(161) Impact of population opt-outs on rare disease prevalence: a simulation study

Publication Number: 1156

Background: Prevalence is one criteria that may determine orphan designation. Population opt-outs are discussed as a way to gain implied consent for the secondary use of routinely collected health data. People who opt-out may be unrepresentative of the general population in terms of demographics and chronic disease, with such populations often assumed to be healthier. In contrast, patients with rare diseases may see the value of data sharing to support research into their condition.

Objectives: To estimate the impact of different population opt-out rates on the estimated prevalence of a rare disease under the assumption that people who opt out are healthy.

Methods: Setting: Simulation study.
Outcome: The estimated prevalence (EP) of a hypothetical rare disease is calculated in the presence of different opt-out rates, where the true prevalence (TP) ranges from 3.75 to 4.99 per 10,000. Rare disease prevalence is estimated for opt-out rates ranging from 1% to 20% of the population, under the assumption that people who opt-out are healthy and do not have the rare disease.
Statistical analysis: EP is calculated using the number of people with each rare disease in the sample size (based upon true prevalence), divided by the sample size*(1-the opt-out rate). Sample sizes of 1 million and 5 million will be examined 95% confidence intervals for the proportion are calculated using the Wilson score method without continuity correction. The EP and its 95% confidence interval is then compared to a 5 per 10,000 threshold used as one criteria for orphan designation in Europe.

Results: Using a sample size of 1million, the level at which TP and 95%CI are below 5 per 10,000 is 4.50 (95%CI 4.10 to 4.94). The opt-out rate at which the 95%CIs of the EP cross the 5 per 10,000 threshold is 2% (EP 4.59, 95%CI 4.18 to 5.04), and 18% when both EP and 95%CIs are greater than 5 per 10,000 (EP 5.49, 95%CI 5.00 to 6.02). When TP is 4.25 (95%CI 3.86 to 4.67) per 10,000, the opt-out rate when the 95%CIs of the EP cross the 5 per 10,000 threshold is 7%, and 17% when TP is 3.75 per 10,000. Using a sample size of 5 million the opt-out rate at which the 95%CIs of the EP cross the 5 per 10,000 threshold is 2% for TP of 4.75 per 10,000, 7% for TP of 4.5 per 10,000, and 12% for TP of 4.0 per 10,000.

Conclusions: Population opt-outs for the secondary use of health data may bias rare disease prevalence estimates if the population that opt out are healthy. The degree and relationship to pre-specified thresholds may depend upon the opt-out rate, the sample size used to estimate prevalence, and the true prevalence of a rare disease.