Background: Early esophagogastroduodenoscopy (EGD) within 24 hours is recommended for non-variceal upper gastrointestinal bleeding (UGIB). However, the effect of early EGD on resource utilization remains unclear due to selection bias inherent in observational studies. We sought to identify the unbiased effects of early EGD on hospital charges and length of stay (LOS) using a causal inference framework.
Methods: We conducted a secondary data analysis using the 2019 National Inpatient Sample. We identified a weight sample of 198,015 non-elective adult admissions (≥18 years old) with a primary diagnosis of non-variceal UGIB requiring EGD. The intervention was defined as “early EGD” performed on hospital day 1 or 2 (serving as a proxy for EGD in 24 h). This group was compared against two control groups: (1) “delayed EGD” performed on day 3 or thereafter, and (2) “one-day delayed EGD” performed specifically on day 3. The estimand was the Average Treatment Effect on log-transformed hospital charges. We used a doubly robust estimator, Augmented Inverse Probability Weighting (AIPW), to adjust for covariates including demographics, socioeconomic status, including insurance, Elixhauser Comorbidity Index, bleeding severity indicators (transfusion, intubation), weekend admission, and hospital characteristics. Secondary outcomes included LOS and in-hospital mortality. Sensitivity analysis using E-values was performed to assess robustness against unmeasured confounding.
Results: The weighted cohort included 141,917 (71.7%) discharges in the early EGD group and 56,098 (28.3%) in the delayed EGD group. Unadjusted analysis showed the early EGD group had lower mean total charges (53,084 USD vs. 65,008 in the delayed EGD group) and shorter mean LOS (4.0 days vs. 5.8 days). The AIPW analysis demonstrated statistically significant reductions in resource utilization in the early EGD group. Compared to delayed EGD (day 3+), early EGD was associated with lower total hospital charges by an average of $8,515 USD (risk ratio 0.82, 95%CI [0.80, 0.83], p < .001) and shorter LOS by 1.73 days (95%CI [–1.86, -1.59], p < .001). When compared to delaying EGD by just one day (one-day delayed EGD group), early EGD was associated with lower charges by $3,745 USD and shorter LOS by 0.92 days (p < .001). There was no statistically significant difference in in-hospital mortality between the groups (risk difference 0.001, 95%CI [-0.001, 0.004], p = .29). The calculated E-value was 1.75.
Conclusions: Using a doubly robust estimator within a causal inference framework, our study confirms that early EGD for non-variceal UGIB is associated with significantly lower hospital costs and shorter length of stay without adversely affecting mortality rates. Extrapolating the $8,515 reduction to 56,098 delayed EGD cases implies a potential saving of approximately $478 million annually. These findings validate current guidelines and suggest investing in resources to ensure timely EGD access improves both cost-effectiveness and operational efficiency.