Predicting vaccination levels without accurate or timely vaccination data

Knowing how many people are vaccinated against an existing or re-emerging threat is a key factor guiding public health decisions, but such information is often sparse or non-existent in many regions, according to researchers at Penn State. Now, in collaboration with a team at the World Health Organization, the researchers have developed a new method to estimate and predict regional measles vaccination coverage levels even when accurate or timely survey data on vaccination is not available. The method uses data that is routinely collected when potential measles cases present at clinics to model vaccination coverage and can be used to guide public health interventions to slow or prevent measles outbreaks.

A paper describing the research appeared recently in the journal Vaccine.

“The measles vaccine is highly effective, providing long-lasting protection from the disease, but we still have outbreaks, and the disease causes over 100,000 deaths each year worldwide because of disparities in vaccine distribution,” said Deepit Bhatia, graduate student in biology in the Eberly College of Science and at the Center for Infectious Disease Dynamics at Penn State and first author of the paper. “The Centers for Disease Control and Prevention recently reported over 1,300 confirmed cases in the United States for the first half of 2025 — the highest number in 33 years. Accurate information on vaccination levels is crucial to guide public health interventions but the sources we have for this information are imperfect.”

Researchers use two main sources for information on vaccination coverage. The Demographic and Health Surveys (DHS) collects health data at the household and individual level in 90 low- and middle-income countries. Formerly funded by the United States Agency for International Development (USAID) the program is considered the gold-standard for accuracy, according to the researchers. These surveys are expensive and time-consuming to perform and are therefore only produced every three to five years. Outside of these large-scale surveys, countries produce administrative vaccinations coverage estimates based on the number of vaccine doses administered to a certain age group in the region. These administrative estimates are produced more frequently, but they are not as accurate as the DHS and can be biased.

“The DHS produce amazing data, but it’s analogous to U.S. Census data in that it is only done every few years,” said Matt Ferrari, director of the Penn State Center for Infectious Disease Dynamics, professor of biology and leader of the research team. “The census is done every 10 years and takes two years to complete. By the time it’s done, it’s out of date. But it’s too expensive to do more frequently. This is how vaccination coverage has been evaluated, particularly in low- and middle-income countries, places where these diseases have the largest impact.”

When developing their new method, the researchers said they wanted to find a way to split the difference between the highly accurate but expensive and potentially out-of-date surveys, and the more timely but less accurate administrative coverage estimates. They built a model using data that is routinely collected when patients are treated for potential measles cases at clinics. They used the mean age of the patients, their vaccination status as reported when at the clinic and whether the suspected cases actually were measles, rather than another disease with similar symptoms.

“We know that these measures are associated with vaccination coverage levels,” Bhatia said. “For example, in regions with high vaccination levels, young children are less likely to come in contact with the disease and the mean age of cases at the clinic will be higher.”

The research team used the three indicators as predictors to train a regression model that could best predict the gold-standard DHS data. Importantly, they withheld the most recent DHS data to later use as a stronger test of the predictive power of their method. They then used their model to predict vaccination coverage for the period covered by the latest DHS data and found that it was highly correlated.

“We found that the predictions of our method fit better with the DHS data than the administrative vaccination coverage estimates did,” Bhatia said. “Since our method uses routinely collected information that is readily available to researchers and public health officials, it provides a cheap and more easily accessible methodology to estimate vaccination coverage for a region that can be done quickly and can help inform policy in a timelier way.”

Recent changes to funding for the DHS have increased the relevance of the new method, according to Ferrari.

“Although this wasn’t the case when we began this research, the DHS program is currently on pause,” Ferrari said. “DHS was primarily funded by USAID, and we don’t know when or if they will be started again. Our method can hopefully help provide a stopgap.”

In addition to Ferrari and Bhatia, the research team included Natasha Crowcroft, Sébastien Antoni, M. Carolina Danovaro-Holliday, Anindya Sekhar Bose, Anna Minta and Balcha Masresha at the World Health Organization. The Bill & Melinda Gates Foundation and an Ecology and Evolution of Infectious Disease award, funded by the U.S. National Science Foundation, the National Institutes of Health and the National Institute of Food and Agriculture, supported the research. The Penn State Huck Institutes for the Life Sciences provided additional support.

At Penn State, researchers are solving real problems that impact the health, safety and quality of life of people across the commonwealth, the nation and around the world.

For decades, federal support for research has fueled innovation that makes our country safer, our industries more competitive and our economy stronger. Recent federal funding cuts threaten this progress.

Learn more about the implications of federal funding cuts to our future at Research or Regress.