This item discusses measurement issues related to demographic disparities in immunization. But the points made could apply to a wide range of healthcare disparities issues, particularly those that are studied using a variety of methods and those that involve rate ranges where the statistical forces described in the Scanlan’s Rule page could tend toward causing absolute differences and differences measured by odds ratio to either increase or decrease as overall prevalence of the outcome changes and where the changes in prevalence may cause the determination of which relative difference is larger to change over time.[i] Among particularly relevant related items on this site are the Pay-for-Performance sub-page of the Measuring Health Disparities page (MHD) and the Mortality and Survival page. The former discusses the way that reliance on absolute differences between rates as a measure of healthcare disparities has led observers in the United States to believe that pay-for-performance will tend to increase healthcare disparities and observers in the United Kingdom to believe that pay-for-performance will tend to reduce healthcare disparities. The latter discusses a variety of issues (both as to survival/mortality and screening/non-screening) concerning confusion regarding disparities in adverse and favorable outcomes including the tendency to discuss relative differences in one outcome (both as to survival issues and screening) while in fact analyzing relative differences in the opposite outcome and to do so without recognizing that the two relative differences tend to change systematically in opposite directions as the prevalence of an outcome changes. Also relevant are Sections E.4 and E.7 of MHD and A.6 of the Scanlan’s Rule page, as well the APHA 2007 presentation,[ii] which discuss the measurement approaches of National Center for Health Statistics (NCHS) and the Agency for Healthcare and Research and Quality (AHRQ).
Immunization disparities are studied in terms of relative differences in immunization rates, relative differences in failure to be immunized, absolute differences, and odds ratios. Universally, however, researchers employing these methods have done so without recognizing the way that the measure tends to be affected by the overall prevalence of an outcome and almost invariably without regard to whether other measures would yield contrary results as to direction of change over time.
The issue is nicely illustrated with data from a 2008 Pediatrics study by Morita et al[iii] that examined the effects on racial and ethnic disparities of a school-entry Hepatitis B vaccination requirement. Relying on relative differences in vaccination rates to measure disparities, the authors found that the requirement, which dramatically increased overall vaccination rates, also dramatically reduced racial and ethnic vaccination disparities. But the National Center for Health Statistics (NCHS), which measures all disparities in terms of relative differences in adverse outcomes (including, in the case of procedures like vaccination, failure to receive the procedure), would find the disparity to have increased dramatically.
The patterns are illustrated in the Tables A and B below. They are taken from Comment on Morita,[iv] which discusses the patterns in greater detail, including the changes in absolute differences and the way different researchers would reach different conclusions based on the choice of measure. It also appraises the changes in the disparities according to the procedure describe on the Solutions sub-page of MHD.
Table A: White and Black Vaccination Rates and Rates of not being Vaccinated Before and After Implementation of School-Entry Vaccination Requirement, with Relative and Absolute Differences, Odds Ratios, and Estimated Effect Size (based on data from Morita et al.)
Table B: White and Hispanic Vaccination Rates and Rates of not being Vaccinated Before and After Implementation of School-Entry Vaccination Requirement, with Relative and Absolute Differences, Odds Ratios, and Estimated Effect Size (based on data from Morita et al.)
A few other studies that recently caught my attention warrant mention as well. I note, however, that the points made about them could be made about countless other studies of disparities in immunization and other healthcare procedures. In fact, the points generally could be made about almost all health and healthcare disparities studies.
A recent study by Zhao and Luman[v]) in the American Journal of Preventive Medicine relied on absolute differences between rates as a measure of immunization disparities in finding general decreases in disparities during a period of overall increases in immunization rates. The study is of particular note because the authors are from the Centers for Disease Control and Prevention (CDC), of which NCHS is a part. Relying on relative differences in failure to be vaccinated, NCHS would have found increasing disparities in 8 of the 12 unadjusted analyses and 7 of the 12 adjusted analyses.
A table that may be found here (which is based on data in Table 1 from Zhao and Luman) illustrates these patterns as well as the other patterns of changes in standard measures. It also shows, using the method described in the Solutions sub-page of MHD, that usually the disparities decreased in a meaningful sense. But in a few cases, though the absolute differences decreased slightly, the disparity actually increased slightly.
Apparently, the ignoring of the NCHS recommendation to measure all disparities in terms of relative differences in adverse outcomes is common at the CDC. On January 14, 2010, CDC released the CDC Health Disparities and Inequalities Report – United States, 2011. The report generally relies on absolute differences between rates for the measurement of healthcare disparities, including immunization disparities (and screening disparities). There is some overlap of the report with AHRQ’s National Healthcare Disparities Report. Given that AHRQ relies on the larger of two relative differences to measure disparities, and that as the prevalence of an outcome changes the absolute difference tends to change in the same direction as the smaller of the two absolute differences, there will be a tendency for CDC to reach conclusions about the directions of changes in the size of disparities over time that are the opposite of those AHRQ would reach.[vi]
A 2010 study by Yoo et al.[vii] in American Journal of Preventive Medicine also relied on absolute differences between rates as a measure of disparity without consideration of the implications of overall prevalence. In the main, the authors found that during periods of vaccine shortage, when overall vaccination rates decreased, absolute differences between rates increased; during period periods of adequate supply when overall vaccination rates increased, absolute differences decreased. But, for the most part, the vaccination rates for the affected groups were in ranges where, as discussed in the introduction to the Scanlan’s Rule page, overall increase will tend to decreases absolute differences and overall decreases will tend to increase absolute differences.
