2367. Selecting Testing Frequency for Estimation of Incubation Periods: A Simulation Study Based on Clostridioides difficile Observations

Abstract

BackgroundEstimates of the incubation period (time between pathogen transmission and symptom onset) for an infection inform infection control and prevention measures. However, observation of the exact transmission and onset times rarely occurs and “coarse,” or doubly interval-censored, data about these exact times are typically used for estimation. The effect of coarseness on the required number of symptomatic cases and the uncertainty of the estimates is unknown, prompting a simulation study informed by data from an investigation of the incubation period of Clostridioides difficile.MethodsWe simulated incubation period data assuming a log-normal distribution, a true median incubation period of 7 days, and a standard deviation of 1 day for sample sizes of 50 to 300 symptomatic cases. For each sample size, we simulated 1000 datasets and examined the impact of testing frequencies, considering intervals between tests of 0.25 to 2 times the median incubation period (1.75 to 14 days) about both transmission and symptom onset times. With these doubly interval-censored observed values, we fit accelerated failure time models to estimate the median incubation time and its 95% confidence interval (CI). Comparing the coverage of the true median and the widths of the CIs, we summarized simulation results across sample sizes and testing frequencies.ResultsModel results from all combinations of sample sizes and testing frequencies yielded median incubation period CIs close to the target 95% coverage level (Figure 1). The width of the 95% CI about the median decreased with larger sample sizes and shorter times between tests (Figure 2). Thus, similar estimates and confidence intervals would be observed from 100 symptomatic cases with a testing frequency of 3.5 days as from 200 symptomatic cases tested every 14 days.ConclusionThe frequency of testing is a key factor in planning studies to estimate incubation periods for infectious diseases. To achieve a desired degree of certainty in estimation, increased frequency of testing can reduce the number of symptomatic cases required. We showed that simulations can assist in planning natural history studies, and these methods could be extended to include population data (e.g., transmission incidence) and cost constraints. Disclosures All authors: No reported disclosures.