logo ecco
Portal

ECCO'24 Abstracts

ECCO'24 Abstracts on the JCC Website

ECCO'24 Abstract Book PDF Version

ECCO'24 Abstracts COI Disclosure

Poster presentations: Epidemiology 2020

P802 Probability of vedolizumab response as defined by clinical decision support tool is associated with lower healthcare utilisation in patients with Crohn’s disease

P. Dulai1, Y. Wan2, Z. Huang2, M. Luo2

1Department of Gastroenterology, University of California San Diego, San Diego, USA, 2Gastroenterology, Takeda Pharmaceuticals, Deerfield, USA

Background

A clinical decision support tool (CDST) has been developed and validated for predicting treatment effectiveness of vedolizumab (VDZ) in Crohn’s disease (CD). We evaluated this tool’s ability to predict healthcare resource utilisation using real-world data.

Methods

CD patients treated with VDZ were identified using real-world data from the Optum (n = 358) and Truven MarketScan (n = 1445) databases. The full CDST with 3 clinical variables (prior bowel surgery, prior fistulising disease, prior TNF-antagonist exposure) and 2 lab values (albumin and CRP) was applied to the Optum dataset with available lab data, and patients were stratified into low (n = 27), intermediate (n = 152), or high (n = 179) probability of response groups. A modified weighting of the 3 clinical variables was created for application in Truven MarketScan due to the lack of availability of lab data. To assess the coherence with the full 5-variable CDST, the modified 3-variable CDST was used to stratify Optum patients into low (n = 137) or high (n = 221) probability of response, and then subsequently validated in the Truven MarketScan dataset (low, n = 510; high n = 935; response probability groups). Rates of hospitalisation, surgery, emergency department (ED) visits and annualised per patient costs were compared across response probability groups.

Results

Using the full 5-variable CDST, we observed a linear trend of significantly lower rates of ED visits, hospitalisation, surgery, and healthcare-related costs with increasing probability of response (Table). These results were consistent when using the modified 3 variable CDST in the Optum patient dataset where we observed a 3-fold higher rate of per patient healthcare costs in the low probability of response group compared with the high probability of response group ($6535 vs. $1900, p = 0.016). In the Truven patient dataset, the modified 3 variable CDST was also associated with a significantly lower rate of hospitalisation, surgery, and healthcare costs in the high probability of response group (Table). The majority of cost differences between high and low response groups was driven by costs related to hospitalisation and surgery.

Conclusion

A simple CDST can identify CD patients treated with VDZ who present a higher risk for healthcare resource utilisation, especially for hospitalisation and surgery. This tool could be integrated into population health monitoring algorithms using real-world data and offer flexibility on requirements of lab data.