Abstract
BACKGROUND:
Percutaneous ventricular assist devices (PVADs) have been replacing intra-aortic balloon pumps for hemodynamic support during percutaneous coronary intervention (PCI), even though data supporting a benefit for hard clinical end points remain limited. We evaluated diffusion of PVADs across US markets and examined the association of market utilization of PVAD with mortality and cost.
METHODS:
Using the 2013 to 2019 Medicare data, we identified all patients aged ≥65 years who underwent PCI with either a PVAD or intra-aortic balloon pump. We used hospital referral region to define regional health care markets and categorized them in quartiles based on the proportional use of PVADs during PCI. Multilevel models were constructed to determine the association of patient, hospital, and market factors with utilization of PVADs and the association of PVAD utilization with 30-day mortality and cost.
RESULTS:
A total of 79 176 patients underwent PCI with either intra-aortic balloon pump (47 514 [60.0%]) or PVAD (31 662 [40.0%]). The proportion of PCI procedures with PVAD increased over time (17% in 2013 to 57% in 2019; P for trend, <0.001), such that PVADs overtook intra-aortic balloon pump for hemodynamic support during PCI in 2018. There was a large variation in PVAD utilization across markets (range, 0%–85%), which remained unchanged after adjustment of patient characteristics (median odds ratio, 2.05 [95% CI, 1.91–2.17]). The presence of acute myocardial infarction, cardiogenic shock, and emergent status was associated with a 45% to 50% lower odds of PVAD use suggesting that PVADs were less likely to be used in the sickest patients. Greater utilization of PVAD at the market level was not associated with lower risk mortality but was associated with higher cost.
CONCLUSIONS:
Although utilization of PVADs for PCI continues to increase, there is large variation in PVAD utilization across markets. Greater market utilization of PVADs was not associated with lower mortality but was associated with higher cost.
Keywords: health services research, heart-assist devices, percutaneous coronary intervention
Graphical Abstract
Mechanical circulatory support (MCS) devices, which include intra-aortic balloon pump (IABP; Maquet, Inc) or percutaneous ventricular assist devices (PVADs), including Impella (Abiomed, Inc) and TandemHeart (Cardiac Assist), can be used to provide circulatory support in patients undergoing high-risk percutaneous coronary intervention (PCI). Although studies have found PVAD to provide greater hemodynamic support when compared with IABP,1 randomized controlled trials (RCTs) have not found PVADs to be superior to IABP in reducing the risk of hard clinical end points (eg, mortality). In the PROTECT-II trial (A Prospective, Multi-Center, Randomized Controlled Trial of the IMPELLA RECOVER LP 2.5 System Versus Intra Aortic Balloon Pump in Patients Undergoing Non Emergent High Risk PCI), 452 patients undergoing elective high-risk PCI were randomized to either Impella 2.5 or IABP for hemodynamic support. The trial was stopped early due to futility, and there was no difference in the primary end point in the intention-to-treat analysis.2 Several observational studies including recent studies from the National Cardiovascular Data Registry Cath PCI registry and the Premier Healthcare Database found a marked increase in utilization of PVADs across the United States.3–5 Both studies found Impella to be associated with higher risk of in-hospital mortality and bleeding complications.
While studies have documented continued increase in the use of PVADs, it remains unclear how and where the diffusion of PVADs is occurring in the United States. Previous research examining patterns of diffusion of invasive cardiac services such as PCI and coronary artery bypass graft surgery across the US health care markets found a strong association between market-level factors such as market-level competition and availability of similar services at nearby hospitals.6 These studies found that hospitals tend to adopt new cardiac services for competitive reasons, including generous reimbursement, and attract patients and doctors. However, as these cardiac services tend to be adopted more rapidly in markets with greater competition, it also could lead to the dilution of expertise with greater diffusion. Although similar incentives may drive the utilization of PVADs, a critical difference is that unlike revascularization,7,8 the evidence of effectiveness for PVADs remains limited. Therefore, it is likely that discretionary use contributes to the variability in utilization and diffusion of PVADs across US markets.
Accordingly, we examined diffusion of PVAD into US health care markets and evaluated the association of patient-, hospital-, and market-level characteristics with utilization of PVAD among Medicare beneficiaries. We also examined whether health care markets with a higher proportion of PVAD use achieve better outcomes for patients undergoing PCI with MCS and the cost of hospitalization.
METHODS
The authors will not make the data, methods used in the analysis, and materials used to conduct the research available to any researcher for purposes of reproducing the results or replicating the procedure.
Data Sources and Study Participants
We used the Medicare Provider Analysis and Review from 2013 to 2019 to identify patients aged ≥65 years who underwent inpatient PCI with placement of either an IABP or a PVAD on the same date as the PCI procedure. Patients receiving both IABP and PVAD on the same day were classified as receiving PVAD as PVAD may represent escalation in care. For patients with multiple admissions during the study period, we only included the first admission for PCI. Patients with trauma, unknown admission type, or missing health care market information were excluded.
