Abstract
Background
This study aimed to identify granular discharge barriers after congenital heart surgery and compare the contributions of discharge barriers versus clinical factors to postoperative length of stay (LOS) variation among infants with severe congenital heart defects.
Methods
This was a retrospective cohort study of patients who underwent initial cardiac surgery at ≤12 months, from 2015 to 2021, at a high‐volume congenital heart defect surgical center. The outcome was postoperative LOS. The exposure was discharge barrier category, a 7‐level categorical variable determined a priori using a mixed‐methods approach. Covariates included surgical/anatomic complexity, postoperative complications, sociodemographics, and comorbidities. Associations between discharge barrier categories and postoperative LOS were estimated with log‐linear models. A Shorrocks‐Shapley decomposition estimated each covariate's contribution to outcome variation.
Results
A total of 752 infants met the inclusion criteria; 24.3% experienced a discharge barrier. The most common discharge barrier category was medical equipment delivery or prior authorization delays (52/752, 6.9%). Median postoperative LOS for those without and with barriers was 8 (interquartile range, 5–15) and 26 (interquartile range, 14–52) days, respectively (P<0.001). In a multivariable analysis, 4 of the 6 barrier categories were associated with higher postoperative LOS when compared with patients without barriers (31%–242%, P=0.000–0.030). Surgical complexity and discharge barriers accounted for similar amounts of variation in postoperative LOS (22.2% [95% CI, 17.3%–26.8%] and 21.2% [95% CI, 14.9%–27.0%], respectively).
Conclusions
Discharge barriers are as influential as surgical and anatomic complexity in explaining variation in postoperative LOS among infants undergoing congenital heart surgery. Future research and clinical efforts to mitigate modifiable discharge barriers may yield improvements in postoperative LOS.
Keywords: congenital heart surgery, discharge barriers, length of stay
Subject Categories: Quality and Outcomes, Social Determinants of Health
Nonstandard Abbreviations and Acronyms
- CHD
congenital heart defect
- STAT
Society of Thoracic Surgeons–European Association for Cardio‐Thoracic Surgery
- STS‐CHSD
Society of Thoracic Surgeons Congenital Heart Surgery Database
Research Perspective.
What New Question Does This Study Raise?
Discharge barriers are as influential as surgical and anatomic complexity in determining postoperative length of stay among infants undergoing congenital heart surgery.
What Question Should Be Addressed Next?
What patient‐specific factors are associated with increased discharge barriers?
What is the impact of a standardized, surgery‐specific approach to addressing discharge barriers on patient length of stay?
Congenital heart defects (CHDs) are the most common and resource‐intensive birth defects, 1 affecting ≈1 in 100 live births. 2 Among those who require cardiac surgery, severe CHDs account for ≈15% to 25%, 3 an annual economic burden of ≈$75 billion, and a lifetime burden of $3 trillion. 4 Specifically, prolonged length of stay (LOS) is a well‐established proxy for both economic and clinical burden, given its consistent association with higher inpatient costs 5 , 6 and worse long‐term developmental outcomes 7 , 8 among children with CHDs.
Several studies have further demonstrated that clinical factors, especially surgical and/or anatomic complexity, are significant and consistent predictors of LOS. 9 , 10 , 11 , 12 Yet, surgical/anatomic complexity is not modifiable, and large administrative and registry data lack the granularity to identify the modifiable factors that contribute to variation in LOS. 13 This highlights a clear knowledge gap: how can clinical care be modified to minimize postoperative LOS for patients following congenital heart surgery?
To address this knowledge gap, we aimed to (1) identify discrete discharge barriers using a mixed‐methods analysis of granular electronic medical record data, (2) estimate associations between discharge barriers and postoperative LOS, and (3) compare the relative contribution of discharge barriers and each covariate to variation in LOS. For the third aim, we hypothesized that discharge barriers were less contributory to postoperative LOS compared with clinical factors such as surgical/anatomic complexity. We investigated these aims among infants who received cardiac surgery for severe CHDs in the first year of life as a case example of a pediatric cohort with 1 of the highest degrees of resource use. 1
METHODS
Study Design and Data Source
This is a mixed‐methods cross‐sectional analysis of retrospective data obtained from the Ann & Robert H. Lurie Children's Hospital of Chicago, a high‐volume pediatric cardiac surgical center. 14 Infants with CHDs who received their first cardiac surgery at ≤12 months of age between January 1, 2015, and December 31, 2021, were included in the study. 15 , 16 We chose 12 months as the age cutoff to define severe CHD, in accordance with the Centers for Disease Control and Prevention's definition. 3 The year 2021 was the upper bound of the inclusion period to maximize data accuracy; at the time of data acquisition, data accuracy after 2021 may have been substantially affected by post–COVID‐related disruptions and changes to hospital operation and discharge planning. Patients were excluded if they were diagnosed with connective tissue diseases leading to cardiac disease (eg, aortic root dilation) or acquired cardiac conditions (eg, endocarditis), or if their surgery did not require cardiopulmonary bypass (eg, ligation of patent ductus arteriosus), as these procedures do not reflect severe CHDs.
