Skip to:

    • Sudipto Banerjee

Making Sense of Asthma-Related Hospitalizations


Using geographic information system technology to show patterns by place and time in California, the Fielding School’s biostatistics chair provides an invaluable tool for hospital administrators and epidemiologists.

THE DATABASE OFFERED FOR ANALYSIS by a California state health official to Dr. Sudipto Banerjee, then a professor at the University of Minnesota, represented a potential treasure trove of information for hospital administrators and epidemiologists: the records of asthma-related hospitalizations from each of California’s 58 counties, collected daily from 1991 to 2008.

Chart of asthma rates averaged by season
The model Dr. Sudipto Banerjee created with his colleagues reveals when spikes in asthma hospitalization tend to occur within each California county. (click graph to enlarge)

But as with many of the complex data sets tackled by biostatisticians, this one was “messy.” Daily records from multiple hospitals are considered too unwieldy and error prone for rigorous statistical analysis. So Banerjee, now professor and chair of the Department of Biostatistics at the Fielding School, proposed to work with his colleagues to create something new in the era of geographic information systems (GIS) – a method to aggregate the hospitalization data by month, per county, over the 18-year period, then convert to rates per 100,000 residents, as a way to better understand spatial and temporal patterns up and down the state.

Chart of asthma rate falling over many years
The model also allowed Dr. Banerjee and colleagues to show a decline in overall asthma-related hospitalization rates over 17 years. (click graph to enlarge)

The result is a series of models created by Banerjee and his colleagues – Dr. Harrison Quick, now at the Centers for Disease Control and Prevention; and Dr. Bradley P. Carlin, professor and head of the Division of Biostatistics at the University of Minnesota School of Public Health – that can be used to predict hospitalization rates in any county for any day of the year. The models incorporate socio-demographic variables (taking into account that children and African-Americans, for example, have disproportionately high rates of asthma attacks), environmental factors (including ozone data from the California Air Resources Board), and seasonal patterns. The spatial and temporal trends can be visualized using GIS maps of the state, with rates color-coded by county.

By showing when spikes in hospitalization tend to occur within each county, Banerjee’s work enables hospital administrators to better prepare for the influxes by adjusting staffing levels accordingly. While the vast majority of counties’ hospitalizations are fewer than 20 per 100,000 residents per month, the range extends from zero to as high as 90.

Equally important, the models can be used by epidemiologists and other public health professionals to learn more about the environmental and seasonal factors associated with asthma exacerbations, potentially pointing the way toward preventive measures. “For hospital planners the key question is ‘when,’ ” says Banerjee. “For epidemiologists, it’s more about ‘why.’ These models provide epidemiologists with tools to map spatial patterns that are more reliable than mapping raw data, so that they can see how these patterns evolve over time.”

Showing asthma hospitalizations by time and place at a much higher resolution than was previously possible, Banerjee’s group has provided a tool for revealing previously inaccessible details. For example, although efforts to associate asthma incidence and ozone levels may not produce statistically significant results, Banerjee and his colleagues used their models to demonstrate that ozone’s effects are seasonal – while levels may be higher in the summer, the impact on asthma hospitalizations is more pronounced in the winter. At the Fielding School, Banerjee plans to continue monitoring asthma hospitalization rates to see whether the trends his group has observed persist. He also intends to apply the models to other diseases, such as certain types of cancers.

Advances in GIS technology have led to a burgeoning of spatial-temporal databases and the need for statisticians and data analysts with expertise in modeling and analysis to assist professionals in making sense of them. For Banerjee, the work is particularly meaningful. Growing up in Calcutta, India, he suffered from asthma, as did both his paternal grandmother and paternal grandfather. “I have a young son, and would certainly not want him to experience what I did with this condition,” Banerjee says. “It’s rewarding to be able to contribute something that may help to prevent some of these problems in the future.” •