The Dartmouth Atlas Project files were used to map each hospital to the corresponding hospital referral region (HRR), which represent regional health care markets for tertiary medical care with geographic contiguity.9 By definition, each HRR contains at least 1 hospital that performs major cardiovascular procedures and neurosurgery and has a minimum population of 120 000. We mapped our study patients to the 306 HRRs as defined by the Dartmouth Atlas Project to assess diffusion of PVADs.10
Study Variables
The Medicare Provider Analysis and Review data set includes information on demographic variables, including age, sex, race; calendar year, primary and secondary diagnoses; source of hospital admission; and mortality. Comorbidities were defined using algorithms developed previously.11 Hospital characteristics were obtained from the American Hospital Association Annual Survey and included hospital teaching status, ownership, and bed count. Data on market-level variables were market size, number of hospitals, health maintenance organization penetration, and market-level competition. Market size was defined as the number of Medicare enrollees within each HRR. Health maintenance organization penetration was defined as the proportion of enrollees within each market who were enrolled in a Medicare-managed care plan (ie, Medicare Advantage), as used in previous studies.12 Market-level competition was determined using the Herfindahl-Hirschman Index (HHI), which is calculated as the sum of the squared market share of each hospital within each HRR and has been widely used to evaluate competition within markets.13 A high HHI represents high concentration (ie, low competition) within an HRR, and a low HHI represents low concentration (ie, high competition). For example, a market with only 1 hospital (ie, no competition) will have an HHI of 1. In contrast, highly competitive markets with multiple hospitals, each with a small market share and, therefore, unconcentrated market, will have a small HHI.14 Each of the above hospital variables (market size, health maintenance organization penetration, and market competition) were categorized based on the quartiles of their distribution as follows: low (first quartile), medium (second and third quartiles), and high (fourth quartile).
The primary outcome was 30-day mortality. The secondary outcome was cost of hospitalization, which was estimated from a payer’s perspective and reflected the amount Medicare would have paid before applying deductibles, coinsurance, or other insurance payments. Since hospital payment for procedures is based on diagnosis-related groups, which accounts for disease severity, we did not adjust for patient variables in this analysis.
Statistical Analysis
For each HRR, we calculated the utilization of PVADs during PCI as the number of patients who received a PVAD during PCI divided by the total number of number of patients who received either an IABP or a PVAD during PCI. We ranked HRRs into quartiles based on the proportion of all PCIs that were performed with a PVAD. We compared patient, hospital, and market-level characteristics across HRR quartiles using the Mantel-Haenszel test for trend for categorical variables and linear regression for continuous variables.
We examined the association of patient-, hospital-, and market-level characteristics with use of PVADs using multivariable logistic regression models. To evaluate the association of market utilization of PVAD with patient outcomes, we used a 2-step modeling approach. First, to estimate the likelihood that a patient received PVAD (compared with IABP), we created a logistic regression model using patient, hospital, and market characteristics, as well as the year of the procedure as independent variables using a stepwise selection strategy. After identifying variables that were associated with use of PVAD from the logistic regression model, we constructed a hierarchical model (3 level, with patients nested within hospitals and hospitals nested within HRRs) to evaluate the adjusted association of patient, hospital, and market variables with PVAD use. In these models, hospital site and the market were included as a random variable. Such models avoid overestimation of significance of statistical associations and account for clustering of patient complexity both at the hospital and the HRR level.15 Models sequentially adjusted for patient, hospital, and market characteristics as fixed effects.
We also quantified the extent of variation in PVAD utilization across HRRs by calculating the median odds ratio (MOR).16 The MOR is derived from the variance estimate of the random intercept from the hierarchical model and is always >1. Conceptually, the MOR represents the relative odds that a patient undergoing PCI in one randomly selected HRR received a PVAD compared with an identical patient in another randomly selected HRR who did not receive a PVAD during PCI. An MOR of 1.0 would indicate no HRR-level variation in PVAD utilization, whereas an MOR of 2.0 would indicate a 2-fold variation in the use of PVAD in 2 randomly selected HRRs. The MOR for receiving PVAD across HRRs was calculated after adjusting for patient variables, patient and hospital variables, and patient, hospital, and market variables. Only variables that satisfied a P criteria of 0.05 were retained in the model.
Finally, we created similar 3-level hierarchical multivariable regression models to evaluate the adjusted association of market-level utilization of PVADs with 30-day mortality. In these models, 30-day mortality was included as the dependent variable; patient, hospital, and market variables were included as fixed effects; and the hospital site and HRR were included as random effects, to account for clustering of patients within hospitals and markets. These models adjusted for differences in patient characteristics both at the site level and the market level.
In sensitivity analyses, we evaluated market-level variation in PVAD use and its association with 30-day mortality and cost after excluding patients who received both IABP and PVAD. All statistical analyses were performed using SAS (version 9.4; SAS Institute, Cary, NC). The Institutional Review Board at the University of Iowa and the Cleveland Clinic approved the study protocol and waived the requirement for informed consent as the study only included deidentified data.
RESULTS
A total of 79 176 Medicare patients underwent PCI with either an IABP (47 514; 60.0%) or a PVAD (31 662; 40.0%) during 2013 to 2019. During the study period, there was a significant increase in the total number of PCI procedures in which any MCS device (IABP or PVAD) was used, from 8584 in 2013 to 14 459 in 2019 (P for trend, <0.001; Figure 1). The total number of PCI procedures with concomitant IABP declined during this period from 7092 in 2013 to 6225 in 2019 (P for trend, 0.001), whereas the total number of PCIs with PVAD increased by 450%, from 1492 in 2013 to 8231 in 2019 (P for trend, <0.001). In 2018, PVADs overtook IABP for MCS support during PCI, and the trend continued in 2019. The mean cost of hospitalization was higher in patients receiving PVAD compared with IABP ($53 691 versus $25 498; P<0.001).
