Date of Award

January 2022

Document Type

Open Access Thesis

Degree Name

Master of Public Health (MPH)


School of Public Health

First Advisor

Daniel M. Weinberger


The number of COVID-19 related hospitalizations that occur in a particular area is dependent on a variety of sociodemographic and clinical factors, but the percentage of hospitalizations that are identified in that region varies based on the type and quality of surveillance system that is utilized. In Connecticut, an active surveillance system called the COVID-19–Associated Hospitalization Surveillance Network (COVID-NET) conducts COVID-19 associated hospitalization surveillance in two of Connecticut’s eight counties. The other counties in Connecticut rely on a passive surveillance system which could be subject to underreporting. To evaluate possible underreporting, positive SARS-CoV-2 test rates obtained from the Connecticut Department of Public Health (CT DPH) and variables from the CDC’s Social Vulnerability Index (SVI) were used as covariates in a negative binomial regression model. The model was fit to random samples of COVID-NET hospitalization data through an iterative process and ten optimal models were selected using stepwise Akaike’s Information Criterion (AIC). The average of the regression coefficients for each covariate included in the ten optimal models was calculated and multiplied by a model matrix containing the original SVI and SARS-CoV-2 testing covariates for each census tract in Connecticut to produce census tract-level estimates of COVID-19 related hospitalizations in Connecticut. Based on the model estimates, 5,600 excess hospitalizations occurred throughout 2020 compared to the number passively reported to CT DPH. Of note, New London County had the largest discrepancy between observed and estimated hospitalization rates (255 hospitalizations per 100,000), and New Haven and Middlesex counties (the counties which comprise the COVID-NET catchment area) had the lowest discrepancy (45 hospitalizations per 100,000 and 56 hospitalizations per 100,000, respectively). Widespread underreporting of COVID-19 related hospitalizations in Connecticut has broad implications. These surveillance gaps must be addressed to achieve more equitable pandemic planning and response.


This is an Open Access Thesis.

Open Access

This Article is Open Access