Classification and reuse of clinical information in general practice : studies on diagnostic and pharmacological information in electronic patient record systems

Abstract

The reuse of clinical information from the electronic patient record (EPR) for all health care areas - administrative, clinical, teaching and research - win constitute a challenge in coming decades. Classification and coding of clinical information is essential for such data retrieval.

The prevalence, characteristics and management of patients with chronic heart failure were investigated in a four-year retrospective study including 46 general practitioners (GPs) (study I). Out of the registered population, 667 (0.7%) patients had chronic heart failure. About 99% had some chronic co-morbidity, and 27% were treated with angiotensin converting enzyme inhibitors. During the study period 66% had an echocardiograph. Gender differences were found concerning co- morbidity and management. It seemed that these patients could be more adequately managed, and the EPRs were feasible for research.

A working model for continuous medical education based on feedback on prescribing rates and educational outreach visits was evaluated in a randomised study, with three parallel intervention groups/fields, including 40 GPs (study II). In the hypertension field desired trends were recorded, with a significant effect on prescriptions for agents acting on the reninangiotensin system. In the peptic ulcer/dyspepsia field desired trends were recorded, and in the depression group only minor changes. The model was promising, and the EPRs were feasible for retrieval of prescription data.

The Swedish primary health care version of International Statistical Classification of Diseases and Related Health Problems (ICD-10) was reclassified in a preceding study using a concept system approach, and implemented in a web-based browser. In a randomised experimental diagnostic coding trial, six GPs coded each of 152 medical problems in patient vignettes by means of three versions of ICD-10, one with a compositional structure (study III). At code level the reliability was poor and almost identical when comparing the three versions. At aggregated level the reliability was good, and it was somewhat better in the compositional structure.

Necessary conditions for the establishment of a database of diseases and health problems based on EPRs were investigated in a postal questionnaire study including 300 randomly selected GPs in Sweden (study IV). A total of 184 (61%) answered, and 92% used an EPR system. About 88% used an ICD based classification, and the classification in use was computerised for 74%. About 76% reported classifying at least one symptom or disease per encounter. Classification of diseases was considered important for follow-up by 83%. EPRs provide several fundamentals for a database of diagnostic data.

The textual content, health problems and diagnostic codes in EPRs were investigated in a retrospective database study including 20 randomly selected GPs and a total of 400 records (study V). About 15% of the records were problem-oriented. The mean number of words per record was 99.4. The mean number of problems managed per record was 1.3 and there were on average 1.1 diagnostic codes per record. The proportion of correct codes was estimated at 97%. The most frequent disease was essential hypertension (8%). It seemed that problem-oriented EPR systems enforced the coding activities.

In conclusion, data from EPRs in general practice was found feasible for research, education, and the establishment of a database of diagnostic data. Positive aspects of the reliability of diagnostic coding were found in new computerised versions of a traditional classification of diseases.