Figure 1. Calendar year trends in the number of percutaneous ventricular assist devices (PVADs) and intra-aortic balloon pump (IABP) by calendar year.
The number of percutaneous coronary intervention (PCI) procedures with a PVAD increased significantly from 1492 in 2013 to 8231 in 2019. P for trend, <0.001. In contrast, the number of PCI procedures with IABP decreased from 7092 in 2013 to 6225 in 2019. P for trend, <0.001. MCS indicates mechanical circulatory support.
At a market level, there was large variation in the use of PVAD during PCI across HRRs, ranging from 0% to 85% (median, 33%; interquartile range, 23%–43%; Figure 2). When HRRs were categorized into quartiles based on utilization of PVAD during PCI, the proportions were as follows: quartile 1, lowest, <22%; quartile 2, 22% to 33%; quartile 3, 34% to 43%; and quartile 4, highest, >43%. The geographic distribution of HRRs is displayed in Figure 3.
Figure 2.
Market variation in utilization of percutaneous ventricular assist devices (PVADs) for supported percutaneous coronary intervention (PCI).
There was a large market variation in the use of PVADs among patients undergoing PCI with a mechanical circulatory support device, ranging from 0% to 85% with a median odds ratio of 2.05 (95% CI, 1.91–2.17). HRRs indicates hospital referral regions.
Figure 3. Geographic distribution of markets according to average percutaneous ventricular assist device use for supported percutaneous coronary interventions from 2013 to 2019 across hospital referral regions.
The US map shows each of the 306 hospital referral regions based on the Dartmouth Atlas according to their utilization of percutaneous ventricular assist devices for supported percutaneous coronary interventions. The median utilization of percutaneous ventricular assist device was 28% (interquartile range, 17%–38%; range, 0%–85%).
Table 1 shows the characteristics of included patients, overall and stratified by quartiles of market-level PVAD utilization. The mean age was 76.1 years, the majority of patients were White (84.6%), and two-thirds (64.2%) were male. More than two-thirds (67.9%) of patients were admitted with a diagnosis of acute myocardial infarction, and 51.7% had cardiogenic shock. There was a high prevalence of major comorbidities including hypertension (83.5%), chronic heart failure (57.4%), diabetes (44.9%), and chronic pulmonary disease (25.0%). Patients in high PVAD utilization HRR (quartile 4) were more likely to be of non-White race (16.6% versus 14.4%) and have dual Medicare/Medicaid eligibility (19.1% versus 14.6%) but less likely to have acute myocardial infarction (62.5% versus 74.1%), cardiogenic shock (46.5% versus 56.8%) or an emergent hospital admission (57.1% versus 64.8%; P<0.001 for all), compared with patients in low PVAD utilization HRR (quartile 1). There was a higher prevalence of comorbidities such as chronic heart failure, diabetes, and chronic kidney disease among patients in markets with higher utilization of PVAD, but the absolute differences were small in magnitude.
Table 1.
Patient Characteristics Stratified by Quartile of PVAD Use Among Patients Undergoing PCI With MCS Across Hospital Referral Regions
| Patient characteristics | Overall (n=79 176) | Quartile 1 (n=12 402) | Quartile 2 (n=17 909) | Quartile 3 (n=26 085) | Quartile 4 (n=22 780) | P value for trend |
|---|---|---|---|---|---|---|
| MCS device, n (%) | ||||||
| IABP | 47 514 (60.0) | 10 112 (81.5) | 12 206 (68.2) | 15 2 74 (58.6) | 9922 (43.6) | |
| PVAD | 31 662 (40.0) | 2290 (18.5) | 5703 (31.8) | 10 811 (41.4) | 12 858 (56.4) | |
| Age, y; mean (SD) | 76.1 (7.4) | 76.1 (7.4) | 76.3 (7.5) | 76.2 (7.5) | 75.9 (7.3) | <0.001 |
| Race, n (%) | <0.001 | |||||
| White | 66 944 (84.6) | 10 613 (85.6) | 16 266 (90.8) | 21 058 (80.7) | 19 007 (83.5) | |
| Black | 5884 (74) | 919 (74) | 799 (4.5) | 2154 (8.3) | 2012 (8.8) | |
| Other | 6348 (8.0) | 870 (7.0) | 844 (4.7) | 2873 (11.0) | 1761 (7.7) | |
| Male, n (%) | 50 818 (64.2) | 7826 (63.1) | 11 449 (63.9) | 16 787 (64.4) | 14 756 (64.8) | 0.01 |
| Acute MI, n (%) | 53 719 (67.9) | 9195 (74.1) | 12 732 (71.1) | 17 552 (67.3) | 14 240 (62.5) | <0.001 |
