They created a logistic regression model with screening (yes/no) as the dependent variable and multiple independent variables such as gestational age, sex, birth weight, and presence of respiratory distress syndrome. These are potential confounders that could affect both whether the baby gets screened and whether the baby dies in the hospital. The logistic regression model combines these variables to create a propensity for being screened. Then they identified 605 pairs of infants with the same propensity score, one who was screened and one who wasn't. Finally, they compared the mortality of the 605 screened infants to the mortality of the 605 unscreened infants.