Statistical Analyses. Author Manuscript To make variables more clinically sensible for decision making in the ED, all patient history and physical examination variables were assessed as dichotomous (ie, presence or absence of a finding). We determined the interobserver agreement for each clinical finding by calculating the unweighted Xxxxx kappa (κ) statistic with two-sided 95% confidence intervals (CIs) as well as the percent agreement. The interobserver agreement was categorized based on the κ point estimates as slight (0–0.2), fair (0.21–0.4), moderate (0.41– 0.6), substantial (0.61–8), and almost perfect (0.81–1.0). (25,26) For each variable, we excluded from the κ analysis any paired observations for which data were missing or at least one assessor marked the variable as “unsure.” Author Manuscript
Statistical Analyses. The concordance rates for responses to the clinical questions raised by the referring physicians, clinically pertinent findings, incidental findings, and proposed investigations were tabulated to in 88.1% of cases (95% CI, 80.6%–92.9%). The Xxxxx n for pos- itive studies was 0.86 (0.77– 0.95) and 0.59 (0.45– 0.74) for the presence/absence of any clinically pertinent findings and for inci- dental findings, respectively. Readings were in agreement for 75.6% (67.2%– 82.5%), 65.5% (56.6%–73.5%), and 86.6% (79.2%–91.6%) of clinically pertinent findings, incidental findings, and recommendations for further in- vestigations, respectively (Table 1). Rates were similar for the 2 sub- groups (referred from the neuro-oncology clinic or from all other clinics). Class 2 discrepancies in reporting clinically pertinent findings were found in 18 cases (15.1%). They are summarized in Table 2. Examples include the presence or absence of a tumor recurrence FIG 2. Class 2 discrepancies in incidental findings. A, One reader reported a polypoid posterior nasal lesion (arrow), whereas the other observer did not mention this incidental finding. B, A pineal cyst (circle) was mentioned in only 1 of the 2 reports. independent observers. Their study showed agreement in 51%, 61%, and 74% of abdominal, chest, and skeletal x-rays, respectively. They also assessed performance by calculating n statistics of interobserver agreement. Weighted n values between pairs of observers were higher with skeletal (0.76 – 0.77) than with chest (0.63– 0.68) or abdominal (0.50 – 0.78) examinations. In a meta- analysis conducted by Xx et al,2 the global discrepancy rate was 7.7% (in- cluding a major discrepancy rate of 2.4%). The major discrepancy rate var- ied according to body region: It was lower for head (0.8%) and spine CT (0.7%) than chest (2.8%) and abdomi- nal CT (2.6%). Blinding of the reference (n = 4, Fig 1A), the growth of a meningioma (n = 2), the evolu- tion of chronic subdural hematomas (n = 2), and the presence of a lytic bone lesion (Fig 1B). These discrepancies were normally distributed between readers (n = 1, 3, 5, 7, 11, 5, 3, 1 for 36 discrepant reports). There was no significant difference between contrast-enhanced (n = 9 of 53) and nonenhanced studies (n = 9 of 66, P = .62). Class 1 discrepancies in clinically pertinent findings were seen in the interpretation of 11 cases. Examples include the location of recent ischemic lesions (n = 2), tumor extensions (n = 2), or the disconnecti...
