Model Predicts Neurodevelopmental Outcomes, Death in Preterm Infants

A study published earlier this month in JAMA Network Open demonstrates that a newly-developed multimodal model using brain function information and other risk factors can improve the prediction of neurodevelopmental impairment (NDI) or death at 2 years in extremely preterm newborns.

The researchers noted that the risk of NDI or death in extremely preterm infants remains high despite recent significant improvements in neonatal care. To help address this, early assessment of a preterm infant’s prognosis is critical to guide considerations of treatment options and drive decision-making. More broadly, the ability to identify high-risk newborns could facilitate the implementation of earlier, more personalized rehabilitation interventions to reduce the impact of NDI.

However, the researchers explained that current prognostic tools for quantifying NDI and death in extremely preterm infants are limited by “(1) their use of conventional analytical methods, (2) their reliance on a small number of clinical and laboratory variables, and (3) their failure to take account of prognostically valuable information on the brain’s structure and function.” These assessments also often do not incorporate functional brain information, which could aid prognosis.

To close these gaps, the researchers developed five prediction models and compared their performance. One of the models was multimodal, simultaneously considering brain function information, brain structure information obtained via cranial ultrasonography, and perinatal and postnatal risk factors. The other four models were unimodal. They considered each category of these variables independently.

Information from preterm infants, defined in this study as those at 23-28 weeks gestational age, admitted to the neonatal intensive care unit (NICU) at Amiens-Picardie University Hospital in France from Jan. 1, 2013, to Jan. 1, 2018, and who underwent conventional electroencephalography (cEEG) and cranial ultrasonography (cUS) within two weeks of delivery were included in the study.

In terms of outcomes measured, the researchers indicated no or moderate NDI as a favorable prediction outcome, while unfavorable outcomes included severe NDI or death before discharge from the NICU.

Researchers determined NDI presence and severity for newborns discharged from the NICU alive using results from the Denver Developmental Screening Test II (DDST II), which was administered by pediatricians when each infant reached 2 years of age. NDI was considered absent for scores over 75 percent, moderate for scores between 50 and 75 percent, and severe for scores below 50 percent.

A total of 109 infants met the inclusion criteria for the study. Of these, 52, or 47.7 percent, had a favorable outcome of no or moderate NDI at 2 years of age, while 57, or 52.3 percent, had an adverse outcome.

The performance of each model was measured using area under the curve, with the multimodal model achieving the highest performance of any model. The multimodal approach achieved an area under the curve of 91.7 percent, while the unimodal models scored significantly lower: the perinatal model, the postnatal model, the brain structure model, and the brain function model reached 80.6, 81.0, 76.6, and 78.8 percent, respectively.

The researchers concluded that these findings highlight the usefulness of brain information in prognostic models for quantifying NDI and death in extremely preterm newborns, which may reflect how the evaluated risk factors work together to impact brain maturation and other outcomes. Further, the authors posited that their multimodal model, or a similar approach, may have potential as a clinical decision support tool in critical care management for these infants.