Using Machine-learning Algorithms to Assess Acute Pain in Nonverbal Infants

Pain assessments in non-verbal, critically ill infants represent an important clinical challenge. Older children or adults can easily express their pain, but infants lack that capability. They frequently experience repetitive acute pain during routine ICU care, but their analgesic management flounders on the horns of a dilemma: (a) failure to treat infant pain leads to immediate clinical instability and potentially long-term physical, behavioral, and cognitive sequelae, vs. (b) strong analgesics may increase risks for medical complications and/or impaired brain growth. Bedside nurses currently assess pain using pain scores, before taking action to ameliorate pain. Pain scores increase nursing workload and provide subjective assessments, rather than objective data for evaluating infant pain. Consequently, infants exposed to skin-breaking procedures, surgery, or other painful conditions often receive variable and inconsistent pain management in the ICU. The investigators aim to develop a multimodal pain assessment system, using sensor fusion and novel machine learning algorithms to provide an objective measure of pain that is context-dependent and rater-independent. This will enhance the quality of pain management in ICUs and allow continuous pain monitoring in real-time.