Measurement Homework

Measurement Homework

by Maria Glymour -

A two-parter:

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). 
2.What was the definition of the construct?
3.How did the authors provide evidence on the validity of the measure? Could you think of additional approaches to validating the 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?
5.Describe the implications of a lack of measurement reliability for future research applications. 

Part 2:

1.Find a paper describing a health disparity (please give the full citation)
2.Summarize the construct and measurement of the dimension of disparity (e.g., race, SES) and the outcome measured (e.g., self-rated health).
3.What is the evidence for the validity and reliability of the measures?
4.What is the reference category used for the disparity measure? Why does this reference category make sense (or not) for this research question?
5.How is the disparity quantified?  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. 

Re: Measurement Homework

by Maria Garcia -

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). 

2.What was the definition of the construct?

 

The CES-D scale is a 20-item self-report instrument designed to assess current levels of depressive symptoms (within the past week) in the general population (Radloff, 1977).

 

3.How did the authors provide evidence on the validity of the measure? Could you think of additional approaches to validating the measure?

From the abstract:

“Principal components and exploratory factor analysis were used. The participants responded to the CES-D scale and Spielberger’s State-Trait Anger Expression Inventory.”

Additional approaches that would have been useful, particularly given the subject matter, would have been to compare the CES-D scale with the gold standard, full diagnostic psychiatric interview, at least for a subset of participants.

 

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?

 

 From the abstract and manuscript:

“The final sample consisted of 208 Black single mothers aged 18–45 years. A 2-factor structure was accepted. Construct validity was confirmed via significant correlations with the anger scales. A method artifact for the 2-factor solution was ruled out.”

 

“Factor analysis was conducted using SPSS 20.1 computer program. In addition, Cronbach’s alpha coefficient was used to determine the reliability of the total scale and the accepted factors of the CES-D.”

 

“A forced two-factor solution, using PCA with varimax rotation, resulted in the most interpretable solution. Both factors yielded high coefficient alpha reliabilities.”

Again, comparing to gold standard would have been useful.

 

5.Describe the implications of a lack of measurement reliability for future research applications. 

If the scale significantly varies in what it is measuring in different populations, this would have significant implications for research on depression/depressive symptoms in diverse populations and describing disparities. For example, Latinos have greater incidence of depressive symptoms that the white population. If this is due to the fact that the scales are measuring other things besides depression, such as distress or cultural expressions of low moods that don’t qualify as depression, then there are important public health and clinical implications.

 

Part 2:

1.Find a paper describing a health disparity (please give the full citation)

2.Summarize the construct and measurement of the dimension of disparity (e.g., race, SES) and the outcome measured (e.g., self-rated health).

 

Construct and measurement of the dimension of disparity: race, as self-reported to N-HANES

Outcome: Age-specific cardiovascular disease mortality rates.

 

3.What is the evidence for the validity and reliability of the measures?

 

Self-identification, as used in N-HANES is the ‘gold standard’ for race measurement.

To obtain mortality rates for CVD, the authors used primary underlying cause of death per the WHO regulations by ICD-10 codes, which is an imperfect, but well-regarded way to measure mortality rates.

 

4.What is the reference category used for the disparity measure? Why does this reference category make sense (or not) for this research question?

The reference category was non-Hispanic white.  This seems to be an adequate control group given that the research question is specifically to examine differences in CVD prevalence and mortality between blacks and whites over a 7-year time period by defined age categories across and adult age range.

 

5.How is the disparity quantified?  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 authors reported prevalence ratios, which are relative measures and which seem an adequate way to present the differences between these two populations.

Re: Measurement Homework

by Maria Glymour -

Maria:

Nice examples.  

Distinguish between the construct they are trying to measure (current levels of depressive symptoms) and the instrument they used (CES_D).  

When they compared to the Speilberger, they were trying to demonstrate construct validity, i.e., this measure correlates with other things we expect depressive symptoms to correlate with - it's probably the most common approach to providing evidence of validity but the extent to which it is convincing depends on how well you've measured the other construct and how strong your understanding of "what should be correlated" is.  

