Impact of Pacemaker Implantation on 12-Month Resource Utilization Following TAVR Hospitalization

Abbreviations and Acronyms: DRG: Diagnosis Related Group; ICD: Implantation of VardioverterDefibrillator; ICD-9-CM: International Classification of Disease, 9th Edition Clinical Modifications; IPSAF: Inpatient Standard Analytical Files; LOS: Length of Stay; MBs: Medicare Beneficiaries; OPSAF: Outpatient Standard Analytical Files; PPM: Pacemaker Implantation; SAVR: Surgical Aortic Valve Replacement; TARV: Transcatheter Aortic Valve Replacement


Introduction
Transcatheter aortic valve replacement (TAVR) offers a less invasive approach than surgical aortic valve replacement (SAVR) to treat severe aortic stenosis in patients with intermediate or high surgical risk from conventional SAVR, or for whom SAVR is contraindicated [1][2][3][4][5]. With similar primary clinical outcomes for TAVR as SAVR in inoperable, high and intermediate risk patients, there has been a sharp rise in the number of TAVR procedures being performed [4,6,7], especially among Medicare Bene iciaries (MBs) [8].
TAVR's cost effectiveness and value in high risk patients has been evaluated [9][10][11][12]. Patients with a high risk pro ile undergoing an index TAVR procedure experience shorter inpatient lengths of stay (LOS) than SAVR, however, procedural costs including the acquisition cost for the transcatheter valve, have resulted in higher index hospital costs [12]. Periprocedural complications experienced during an index TAVR hospitalization have been described, and impact TAVR's effective cost [11].
is atrioventricular conduction disturbance that requires permanent pacemaker implantation (PPM) [13,14]. PPM following TAVR is associated with higher index hospitalization cost due to both procedural cost and increased LOS [11,14]. In addition; PPM post-TAVR has been reported to increase the risk of unexpected hospital re-admission [13,15,16].
From a hospital's inancial risk perspective, the U.S. Centers for Medicare and Medicaid Services has identi ied cardiac valve procedures for voluntary bundled payment beginning in October 2018 and extending to December 2023 [17]. Reimbursement under a shared risk arrangement merits an understanding of resources consumed by MBs during an episode of care that includes an index TAVR procedure and extends through post-discharge follow-up.
In light of this context, this study divides MBs into two study cohorts: those who received a PPM during their index TAVR hospitalization and those who did not. Resource utilization is reported for both groups using the metrics of hospital length of stay and Medicare reimbursement for three time periods: the MBs' index TAVR hospitalization, the 365day post-discharge follow-up period, and the total episode of care. Additionally, the most common reasons for hospital readmission are reported for all MBs, which may be of use to hospitals seeking to reduce costly readmissions. Of note, during the study period approval for MB TAVR procedures was limited to the "high risk" patient cohort.

Data source
Centers for Medicare and Medicaid Service's Inpatient and Outpatient Standard Analytical Files (IPSAF and OPSAF) linked data for calendar years 2014, 2015, and 2016 are the data sources for this retrospective analysis. These iles allow researchers to link all acute care and outpatient services utilized for an individual MB. The IPSAF data ile contained information to calculate length of stay in days, discharge status, total Medicare reimbursement, International Classi ication of Disease, 9 th or 10 th Edition Clinical Modi ication (ICD-9-CM or ICD-10-CM) diagnoses and procedures codes, and diagnosis related category (DRG). While the OPSAF data ile contains procedure codes to identify PPM or cardioverter-de ibrillator (ICD) implantation performed in the outpatient setting during the study period, but Medicare reimbursement for outpatient episodes was not collected.

Study population
The population in this study consists of MBs in the IPSAF who underwent TAVR in a US hospital between January 1, 2014 and June 30, 2015. MBs undergoing TAVR were identi ied using the following ICD-9-CM procedure code: 35.05. A total of 20,682 TAVR hospitalizations were identi ied as meeting the inclusion criteria. A MB's TAVR hospitalization was excluded from the study population for four reasons: 1) the MB's TAVRs hospitalization's discharge date was missing (n = 7); 2) the TAVR procedure identi ied was not the MBs irst TAVR hospitalization in the study period (n = 13); 3) the MB had a previous PPM or ICD (n = 2,844); or 4) the MB died during the index hospitalization or the follow-up period (n = 2,292). Because seven MBs were excluded for multiple reasons, the inal study sample consists of 15,533 who survived for 365 days following their irst TAVR procedure between January 1, 2014 and June 30, 2015. Two study cohorts were created based on whether or not the MBs received a permanent pacemaker implantation (PPM) identify by ICD-9-CM codes (37.80, 37.81. 37.82, 37.82, 37.83, 37.85, 37.86, 37.87, 00.50 or 00.53) during their index TAVR hospitalization.

