Analyzing data from computerized medical records: Possibilities vs. limitations

Objective: To assess the possibilities and limitations of analyzing pediatricians' computerized medical records.

Methods: Data analysis of computerized medical records of all visits to pediatricians in a sample of 13 community-based clinics, during a two year period.

Results: Approximately 527,000 visits of 63,000 children to 57 pediatricians (59.6% board-certified) were analyzed. The mean number of visits over the two year period was 8.3. With increasing age, the number of visits decreased, with the sharpest drop between ages 0-4 years. Over half of all visits resulted in a drug being prescribed (56.8%) and board-certified pediatricians prescribed significantly fewer drugs than non-certified pediatricians. The most common diagnostic groupings in all sample strata were: "Respiratory illness", "Symptoms and Ill-defined conditions" and "Administrative" diagnoses. Only 0.93% of visits (n=4,928) of 5.3% of children (n=3,357) received a diagnosis from the new morbidity (NM) area (developmental, behavioral, and psychosocial issues). Board-certified compared to non-certified pediatricians gave a higher proportion of NM diagnoses. Examination of pediatricians free text, in a random sample of medical records of 1,200 children, nearly doubled the percentage of visits with a NM diagnosis (1.9%), and increased by one and a half the number of children with these diagnoses (8.9%).

Conclusions: Our findings demonstrate the possibilities and limitations of analyzing and processing data from computerized medical records. Analyzing computerized data allows collection of large volumes of data, but presents problems regarding the quality of the data and the complexity of analyzing un-coded data to complement the collected data.