Statistical Analyses. The multirater n statistics were computed by using the macro MAGREE with SAS, Version 9.3 (SAS Institute, Xxxx, North Car- olina). This macro implements the methodology of Fleiss et al,27 measuring the agreement when the number of raters is >2. This method also allowed identifying, for each scale, the categories in Table 1: Interobserver agreement using the TICI reperfusion scale Ob2 Ob3 Ob4 Ob5 Ob6 Ob6a Ob7 Ob8 Ob9 Ob1 0.497 ± 0.098 0.411 ± 0.102 0.478 ± 0.103 0.544 ± 0.101 0.508 ± 0.104 0.315 ± 0.114 0.517 ± 0.100 0.580 ± 0.100 0.519 ± 0.094 Ob2 0.419 ± 0.100 0.286 ± 0.102 0.576 ± 0.096 0.506 ± 0.102 0.320 ± 0.115 0.458 ± 0.096 0.538 ± 0.099 0.330 ± 0.089 0.330 ± 0.089 Ob3 0.197 ± 0.088 0.284 ± 0.100 0.513 ± 0.103 0.191 ± 0.096 0.339 ± 0.103 0.345 ± 0.105 0.404 ± 0.103 Ob4 0.510 ± 0.100 0.343 ± 0.103 0.352 ± 0.103 0.384 ± 0.098 0.583 ± 0.098 0.297 ± 0.087 Ob5 0.602 ± 0.101 0.594 ± 0.102 0.712 ± 0.091 0.752 ± 0.082 0.397 ± 0.094 Ob6 0.525 ± 0.107 0.465 ± 0.108 0.610 ± 0.101 0.425 ± 0.093 Ob6a 0.542 ± 0.096 0.421 ± 0.106 0.283 ± 0.085 Ob7 0.511 ± 0.102 0.442 ± 0.102 Ob8 0.423 ± 0.095 All observers n = 0.44570 ± 0.013176; P < 0.001 Note:—Inter-observer Kappa values ≥ 0.6 are highlighted in bold type.
Statistical Analyses. Statistical analyses of the registry data will be performed in regular intervals. For qualitative parameters, descriptive statistics such as the population size and the percentage of available data for each class of the parameter will be presented. Quantitative parameters will be summarized by presenting, for example, the population, the mean, standard deviation (SD), median, minimum and maximum values, or interquartile ranges. Statistics may be presented, if sample size permits, for cohorts of interest. Due to the observational nature of the registry, all analyses will be considered exploratory. Analysis Populations All patients entered in CARE are intended to be included in the analyses. Patients with missing data will not be excluded from the patient analysis population, but will be included to the extent that evaluable data are present. However, some patients with missing values may be excluded from specific analyses.
Statistical Analyses. During the sample size calculations, we anticipated a normal distribution of the arthritis scores. Nevertheless, upon visual inspection of the data, the standards required for parametric statistical testing were not met. Therefore, the non-parametric Xxxx-Xxxxxxx U test was applied for all statistical analyses to determine significance of observed differences. To determine whether TP administration had an effect on Pred treatment in the CAIA model and side effects of Pred, significant differences were only determined between the Veh-treated group and the Pred-treated group, with or without TP. Important to note is that, with this study, we want to find out whether analgesia can be applied and would provide us with a similar extent in parameters required to evaluate whether alternatives to Pred are superior. Therefore, no analysis has been performed between the non-TP and TP-treated groups, except when the effect of TP on GR target-gene expression was evaluated. All statistical analyses were performed using GraphPad Prism 9. Notion should be taken when interpreting the p-values since the Veh + TP-treated group included only 3 mice and therefore comparisons to this group cannot reach statistical significance.
Statistical Analyses. 2.8.1 Selection and missing values We excluded test and/or retest patient questionnaires if they had been completed >30 days post-consultation, and physician questionnaires if they had been completed >7 days post-consultation (Figure 1). We assumed that a longer period would be detrimental to participants’ recollection of the decision-making process. We handled missing values according to authors’ recommendations, if provided in the original or Dutch validation paper (see section 2.5).12, 13, 34 For the other questionnaires and the iSHARE questionnaires (see section 2.4), we only report scores when all respective items had been completed. We report sample sizes per analysis, since these may differ due to missing values.
Statistical Analyses. Categorize stormwater analytes in each basin by relative likelihood of having significant uncontrolled sources.
