Abstract
Background
To compare catheter directed thrombolysis (CDT) and endovascular mechanical thrombectomy (MT) for acute intermediate-risk and high-risk pulmonary embolism (PE) with respect to unplanned 6-month readmissions and in-hospital outcomes using a national database.
Methods
The NRD was queried from 2016 to 2022 for adult patients with acute nonseptic PE who received either CDT or MT. Inverse probability weighted regression adjustment was used to compare CDT and MT with regard to in-hospital mortality, unplanned 6-month readmissions, discharge other than home (DOTH), total cost as well as gastrointestinal bleeding (GIB), intracranial hemorrhage (ICH), blood transfusion at index admission, and 6-month readmission. Results are depicted as average treatment effect (ATE).
Results
A total of 28,376 patients were identified; 18,168 (64%) underwent CDT and 10,208 (36%) underwent MT. There was a total of 3,945 patients with high risk PE (13.9%). Compared to CDT, the ATE of MT was higher with regards to in-hospital mortality (ATE: 0.007; p = 0.048), DOTH (ATE: 0.031; p < 0.001), ICH (ATE: 0.006; p < 0.001), total cost at index admission (ATE: 0.118; p < 0.001), GIB at 6 months (ATE: 0.004; p = 0.023), blood transfusions at 6 months (ATE: 0.006; p = 0.017) as well as death at final follow-up (ATE: 0.009; p = 0.025). The ATE was similar between CDT and MT with regard to GIB (p = 0.097) at index admission as well as unplanned 6-month readmission (p = 0.10), ICH (p = 0.11), and recurrent PE (p = 0.39) at 6 months.
Conclusion
MT was associated with higher rates of mortality, DOTH, ICH, and total cost at index admission as well as GIB, blood transfusions, and mortality at 6 months. MT was associated with similar rates of GIB at index admission, as well as unplanned readmission, ICH, and recurrent PE at 6 months.
Supplementary Information
The online version contains supplementary material available at 10.1186/s42155-026-00752-1.
Keywords: Pulmonary embolism, Thrombectomy, Catheter-directed thrombolysis, ICH, Unplanned readmissions, Comparison
Introduction
Acute pulmonary embolism (PE) is a major cause of morbidity and mortality and the third leading cause of cardiovascular death [1]. Catheter directed thrombolysis (CDT) was utilized to minimize bleeding complications through localized delivery of thrombolytics using lower dosages of thrombolytics over a prolonged duration. CDT is currently used primarily in patients with intermediate-high and high-risk pulmonary embolism [2]. These therapies were not developed to replace systemic thrombolysis but rather to provide alternative reperfusion strategies using localized thrombolytic delivery or large-bore aspiration systems adapted from technologies long used in other vascular territories.
Systemic thrombolysis and CDT have historically been used to improve RV dysfunction compared to anticoagulation alone [3–5]. Mechanical thrombectomy was introduced as an endovascular option to avoid thrombolytics and may be considered in patients with shock, failed thrombolysis, or bleeding risk [6].
Despite increased adoption of mechanical thrombectomy for both intermediate- and high-risk PE, data directly comparing mechanical thrombectomy to CDT with respect to clinical outcomes and adverse events have been limited. Several retrospective cohort studies, the PEERLESS randomized controlled trial, and subsequent real-world analyses have addressed this comparison [7–11]. Concurrently, the HI-PEITHO trial and others have provided new evidence for catheter-based reperfusion strategies against anticoagulation alone [12]. The primary objective of this study was to utilize the Nationwide Readmissions Database (NRD) to compare CDT and mechanical thrombectomy for acute high- and intermediate-risk PE with respect to unplanned 6-month readmissions and outcomes.
Methods
Institutional review board statement
Institutional review board approval was not required for this study because it only contains deidentified patient data from a national public database.
Data source
The NRD database records about 35 million discharges annually, encompassing all types of payers and uninsured patients, from over 2000 hospitals across 30 states. It analyzes readmission trends with both clinical and nonclinical factors, accounting for approximately 60% of all hospitalizations in the USA. Moreover, the NRD includes sampling weights that allow for the estimation of national rates. Further information about the NRD can be found at https://www.hcup-us.ahrq.gov/db/nation/nrd/nrddbdocumentation.jsp [13].
Study cohort
All adult patients with a diagnosis of acute, nonseptic PE who underwent CDT or thrombectomy were identified using International Classification of Diseases, 10th Edition (ICD-10), codes. Admissions involving septic PEs were excluded using corresponding ICD codes. Additionally, patients with acute ST-elevation myocardial infarction and ischemic stroke were excluded to limit the misclassification of indication for thrombolytic use in addition to pre-procedural intracranial hemorrhage (ICH) and hemorrhagic conversion. ICD codes for shock, vasopressor usage and mechanical ventilation as well as saddle PE were included to distinguish between high risk PE from low risk which was excluded. This classification was described in previous studies [14–17]. The final cohort was split into CDT and mechanical thrombectomy (MT) based on the ICD-10 procedural codes. Patients that received intravenous thrombolytics (IVT) in combination of these procedures were excluded from our cohort. Patients with procedure codes for both CDT and MT during the index admission were excluded to maintain mutually exclusive treatment groups (n = 790 unweighted; weighted national estimate 1436). All codes are listed in Supplementary Table 1.
Covariates of interest
Several demographic variables, such as age, sex, insurance status, and quartile of median household income, were gathered, alongside information on hospital characteristics such as teaching status and size. Using corresponding ICD codes, Patients with an inferior vena cava (IVC) filter were also identified. Hemodynamic shock, mechanical ventilation, and vasopressor use were used to classify high-risk PE based on prior definitions [14–16]. To assess the risks of unfavorable outcomes, the Elixhauser comorbidity indices were utilized, which consider 31 predefined comorbidities and make use of ICD-9 and ICD-10 codes [18]. In addition, the length of stay, days until readmission, and total cost for the index admission (in US dollars adjusted for inflation in 2024) were among the other variables collected. Hospital charges were converted to cost estimated using NRD-provided cost-to-charge ratios and standardized to 2024 U.S. dollars using inflation adjustment.
Outcomes of interest
Three variables were collected based on diagnosis at discharge during the index admission and readmissions using the ICD codes: intracranial hemorrhage (ICH), gastrointestinal bleeding (GIB), and blood transfusion (BT) and the presence of recurrent PE. Discharges other than home (DOTH), including discharges to short-term care, skilled nursing facility, and home health care, were also captured for the index admissions. Cardiovascular-related unplanned readmissions were derived by selecting readmissions that were nonelective. Of note, readmission outcomes do not include the outcomes from the index admissions.
