- Poster presentation
- Open Access
Data completeness in the Finnish Intensive Care Quality Consortium database
© BioMed Central Ltd. 2007
- Published: 22 March 2007
- Data Completeness
- Intensive Care Medicine
- Scheme Variable
- Admission Data
- Clinical Information System
Benchmarking has been an essential part of intensive care medicine in Finland since 1994. At present, web-based quality/performance reports are shared with the 24 members of the Finnish Intensive Care Quality Consortium (FICQC). Thirteen ICUs collect FICQC data manually and 11 ICUs utilize data collection software (IVT) integrated with the Clinical Information System (CIS). In recent literature, the completeness of data between centralized medical benchmarking registries varies widely [1, 2]. We hypothesized that: (1) the completeness of data in FICQC has increased over the years and (2) the variation between the different units still exists.
We assessed data completeness of a Finnish ICU quality benchmarking database from 1998 to March 2006 containing 93,964 admission records. The data completeness was defined as a ratio of available and required data at ICU admission level. We evaluated the dataset and selected 19 most significant admission scheme variables to be included in completeness ratio calculations.
The majority of data (77.5%) was collected with the manual system and the remaining 22.5% with an integration software. The mean admission data completeness ratio (CR) increased from 85.3% at 1998 to 97.9% at 2005 (P = 0.01). Between the ICUs, the mean CR varied from 91.6% to 99.6% (P = 0.01). The mean CR of the data collected with the IVT was 98.7% and with the manual system was 95.1% (P = 0.01). The rate of 100% complete records per patient was 48.7% and it increased from 0.0% in 1998 to 71% in 2005.
Data completeness in the FICQC has improved during the study period, although there is still significant variation between ICUs. Improved data completeness and decreased proportion of missing data are most likely due to the increasingly common use of CIS and automated data collection. We conclude that measuring/reporting the amount of missing data is mandatory when data collection and data management procedures for benchmarking are being developed.