They report very good coefficient alpha (Cronbach's or internal consistency reliability) or .94 for "factor 1" and pretty good reliability (.810) for factor 2 (positive affect).  The imperfect reliability could attenuate associations, especially with factor 2, and, in models trying to use the CESD to control for depressive symptoms (ie. if they are potential confounders), the control might not be adequate because of imperfect reliability.  Similarly, in a mediation model, imperfect reliability of the measure of the mediator will lead to an underestimate of the importance of the mediating pathway.

In the Jolly paper, as you note they are using relative measures (the prevalence ratio).  In my view, this would be a great setting to present both the absolute and relative disparities.  

Note for example that there were 1,759 /100,000 excess CVD cases in blacks compared to whites among 35-44 years olds (Table 1: 3718-1959), the age group with by far the largest relative disparity (PR=1.9, per figure 1).  In contrast, among people ages 65-74, there were 2,618/100,000 excess CVD cases in blacks compared to whites among people ages 65-74, a group with a much smaller relative disparity (PR=1.2 per figure 1).  Often (as in this case) the relative and absolute measures would imply really different age groups as having the largest disparity.  For public health impact, there is an emerging consensus that the absolute measures are most relevant (note that public health impact is not the only goal, and in this paper there is much more emphasis on questions of when you should try to start interventions/prevention).  

-Maria

 

Re: Measurement Homework

by Bliss Temple -

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). 

 

2.What was the definition of the construct?

 

The construct here is referred to as “social participation” by the authors and “participation” more generally in the literature.  It is poorly defined in this paper, but refers to a subject with a disability’s degree of involvement in economic, social, and community life; this is what all three instruments used by the authors are designed to measure.

 

3.How did the authors provide evidence on the validity of the measure? Could you think of additional approaches to validating the measure?

 

The authors assess the content and criterion validity of the measures by comparing to the domains of participation laid out in the ICF (WHO’s International Classification of Functioning and Disability), as well as assessing face validity by asking study subjects to assess the relevance of the instruments for measuring their participation.  It is interesting and laudable that they involved the subjects in validating the measures in this way.  It would have been powerful to look at concurrent validity by doing objective assessments of the subjects’ participation and comparing this to the instruments used (which all relied exclusively on self-report).

 

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?

 

The authors tested the instruments twice in each subject, with the administrations being two weeks apart, to test the reliability of the measures.  Another way to assess reliability would be to rate participation objectively, then see whether the instruments produced similar scores for subjects with similar objective measures of participation.

 

5.Describe the implications of a lack of measurement reliability for future research applications. 

 

If a measure is not reliable, it is hard to use it to do meaningful research.  It’s a case of “garbage in, garbage out” because a measure can’t be considered meaningful if it isn’t sufficiently reliable.

 

Part 2:

1.Find a paper describing a health disparity (please give the full citation)

 

2.Summarize the construct and measurement of the dimension of disparity (e.g., race, SES) and the outcome measured (e.g., self-rated health).

The two constructs examined here for disparities are disability and living in rural vs urban areas.  Disability was defined as any affirmative response to any Medical Panel Expenditure Survey (the data set used) question about limitations in basic actions involving physical functions, vision, hearing, or cognition.  Subjects were characterized as living in an urban area if they resided in a Metropolitan Statistical Area, and otherwise were considered to live in a rural area.

The outcomes measured were up-to-date breast cancer screening, defined as mammogram within past two years (for women aged 40-64); and up-to-date cervical cancer screening, defined as Pap smear within the past three years (for women aged 18-64).  Women were otherwise considered not up-to-date.

3.What is the evidence for the validity and reliability of the measures?

 

 

4.What is the reference category used for the disparity measure? Why does this reference category make sense (or not) for this research question?

 

The reference category for the disability measure is those who reported no limitation in basic actions.  This makes basic sense, since these are people who report none of the limitations associated with disability.

 

The reference category for the rural living measure was those who resided in a Metropolitan Statistical Area.  This is a reasonable and straightforward, though crude, way to construct the reference category.  However, it misses a lot of nuance since, for example, there may in fact not be big differences in health services in some exurbs vs medium-sized towns.