Statistical analysis
The cohort was divided into CDT and MT groups. Categorical variables were compared using χ2 or Fisher exact tests, and continuous variables using t tests. Multivariable hierarchical regression adjusted for demographics, hospital characteristics, and IVC filter status was performed to evaluate associations between treatment modality and index outcomes (in-hospital mortality, DOTH, GIB, ICH, BT). Negative binomial regression was used for cost analysis. 7-day, 1-month and 6-month outcomes included unplanned readmission, GIB, ICH, BT, recurrent PE, and mortality, reported as odds ratios (ORs) with 95% confidence intervals (CIs).
Inverse probability weighted regression adjustment (IPWRA) was the primary analytic approach, estimating average treatment effects (ATE). Treatment and outcome models were adjusted for age, sex, insurance, income quartile, massive PE, IVC filter, and Elixhauser Comorbidity Index. Covariate balance before and after weighting was assessed using standardized mean differences (SMDs) and variance ratios. Adequate balance was defined as SMD < 0.1 and variance ratios near 1. This method reduces bias from measured confounders in observational studies [19]. Covariates were selected based on clinical relevance, prior literature, and univariate associations. Variables were removed only in cases of significant multicollinearity or model instability [17, 20]. Subgroup analysis for intermediate-risk PEs was completed for index admission and 6-month readmission using IPWRA. The primary analytic approach was IPWRA estimating average treatment effects (ATE), while the multivariable regression analysis was secondary. To aid interpretation, adjusted odds ratios with 95% confidence intervals were derived from the IPWRA potential-outcome means for each binary outcome and are reported alongside the average treatment effects in Table 2. All analyses were conducted using population weights provided by the NRD with “DISCWT” variable. p values of < 0.05 were considered statistically significant. Statistical analysis was performed using STATA 17 (StataCorp, College Station, Texas).
Table 2.
Adjusted odds ratios and average treatment effects of mechanical thrombectomy compared with CDT for outcomes in the index admission and readmission at 7 days, 30 days, and 6 months
| Outcome | Adjusted OR (95% CI) | Average treatment effect | 95% lower bound | 95% upper bound | p-value |
|---|---|---|---|---|---|
| Index admission | |||||
| In-hospital mortality | 1.20 (1.00–1.41) | 0.007 | 0.0001 | 0.014 | 0.048 |
| Discharge other than home | 1.17 (1.09–1.26) | 0.031 | 0.016 | 0.046 | < 0.001 |
| GI bleed | 1.21 (0.95–1.52) | 0.004 | − 0.001 | 0.010 | 0.097 |
| ICH | 2.20 (1.60–3.00) | 0.006 | 0.003 | 0.010 | < 0.001 |
| Blood transfusion | 1.55 (1.32–1.78) | 0.017 | 0.010 | 0.024 | < 0.001 |
| Total cost1 | Mean $34,788 (CDT) vs. $41,884 (MT); median $28,827 vs. $32,615 | 0.118 | 0.098 | 0.137 | < 0.001 |
| 7-day readmission | |||||
| Unplanned readmission | 0.90 (0.68–1.11) | − 0.003 | − 0.009 | 0.003 | 0.355 |
| GI bleed | 1.09 (0.09–2.09)2 | 0.002 | − 0.002 | 0.007 | 0.354 |
| ICH | 1.90 (0.20–3.61)2 | 0.001 | − 0.001 | 0.003 | 0.293 |
| PE | 1.16 (0.83–1.49) | 0.000 | − 0.001 | 0.001 | 0.871 |
| Blood transfusion | 1.57 (0.50–2.65)2 | 0.001 | − 0.001 | 0.002 | 0.285 |
| Death | 1.13 (0.90–1.36) | − 0.003 | − 0.009 | 0.003 | 0.355 |
| 30-day readmission | |||||
| Unplanned readmission | 1.10 (0.95–1.26) | 0.006 | − 0.003 | 0.015 | 0.21 |
| GI bleed | 1.25 (0.61–1.93) | 0.003 | − 0.003 | 0.009 | 0.32 |
| ICH | 1.42 (0.42–2.43)2 | 0.001 | − 0.001 | 0.003 | 0.42 |
| PE | 1.12 (0.88–1.37) | 0.001 | − 0.001 | 0.003 | 0.44 |
| Blood transfusion | 2.98 (1.93–4.03) | 0.005 | 0.002 | 0.008 | < 0.001 |
| Death | 1.21 (1.01–1.41) | 0.008 | 0.001 | 0.015 | 0.036 |
| 6-month readmission | |||||
| Unplanned readmission | 1.09 (0.98–1.20) | 0.011 | − 0.002 | 0.024 | 0.09 |
| GI bleed | 1.59 (1.15–2.04) | 0.004 | 0.001 | 0.007 | 0.023 |
| ICH | 1.72 (0.86–2.44) | 0.002 | − 0.0004 | 0.004 | 0.11 |
| PE | 1.07 (0.90–1.27) | 0.003 | − 0.004 | 0.011 | 0.39 |
| Blood transfusion | 1.53 (1.09–1.97) | 0.006 | 0.001 | 0.011 | 0.017 |
| Death at final follow-up | 1.23 (1.02–1.40) | 0.009 | 0.001 | 0.016 | 0.025 |
Adjusted odds ratios were derived from the inverse-probability-weighted regression adjustment (IPWRA) potential-outcome means for mechanical thrombectomy versus CDT, with the corresponding average treatment effect (risk difference) shown alongside
1Total cost is reported as survey-weighted absolute mean and median values; the average treatment effect is on the natural-log cost scale, corresponding to approximately 12.5% higher cost with mechanical thrombectomy. Odds ratios are not reported for total cost (continuous). Readmission-linked outcomes were evaluated in 14,104 records with valid readmission linkage; in-hospital mortality reflects the full cohort (14,687)
2Odds ratio based on fewer than 30 events; the estimate and confidence interval are imprecise and should be interpreted with caution
Results
Patient characteristics
A total of 28,376 patients were identified; 18,168 (64%) underwent CDT and 10,208 (36%) underwent MT. There was a total of 3945 patients with high risk PE (13.9%) (Table 1). After inverse probability weighting, covariates were well balanced (SMD < 0.10; Supplementary Table 2).
