Objectives: (1) To develop methods to describe autopsy utilization patterns in a neonatal intensive care unit. (2) To identify classes of patients likely to have clinicopathologic concordance or discordance.
Methods: Five hundred forty-five consecutive neonatal intensive care unit deaths (338 autopsied, rate 62%) in a regional tertiary/quaternary care neonatal intensive care unit for referred infants (65,000 annual births) were classified in six clinical diagnostic groups (anomalies, cardiac anomalies, hypoxic ischemic encephalopathy, prematurity and its complications, infections, and other) and rated in three levels of certainty of clinical diagnosis as "gold standard" certainty, almost complete certainty, and less certain than the latter. Clinicopathologic discordances were rated in three classes using clinical, pathologic, and multidisciplinary mortality conference records. The proportions of autopsied cases, cases with major discordances, and cases with no discordances were compared and analyzed in relation to diagnostic group and level of certainty.
Results: Performance of autopsy was associated with clinical diagnostic uncertainty (p = 0.008). Major discordances with implications for outcome (Class I) were found in 3%, and without implications for outcome (Class II) in 15% of cases; 42% of cases had no discordances. Major discordance rate varied inversely with the degree of diagnostic certainty (p = 0.000) and varied among clinical groups.
Conclusions: (1) Autopsy was used most for cases with potential for high yields. (2) Clinicopathologic discordances were more frequent and important in certain clinical diagnostic groups (prematurity, other) and with high levels of diagnostic uncertainty. When the diagnostic "gold standard" is available during life, autopsy will provide little information.