Statistical Analyses. A statistical mixture model was used for the analysis of the harborwide stormwater outfall data. A detailed description of the model (including the key assumptions, limitations, and statistical output) is provided in Appendix C. The model was used to analyze the distribution of the geometric mean concentrations for harborwide outfall basins and to assign each outfall basin to one of three concentration groups (i.e., lower, moderate, or higher). This concept is consistent with the JSCS method for identifying low, medium, and high priority sites. Statistical analysis was performed for the analytes selected based on the data screening described in Section 4.2 (unless the number of samples was insufficient to conduct a valid statistical analysis; see Table 4-1). Graphs of the results of the statistical evaluation are included in Appendix C. Use of the full harborwide data set in the statistical evaluation allowed comparison of stormwater concentrations in City outfall basins to other stormwater discharges in the Portland Harbor Study Area. Source tracing categories were statistically identified based on specific outfall concentrations relative to harborwide concentration levels. For 5 See data tables in Appendices A and B for data collected by the City; see Anchor and Integral 2008b for City outfall basin data collected by others. each analyte included in the statistical evaluation, each outfall basin was assigned to one of three source tracing categories:
Statistical Analyses. We first described the study population by presenting their demographic, socioeconomic, and health-related characteristics by their disclosure status and residence. To examine the exposure and outcome associations, we conducted multilevel, multivariable modeling using generalized estimating equations (GEE) with a binomial distribution and logit link and an exchangeable correlation structure, which accounted for clustering by ZCTA. Subjects with missing data for the outcome and/or covariates were excluded from the analyses. Our modeling strategy first examined the unadjusted association between exposure and outcome and then proceeded to examine interaction between urban-rural residence and perceived neighborhood tolerance perception on the multiplicative scale (interaction-only model). In the fully adjusted model, we adjusted for individual-level (age, race/ethnicity, education, sexual identity, and experienced sexual identity-related stigma) and ZCTA-level covariates (region, median household income, percent with less than a high school education, and percent of same-sex households) to address potential confounding. For the residence and disclosure to a HCP association, the models were adjusted for the same covariates, in addition to health insurance status, seeing a HCP in the past year, and sexual identity-related stigma in health care. These covariates were selected based on their associations with sexual identity disclosure, overall and to HCPs specifically, from previous literature (30,50,63,74,78,88,89). We also examined collinearity among the covariates, and the ZCTA-level percent of individuals with health insurance was not included in the models due to collinearity issues. Missing data and sensitivity analyses There were 10,023 participants who had disclosed their sexual identity, and 28% (2,787) did not respond to the subsequent question asking whether they had disclosed their sexual identity to a HCP, the second outcome of interest. We did not find any meaningful differences between those who responded and those who did not by age, education, or sexual identity. Those who were missing disclosure to HCP data had a higher proportion of missing for health insurance status (16% versus 10%) and having seen a health provider in the past year (13% versus 4%), compared to those who did not have missing data. There were 4,105 participants (41%) who were missing data on at least one covariate. We conducted sensitivity analyses to assess the impact of the ...
Statistical Analyses. Case information including sex, age, vaccination status, clinical manifestation of disease, underlying diseases, hospitalization status, admission to the ICU, length of hospital stay, and S. pneumoniae strain was stratified on race. Statistical differences between white and black cases were examined using chi-square tests for categorical variables and t-tests for continuous variables. Average annual incidence rates were calculated by dividing the average number of cases in 2001 to 2009 by the population of children under five years of age in the surveillance area according to the 2000 U.S. Census data. Stratified incidence rates were calculated for race, sex, and socioeconomic variable categories. The chi-squared test for trend was used to analyze trends in incidence for each socioeconomic variable. Incidence rates could not be stratified by age because the 2000 U.S. Census did not contain population denominators stratified on both age and race for children under five years of age. Poisson regression was used to calculate rate ratios (RRs) and 95% confidence intervals (CIs). The dependent variable for Poisson regression was the number of invasive pneumococcal disease cases for each racial, gender, and socioeconomic category. Independent variables included race, gender, and a census-tract level socioeconomic measure. Only one socioeconomic variable was included per model. Interactions between race and socioeconomic variables were analyzed for each model, and race-specific rates were also reported. In order to identify if any of the socioeconomic measures explained racial disparities in invasive pneumococcal disease in children, ten independent models were created containing race, sex, one socioeconomic measure, and an interaction term for race and the socioeconomic measure if significant. Other race was excluded from models due to low numbers of cases. Adjusted RRs for all variables were reported. Vaccination status and underlying disease could not be included in the model because denominator data did not exist for those variables at the census tract level. To assess the effect of those variables, cases were stratified on vaccination status or underlying disease, and race- stratified RRs were reported. All statistical analyses were performed in SAS Version 9.3 (SAS Institute Inc, Cary, NC). Significance was defined as a 2-tailed p-value < 0.05. RESULTS From 2001 to 2009, there were 935 cases of invasive pneumococcal disease in children less than five years ...