Table 1.
Patient demographics, in n (%) or mean (SD)
| Variables | Treatment type | |||
|---|---|---|---|---|
| CDT 18,168 (64.0%) | Thrombectomy alone 10,208 (36.0%) | Total 28,376 (100.0%) | p-value | |
| Demographics | ||||
| Age in years at admission | 60.01 (15.19) | 61.40 (14.95) | 60.51 (15.12) | < 0.001 |
| Number of readmissions | 0.22 (0.60) | 0.24 (0.61) | 0.23 (0.60) | 0.104 |
| Final follow-up time, in days | 16.76 (34.69) | 17.83 (35.90) | 17.15 (35.13) | 0.093 |
| Gender | 0.165 | |||
| Male | 9532 (52.47%) | 5492 (53.80%) | 15,024 (52.95%) | |
| Female | 8636 (47.53%) | 4716 (46.20%) | 13,352 (47.05%) | |
| Primary expected payer | 0.115 | |||
| Medicare | 8100 (44.66%) | 4813 (47.23%) | 12,913 (45.58%) | |
| Medicare | 1698 (9.36%) | 963 (9.45%) | 2660 (9.39%) | |
| Private insurance | 6849 (37.76%) | 3605 (35.38%) | 10,454 (36.90%) | |
| Self-pay | 713 (3.93%) | 412 (4.05%) | 1126 (3.97%) | |
| No charge | 105 (0.58%) | 48 (0.47%) | 153 (0.54%) | |
| Other | 673 (3.71%) | 349 (3.43%) | 1023 (3.61%) | |
| Bed size of hospital | 0.027 | |||
| Small | 2387 (13.14%) | 1181 (11.57%) | 3568 (12.57%) | |
| Medium | 4965 (27.33%) | 2481 (24.31%) | 7446 (26.24%) | |
| Large | 10,816 (59.54%) | 6545 (64.12%) | 17,362 (61.19%) | |
| Teaching status of urban hospitals | 0.13 | |||
| Urban Nonteaching | 3460 (19.05%) | 1584 (15.51%) | 5044 (17.78%) | |
| Urban Teaching | 13,752 (75.69%) | 8188 (80.22%) | 21,940 (77.32%) | |
| Rural | 956 (5.26%) | 436 (4.27%) | 1392 (4.90%) | |
| Median household income national quartile for patient ZIP Code | 0.053 | |||
| 0–25 percentile | 5194 (28.87%) | 3183 (31.52%) | 8377 (29.82%) | |
| 26–50 percentile | 5070 (28.18%) | 2913 (28.85%) | 7983 (28.42%) | |
| 51–75 percentile | 4556 (25.32%) | 2347 (23.24%) | 6903 (24.58%) | |
| > 75 percentile | 3172 (17.63%) | 1655 (16.39%) | 4827 (17.18%) | |
| IVC filter | 2572 (14.16%) | 1486 (14.55%) | 4058 (14.30%) | 0.683 |
| Massive PE | 2232 (12.29%) | 1713 (16.79%) | 3945 (13.90%) | < 0.001 |
| Mechanical ventilation | 1614 (8.88%) | 1139 (11.16%) | 2753 (9.70%) | < 0.001 |
| Vasopressor | 226 (1.25%) | 236 (2.31%) | 462 (1.63%) | < 0.001 |
| Non-septic shock | 1060 (5.84%) | 967 (9.47%) | 2027 (7.14%) | < 0.001 |
| Elixhauser comorbidities | ||||
| Elixhauser Scaled Score (AHRQ Algorithm, Readmission) | 19.15 (14.79) | 20.74 (15.67) | 19.72 (15.13) | < 0.001 |
| Congestive heart failure | 3145 (17.31%) | 1877 (18.38%) | 5021 (17.70%) | 0.140 |
| Cardiac arrhythmias | 4472 (24.61%) | 2583 (25.30%) | 7054 (24.86%) | 0.394 |
| Valvular disease | 1419 (7.81%) | 681 (6.67%) | 2099 (7.40%) | 0.028 |
| Peripheral vascular disorders | 1108 (6.10%) | 503 (4.93%) | 1611 (5.68%) | 0.004 |
| Paralysis | 130 (0.72%) | 169 (1.66%) | 300 (1.06%) | < 0.001 |
| Other neurological disorders | 1257 (6.92%) | 969 (9.50%) | 2227 (7.85%) | < 0.001 |
| Chronic pulmonary disease | 3699 (20.36%) | 1880 (18.42%) | 5579 (19.66%) | 0.015 |
| Diabetes, uncomplicated | 2456 (13.52%) | 1338 (13.11%) | 3794 (13.37%) | 0.518 |
| Diabetes, complicated | 2702 (14.87%) | 1542 (15.10%) | 4243 (14.95%) | 0.742 |
| Hypothyroidism | 2180 (12.00%) | 1211 (11.86%) | 3391 (11.95%) | 0.829 |
| Renal failure | 1990 (10.95%) | 1145 (11.22%) | 3135 (11.05%) | 0.690 |
| Liver disease | 865 (4.76%) | 589 (5.77%) | 1454 (5.12%) | 0.010 |
| Peptic ulcer disease excluding bleeding | 52 (0.29%) | 54 (0.53%) | 107 (0.38%) | 0.032 |
| AIDS/HIV | 38 (0.21%) | 19 (0.18%) | 57 (0.20%) | 0.750 |
| Lymphoma | 122 (0.67%) | 95 (0.93%) | 217 (0.77%) | 0.100 |
| Metastatic cancer | 460 (2.53%) | 449 (4.40%) | 909 (3.20%) | < 0.001 |
| Solid tumor without metastasis | 890 (4.90%) | 850 (8.33%) | 1741 (6.13%) | < 0.001 |
| Rheumatoid arthritis/collagen vascular | 564 (3.10%) | 291 (2.85%) | 855 (3.01%) | 0.394 |
| Coagulopathy | 3175 (17.48%) | 1634 (16.01%) | 4809 (16.95%) | 0.039 |
| Obesity | 7595 (41.81%) | 4050 (39.67%) | 11,645 (41.04%) | 0.035 |
| Weight loss | 597 (3.28%) | 460 (4.51%) | 1057 (3.72%) | < 0.001 |
| Fluid and electrolyte disorders | 5340 (29.39%) | 3327 (32.59%) | 8667 (30.54%) | < 0.001 |