| Cardiogenic shock, n (%) | 40 895 (51.7) | 7048 (56.8) | 9633 (53.8) | 13 614 (52.2) | 10 600 (46.5) | <0.001 |
| Admission type, n (%) | <0.001 | |||||
| Elective | 13 471 (17.0) | 1765 (14.2) | 2476 (13.8) | 4336 (16.6) | 4894 (21.5) | |
| Urgent | 17 305 (21.9) | 2606 (21.0) | 4020 (22.5) | 5804 (22.3) | 4875 (21.4) | |
| Emergent | 48 400 (61.1) | 8031 (64.8) | 11 413 (63.7) | 15 945 (61.1) | 13 011 (57.1) | |
| Dual Medicare/Medicaid eligibility | 13 492 (17.0) | 18.7 (14.6) | 2542 (14.2) | 4788 (18.4) | 4355 (19.1) | <0.001 |
| Comorbidities, n (%) | ||||||
| Chronic heart failure | 45 438 (57.4) | 6578 (53.0) | 9647 (53.9) | 15 614 (59.9) | 13 599 (59.7) | <0.001 |
| Valve disease | 20 085 (25.4) | 3000 (24.2) | 4650 (26.0) | 6682 (25.6) | 5753 (25.3) | 0.004 |
| Peripheral vascular disease | 15 780 (19.9) | 2379 (19.2) | 3284 (18.3) | 5399 (20.7) | 4718 (20.7) | <0.001 |
| Hypertension | 66 115 (83.5) | 10 070 (81.2) | 14 710 (82.1) | 21 913 (84.0) | 19 422 (85.3) | <0.001 |
| Chronic pulmonary disease | 19 826 (25.0) | 2982 (24.0) | 4454 (24.9) | 6406 (24.6) | 5984 (26.3) | <0.001 |
| Diabetes | 35 545 (44.9) | 5210 (42.0) | 7529 (42.0) | 1 1 987 (46.0) | 10 819 (47.5) | <0.001 |
| Anemia | 20 782 (26.3) | 3004 (24.2) | 4253 (23.8) | 7190 (27.6) | 6335 (27.8) | <0.001 |
| Stroke | 6178 (78) | 866 (70) | 1339 (75) | 2103 (8.1) | 1870 (8.2) | <0.001 |
| Chronic kidney disease | 20 486 (25.9) | 2868 (23.1) | 4237 (23.7) | 7216 (27.7) | 6168 (27.1) | <0.001 |
| Smoking | 19 431 (24.5) | 2941 (23.7) | 4444 (24.8) | 6155 (23.6) | 5891 (25.9) | <0.001 |
| Previous PCI | 15 650 (19.8) | 2268 (18.3) | 3394 (19.0) | 5152 (19.8) | 4836 (21.2) | <0.001 |
IABP indicates intra-arterial balloon pump; MCS, mechanical circulatory support; MI, myocardial infarction; PCI, percutaneous coronary intervention; and PVAD, percutaneous ventricular assist device.
Table 2 shows the distribution of hospital and market characteristics. Hospitals in high PVAD utilization HRRs (quartile 4) were more likely to be for profit (32.7% versus 17.8%; P<0.001) and located in the South (55.7% versus 36.3%; P<0.001) when compared with low PVAD utilization HRRs (quartile 1). There was no significant difference in hospital bed count or teaching status across quartiles of HRR based on PVAD utilization. Markets with high utilization of PVAD (quartile 4) were larger in size (29.0% versus 18.4%; P<0.001), had more hospitals (mean, 12 versus 8; P<0.001), and were more likely to be unconcentrated (ie, highly competitive, 26.3% versus 17.1%; P=0.01) when compared with markets with the lowest utilization of PVADs (quartile 1).
Table 2.
Hospital and Market Characteristics Stratified by Quartile of PVAD Use Among Patients Undergoing PCI Mechanical Circulatory Support Across Hospital Referral Regions
| Overall (n=1748) | Quartile 1 (n=311) | Quartile 2 (n=382) | Quartile 3 (n=590) | Quartile 4 (n=465) | P value for trend | |
|---|---|---|---|---|---|---|
| Hospital characteristics | ||||||
| Bed count, mean (SD) | 323 (246) | 298 (196) | 321 (231) | 321 (247) | 332 (281) | 0.3 |
| Hospital teaching status* n (%) | 0.6 | |||||
| Teaching | 1147 (66.6) | 200 (64.7) | 254 (68.3) | 382 (65.5) | 311 (67.8) | |
| Nonteaching | 576 (33.4) | 109 (35.3) | 118 (31.7) | 201 (34.5) | 148 (32.2) | |
| Hospital ownership* n (%) | <0.001 | |||||
| For profit | 376 (21.8) | 55 (17.8) | 60 (16.1) | 111 (19.0) | 150 (32.7) | |
| Nonprofit | 1191 (69.1) | 227 (73.5) | 282 (75.8) | 420 (72.0) | 262 (57.1) | |
| Government | 156 (9.1) | 27 (8.5) | 30 (8.1) | 52 (8.9) | 47 (10.2) | |
| Region, n (%) | <0.001 | |||||
| West | 359 (20.5) | 60 (19.3) | 46 (12.0) | 160 (271) | 93 (20.0) | |
| Northeast | 249 (14.2) | 29 (9.3) | 121 (31.7) | 58 (9.8) | 41 (8.8) | |
| Midwest | 436 (25.0) | 109 (35.1) | 120 (31.4) | 135 (22.9) | 72 (15.5) | |
| South | 704 (40.3) | 113 (36.3) | 95 (24.9) | 237 (40.2) | 259 (55.7) | |
| Market characteristics | Overall (n=306) | Quartile (n=76) | Quartile 2 (n=78) | Quartile 3 (n=76) | Quartile 4 (n=76) | P value for trend |
| Market size, n (%) | 0.003 | |||||
| Small | 77 (25.2) | 30 (39.5) | 17 (21.8) | 14 (18.4) | 16 (21.0) | |
| Medium | 153 (50.0) | 32 (42.1) | 47 (60.2) | 36 (47.4) | 38 (50.0) | |
