RESEARCH APP

Calculator to account for test accuracy when estimating SARS Cov-2 prevalence

overview

This app is a tool for re-calculating the prevalence of SARS-CoV-2 in studies based on the accuracy of the test used. The output can help inform the true prevalence or seroprevalence of SARS-CoV-2. Additionally, the calculator can help educate on how the accuracy of SARS-CoV-2 tests affects our understanding of the prevalence of SARS-CoV-2 in different communities. 

To use this app, select a SARS-CoV-2 test used to estimate the prevalence of SARS-CoV-2 in the study of interest. Then, type in the sample size of the study and the upper and lower confidence bounds (see definitions below). In the bar chart, the calculator outputs the original study prevalence and respective confidence interval and the updated study prevalence and respective confidence interval accounting for test accuracy. The calculator also reports the sensitivity and specificity of the test. You can select the “Testing Data” tab to obtain details on the test selected.  This calculator assumes sensitivity and specificity are known. You can input your own sensitivity and specificity by selecting “User Specified Values” under Test Name. Data were last updated on December 22, 2020.

To see how to use, FAQs and definitions, click here.

mood-warning
Please view on a larger screen to interact with this feature.
What is the purpose of this app?

The purpose of this app is to calculate the optimal pool size of hierarchical or square matrix group testing. We have included three different pooling algorithms in our app, enabling users to compare various methods. 

Not all SARS-CoV-2 tests are created equal. This app is designed to calculate the true prevalence of SARS-CoV-2 given an original study population and SARS-CoV-2 test used to identify infection. Understanding the true prevalence can improve interpretation of study generalizability and limitations. 

The ability to rapidly identify people likely to be infected with SARS-CoV-2 both speeds infection control efforts and results in better-powered clinical trials and observational studies. This app is a tool for identifying groups of people likely to be infected with SARS-CoV-2, including asymptomatic and pre-symptomatic cases. The purpose of the app is to help institutional decision-makers and researchers compare SARS-CoV-2 testing strategies to determine who in a given population should be approached for testing or recruited for a study on SARS-CoV-2.

Who should use this app?

Laboratories that are interested in conducting group testing for COVID-19 should use this app to find the optimal pool size. Anyone interested in understanding the usage of group testing should use this app for reference. 

Anyone! This app may be used by researchers and the general public to understand SARS-CoV-2 prevalence better from study data and to understand the impact test accuracy has in understanding the penetrance of SARS-CoV-2 in our community.

This app may be used by decision-makers wishing to optimize transmission control effects through testing in settings where universal testing is not feasible. The method implemented by this app can apply to a variety of institutional settings, workplaces, or schools. In addition, it is often impractical to test every single individual in a randomized or observational study of SARS-CoV-2. Therefore, this app is also useful for researchers wishing to optimize statistical power in any study where people at higher risk of infection must be preferentially sampled.

What information do I need to use this app?
  • Prevalence is the proportion of the pooled population with COVID-19. (The user can select a number or a range.)
  • Assay Sensitivity is the probability that a truly positive individual is correctly classified as positive by a specific test. (Will be between 0 and 1)
  • Assay Specificity is the probability that a truly negative individual is correctly classified as negative by a specific test. (Will be between 0 and 1)
  • Maximum allowable dilution (MAD) is the maximum level of dilution of assay sensitivity that the user will allow, and is typically determined a priori. For example, if the user is comfortable with a 20% reduction in sensitivity due to pooling, but no more, then they should set MAD to 20%. Along with information on the length of the sensitivity window, and the exponential rate of increase of COVID-19 viral load early in acute infection, the MAD is used to determine the maximum allowable pool size with which the user is comfortable. (Given our model of viral dynamics, for example, a MAD of 20% yields a maximum allowable pool size of about 25.)

At minimum, you need an estimate of SARS-CoV-2 prevalence in a population and the name of a SARS-CoV-2 test that is used to obtain that estimate of prevalence. Additionally, you can incorporate information about the precision of the result if you have confidence intervals for the estimated study prevalence. If the study you’re examining has a test not listed in our app, you may use the calculator by providing estimates of the sensitivity and specificity of the SARS-CoV-2 test used in the study.

This app assumes that some testing has already been performed in the target population, either as part of a specific data collection effort (e.g. a random sample of employees at a company were tested in a study on workplace exposure to SARS-CoV-2) or a universal testing activity (e.g. at the beginning of the semester, all students at a university were tested for SARS-CoV-2). Use of the app requires data on characteristics that could be used to propose testing strategies that would prioritize individuals with certain characteristics for testing. For example, data from housing contracts can be used to estimate the testing efficiency and yield of a strategy that prioritizes on-campus students for SARS-CoV-2 testing.

What types of datasets can I upload?

This app is designed to be used with either of the following two types of datasets:

  1. Census/Convenience Sample – If the study sample represents a census of the target population, or a convenience sample nested in the target population, the dataset should contain one row per person in the target population. No sampling weights or clustering variables are required.
  2. Probability-Based Sample – If the study sample represents a probability-based sample of the target population, the dataset should include one row per person in the probability-based sample. If sampling probabilities among individuals differed (i.e., if the sample was not a simple random sample), include a weight variable. If clusters were used to identify individuals to test, include the cluster variable.

For more information about dataset file structure, click to the "How to use" tab above.

What can I learn by using this app?

From this app, one can find the optimal pool size for the different pooling algorithms available. Users can compare and contrast the optimal pool size, efficiency, positive predicted values, and total sensitivity of each pooling algorithm with varying prevalence. Furthermore, one can visualize the relationship between prevalence and optimal pool size, and prevalence and efficiency for different algorithms. 

This app is a tool for re-calculating the prevalence of SARS-CoV-2 in studies based on the accuracy of the test used. The output can help inform the true prevalence or seroprevalence of SARS-CoV-2. Additionally, the calculator can help educate on how the accuracy of SARS-CoV-2 tests affects our understanding of the penetrance of SARS-CoV-2 in different communities. 

This app can help decision-makers evaluate candidate SARS-CoV-2 testing strategies in terms of their efficiency and yield. Comparing the efficiency and yield of potential testing strategies can help stakeholders make informed decisions about which testing strategy or strategies to pursue, given available resources.

What do I do if the calculator outputs an error?

If you do not input an estimated prevalence or study population, the app will not be able to compute an adjusted prevalence and corresponding confidence interval. If the estimated prevalence is too low for a given study population, the calculator may not be able to calculate an adjusted prevalence if the sensitivity and specificity is low enough. Finally, some tests have sparse cells for the validation data, which may produce an error. You can check the validation data in the testing tab to diagnose if this is the case. Feel free to contact us if you have additional questions.

Limitation

This app currently has only one viral dynamic model as from (Pilcher, Westreich, Hudgens; JID 2020).

The efficiency, adjusted sensitivity (due to larger pool size), and positive predictive value outputs should be regarded as conservative estimates. The optimal pool size output should be regarded as approximate. Finally, we assume that if a sample with a given viral load can be detected in the master pool, it can be detected in any smaller pools.

  1. Identify the estimated study prevalence and type the value into the app.
  2. Select the type of test used to estimate the study prevalence in Step 1.
  3. If the test is not available in the drop-down, specify the sensitivity and specificity of the SARS-CoV-2 test used in the study.
  4. If a confidence interval for the study prevalence is available, type the value into the app.
No items found.