| Blood loss anemia | 194 (1.07%) | 107 (1.05%) | 301 (1.06%) | 0.927 |
| Deficiency anemia | 871 (4.79%) | 583 (5.71%) | 1454 (5.12%) | 0.019 |
| Alcohol abuse | 783 (4.31%) | 345 (3.38%) | 1128 (3.97%) | 0.008 |
| Drug abuse | 582 (3.21%) | 311 (3.04%) | 893 (3.15%) | 0.610 |
| Psychoses | 189 (1.04%) | 99 (0.97%) | 288 (1.01%) | 0.695 |
| Depression | 2268 (12.48%) | 1298 (12.72%) | 3567 (12.57%) | 0.708 |
| Index hospitalization outcomes | ||||
| Length of stay, in days | 5.88 (6.42) | 6.39 (8.73) | 6.06 (7.34) | 0.001 |
| Total charge (in 2024 dollars) | 34,787.89 (27,233) | 41,884.06 (39,936.06) | 37,339.52 (32,556.97) | < 0.001 |
| Abdominal bleeding | 346 (1.91%) | 258 (2.53%) | 605 (2.13%) | 0.015 |
| Intracranial hemorrhage | 89 (0.49%) | 121 (1.19%) | 210 (0.74%) | < 0.001 |
| Discharge type | ||||
| Routine | 13,158 (75.40%) | 6836 (70.43%) | 19,994 (73.62%) | < 0.001 |
| Short-term hospital | 113 (0.65%) | 63 (0.65%) | 176 (0.65%) | |
| Skilled nursing/intermediate care facility | 1700 (9.74%) | 1165 (12.00%) | 2865 (10.55%) | |
| Home health care (HHC) | 2480 (14.21%) | 1643 (16.93%) | 4123 (15.18%) | |
| Discharge other than home | 4293 (24.60%) | 2871 (29.57%) | 7164 (26.38%) | < 0.001 |
| Died during hospitalization | 635 (3.50%) | 464 (4.54%) | 1,099 (3.87%) | 0.005 |
| 1-week readmission outcomes | ||||
| Unplanned readmission | 502 (2.86%) | 258 (2.65%) | 760 (2.79%) | 0.487 |
| PE | 245 (1.40%) | 157 (1.61%) | 402 (1.47%) | 0.318 |
| ICH | 16 (0.09%) | 19 (0.19%) | 35 (0.13%) | 0.195 |
| GI bleed | 18 (0.10%) | 12 (0.12%) | 30 (0.11%) | 0.793 |
| Blood transfusion | 24 (0.14%) | 23 (0.23%) | 47 (0.17%) | 0.209 |
| Death | 502 (2.76%) | 335 (3.29%) | 838 (2.95%) | 0.107 |
| 1-month readmission outcomes | ||||
| Unplanned readmission | 1,094 (6.24%) | 698 (7.16%) | 1,792 (6.57%) | 0.058 |
| PE | 452 (2.58%) | 288 (2.95%) | 740 (2.71%) | 0.227 |
| ICH | 32 (0.19%) | 27 (0.28%) | 60 (0.22%) | 0.301 |
| GI bleed | 47 (0.27%) | 36 (0.37%) | 83 (0.30%) | 0.274 |
| Blood transfusion | 44 (0.25%) | 79 (0.81%) | 122 (0.45%) | < 0.001 |
| Death | 675 (3.72%) | 498 (4.88%) | 1173 (4.13%) | 0.002 |
| 6-month outcomes | ||||
| Unplanned readmission | 2453 (14.00%) | 1544 (15.85%) | 3997 (14.66%) | 0.006 |
| All-cause readmission | 2791 (15.92%) | 1673 (17.17%) | 4463 (16.37%) | 0.073 |
| Pulmonary embolism | 739 (4.21%) | 455 (4.67%) | 1194 (4.38%) | 0.238 |
| ICH | 47 (0.27%) | 48 (0.49%) | 95 (0.35%) | 0.049 |
| GI bleed | 115 (0.66%) | 110 (1.13%) | 225 (0.83%) | 0.004 |
| Death at final follow-up | 737 (4.06%) | 546 (5.35%) | 1283 (4.52%) | < 0.001 |
Multivariable regression
At index admission, MT was associated with higher odds of DOTH (OR 1.15; 95% CI 1.04–1.27; p = 0.006), ICH (OR 1.84; 95% CI 1.20–2.83; p = 0.005), and BT (OR 1.51; 95% CI 1.24–1.82; p < 0.001). MT was also associated with 13% higher cost (Coef. 0.12; 95% CI 0.09–0.15; p < 0.001) (Supplementary Table 3).
At 6 months, MT was associated with higher odds of GIB (OR 1.55; 95% CI 1.06–2.28; p = 0.025) and BT (OR 1.46; 95% CI 1.01–2.10; p = 0.041). No significant differences were observed for mortality (OR 1.07; 95% CI 0.88–1.29; p = 0.52), ICH (OR 1.70; 95% CI 0.93–3.11; p = 0.082), recurrent PE (OR 1.10; 95% CI 0.92–1.33; p = 0.29), or unplanned readmission (OR 1.08; 95% CI 0.97–1.20; p = 0.18) (Supplementary Table 4).
IPWRA treatment effects (primary analysis)
At index admission, MT was associated with higher average treatment effects (ATEs) for in-hospital mortality (0.007; p = 0.048), DOTH (0.031; p < 0.001), ICH (0.006; p < 0.001), BT (0.017; p < 0.001), and cost (0.118; p < 0.001), with similar GIB (0.004; p = 0.097).
At 30 days, MT was associated with higher mortality (ATE 0.008; p = 0.036) and BT (0.005; p < 0.001), while other outcomes were similar.
At 6 months, MT was associated with small absolute increases in mortality (ATE 0.009; p = 0.025), GIB (0.004; p = 0.023), and BT (0.006; p = 0.017), with no significant differences in unplanned readmission (0.011; p = 0.09), ICH (0.002; p = 0.11), or recurrent PE (0.003; p = 0.39) (Table 2; Fig. 1).
Fig. 1.
