Certainty, Uncertainty, and Accuracy of Daily Predictions of Non-Survival in the NICU: We're Often Wrong but Rarely in Doubt ♦ 156

Care provided in the neonatal intensive care unit (NICU) is expensive and agonizing. In order to optimize parental decisions to provide or forego medical interventions, physicians and nurses are obliged to offer their best estimates of the likelihood of survival for the infants in their care. But how good are these estimates?

Methods: We surveyed M.D.s and R.N.s caring for ventilated infants in the NICU every day for 48 weeks, asking whether they thought the infants in their care would survive or die before hospital discharge. We correlated these daily estimates against each other and against the outcome of the patients.

Results: During the 48 weeks of our study, roughly 14,000 predictions were offered for 255 infants who required mechanical ventilation for 3070 cumulative bed days in our NICU. For 190/216 (88%) survivors, there was 100% agreement on each day of ventilation that the infant would survive. 18/39 (46%) non-survivors had 100% agreement on each day that they would not survive. 47 (18%) of patients had days marked by disagreements -- that is, some caretakers predicted “die”, while others predicted“live”. 21 (45%) of these infants died. 517/3070 (17%) patient days were marked by a prediction of “die”. However, 311/517 (60%) of these non-survival predictions were inaccurate -- that is, the patient eventually survived. 332 patient days were associated with > 50% prediction of death -- 49% were inaccurately attributed. 141 patient days were associated with uniform (100%) prediction of death -- 32% were wrong. 77 patient days were associated with 72 consecutive hours of predicted death -- 18% were wrong. Only 63/412 (15%) of all bed-days occupied by non-survivors were characterized by 72 consecutive hours of predicted death.

Conclusions: 1) The vast majority of ventilated NICU patients and patient days are accurately associated with uniform predictions for survival. 2) The majority of NICU patient days associated with predictions of non-survival are inaccurately attributed; 3) Predictions of non-survival for ventilated infants in the NICU, even if highly collaborated and consistent, appear empirically and ethically tenuous.