Assessment of asthma using automated and full-text medical records

Automated medical records systems are used to study clinical outcomes and quality of care, but this requires accurate disease identification and assessment of severity. We sought to determine the reliability of identifying asthmatics through automated medical and pharmacy records, and the adequacy of such data for severity assessment. All adult health maintenance organization (HMO) members who received at least one asthma drug and an asthma diagnosis between April 1988 and September 1991 were identified. Records of a random sam pie were reviewed to validate the diagnosis and extract clinical information. Asthma drugs were dispensed to 15,491 individuals; 7583 (49%) also received an asthma diagnosis. Asthma drug use was three times greater for persons with diagnosed asthma compared to those with no diagnosis. Record review revealed that a coded asthma diagnosis had a positive predictive value of 86%. Nearly 4000 ambulatory encounters were reviewed, 10% of which were for asthma; the median number of encounters was two. Asthma symptoms were mentioned in 9% of all encounters; wheezing was most common. Peak flow and spirometry were measured in 4% and 1% of encounters, respectively. Records from recipients of asthma drugs who racked an asthma diagnosis showed that 79% did not have asthma. Automated medical and pharmacy records from an HMO were relatively accurate when used to identify individuals with asthma. Similarly, most asthma drug recipients who lacked a coded diagnosis of asthma did not have asthma. However, convention al full-text records usually do not con ta in sufficient information to assess asthma severity, limiting the utility of such records for research and quality improvement.