 

5.How is the disparity quantified?  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. 

 

Absolute measures are given for all the groups, but risk differences are not actually calculated to quantify the disparity.  The disparity is quantified in relative terms using odds ratios.  I think it is important to have a sense of the basic magnitude of the absolute measures so as to be able to put the disparity into context.  However, using an odds ratio is a good way to actually quantify the disparity because it helps give a better intuitive grasp of the magnitude of the disparity when comparing the two groups.

Re: Measurement Homework

by Dominika Seidman -

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). 

Am J Public Health. 2009 June; 99(6): 1023–1031.

Concurrent Partnerships and HIV Prevalence Disparities by Race: Linking Science and Public Health Practice

Martina Morris, PhD, Ann E. Kurth, PhD, Deven T. Hamilton, MPH, James Moody, PhD, and Steve Wakefield for the Network Modeling Group

2.What was the definition of the construct?

This paper used a data-driven network simulation model (“network epidemiology”) to see if levels of sexual concurrency (having concurrent sexual partners) and network segregation can explain disparities seen in STI and specifically HIV infection rates. This approach was novel, because rather than focusing on individual level behaviors accounting for disparities in infection rates, it describes sexual networks as a structural cause of disparities in infections.

3.How did the authors provide evidence on the validity of the measure? Could you think of additional approaches to validating the measure?

This paper was a modeling exercise to demonstrate that increasing levels of sexual concurrency and sexual network segregation can replicate disparities in the HIV epidemic. Because this was a simulation, no actual validation was performed. Next steps might be to conduct a study collecting prospective information on sexual concurrency/network segregation and predicting HIV prevalence based on the model and comparing it to incidence rates.

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?

Sexual concurrency and network segregation are based on self-report. Unfortunately getting confirmatory information for this type of measure is very challenging. One way to confirm sexual partners in people who are infected with HIV is to see if the strain of the infection matches that of their partner. However, in uninfected individuals, it’s very hard to reliably measure sexual concurrency levels.

5.Describe the implications of a lack of measurement reliability for future research applications. 

Perhaps different communities are more/less likely to report sexual concurrency. Consequently, the association between concurrency and infection might be confounded by measurement bias.

 

Part 2:

1.Find a paper describing a health disparity (please give the full citation)

Am J Public Health. 2007 January; 97(1): 125–132.

Sexual and Drug Behavior Patterns and HIV and STD Racial Disparities: The Need for New Directions

Denise Dion Hallfors, PhD, Bonita J. Iritani, MA, William C. Miller, MD, PhD, MPH, and Daniel J. Bauer, PhD

2.Summarize the construct and measurement of the dimension of disparity (e.g., race, SES) and the outcome measured (e.g., self-rated health).

Disparity dimension: race

Outcome: odds of STI/HIV infection stratified by risk behavior among young people in the US

3.What is the evidence for the validity and reliability of the measures?

Risk behaviors were self-reported on a confidential laptop. There was no validation of these self-reports. However, the discussion notes that there was previous data suggesting that the survey was validated in different ethnic groups. Infections were prospectively identified via serial infection screening.

4.What is the reference category used for the disparity measure? Why does this reference category make sense (or not) for this research question?

White young people were the reference. This makes sense because infection rates are relatively low in white US youth, demonstrating the potential for interventions/areas for improvement. However, this reference group is also diverse and hides other disparities, such as higher rates of infections among MSM (and highest rates among MSM of color) and transgender individuals.

5.How is the disparity quantified?  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. 

Differences in risk of HIV/STI acquisition among youth in different racial groups, stratified by risk, were compared with odds ratios. These statistics made for powerful public health messages: for example among black youth engaging in only low risk activities, the odds of HIV/STIs was almost 8 times that of white youth. This is a relative measurement. I think using a relative measurement is helpful in publicizing the disparity, but an absolute measure might be more useful if (for example) an intervention occurred to decrease infection risks and the goal was to show changes over time (and importantly, disparities in impact of an intervention).