Data were obtained from the Society of Thoracic Surgeons Congenital Heart Surgery Database (STS‐CHSD) at Ann & Robert H. Lurie Children's Hospital of Chicago. The STS‐CHSD is the largest congenital heart surgery clinical registry in the world. It captures >90% of all US congenital heart operations, with detailed information on patient demographics, hospital, surgeon, 110 anatomically specific cardiac diagnoses, 187 procedures and procedural subtypes, 22 comorbid conditions and preoperative risk factors, as well as postoperative complications. 17 In addition, more granular clinical variables that were not available in the STS‐CHSD were collected manually from patients' electronic health records. Because of the presence of protected health information in the data set, data will not be shared or made available on request.
Outcome Variable
The outcome was postoperative LOS, in days, starting from date of surgery to date of discharge. Postoperative LOS, rather than total LOS, was chosen as the outcome to avoid confounding from preoperative events. 18
Exposure Variable: Mixed‐Methods Determination of Discharge Barrier Groups and Categories
The exposure variable was discharge barrier categories, which were defined a priori using a mixed‐methods approach: first, barriers to discharge were manually extracted from the electronic health record (Z.Z.) as excerpts from any clinical provider note (eg, physician, social worker, nurse, or case manager) that indicated a delay in discharge as perceived by the provider. Example excerpts included “identifying a Department of Child and Family Services–appointed guardian” or “insurance approval of enalapril solution.” Next, each excerpt was codified into 13 groups based on clinical consensus (Z.Z., J.W., and A.L.): delay because of parental preference, need for supervised parental overnight stay, availability of home nursing, long‐term acute care center bed availability, unexpected clinical instability, delayed medical equipment delivery, delayed medical equipment teaching, prior authorization, delay because of interpreter assistance, lack of transportation, involvement of the Department of Child and Family Services, lack of home resources (eg, utility assistance), and other.
Finally, the 13 groups were collapsed into 6 discharge barrier categories to achieve adequate cell size for analysis and to account for the fact that some patients experienced multiple discharge barriers. If a patient experienced multiple discharge barriers, his/her clinical documentation of the discharge process was rereviewed by 3 clinical cardiologists (J.W., Z.Z., and A.L.), and a single, predominant discharge barrier was assigned. Three categorizations were separately performed (Z.Z., J.W., and A.L.), then each was tested using ordinary least squares regression (C.L.). The final categorization was determined first based on the magnitude/rank of each of the coefficients for the 13 groups, followed by discussion and consensus by all coauthors. The final 6 discharge barrier categories were as follows: home nursing or long‐term acute care center bed availability, parental overnight stay, delay because of medical equipment delivery or prior authorization approval, delay because of parental preference or equipment teaching, delay because of social determinants of health (ie, transport need, utility assistance, or involvement of the Department of Child and Family Services), and unexpected clinical instability or other.
Additional Covariates
Additional covariates included clinical factors, sociodemographic variables, and comorbidities. Clinical factors included CHD surgical complexity measured by the Society of Thoracic Surgeons–European Association for Cardio‐Thoracic Surgery (STAT) category, a previously validated 19 , 20 5‐level categorical variable within the STS‐CHSD, where STAT 1 is defined as the lowest surgical mortality risk, and STAT 5 is defined as the highest surgical mortality risk; poor weight gain; postoperative complications; and discharge equipment needs (eg, oxygen or nasogastric tube). A patient was identified as having poor weight gain if postoperatively he/she was documented to have “poor weight gain,” “slow weight gain,” “inadequate weight gain,” or “failure to thrive” at least once. Postoperative complications included any complication listed in the STS‐CHSD, which includes, but is not limited to, arrhythmia, respiratory failure, and acute renal failure. Although each patient's respiratory needs were noted at discharge, his/her respiratory support need before his/her surgery, such as length of ventilatory support before the operation, was not noted. Sociodemographic variables included race, ethnicity, payor (commercial/military/Tricare versus public), and transportation aid received (parking only versus transportation only versus both). Military/Tricare insurances were grouped with commercial insurance given that their LOSs were similar to commercial rather than public insurance. Comorbidities included presence of a genetic syndrome, nonchromosomal syndrome, or extracardiac structural abnormality.
Statistical Analysis
Standard descriptive statistics were used to analyze each descriptive variable. Bivariate comparisons were performed using χ2 or Wilcoxon rank‐sum test.
A log‐transformed ordinary least squares regression model with robust SEs was used to estimate the adjusted associations between discharge barrier categories and postoperative LOS. The model included discharge barrier categories, STAT category, and additional covariates (postoperative complications, equipment needs at discharge, poor postoperative weight gain, chromosomal/noncardiac/genetic anomaly, race and ethnicity, insurance type, parking or transportation aid received, and a COVID‐19 indicator variable).
