Outcomes for transcatheter pulmonary valve replacement (TC-PVR) and operative pulmonary valve replacement (S-PVR) are excellent. Thus, their respective cost is a relevant clinical outcome. We performed a retrospective cohort study of children and adults who underwent PVR at age ≥8 years from January 1, 2011, to December 31, 2013, at 35 centers contributing data to the Pediatric Health Information Systems database to address this question. A propensity score–adjusted multivariable analysis was performed to adjust for known confounders. Secondary analyses of department-level charges, risk of re-admission, and associated costs were performed. A total of 2,108 PVR procedures were performed in 2,096 subjects (14% transcatheter and 86% operative). The observed cost of S-PVR and TC-PVR was not significantly different (2013US $50,030 vs 2013US $51,297; p = 0.85). In multivariate analysis, total costs of S-PVR and TC-PVR were not significantly different (p = 0.52). Length of stay was shorter after TC-PVR (p <0.0001). Clinical and supply charges were greater for TC-PVR (p <0.0001), whereas laboratory, pharmacy, and other charges (all p <0.0001) were greater for S-PVR. Risks of both 7- and 30-day readmission were not significantly different. In conclusion, short-term costs of TC-PVR and S-PVR are not significantly different after adjustment.
Right ventricular outflow tract reconstruction frequently results in stenosis and/or insufficiency. This may not cause symptoms for years but has been implicated in insidious development of RV dysfunction and increased risk of mortality. Operative replacement of the pulmonary valve (S-PVR) has previously been the single definitive treatment for dysfunctional right ventricular outflow tract reconstruction. Transcatheter pulmonary valve replacement (TC-PVR) is an alternative to S-PVR with excellent safety and efficacy, which potentially avoids the morbidity and mortality associated with open-heart surgery. With clinical equipoise, the economic costs of TC-PVR and S-PVR become relevant as outcomes. Although several studies have assessed the costs of TC-PVR and S-PVR, small volumes and homogenous practice at single centers make a traditional cohort study challenging. Using administrative data from the Pediatric Health Information System (PHIS) database, we performed a multicenter, retrospective cohort study to overcome these obstacles.
Methods
The PHIS database is an administrative database containing data from inpatient, emergency department, ambulatory surgery, and observation encounters from 42 not-for-profit, tertiary care pediatric hospitals in the United States, affiliated with the Children’s Hospital Association (CHA) (Overland Park, Kansas). Data quality and reliability are assured through a joint effort between the CHA and participating hospitals. The data warehouse function for PHIS is managed by Truven Health Analytics (Ann Arbor, Michigan). For the purposes of external benchmarking, participating hospitals provide discharge/encounter data including demographics, diagnoses, and procedures. Forty-two of these hospitals also submit resource utilization data (e.g., pharmacy products, radiologic studies, and laboratory studies) to PHIS. Data are de-identified at the time of data submission and are subjected to a number of reliability and validity checks. A data-use agreement was signed between study investigators and CHA. The institutional review board of The Children’s Hospital of Philadelphia reviewed the proposed project and determined that it did not represent human subjects research in accordance with the Common Rule (45 CFR 46.102(f)).
We included children and adults of age ≥8 years who underwent pulmonary valve replacement at any of the 43 PHIS centers from January 1, 2011, to December 31, 2013. The age restriction was added to eliminate relatively young subjects who would not be considered candidates for TC-PVR as described previously. Subjects were identified by the International Classification of Disease, Ninth Revision ( ICD-9 ) code and divided between those who underwent (1) transcatheter valve pulmonary valve replacement ( ICD-9 : 35.07) or (2) open-heart surgery for pulmonary valve replacement ( ICD-9 : 35.25 or 35.26). To restrict analysis to centers with stable reporting and procedural volumes, we also excluded subjects from centers reporting (1) ≤25 cardiac catheterization or cardiac operative procedures per year over the study period or (2) failure to report catheterizations and cardiac operations 2 of 3 years of the study period.
Data were extracted from the PHIS database using ICD-9 codes for diagnoses and procedures. Medical co-morbidities were divided by system. Several steps were taken to generate costs that were comparable across the study population. Costs were extracted from PHIS. The PHIS database calculates cost by converting charges digitally transferred from individual hospital’s billing records using ratios of costs to charges (RCC). The PHIS registry also adjusts costs to account for regional differences (based on wage/price indexes). The results of this conversion are referred to in this report as “adjusted costs.” We also accounted for inflation by converting all costs to 2013 US dollars (2013US$) using the consumer price index for medical care ( http://data.bls.gov/cgi-bin/dsrv ). PHIS also reports cost without adjustment for local wage/price indexes (heretofore referred to as “unadjusted costs”). To be complete, we repeated all analyses with unadjusted costs ( Supplementary Tables 1 and 2 ).
The primary exposure was method of PVR, dividing the study population into TC-PVR and S-PVR cohorts. TC-PVR and S-PVR were not both performed in a single admission (i.e., no crossover was recorded). Because cohorts were identified by ICD-9 codes, there is the possibility of subjects who underwent catheterization and failed to receive TC-PVR and subsequently underwent S-PVR would be included in the S-PVR cohort. The rate at which this occurred could not be measured. Therefore, analysis was performed as treated. Continuous variables were expressed as mean ± SD or median (range and interquartile range [IQR]) as appropriate. Categorical variables were expressed as percent (count).