Re: Measurement Homework

by Maria Glymour -

Dominika,

The network simulation is a very unusual way to approach measurement development!  It has some potential to be useful for certain research areas, though, as noted here. 

I completely agree with your response to item 5 - my suspicion is that people often report only the relative measures because they make for a more splashy result (which in some situations is probably very appropriate) - but the absolute measure should also be provided so we can think about the total public health impact, contrast groups, and consider trends.

Maria

Re: Measurement Homework

by Maria Glymour -

Bliss,

Thanks for the nice examples.  The "reliability" issue is really a spectrum: almost no measures are perfectly reliable, so it's important for any given measure to consider how much the imperfect reliability will bias your particular analysis and if the measurement error is systematic (e.g., is the error worse for women - or is there differential item functioning for some subgroups).  

We often have rules of thumb about "acceptable" reliability but in truth what is "acceptable' should depend on how the measure is being used (as a predictor, an outcome, to control confounding, to evaluate mediation, etc).

For the disparities paper: I agree it would have been really nice to see risk differences, especially after adjustment for various other factors.  Also it is generally problematic to use odds ratios with very common outcomes because as the outcome becomes common, the odds ratios diverge from the prevalence ratios and the ORs look more and more extreme.  If, for example, you take their most extreme contrast (disabled rural vs no disability MSA for a pap), the crude prevalence ratio is .77/.87=0.88, compared to the aOR of .69 (and maybe this is partly the adjustment but not entirely).  

Thanks for these.

Maria

Re: Measurement Homework

by Sarah Averbach -

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).  Scheerhagen et al. http://www.ncbi.nlm.nih.gov/pubmed/25671310 (too large to attached).

 

2.What was the definition of the construct?

 

The construct was the experience of dignified maternity care. Dignified maternity care has been defined by the WHO as a basic human right. Measuring maternity care has traditionally focused only on maternal and infant morbidity and mortality. How the construct was measured was the creation of a patient-centered outcomes questionnaire, the “Repro-Q”.

The Repro-Q questionnaire was developed to measure self-reported experience of maternity care including the experience of dignity, confidentiality and support).

 

3.How did the authors provide evidence on the validity of the measure? Could you think of additional approaches to validating the measure?

 

They assessed the measure using 40 structured interviews and then validated it with ~900 dutch women using correlation with a VAS (construct validity).

Another approach could have been to specifically recruit women representing racial and ethnic minorities in order to validate the instrument among a more diverse group than the primarily well educated, white and dutch population that the instrument was validated in.

 

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?

 

Multiple questions dealt with each primary domain and they assessed for consistency within the domains (inter-item reliability). There was no specific measure of inter-rater reliability or test-re-test reliability given.

 

5.Describe the implications of a lack of measurement reliability for future research applications. 

 

First a confirmatory analysis is needed. Second, per the manuscript women with “low education” (<= 6 yrs), “migrant women”  (<=8 yrs  living in Holland), and women of “non-western origin” (women of color?) were poorly represented in this cohort.  If the instrument varies in different populations then it limits the populations in which it can be used.

 

Part 2:

1.Find a paper describing a health disparity (please give the full citation) attached Finer et al.

 

2.Summarize the construct and measurement of the dimension of disparity (e.g., race, SES) and the outcome measured (e.g., self-rated health).

 

Construct: Race

Measurement of the disparity: Data collected by the national center for health statistics and national survey of family growth and CDC surveillance.  Race and ethnicity data came from the US Census bureau.

Outcome measured: proportion of pregnancies that ended in abortion out of the total number of births (compared to birth and miscarriage).

 

3.What is the evidence for the validity and reliability of the measures?

 

There is evidence to support the Census questions on race and ethnicity. They are, however, controversial and may change in the coming years. The census questions are based on self-identification which is the gold standard. The categories may be too simple and may not reflect current complexities. For instance, Caribbean Black and African immigrants are not the same. Given the lumping there has been shown to be misclassification in administrative data of race and ethnicity. This is not discussed in the paper attached. 

