Abstract Body (Do not enter title and authors here): Introduction: Previous studies using national administrative data show that among patients who undergo congenital heart surgery, those with public insurance have longer length of stay (LOS) than those with private insurance. This reveals a knowledge gap that requires granular data to investigate: do discharge barriers explain this insurance-based difference in LOS?

Objective: To determine whether discharge barriers mediate, or explain, the association between insurance type and postoperative LOS.

Methods: Retrospective cohort study of all infants who received interventional catheterization or cardiac surgery between 0-60 days of life at Lurie Children’s Hospital of Chicago, 2015-2021. The exposure was insurance type (public versus private). The outcome was postoperative LOS (number of days between cardiac surgery requiring bypass and discharge). The mediator was whether a discharge barrier occurred (none versus at least one barrier). Our mixed-methods approach to defining discharge barriers has been previously described. In brief, barriers impeding discharge were manually extracted from electronic health records, then collapsed and ranked using regression. Example barriers included prior authorization or home equipment delivery delays. Additional covariates included weight at surgery, presence of syndrome or non-cardiac abnormality, discharge equipment needs (e.g., nasal cannula), postoperative intubation duration, postoperative complications, surgical complexity (STAT category), and a COVID-19 indicator. Linear probability and log-linear regressions with robust standard errors were used for mediation analysis.

Results: In total, 372 infants met inclusion; 30.9% had a discharge barrier. Median postoperative LOS was 27 (IQR 15-55) days for those with a discharge barrier versus 13 (IQR 8-22) days for those without (p<0.001). In multivariable models, those with public insurance had a 14.4 percentage-point higher probability of having a discharge barrier than those with private insurance (p=0.004). After adjusting for discharge barriers, insurance type was no longer associated with postoperative LOS (p=0.140), consistent with a mediating effect. In mediation analysis, presence of a discharge barrier explained 50% of the association between insurance type and postoperative LOS (indirect effect 4.9%, p=0.013, Figure 1).

Conclusion: Discharge barriers, a modifiable aspect of care, are a key driver of insurance-based differences in postoperative LOS.