Continuous Noninvasive Measurement of Pulsus Paradoxus Complements Medical Decision Making in Assessment of Acute Asthma Severity: Relationship Between Objective Scoring and PP
All variable distributions were assessed for violation of the assumption of normality based on skewness, the Shapiro-Wilk statistic (a = 0.01), and visualization. Variables having a significant deviation from normal via the Shapiro-Wilk statistic were submitted to the following three linear transformations: square root; natural logarithm; and inverse. The linear transformation that improved the distribution the most was selected. In addition, both the untransformed and transformed distributions were visually inspected to verify normality.
The interrater reliability of the objective scoring composite and subscales (transformed where necessary) was estimated using the intraclass correlation coefficients (ICCs) as described by Shrout and Fleiss. A mixed model was used, with “rater” treated as a random variable since each patient was rated by a pair of physicians who were pulled from a sample of possible physicians (the same two physicians were not always used for each patient, though the same two physicians were used for both pretreatment and posttreatment time points within a given patient). The ICC of the raters was used as an index of reliability of actual rater judgments. The estimated ICC of the mean of the two raters (n = 2) was used throughout the analysis.
For objective scoring measures (composite and subscales) that met or exceeded an ICC of 0.80 for the mean of the ratings at both time points, the mean of the two raters for each patient was assessed for its relationship to AT-PP using a repeated measures (ie, pretreatment/posttreatment) general linear model with the score (continuous) as a fixed effect. In addition, each objective scoring measure (including those that failed to meet the ICC criterion) was evaluated for predicting AT-PP using hierarchical linear models (PROC MIXED, SAS Version 9.1; SAS Institute) to assess whether or not on average there was a relationship between observer ratings and AT-PP (ie, the mean slope within rater). Residuals were examined for systematic deviations and for overall model fit, and scatter-plots were examined to verify and assist in interpreting the model parameters. The cost of care was based on hospital and physician charges for outpatient and inpatient treatment of asthma. The cost of appropriate inpatient care was determined by the average level of service, the cost of care per day, and the average length of stay for patients with conditions denoted by International Classification of Diseases, ninth revision, (ICD-9) codes 49390 and 49392 from inpatient billing records for 2004. This cost also included the ED charges.