| Large | 76 (24.8) | 14 (18.4) | 14 (18.0) | 26 (34.2) | 22 (29.0) | |
| No. of hospitals, mean (SD) | 11 (11) | 8 (7) | 10 (9) | 14 (14) | 12 (11) | <0.001 |
| No. of hospitals, n (%) | <0.001 | |||||
| 1–5 | 113 (36.9) | 37 (48.7) | 34 (43.6) | 23 (30.3) | 19 (25.0) | |
| 6–13 | 119 (38.9) | 29 (38.1) | 30 (38.5) | 25 (32.9) | 35 (46.1) | |
| >14 | 74 (24.2) | 10 (13.2) | 14 (17.9) | 28 (36.8) | 22 (28.9) | |
| HMO penetration, n (%) | 0.1 | |||||
| Low | 76 (24.8) | 25 (32.9) | 20 (25.6) | 18 (23.7) | 13 (171) | |
| Medium | 154 (50.4) | 36 (47.4) | 36 (46.2) | 36 (47.3) | 46 (60.5) | |
| High | 76 (24.8) | 15 (19.7) | 22 (28.2) | 22 (29.0) | 17 (22.4) | |
| Market-level competition, n (%) | 0.01 | |||||
| Highly concentrated | 76 (24.8) | 24 (31.6) | 21 (26.9) | 14 (18.4) | 17 (22.4) | |
| Moderately concentrated | 154 (50.4) | 39 (51.3) | 44 (56.4) | 32 (42.1) | 39 (51.3) | |
| Unconcentrated | 76 (24.8) | 13 (17.1) | 13 (16.7) | 30 (39.5) | 20 (26.3) | |
Market size was defined based on the total number of Medicare beneficiaries in each market and categorized as small (<54 925), medium (54 925–191 805), and large (>191 805). HMO penetration was defined as the percentage of Medicare beneficiaries with HMO coverage and categorized as low (<20%), medium (20%- 40.7%), and high (>40.7%). Market-level competition was defined using HHI and categorized as highly concentrated markets (HHI, >0.41) suggesting low competition, moderately concentrated markets (HHI, 0.15–0.41), and unconcentrated markets (HHI, <0.15) suggesting a highly competitive market. HHI indicates Herfindahl- Hirschman Index; HMO, Health Maintenance Organization; PCI, percutaneous coronary intervention; and PVAD, percutaneous ventricular assist device.
Information missing from 25 hospitals.
The association of patient-, hospital-, and market-level variables with use of PVAD compared with IABP is reported in Table S1. Older age and male sex were strongly associated with the use of PVAD. Although most comorbidities were positively associated with the use of PVAD compared with IABP, acute myocardial infarction and cardiogenic shock were associated with lower odds of receiving a PVAD (odds ratio, 0.45 and 0.50; P<0.001). Similarly, compared with elective admission, urgent and emergent admission status was associated with lower odds of PVAD compared with IABP. There was a strong association of PVAD use with calendar year, with an odds ratio of 12.28 (95% CI, 9.20–16.40) for patients in 2019 compared with 2013. Among hospital and market variables, teaching status and medium health maintenance organization penetration within a market were significantly associated with the use of PVAD.
After adjusting for differences in patient characteristics, the MOR for PVAD utilization was 2.05 (95% CI, 1.91–2.17), which suggests that the relative odds of using a PVAD compared with an IABP for a patient undergoing PCI in one randomly selected HRR were ≈2× higher when compared with an identical patient in another randomly selected HRR. The MOR remained largely unchanged after the addition of hospital (MOR, 2.05 [95% CI, 1.92–2.18]) and market variables (MOR, 2.07 [95% CI, 1.93–2.20]), respectively.
Overall, 30-day mortality was 32.1% and decreased across quartiles. Risk-adjusted mortality was 31.5%, 31.2%, 31.3%, and 31.0% from HRRs with low-to-high PVAD utilization, respectively. Higher market utilization of PVAD was not associated with lower 30-day mortality (P for trend, 0.074; Table 3). The full model results are reported in Table S2. The mean cost of hospitalization was $36 787. Higher utilization of PVAD was associated with higher hospitalization cost (P for trend, <0.001; Table 3).
Table 3.
Association of PVAD Utilization With 30-Day Mortality and Cost of Hospitalization
| PVAD use within HRRs | Quartile 1 (<17%) | Quartile 2 (17%-28%) | Quartile 3 (28%-38%) | Quartile 4 (>38%) | P value for trend |
|---|---|---|---|---|---|
| 30-d mortality | |||||
| Unadjusted | 32.9% | 32.7% | 32.5% | 30.6% | <0.001 |
| Risk adjusted | 31.5% | 31.2% | 31.3% | 31.0% | 0.074 |
| Adjusted cost* | 32 670 | 35 316 | 36 726 | 40 252 | <0.001 |
HRR indicates hospital referral region; and PVAD, percutaneous ventricular assist device.
