Attempts to Infer Observation Bias from Chart Entry Behavior
- VernacularTitle:カルテへの記載行動から観察バイアスを推察する試み
- Author:
Kotonari AOKI
1
;
Yosuke NISHIDA
2
;
Suguru NOZUE
2
Author Information
- Keywords: pharmacoepidemiology; medical information databases; observational bias; post-marketing safety measures
- From:Japanese Journal of Pharmacoepidemiology 2025;():30.e2-
- CountryJapan
- Language:Japanese
- Abstract: Objective：Compared to research conducted under a controlled system such as a clinical trial, differences in characteristics among observers (medical care providers) can be an issue when conducting observational research in general or research using RWD in particular. In particular, systematic differences in observational behaviors (attitudes) by department may cause confusion in interpreting study results, so we aimed to quantify the differences in behaviors toward patient observation by department. Design：Descriptive aggregation using “Millennium medical record” Database. Basic statistics such as median values for the number of characters in the clinical summary, which is a free entry column, are obtained for each department. Based on the purpose of this study, it was judged that it was appropriate to omit the description after hospitalization because the number of characters described in the part of the clinical course up to hospitalization in the clinical summary was included in the character count. Results：The median number of letters in internal medicine was 503. The ratios of the median number of letters to the median number of letters in various departments were surgery (0.55), ophthalmology (0.57), psychiatry/psychosomatic medicine (2.85), pediatrics (1.19), obstetrics/gynecology (1.04), and dermatology, orthopedic surgery/plastic surgery (0.41), respectively. Conclusion：There are characteristic differences in the number of letters in the free entry items depending on the medical department.