 

 

 

4.What is the reference category used for the disparity measure? Why does this reference category make sense (or not) for this research question?

 

In the primary analysis they presented a simple comparison of proportions so they did not use a reference category. In a secondary analysis Non-Hispanic white women were used as the reference category for this research question. I think it made sense to choose one group to use as the comparator throughout the analyses because it would have been confusing to switch from one comparator to another.

I think here the use of whites as the comparator is reasonable given the because they are the “default” or the most advantaged group and deviations from this highlight the disparities in unintended pregnancy rates and births following unintended pregnancy between whites and other groups. Given the large n available in this study the use of whites as the comparator because they are the largest group and are most precisely estimated is not a good argument.

 

5.How is the disparity quantified?  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. 

 

Both absolute and relative measures are provided. This study primarily used absolute measures “The proportion of unintended pregnancies ending in abortions was highest among black women” “Hispanic women had the highest unintended birth rate.” The study also made relative comparisons. For instance, Blacks had the highest rate of unintended pregnancy, “more than double that of non-Hispanic whites”.

Re: Measurement Homework

by Maria Glymour -

Great examples Sarah!

The measurement example is very carefully done and it's impressive to see them take it through all of the steps.  You can see that for validation they are also struggling a bit with how to provide quantitative evidence. They attempt with convergent and discriminant validity, but these really depend on a theoretical understanding of how the construct relates to other things they measured.  Nice paper.

The unintended pregnancy paper provides a nice illustration of using both relative and absolute measures.  

Re: Measurement Homework

by Valy Fontil -

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).  Fernandez et al. http://www.ncbi.nlm.nih.gov/pmc/articles/PMC3763496/

2.What was the definition of the construct?

 The construct was medication adherence  in African Americans with hypertension.  African Americans have lower rates of adherence to antihypertensive medications which may contribute poorer hypertension control in this population.  Self-efficacy is component of the social cognitive model that refers to an individual’s judgment of her confidence to carry out a task. The stronger  one’s self-efficacy beliefs, the more likely she will initiate and maintain recommended health behaviors. The authors developed and evaluated the reliability Medication Self-Efficacy Scale (MASES) in hypertensive African Americans. MASES is a patient-derived, self-report measure designed to assess efficacy beliefs regarding adherence to prescribed anti-hypertensive medications.

 

3. How did the authors provide evidence on the validity of the measure? Could you think of additional approaches to validating the measure?

 They  assessed unidimensionality using confirmatory and exploratory models in a study sample of 168 African Americans followed in primary care practices. They also assessed internal consistency and reliability with item total correlations, item means, and coefficient alpha in a classic test theory analysis. Lastly, they examined the predictive validity of MASES by assessing its correlation with both self-report medication adherence electronic adherence.

I think the authors did a good job establishing the internal validity of MASES. The next step to assess the external validity by sampling other patient populations across different clinical settings and other socio-economic groups associated with lower medication adherence.

 

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?

 

They reported good alpha coeffiticients, item total correlations and test-retest coefficients  in addition to unidimensional factor structure.

 

5.Describe the implications of a lack of measurement reliability for future research applications. 

 

If an instrument is unreliable,  it is difficult to use it to predict outcomes or inform any intervention.

 

Part 2:

1.Find a paper describing a health disparity (please give the full citation) attached  Thorpe et al http://www.ncbi.nlm.nih.gov/pubmed/25065066

 

2.Summarize the construct and measurement of the dimension of disparity (e.g., race, SES) and the outcome measured (e.g., self-rated health).

 

Construct: self-reported race;

The authors examined African American and Whited adults living in the same racially integrated community sharing a similar healthcare market.

Measurement of the disparity:  Odds ratio.  relative point estimate of the outcomes of interest by logistic regression with white race as the reference group.

Outcome measured:  hypertension awareness, treatment, and control

 

3.What is the evidence for the validity and reliability of the measures?