Inverse propensity weighting with regression adjustment analysis for outcomes at index admission and 6-month readmissions depicted as forest plot
The mean total hospitalization cost was $41,884 (SD $39,932) for MT compared with $34,788 (SD $27,232) for CDT (median $32,615 vs. $28,827). In IPWRA analysis, MT was associated with a 0.118 higher ln-transformed total cost (p < 0.001), corresponding to approximately 12.5% higher cost. Survey-weighted model-adjusted mean costs were $41,215 for MT versus $35,032 for CDT. Mean length of stay was 6.4 days for MT versus 5.9 days for CDT (p = 0.001).
At the first unplanned readmission within 6 months, circulatory diagnoses were most common in both groups (31.0% CDT vs. 31.6% MT). Sepsis (A419) and unspecified PE (I2699) were among the leading primary diagnoses in both cohorts, with sepsis more frequent in the MT group (7.9% vs. 5.7%) (Supplementary Table 6).
Subgroup analysis—intermediate-risk PE
MT was associated with higher index ICH (ATE 0.004; p = 0.003), BT (0.012; p < 0.001), and cost (0.086; p < 0.001), with similar mortality and DOTH.
At 6 months, MT was associated with higher unplanned readmission (0.014; p = 0.044), GIB (0.004; p = 0.027), recurrent PE (0.010; p = 0.020), BT (0.005; p = 0.039), and mortality (0.005; p = 0.045), with similar ICH (0.001; p = 0.245) (Supplementary Table 5).
Subgroup analysis—high-risk PE
In the high-risk PE subgroup (n = 3945 weighted), MT was associated with higher rates of discharge other than home (ATE 0.101, p < 0.001) and higher ln-transformed total cost (ATE 0.222, p < 0.001), while CDT was associated with higher 6-month PE readmission (ATE − 0.037, p = 0.002); in-hospital mortality did not differ significantly (ATE − 0.025, p = 0.19) (Supplementary Table 7). In this subgroup, the absence of a significant mortality difference and the higher PE readmission rate after CDT suggest that treatment-associated patterns may differ by PE severity.
Discussion
The present study queried a national database to compare the risks of index hospitalization complications and 7-day, 30-day, and 6-month readmissions between CDT and MT for intermediate- and high-risk PEs. MT was associated with higher rates of DOTH, ICH, BT, and total cost at index admission, as well as a small absolute increase in 1-month and 6-month mortality in IPWRA analysis. In the subgroup analysis of intermediate-risk PE, MT demonstrated higher ATEs for ICH, BT, and total cost at index admission, and for unplanned readmission, GIB, recurrent PE, BT, and mortality at 6 months. Although inverse probability weighting improved covariate balance, residual confounding by indication remains likely, as patients undergoing mechanical thrombectomy had higher rates of massive PE and hemodynamic instability at baseline. While IPWRA demonstrated a small absolute increase in 6-month mortality, this association was not observed in multivariable regression, underscoring sensitivity to modeling approach and the potential influence of residual confounding. Long-term outcomes likely reflect cumulative patient trajectory, post-discharge anticoagulation management, and residual confounding rather than procedural effects alone.
Prospective device studies have reported favorable short-term safety in selected populations, including low major bleeding in the FLARE trial evaluating the FlowTriever system and reduced ICH rates in the SEATTLE II trial evaluating ultrasound-assisted CDT [21, 22]. Similarly, short-term thrombectomy outcomes through 30 days were reported in the EXTRACT-PE trial evaluating the Indigo Aspiration System [23]. In contrast, our real-world NRD analysis demonstrated higher adjusted ICH and BT at index admission with MT and higher 6-month GIB, underscoring differences between trial-selected populations and routine clinical practice [21–23]. Furthermore, the findings suggest that MT was associated with higher rates of DOTH, BT, ICH, and total cost at index admission, and a small absolute increase in mortality at 1- and 6-month follow-up in IPWRA analysis, while demonstrating similar rates of unplanned readmission at early time points and higher GIB and unplanned readmission at 6 months.
The higher rates of ICH, blood transfusion, and GI bleeding in the MT cohort merit discussion of alternative explanations. Residual confounding may be important, as patients selected for MT had greater baseline illness severity and a higher cancer burden, both of which carry independent bleeding risk. ICD-10 coding misclassification is a relevant consideration, as procedure-related blood loss from large-bore vascular access may be variably coded across institutions. Variation in periprocedural anticoagulation practices, which cannot be captured in administrative data, may also contribute. Finally, MT is often selected for patients considered unsuitable for thrombolysis due to elevated bleeding risk, which may concentrate bleeding-prone patients in the MT cohort. These findings are consistent with REAL-PE and REAL-PE II. As with all observational analyses, causality cannot be inferred from these data alone.
The PEERLESS randomized controlled trial compared large-bore MT directly to CDT in intermediate-risk PE and reported favorable outcomes for MT using a hierarchical composite win ratio endpoint [7]. The composite included both softer measures such as ICU admission duration, need for rescue therapy, and early clinical deterioration, alongside harder endpoints like mortality and intracranial hemorrhage. Notably, when individual components were examined, no significant differences were identified between groups for mortality or ICH [7]. This distinction matters: win ratio composites can convey an impression of broad superiority that does not always translate into patient-centered benefit when softer, resource-based outcomes account for much of the signal. PEERLESS also enrolled hemodynamically stable intermediate-risk patients under controlled trial conditions, likely excluding those with the highest baseline risk profile where procedural safety margins are narrowest. Our results, alongside REAL-PE and other real-world observational analyses, suggest that outcomes observed in trial populations may not fully generalize to the broader spectrum of patients receiving MT in routine clinical practice [9, 11, 25–27].
The recently published HI-PEITHO trial offers additional context for evaluating catheter-based reperfusion strategies in intermediate-high-risk PE. In this multicenter randomized trial, ultrasound-facilitated CDT reduced the composite outcome of hemodynamic decompensation or death at 7 days by approximately 61% compared with anticoagulation alone [12]. Rates of major bleeding and intracranial hemorrhage were not significantly different between groups, supporting the acceptable safety profile of catheter-directed thrombolytics at reduced doses. These findings reinforce the continued therapeutic value of CDT and raise reasonable questions about whether the added procedural complexity and cost of large-bore mechanical thrombectomy are justified across all intermediate-risk PE presentations, particularly in hemodynamically stable patients. Additionally, our subgroup findings should be interpreted cautiously given ICD-based risk classification and multiple comparisons.
