Resolving fundamental quality issues in linked datasets for clinical care

The objective of this study was to resolve the problem whereby data entry error rates are compounded within a comprehensive Electronic Health Record (EHR). This is a prerequisite if quality issues are to be addressed through electronic record linkage as a step towards integrating personal health data. In our hospital, electronic record linkage of records included data collected through systematic validation of primary care notes and all other regional diabetes-specific data sources. Using manual validation (the systematic collection of data from paper records) as a gold standard, data error rates from individual clinical sources were calculated. An automated system was designed for managing the results of these comparisons allowing the practical ‘weighting’ of data from disparate sources and hence the summarization of a patient record accordingly. By applying ‘weighting’ to the patient data, a single summary for each patient was generated. Thirty-two per cent of patients had ambiguities (potential errors) for type of diabetes alone and these were fed back to their clinical sources. In almost all cases, the clinic database was found to be in error and could be subsequently corrected. This method has proven itself through the feedback of numerous data entry errors to clinic databases. Systems that include such an enterprise-wide negative feedback loop can actually improve data quality in clinical systems and hence reduce the risk of inappropriate treatment.