Receipt of cancer screening procedures among Hispanic and non-Hispanic health maintenance organization members

Equal accuracy for Andrew and Abubakar-detecting and mitigating bias in name-ethnicity classification algorithms

Uncovering the world's ethnic inequalities is hampered by a lack of ethnicity-annotated datasets. Name-ethnicity classifiers (NECs) can help, as they are able to infer people's ethnicities from their names. However, since the latest generation of NECs rely on machine learning and artificial intelligence (AI), they may suffer from the same racist and sexist biases found in many AIs. Therefore, this paper offers an algorithmic fairness audit of three NECs. It finds that the UK-Census-trained EthnicityEstimator displays large accuracy biases with regards to ethnicity, but relatively less among gender and age groups. In contrast, the Twitter-trained NamePrism and the Wikipedia-trained Ethnicolr are more balanced among ethnicity, but less among gender and age. We relate these biases to global power structures manifested in naming conventions and NECs' input distribution of names. To improve on the uncovered biases, we program a novel NEC, N2E, using fairness-aware AI techniques. We make N2E freely available at www.name-to-ethnicity.com.

Supplementary Information

The online version contains supplementary material available at 10.1007/s00146-022-01619-4.

Low rates of colorectal, cervical, and breast cancer screening in Asian Americans compared with non-Hispanic whites: Cultural influences or access to care?

BACKGROUND

Asian Americans have lower cancer screening rates compared with non-Hispanic whites (NHWs). Little is known about mechanisms that underlie disparities in cancer screening. The objectives of the current study were 1) to determine the relation between nativity, years in the United States, language, and cancer screening in NHWs and Asian Americans, independent of access to care and 2) to determine whether Asians reported different reasons than NHWs for not obtaining cancer screening.

METHODS

This population-based study included 36,660 NHWs, 1298 Chinese, 944 Filipinos, 803 Koreans, 857 Vietnamese, and 1036 Other Asians from the 2001 California Health Interview Survey. The main study outcomes were 1) self-reported colorectal, cervical, and breast cancer screening and 2) reasons for not obtaining cancer screening.

RESULTS

After adjusting for access to care, several Asian subgroups still had significantly lower rates of all types of cancer screening compared with NHWs. Adjusting for nativity, years in the United States, and English language attenuated the relation between Asian ethnicity and lower rates of colorectal and breast cancer screening. When they were asked what the most important reason was for not having each screening test, foreign-born Asians were significantly more likely than United States-born NHWs to report that they "didn't have problems/symptoms" (P < .01).

CONCLUSIONS

Nativity, years in the United States, and English language may be markers of cultural differences that are mediating cancer screening disparities. Foreign-born Asians may believe that cancer screening is in response to symptoms rather than tests that are used prior to the development of symptoms. Health education messages must consider how to communicate effectively that "cancer screening is valuable, because it finds cancer before it is advanced enough to cause symptoms."

Use of geocoding and surname analysis to estimate race and ethnicity

OBJECTIVE

To review two indirect methods, geocoding and surname analysis, for estimating race/ethnicity as a means for health plans to assess disparities in care.

STUDY DESIGN

Review of published articles and unpublished data on the use of geocoding and surname analyses.

PRINCIPAL FINDINGS

Few published studies have evaluated use of geocoding to estimate racial and ethnic characteristics of a patient population or to assess disparities in health care. Three of four studies showed similar estimates of the proportion of blacks and one showed nearly identical estimates of racial disparities, regardless of whether indirect or more direct measures (e.g., death certificate or CMS data) were used. However, accuracy depended on racial segregation levels in the population and region assessed and geocoding was unreliable for identifying Hispanics and Asians/Pacific Islanders. Similarly, several studies suggest surname analyses produces reasonable estimates of whether an enrollee is Hispanic or Asian/Pacific Islander and can identify disparities in care. However, accuracy depends on the concentrations of Asians or Hispanics in areas assessed. It is less accurate for women and more acculturated and higher SES persons due intermarriage, name changes, and adoption. Surname analysis is not accurate for identifying African Americans. Recent unpublished analyses suggest plans can successfully use a combined geocoding/surname analyses approach to identify disparities in care in most regions. Refinements based on Bayesian methods may make geocoding/surname analyses appropriate for use in areas where the accuracy is currently poor, but validation of these preliminary results is needed.

