Abstract 1572P
Background
Using "big data" from real-world evidence (RWE) among cancer regimen-related adverse events is an opportunity by using techniques and tools from information gathering. We show that RWE is useful in predicting the outcome of opioid use for chemotherapy-induced ulcerative mucositis (CTUM) in cancer patients hospitalized for treatment.
Methods
To find CTUM and the opioid intervention, we used the National Inpatient Sample database in the United States. The total charge and length of stay (LOS) in the hospital were evaluated outcomes. The propensity score (PS) model was built between variables, treatment assignment, and outcomes using PS estimation. PS was calculated using computational methods: tree-based machine learning, regression-based approaches, and covariate balancing. Treatment effects were estimated using PS weighting.
Results
We identified 365 opioid users among 11,765 CTUM patients hospitalized for cancer treatment. Estimating treatment effects (TE) with PS weighting showed that CTUM patients with opioid use had a total charge of 46243 USD and LOS of 5.8 days compared to those without opioids: 89622 USD and 9.2 days. The average treatment effect on the treated (ATT) is -0.95, standard error (SE) of 0.16 with a 95%CI of (–1.27) –(–0.64). The final estimate and the SE are similar to those with a single imputed dataset and the five imputed datasets. Doubly robust estimation of TE with PS weighting is similar to estimates obtained with non-robust estimators, either single or multiple data sets.
Conclusions
Observational data from RWE can be used to assess different treatment approaches in the absence of a clinical trial. The potential to use "big data" from RWE in the context of cancer regimen-related adverse events will lead to the expansion of methodologies and systems for data gathering to improve cancer research outcomes. Additional comparative efficacy research on CTUM patients is currently being investigated.
Clinical trial identification
Editorial acknowledgement
Legal entity responsible for the study
The authors.
Funding
Has not received any funding.
Disclosure
All authors have declared no conflicts of interest.