Relationship between emotional words in electronic medical records and leave periods of users of a return-to-work program with depression

Abstract

Introduction We attempted to score data extracted from written medical records containing assessment results using natural language processing, and to clarify the relationship between duration of sick leave and the use of emotional words among return-to-work (RTW) program users on sick leave due to mental health problems. Method Participants were users of an RTW program. We extracted textual data from their electronic medical records, and gave all words a score based on the following two considerations: positivity score (the degree of positive emotion a word has) and emotion score with respect to seven emotions (sadness, anxiety, anger, disgust, trust, surprise, and joy), with the score for each emotion measured for each word. We analyzed relationships between duration of sick leave and each score. Results Forty-two users participated. The results showed that high positive scores (β = −0.42, p < 0.00) and high sadness scores (β = −0.60, p < 0.00) were related to a shorter duration of sick leave, and high anger score (β = 0.52, p < 0.00) was related to a longer duration of sick leave. Conclusion Professional assessments based on occupational therapy and natural language processing of medical records may predict the appropriate timing of RTW.