Return To Abstract Listing

Scientific Session 23 — SS23: Efficacy/Administration/Informatics - Administration and Appropriateness

Thursday, May 9, 2019

Abstracts 1183-2994



1670. Medicare Referral Leakage Among Radiology Groups

Pyrros A1*,  Nikolaidis P2,  Siddiqui N1,  Garg N3,  Flanders A4 1. Dupage Medical Group, Downers Grove, IL; 2. Northwestern University, Chicago, IL; 3. MD Anderson Cancer Center, Houston, TX; 4. Thomas Jefferson Unversity, Philadelphia, PA

Address correspondence to A. Pyrros (ayis@ayis.org)

Objective: Referral or patient leakage is the concept that patients may go out of network or "leak" for services to other groups or networks. This phenomemnon can be thought of as how many referrals a radiology group is receiving and how many referrals they are losing out of their potential market. The purpose of this study was to calculate "leakiness" nationwide with a publicly available Medicare dataset called DocGraph and open-source analytical tools.

Materials and Methods: The radiology referral network was derived from the DocGraph dataset (2015), containing over 1 million nodes. The dataset describes services billed between two CMS providers sharing a unique beneficiary within a window of time. Files were imported into Postgres (v9.6) and National Provider Identifiers were used to select providers. Referral pairings were weighted with the CMS public use file and the percentage of office visits. Radiology groups were then grouped by the billing provider's telephone numbers. Analysis was performed with R (v3.4.1) and Gephi (v0.9.1).

Results: In total, 1,116,414 referrer-to-radiologist pairs were identified, and the leakage was calculated for 8,288 radiology groups. Referral leakage had a mean of 84%, median of 91%, standard deviation of 18.37, estimated skewness of -2.27, and estimated of kurtosis 8.66. Average node degree in the referral network was 7.55, and average weighted degree was 3.77. Subset analysis of the top ten groups by estimated actual and estimated potential referrals was performed.

Conclusion: Referral leakage was remarkably high among radiology groups, with an average of 84%, whereas it is estimated to be between 24-30% and 55-65% in other publications. Subset analysis of the top ten groups by estimated actual and estimated potential referrals demonstrated an interesting trend, in which the largest estimated potential referral groups were dominated by tertiary hospital systems, and the largest estimated actual referrals by large private radiology groups. There was a statistically significant difference between the two top ten groups’ referral leakage (p value <0.01). Larger private groups have lower leakage than larger tertiary referral centers. In addition, there is extensive national variation in patient leakage, which should be investigated by all radiology groups as a potential business opportunity.