The etiology of higher DOTH after MT is uncertain and may reflect unmeasured severity or real-world practice variation [28]. MT was associated with higher DOTH (ATE 0.031; p < 0.001) and higher BT/ICH, which could relate to operator experience, device learning curves, or incomplete thrombus extraction, particularly early in adoption [21]. Although IPWRA reduces bias from measured confounders, residual unmeasured confounding likely persists [28]. While large retrospective database analyses provide valuable real-world insights, future studies in this field should ideally be prospective in design to better address confounding and causal inference.
The economic implications of device selection deserve careful consideration in the context of PE care. In the present study, MT was associated with significantly higher total hospitalization costs compared with CDT, with an average treatment effect of 0.118 (p < 0.001) in IPWRA analysis. This cost differential persisted in the intermediate-risk PE subgroup (ATE 0.086; p < 0.001), where procedural urgency is often lower and patient trajectories are generally favorable with anticoagulation or reduced-intensity reperfusion alone. These findings are consistent with prior cost-comparison analyses demonstrating substantially higher expenditures for large-bore thrombectomy systems relative to catheter-directed thrombolytic delivery [10]. Large-bore mechanical thrombectomy devices carry per-case costs that can exceed $8000 to $12,000, while catheter-directed administration of recombinant tissue plasminogen activator uses generically available pharmaceuticals at a fraction of that cost [29]. The total cost differential likely also reflects downstream effects, including higher rates of mechanical ventilation (11.2% vs. 8.9%; p = 0.005), periprocedural complications, and longer hospital stays (mean 6.4 vs. 5.9 days; p = 0.001). Captured across more than 28,000 procedures representing approximately 60% of U.S. hospitalizations, the aggregate economic burden of routine MT adoption in intermediate-risk patients without hemodynamic compromise could be substantial at a national level. Prospective analyses incorporating quality-adjusted outcomes and long-term resource utilization will be important for guiding rational device selection across PE risk strata.
Strengths and limitations
Although this study was a comprehensive national-level analysis, there are several limitations. Administrative claims data was utilized to identify diagnoses and procedures, which is prone to human error and misclassification of the ICD codes [30]. Specifically, the stratification of risk of PE in this current study was based on ICD codes. However, codes selected for this study were curated by clinicians and billing code experts as well as industry billing guides [17, 29, 31, 32]. Moreover, heparin use and dosage rates were not able to be recorded due to lack of corresponding ICD codes. Those who underwent IVT were excluded, which limits the analysis to admission that did not require IVT either before or after the studied intervention. The NRD does not capture in-hospital fibrinolytic dosing, and some patients undergoing MT may have received adjunctive thrombolytics, potentially influencing bleeding outcomes. Claims-based data cannot distinguish procedural complications from bleeding related to anticoagulation or comorbidities. Additionally, there are less than 11 patients with brain tumor, and among them with ICH. These patients are unlikely to meaningfully influence overall results. Due to the nature of retrospective data, there may be selection bias with procedural choices based on risk factors and that patients who received MT may have required faster intervention than the slower thrombolytic infusion of CDT. As with other claims-based analyses, individual thrombectomy devices and CDT delivery systems cannot be distinguished. The MT cohort likely encompasses multiple large-bore aspiration, rheolytic, and combined thrombectomy platforms with differing technical profiles. The CDT cohort may similarly include both standard infusion catheters and ultrasound-facilitated delivery systems. Thrombectomy technology evolved across the 2016 to 2022 study period, and these results reflect the composite real-world experience during this period rather than any specific device or technique. Although IPWRA reduced bias from measured confounders, unmeasured clinical severity variables and selection factors may still influence outcomes. Additionally, although we utilized the ECI score for risk stratification, the database does not capture PE-specific physiologic severity markers such as RV dysfunction, clot burden, laboratory values, or operator decision-making factors. Lastly, cause of mortality is not recorded. Mortality in this study reflects in-hospital death during index admission or subsequent readmissions within 7, 30, and 180 days, and does not capture out-of-hospital mortality.
As with all observational database studies, residual confounding by indication cannot be excluded. IPWRA achieved standardized mean differences below 0.10 for all measured covariates; however, key clinical parameters that influence treatment selection, including RV/LV ratio, clot burden, troponin, BNP, and oxygen requirement, are not available in administrative claims data. Patients undergoing MT had higher baseline rates of massive PE, vasopressor use, and mechanical ventilation, and these baseline differences should be considered when interpreting the observed associations. Real-world database analyses nonetheless provide important evidence across the broad spectrum of patients encountered in clinical practice, complementing findings from randomized trials conducted in more selected populations [24].
The classification of PE severity using administrative ICD-10 codes does not fully replicate contemporary guideline-based risk stratification. Key clinical parameters including RV/LV ratio, troponin, BNP, echocardiographic findings, and oxygen requirements are not available in the NRD. High-risk PE was approximated using markers of hemodynamic compromise consistent with prior database studies; however, intermediate-risk patients in this cohort may represent a heterogeneous population that would not uniformly satisfy contemporary definitions of intermediate-high-risk PE. This is a known constraint of administrative database research in this field.
Conclusion
In this national analysis, mechanical thrombectomy was associated with higher rates of adverse in-hospital events and total cost compared with catheter-directed thrombolysis, as well as a small absolute increase in 6-month mortality on primary analysis. These findings reflect real-world patterns across a nationally representative cohort and should be considered alongside the observational study design. Prospective randomized data are needed to define the optimal role of each strategy in contemporary PE care.
Supplementary Information
Supplementary Material 1: Supplementary Table 1: ICD codes used in this study. Supplementary Table 2: Covariate balance before and after weighting. Supplementary Table 3: Multivariable Regression for mortality, DOTH, GI bleed, ICH, blood transfusion and total cost at index admission. Supplementary Table 4: Multivariable Regression for unplanned readmission, GI bleed, ICH, recurrent pulmonary embolism, blood transfusion, and mortality at 6-month follow-up. Supplementary Table 5: Average treatment effect of Thrombectomy alone compared to CDT for outcomes in the index admission and admission at 6 months for intermediate-risk PEs. Supplementary Table 6. Primary Diagnoses at First Unplanned Readmission Within 6 Months, by Treatment Group. Supplementary Table 7. Inverse Probability Weighted Regression Adjustment (IPWRA) Results — High-Risk Pulmonary Embolism Subgroup.
Acknowledgements
Not applicable.
