Assessing Data Quality in Manual Entry of Ventilator Settings

From Clinfowiki
Jump to: navigation, search

Vawdrey DK, Gardner RM, Evans RS, Orme JF Jr, Clemmer TP, Greenway L, Drews FA. Assessing data quality in manual entry of ventilator settings. J Am Med Inform Assoc. 2007 May-Jun;14(3):295-303.

INTRODUCTION: This article compares data acquired directly from ventilators with manually entered data by respiratory therapists. Disclosures: Although one of the authors was funded by a training grant, there appears to be no conflict of interest.

Background: Traditionally data from ventilators is entered on a flow sheet by a respiratory therapist. Depending on the hospital, this data may be entered into a hand written chart or entered using a computer-charting program. Today we have the option of directly uploading the data from the ventilator to the computerized chart. This study compares the manually entered data with direct output from the ventilator.

Question: Is manually entered data as good as direct output from the ventilator to the chart? If the answer is No, this can seriously affect computerized management protocols and decision support. Methods: Prospectively collected data from ventilators using a MIB interface in four intensive care units over a 12-month period. This data was compared with manually entered data by respiratory therapist. Data quality was assessed for accuracy and delay.

Results: The percentage of times the settings matched ranged from 99% to 75.9% (errors of manual charting from 1% to 24.1%). Average charting delay was 6.1 minutes. 3.9% of the recommendations for computerized ventilator management protocols were based on wrong data.

Conclusion: Poor data quality effects human and computer decision-making. Utilizing MIB interface to directly input data to the chart eliminates delay and reduces errors

Commentary: The buzzword today is “Decision Support”. This is based on the data you are providing the decision support system. Improving data capture in a timely fashion and with no errors is the cornerstone of this function. Using interfaces and reducing the human input variable can improve the system. The human factor is necessary to ultimately approve the decision and be legally responsible for it.