Notably, weight, operative duration, residential distance from the hospital, residential rurality, spoken language, and number of adults in the family unit were collected but excluded from the model either because of collinearity or lack of explanatory power. This was also found to be true for sex, age at surgery, and gestational age. There was some overlap between the discharge barrier categories and the remaining covariates; for instance, patients with discharge barriers related to delays in equipment delivery also had discharge equipment needs. However, testing between the covariates and the exposure variable did not suggest collinearity (variance inflation factor <2). To provide estimates that may be more clinically meaningful, we report differences in postoperative LOS in 2 ways: (1) differences in predicted mean LOS in days (ie, accounting for bias in prediction using a Duan estimator 21 and assuming averages of other covariates) and (2) percentage difference in median LOS (ie, log‐transformed regression coefficient). Because we analyzed postoperative LOS on a log scale, the estimated differences between groups depend on input values of the other covariates within the model. Therefore, differences in predicted mean postoperative LOS should be interpreted as relative, rather than absolute, differences.
Results were compared using the Shorrocks‐Shapley decomposition, 22 a postestimation that provides an additive decomposition of the relative contribution of each covariate (or groups of covariates) to the explained variation (R 2) in the model. We used the Stata command shapley2 to calculate the Shapley decomposition. 23
Sensitivity Analyses
We performed 3 sensitivity analyses. First, to determine how the Shorrocks‐Shapley decomposition might change based on the specification of the exposure variable, we tested a binary indicator for perceived discharge barriers, rather than a 6‐level categorical variable. Second, to test whether the 6 discharge barrier categories were too broad, we ran a model with the 13 discharge barrier groups individually and simultaneously. Third, to determine if results were significantly different based on the specification of the outcome variable, we replaced postoperative LOS with total LOS (number of days between admission and discharge). Fourth, to determine if results differed between high‐ versus low‐complexity surgical cases, we repeated the regression and Shorrocks‐Shapley analysis, stratifying by low (STAT category 1–2) and high (STAT category 3–5) cases. All analyses were conducted using Stata18 (StataCorp, College Station, TX). The study was approved by the Institutional Review Board at Ann & Robert H. Lurie Children's Hospital of Chicago (number 2022‐5367). Informed consent from the patients and their caregivers was waived.
RESULTS
Cohort Characteristics
In total, 752 patients met inclusion criteria. Of these, 24.3% (n=183) experienced at least 1 discharge barrier, and 8.5% (n=64) experienced >1 discharge barrier. The most prevalent discharge barrier categories were medical equipment delivery or prior authorization (6.9%, n=52) and discharge delay because of supervised overnight stays (6.5%, n=49). The least prevalent discharge barrier category was social determinants (1.9%, n=14; Table 1). The median postoperative LOS for the whole cohort was 11 (interquartile range, 6–21) days and was ≈3 times longer among those with versus without discharge barriers (8 [interquartile range, 5–15] days versus 26 [interquartile range, 14–52 days]; P<0.001; Table 2).
Table 1.
Perceived Discharge Barriers (N=183)
| Discharge barrier category | Description | No.* | Discharge barrier group (No.†) | Example excerpt from electronic health record |
|---|---|---|---|---|
| Medical equipment delivery or prior authorization | Delay because of medical equipment delivery (eg, feeding supplies) or insurance prior authorization | 52 | Medical equipment delivery (37) | “Backordered tracheostomy supplies” |
| Prior authorization (39) | “Prior authorization for sildenafil using Indiana Medicaid” | |||
| Need for supervised overnight stay | Nursing‐supervised parental stay to demonstrate comfort with medication schedule or medical equipment | 49 | Overnight stay (49) | “12 h [supervised] stays for both mom and dad, discharge delayed pending availability of parents for stay” |
| Need for equipment teaching, or parental preference | Delay in discharge because of need for parent teaching, or parent request for prolonged stay because of teaching | 28 | Equipment teaching (58) | “Nephrostomy tube and pump teaching” |
| Parent preference (3) | “Family staying for further comfort with nasogastric tube” | |||
| Home nurse or long‐term acute center bed availability | Delay in discharge because of availability of home nurse or bed at receiving hospital | 24 | Home nurse delay (24) | “20 h of nursing for home not appropriate for discharge on trach/vent” |
| Bed availability (14) | “Pending [long‐term acute care facility] bed availability” | |||
| New clinical instability | Unexpected clinical instability after discharge readiness was already determined, or miscellaneous discharge barrier | 16 | New clinical instability (10) | “Failed car seat challenge due to increased work of breathing and tachypnea” |
| Other (11) | “Did not have social security card and had difficult time finding primary care provider” | |||
| Social determinants | Delay because of lack of interpreter, resources, or transportation; or delay because of involvement of DCFS | 14 | Interpreter (9) | “Used Nepalese interpreter for discharge teaching” |
| Lack of home resources (10) | “Mother reporting utilities shut off 2 d prior to discharge” | |||
| Transportation (12) | “Illinois Medicaid ride did not show” | |||
| Involvement of Department of Child and Family Services (3) | “Coordinating DCFS discharge with emergency foster parent, was originally supposed to go home with aunt” |
DCFS indicates Department of Child and Family Services.
Each patient could only experience discharge barrier category, and values are additive.
Because some patients experienced multiple discharge barrier groups, values are not additive.
Table 2.
