Currently Viewing:
The American Journal of Managed Care September 2015
Do Patient or Provider Characteristics Impact Management of Diabetes?
Erin S. LeBlanc, MD, MPH; A. Gabriela Rosales, MS; Sumesh Kachroo, PhD; Jayanti Mukherjee, PhD; Kristine L. Funk, MS; and Gregory A. Nichols, PhD
The Utility of Cost Discussions Between Patients With Cancer and Oncologists
S. Yousuf Zafar, MD, MHS; Fumiko Chino, MD; Peter A. Ubel, MD; Christel Rushing, MS; Gregory Samsa, PhD; Ivy Altomare, MD; Jonathan Nicolla, MBA; Deborah Schrag, MD; James A. Tulsky, MD; Amy P. Abernethy, MD, PhD; and Jeffery M. Peppercorn, MD, MPH
Building Upon the Strong Foundation of National Healthcare Quality
Charles N. Kahn III, MPH, President and CEO, Federation of American Hospitals
Improving Partnerships Between Health Plans and Medical Groups
Howard Beckman, MD, FACP, FAACH; Patricia Healey, MPH; and Dana Gelb Safran, ScD
Innovative Approach to Patient-Centered Care Coordination in Primary Care Practices
Robin Clarke, MD, MSHS; Nazleen Bharmal, MD, PhD; Paul Di Capua, MD, MBA; Chi-Hong Tseng, PhD; Carol M. Mangione, MD, MSPH; Brian Mittman, PhD; and Samuel A. Skootsky, MD
Private Sector Risk-Sharing Agreements in the United States: Trends, Barriers, and Prospects
Louis P. Garrison, Jr, PhD; Josh J. Carlson, PhD; Preeti S. Bajaj, PhD; Adrian Towse, MA, MPhil; Peter J. Neumann, ScD; Sean D. Sullivan, PhD; Kimberly Westrich, MA; and Robert W. Dubois, MD, PhD
Developing Evidence That Is Fit for Purpose: A Framework for Payer and Research Dialogue
Rajeev K. Sabharwal, MPH; Jennifer S. Graff, PharmD; Erin Holve, PhD, MPH, MPP; and Robert W. Dubois, MD, PhD
Predicting Adherence Trajectory Using Initial Patterns of Medication Filling
Jessica M. Franklin, PhD; Alexis A. Krumme, MS; William H. Shrank, MD, MSHS; Olga S. Matlin, PhD; Troyen A. Brennan, MD, JD, MPH; and Niteesh K. Choudhry, MD, PhD
Reply to "Transforming Oncology Care": Advancing Value, Accessing Innovation
Rebecca Paradis, MPA
Payer Source Influence on Effectiveness of Lifestyle Medicine Programs
Joseph Vogelgesang, BS; David Drozek, DO; Masato Nakazawa, PhD; Jay H. Shubrook, DO
High-Risk Centers and the Benefits for Lower-Risk Transplants
Schelomo Marmor, PhD, MPH; James W. Begun, PhD; Jean Abraham, PhD; and Beth A. Virnig, PhD, MPH
Targeting a High-Risk Group for Fall Prevention: Strategies for Health Plans
Lee A. Jennings, MD, MSHS; David B. Reuben, MD; Sung-Bou Kim, MPhil; Emmett Keeler, PhD; Carol P. Roth, RN, MPH; David S. Zingmond, MD, PhD; Neil S. Wenger, MD, MPH; and David A. Ganz, MD, PhD
Currently Reading
Socioeconomic Disparities Across Ethnicities: An Application to Cervical Cancer Screening
Brendan Walsh, PhD; and Ciaran O'Neill, PhD

Socioeconomic Disparities Across Ethnicities: An Application to Cervical Cancer Screening

Brendan Walsh, PhD; and Ciaran O'Neill, PhD
An examination of socioeconomic disparities in cervical cancer screening across ethnic groups in the United States using concentration indices and probit regression analyses.
Objectives: Our aim is to investigate socioeconomic disparities in cervical cancer screening utilization among and between ethnic groups in the United States.

Study Design: Observational study.

Methods: Data on 26,338 women aged 21 to 64 years were obtained from the 2007 to 2011 years of the Medical Expenditure Panel Survey. Data on cervical cancer screening utilization in the preceding 12 months and 3 years, and a range of sociodemographic characteristics were included. Analyses were undertaken for all women and across racial/ethnic grouping (ie, white, black, Hispanic, and other). Concentration indices were used to measure the socioeconomic gradient across ethnic groups. Probit regression analyses were used to examine variations in utilization related to socioeconomic factors across ethnic groups controlling for a range of pertinent characteristics. 

