Abstract

Introduction

Routine primary care data contains prescription data for treatments for respiratory diseases these include trade names for corticosteroid drugs, both on their own and combined with Long Acting Beta Agonists (LABA).  This data usually contains the treatment and dose in one field and the direction of treatment in another.  There is a need to obtain usable analysable data for standardised corticosteroid dose from routine data.

Aim

To obtain and test text matching algorithms that can mine bdp equivalent corticosteroid dose from routine primary care data.     

Methods

We created string matching algorithms in R language to determine which inhaled corticosteroid a patient was on.  We then mined the dose from the text field containing the description of the inhaled corticosteroid.  We further used string matching algorithms to determine how many times the patient took their inhaler in a day, i.e. “two puffs, twice a day”.    

Results

The algorithms worked on all fields without missing data. For evaluation the data will be compared to 100 new records using the reference standard of a trained Asthma professional looking through the records and obtaining the bdp using written down rules.  Thus the predictive sensitivity and specificity will be calculated for the algorithms.

Conclusions

It is possible to obtain bdp equivalent from routine primary care using string matching algorithms to obtain trade name of drug, dose of drug and number of puffs per day to obtain total dosage per day and converting to bdp equivalent.