Quantifying under-reporting of pathology tests in Medical Benefits Schedule claims data

Abstract

Objective We investigated the completeness of recording of pathology tests in Australian Medical Benefits Schedule (MBS) claims data, using the example of the prostate-specific antigen (PSA) test. With some exceptions, MBS claims data records only the three most expensive pathology items in an episode of care, and this practice (‘episode coning’) means that pathology tests can be under-recorded. Methods The analysis used MBS data for male participants in the 45 and Up Study. The number and cost of items in each episode of care were used to determine whether an episode contained a PSA screening test (Item 66655), or could have lacked a record of this item because of episode coning. Results MBS data for 1 070 392 episodes involving a request for a pathology test for 118 074 men were analysed. Of these episodes, 11% contained a request for a PSA test; a further 7.5% may have been missing a PSA request that was not recorded because of episode coning. Conclusions It is important to consider under-reporting of pathology tests as a result of episode coning when interpreting MBS claims data. Episode coning creates uncertainty about whether a person has received any given pathology test. The extent of this uncertainty can be estimated by determining the proportion of episodes in which the test may have been performed but was not recorded due to episode coning. What is known about the topic? Medical Benefits Schedule (MBS) claims data are a key resource for Australian health researchers. What does this paper add? We investigated a feature of MBS claims data known as episode coning, which may cause some pathology tests to be under-reported. Using the example of requests for PSA tests, we estimated the uncertainty in the amount of use of PSA tests introduced by episode coning. What are the implications for practitioners? Researchers using MBS data to identify use of specific pathology tests need to be aware of under-reporting caused by episode coning, and to estimate and report the uncertainty that this introduces into their findings.