Information missing from 8 patients. The cost data represent the amount Center for Medicare and Medicaid Services would have paid for each patient before applying deductible, coinsurance, or other insurance. Hospital cost was derived from reimbursement based on diagnosis-related groups, which already account for severity of patient illness.
A total of 2931 (3.7%) patients received both IABP and PVAD on the same day as the PCI procedure. To determine whether these patients influenced the results, a sensitivity analysis was performed in which patients receiving both IABP and PVAD were excluded. As shown in Tables S3 and S4, our findings were similar even after exclusion of these patients. Higher utilization of PVAD was not associated with lower 30-day mortality (P for trend, 0.14) but was associated with higher cost (P for trend, 0.01), similar to the main analysis.
DISCUSSION
In this contemporary study of Medicare beneficiaries, we found continued growth in the use of PVADs for providing MCS during PCI, such that PVADs overtook IABP for PCI support in 2018 and the upward trend continued in 2019. Notably, utilization of PVADs during PCI varied markedly across health care markets ranging from 0% to >80%, and this variation was largely unexplained by differences in patient, hospital, and market variables. Notably, higher utilization of PVADs in some markets was not associated with lower 30-day mortality but was associated with higher cost of hospitalization. A number of these findings merit further discussion.
In recent years, studies have shown an exponential growth in utilization of PVADs over IABP during PCI, with one study reporting the percentage of PVAD use increasing from 9.9% in 2008 to 31.9% in 2016.5 Our findings are consistent with these prior studies and demonstrate that PVAD use continues to grow. Of note, this dramatic growth has occurred even though no RCT has demonstrated a reduction in hard clinical end points with the use of PVADs during PCI. The PROTECT-II trial, which was designed to evaluate the efficacy of Impella 2.5 compared with IABP for elective high-risk PCI, was stopped early due to futility. Use of PVADs in the setting of acute myocardial infarction has also been studied in 2 small RCTs, which found improved hemodynamics with the use of PVADs but no difference in mortality.17,18 Although data from the National Cardiogenic Shock Initiative suggests improved survival in patients with AMI and cardiogenic shock using a structured protocol that included PVADs, the study lacked a comparator group and used a bundled approach, which makes it difficult to determine the contribution of PVADs in improved outcomes.19 Given the paucity of data from RCTs, comparative observational studies provide key data on PVADs. Data from the National Cardiovascular Data Registry found PVADs to be associated with a higher risk of in-hospital death and major bleeding compared with IABP4—a finding that was also seen in a separate study from the Premier health care database.5
Greater hemodynamic support has been touted as the main benefit of PVADs and was the basis for the initial clearance of these devices by the Food and Drug Administration. Therefore, it was surprising that acute myocardial infarction, cardiogenic shock, and emergent admission status were associated with 40% to 50% lower odds of PVAD compared with an IABP. These findings are counterintuitive and highlight the existence of a risk-treatment paradox,20 that is, patients with the greatest potential for hemodynamic collapse were less likely to receive PVAD, even though the latter provides greater hemodynamic support. These findings are contrary to expectation and suggest that comparative effectiveness of PVAD versus IABP in real-world observational studies may have been biased in favor of PVAD due to the preferential use of PVADs in lower risk patients. A notable finding of our study was the marked variation in use of PVADs across HRRs ranging from 0% to 84% with an MOR of >2, which remained largely unchanged after adjusting for patient characteristics, comorbidities, and additional adjustment of hospital and market variables, suggesting that variation in PVAD use is likely driven by discretionary use.
An important contribution of our study is the lack of association between utilization of PVADs and 30-day mortality at a market level. In contrast, higher utilization of PVADs was associated with higher cost. Although PVADs are more expensive compared with IABP, generous reimbursement from payors more than offset the higher cost of PVADs, which is likely a major driver of the large growth in utilization in recent years. Thus, in the absence of strong data demonstrating effectiveness of PVADs in reducing mortality or other clinical end points, our findings raise concerns that the current growth in PVAD utilization has contributed to an increase in cost without a meaningful improvement in patient outcomes. Moreover, the large variation in PVAD utilization across markets highlights the uncertainty regarding their overall value and suggests an urgent need for RCTs to guide clinical practice.
The findings of our study should be interpreted in the context of these limitations. First, because of reliance on International Classification of Diseases codes within administrative data, we are unable to distinguish whether a patient received an Impella or TandemHeart because both devices have the same International Classification of Diseases procedure code. However, in our experience, most of the growth in recent years has occurred with Impella. In 2021, Abiomed announced that they had surpassed 200 000 patients treated with Impella.21 Second, as with any observational study, there is potential for unmeasured confounding factors due to unmeasured variables such as coronary anatomy and severity of heart failure. However, we found that PVADs were used less often in patients undergoing PCI in the setting of acute myocardial infarction (higher risk) and were used more often in elective admissions. Moreover, we also found differences in measured patient characteristics across quartiles of PVAD utilization, which are unlikely to explain the large difference in utilization of PVADs. Second, the Medicare Provider Analysis and Review data set includes only Medicare beneficiaries who are ≥65 years of age. Therefore, results may not be generalizable to a younger population. Third, we were unable to differentiate between the different types of PVADs, including Impella 2.5, Impella CP, Impella 5.0, or TandemHeart, that provide different degree of hemodynamic support.
