OBJECTIVES: We evaluated: a) whether the frequency of variable measurement could influence the performance of the Pediatric Risk of Mortality (PRISM) score; b) whether measurement frequency of physiologic variables varied between individual pediatric intensive care units (ICUs), and c) if so, how much of this variability could be attributed to institution-level and patient-level factors. DESIGN: Prospective cohort. SETTING: Sixteen pediatric ICUs, chosen for their diversity. PATIENTS: Consecutive admissions (n = 5,415). INTERVENTIONS: None. MEASUREMENTS AND MAIN RESULTS: First, the measurement frequency of the 14 physiologic variables in the PRISM score was included in the logistic regression model predicting mortality risk. Measurement frequency was not significant, alone or in its interaction with the PRISM score. Second, the presence or absence of measurement of each physiologic variable was included in the logistic model using indicator variables; none was significant. Finally, the contribution of the individual pediatric ICUs and patient factors in explaining the variability in the frequency of physiologic variable measurement were investigated with linear regression analysis. In this analysis, the separation of severity of illness from measurement frequency was accomplished by computing the PRISM score from the first 4 hrs and measurement frequencies from hours 5 through 24. Overall, 70.22% (r2) of the variability of measurement frequency could be explained by the factors included in the linear regression model. The individual ICUs accounted for a total of only 6.23% of the explained variability and no individual hospital accounted for > 1.44% of the variability. Other variables positively correlated with measurement frequency included the presence or absence of a pediatric intensivist, and whether the institution was a children's hospital or not. Variables negatively correlated with measurement frequency included larger ICUs and house officers assigned to the ICU. CONCLUSIONS: Although measurement frequency is associated with unit-level factors, their contribution to the overall variability is small and unlikely to influence the accuracy or reliability of the PRISM score. It is unlikely that there are routine biases associated with differences in measurement frequency of PRISM variables within the spectrum of care practices that now exist.
OBJECTIVES: We evaluated: a) whether the frequency of variable measurement could influence the performance of the Pediatric Risk of Mortality (PRISM) score; b) whether measurement frequency of physiologic variables varied between individual pediatric intensive care units (ICUs), and c) if so, how much of this variability could be attributed to institution-level and patient-level factors. DESIGN: Prospective cohort. SETTING: Sixteen pediatric ICUs, chosen for their diversity. PATIENTS: Consecutive admissions (n = 5,415). INTERVENTIONS: None. MEASUREMENTS AND MAIN RESULTS: First, the measurement frequency of the 14 physiologic variables in the PRISM score was included in the logistic regression model predicting mortality risk. Measurement frequency was not significant, alone or in its interaction with the PRISM score. Second, the presence or absence of measurement of each physiologic variable was included in the logistic model using indicator variables; none was significant. Finally, the contribution of the individual pediatric ICUs and patient factors in explaining the variability in the frequency of physiologic variable measurement were investigated with linear regression analysis. In this analysis, the separation of severity of illness from measurement frequency was accomplished by computing the PRISM score from the first 4 hrs and measurement frequencies from hours 5 through 24. Overall, 70.22% (r2) of the variability of measurement frequency could be explained by the factors included in the linear regression model. The individual ICUs accounted for a total of only 6.23% of the explained variability and no individual hospital accounted for > 1.44% of the variability. Other variables positively correlated with measurement frequency included the presence or absence of a pediatric intensivist, and whether the institution was a children's hospital or not. Variables negatively correlated with measurement frequency included larger ICUs and house officers assigned to the ICU. CONCLUSIONS: Although measurement frequency is associated with unit-level factors, their contribution to the overall variability is small and unlikely to influence the accuracy or reliability of the PRISM score. It is unlikely that there are routine biases associated with differences in measurement frequency of PRISM variables within the spectrum of care practices that now exist.
Authors: S M Tibby; D Taylor; M Festa; S Hanna; M Hatherill; G Jones; P Habibi; A Durward; I A Murdoch Journal: Arch Dis Child Date: 2002-11 Impact factor: 3.791
Authors: Murray M Pollack; J Michael Dean; Jerry Butler; Richard Holubkov; Allan Doctor; Kathleen L Meert; Christopher J L Newth; Robert A Berg; Frank Moler; Heidi Dalton; David L Wessel; John Berger; Rick E Harrison; Joseph A Carcillo; Thomas P Shanley; Carol E Nicholson Journal: Pediatr Crit Care Med Date: 2013-06 Impact factor: 3.624
Authors: Julie C Leonard; Cassandra D Josephson; James F Luther; Stephen R Wisniewski; Christine Allen; Fabrizio Chiusolo; Adrienne L Davis; Robert A Finkelstein; Julie C Fitzgerald; Barbara A Gaines; Susan M Goobie; Sheila J Hanson; Hilary A Hewes; Laurie H Johnson; Mark O McCollum; Jennifer A Muszynski; Alison B Nair; Robert B Rosenberg; Thomas M Rouse; Athina Sikavitsas; Marcy N Singleton; Marie E Steiner; Jeffrey S Upperman; Adam M Vogel; Hale Wills; Margaret K Winkler; Philip C Spinella Journal: Crit Care Med Date: 2021-11-01 Impact factor: 9.296