Awareness of hypertension measured  by self  report of having been  told/diagnosed by a doctor is widely used by national survey and is correlated with hypertension treatment and control;

Hypertension control is  validated health measure associated with cardiovascular  outcomes.  Hypertension treatment is defined by self report of taking antihypertensive medications. I am not aware of any formal validation studies for this measure.

 

Re: Measurement Homework

by Maria Glymour -

Valy, I glanced back at the original paper they published on this scale and they noted they generated the items with open ended interviews, using a fairly formal qualitative analysis of barriers and facilitators - starting with a 43 item pool and dropping many based on reliability and "clinical relevance", so some work on face/content validity prior to the current paper.   

The consequences of low reliability of the instrument depend on how it is being used: as a predictor, an outcome, a control variable, a mediator, etc. One thing that was interesting here is the large discrepancy between Cronbach's alpha 90.91) and the test retest (0.56).  It's possible that self-efficacy really changes that much over the 3 months, so the test-retest is accurately reflecting changes in the construct, but it does make you worry a little.  It would be nice to see a test-retest with a shorter time interval. If the reliability is only 0.56, it's going to be unusable for many purposes.

Nice examples. 

Maria

Re: Measurement Homework

by Brian -

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). 

 

 

Zimet G. The multidimensional scale of perceived social support. Journal of personality assessment, 1988, 52(2), 30-41.

 

2.What was the definition of the construct?

 

This measure seeks to characterized perceived social support. Social support can be thought of one’s ability to access resources through one’s social network. Social support has been characterized in multiple ways including educational support, tangible/resource support, and emotional support.

 

The multi-dimensional scale of perceived social support construct was developed and measured by creating a 24 item questionaire. The scale was  intended to focus on subjective social support adequacy, perception of social support received by family, friends, and significant others.

 

3.How did the authors provide evidence on the validity of the measure? Could you think of additional approaches to validating the measure?

 

The investigators assessed measure using 275 college students who were administered the questionnaire, as well as the Hopkins Symptom Checklist (HSCL). They then used factor analysis for validation, that led to paring of the questions from 24 to 12 question. They then used the “Kaiser normalization test” as a confirmatory factor analysis.

 

They compared the results of the scale to HSCL for construct validity

 

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 assessed using cronbach’s alpha-for the whole scale they reported a cronbach alpha of .88

 

They also re-tested subjects  2-3 months after the initial administration and found reliability estimate of .85.

 

 

5.Describe the implications of a lack of measurement reliability for future research applications. 

 

The biggest question that this and other perceived/subjective scales of social support I have is what it actually measures. The validity of this instrument was tested against the Hopkins Symtom Check list, and against things like depression, anxiety subscales—so my concern is whether this scale is measuring things like anxiety/depression, rather than social support.

 

 

Part 2:

1.Find a paper describing a health disparity (please give the full citation) attached

 

Kind A, Jencks, et al. Neighborhood Socioeconomic Disadvantage and 30-day Rehospitalization. A retrospective cohort study. Ann Intern Med. 2014;161:765-774.

 

 

2.Summarize the construct and measurement of the dimension of disparity (e.g., race, SES) and the outcome measured (e.g., self-rated health).

 

Construct: SES

Measurement: The investigators assessed SES using the “Area Deprivation Index (ADI)”, created by Singh- a composit measure of socioeconomic disadvantage that uses 17 US Census poverty, housing, and employment indicators to characterize an area level/census based regions. In this case they linked the ADI to zip-code status for participants.

 

Outcome measured: all-cause rehospitalization within 30 days of discharge from Medicare claims data.

 

3.What is the evidence for the validity and reliability of the measures?

 

Citation for original paper is below- in this paper, they checked validity of their measure by correlations with many county-level health outcomes including birthweight, infant mortality, mortality from all causes.

 

 

25. Singh GK. Area deprivation and widening inequalities in U.S. mortality, 1969-1998. Am J Public Health. 2003;93:1137-43. [PMID: 12835199]


 

 

4.What is the reference category used for the disparity measure? Why does this reference category make sense (or not) for this research question?

