Skip to Main Content
 

Global Search Box

 
 
 
 

Files

File List


ucin1177418766.pdf (2.12 MB)

ETD Abstract Container

Abstract Header

A Special Inference Problem in Repeated Measures Design with Applications to Pulse Oximetry

Ndikintum, Nfii Kangong
http://rave.ohiolink.edu/etdc/view?acc_num=ucin1177418766

Abstract Details

2007, PhD, University of Cincinnati, Medicine : Biostatistics (Environmental Health).
The objective of pulse oximetry studies is to demonstrate that the paired repeated measurements from a pulse oximeter (SpO2) and a co-oximeter (SaO2), agree sufficiently with respect to a pre-specified agreement score. The measure of agreement in pulse oximetry studies is the Accuracy Root-Mean Square (ARMS). To account for both systematic and random components of error, pulse oximeter manufacturers have defined accuracy to represent the root-mean square between the pulse oximeter and co-oximeter values. The current methods for analyzing pulse oximetry method comparison studies are mostly descriptive. Pennello (2002) initiated the discussion on sample size calculation and test of statistical hypotheses applied to the ARMS. Using Pennello’s results as a springboard, the purpose of this dissertation was to further study the distributional properties of the ARMS and recommend sample size and test of hypothesis methodologies appropriate to this agreement score. We proposed and evaluated via simulation studies a more robust method for testing the statistical hypothesis of the accuracy of a pulse oximeter. The proposed Z-test, based on restricted maximum likelihood estimate of the between-subject variance is superior to the Pennello test; the alpha level is maintained and the power of the test is superior to the power of the Pennello test. For all configurations studied, the Pennello Z-test is conservative (undersized). The proposed Z-test is therefore highly recommended to be incorporated in future pulse oximetry studies as a test statistic of choice. The proposed Z-test would obviate the need for regulatory agencies (e.g., the FDA) to base device approval decisions on subjective assessments of accuracy root-mean square (i.e., comparing ARMS to a target of 3%). A Likelihood Ratio test was also derived and evaluated via simulation studies. It is highly recommended not to use the LRT in studies aimed at assessing the accuracy of a pulse oximeter, since the test is liberal (oversized). We also derived a sample size algorithm for testing the accuracy of a pulse oximeter and via simulation studies we evaluated the performance of the proposed algorithm. The resulting sample size should be inflated if an effect size greater than one is anticipated.
Jeffrey Welge (Advisor)
Richard Hornung (Other)
Marepalli Rao (Other)
Rakesh Shukla (Other)
Paul Succop (Other)
169 p.
Repeated Measures; Accuracy; Root-Mean Square; Pulse Oximeter

Recommended Citations

Citations

  • Ndikintum, N. K. (2007). A Special Inference Problem in Repeated Measures Design with Applications to Pulse Oximetry [Doctoral dissertation, University of Cincinnati]. OhioLINK Electronic Theses and Dissertations Center. http://rave.ohiolink.edu/etdc/view?acc_num=ucin1177418766

    APA Style (7th edition)

  • Ndikintum, Nfii. A Special Inference Problem in Repeated Measures Design with Applications to Pulse Oximetry. 2007. University of Cincinnati, Doctoral dissertation. OhioLINK Electronic Theses and Dissertations Center, http://rave.ohiolink.edu/etdc/view?acc_num=ucin1177418766.

    MLA Style (8th edition)

  • Ndikintum, Nfii. "A Special Inference Problem in Repeated Measures Design with Applications to Pulse Oximetry." Doctoral dissertation, University of Cincinnati, 2007. http://rave.ohiolink.edu/etdc/view?acc_num=ucin1177418766

    Chicago Manual of Style (17th edition)

ucin1177418766
1,041
© 2007, all rights reserved.
This open access ETD is published by University of Cincinnati and OhioLINK.