In conclusion, among patients undergoing PCI with MCS, we found large market-level variation in the use of PVADs in the United States. At a market level, higher utilization of PVAD was not associated with lower 30-day mortality but was associated with higher cost.
Supplementary Material
WHAT IS KNOWN
Use of percutaneous ventricular assist devices (PVADs) during percutaneous coronary intervention is growing despite limited benefit for hard clinical end points.
Diffusion of PVADs during percutaneous coronary intervention across US markets and factors associated with PVAD utilization remain unclear.
WHAT THE STUDY ADDS
Use of PVADs during percutaneous coronary intervention continues to increase with large variation across US health care markets.
Patients with acute myocardial infarction, cardiogenic shock, and admitted urgently/emergently were less likely to receive PVAD during percutaneous coronary intervention, suggesting improved patient selection for PVAD is needed.
Higher utilization of PVAD at a market level was not associated with lower mortality but was associated with higher cost.
Nonstandard Abbreviations and Acronyms
- HHI
Herfindahl-Hirschman index
- HRR
hospital referral region
- IABP
intra-aortic balloon pump
- MCS
mechanical circulatory support
- MOR
median odds ratio
- PCI
percutaneous coronary intervention
- PVAD
percutaneous ventricular assist device
- RCT
randomized controlled trial
Footnotes
Disclosures
Dr Dhruva reports receiving research funding from the National Heart, Lung, and Blood Institute (K12HL138046), the National Evaluation System for Health Technology Coordinating Center, the Food and Drug Administration, Arnold Ventures, the Greenwall Foundation, and the National Institute for Health Care Management. Dr Desai works under contract with the Centers for Medicare and Medicaid Services. He reports research grants and consulting for Amgen, AstraZeneca, Bristol Myers Squibb, Boehringer Ingelheim, Cytokinetics, MyoKardia, Relypsa, Novartis, and SC Pharmaceuticals. Dr Girotra is supported by a grant from the NIH (R56HL158803). He also receives funding from the American Heart Association for editorial work. The other authors report no conflicts.
Contributor Information
Thorarinn A. Bjarnason, Division of Cardiovascular Medicine, Department of Internal Medicine, University of Iowa Carver College of Medicine, Iowa City..
Amgad Mentias, Department of Cardiovascular Medicine, Heart, Thoracic and Vascular Institute, Cleveland Clinic, OH.
Sidakpal Panaich, Division of Cardiovascular Medicine, Department of Internal Medicine, University of Iowa Carver College of Medicine, Iowa City..
Mary Vaughan Sarrazin, University of Iowa Carver College of Medicine and Center for Access and Delivery Research and Evaluation, Iowa City; General Internal Medicine, Department of Internal Medicine, University of Iowa Carver College of Medicine, Iowa City..
Yubo Gao, General Internal Medicine, Department of Internal Medicine, University of Iowa Carver College of Medicine, Iowa City..
Milind Desai, Department of Cardiovascular Medicine, Heart, Thoracic and Vascular Institute, Cleveland Clinic, OH.
Ambarish Pandey, Division of Cardiovascular Medicine, Department of Internal Medicine, University of Texas Southwestern Medical Center, Dallas.
Sanket S. Dhruva, Department of Medicine, University of California, San Francisco.
Nihar R. Desai, Center for Outcomes Research and Evaluation and Department of Cardiovascular Medicine, Yale School of Medicine, New Haven, CT.
Saket Girotra, Division of Cardiovascular Medicine, Department of Internal Medicine, University of Texas Southwestern Medical Center, Dallas.
REFERENCES
- 1.Shah NR, Bieniarz MC, Basra SS, Paisley RD, Loyalka P, Gregoric ID, Mann DL, Kar B. Serum biomarkers in severe refractory cardiogenic shock. JACC Heart Fail. 2013;1:200–206. doi: 10.1016/j.jchf.2013.03.002 [DOI] [PubMed] [Google Scholar]
- 2.O’Neill WW, Kleiman NS, Moses J, Henriques JP, Dixon S, Massaro J, Palacios I, Maini B, Mulukutla S, Dzavík V, et al. A prospective, randomized clinical trial of hemodynamic support with Impella 2.5 versus intra-aortic balloon pump in patients undergoing high-risk percutaneous coronary intervention: the PROTECT II study. Circulation. 2012;126:1717–1727. doi: 10.1161/CIRCULATIONAHA.112.098194 [DOI] [PubMed] [Google Scholar]
