This study examined practice patterns, before and after implementation of the ESRD PPS, related to HD catheter thrombosis management in a series of large national cohorts of Medicare beneficiaries receiving maintenance HD. We found that implementation of the PPS was associated with several shifts in care, including a decline in within-HD-unit thrombolytic claims and an increase in thrombus/fibrin sheath removal claims. Although we did not find evidence of an association of the PPS implementation with a shift in outside-HD-unit thrombolytic claims, claims did appear to increase starting in 2013. We also found that the rate of delayed HD sessions increased over the entire study period; implementation of the PPS was associated with a reduction, but not a reversal, of this increasing trend. Finally, among patients receiving thrombolytic therapy, drug administration outside, vs. within, the HD unit was associated with increased risk of delayed HD treatment in both the pre-PPS and post-PPS periods.
We observed a decline, by approximately 55%, in the rate of thrombolytic use in the dialysis unit, and the coincident 56% increase in the rate of thrombus/fibrin sheath removal procedures (i.e., those relying on a skilled interventional specialist—nephrologist, radiologist, or surgeon), comparing the post-PPS and the pre-PPS periods. A possible explanation for these findings is that, under the PPS, a payment model that eliminated separate reimbursement for injectable medications, some cases of catheter thrombosis, which previously may have been resolved directly within the HD unit by administering thrombolytic therapy, were more likely to be shifted to another setting (e.g., a vascular access center or other radiology suite) to perform a more intensive procedure. Other studies have reported similar post-PPS shifts in care in maintenance HD patients. For example, with respect to anemia management, a decline in use of erythropoiesis-stimulating agents (ESAs) coincided with an increase in red blood cell transfusions after PPS implementation (and an updated ESA drug label) in 2011 [2, 3, 13].
Furthermore, many of the additional transfusions required care in more intensive settings (i.e., inpatient hospital and emergency department) [14]. Also, it is worth noting that thrombolytic use in the dialysis unit began declining in the quarter prior to PPS implementation (19% decrease between Q3 and Q4 of 2010). Since CMS published the final rule of PPS regulations in August 2010 [1], the decline likely represents an anticipatory reaction, similar to the decline in use of ESAs in the months prior to PPS implementation [3], and the decline in the rate of hospital readmissions in the 2 years preceding the start of the Hospital Readmissions Reduction Program in October 2012 [15].
Along with the PPS-associated decrease in within-HD-unit thrombolytic use and increase in thrombus/fibrin sheath removal procedures, we also expected to observe clear PPS-associated shifts toward increased use of thrombolytics outside the HD unit. While thrombolytic use outside the HD unit appeared to increase by nearly 40% from 2012 to 2013, we found no clear evidence that this was associated with the implementation of the PPS given that the increase began to occur several years after PPS implementation. One possibility is that PPS implementation and the apparent increase in non-HD-unit use of thrombolytics are entirely unrelated. An alternative explanation, which we favor, is that thrombolytic use within the HD unit has slowly become more disfavored over time, perhaps due to cost constraints or the reimbursement environment. Anecdotal experience suggests that many HD units order fewer monthly doses of thrombolytics than in years past, creating a less liberal environment for its use than before. Future studies examining this trend in present and future years may be in a position to offer more insights into this issue.
The findings regarding trends in delayed treatment were somewhat nuanced. Delayed HD sessions, perhaps surprisingly, continued to increase throughout the study period, although the PPS implementation was associated with a slowing of this increasing trend. We observed nearly a 50% increase in delayed HD, from 1.6 sessions per patient per quarter in Q2 of 2008 to 2.3 sessions per patient per quarter by Q3 of 2015, which is similar in magnitude to the 7.1 days of missed treatment per patient per year reported from a large cohort of HD patients from a national dialysis chain in 2005–2009 [16]. We find this potential increasing trend of delayed HD (which has also been reported in other US HD cohorts [17]) concerning, particularly since missed HD treatment is associated with greater risk of numerous adverse outcomes, including hospitalization and death [16, 18]. Future investigations should determine whether implementation of newer care models that incentivize coordinated ESRD care, such as the 2015 CMS Comprehensive ESRD Care Model and ESRD seamless care organizations, has led to reductions in delayed HD.
In secondary analyses, we found evidence that thrombolytic administration outside, as opposed to within, the HD unit was associated with greater risk of delayed HD in the pre- and post-PPS periods. Our rationale for this comparison was that referral outside of the HD unit for thrombolytic therapy would more likely cause a disruption in the HD schedule. However, disentangling the direction of this association using observational data is somewhat complicated, as patients with access thrombosis may experience delayed treatment either before or after receiving thrombolytics. For example, clinical scenarios whereby thrombosis may lead to delayed HD treatment include: (i) thrombolytics are not immediately available, (ii) thrombolytics are available, but the time required for instillation delays treatment until the following day, or (iii) thrombolytics are available, but instillation is unsuccessful and a more invasive declotting procedure (e.g., thrombectomy or catheter replacement) is required. The first scenario would be expected to cause treatment delays prior to administering thrombolytics, while the latter two would cause delays after. We found that the location of administration was more strongly associated with delay in the week before thrombolytic administration than in the week after. Thus, we suspect the former association may be related primarily to immediate availability (i.e., having thrombolytics on hand in the HD unit may decrease risk for treatment delays). Nevertheless, we cannot rule out the possibility that this association is confounded by occlusion history, whereby patients with prior access complications are more likely to be referred outside of the HD unit to treat the thrombosis.
This study has several limitations. First, it can be challenging to assess the impact of a policy change using observational data. Our findings may be biased due to other unaccounted changes, such as with respect to patient characteristics or treatment practices, between the pre- to post-PPS periods. We generally observed similar patient characteristics across the quarterly cohorts, and standardized or adjusted for these covariates in our models, but residual confounding is still a possibility. Second, start and end dates for patient follow-up were defined, in part, on procedure codes for insertion, replacement, and removal of tunneled central venous catheters, but not using codes for non-tunneled catheters, which we could not be sure were specific to the vascular access for HD. Non-tunneled (temporary) catheters are used relatively infrequently for maintenance HD, at least in the US, so we expect this should have a minimal impact on our findings. Third, given our use of administrative claims data, we were limited in our ability to fully characterize catheter thrombosis, both with regard to its exact date of occurrence and severity. For this reason, we were able to describe only rates of thrombosis-related care in a broad sense, but not rates of thrombosis itself. Finally, our findings are limited to the ESRD population with Medicare Part A and B (fee-for-service) coverage.