The Yoo data show some exceptions to this pattern. And no doubt in many cases the directions of changes in absolute differences would comport with the directions of changes in disparities that one would find by the method described on the Solutions sub-page of MHD. But one cannot usefully measure changes in disparities by simply relying on changes in absolute differences – or changes in any other standard measures of difference between rates – without consideration of the way the measure tends to be affected by the overall prevalence of an outcome.
An article by Niederhauser and Stark,[viii] which broadly discussed immunization disparities, presented in its Table 2 information (drawn from Klevens and Luman[ix] on various vaccination rates for children 19-35 months (a) above the poverty line, (b) in near poverty, (c) in intermediate poverty, and (d) in severe poverty. Table C below presents the vaccination rates for those above the poverty line (AP) and for those in (SP) severe poverty, while presenting the same measures shown Tables A and B above.
Table C: Vaccination Rates and Rates of not being Vaccinated by Type of Vaccination for Children 19-35 Months above the Poverty Line (AP) and in Severe Poverty (SP), with Relative and Absolute Differences, Odds Ratios, and Estimated Effect Size (based on data from Niederhauser and Stark et al.)
Hepatitis B-3 doses
A recent book published by the World Bank, Attacking Inequality in the Health Sector: A Synthesis of Evidence and Tools,[x] warrants mention for its analysis of immunization disparities in terms of relative differences in immunization rates. Like Morita et al., it employs what many would regard as the standard approach to measurement of immunization disparities, and is reflective of the unlikelihood that the NCHS approach will have great influence (as discussed with regard to the United Kingdom in Section  of the Mortality and Survival page). Of, course as discussed above, even CDC seems not to be following the NCHS recommendation.
A paper presented at the 2010 European Population Conference by Gupta and Sekher[xi] is interesting because the authors analyzed both relative differences in favorable outcomes and relative differences in adverse outcomes. The authors analyzed relative differences in favorable outcomes with respect to the receipt of full immunization and relative differences in adverse outcomes with respect to failure to receive any immunization. The patterns in data they provide in an extended abstract are not particularly consistent with the patterns I discuss here and elsewhere on this site. But there nevertheless is reason to expect that generally overall increases in immunization will tend to reduce relative differences in full immunization but increase relative differences in failure to receive any immunization (while increasing relative differences in the failure to receive full immunization and reducing relative differences in receiving some immunization).
[i] By way of broad summary, as an outcome increases in overall prevalence, relative differences in experiencing it tend to decrease while relative differences in failing to experience it tend to tend to increase. As rates that are less than 50% for both groups increase, absolute differences between rates tend to increase; as rates that are greater than 50% for both groups increase, absolute differences tend to decrease. Differences measured by odds ratios tend to change in the opposite direction of the absolute difference.
[iii] Morita JY, Ramirez E, Trick WE. Effect of school-entry vaccination requirements on racial and ethnic disparities in Hepatitis B immunization coverage among public high school students. Pediatrics 2008;121:e547-e552
[iv] Scanlan JP. Study illustrates ways in which the direction of a change in disparity turns on the measure chosen. Pediatrics Mar 27, 2008 (responding to Morita JY, Ramirez E, Trick WE. Effect of school-entry vaccination requirements on racial and ethnic disparities in Hepatitis B immunization coverage among public high school students. Pediatrics 2008;121:e547-e552):http://pediatrics.aappublications.org/cgi/eletters/121/3/e547
[v] Zhao Z, Luman ET. Progress toward eliminating disparities in vaccination coverage among U.S. Children 2000-2008. Am J Prev Med 2010;xx(x);xxx
[vi] Extended discussion of another instance of reliance on absolute differences as a measure of immunization disparities may be found in Scanlan JP. Understanding patterns of absolute differences in vaccination rates in different settings. Journal Review Apr. 22, 2008 (responding to Schneider EC, Cleary PD, Zaslavsky AM, Epstein AM. Racial disparity in influenza vaccination: Does managed care narrow the gap between blacks and whites? JAMA 2001;286:1455-1460): http://journalreview.org/v2/articles/view/11572737.html
[vii] Yoo B-K, Kasajima M, Phelps CE, et al. Influenza vaccine supply and racial/ethnic disparities in vaccination rates. Am J Prev Med 2011;40(1):1-10.