Results: Annual utilization rates are high in the United States (60.15%) and greatest among black women (66.25%). Disparities, as measured by concentration indices (CIs), are large in the United States, with the largest being for white women (CI, 0.179) relative to black (CI, 0.103) and Hispanic (CI, 0.081) women. Screening differences across income, education and insurance status are also greater amongst white women. 
Conclusions: Uptake of cervical cancer screening is common in the United States, with large socioeconomic disparities also evident. Those from lower socioeconomic or uninsured groups who are most likely to have, and to die from, cervical cancer, are least likely to use preventive screening. Disparities differ across ethnic groups and are greatest amongst white women. Incorporating organized screening may serve to improve both the systems efficiency and address disparities between and within groups.
Am J Manag Care. 2015;21(9):e527-e536
Take-Away Points
Annual utilization of cervical cancer screening is high in the United States, but significant socioeconomic disparities are especially evident among white women. 
  • Factors underlying utilization differ between whites and other ethnicities. 
  • Those least well-served by current arrangements appear to be poor white women. 
  • Expanding access may be neither efficient nor serve to address the relative disadvantage experienced by poorer whites. 
  • An organized national screening program may better serve to address both efficiency and equity issues.
In the United States, age-standardized mortality rates for cervical cancer fell from 5.6 deaths per 100,000 women in 1975 to 2.3 per 100,000 in 2010.1 Preventive screening was instrumental in this reduction in mortality. The differences in mortality rates across socioeconomic groups2,3 and racial/ethnic groups4,5 have also been linked to differences in screening utilization across these groups. Age-standardized cervical cancer mortality rates among non-Hispanic black (termed “black”) and Hispanic women were 4.2 per 100,000 individuals and 2.9 per 100,000 individuals, respectively, compared with 2.2 per 100,000 individuals among non-Hispanic white (termed “white”) women.1 Although disparities across ethnic groups in cervical cancer treatment may explain some of the differences in mortality,6 lower screening rates among nonwhite women in the past may also explain these mortality differences.

Examining screening disparities between groups can aid in understanding the nature of barriers to screening and in developing appropriate policy responses. Although evidence shows that cervical screening utilization differs between both ethnic and socioeconomic groups in the United States,7-10 there is a dearth of research examining how socioeconomic disparities may differ across ethnicities, with limited research on socioeconomic status on its own. There is evidence that higher income and private health insurance are predictors of screening for white and Hispanic women, but these play a small (insignificant) role for black women.7 Income and education disparities have also been shown to be greater among white women relative to other groups.11

More detailed analyses of these relationships have not been pursued in these studies, in part, perhaps, because of limitations in the data used with respect to a fuller range of sociodemographic variables. Similarly, no attempt has been made to quantify or compare socioeconomic disparities between groups to inform discussions on relationships. If differences in socioeconomic disparities exist between ethnicities, this would add to our understanding and prompt more effectively tailored policy instruments. In this paper, we augment an analysis using regression techniques with an examination of the socioeconomic gradient in cervical cancer screening. We separately examined concentration indices (CIs) to shed further light on the issue of disparities; whereas CIs measure disparities across income, complementary probit regressions measure disparities controlling for a range of other pertinent variables.

Data Sources and Participants

Data from the 2007 to 2011 years of the Medical Expenditure Panel Survey (MEPS) were used in the analysis. MEPS is a nationally representative survey of respondents’ health, healthcare usage, and range of sociodemographic characteristics. The years 2007 to 2011 were chosen in order to increase the number of observations available. Years prior to 2007 were not included as they do not incorporate the same explanatory variable relating to total household income, and 2011 corresponds to the most recent year available. In the MEPS survey design, individuals are included in 2 consecutive waves of the survey. To prevent double counting in our pooled sample, only individuals in their second wave were included.

Main Dependent Variables

Within MEPS, women are asked if they screened for cervical cancer: 1) in the past year, 2) within the past 2 years, 3) within the past 3 years, 4) within the past 5 years, 5) more than 5 years ago, or 6) never screened. Using this variable, screening in the preceding 12 months and 3 years were analyzed among women aged 21 to 64 years, producing 27,238 observations. There were 12,282 women categorized as white; 5451 as black; 7248 as Hispanic; and 2257 as other ethnicities (including Asian and Native American, native Hawaiian/Pacific Islander, and multiple races). This age group is recommended for screening by the American Cancer Society (ACS) and the US Preventive Services Task Force (USPSTF).12,13 Survey weights based on the probability of being sampled in the survey were used to facilitate the robust extrapolation of results to the overall population. We examined utilization at 12 months and 3 years to ascertain the stability of observed relationships at different intervals. Furthermore, although the USPSTF and ACS recommend screening using Pap tests every 3 years for women aged 21 to 64 years, annual screening is common in the United States.14 Under the Affordable Care Act (ACA), 26 states mandate annual screening for privately insured women.15

Concentration Indices

Socioeconomic disparities were calculated using CIs, which allowed us to measure the extent to which screening utilization in different income groups reflects their proportionate representation in the population. CIs are one of the foremost methods for calculating and comparing socioeconomic disparities,16 including for cervical cancer screening17,18 and other forms of screening.19 As CIs present a summary statistic of disparity, they should, however, be handled with some care as the population level at which the statistic is calculated may serve to conceal material differences in disparities at the subgroup level. Nevertheless, the CI in its own right allows for disparities to be easily conveyed to policy makers and also allows for disparities to be compared across population groups.

