Reliability of reagent strips for semi-quantitative measurement of glucosuria in a neonatal intensive care setting

Glucosuria in preterm infants is often measured using a visually readable reagent strip, e.g., when monitoring total parenteral nutrition or during sepsis or when treating with corticosteroids. However, the specific circumstances in a neonatal intensive care unit (NICU), such as the use of diapers and the high temperature in incubators, could affect its reliability.

To evaluate the reliability of the semi-quantitative measurement of glucosuria under the specific circumstances of a NICU setting.

Nine hundred assessments of artificially supplemented (contrived) urine samples, intended to simulate pathological specimens, were performed under the following varying conditions: environmental temperature (21°C and 34°C); different times of contact of the urine with the diaper; and using two different methods of collecting urine from the diaper. Each reagent strip was read independently by three observers. The test strips scores were categorized as 0, 1+, 2+, 3+, or 4+ in ascending degree of glucosuria.

Agreement was excellent under all the different conditions (temperature, weighted kappa (κ(w)) = 0.92; method of urine collection, κ(w) = 0.88; time, p = 0.266). Inter-observer reliability was very good (multi-rater κ = 0.81). The deviation between the different conditions was seldom larger than one category (2.9%). The reagent strip readings were concordant with the true urinary glucose concentrations in 79.0% of assessments. The discordance was never larger than one category.

The reliability of the semi-quantitative measurement of glucosuria in newborn infants using reagent strips is good, even under the conditions of a NICU. Changes in the rating of reagent strips of more than one category are most likely to be beyond measurement error.