A modeling approach developed by the US Forest Service for detecting rare wildlife and repeated testing can help monitor the spread of COVID-19.
The ability to track an emerging infectious disease quickly and accurately within a population is a critical piece of the public health response; for example, mask mandates and other restrictions may be tied to COVID-19 prevalence within communities. A primary hurdle to achieving this community-level monitoring is the tradeoff between speed and accuracy in disease tests. In the case of COVID-19, rapid tests are more error prone and have a lower sensitivity for disease detection, which can lead to lower accuracy of prevalence estimates if these errors are not accounted for within models.
Modeling techniques developed for sampling wildlife may be able to address this problem. Because many wildlife species of concern are rare and difficult to find, wildlife scientists have spent decades developing tools to improve population estimates with imperfect detection.
Occupancy models are one such tool. These models address imperfect detection by collecting repeated observations of whether a species is present or absent at a site. These observations can then be used to model how detectable the species is – in other words, how likely it is to see a species at a site given it is actually there – which can then be used to improve overall estimates of the proportion of sites where species are present or absent.
When applied to disease monitoring, this translates to repeatedly testing individuals to determine the proportion of people within the population that are infected (prevalence), given imperfect detection from false negative tests and/or false positive tests.
Th Forest Service study, published in late March by BMC Public Health, included scientists from Rocky Mountain Research Stations in Flagstaff, Arizona, and Missoula, Montana. They used simulations to investigate whether occupancy modeling and repeated testing could overcome low test sensitivity in COVID-19 rapid tests to produce accurate estimates of disease prevalence more quickly and cheaply. A total of 108 scenarios were simulated, with varying values for disease prevalence, test sensitivity, proportion of the population initially tested, proportion of the population repeatedly tested, and number of repeat tests.
* Occupancy modeling and repeated testing can overcome the low sensitivity of rapid tests to provide accurate COVID-19 prevalence estimates. When occupancy models were applied with repeated testing, prevalence accuracy was similar for low-sensitivity rapid compared to higher-sensitivity, slower tests.
* Across all disease prevalence levels, the optimal sampling strategy in terms of both accuracy and cost involved initially testing 1% of the population and repeating the tests four additional times.
* These methods can help support real-time, informed decision making, even at low disease prevalence levels when decisive action is most meaningful.