This work contemplates the main difficulties related to the capture of clinical data into the computer. We have designed an experience that consists of capturing into the computer, in absence of automatic protections, 335 structured interviews coded on paper by trained interviewers. The general objective of this study consists of rating the efficiency of a new computerized system (based on the feedback principle) for controlling and improving the capture: the random verification (RV) of a percentage of questionnaires. To accomplish this objective, we have measured and classified the mistakes that professional operators commit during the data capture. The specific objectives are: 1) to rate the efficiency of the RV compared to double entry (DE) -this last strategy is the most commonly used for improving the quality of the capture-; 2) to check the types of mistakes contained in the recorded data; and 3) to rate the operators' efficiency for resolving the discrepancies that appeared during verification (through the measurement of the mistakes that persist in the fields with disagreements during the second capture of the RV and the DE). The principle results obtained indicate that the RV and the DE with different operators offer similar performance, while the simple entry and the DE with only one operator do not constitute effective systems for guaranteeing the quality of the capture. In the discussion we emphasize what aspects of the measurement and the data capture are improved using computerized controls, what minimum quality requirements should be present during the recording and what elements the computerized systems require for operating efficiently.
|Publication status||Published - 1 Feb 2001|