- Dr. Hays' response to
"What is Important to Patients? Quantity or Quality of Life? By
Allen Nissenson, MD, posted in Quality"
Nissenson's blog says the following: “As shown recently by Tracy Mayne, an international authority in this area and a member of DaVita Clinical Research® (DCR®). KDQOL cannot be validated as truly predictive of QOL with current ESRD patients.”
My response (comment) to this statement was submitted to Nissenson's blog on June 26, 2016 @ 2:10 pm (does not appear that it was accepted as a post yet by Nissenson). The response is reproduced below.
Not sure what you are referring to by this vague comment, but this presentation of results from 32,926 KDQOL-36 surveys reveals a superficial understanding of psychometric methods: http://www.davitaclinicalresearch.com/wp-content/pdfs/ASN_2010/SA-PO2616-ASN2010_KDQOL_Valid_8Nov10DVW.pdf
The poster summarizes results of a factor analysis, item-scale correlations, and internal consistency reliability. The summary of results states that the data “do not confirm 2-factor solution” for the SF-12 in dialysis patients.
The methodology is described as a confirmatory factor analysis, but apparently this was really an exploratory principal components analysis. Despite the claim that a 2-factor solution was not confirmed, the scree plot for the SF-12 in Figure 1 can be interpreted as 2 principal component eigenvalues prior to the “scree.” The third principal component eigenvalue is 1.03–barely larger than eigenvalue > 1.0 criteria for possible number of factors (i.e., Guttman’s weakest lower bound rule of thumb).
Similarly, the second principal component eigenvalue for the burden scale is 1.09, again not much larger than Guttman’s weakest lower bound and the scree plot supports a single underlying dimension.
In addition, it is stated that a Varimax (uncorrelated) factor rotation was conducted despite the fact that the SF-12 physical and mental health scores are know to be associated significantly and substantially (correlations of 0.40-0.60).
The statement that MCS explains only 12% of the scale variance is inaccurate. The second principal component eigenvalue accounts for that proportion of variance but the second component is not the same as the MCS.
Cronbach’s alphas are reported for the SF-12 physical and mental health summary scores, but coefficient alpha is not appropriate for a weighted combination of items. The fact that negative item-total correlations are reported suggests a potential problem with how the data were scored.
There is a good distribution of scores and ceiling effects do not appear to be excessive, despite the note about “significant” floor and/or ceiling effects.
Alpha is mispelled as “alfa” a few times. The term, “Internal Validity,” in the title does not make sense.
This study does not provide information on validity (whether the KDQOL-36 measures what it is supposed to assess).
The KDQOL-SF scoring program is available for download - please visit our downloads page for information.
The KDQOL-36 is short form that includes the SF-12 as generic core plus the symptoms/problem, effects of kidney disease, and burden of kidney disease scales from the KDQOL-SF. The PDF version is now available on our downloads page.
(To access download page, register first to get username/password).