Progress note length and structure in the intensive care unit: the unabridged patient medical record

SESSION TITLE: Practice Management and Administration Posters SESSION TYPE: Original Investigation Posters PRESENTED ON: 10/10/2023 12:00 pm - 12:45 pm PURPOSE: Electronic health records (EHRs) allow users to copy content from prior notes or use system-generated text when writing clinical notes. A growing corpus of work suggests this practice disperses or obscures useful information and is increasing over time. There is rising recognition that notes are growing increasingly longer, compounding the problem of obscured information. The ICU may be particularly sensitive to the pitfalls of these trends in note structure and length. This analysis was undertaken to quantify trends in note length, note content sources, and the impact of clinical load in a critical care setting at an academic health facility. METHODS: 58,548 progress notes written by residents or attending physicians between February 2017 and January 2023 in the Oregon Health & Science University Medical ICU were included in the study. Total note character counts were extracted from our EHR (Epic Systems) audit log data to assess changes in average note length over time. Metadata describing how note content was derived was also extracted, allowing the grouping of characters written into four major categories: “system-generated”, “copied”, “manual”, and “other”. Data were aggregated at the user level for individual years. Correlations between the proportion of characters written in each of these categories to average note length and number of notes written in a year were calculated. RESULTS: Resident average note length increased from 13,173 to 16,449 characters (25%; P<.0001) over the study period. The proportion of text which was copied increased significantly over the period (34% vs 46%; P<.0001) while the proportion of manually generated text decreased (24% vs 12%; P<.0001), with no significant change in proportion of system-generated text. On a physician level, increasing average note length was associated with an increase in the proportion of duplicated text (R=0.24, P<.0001), a decrease in the proportion of manually entered text (R=-0.45; P<.0001), and a decrease in the proportion of system-generated text (R=-0.30; P<.0001). Among residents, the total number of notes written per year was positively correlated with higher proportion of duplicated content (R=0.33, P<.0001) and negatively correlated with proportion of manually generated text (R=-0.27, P<.0001). In contrast, attending average note length showed no significant change over the period. However, increasing average attending physician note length was associated with an increase in the proportion of duplicated text (R=0.33; P<.0001) and a decrease in the proportion of manually generated text (R=-0.31; P<.001). CONCLUSIONS: Residents are writing longer and longer notes. This increase in length appears to be driven by passively generated content and at the expense of more actively generated, novel content. This behavior is exacerbated as clinical load increases. Attending notes also show that longer notes are more likely to be the result of copied content at the expense of novel content. However, attending notes are not growing longer over time. CLINICAL IMPLICATIONS: Increasing note length and prevalence of copied content has significant implications for clinical cognitive load and, by extent, patient safety. Further studies are ongoing to assess this impact and the underlying perceptions driving the differences in behavior found in this analysis. DISCLOSURES: No relevant relationships by Tanuj Devara No relevant relationships by Sarah Florig No relevant relationships by Jeffrey Gold No relevant relationships by Ryan Grattan No relevant relationships by Vishnu Mohan