 

 

 

The investigators split neighborhood disadvantage into 4 groups. Based on “empirical data,” the most disadvantaged neighborhoods made up the top 15% of the ADI distribution. They then divided groups in to the least 85% disadvantaged, then 3 groups of 5% of increasing disadvantage.

 

Thought the comparator of using least disadvantaged 85% group makes sense in this paper because it the investigators sought to highlight quality of care disparities compared to the idealized/”best case” group.

 

5.How is the disparity quantified?  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 investigators provided relative measures (OR) of disparity outcome, and also presented absolute rates of 30-day rehospitalization for each disadvantaged group. ORs may have been presented because a clinical audience understands them better (odds of readmission increased for areas of high economic disadvantage). On closer read, the differences in 30-day rehospitalization in the least disadvantaged group (21%) and the most disadvantaged group (24%), is an absolute difference of 3%-- whether this is clinically or policy relevant difference, is unclear. 

Re: Measurement Homework

by Niharika Dixit -

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). 

 

García-Jimenez M, Santoyo-Olsson J, Ortiz C, Lahiff M, Sokal-Gutierrez K, Nápoles AM. Acculturation, Inner Peace, Cancer Self-efficacy, and Self-rated Health among Latina Breast Cancer Survivors. Journal of health care for the poor and underserved. 2014;25(4):1586-1602. doi:10.1353/hpu.2014.0158.

 

 

2.What was the definition of the construct?

Psychoscoial model of adaptation to cancer.

 

 “Constructs of self-efficacy and spiritual well beings ( religious and non religious) are rooted in cultural beliefs and are associated with acculturation. “

 

Authors hypothesized that greater acculturation is associated positively with self reported health and that this relationship is partial mediated by cancer self efficacy and spiritual well being.

 

Predictor Variable

Author used SASH , Short Acculturation Scale for Hispanics. Crohnbach’s alpha for acculturation scale was .95.

 

Mediators

Cancer self-efficacy was measured with Cancer Behavior Inventory – Breast cancer

Cronbach’s alpha was .80

 

 

 

Functional Assessment of Cancer Therapy Quality of Life Measurement System-Spiritual Well-being Scale  was used for spiritual well being.

Cronbach's alpha for the scale was 0.82 in our sample

 

The outcome variable, SRH, was measured by the item “In general, how would you rate your health?

A mediation analysis was used to confirm mediators and set of predictors.

 

 

3.How did the authors provide evidence on the validity of the measure? Could you think of additional approaches to validating the measure?

Using multivariate logistic regression, the four mediation steps were modeled separately for each mediator of interest that was related to SRH at p <.15 in bivariate analyses.

4 requirements of mediation analysis were fulfilled as well.

 

 

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?

 

Crohnnach’s alpha was used as the reliability of the measure

 

5.Describe the implications of a lack of measurement reliability for future

research applications. 

 

While the modeling takes in to account several variable there are other issues such as ether greater self reported health is associated with greater self efficacy and well being.

It is also challenging to consider intervention. Language proficiency was an important predictor. However we do not know if language concordance may have been associated with higher self efficacy and self reported health.

 

 

 

Part 2:

1.Find a paper describing a health disparity (please give the full citation)

 

  1. Racial disparity in in-hospital mortality after lobectomy for lung cancer

Harrison, Meredith A. et al.

The American Journal of Surgery , Volume 209 , Issue 4 , 652 - 658

 

 

2.Summarize the construct and measurement of the dimension of disparity (e.g., race, SES) and the outcome measured (e.g., self-rated health).

 

Race was the important predictor measure, SES was also used with insurance income level, education, type of hospital and annual lobectomy volume.

 

 

3.What is the evidence for the validity and reliability of the measures?

This was a database study,

 

4.What is the reference category used for the disparity measure? Why does this reference category make sense (or not) for this research question?

Reference category was non-black. This makes sense as the African Americans are more like to have worse outcomes than other groups.

 

5.How is the disparity quantified?  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. 

Disparities is identified by odd ratio which is a relative measures. I think both relative and absolute measures are important here. Absolute measure are significant in defining the extent of the problem and its impact on cancer outcomes.