CONCLUSIONS

Geocoding and surname analysis show promise for estimating racial/ethnic health plan composition of enrollees when direct data on major racial and ethnic groups are lacking. These data can be used to assess disparities in care, pending availability of self-reported race/ethnicity data.

A population-based study of racial and ethnic differences in survival among women with invasive cervical cancer: analysis of Surveillance, Epidemiology, and End Results data

OBJECTIVE

The incidence of cervical cancer is higher in Hispanic than in non-Hispanic or African American women in the United States, but few studies have examined differences in survival between these groups. The objective of this study was to examine racial/ethnic differences in survival after diagnosis with invasive cervical cancer in a population-based sample of patients while adjusting for patient and tumor characteristics and treatment types.

METHODS

We identified 7267 women (4431 non-Hispanic Caucasians, 1830 Hispanic Caucasians, and 1006 non-Hispanic African Americans) diagnosed with primary invasive cervical cancer from 1992 to 1996 (with follow-up through 2000) from the Surveillance, Epidemiology and End Results (SEER) Program. Kaplan-Meier and Cox proportional hazards survival methods were used to assess differences in survival by race/ethnicity.

RESULTS

After adjusting for age at diagnosis, histology, stage, first course of cancer-directed treatment (surgery and radiation therapy), and SEER registry, Hispanic Caucasian women were at 26% decreased risk of death from any cause (hazard ratio (HR) = 0.74, 95% confidence interval (CI): 0.66-0.83) and non-Hispanic African American women were at 19% increased risk of death (HR = 1.19, 95% CI: 1.06-1.33) compared to non-Hispanic Caucasian women over the follow-up period.

CONCLUSION

Analysis of population-based SEER data indicates significant survival differences by race/ethnicity for women with invasive cervical cancer. Hispanic Caucasian women in SEER had improved survival compared to non-Hispanic Caucasian or non-Hispanic African American women.

Variation in screening mammography and Papanicolaou smear by primary care physician specialty and gatekeeper plan (United States)

OBJECTIVE

To assess whether the specialty of a patient's primary care physician or being part of a gatekeeper plan influence breast and cervical cancer screening.

METHODS

Cross-sectional study of women in a national sample. For mammography, we studied women aged 40 and above, and for Papanicolaou (Pap) smear, women aged 18-65 years. Screening mammography or Pap smear within the previous two years was measured by patient self-report. The key independent variables were primary care physician specialty and whether the patient had a gatekeeper.

RESULTS

Among women seen by a family practice physician, there was a higher probability of being screened if the patient was part of a gatekeeper plan than if the patient was not part of a gatekeeper plan: mammography (OR = 1.35; 95% CI = 1.20-1.52) and Pap smear (OR = 1.60; 95% CI = 1.34-1.91). Among women seen by an internal medicine physician, cancer screening did not vary significantly by gatekeeper status.

CONCLUSIONS

The impact of gatekeeper plans upon cancer screening varies according to the primary care physician's specialty. Policy interventions designed to increase cancer screening should take into account different responses to gatekeeper requirements among different types of providers.

Report from the CDC. Use of Preventive Health Services by Hispanic/Latino Women in Two Urban Communities: Atlanta, Georgia and Miami, Florida, 2000 and 2001

PURPOSE

During the 1990s, a 58% increase in the Hispanic/Latino population, fueled by the century's largest immigration wave and the highest fertility of any group, resulted in Hispanics becoming the largest U. S. minority group. To assess use of preventive services by Hispanics in Atlanta, Georgia, the largest Hispanic new destination, and Miami, Florida, the largest established Hispanic community in the Southeast, survey data were analyzed.

METHODS

Miami-Ft. Lauderdale and Atlanta metropolitan area data from the 2000 National Health Interview Survey (NHIS) and from anonymous surveys conducted at health festivals in Miami and Atlanta in 2001 were analyzed.