Before analysis, we suspected that subject characteristics would differ systematically between transcatheter and operative cohorts and also influence cost. Wilcoxon rank-sum test, chi-square test, and Fisher’s exact test were used to test for differences in the distribution of subject characteristics. To adjust for confounding by indication, a propensity score for each subject defining the relative probability of assignment to TC-PVR versus S-PVR was calculated. The factors included in the propensity score were gender, noncardiac congenital anomaly, known genetic syndrome, age (centered on mean), and each of the following: hematologic disorder, malignancy, metabolic disorder, neurologic disorder, or pulmonary disorder. The propensity score was then included in subsequent models as a covariate.
Analysis of cost included direct medical costs (determined from hospital charges and professional charges) with the time horizon of analysis restricted to hospitalizations during which valve replacement was performed. Preprocedural costs were assumed to be equal between both cohorts. Median costs (total and according to categories) were calculated for both cohorts and compared using Wilcoxon rank-sum tests. Subsequently generalized linear models were used to account for measurable covariates (including propensity score as a covariate) and compare costs between the 2 cohorts. A log gamma distribution was used to account for the skewed distribution of costs. Length of stay (LOS) was analyzed similarly, but because it includes zero as a possible value, a Gaussian probability distribution was used. For all models, conditional standardization was used to generate an adjusted estimate of the outcome of interest. An analysis adding a random intercept for center to adjust for correlation in centers was performed, without any change from reported results (data not shown).
An analysis of differences in department-level charges for S-PVR and TC-PVR was performed using the Wilcoxon rank-sum test. Two sensitivity analyses were performed (1) varying the minimum age of subjects (1-year increment from 8 to 18 years) to ensure that this did not bias results and (2) restricting the analysis to hospitals that performed both S-PVR and TC-PVR (eliminating 6 of 36 centers and 8% [172 of 2,133] of total procedures). The results of main effects did not change in either analysis (data not shown).
As a secondary analysis, we studied the distribution of costs between hospital departments to identify aspects of care that could be optimized to improve value. PHIS does not release department-level ratios of costs to charges or costs, so only department-level hospital charges could be analyzed. These charges were inflation adjusted, but no further normalization was possible. This is a significant limitation of the data source. Charges are strongly influenced by individual hospital charging practices and are imperfect surrogates for cost, but they provide a measure of the relative magnitude of resource utilization that influence cost for each method.
A third analysis was performed assessing the relative risk of readmission within 7 and 30 days and the cost of these admissions. Other clinical outcomes (inhospital death and LOS) were also collected and compared between the TC-PVR and S-PVR.
There were no patients with missing data about their age or gender. For patients with missing data about their race or cardiac diagnosis, a separate categorical variable was generated; this allowed them to be included in the analysis and mitigated bias. A sensitivity analysis that did not include cardiac diagnosis (for which there was a high rate of missing data) was performed with no change in coefficients for cost or significance (data not shown). No formal measures were applied to account for multiple comparisons. The primary analysis is specified, and other analyses should be considered exploratory. All data analyses were performed using Stata MP 13 (StataCorp, College Station, Texas). The threshold for statistical significance was p <0.05.
Results
From January 1, 2011, to December 31, 2013, 2,294 procedures were performed in 2,274 eligible subjects at the 42 centers included in the PHIS database ( Figure 1 ). Data from 6 centers accounting for 185 procedures in 178 subjects were excluded because they failed to meet inclusion criteria. Multiple PVR procedures were performed in a single hospitalization for 2 subjects. For the remaining 2,108 hospitalizations in 2,096 subjects, the median age of subject at PVR was 16.3 years, 39% women, and 73% white.
The cohort was 86% S-PVR (1,816 of 2,108 procedures in 1,806 of 2,096 subjects) and 14% TC-PVR (292 of 2,108 procedures in 290 of 2,096 subjects). Several significant differences in demographics were seen between the S-PVR and TC-PVR cohorts. The operative cohort was younger (p <0.0001) with a different distribution of cardiac diagnoses (p <0.001). The distribution of other patient factors was not significantly different between the cohorts ( Table 1 ).
