The impact of surgical volume on hospital ranking using the standardized infection ratio

Figures

• 

  Figure 1. Estimated classification accuracy measures by the number of predicted events. Each hospital corresponds to a triangle. (a) Estimated power for all hospitals. Hospitals highlighted in blue correspond to those with the number of predicted events greater than 1 and power less than 80%. (b) Estimated false positive rate (FPR) for all hospitals. Hospitals in blue correspond to those with the number of predicted events greater than 1 and FPR higher than 10%. (c) Estimated positive predictive value (PPV) for hospitals being ranked into the worst quartile. Hospitals in blue correspond to those with the number of predicted events greater than 1 and PPV lower than 80%. (d) Estimated negative predictive value (NPV) for hospitals not being ranked into the worst quartile. Hospitals in blue correspond to those with the number of predicted events greater than 1 and NPV lower than 90%.

• 

  Figure 2. Estimated classification accuracy measures by the hospital surgical volume (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$n_i$$\end{document}ni). Each hospital corresponds to a triangle. (a) Estimated power for all hospitals. Hospitals highlighted in blue correspond to those with the number of predicted events greater than 1 and power less than 80%. (b) Estimated false positive rate (FPR) for all hospitals. Hospitals in blue correspond to those with the number of predicted events greater than 1 and FPR higher than 10%. (c) Estimated positive predictive value (PPV) for hospitals being ranked into the worst quartile. Hospitals in blue correspond to those with the number of predicted events greater than 1 and PPV lower than 80%. (d) Estimated negative predictive value (NPV) for hospitals not being ranked into the worst quartile. Hospitals in blue correspond to those with the number of predicted events greater than 1 and NPV lower than 90%.

• 

  Figure 3. True and estimated power (a), false positive rate (b), positive predictive value (c), and negative predictive value (d) by surgical volume. True accuracy measures are obtained from simulation (black curves). The corresponding estimates for each of the 100 simulated datasets are obtained by the proposed algorithm (blue curves).

• 

  Figure 4. Empirical power (a), false positive rate (b), positive predictive value (c), and negative predictive value (d) by surgical volume for varying overall event rates.

• 

  Figure 5. Empirical power (a), false positive rate (b), positive predictive value (c), and negative predictive value (d) by overall event rates for two individual hospitals.

• 

  Figure 6. Empirical power (a), false positive rate (b), positive predictive value (c), and negative predictive value (d) by surgical volume for different random effects variances.

• 

  Figure 7. Empirical power (a), false positive rate (b), positive predictive value (c), and negative predictive value (d) by hospital volume when the CMS model was set as the correct model.

• 

  Figure 8. Empirical power (a), false positive rate (b), positive predictive value (c), and negative predictive value (d) by hospital volume when the Claims-plus-EHR model was set as the correct model.