Physiological patient data collected in the neonatal intensive care unit (NICU) often does not lend itself to straightforward analysis. It is common for NICU data to be missing not at random (MNAR), as only abnormal values may be recorded while healthy babies are passed over. Strategies are needed to handle this missing data framework both in the measures that are directly afflicted with missingness and in composite diagnostic scores that are calculated from these measures. The Pediatric Logistic Organ Dysfunction (PELOD) score and the Proulx and Goldstein criteria, used to diagnose multiple organ dysfunction syndrome (MODS), are composite scores based on dysfunction of various organ systems as indicated by abnormal lab results and vitals. A missing result for any organ system produces a missing overall score. We will examine the performance of various imputation methods at both the item (organ) level and the scale (score) level in the context of NICU data MNAR. Imputation performance will be assessed by examining the bias and efficiency of scale-level parameter estimates, including scale score means, between-scale correlations, and regression coefficients.