An analysis of hospital coding in Portugal: Detection of patterns, errors and outliers in female breast cancer episodes

Many studies use coded clinical data as a source to investigate some diseases and its demographic impact. In Portugal, according to the National Statistics Institute, 1,635 deaths due to female breast cancer occurred in the year 2009, accounting for near 4,900 new cases by year. This study aimed to describe and analyze data from hospital female breast cancer episodes, coded in Portuguese hospital databases. Specifically, it aimed to profile data quality and to identify inconsistencies, errors and outliers. The database used includes inpatient and outpatient episodes from all mainland acute care public hospitals of the Portuguese National Health Service, within years 2000 to 2008, corresponding to more than 10 million records. Empirical data analyses were performed to understand the data distribution and the profile of possible outliers. Many problems and inconsistencies were found as, for instance, the existence of 40% of female breast cancer episodes with a secondary diagnosis of “admission for radiotherapy and chemotherapy” and, also, 783 records with the variable gender filled with “male”. Systematic audits to the data and to the information produced, performed by competent agencies, can minimize some of the found problems. In fact, the knowledge of anomalous situations should allow, when feasible, an immediate improvement in the quality of healthcare data.