RESULTS

Female non-Hispanic white and black NHIS respondents were more likely than Hispanic counterparts to report annual household income >$20,000 (77.3%, 70.8% vs. 67.7%), usual source of healthcare (61.5%, 56.4% vs. 50.2%), or ever having had Pap screening (88.8%, 86.7% vs. 80.7%) or oral contraceptive use (55.7%, 59.7% vs. 33.7%). Miami-Ft. Lauderdale Hispanics were less likely than Atlanta respondents to be monolingual Spanish speakers, to lack usual source of healthcare, or to have less than 12 years of education. Of 295 female health festival respondents, the 160 Miami participants were more likely than Atlanta participants to have health insurance, monthly income >$1000, and prior Pap screening (p < 0.01) but less likely to have used contraception (p = 0.07). Most Hispanics felt they had inadequate healthcare; 15.0% reported being denied healthcare because of inability to pay.

CONCLUSIONS

Low income, uninsured status, and language barriers were associated with lower use of preventive services among Hispanics in these Southeastern communities, particularly Atlanta, a new destination.

Use of Pap test among Hispanics and non-Hispanic whites in a rural setting

BACKGROUND

Hispanic women in the United States have a higher incidence of invasive cervical cancer than non-Hispanic whites. A key factor related to the relatively high incidence is the underutilization of cancer-screening services. Several previous investigations have reported that Hispanic women are less likely to comply with screening recommendations for Pap testing; some have identified factors that are associated with non-compliance with screening recommendations.

METHODS

Using baseline personal interview data from a randomized community trial on cancer prevention in a rural area, we assessed the association between health barriers and non-compliance with cervical screening recommendations.

RESULTS

Among Hispanic women (n = 382), those with a lower level of acculturation were more likely than highly acculturated Hispanic and non-Hispanic white women (n = 385) to report personal barriers as reasons for not getting an initial or subsequent screening exam. Such barriers include fear of finding cancer, fear of finding diseases other than cancer, and embarrassment about receiving a physical exam. Certain structural barriers, such as cost of care, no time off work, and lack of transportation, were reported to be associated with non-compliance with routine cervical cancer screening, especially among non-Hispanic white women.

CONCLUSIONS

Future research should examine barriers related to initial and subsequent screening among Hispanics with varying levels of acculturation.

Prostate cancer mortality in Connecticut, Iowa and New Mexico African American men

We sought to assess trends in prostate cancer incidence, treatment and mortality in African American men by means of analysis of prostate cancer data from three states, Connecticut, Iowa and New Mexico, all participants in the Surveillance, Epidemiology, and End Results (SEER) Program. Compared with levels before prostate specific antigen (PSA) testing, prostate cancer incidence increased in all three states after widespread testing. For men diagnosed with localized or regional prostate cancer, the respective increases in radical prostatectomy in Connecticut, Iowa, and New Mexico were 3.2, 2.3, and 4.9 times pre-test levels. Age-standardized mortality in Connecticut and Iowa increased slightly; in New Mexico the 104.7 deaths per 100,000 in 1979-1986, 62.1 in 1987-1990, dropped to 47.6 in 1991-1998, an amount of decline that was statistically significant. Introduction of PSA testing influenced early detection and treatment of prostate cancer in all three states. Although decline in prostate cancer mortality in New Mexico over time may be linked with use of the PSA test and definitive therapy, the relationship among these factors, and thus the proper treatment for the early stages of this condition, is unclear on the basis of these data.

Analysis of the French National Registry of unrelated bone marrow donors, using surnames as a tool for improving geographical localisation of HLA haplotypes

The first statistical analysis of the French National Registry of volunteer bone marrow donors estimated the probabilities of haplotype frequencies separately for each of the 20 administrative regions of France. Here we propose to use donors' surnames to increase the accuracy of location of the donor's geographical origin. This approach allows us to estimate haplotype frequencies for administrative entities (90 departments) smaller than regions and to correct for bias resulting from recent mobility. We analysed 30,777 donors typed for HLA-A,B and 17,745 donors typed for HLA-A,B,DR,DQ. By using the donors' surnames, we identified common and rare haplotypes (those found in only one department) and estimated the degree of HLA polymorphism at the department level. We also identified departments with a distinctive genetic structure (for example, Paris, Corsica, Pyrenees and Meurthe-et-Moselle). By providing a more accurate geographical distribution of HLA polymorphisms in France, this study will enable us to optimise policies for recruiting bone marrow donors and to improve the fit between the donor file and patients' needs.