Cohort Characteristics
| Characteristic | Total | No discharge barriers | At least 1 discharge barrier | P value |
|---|---|---|---|---|
| (N=752; 100%) | (n=569; 75.7%) | (n=183; 24.3%) | ||
| Postoperative length of stay, median (IQR), d | 11 (6–21) | 8 (5–15) | 26 (14–52) | <0.001 |
| STAT category, n (%)* | <0.001 | |||
| 1 (Least complex) | 291 (38.7) | 249 (43.8) | 42 (23.0) | |
| 2 | 207 (27.5) | 158 (27.8) | 49 (26.8) | |
| 3 | 105 (14.0) | 70 (12.3) | 35 (19.1) | |
| 4 | 85 (11.3) | 54 (9.5) | 31 (16.9) | |
| 5 (Most complex) | 64 (8.5) | 38 (6.7) | 26 (14.2) | |
| Postoperative complication, n (%) | <0.001 | |||
| No | 285 (37.9) | 250 (43.9) | 35 (19.1) | |
| Yes | 467 (62.1) | 319 (56.1) | 148 (80.9) | |
| Equipment needs at discharge, n (%) | <0.001 | |||
| No | 340 (45.2) | 309 (54.3) | 31 (16.9) | |
| Yes | 412 (54.8) | 260 (45.7) | 152 (83.1) | |
| Poor postoperative weight gain, n (%) | <0.001 | |||
| No | 567 (75.4) | 466 (81.9) | 101 (55.2) | |
| Yes | 185 (24.6) | 103 (18.1) | 82 (44.8) | |
| Chromosomal, noncardiac, or genetic abnormality, n (%) | <0.001 | |||
| No | 401 (53.3) | 328 (57.6) | 73 (39.9) | |
| Yes | 351 (46.7) | 241 (42.4) | 110 (60.1) | |
| Race and ethnicity, n (%) | <0.001 | |||
| Asian, non‐Hispanic | 50 (6.7) | 40 (7.0) | 10 (5.5) | |
| Black, non‐Hispanic | 95 (12.6) | 57 (10.0) | 38 (20.8) | |
| Other, non‐Hispanic | 27 (3.6) | 22 (3.9) | 5 (2.7) | |
| White, non‐Hispanic | 366 (48.7) | 297 (52.2) | 69 (37.7) | |
| Hispanic | 213 (28.3) | 152 (26.7) | 61 (33.3) | |
| Insurance type at time of surgery, n (%) | <0.001 | |||
| Commercial/military/Tricare | 431 (57.3) | 352 (61.9) | 79 (43.2) | |
| Medicaid | 305 (40.6) | 205 (36.0) | 100 (54.6) | |
| Parking or transportation aid received, n (%) | <0.001 | |||
| None | 368 (48.9) | 321 (56.4) | 47 (25.7) | |
| Parking aid | 141 (18.8) | 94 (16.5) | 47 (25.7) | |
| Transportation aid | 110 (14.6) | 75 (13.2) | 35 (19.1) | |
| Parking and transportation aid | 133 (17.7) | 79 (13.9) | 54 (29.5) | |
| Surgery during COVID‐19† pandemic, n (%) | 0.229 | |||
| Before COVID‐19 | 520 (69.2) | 400 (70.3) | 120 (65.6) | |
| During COVID‐19 | 232 (30.8) | 169 (29.7) | 63 (34.4) | |
| Sex, n (%) | 0.569 | |||
| Female | 352 (46.8) | 263 (46.2) | 89 (48.6) | |
| Male | 400 (53.2) | 306 (53.8) | 94 (51.4) | |
| Age at surgery, median (IQR), d | 97 (8–161) | 113 (11–169) | 62 (7–143) | 0.001 |
IQR indicates interquartile range; and STAT, Society of Thoracic Surgeons–European Association for Cardio‐Thoracic Surgery.
The STAT category is a previously validated score to stratify surgical mortality risk.
SARS‐CoV‐2 pandemic of 2019 (COVID‐19).
Adjusted Associations of Postoperative LOS
When compared with those without discharge barriers, those with delays because of home nursing or bed availability had a 33‐day longer predicted postoperative LOS (95% CI, 14–51 days) (ie, an adjusted 242% longer LOS [95% CI, 127%–415%]); those who required a supervised overnight parental stay had a 12‐day longer mean postoperative LOS (95% CI, 6–18 days) (ie, an adjusted 87% longer LOS [95% CI, 46%–139%]); those with unexpected clinical instability/other discharge barriers had a 10‐day longer mean postoperative LOS (95% CI, 4–17 days) (ie, an adjusted 78% longer LOS [95% CI, 34%–135%]); and those with delays because of medical equipment delivery and insurance prior authorization had a 4‐day longer mean postoperative LOS (95% CI, 0–8 days) (ie, an adjusted 31% longer LOS [95% CI, 3%–66%]) (Table 3).
Table 3.
