Introduction

• ✨ Remove sensitive information from clinical text written in Dutch

• 🔍 Rule based logic for detecting e.g. names, locations, institutions, identifiers, phone numbers

• 📐 Useful out of the box, but customization higly recommended

• 🌱 Originally validated in Menger et al. (2017), but further optimized since

❗ Deduce is useful out of the box, but please validate and customize on your own data before using it in a critical environment. Remember that de-identification is almost never perfect, and that clinical text often contains other specific details that can link it to a specific person. Be aware that de-identification should primarily be viewed as a way to mitigate risk of identification, rather than a way to obtain anonymous data.

Currently, deduce can remove the following types of Protected Health Information (PHI):

• 👤 person names, including prefixes and initials

• 🌎 geographical locations smaller than a country

• 🏥 names of hospitals and healthcare institutions

• 📆 dates (combinations of day, month and year)

• 🎂 ages

• 🔢 BSN numbers

• 🔢 identifiers (7+ digits without a specific format, e.g. patient identifiers, AGB, BIG)

• ☎️ phone numbers

• 📧 e-mail addresses

• 🔗 URLs

Citing

If you use deduce, please cite the following paper:

Menger, V.J., Scheepers, F., van Wijk, L.M., Spruit, M. (2017). DEDUCE: A pattern matching method for automatic de-identification of Dutch medical text, Telematics and Informatics, 2017, ISSN 0736-5853