| Operative | Trans-catheter | p | |
|---|---|---|---|
| Procedures | 1,816 | 292 | n/a |
| Individuals | 1,806 | 290 | n/a |
| Age (years) | 16.2 (IQR: 12.5-21.3) | 17.7 (IQR: 13.1-24.8) | <0.0001 |
| Female sex | 39% (703) | 38% (112) | 0.96 |
| Race | |||
| White | 73% (1,334) | 72% (209) | 0.70 |
| Black | 8% (153) | 10% (28) | |
| Asian | 4% (66) | 5% (15) | |
| Other | 11% (198) | 10% (29) | |
| Missing | 4% (65) | 4% (11) | |
| Payor | |||
| Private | 51% (935) | 44% (130) | <0.001 |
| Public | 33% (596) | 33% (96) | |
| Other | 14% (252) | 23% (66) | |
| Missing | 2% (33) | 0.3% (1) | |
| Primary cardiac diagnosis | |||
| Tetralogy of Fallot | 14% (255) | 14% (42) | <0.001 |
| Transposition of the great arteries | 0% (0) | 0.7% (2) | |
| Truncus arteriosus | 0.2% (3) | 2% (7) | |
| Double outlet right ventricle | 4% (64) | 2% (5) | |
| Other pulmonary valve disease | 1% (18) | 0% (0) | |
| Other | 30% (549) | 11% (33) | |
| Missing | 51% (927) | 70% (203) | |
| Genetic syndrome | |||
| None recorded | 97.0% (1,762) | 97.3% (284) | 0.41 |
| 22q11.2 microdeletion | 1.5% (28) | 0.7% (2) | |
| Trisomy 21 | 1.3% (23) | 2.1% (6) | |
| Non-cardiac congenital anomaly | 3% (60) | 2% (7) | 0.41 |
| History of gastrointestinal condition | 0.7% (24) | 1.4% (4) | 0.20 |
| History of hematologic condition | 5.0% (91) | 7.5% (22) | 0.08 |
| History of malignancy | 0.4% (7) | 0.3% (1) | 0.91 |
| History of metabolic condition | 3.3% (60) | 2.1% (6) | 0.25 |
| History of neurologic condition | 3.6% (66) | 4.8% (14) | 0.34 |
| History of renal condition | 1.2% (22) | 0.3% (1) | 0.18 |
| History of respiratory condition | 0.4% (8) | 0.3% (1) | 0.81 |
The observed total costs of S-PVR (2013US$50,030, IQR 38,639 to 65,932) and TC-PVR (2013US$51,297, IQR 37,858 to 73,328, p = 0.85; Table 2 ) were not significantly different. Observed inhospital mortality was not significantly different after S-PVR than TC-PVR (p = 0.24). LOS was longer after S-PVR than TC-PVR (p <0.0001). In terms of department-level charges, clinical charges and supply charges were greater in the TC-PVR cohort (p <0.0001 for both). Other charges (including room charges) and laboratory and pharmacy charges were all greater for the S-PVR cohort (p <0.0001 for all). Imaging costs were not significantly different between the 2 cohorts (p = 0.27).
| Operative | Transcatheter | p | |
|---|---|---|---|
| Total cost (2013US$) | 50,030 (IQR: 38,639-65,932) | 51,297 (IQR: 37,858-72,328) | 0.85 |
| Hospital length of stay (days) | 4 (Range: 1-198, IQR: 3-5.5) | 1 (Range: 0-49, IQR: 1-1) | <0.0001 |
| In-hospital mortality | 0.8% (36) ∗ | 0% (0) ∗ | 0.24 † |
| Department level charges (2013$US) | |||
| Clinical | 7,316 (IQR: 3,602-23,820) | 23,753 (IQR: 12,805-30,727) | <0.0001 |
| Imaging | 7,475 (IQR: 4,853-11,808) | 6,578 (IQR: 2,966-17,737) | 0.27 |
| Lab | 12,661 (IQR: 6,619-20,303) | 1,502 (IQR: 692-2,941) | <0.0001 |
| Pharmacy | 10,329 (IQR: 5,797-18,768) | 1,724 (IQR: 1,018-3,901) | <0.0001 |
| Supply | 33,127 (IQR: 20,796-48,234) | 83,288 (IQR: 51,992-116,246) | <0.0001 |
| Other | 51,943 (IQR: 32,651-73,870) | 10,383 (IQR: 6,999-14,776) | <0.0001 |
| 7-day readmission | 8.2% (163) | 6.4% (20) | 0.26 |
| 30-day readmission | 18.1% (401) | 15.1% (52) | 0.17 |
| With operation | 0% (0) | 0.3% (1) | 0.13 |
| With catheterization | 0.2% (3) | 0.7% (2) | 0.14 |
∗ 95% CI for S-PVR is 0.4% to 1.2%. One-sided 97.5% CI for TC-PVR is 0% to 1.3%.
In a propensity score–adjusted multivariate model, there was no significant difference in total cost between S-PVR and TC-VR (p = 0.52, Table 3 ). Using conditional standardization, cost of S-PVR was US2013$47,635 (95% confidence interval [CI] 27,537 to 82,404) versus cost of TC-PVR, which was 2013US$51,409 (95% CI 28,978 to 91,201). A history of malignancy (coefficient 2.29, 95% CI 1.26 to 4.18, p = 0.007) was associated with increased cost. Previous pulmonary (coefficient 2.01, p = 0.10), gastrointestinal (coefficient 1.43, p = 0.06), and hematologic (coefficient 1.30, p = 0.07) conditions had coefficients consistent with higher cost, but the associations were not statistically significant. History of 22q11.2 microdeletion syndrome (coefficient 0.50, p = 0.05) was suggestive of decreased cost, but this association was not statistically significant.