Adjusted Associations With Postoperative LOS
| Variable | % Difference in postoperative LOS (95% CI) | % Contribution to explained variation in postoperative LOS (95% CI) |
|---|---|---|
| Discharge barrier category |
21.2 † (14.9 to 27.0) |
|
| None | Reference | |
| Medical equipment delivery or prior authorization | 31 (3 to 66)* | |
| Need for supervised overnight stay | 87 (46 to 139)* | |
| Need for equipment teaching, or parental preference | 11 (−10 to 37) | |
| Home nurse or bed availability | 242 (127 to 415)* | |
| New clinical instability | 78 (34 to 135)* | |
| Social determinants | 22 (−6 to 57) | |
| STAT category (surgical complexity)‡ |
22.2 (17.3 to 26.8) |
|
| 1 (Least complex) | Reference | |
| 2 | 27 (13 to 43)* | |
| 3 | 39 (20 to 62)* | |
| 4 | 77 (52 to 106)* | |
| 5 (Most complex) | 149 (100 to 209)* | |
| Postoperative complications |
20.2 (16.1 to 23.6) |
|
| 0 | Reference | |
| ≥1 | 70 (55 to 85)* | |
| Equipment needs at discharge |
18.2 (14.7 to 21.5) |
|
| No | Reference | |
| Yes | 51 (38 to 66)* | |
| Poor postoperative weight gain |
9.7 (6.0 to 13.7) |
|
| No | Reference | |
| Yes | 40 (23 to 59)* | |
| Chromosomal, noncardiac or genetic anomaly |
1.9 (0.1 to 3.6) |
|
| No | Reference | |
| Yes | 11 (2 to 22)* | |
| Race and ethnicity |
6.9 (5.3 to 10.8) |
|
| White, non‐Hispanic | Reference | |
| Asian, non‐Hispanic | 13 (−7 to 36) | |
| Black, non‐Hispanic | 12 (−4 to 31) | |
| Other, non‐Hispanic | 33 (−3 to 82) | |
| Hispanic | 0 (−10 to 12) | |
| Insurance type | ||
| Commercial/military/Tricare | Reference | |
| Medicaid | −5 (−15 to 6) | |
| Parking or transportation aid received | ||
| None | Reference | |
| Parking aid | 19 (4 to 35)* | |
| Transportation aid | 12 (−3 to 30) | |
| Parking and transportation aid | 19 (4 to 37)* | |
| Cardiac surgery during COVID‐19§ pandemic | ||
| Before COVID‐19 | Reference | |
| During COVID‐19 | −12 (−21 to −2)* | |
Regression R 2=0.64. LOS indicates length of stay; and STAT, Society of Thoracic Surgeons–European Association for Cardio‐Thoracic Surgery.
Statistical significance at P<0.05. Coefficient interpretation: infants who experienced a delay in medical equipment delivery or prior authorization had a 31% (95% CI, 3%–66%) higher average postoperative LOS compared with those without discharge barriers.
Interpretation: Discharge barrier categories account for 21.2% of the explained variation in postoperative LOS.
The STAT category is a previously validated score to stratify surgical mortality risk.
SARS‐CoV‐2 pandemic of 2019 (COVID‐19).
Associations between equipment teaching, parental preference for extended stay, social determinants, and postoperative LOS did not reach statistical significance. Although other clinical factors, such as STAT category, postoperative complications, equipment needs at discharge, chromosomal, noncardiac or genetic anomaly, and poor postoperative weight gain were associated with longer postoperative LOS, associations between race and ethnicity, insurance type, and postoperative LOS did not reach statistical significance in the adjusted model (Table 3).
Contributors to Variation in Postoperative LOS
The Shorrocks‐Shapley decomposition estimated that discharge barrier category and STAT category accounted for similar, and collectively the highest, amounts of explained variation in postoperative LOS (21.2% [95% CI, 14.9%–27.0%] and 22.2% [95% CI, 17.3%–26.8%], respectively). Sociodemographic factors accounted for the least amount of explained variation, 6.9% (95% CI, 5.3%–10.8%; Table 3).
Sensitivity Analyses
Sensitivity analyses 1 and 2 yielded similar results to the main model (Tables S1 and S2). In sensitivity analysis 3, regression coefficients for each discharge barrier category were higher than in the main model, suggesting that postoperative LOS yielded more conservative estimates compared with total LOS (Table S3). In sensitivity analysis 4, stratification of the regression into low and high surgical complexity populations yielded an even greater contribution of discharge barriers to postoperative LOS relative to STAT category (Table S4).
DISCUSSION
In this analysis, we demonstrate that discharge barriers have a significant influence on variation in postoperative LOS for infants requiring congenital heart surgery. “Anatomy is destiny” is a common adage to describe the drivers of CHD outcomes, 24 , 25 , 26 yet we find that discharge barriers explain as much variation in postoperative LOS as STAT category, a measure of surgical or anatomic complexity. Although STAT category is often not modifiable, many discharge barriers might be: supervised overnight stays could be scheduled further in advance of planned discharge if the need for one is identified early in the patient's clinical course. Proactive identification of patient caregivers could prevent delays in the completion of discharge education, especially if certain caregivers are only available for teaching in brief intervals or teaching can only occur during daytime hours on a weekday. Health care providers can prescribe home medications or order home supplies once they are indicated rather than waiting until the patient is near discharge, thereby mitigating delays because of prior authorizations or logistics of supply delivery, such as delivering supplies before a weekend. This highlights a clear direction for future research: the associations between patient‐level factors and discharge barriers. Nonetheless, approximately one‐quarter of this cohort experienced at least 1 discharge barrier, which suggests that future interventions that reduce or remove these barriers to discharge could reduce postoperative LOS.
