O2098: More Than Words: Language Patterns Indicative of Unrecognized Shock in Children Undergoing Interhospital Transfer

Background: Eighty-nine percent of pediatric emergency department (ED) visits in the United States occur in general rather than specialized pediatric EDs. Early detection of shock is necessary to reduce morbidity and mortality, yet shock is frequently unrecognized in children. The aim of this pilot study was to identify specific terms frequently used by referring providers during transfer request calls to describe pediatric patients with unrecognized shock (URS) compared to those with recognized shock (RS).

Methods: Transcripts of transfer request calls were obtained for patients aged ≤21 years who were transferred to the Johns Hopkins Children’s Center (JHCC) between January 1, 2015 and December 31, 2021 with the primary diagnosis of shock upon evaluation at JHCC. The referring facility’s diagnosis was compared to the diagnosis at JHCC to determine if shock was recognized vs. unrecognized pre-transfer. Transcripts were reviewed to identify descriptive terms most frequently used by referring providers. Continuous variables were summarized using medians and interquartile ranges (IQR) and compared using the Mann-Whitney U-test. Proportions were compared using a two-sample test of proportions.

Results: We analyzed 23 transcripts: 11 URS and 12 RS. Median ages (IQR) of patients with URS and RS were 13 (0-204) months and 42 (9-177) months, respectively. Compared to the RS cohort, patients with URS had longer hospitalizations (median (IQR): 9 (5-30) vs. 3.5 (2-6.8) days, p=.01). Patients’ characteristics are included in Table 1. The descriptive terms frequently used by referring providers are included in Table 2. Notably, “shock” was used in 91.7% of RS vs. 0% of URS transcripts (p <.001), while “appearing” was used in 45.5% of URS vs. 0% of RS transcripts (p=.004). Several terms approached statistical significance in being used more frequently to describe patients with URS vs. RS: “pale” (36.4% URS vs. 8.3% RS, p=.05), as well as “tired”, “sick”, “mottled”, and “fast breathing” (18.2% URS vs. 0% RS, p=.06 each).

Conclusion: In this pilot sample, there were discretely identifiable terms used to describe patients with URS compared to those with RS. Compared to the clinically unambiguous term “shock” used to describe patients with RS, providers describing patients with URS more frequently used ambiguous descriptors of patient presentation related to overall appearance, coloration, mental status, and respiratory patterns. Analysis of additional qualifying transcripts is expected to reveal more significant differences between URS and RS transcripts. Based on these and future results, we aim to develop natural language processing-based predictive models to identify shock in children undergoing interhospital transfer.