Part 1:
1. Choose a paper describing the development or validation of a measure of relevance in health disparities research (please give the full citation and/or upload the paper if that's possible).
van Bergen et al. Social Exclusion Index-for Health Surveys (SEI-HS): a prospective nationwide study to extend and validate a multidimensional social exclusion questionnaire. BMC Public Health (2017) 17:253
DOI 10.1186/s12889-017-4175-1
(file size was too large to upload, sorry!)
2. What was the definition of the construct?
Van Bergen et. al. define Social Exclusion (SE) as “the inability of certain groups or individuals to fully participate in society.”
3. How did the authors provide evidence on the validity of the measure? Could you think of additional approaches to validating the measure?
Items were derived from a previously developed social exclusion index (the 15-item SCP instrument developed by the Netherlands Institute for Social Research). This index had been validated in a very similar population to this study (Dutch adults).
Their validation analyses involved use of nonlinear canonical correlation analysis (OVERALS module in SPSS). This algorithm compares the variable sets (dimensions of SE) to an unknown comprise set that is defined by the object scores. The dataset was randomly split with SPSS into a development sample a validation sample; analyses were carried out in both samples. This canonical correlation measures the degree to which the items contribute to the underlying construct of SE.
To test construct validity, the authors tested a number of hypotheses using linear regression models comparing risk factors with SE score (risk factors that were known to be associated with SE according to previous research/measures). They arbitrarily defined adequate construct validity as >=75% hypothesis confirmation.
An additional approach the authors did not include would be to include a second, independent measure of SE which had been previously validated, which would be expected to be highly correlated with this measure.
4. How did the authors provide evidence on the reliability of the measure? Could you think of additional approaches to evaluating the reliability of the measure?
Reliability was demonstrated by repeating their analyses in the split sample, but a more convincing demonstration of reliability would be a temporal split – repeating their study at a later date in a similar population.
5. Describe the implications of a lack of measurement validity or reliability for future research applications.
Lack of measurement validity would imply that the SE instrument does not adequately capture the construct of interest (the inability of certain groups or individuals to fully participate in society). This could perhaps lead to Type 1 or 2 errors in future research depending on the directionality of the false association. Lack of reliability would increase the variability of any data collected and would likely decrease the statistical power of any future research using the instrument.
Part 2:
1. Find a paper describing a health disparity (please give the full citation or, even better, upload the paper so everyone else can look at it too)
Please see attached (Ramirez, 2019).
2. Summarize the construct and measurement of the dimension of disparity (e.g., racial inequalities?, SES inequalities?) and the outcome measured (e.g., self-rated health).
They were looking at two constructs: 1) Race, and 2) Insurance status as recorded in the Florida Cancer Data System (FCDS), grouped into 4 large categories of uninsured, insured, Medicaid, and Medicare. The outcome of interest was stage of previously undetected prostate cancer at time of diagnosis (their hypothesis stated that insurance status was a mediator between race and cancer stage at diagnosis).
3. What is the evidence for the validity and reliability of the measures?
The FCDS database recorded insurance status at the time of diagnosis with presumably good accuracy, though the study authors did not provide any evidence regarding the measurement’s validity and reliability. Race was recorded into the FCDS database following specific guidelines found in the “Race and Nationality Descriptions from the 2000 Census and Bureau of Vital Statistics” to identify both race and nationality (race was inferred from nationality if no race was stated). In this study, only “white” and “black” race patients were included.
4. What is the reference category used for the disparity measure (ie, who is the comparison group)? Why does this reference category make sense (or not) for this research question?
The reference category were white individuals (vs black). It would have been better to look at multiple race categories as there are a substantial number of confounders when looking at black vs. white, but in the context of their question, this reference category makes sense.
5. How is the disparity quantified or measured? Is this an absolute or relative measure or are both provided? Describe which type of measure you would prefer for this research area, or, if both, why.
The disparity of race is relative (black vs white), as are the insurance categories – the authors were interested in comparing categories against one another to find differences in cancer stage at diagnosis. This makes sense, given the differences in these categories and the lack of a binary “yes/no” situation for these constructs.