Similar to previous studies, our analysis found clinical contributors to LOS, including postoperative complications, equipment needs at discharge, and poor postoperative weight gain. Foote et al used clinical registry data from the National Pediatric Cardiology Quality Improvement Collaborative database to demonstrate that among patients following stage 1 single‐ventricle palliation, those surgical centers with the lowest LOS were more likely to care for patients with fewer additional cardiac operations and who were discharged home with nasogastric rather than surgical feeding tubes. 27 Using the same clinical registry, Baker‐Smith et al similarly found that among patients who underwent stage 2 single‐ventricle palliation, those with the longest LOS were more likely to have a cardiac reoperation, have a postoperative complication, or need a nonoral method of nutrition. 28
Other studies have leveraged administrative, rather than clinical registry, data, such as the Healthcare Cost and Utilization Project Kids Inpatient Database and the Pediatric Health Information Systems database, to demonstrate that social determinants, such as race and individual and neighborhood‐level socioeconomic status, are associated with increased LOS. 29 , 30 Yet, our analysis did not detect differences in LOS based on these social determinants. We suspect that this is because of the use of more clinically granular variables that are not available in administrative data sets (ie, discharge barriers). In other words, discharge barriers are a reflection of social determinants: patients of lower socioeconomic status may have longer postoperative LOS because they experience more discharge barriers, and therefore a model that includes discharge barriers may no longer demonstrate statistical significance for socioeconomic factors.
The impacts of reducing discharge barriers for the population with CHD is an area that warrants significantly more study. Madsen et al also used highly granular, single‐center data of patients in a cardiac care unit (ie, patients who did and did not require cardiac surgery) and identified similar discharge barriers to those identified in our study: caregiver unavailability for discharge teaching, picking up medications, or transportation home. 31 Although they did not investigate the time necessary to achieve discharge readiness, they found that the time after discharge readiness was achieved (ie, discharge within 2 hours of achieving discharge readiness) was only associated with a modest improvement in LOS. Based on these findings and our analysis, we hypothesize that (1) mitigating discharge barriers is more impactful for the surgical than the nonsurgical patient population, and (2) the time to achieve discharge readiness has a greater impact on LOS than the time to discharge after a patient is deemed ready. Together, our analysis and the analysis of Madsen et al highlight the immense potential for an understudied area of CHD research pertaining to discharge barriers. Such research could span evidence‐based interventions (eg, identification of surgery‐specific discharge barriers and development of resultant quality improvement initiatives), to the implementation of quality improvement initiatives with quantitative analysis on the impact on local outcomes, and, finally, to prospective dissemination studies across surgical centers.
Limitations
A key strength of this analysis was the leverage of highly granular data to better understand LOS; however, this also introduces the limitation of reliance on the electronic health record documentation of discharge barriers. It is possible that other patients with shorter LOSs had discharge barriers present that we were unable to capture, an example of exposure misclassification, or that our discharge barrier categories mask heterogeneity following adjustment with covariates. To mitigate these limitations of the exposure variable, we adjusted for social determinants of health and clinical factors that could confound the relationship between discharge barriers and LOS. This retrospective study is also limited to a single cardiac center, with the impact of discharge barriers potentially being unique in our setting compared with others. Although other studies using national clinical registry data that did not contain discharge variables as a variable estimated similar clinical associations with LOS, 18 , 27 , 28 the observational nature of this study still limits the generalizability and the ability for causal inference. Future prospective studies may provide stronger evidence for quality improvement initiatives that lead to significant changes in postoperative outcomes. Finally, this study contained data from 2015 to 2021. These data may not account for post–COVID‐19 changes in operations and their influence on discharge barriers.
CONCLUSIONS
Discharge barriers are as influential as surgical and anatomic complexity in explaining variation in postoperative LOS among infants undergoing congenital heart surgery. Although surgical and anatomic complexity is often not modifiable, many discharge barriers are. Therefore, future research should aim to understand the causal effects of mitigating discharge barriers on postsurgical outcomes, which will inform the development and implementation of clinical interventions to mitigate discharge barriers.
Sources of Funding
This work was supported by the Stanley Manne Children's Research Institute, affiliated with Ann & Robert H. Lurie Children's Hospital of Chicago (AWD001780; principal investigator [PI]: Woo). Drs. Woo and Kan also receive salary support from the National Institutes of Health (K01HL171819, PI: Woo; K23HL157615, PI: Kan).
Disclosures
None.
Supporting information
Tables S1–S4
STROBE Checklist
Acknowledgments
Drs. Zofia Zdanowicz, Joyce L. Woo, and Christina Laternser conceptualized and designed the study and data collection instruments, collected data, conducted formal analysis, drafted the initial manuscript, and critically reviewed and revised the manuscript. Dr. Colin Crilly drafted the initial manuscript, interpreted the analysis, and critically reviewed and revised the manuscript. Drs. Amy S. Lay and Kristin Kan conceptualized and designed the study, interpreted the analysis, and critically reviewed and revised the manuscript. All authors approved the final manuscript as submitted and agree to be accountable for all aspects of the work.
