RATIONALE, AIMS AND OBJECTIVES: Graphical monitoring tools are needed for real-time quality evaluation in intensive care unit. The variable life-adjusted display (VLAD) provides a directly interpretable assessment indicating whether the overall performance is better or worse than expected on the basis of the predicted risk of failure. The aim of this study is to quantify the ability of VLAD charts to early recognize a worsening in clinical performance. METHODS: A Monte Carlo experiment simulating the sequence of successes and failures of an intensive care unit is performed; the predicted mortality is calculated by means of the Simplified Acute Physiology Score 3 admission score. From a given position in the admissions sequence, we increased the probability of death; we calculated: (i) the surveillance system delay in responding to the mortality increase; (ii) the percentage of cases where the VLAD has been able to give an alarm within the first 5, 10, 20 and 60 deaths occurred after the increase of probability of death; and (iii) the percentage of false declarations of increase (anticipated alarms). RESULTS: The frequency distribution of the alarm delays shows VLAD was not always able to early detect mortality increase. Only a very small number of anticipated alarms were given. CONCLUSIONS: Variable life-adjusted display ability to signal is mild and strictly correlated with the institution volume of activity. Therefore, the use of VLAD seems to be not always advisable, and an integration between VLAD and other well-documented tools as CUSUM charts could be preferable.
RATIONALE, AIMS AND OBJECTIVES: Graphical monitoring tools are needed for real-time quality evaluation in intensive care unit. The variable life-adjusted display (VLAD) provides a directly interpretable assessment indicating whether the overall performance is better or worse than expected on the basis of the predicted risk of failure. The aim of this study is to quantify the ability of VLAD charts to early recognize a worsening in clinical performance. METHODS: A Monte Carlo experiment simulating the sequence of successes and failures of an intensive care unit is performed; the predicted mortality is calculated by means of the Simplified Acute Physiology Score 3 admission score. From a given position in the admissions sequence, we increased the probability of death; we calculated: (i) the surveillance system delay in responding to the mortality increase; (ii) the percentage of cases where the VLAD has been able to give an alarm within the first 5, 10, 20 and 60 deaths occurred after the increase of probability of death; and (iii) the percentage of false declarations of increase (anticipated alarms). RESULTS: The frequency distribution of the alarm delays shows VLAD was not always able to early detect mortality increase. Only a very small number of anticipated alarms were given. CONCLUSIONS: Variable life-adjusted display ability to signal is mild and strictly correlated with the institution volume of activity. Therefore, the use of VLAD seems to be not always advisable, and an integration between VLAD and other well-documented tools as CUSUM charts could be preferable.
Authors: Danilo Fusco; Anna P Barone; Chiara Sorge; Mariangela D'Ovidio; Massimo Stafoggia; Adele Lallo; Marina Davoli; Carlo A Perucci Journal: BMC Health Serv Res Date: 2012-01-27 Impact factor: 2.655
Authors: Nick Andrianopoulos; Damien Jolley; Sue M Evans; Caroline A Brand; Peter A Cameron Journal: BMC Health Serv Res Date: 2012-08-28 Impact factor: 2.655