Start Date

22-5-2012 4:30 PM

End Date

22-5-2012 6:00 PM

Document Type

Event

Description

BACKGROUND: Patient-reported outcome (PRO) research requires valid and sensitive measures. Relative Validity (RV) offers an objective way to compare the validity of different PRO measures in discriminating groups of patients or occasions.

There is no significance test associated with RV. We applied bootstrap to estimate the confidence interval (CI) of RV to better interpret the differences in RV.

METHODS: The CKD-specific legacy (KDQOL Burden, Symptom, and Effect), generic health scales (SF-12), and Kidney Disease Impact Scale (KDIS) were administrated to 453 CKD patients. ANOVA-based RV coefficients were computed to compare how well each scale discriminated between three clinically-defined severity groups (Dialysis > Stage 3-5 > Transplant). Bootstrap was used to construct CI to determine whether the differences in RV were significant in comparisons between each scale and the best legacy standard- KDQOL Burden. Factors of sample size, number of bootstrap replications, bootstrap method were varied to investigate their impacts.

RESULTS: In comparison with KDQOL Burden (RV=1), using 95% CI, differences were non-significant for KDIS (RV=1.13), KDQOL Effect (RV=.99), SF-12 RP (RV=.77) and PF (RV=.70). SF-12 PCS (RV=.60) was at borderline. The other measures were significantly poorer in discriminating the patients.

Sample size played a substantial role. 300 patients for 3 groups greatly reduced the standard errors compared to 100 patients. A larger sample size greatly increased the power of detecting the differences.

The number of replications did not have consequential influence. The types of BCa and percentile intervals were preferred as all bootstrap distributions were skewed. The magnitude of chosen standard measure’s F-statistics appeared to have a noticeable impact on CI too.

CONCLUSIONS: Bootstrapping appears to be valuable in comparing the validity of PRO measures from a statistical perspective. The significance test of RV was affected by the sample size, magnitude of RV, and F-statistic of standard measure.

Creative Commons License

Creative Commons Attribution-Noncommercial-Share Alike 3.0 License
This work is licensed under a Creative Commons Attribution-Noncommercial-Share Alike 3.0 License.

Share

COinS
 
May 22nd, 4:30 PM May 22nd, 6:00 PM

Using bootstrap to compare the validity of PRO measures in discriminating among CKD patients

BACKGROUND: Patient-reported outcome (PRO) research requires valid and sensitive measures. Relative Validity (RV) offers an objective way to compare the validity of different PRO measures in discriminating groups of patients or occasions.

There is no significance test associated with RV. We applied bootstrap to estimate the confidence interval (CI) of RV to better interpret the differences in RV.

METHODS: The CKD-specific legacy (KDQOL Burden, Symptom, and Effect), generic health scales (SF-12), and Kidney Disease Impact Scale (KDIS) were administrated to 453 CKD patients. ANOVA-based RV coefficients were computed to compare how well each scale discriminated between three clinically-defined severity groups (Dialysis > Stage 3-5 > Transplant). Bootstrap was used to construct CI to determine whether the differences in RV were significant in comparisons between each scale and the best legacy standard- KDQOL Burden. Factors of sample size, number of bootstrap replications, bootstrap method were varied to investigate their impacts.

RESULTS: In comparison with KDQOL Burden (RV=1), using 95% CI, differences were non-significant for KDIS (RV=1.13), KDQOL Effect (RV=.99), SF-12 RP (RV=.77) and PF (RV=.70). SF-12 PCS (RV=.60) was at borderline. The other measures were significantly poorer in discriminating the patients.

Sample size played a substantial role. 300 patients for 3 groups greatly reduced the standard errors compared to 100 patients. A larger sample size greatly increased the power of detecting the differences.

The number of replications did not have consequential influence. The types of BCa and percentile intervals were preferred as all bootstrap distributions were skewed. The magnitude of chosen standard measure’s F-statistics appeared to have a noticeable impact on CI too.

CONCLUSIONS: Bootstrapping appears to be valuable in comparing the validity of PRO measures from a statistical perspective. The significance test of RV was affected by the sample size, magnitude of RV, and F-statistic of standard measure.

 

To view the content in your browser, please download Adobe Reader or, alternately,
you may Download the file to your hard drive.

NOTE: The latest versions of Adobe Reader do not support viewing PDF files within Firefox on Mac OS and if you are using a modern (Intel) Mac, there is no official plugin for viewing PDF files within the browser window.