Unit of analysis and analytical fi le
The unit of analysis is a MB. To construct the analytical ile, the hospital ile was searched to identify each MB's initial TAVR hospitalization, including all relevant utilization, reimbursement, and discharge destination information associated with the index TAVR. Next, the IPSAF and OPSAF hospital iles for 2014, 2015, and 2016 were searched for all encounters within 365 days of the discharge date of the index TAVR hospitalization. A MB's claims were converted into a bene iciary level ile by summing, averaging or counting the relevant data information obtained from all follow-up encounters. If a MB did not have any hospital readmissions or outpatient procedures all relevant study information was set equal to zero for that MB.

Statistical analysis
Univariate differences between MBs who did and did not receive a PPM during their index TAVR hospitalization were assessed using χ 2 analysis or the Fisher exact test when χ 2 analysis could not be performed due to expected counts less than ive. Observed resource utilization statistics were reported as mean ± SD, median, irst quartile, and third quartile values. Differences in resource utilization were tested using one-way ANOVA statistic with median score (number of points above the median). Differences between study groups were considered statistically different if the p-value was less than or equal to 0.001. Median regression models were run to estimate risk-adjusted differences in median resource usage between the two study groups after controlling for demographic characteristics and 47 comorbid conditions. All analyses were performed with SAS 9.4 (SAS Institute, Cary, North Carolina).

Demographic and comorbidity controls
All demographic and comorbid conditions were created based on information contained in the claim information associated with the index TAVR hospitalization. Demographic variables of interest included: age group (under 65, 65 to 69, 70 to 74, 75 to 79, and 80 plus), gender, and race (white, or non-white). All comorbid conditions were identi ied using ICD-CM-9 or ICD-CM-10 codes that were present on admission during the index hospitalization.

Results
Overall, MBs undergoing a TAVR during the study period were most likely to be older than 80 (69%), white (93%) and male (51%) ( Table 1). In addition, MBs during their index TAVR reported a variety of comorbid conditions, Table  1 reports on 19 different conditions experienced by more than 10% of MBs in the study populations. A comparison of demographic conditions between the two study cohorts indicated signi icant differences in the age distribution (MBs receiving a PPM were more likely to be over 80 years of age (72.95% versus 68.66%) and male (54.93% vs 50.50%).
The average length of stay (LOS) for all MBs during their index TAVR hospitalization was 6.4±5.7 days while the median length of stay was 5.0 days (interquartile range for LOS was 3.0 to 7.0 days) ( Table 2). Medicare reimbursed hospitals an average of $50,822 ± $19,834 for the index hospitalization and the median Medicare reimbursement was $48,530 (interquartile range for reimbursement was $39,574 to $59,307). The vast majority of MBs were discharged from their index TAVR hospitalization to one of three destinations: home (39.9%), home with a home health agency (32.5%), or skilled nursing facility (20.4%).
A comparison of the two study cohorts indicated that MBs in the PPM cohort consumed more resources than those MBs not receiving a PPM during their index TAVR hospitalization. The observed differences were 7.9 vs 6.1 days for index LOS and $55,597 vs $49,996 for Medicare reimbursement, respectively, while the median values were 6.0 vs 4.0 days LOS and $55,597 vs $49,996 for Medicare reimbursement. The one-way ANOVA test indicates that signi icantly more observations were above the median value for MBs in the PPM cohort for both resource measures. Finally, MBs in the PPM cohort were signi icantly less likely to be discharged home (29.4% vs 41.7%), and signi icantly more likely to be discharged to a home health agency (35.5% vs 32.0%), a skill nursing facility (25.9% vs 19.5%) or a rehabilitation facility (6.9% vs 4.8%).     3 reports resource utilization on all hospitalizations during the 365-day follow-up period among MBs that experienced at least one hospitalization. The average total LOS during the entire follow-up period among MBs with a hospitalization was 10.5 ± 11.5 (median 7 days, interquartile range 3 to 13 days) and average total Medicare reimbursement was $21,343 ± $23,034 (median $14,113, interquartile range $7,451 to $26,587). The one-way AVOVA test found no signi icant differences in the distribution of any of these resource measures during the follow-up period between the two study cohorts in table 3. It is interesting to note that during the follow-up period, 107 MBs (1.5%) had an additional valve procedure, of which 76 had a second TAVR procedure. Further, the four most common reasons (based on DRG categories) for hospital readmissions during the followup period were: pulmonary edema and respiratory failure (23.1%), heart failure (21.5%), sepsis (11.2%) and COPD or pneumonia (10.5%). MBs in the PPM cohort were signi icantly more likely to have a readmission associated with pulmonary edema and respiratory failure (28.1% vs 22.2%) and heart failure (27.2% vs 20.5%). Part B of table 4 reports estimated incremental resource utilization between the two study cohorts obtained from the risk-adjusted median regression models for total episode hospital LOS and Medicare reimbursement. The results of the median regression models indicate that median Medicare reimbursement was signi icantly higher ($5,132) and median LOS was signi icant longer (1.84) days in the PPM cohort after controlling for difference in demographic characteristics and observed comorbid conditions between the two study cohorts.