Supplemental Material is available at https://www.ahajournals.org/doi/suppl/10.1161/JAHA.125.047918
This work was presented as an oral presentation at the American Academy of Pediatrics National Conference, September 26 to 30, 2025, in Denver, CO.
This manuscript was sent to Daniel E. Clark, MD, MPH, Associate Editor, for review by expert referees, editorial decision, and final disposition.
For Sources of Funding and Disclosures, see page 8.
References
- 1. Arth AC, Tinker SC, Simeone RM, Ailes EC, Cragan JD, Grosse SD. Inpatient hospitalization costs associated with birth defects among persons of all ages—United States, 2013. MMWR Morb Mortal Wkly. 2017;66:41–46. doi: 10.15585/mmwr.mm6602a1 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2. Hoffman JI, Kaplan S. The incidence of congenital heart disease. J Am Coll Cardiol. 2002;39:1890–1900. doi: 10.1016/s0735-1097(02)01886-7 [DOI] [PubMed] [Google Scholar]
- 3. CDC . Congenital heart defects (CHDs): data and statistics . 2024. Accessed June 19. Available from: https://www.cdc.gov/heart‐defects/data/index.html#cdc_data_surveillance_section_5‐cost.
- 4. Pickles DM, Keller K. The cost of complex congenital heart disease in the US . 2024.
- 5. Simeone RM, Oster ME, Cassell CH, Armour BS, Gray DT, Honein MA. Pediatric inpatient hospital resource use for congenital heart defects. Birth Defects Res A Clin Mol Teratol. 2014;100:934–943. doi: 10.1002/bdra.23262 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6. Pasquali SK, He X, Jacobs ML, Shah SS, Peterson ED, Gaies MG, Hall M, Gaynor JW, Hill KD, Mayer JE. Excess costs associated with complications and prolonged length of stay after congenital heart surgery. Ann Thorac Surg. 2014;98:1660–1666. doi: 10.1016/j.athoracsur.2014.06.032 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7. Butler SC, Rofeberg V, Wypij D, Ferreira R, Singer J, Stopp C, Wood L, Ware J, Newburger JW, Sadhwani A. Inpatient screening for early identification of developmental risk in infants with congenital heart defects. J Pediatr. 2023;263:113687. doi: 10.1016/j.jpeds.2023.113687 [DOI] [PubMed] [Google Scholar]
- 8. Newburger JW, Wypij D, Bellinger DC, du Plessis AJ, Kuban KC, Rappaport LA, Almirall D, Wessel DL, Jonas RA, Wernovsky G. Length of stay after infant heart surgery is related to cognitive outcome at age 8 years. J Pediatr. 2003;143:67–73. doi: 10.1016/S0022-3476(03)00183-5 [DOI] [PubMed] [Google Scholar]
- 9. Ungerleider RM, Bengur AR, Hart EM, Miller CE, Lockwood NW, Knauss SA, Jaggers J, Sanders SP, Greeley WJ. Risk factors for higher cost in congenital heart operations. Ann Thorac Surg. 1997;64:44–49. doi: 10.1016/s0003-4975(97)00503-1 [DOI] [PubMed] [Google Scholar]
- 10. Brennan A, Gauvreau K, Connor J, Almodovar M, J DN, Banka P, Nathan M, Mathieu D, Kaza A, Mayer JJE. A method to account for variation in congenital heart surgery length of stay. Pediatr Crit Care Med. 2017;18:550–560. doi: 10.1097/PCC.0000000000001168 [DOI] [PubMed] [Google Scholar]
- 11. Pagowska‐Klimek I, Pychynska‐Pokorska M, Krajewski W, Moll JJ. Predictors of long intensive care unit stay following cardiac surgery in children. Eur J Cardiothorac Surg. 2011;40:179–184. doi: 10.1016/j.ejcts.2010.11.038 [DOI] [PubMed] [Google Scholar]
- 12. Pasquali SK, Jacobs ML, He X, Shah SS, Peterson ED, Hall M, Gaynor JW, Hill KD, Mayer JE, Jacobs JP. Variation in congenital heart surgery costs across hospitals. Pediatrics. 2014;133:e553‐e560. doi: 10.1542/peds.2013-2870 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13. Sacristán JA, Dilla T. No big data without small data: learning health care systems begin and end with the individual patient. J Eval Clin Pract. 2015;21:1014–1017. doi: 10.1111/jep.12350 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14. Congenital heart surgery public reporting. Society of Thoracic Surgeons. Accessed June 19. Available from: https://publicreporting.sts.org/chsd; 2025. [Google Scholar]
