PT - JOURNAL ARTICLE AU - A Gray AU - S Goodacre AU - M Braidburn AU - J Cohen AU - J Benger AU - T Coats TI - PREDICTION OF UNSUCCESSFUL TREATMENT IN PATIENTS WITH SEVERE ACUTE ASTHMA: AN ANALYSIS FROM THE 3MG TRIAL AID - 10.1136/emermed-2013-203113.15 DP - 2013 Oct 01 TA - Emergency Medicine Journal PG - 872--873 VI - 30 IP - 10 4099 - http://emj.bmj.com/content/30/10/872.short 4100 - http://emj.bmj.com/content/30/10/872.full SO - Emerg Med J2013 Oct 01; 30 AB - Objectives & Background Clinical assessment can be used to identify which patients with acute asthma are at risk of unsuccessful initial treatment. We aimed to determine, using data from the 3MG trial, which elements of clinical assessment predict unsuccessful treatment, defined as needing critical care or any unplanned additional treatment. Methods We analysed data from a large multicentre trial (the 3Mg trial). Adults with severe acute asthma underwent standardised clinical assessment, including peak expiratory flow rate (PEFR), up to two hours after initiation of treatment. Standard care was provided other than blinded random allocation to trial treatment or placebo. Patients were followed up by record review up to 30 days. Unsuccessful treatment was defined as needing (1) critical care or (2) critical care or any unplanned additional treatment within seven days of presentation. Logistic regression was used to identify predictors and derive a prediction model for each outcome. Results Out of 1084 patients analysed, 81 (7%) received critical care and 157 (14%) received critical care or unplanned additional treatment. Baseline PEFR (p=0.017), baseline heart rate (p<0.001), other serious illness (p=0.019), PEFR change (p=0.015) and heart rate change (p<0.001) predicted need for critical care. Baseline PEFR (p=0.010), baseline heart rate (p<0.001), baseline respiratory rate (p=0.017), other serious illness (p=0.023), PEFR change (p=0.003) and heart rate change (p=0.001) predicted critical care or additional unplanned treatment. Models based on these characteristics had c-statistics of 0.77 and 0.69 respectively. Conclusion PEFR, heart rate and other serious illness are the best predictors of unsuccessful treatment, but models based on these variables provide only modest predictive value.