Abbreviations
- ATE
Average treatment effect
- BT
Blood transfusion
- CDT
Catheter-directed thrombolysis
- DOTH
Discharge other than home
- GIB
Gastrointestinal bleeding
- ICH
Intracranial hemorrhage
- MT
Mechanical thrombectomy
- PE
Pulmonary embolism
- NRD
Nationwide Readmissions Database
- OR
Odds ratio
- IPWRA
Inverse propensity weighted regression adjustment
- CI
Confidence interval
Authors’ contributions
WW had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis, including and especially any adverse effects. WW, HH, OS, and DI contributed substantially to the study design, data analysis and interpretation, and the writing of the manuscript. PM, JP, and OS contributed substantially to the review and editing of the manuscript, and final approval. All authors read and approved the final manuscript.
Funding
This work was not supported by funding agencies in the public, commercial, or not-for-profit sectors.
Data availability
The data that support the findings of this study are available from the Healthcare Cost and Utilization Project (HCUP) Nationwide Readmissions Database (NRD). Data are available for purchase at https://www.hcup-us.ahrq.gov/db/nation/nrd/nrddbdocumentation.jsp.
Declarations
Ethics approval and consent to participate
This manuscript reports on de-identified patient data that is publicly available for purchase. Institutional review board approval was not required.
Consent for publication
Not applicable.
Competing interests
The authors declare no competing interests.
Footnotes
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
References
- 1.Bĕlohlávek J, Dytrych V, Linhart A. Pulmonary embolism, part I: Epidemiology, risk factors and risk stratification, pathophysiology, clinical presentation, diagnosis and nonthrombotic pulmonary embolism. Exp Clin Cardiol. 2013;18:129–38. Available from: https://www.ncbi.nlm.nih.gov/pubmed/23940438. [PMC free article] [PubMed]
- 2.Konstantinides SV, Meyer G, Becattini C, Bueno H, Geersing G-J, Harjola V-P, et al. 2019 ESC Guidelines for the diagnosis and management of acute pulmonary embolism developed in collaboration with the European Respiratory Society (ERS). Eur Heart J. 2020;41:543–603. Available from: 10.1093/eurheartj/ehz405. [DOI] [PubMed]
- 3.Toma C, Bunte MC, Cho KH, Jaber WA, Chambers J, Stegman B, et al. Percutaneous mechanical thrombectomy in a real-world pulmonary embolism population: interim results of the FLASH registry. Catheter Cardiovasc Interv. 2022;99:1345–55. Available from: https://onlinelibrary.wiley.com/doi/abs/10.1002/ccd.30091. [DOI] [PMC free article] [PubMed]
- 4.Meyer G, Vicaut E, Danays T, Agnelli G, Becattini C, Beyer-Westendorf J, et al. Fibrinolysis for patients with intermediate-risk pulmonary embolism. N Engl J Med. 2014;370:1402–11. Available from: 10.1056/NEJMoa1302097. [DOI] [PubMed]
- 5.Marti C, John G, Konstantinides S, Combescure C, Sanchez O, Lankeit M, et al. Systemic thrombolytic therapy for acute pulmonary embolism: a systematic review and meta-analysis. Eur Heart J. 2015;36:605–14. Available from: 10.1093/eurheartj/ehu218. [DOI] [PMC free article] [PubMed]
- 6.Desai KR. Mechanical Thrombectomy in Pulmonary Embolism. JACC: Cardiovascular Interventions. 2021. p. 330–2. Available from: 10.1016/j.jcin.2020.11.002. [DOI] [PubMed]
- 7.Jaber WA, Gonsalves CF, Stortecky S, Horr S, Pappas O, Gandhi RT, et al. Large-bore mechanical thrombectomy versus catheter-directed thrombolysis in the management of intermediate-risk pulmonary embolism: primary results of the PEERLESS randomized controlled trial. Circulation. 2024. Available from: 10.1161/CIRCULATIONAHA.124.072364. [DOI] [PMC free article] [PubMed]
- 8.Inci EK, Khandhar S, Toma C, Licitra G, Brown MJ, Herzig M, et al. Mechanical thrombectomy versus catheter directed thrombolysis in patients with pulmonary embolism: a multicenter experience. Catheter Cardiovasc Interv. 2023;101:140–6. Available from: 10.1002/ccd.30505. [DOI] [PubMed]
- 9.Graif A, Patel KD, Wimmer NJ, Kimbiris G, Grilli CJ, Upparapalli D, et al. Large-bore aspiration thrombectomy versus catheter-directed thrombolysis for acute pulmonary embolism: a propensity score-matched comparison. J Vasc Interv Radiol. 2020;31:2052–9. Available from: 10.1016/j.jvir.2020.08.028. [DOI] [PubMed]
- 10.Tran A, Toma C, Jaber W, Heintz J, Matthai WH, Palevsky H, et al. Cost-comparison of mechanical thrombectomy and catheter-directed thrombolysis in intermediate-risk pulmonary embolism. J Soc Cardiovasc Angiogr Interv. 2024;3:101187. Available from: 10.1016/j.jscai.2023.101187. [DOI] [PMC free article] [PubMed]
- 11.Abrahamian A, Khokher W, Ahmad R, Holtzapple Z, Patel R, Assaly RA, et al. Catheter-directed thrombolysis vs mechanical thrombectomy in treatment of acute pulmonary embolism: a systemic review and meta-analysis of dual arm studies. Chest. 2023;164:A5245–6. Available from: 10.1016/j.chest.2023.07.3401.
- 12.Rosenfield K, Klok FA, Piazza G, Sharp ASP, Ní Áinle F, Jaff MR, et al. Ultrasound-facilitated, catheter-directed fibrinolysis for acute pulmonary embolism. N Engl J Med. 2026. Available from: 10.1056/NEJMoa2516567. [DOI] [PubMed]
- 13.NRD Database Documentation. Available from: https://hcup-us.ahrq.gov/db/nation/nrd/nrddbdocumentation.jsp. Cited 2 Mar 2023.