Discussion
This analysis reports a set of nationally representative Medicare benchmarks for a MB's index TAVR hospitalization and all hospital encounters during a 365-day follow-up period. First, total average Medicare reimbursement to hospitals among the 15,533 MBs undergoing a TAVR procedure was $60,638 ± $28,974 for the entire study period. Average Medicare reimbursement for the index TAVR hospitalization accounted for 83.8% of total average reimbursement for the entire study period. Second, 54.0% (8,390) of the MBs undergoing a TAVR procedure did not have any hospital readmissions during the 365-day follow-up period. Third, MBs in the PPM cohort had higher average Medicare reimbursement during both the index TAVR hospitalization ($55,597 ± $19,781 versus $49,996 ± 19,727) and for the entire study period ($65,473 ± $29,053 versus $59,801 ± $28,879) than MBs in the non-PPM cohort. Finally, this study provides insight into the clinical reasons associated with MBs having a hospitalization following TAVR. The two most common DRG categories for readmission were pulmonary edema/respiratory failure and heart failure.
This paper provides insights into the inancial risks that healthcare providers will incur if a provider proceeds with a bundle payment program for TAVR procedures. After controlling for demographics and 47 comorbid conditions, this paper inds statistically signi icant longer total lengths of stay (1,8 days) and higher Medicare reimbursements ($5,132) for MBs receiving PPM implantations. This paper inds that nearly 55% of MBs irst hospital readmission occurred within 90 days of the index TAVR hospitalization. Further, average Medicare reimbursement during follow-up hospitalization for all MBs with at least one readmission in this study was $21,343 ± $23,034, approximately 42% of observed average Medicare reimbursement during the index TAVR hospitalization.
Furthermore, this paper provides insight into the clinical problems that resulted in readmissions during the follow-up period. In particular, this paper indings that over 4.0% of TAVR patients not receiving PPM during their index hospitalization underwent PPM procedures during the follow-up period. This inding supports concerns related to atrioventricular block following TAVR [18] and gives providers insight into potential bundling of devices from manufacturers to cover this additional cost. In addition, under bundled payments, it will be inancially advantageous to manage the comorbid conditions associated with readmissions. Given that 44.6% of readmissions were due to pulmonary edema and respiratory failure or heart failure, there appears to be opportunity for outpatient intervention, remote monitoring, telemedicine follow-up or other preemptive maneuvers to help avoid these readmissions. Finally, this study found MBs in the PPM cohort were signi icantly more likely to use post-acute care, including home health agencies (35.5% vs 32.0%), skilled nursing Several limitations warrant discussion. First, this analysis applies only to MBs in the fee-for service program. A second limitation is that this study does not have any information concerning the resources consumed for using post-acute care services or outpatient procedures during the follow-up period. Another limitation of this study is Medicare reimbursement in this study is observed based on the payment rule and inancial incentives in the Medicare program during 2014 to 2016. It is not possible to speculate how hospitals and other healthcare providers will change their patterns of care, in response to the new inancial incentives associated with future Medicare's bundle payment programs and other value-based delivery models.