Targeted testing reaching towards people

COVID-19 Insights: How Accurate Testing Can Guide Us

Lena Kashyap in Health & Wellness November 2025 • 4 min read.

"Unlock the Secrets: Understanding the True Scope of Infection Through Better Testing Strategies"

In the face of a novel infection like SARS-CoV-2, the virus behind COVID-19, understanding the actual spread is a monumental challenge. Testing, while a crucial tool, only captures a fraction of the population, and this fraction isn't random. It's a selective subset, making it difficult to grasp the big picture accurately. We often lack complete knowledge about how accurate these tests are, and the initial understanding of a pandemic is frequently too hazy to lean heavily on complex, detailed models.

This is where the power of partial identification analysis comes in. It allows us to define the bounds of what's possible, to infer parameter values from imperfect data, using assumptions that are credible without needing to force statistical identifiability. The goal is to create a framework for analyzing disease prevalence, based on what we can reasonably ascertain about the tests we use, and how selective they are.

This article explores a general framework for analyzing how widespread a disease is by looking at how selective and sensitive our diagnostic tests are. We'll explore how to refine the worst-case estimates by setting limits for how sensitive and selective tests can be. These restrictions link easily to existing research and, unlike relying on predictive values alone, allow us to set realistic prior bounds without skewing our prevalence estimates. By applying these methods to data from the early stages of the COVID-19 pandemic, we discover insights that challenge earlier speculations, particularly regarding infection fatality rates.

Understanding Test Sensitivity and Selectivity

Targeted testing reaching towards people

Let's begin with the basics. Imagine trying to determine how prevalent an infection is within a group, armed only with the rate at which tests are conducted and the yield of those tests. We'll call the infection COVID-19, but the principles apply more broadly. C represents the true infection status (1 for infected), T indicates whether a person has been tested (1 for tested), and R represents the test result (1 for positive). Crucially, we only see R when T is 1.

The testing rate, denoted as ", is the probability of being tested (Pr(T = 1)), and the test yield, denoted as y, is the probability of a positive result given that a test was taken (Pr(R = 1|T = 1)). Both of these values are directly observable from the data.

Test Sensitivity: How well the test correctly identifies those who have the infection.
Test Selectivity: The degree to which testing targets specific groups within the population, for example, whether tests are more likely to be administered to those already showing symptoms.

Introducing bounds on sensitivity refines these estimates. Test sensitivity is a critical parameter in COVID-19 research, with ongoing efforts to measure and improve it. By setting a range for test sensitivity, we acknowledge the inherent uncertainty in test performance. Selectivity is equally important. If tests are primarily given to those with symptoms, the results will skew towards higher prevalence. Understanding and bounding this selectivity helps us correct for potential bias.

Navigating the Future with Better Data

The methods discussed provide a clearer, more nuanced understanding of infection prevalence by accounting for the selectivity and sensitivity of testing. This approach not only refines our estimates but also challenges us to think critically about the assumptions that underpin our models. As we continue to grapple with ongoing and future pandemics, improving our data interpretation tools will be essential for effective public health strategies.

About this Article -

This article was crafted using a human-AI hybrid and collaborative approach. AI assisted our team with initial drafting, research insights, identifying key questions, and image generation. Our human editors guided topic selection, defined the angle, structured the content, ensured factual accuracy and relevance, refined the tone, and conducted thorough editing to deliver helpful, high-quality information.See our About page for more information.

Everything You Need To Know

Why is it difficult to determine the actual spread of an infection like SARS-CoV-2, and how does testing contribute to this challenge?

Determining the actual spread of an infection like SARS-CoV-2 is difficult because testing only captures a fraction of the population, and this fraction isn't random. It's a selective subset of the population. The accuracy of these tests is not always completely known, and early data can be too uncertain to rely on complex models. This creates challenges in accurately assessing the true extent of the infection within the broader population, making it difficult to grasp the big picture accurately.

What is partial identification analysis, and how can it help address the challenges of imperfect data in understanding disease prevalence, particularly in the context of COVID-19?

Partial identification analysis helps define the bounds of what's possible when working with imperfect data. It enables the inference of parameter values from such data, using credible assumptions without forcing statistical identifiability. In the context of COVID-19, this approach helps create a framework for analyzing disease prevalence by considering what can be reasonably ascertained about the tests used, including their selectivity, and how they target different groups within the population.

How do test sensitivity and test selectivity affect the estimates of infection prevalence, and why is it important to consider both in analyzing COVID-19 data?

Test sensitivity, which is how well a test correctly identifies those who have the infection, and test selectivity, which is the degree to which testing targets specific groups within the population, both significantly affect estimates of infection prevalence. If tests are primarily given to those with symptoms, the results will skew towards higher prevalence. By considering and bounding both sensitivity and selectivity, it helps to correct for potential biases and refine prevalence estimates, providing a more nuanced understanding of infection spread.

How can setting limits or ranges for test sensitivity and test selectivity improve our understanding of disease prevalence, and what advantages does this approach offer compared to relying solely on predictive values?

Setting limits for test sensitivity and test selectivity refines estimates of disease prevalence by acknowledging the inherent uncertainty in test performance and accounting for potential biases in testing practices. This approach allows for setting realistic prior bounds without skewing prevalence estimates, unlike relying on predictive values alone. This also allows the application of research, and application of existing research.

Can you explain how the testing rate, test yield, true infection status, and test results are interconnected and what role they play in the framework for analyzing disease prevalence?

In the framework, 'C' represents the true infection status (1 for infected), 'T' indicates whether a person has been tested (1 for tested), and 'R' represents the test result (1 for positive). The testing rate (Pr(T = 1)) and the test yield (Pr(R = 1|T = 1)) are directly observable from the data. The testing rate and yield, combined with the understanding of test sensitivity and selectivity, allow for estimating the true prevalence of the infection. By analyzing these factors together, a more accurate understanding of the disease's spread can be achieved.