- 15. Woo JL, Gotteiner N, Anderson BR, Yee LM, Grobman WA, Davis MM, Laternser C. Prenatal diagnosis mediates the relationship between disease complexity and intervention timing for congenital heart defects. Pregnancy. 2025;1:e70011. doi: 10.1002/pmf2.70011 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16. Tseng SY, Truong VT, Peck D, Kandi S, Brayer S, Jason DP III, Mazur W, Hill GD, Ashfaq A, Goldstein BH. Patent ductus arteriosus stent versus surgical aortopulmonary shunt for initial palliation of cyanotic congenital heart disease with ductal‐dependent pulmonary blood flow: a systematic review and meta‐analysis. J Am Heart Assoc. 2022;11:e024721. doi: 10.1161/jaha.121.024721 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17. Kumar SR, Gaynor JW, Heuerman H, Mayer JE Jr, Nathan M, O'Brien JE Jr, Pizarro C, Subačius H, Wacker L, Wellnitz C. The society of thoracic surgeons congenital heart surgery database: 2023 update on outcomes and research. Ann Thorac Surg. 2024;117:904–914. doi: 10.1016/j.athoracsur.2024.03.018 [DOI] [PubMed] [Google Scholar]
- 18. Baker‐Smith CM, Wilhelm CM, Neish SR, Klitzner TS, Beekman RH III, Kugler JD, Martin GR, Lannon C, Jenkins KJ, Rosenthal GL. Predictors of prolonged length of intensive care unit stay after stage I palliation: a report from the National Pediatric Cardiology Quality Improvement Collaborative. Pediatr Cardiol. 2014;35:431–440. doi: 10.1007/s00246-013-0797-y [DOI] [PubMed] [Google Scholar]
- 19. O'Brien SM, Jacobs JP, Pasquali SK, Gaynor JW, Karamlou T, Welke KF, Filardo G, Han JM, Kim S, Shahian DM. The society of thoracic surgeons congenital heart surgery database mortality risk model: part 1—statistical methodology. Ann Thorac Surg. 2015;100:1054–1062. doi: 10.1016/j.athoracsur.2015.07.014 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20. Jacobs JP, O'Brien SM, Pasquali SK, Gaynor JW, Mayer JE Jr, Karamlou T, Welke KF, Filardo G, Han JM, Kim S. The Society of Thoracic Surgeons congenital heart surgery database mortality risk model: part 2—clinical application. Ann Thorac Surg. 2015;100:1063–1070. doi: 10.1016/j.athoracsur.2015.07.011 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21. Duan N. Smearing estimate: a nonparametric retransformation method. J Am Stat Assoc. 1983;78:605–610. doi: 10.1080/01621459.1983.10478017 [DOI] [Google Scholar]
- 22. Shorrocks AF. Inequality decomposition by factor components. Econometrica. 1982;50:193–211. doi: 10.2307/1912537 [DOI] [Google Scholar]
- 23. Wendelspiess Chávez Juárez F. SHAPLEY2: stata module to compute additive decomposition of estimation statistics by regressors or groups of regressors . 2015.
- 24. Athanasuleas CL, Stanley AW, Buckberg GD. Mitral regurgitation: anatomy is destiny. Eur J Cardiothorac Surg. 2018;54:627–634. doi: 10.1093/ejcts/ezy174 [DOI] [PubMed] [Google Scholar]
- 25. Ferrari VA, Fisher SA. Imaging the embryonic heart: how low can we go? How fast can we get? J Mol Cell Cardiol. 2003;35:141–143. doi: 10.1016/S0022-2828(02)00306-1 [DOI] [PubMed] [Google Scholar]
- 26. Sen DG, Mettler BA. Commentary: an idea with some muscle behind it. J Thorac Cardiovasc Surg. 2021;162:989–990. doi: 10.1016/j.jtcvs.2020.07.045 [DOI] [PubMed] [Google Scholar]
- 27. Foote HP, Thibault D, Gonzalez CD, Hill GD, Minich LL, Overbey DM, Tallent SL, Hill KD, McCrary AW. Center‐level factors associated with shorter length of stay following stage 1 palliation: an analysis of the national pediatric cardiology quality improvement collaborative registry. Am Heart J. 2023;265:143–152. doi: 10.1016/j.ahj.2023.08.003 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28. Baker‐Smith CM, Goldberg SW, Rosenthal GL. Predictors of prolonged hospital length of stay following stage II palliation of hypoplastic left heart syndrome (and variants): analysis of the National Pediatric Cardiology Quality Improvement Collaborative (NPC‐QIC) database. Pediatr Cardiol. 2015;36:1630–1641. doi: 10.1007/s00246-015-1208-3 [DOI] [PubMed] [Google Scholar]
- 29. Peterson JK, Chen Y, Nguyen DV, Setty SP. Current trends in racial, ethnic, and healthcare disparities associated with pediatric cardiac surgery outcomes. Congenit Heart Dis. 2017;12:520–532. doi: 10.1111/chd.12475 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30. Sengupta A, Bucholz EM, Gauvreau K, Newburger JW, Schroeder M, Kaza AK, Del Nido PJ, Nathan M. Impact of neighborhood socioeconomic status on outcomes following first‐stage palliation of single ventricle heart disease. J Am Heart Assoc. 2023;12:e026764. doi: 10.1161/JAHA.122.026764 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 31. Madsen NL, Porter A, Cable R, Hanke SP, Hoerst A, Neogi S, Brower LH, White CM, Statile AM. Improving discharge efficiency and charge containment on a pediatric acute care cardiology unit. Pediatrics. 2021;148:e2020004663. doi: 10.1542/peds.2020-004663 [DOI] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Tables S1–S4
STROBE Checklist