DOSES (DOSage Extraction System) extracts and represents free-text medication prescription instruction information in a structured form. In particular, it represents the variability and flexibility in drug directions by including minimum and maximum values for drug dosage, frequency and interval of administration, as well as optional choices. In order to comprehensively represent the information in medication prescriptions, DOSES identifies and represents in structure the following medication prescription dosage attributes:

  • min/max dose number
  • min/max dose frequency
  • min/max dose interval
  • dose unit
  • optional dose

DOSES is implemented in MinorThird for the identification of the targeted dosage attributes, and Python for result post-processing and structured representation.

Currently, two versions are available for DOSES:

In DOSES v.1, the performance has been evaluated on a subset of CPRD (Clinical Practice Research Database) prescription records.

More details on the architecture and the performance can be found in.

DOSES v.2 has been tailored and evaluated on a collection of patient records acquired from the UK Renal Registry (UKRR).

More details on the tailored version can be found in.

  • Alfattni, Ghada, et al. “Integrating text analytics and statistical modelling to analyse the UK Renal Registry data.” Poster presented at: International Population Data Linkage Conference; 2016 AUGUST 24-26; Swansea, United Kingdom.

Contact George Christopher Karystianis, (karystianis@gmail) regarding any questions, bugs and/or suggestions.