- 3.Khera R, Cram P, Lu X, Vyas A, Gerke A, Rosenthal GE, Horwitz PA, Girotra S. Trends in the use of percutaneous ventricular assist devices: analysis of national inpatient sample data, 2007 through 2012. JAMA Intern Med. 2015;175:941–950. doi: 10.1001/jamainternmed.2014.7856 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4.Dhruva SS, Ross JS, Mortazavi BJ, Hurley NC, Krumholz HM, Curtis JP, Berkowitz A, Masoudi FA, Messenger JC, Parzynski CS, et al. Association of use of an intravascular microaxial left ventricular assist device vs intra-aortic balloon pump with in-hospital mortality and major bleeding among patients with acute myocardial infarction complicated by cardiogenic shock. JAMA. 2020;323:734–745. doi: 10.1001/jama.2020.0254 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Amin AP, Spertus JA, Curtis JP, Desai N, Masoudi FA, Bach RG, McNeely C, Al-Badarin F, House JA, Kulkarni H, et al. The evolving landscape of impella use in the united states among patients undergoing percutaneous coronary intervention with mechanical circulatory support. Circulation. 2020;141:273–284. doi: 10.1161/CIRCULATIONAHA.119.044007 [DOI] [PubMed] [Google Scholar]
- 6.Horwitz JR, Nichols A, Nallamothu BK, Sasson C, Iwashyna TJ. Expansion of invasive cardiac services in the United States. Circulation. 2013;8:803–810. doi: 10.1161/CIRCULATIONAHA.112.000836 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Hochman JS, Sleeper LA, Webb JG, Sanborn TA, White HD, Talley JD, Buller CE, Jacobs AK, Slater JN, Col J, et al. Early revascularization in acute myocardial infarction complicated by cardiogenic shock. SHOCK investigators. Should we emergently revascularize occluded coronaries for cardiogenic shock. N Engl J Med. 1999;341:625–634. doi: 10.1056/NEJM199908263410901 [DOI] [PubMed] [Google Scholar]
- 8.Huynh T, Perron S, O’Loughlin J, Joseph L, Labrecque M, Tu JV, Théroux P. Comparison of primary percutaneous coronary intervention and fibrinolytic therapy in ST-segment-elevation myocardial infarction: Bayesian hierarchical meta-analyses of randomized controlled trials and observational studies. Circulation. 2009;119:3101–3109. doi: 10.1161/CIRCULATIONAHA.108.793745 [DOI] [PubMed] [Google Scholar]
- 9.Wennberg JE, Cooper MM. The Dartmouth Atlas of Health Care in the United States. Accessed June 11, 2022. https://www.dartmouthatlas.org
- 10.Dartmouth Atlas Project. Appendix on the Geography of Health Care in the United States. Accessed June 11, 2022. https://data.dartmouthatlas.org/downloads/methods/geogappdx.pdf.
- 11.Quan H, Sundararajan V, Halfon P, Fong A, Burnand B, Luthi JC, Saunders LD, Beck CA, Feasby TE, Ghali WA. Coding algorithms for defining comorbidities in ICD-9-CM and ICD-10 administrative data. Med Care. 2005;43:1130–1139. doi: 10.1097/01.mlr.0000182534.19832.83 [DOI] [PubMed] [Google Scholar]
- 12.Zhang Y, Hollenbeck BK, Schroeck FR, Jacobs BL. Managed care and the dissemination of robotic prostatectomy. Surg Innov. 2014;21:566–571. doi: 10.1177/1553350614524841 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Cutler DM, Scott Morton F. Hospitals, market share, and consolidation. JAMA. 2013;310:1964–1970. doi: 10.1001/jama.2013.281675 [DOI] [PubMed] [Google Scholar]
- 14.Wright JD, Tergas AI, Hou JY, Burke WM, Chen L, Hu JC, Neugut AI, Ananth CV, Hershman DL. Effect of regional hospital competition and hospital financial status on the use of robotic-assisted surgery. JAMA Surg. 2016;151:612–620. doi: 10.1001/jamasurg.2015.5508 [DOI] [PubMed] [Google Scholar]
- 15.Goldstein H Multilevel Statistical Models. Oxford University Press; 1995. [Google Scholar]
- 16.Larsen K, Merlo J. Appropriate assessment of neighborhood effects on individual health: integrating random and fixed effects in multilevel logistic regression. Am J Epidemiol. 2005;161:81–88. doi: 10.1093/aje/kwi017 [DOI] [PubMed] [Google Scholar]
- 17.Schrage B, Ibrahim K, Loehn T, Werner N, Sinning JM, Pappalardo F, Pieri M, Skurk C, Lauten A, Landmesser U, et al. Impella support for acute myocardial infarction complicated by cardiogenic shock. Circulation. 2019;139:1249–1258. doi: 10.1161/CIRCULATIONAHA.118.036614 [DOI] [PubMed] [Google Scholar]
- 18.Ouweneel DM, Eriksen E, Sjauw KD, van Dongen IM, Hirsch A, Packer EJ, Vis MM, Wykrzykowska JJ, Koch KT, Baan J, et al. Percutaneous mechanical circulatory support versus intra-aortic balloon pump in cardiogenic shock after acute myocardial infarction. J Am Coll Cardiol. 2017;69:278–287. doi: 10.1016/j.jacc.2016.10.022 [DOI] [PubMed] [Google Scholar]
- 19.Basir MB, Kapur NK, Patel K, Salam MA, Schreiber T, Kaki A, Hanson I, Almany S, Timmis S, Dixon S, et al. ; National Cardiogenic Shock Initiative Investigators. Improved outcomes associated with the use of shock protocols: updates from the national cardiogenic shock initiative. Catheter Cardiovasc Interv. 2019;93:1173–1183. doi: 10.1002/ccd.28307 [DOI] [PubMed] [Google Scholar]
- 20.Spertus JA, Furman MI. Translating evidence into practice: are we neglecting the neediest? Arch Intern Med. 2007;167:987–988. doi: 10.1001/archinte.167.10.987 [DOI] [PubMed] [Google Scholar]
- 21.Abiomed: Our Story. Accessed June 11, 2022. https://www.abiomed.com/about-us/our-story.
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.