(299) Assessing the agreement between open and closed medical claims in a rare disease cohort

Background: Traditionally, administrative claims data sources have only included closed claims, which are those that sourced from health insurance providers. Closed claims are an attractive source of information because any patient healthcare encounter during a patient’s enrollment period is expected to be captured at some point. Recently, open claims data sources have become available, which are those sourced from providers, billing management systems, and claims clearinghouses. Unlike closed claims, encounters with providers that do not use the systems included in the open claims data source are not captured in open claims. In the setting of rare diseases, open claims may be useful to augment closed claims data sources, however, concerns about completeness and differential capture remain.

Objectives: To assess the correspondence between closed and open claims in a cohort of patients with a rare disease.

Methods: We used a confirmed cohort of patients with a rare bone disease in the US to identify patients. Patients were linked via deidentified tokens to open and closed claims data sources from the HealthVerity Marketplace. Medical claims during patient enrollment periods between January 1, 2018 and December 31, 2022 were included in the analysis. For each closed claim, we determined whether there was an open claim with the same diagnosis code and a) the same start date, b) overlapping service period, and c) service period that overlap within one week of start date of the index claim. The same analysis was conducted in reverse, identifying matching closed claims for each open claim. The analyses were repeated using only adjudicated closed claims and, separately, requiring the place of service or bill type code to match.

Results: A total of 66 patients contributed 57,452 closed claims, 55 of whom contributed 13,909 open claims. Overall, 24%, 39%, and 44% of closed claims had an open claim with the same start date, overlapping service periods, and a service period within one week of the closed claim start date, respectively. Additionally, 74%, 77%, and 79% of open claims had similarly matching closed claims. The proportion with corresponding claims within one week of the claim start date was considerably lower when requiring matching place of service or bill type codes.

Conclusions: More than 75% of closed claims did not have an exactly matching open claim, and more than 50% did not have a temporally close open claim. Additionally, nearly 25% of open claims did not have a temporally close closed claim. These results suggest that open claims under capture events when used on their own but may add additional information when used in conjunction with closed claims. However, because not all closed claims patients contribute open claims, differential capture by patient may be a source of bias.