Copying in medical documentation: developing an evidence-based approach

Abstract

Background: Electronic medical records (EMRs) provide multiple efficiencies in communication to clinicians. The ability to copy and paste text in an EMR can be useful; however, it also conveys a risk of inaccurate documentation. Studies in international settings have described such overuse of copying to result in 'note bloat', with the dilution of relevant clinical information and potential clinical detriment.

Aim: To determine the frequency of erroneous copying, characterise the component of notes in which this occurs and determine the performance of similarity metrics in the prediction of notes likely to have erroneous copying.

Methods: A cross-sectional evaluation of all ward round notes over a 48-h period for all long-stay (>48 h) medical services, except the Acute Medical Unit, at the Lyell McEwin Hospital, a 257-bed tertiary hospital in South Australia. Four similarity metrics were evaluated: longest-sequential series of unchanged characters, similarity score (Difflib SequenceMatcher), Levenshtein distance and the Jaccard index.

Results: One hundred twenty-eight patients were included. The number of patients who had a ward round note on two consecutive days was 97 out of 128 (75.8%). Erroneous copying was found in 8.3% of ward round notes. All (eight out of eight, 100%) of these instances of erroneous copying were in the 'issues list'. A threshold of >850 unchanged sequential characters, when compared with the ward round note the preceding day, demonstrated reasonable performance in the prediction of erroneous copying.

Conclusions: Erroneous copying may occur in up to 8.3% of ward round notes in a variety of medical services. Automated strategies to help address this issue should be explored.

Keywords: detection; documentation of information; electronic health records; improvement; standards.

Figure 1

AUROC of longest‐sequential series of unchanged characters for the prediction of erroneous copying. AUROC, area under the receiver operator characteristic curve.

Figure 2

Calibration graph of transformed longest‐sequential series of unchanged characters for the prediction of erroneous copying.