5A 112 CHAPTER 5A an injury was calculated using a binary logistic regression model that modelled acute workload week categories, chronic workload week categories, the ACWR week categories, the week-to-week and fortnightly ACWR difference categories as independent variables and injury/no injury as dependent variable. The ‘Low’ week category and the ‘Low decrease’ ACWR difference category were the reference categories. The data were statistically analysed using R version 3.4.4 (R Foundation for Statistical Computing, Vienna, Austria) and the caret library, version 6.0.79 2.6. Determining relative risk and prediction A two-by-two table was used to determine basics for the metrics [31]. The two-by-two table consists of four categories: (i) True Positive (TP; i.e., the support for the identified associated categories in relation with the injury incidence), (ii) True Negative (TN; i.e., the support for the non-associated categories and the non-injury incidence), (iii) False Positive (FP; i.e., the support for the identified associated categories which did not result in an injury), and (iv) False Negative (FN; i. e., the support for the non-associated categories which resulted into an injury). The metrics for injury occurrence were the relative risk (RR), the standard error (SE) of log RR, the 95% confidence interval (CI 95%), and the p-value of the relative risk. The RR, its SE, the CI 95% and p-value were calculated accordingly [31], [32]. The predictive power of the significant workload variables and the affiliated categories were calculated by the sensitivity and specificity [33]. The relative risk was calculated as RR = '" &# // (( "' #& %% '" ## )), for which the SE of the log of the RR can be calculated as SE{ln(RR)} = í" 3 # +' 3 &− "#% 3 '# − "&% 3 "# . When a category caused a division by zero in calculation of the RR or the SE, 0.5 was added to all four categories of the two-by-two table [3]. We calculated the 95% CI as } ( ) ±1.96∗ { . We determined the p-value with the calculated z-value, z-value = value = : f2{.2 .2 .2 .( 0( 000) ) }. Finally, we calculated the sensitivity and specificity. The sensitivity was calculated as the proportion of correctly identified injuries, as sensitivity = "# " % # '&. The specificity was calculated as the proportion of correctly identified non-injuries, as specificity = "& " % & '#. The calculations were performed using Microsoft Excel2016. We confirm the study meets the ethical standards of the International Journal of Sports Medicine [34].
RkJQdWJsaXNoZXIy MjY0ODMw