A continuous measure of socioeconomic status enables the most precise calculation of CIs.20 The total household income variable (FAMINC) in MEPS provides the measure of socioeconomic status in our analysis. Using this variable offers greater comparability among women in paid employment and those who are homemakers, for instance. Income is further equivalized to allow for more meaningful comparisons between women with different sized households where disposable income might vary. (In our sample, the family sizes for white, black, Hispanic, and other is 2.76, 2.98, 3.91, and 3.33 members per household, respectively.) The OECD equivalence scale (square root of the number of people in the household) was chosen, and it is this equivalized income variable that is thus used to rank women from the poorest to the richest in the sample.

The CI calculates a disparity estimate between –1 (screening disproportionately higher among poor) and +1 (screening disproportionately higher among rich), with 0 representing an equal distribution of screening across the ranking variable. The CI can be expressed as twice the covariance between screening (y) and income (R) divided by the mean of screening (µ):
CI = 2cov(yR) /  µ

For binary variables (whether a woman was screened in the preceding 12 months or 3 years) the CI above is no longer bounded between –1 and +1. Therefore, it is further divided by 1 minus µ to allow for the inequality to be measured between –1 and +1. This is known as the Wagstaff correction.21 The CI used in our analysis is thus expressed as:
CI = (2cov(yR)/µ) / (1- µ)

Our analyses were undertaken in Stata version 13.0 (StataCorp LP, College Station, Texas) with the CIs computed using applicable Stata code from O’Donnell et al.22

Regression Analyses

Probit regression analyses were also undertaken, allowing for adjusted socioeconomic disparities to be calculated, controlling for other potential confounders, and allowing for the measurement of differences across education, marital status, and health insurance status. Results are presented as marginal effects with standard errors clustered at the region level. Although odds ratios allow for greater ease of interpretation, they cannot accurately be compared across groups within a model23,24; therefore, marginal effects are calculated in this study. The explanatory variables included were educational attainment (college degree or higher, high school degree, and less than high school); marital status (married/cohabiting or not), age (5-year age groups), geographic regions (northeast, midwest, south, and west), and ethnicity (white, black, Hispanic, and other); and whether the woman had a usual source of care. Although the insurance plan of the woman was available, for ease of comparison, the health insurance status is included as any private insurance, only public insurance, and uninsured. Additionally, while continuous equivalized household income is used to calculate the CIs, to allow for ease of interpretation of income, it is partitioned into quintiles in the regression.

Table 1 presents the descriptive statistics for the study sample. Whites constitute 45.43% of the sample, 19.92% are black, 26.47% are Hispanic, and 8.18% are other. Additionally, 33.24% had at least a college degree, 53.82% were married or cohabiting, 61.54% had private health insurance, and 20.89% were uninsured. There were 75.62% women with a usual source of care.

Determinants of Cervical Cancer Screening Utilization

Table 2 presents utilization rates for cervical cancer screening in the preceding 12 months and 3 years, with results partitioned across ethnicities. In the 12-month period, utilization in the United States was 60.15%; black women had higher utilization (66.25%) than whites (60.22%) and Hispanics (58.36%). Utilization was higher among those in the highest income quintile compared with those in the poorest income quintile (68.22% vs 53.17%). Although an income gradient exists for each group, the differential between the richest and poorest women was larger for whites (68.46% vs 49.30%) than for blacks (73.54% vs 61.50%) and Hispanics (70.97% vs 54.78%). Disparities are also greater for white women across education and insurance status. Lower-educated blacks (57.96%) and Hispanics (54.32%) had greater 12-month utilization than similar whites (42.22%). Blacks with private health insurance (71.60%) had higher utilization than whites (64.86%) and Hispanics (64.22%) with private health insurance. Uninsured white women had far lower utilization (33.59%) than uninsured black (43.81%) and Hispanic (47.78%) women. Screening across regions also differed across ethnicities, although utilization for each ethnic group is greatest in the northeast.

Overall utilization in the preceding 3 years in the United States is 85.88%. Utilization is greater for black (90.14%) women than for Hispanic (86.68%) and white (85.53%) women. Although white, black, and Hispanic women in the richest group have similar utilization, the poorest black (88.09%) and Hispanic (85.02%) women have higher utilization than the poorest white (76.92%) women. Once more, lower-educated black (86.17%) and Hispanic (84.95%) women and uninsured black (79.13%) and Hispanic (79.01%) women have greater utilization than similar white women (71.07% had no high school degree; 66.70% were uninsured). Marital status for Hispanic and other ethnic women increases utilization by 9.98 and 13.39 percentage points, respectively, though marital status is not as important for black and white women.

Copyright AJMC 2006-2020 Clinical Care Targeted Communications Group, LLC. All Rights Reserved.
Welcome the the new and improved, the premier managed market network. Tell us about yourself so that we can serve you better.
Sign Up