- 14.Smith SB, Geske JB, Kathuria P, Cuttica M, Schimmel DR, Courtney DM, et al. Analysis of national trends in admissions for pulmonary embolism. Chest. 2016;150:35–45. Available from: 10.1016/j.chest.2016.02.638. [DOI] [PMC free article] [PubMed]
- 15.Kucher N, Goldhaber SZ. Management of massive pulmonary embolism. Circulation. 2005;112:e28–32. Available from: 10.1161/CIRCULATIONAHA.105.551374. [DOI] [PubMed]
- 16.Yamamoto T. Management of patients with high-risk pulmonary embolism: a narrative review. J Intensive Care Med. 2018;6:16. Available from: 10.1186/s40560-018-0286-8. [DOI] [PMC free article] [PubMed]
- 17.Wahood W, Sista AK, Paul JD, Ahmed O. Unplanned 30-day readmissions after management of submassive and massive acute pulmonary embolism: catheter-directed versus systemic thrombolysis. J Vasc Interv Radiol. 2023;34. Available from: https://pubmed.ncbi.nlm.nih.gov/36167297/. Cited 16 Feb 2023. [DOI] [PubMed]
- 18.van Walraven C, Austin PC, Jennings A, Quan H, Forster AJ. A modification of the Elixhauser comorbidity measures into a point system for hospital death using administrative data. Med Care. 2009;47:626–33. Available from: 10.1097/MLR.0b013e31819432e5. [DOI] [PubMed]
- 19.Austin PC. An introduction to propensity score methods for reducing the effects of confounding in observational studies. Multivariate Behav Res. 2011;46:399–424. Available from: 10.1080/00273171.2011.568786. [DOI] [PMC free article] [PubMed]
- 20.Witte J, Didelez V. Covariate selection strategies for causal inference: classification and comparison. Biom J. 2019;61:1270–89. Available from: 10.1002/bimj.201700294. [DOI] [PubMed]
- 21.Tu T, Toma C, Tapson VF, Adams C, Jaber WA, Silver M, et al. A prospective, single-arm, multicenter trial of catheter-directed mechanical thrombectomy for intermediate-risk acute pulmonary embolism: the FLARE study. JACC Cardiovasc Interv. 2019;12:859–69. Available from: 10.1016/j.jcin.2018.12.022. [DOI] [PubMed]
- 22.Piazza G, Hohlfelder B, Jaff MR, Ouriel K, Engelhardt TC, Sterling KM, et al. A prospective, single-arm, multicenter trial of ultrasound-facilitated, catheter-directed, low-dose fibrinolysis for acute massive and submassive pulmonary embolism: the SEATTLE II study. JACC Cardiovasc Interv. 2015;8:1382–92. Available from: 10.1016/j.jcin.2015.04.020. [DOI] [PubMed]
- 23.Sista AK, Horowitz JM, Tapson VF, Rosenberg M, Elder MD, Schiro BJ, et al. Indigo aspiration system for treatment of pulmonary embolism: results of the EXTRACT-PE trial. JACC Cardiovasc Interv. 2021;14:319–29. Available from: 10.1016/j.jcin.2020.09.053. [DOI] [PubMed]
- 24.Monteleone PP. Acute safety of contemporary pulmonary embolism treatment with ultrasoundassisted catheter-directed thrombolysis or mechanical thrombectomy in a large matched sample: real-PE ii. J Am Coll Cardiol. 2025;85:2166. Available from: 10.1016/s0735-1097(25)02650-6.
- 25.Monteleone P, Ahern R, Banerjee S, Desai KR, Kadian-Dodov D, Webber E, et al. Modern treatment of pulmonary embolism (USCDT vs MT): results from a real-world, big data analysis (REAL-PE). J Soc Cardiovasc Angiogr Interv. 2024;3:101192. Available from: 10.1016/j.jscai.2023.101192. [DOI] [PMC free article] [PubMed]
- 26.Monteleone P, Banerjee S, Desai KR, Jaff MR. Stop apologizing for real-world data. Am J Cardiol. 2024;211:374–6. Available from: 10.1016/j.amjcard.2023.10.033. [DOI] [PubMed]
- 27.Majmundar M, Majmundar V, Bhat V, Patel KN, Parmar G, Alli A, et al. Endovascular mechanical thrombectomy versus catheter-directed thrombolysis in pulmonary embolism: insights from the National Readmission Database. Eur Heart J Acute Cardiovasc Care. 2025. Available from: 10.1093/ehjacc/zuaf110. [DOI] [PubMed]
- 28.Miguel A, Robins HERNAN(JAMESM). Causal inference: what if. CRC PRESS; 2023. [Google Scholar]
- 29.Arora S, Panaich SS, Ainani N, Kumar V, Patel NJ, Tripathi B, et al. Comparison of in-hospital outcomes and readmission rates in acute pulmonary embolism between systemic and catheter-directed thrombolysis (from the National Readmission Database). Am J Cardiol. 2017;120:1653–61. Available from: https://www.sciencedirect.com/science/article/pii/S0002914917312390. [DOI] [PubMed]
- 30.Rabinstein AA. Administrative medical databases for clinical research: the good, the bad, and the ugly. Neurocrit Care. 2018;29:323–5. Available from: 10.1007/s12028-018-0625-6. [DOI] [PubMed]
- 31.Reimbursement. Inari Medical. 2021. Available from: https://www.inarimedical.com/reimbursement/. Cited 27 Apr 2023.
- 32.Beyer SE, Shanafelt C, Pinto DS, Weinstein JL, Aronow HD, Weinberg I, et al. Utilization and outcomes of thrombolytic therapy for acute pulmonary embolism: a nationwide cohort study. Chest. 2020;157:645–53. Available from: 10.1016/j.chest.2019.10.049. [DOI] [PubMed]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Supplementary Material 1: Supplementary Table 1: ICD codes used in this study. Supplementary Table 2: Covariate balance before and after weighting. Supplementary Table 3: Multivariable Regression for mortality, DOTH, GI bleed, ICH, blood transfusion and total cost at index admission. Supplementary Table 4: Multivariable Regression for unplanned readmission, GI bleed, ICH, recurrent pulmonary embolism, blood transfusion, and mortality at 6-month follow-up. Supplementary Table 5: Average treatment effect of Thrombectomy alone compared to CDT for outcomes in the index admission and admission at 6 months for intermediate-risk PEs. Supplementary Table 6. Primary Diagnoses at First Unplanned Readmission Within 6 Months, by Treatment Group. Supplementary Table 7. Inverse Probability Weighted Regression Adjustment (IPWRA) Results — High-Risk Pulmonary Embolism Subgroup.
Data Availability Statement
The data that support the findings of this study are available from the Healthcare Cost and Utilization Project (HCUP) Nationwide Readmissions Database (NRD). Data are available for purchase at https://www.hcup-us.ahrq.gov/db/nation/nrd/nrddbdocumentation.jsp.