(BSPOT38) Estimating period prevalence: comparing methods to describe a rare disease population

Background: In the rare disease setting, estimates of prevalence may be unstable due to a number of factors such as the limited sample size, sensitivity to truncation, censoring due to frequent changing of insurance, or late entry to the cohort. Methods used to define period prevalence may be impacted by these factors in different ways, which leads to varying and inconsistent estimates across studies. A standard approach to period prevalence estimation is needed to facilitate evidence synthesis to better understand the clinical characteristics of these populations.

Objectives: To compare three approaches to period prevalence estimation of rare and common conditions in a rare disease cohort.

Methods: We utilized data from a confirmed cohort of patients with a rare bone disease in the US linked via deidentified tokens to open and closed claims data sources from the HealthVerity Marketplace. The period prevalences of dermatologic disorders (which is common in the population) and chronic kidney disease (which is uncommon in the population) between January 1, 2018 and December 31, 2022 were estimated using three different methods. Method 1: the number of patients with a history of the condition among those with any enrollment in the closed claims during the period, using the total number of patients with any enrollment during the period as the denominator to facilitate simple calculation. Method 2: the same numerator as method 1, but with the average number of patients enrolled during the period as the denominator, computed as the total person-time accrued during the period divided by the length of the period, thus accommodating variable follow-up times. Method 3: an estimator that combines the point prevalence at the start of the period and a Kaplan-Meier estimate of the cumulative incidence of the events that occurred throughout the period that accounts for left truncation and censoring.

Results: A total of 66 patients were enrolled during the study period, with an average population of 56.6 patients throughout the study. The period prevalence of dermatologic disorders during the study period was estimated to be 51.5%, 60.1%, and 58.3% using methods 1, 2, and 3, respectively. The period prevalence estimates for chronic kidney disease using methods 1, 2, and 3 were 4.5%, 5.3%, and 5.2%, respectively.

Conclusions: The period prevalence estimates differed substantially depending on the method used, especially for a condition that is common in the study population. Due to this variability, there is a clear need for a well-defined, interpretable definition of period prevalence to allow comparisons between studies and across populations. Ideally, such a definition will properly account for variable follow-up times, censoring, and truncation, such as Methods 2 and 3.