3,4

For years this variation has caught the attention of policy makers with specific concerns about this “two tail” issue of hospice utilization. Consequently, a provision in the Affordable Care Act authorized consideration of revisions to hospice payment policy that better align payment with the cost of providing services. The August 6, 2015 final rule instituted a change to hospice reimbursement that provided higher payment to the routine home care rate for days 1–60 and a reduced payment for days 61 on. ¹ In addition, hospices would receive a service intensity add-on (SIA) payment for nursing and social work visits in the last week of life.

This article offers an initial examination of the impact of the SIA policy on social work visits in the last week of life. Recently, the Medicare Payment Advisory Commission offered a brief analysis of social work visit time before and after SIA, finding limited effect. ² This article extends this analysis to examine the difference in response to this policy based on organizational characteristics, in recognition of the evidence on how organizational implementation challenges and policy readiness may affect policy awareness and uptake.

Methods

Our analysis is based on the 2015–2016 Medicare Public Use Files (PUFs) and 2016 Provider of Service (POS) file for hospices. The PUFs include information at the hospice level on utilization, including social service visit hours provided in the last week of life to hospice recipients. The POS file includes structural characteristics of hospices receiving payment from Medicare, such as ownership status and facility type. We used the Centers for Medicare and Medicaid Services (CMS) certification numbers to merge across files and identify hospices that participated in Medicare in both 2015 and 2016, before and after SIA was introduced.

To capture policy awareness across hospices, we identified hospices approved to participate in the Medicare Care Choices Model (MCCM) using the first year evaluation report. ⁵ MCCM is a model in the Center for Medicare and Medicaid Innovation portfolio that tests the clinical and financial impacts of allowing hospices to provide supportive services outside of the MHB; it was announced in 2014 and initiated in 2016, and hospices had to apply to participate in this model. Thus, we viewed initial selection for the MCCM as a proxy of hospice policy responsiveness.

We descriptively examined the impact of payment reform on social service hours in the last week of life. We looked at the overall impact of SIA on this outcome using pre-post t-tests, and then examined variation by repeating the pre-post t-tests for different subcohorts of hospices to assess how the impact was distributed across different types of organizations. Hospices were classified as to whether, in 2016, they were for-profit (as opposed to nonprofit or government owned), freestanding (as opposed to part of a hospital, nursing facility, skilled nursing facility, or home health agency), size (by quartiles of number of Medicare beneficiaries served), and selected to participate in MCCM.

Results

Table 1 summarizes the level of social service visit hours provided in the last week of life by year. There is substantial variation in this measure across hospices, ranging from 0 to 1.07 hours per day per hospice decedent across 2015–2016. The median is smaller than the mean, suggesting a preponderance of hospices at the lower end of the distribution.

Table 2 depicts the social service visit hours over time, overall and by organizational characteristics. From 2015 (pre-SIA) to 2016 (post-SIA), we see an increase of ∼0.009 hours of social service visit hours per day per hospice decedent in the last week of life, an ∼10% increase. Examining the pre-post differential by hospice size, as captured by the number of Medicare beneficiaries served, unpacks this overall change further. On average, larger hospices tended to provide a higher baseline level of social service hours to decedents in the last week of life. The upper three quartiles of organizations, in terms of number of Medicare beneficiaries served, all showed statistically significant and similar increases in the last week of life social service hours from 2015 to 2016, whereas the smallest hospices showed a substantially smaller increase in this outcome that was not statistically significant. For-profit and not for-profit hospices showed similar increases in this outcome from 2015 to 2016; however, not for-profit hospices had a higher starting level of social service visit hours in the last week of life by ∼18%. In terms of facility type, freestanding hospices had an increase in social services visit hours in the last week almost twice as great as those of nonfreestanding facilities; these freestanding facilities also had a lower starting level of visit hours, by ∼10%. Finally, MCCM participants recorded ∼25% more visit hours in 2015. In this case, the higher baseline also corresponded with a greater increase over time, with the increase in 2016 ∼ 50% larger in magnitude for MCCM participants than for nonparticipants.

Discussion

Although generally supporting the previously identified finding of a modest increase in social service visits in the last week of life before and after the implementation of the SIA policy, this article identifies substantial variation in both the baseline level of social service visits during this timeframe and how different kinds of hospices responded.

This variation may be attributable to a number of factors, including differences in the organizational capacity of hospices to respond to this payment change. For example, the “cascading” effect observed by hospice size in both the 2015 baseline and the 2015–2016 differences shows that larger hospices tended to provide more social service visit hours in the last week of life before the SIA, and they also showed a larger increase in service visit hours postreform. It may be that smaller hospices have a more limited ability to expand. These differences are notable in that they suggest a divergence in social work exposure for hospice beneficiaries, since, based on size, hospices already offering a higher number of end-of-life social work hours appeared to increase these hours, therefore, widening the gap in offerings between the smallest hospices and others.

Payment reforms are premised on a belief that provider organizations will respond to changes in financial incentives. The operational challenges for hospices to be agile adopters of policy changes should be considered, especially heterogeneity in responsiveness. We added a subanalysis based on MCCM selection to proxy for policy receptiveness. There were different baseline levels of social service investments in the last week of life between MCCM participants and nonparticipants, and a resulting change in levels from 2015 to 2016 that suggest a further widening of the gap. This underscores the importance of examining the distribution of impact to identify whether change is being realized uniformly or is associated with certain types of organizations.

Our findings are preliminary and descriptive, not causal, examining trends before and after SIA took effect. Thus, we cannot rule out that other factors in flux during this time may be impacting this outcome of interest. Although we took a structural approach to examining variation in policy impact, focusing on organizational characteristics, further study could explore whether there are process or outcome differences among these organizations that affect their ability to respond to SIA. For example, if case mix is meaningfully related to hospice size, then this may affect visits in the last week of life, since some conditions are more difficult to prognosticate than others; such a difference may further explain why some organizational types appear to be more responsive to SIA than others.

Conclusion

The SIA payment reform was designed to better reflect the resource intensity required to care for terminally ill beneficiaries and to further incentivize the presence of hospice team members at the end of life. Our finding suggesting modest across-the-board improvements in social work visits in the last week of life is juxtaposed with evidence of significant heterogeneity based on organizational characteristics. The promise of payment reform is that altered financial incentives will drive organizational change that will, in turn, translate into better outcomes across the board; cross-cutting barriers to that translational exercise should be explored, with an emphasis on factors that may cause differential responses that could potentially widen access and outcome inequalities.