<b><i>Introduction:</i></b> Premature infants are exceptionally vulnerable to nutrition-related diseases, and the utilization of standardized feeding guidelines may reduce nutritional practice variation, which can promote growth. Nutritional risk screening is the first step for standardized nutrition advice. However, risk screening tools specific for following up preterm infants are scarce. Hence, our study aimed to develop and evaluate a standardized Nutritional Risk Screening Tool for Preterm Infants (NRSP subscale 1) from birth to corrected age four months old . <b><i>Methods:</i></b> This study was a two-phase (the development phase and evaluation phase) study. Initially, we used the Delphi expert consultation method to create NRSP subscale 1. Then, a professional panel interviewed the participated preterm infants using the screening tool, measured anthropometric parameters, and conducted an intellectual development test on the interview day and remeasured anthropometric parameters 2 weeks or 1 month after the first interview. In the development phase, we cross-tabulated the responses to the screening tool with the classifications of <i>z</i>-scores of the body weight, length, or head circumference to identify significant predictors of underweight, stunting, or microcephaly. We then combined significant predictors to produce models for predicting underweight, stunting, or microcephaly by multivariate logistic regression analysis. In the evaluation phase, the area under the curve (AUC), sensitivity, specificity, and correlation coefficient by Spearman’s correlation analysis (<i>r</i><sub>s</sub>) between the risk classifications by NRSP subscale 1 and the classifications of the <i>z</i>-scores of the body weight, length, or head circumference were calculated to assess the validity of the screening tool. Intellectual development levels between high and low nutritional risk infants were statistically compared. <b><i>Results:</i></b> A total of 219 and 244 preterm infants were included to two phases, respectively. AUC was 0.936 (95% CI: 0.860–1.000, <i>p</i> < 0.001), sensitivity was 0.667, specificity was 0.941, <i>r</i><sub>s</sub> = 0.407 (<i>p</i> < 0.001); AUC was 0.794 (95% CI: 0.638–0.951, <i>p</i> = 0.002), sensitivity was 0.500, specificity was 0.953, <i>r</i><sub>s</sub> = 0.339 (<i>p</i> < 0.001); AUC was 0.831 (95% CI: 0.737–0.925, <i>p</i> = 0.001), sensitivity was 0.889, specificity was 0.643, <i>r</i><sub>s</sub> = 0.215 (<i>p</i> = 0.001) in predicting underweight, stunting, and microcephaly on the interview day, respectively. AUC was 0.905 (95% CI: 0.826–0.984, <i>p</i> = 0.006), sensitivity was 0.500, specificity was 0.905, <i>r</i><sub>s</sub> = 0.504 (<i>p</i> < 0.001); AUC was 0.738 (95% CI: 0.515–0.960, <i>p</i> = 0.034), sensitivity was 0.429, specificity was 0.848, <i>r</i><sub>s</sub> = 0.382 (<i>p</i> < 0.001); AUC was 0.664 (95% CI: 0.472–0.856, <i>p</i> = 0.071), sensitivity was 0.455, specificity was 0.809, <i>r</i><sub>s</sub> = 0.169 (<i>p</i> = 0.037) in predicting underweight, stunting, and microcephaly 2 weeks or 1 month after the first interview, respectively. Gross motor development quotients (DQs) (95.85 [32.87] vs. 86.29 [17.19], <i>p</i> = 0.022), fine motor DQs (115.77 [46.03] vs. 102.12 [20.27], <i>p</i> = 0.010), and verbal DQs (110.73 [35.27] vs. 100.63 [21.28], <i>p</i> = 0.042) were higher in low nutritional risk infants than high-risk ones. <b><i>Conclusion:</i></b> NRSP subscale 1 was acceptable and reliable in predicting underweight, but the validity in predicting stunting or microcephaly was slightly mild. Further investigations are required to authenticate NRSP subscale 1’s effectiveness.
