Bayesian estimation on diagnostic performance of Face, Legs, Activity, Cry, and Consolability and Neonatal Infant Pain Scale for infant pain assessment in the absence of a gold standard

Abstract

Background: Nonverbal pediatric patients such as infants are unable to describe their pain, which leads to the lack of a gold standard scale for their pain assessment. The aim of this study was to estimate the diagnostic performance of Face, Legs, Activity, Cry, and Consolability (FLACC) scale and Neonatal Infant Pain Scale (NIPS) for infants' pain in the absence of a gold standard.

Methods: This prospective observational study recruited 202 postoperative infants, aged <12 months. Postoperative pain intensity was evaluated using FLACC and NIPS scales. The diagnostic performance of these two scales was to estimate using a Bayesian latent class model with conditional dependence. McNemar's test was applied to test whether NIPS and FLACC tests differ from each other.

Results: Under a combined model with conditional dependence, the median posterior sensitivity and specificity of the FLACC were 89.94% (95% CI: 78.48-96.83%) and 87.82% (95% CI: 78.6-95.23%), respectively. The sensitivity and specificity of the NIPS were 85.94% (95% CI: 72.15-95.6%) and 92.61% (84.05-97.52%), respectively. McNemar's test demonstrated no significant difference between FLACC and NIPS in either sensitivity or specificity.

Conclusion: Both the FLACC and NIPS have excellent sensitivity and specificity for pain assessment in infants. The comparison test showed that the FLACC scale was no different to the NIPS scale in sensitivity and specificity.

Keywords: Bayesian analysis; infant; pain assessment; pediatric pain.