Benjamin C. Amick’s research while affiliated with University of Arkansas for Medical Sciences and other places

To evaluate algorithmic fairness in low birthweight predictive models. This study analyzed insurance claims (n = 9,990,990; 2013–2021) linked with birth certificates (n = 173,035; 2014–2021) from the Arkansas All Payers Claims Database (APCD). Low birthweight (< 2500 g) predictive models included four approaches (logistic, elastic net, linear discriminate analysis, and gradient boosting machines [GMB]) with and without racial/ethnic information. Model performance was assessed overall, among Hispanic individuals, and among non-Hispanic White, Black, Native Hawaiian/Other Pacific Islander, and Asian individuals using multiple measures of predictive performance (i.e., AUC [area under the receiver operating characteristic curve] scores, calibration, sensitivity, and specificity). AUC scores were lower (underperformed) for Black and Asian individuals relative to White individuals. In the strongest performing model (i.e., GMB), the AUC scores for Black (0.718 [95% CI: 0.705–0.732]) and Asian (0.655 [95% CI: 0.582–0.728]) populations were lower than the AUC for White individuals (0.764 [95% CI: 0.754–0.775 ]). Model performance measured using AUC was comparable in models that included and excluded race/ethnicity; however, sensitivity (i.e., the percent of records correctly predicted as “low birthweight” among those who actually had low birthweight) was lower and calibration was weaker, suggesting underprediction for Black individuals when race/ethnicity were excluded. This study found that racially blind models resulted in underprediction and reduced algorithmic performance, measured using sensitivity and calibration, for Black populations. Such under prediction could unfairly decrease resource allocation needed to reduce perinatal health inequities. Population health management programs should carefully consider algorithmic fairness in predictive models and associated resource allocation decisions.

Purpose This study aimed to evaluate the association of mental health and substance use disorders on the risk of adverse infant outcomes overall and by race/ethnicity and payer. Methods We used birth certificates (2017–2022; n = 125,071) linked with state-wide insurance claims (2016–2022; n = 7,583,488) to assess the risk of an adverse infant outcome (i.e., prematurity [< 37 weeks gestation] or low birthweight [< 2,500 g]) associated with “any mental health” or “any substance use” disorder overall, by race/ethnicity, and by payer using diagnoses during the 9 months of pregnancy. We additionally evaluated seven specific mental health conditions and four specific substance use disorders. Results The rate of having an adverse infant outcome was 13.4%. Approximately 21.5% of birthing individuals had a mental health condition, and 8.7% had a substance use disorder. We found increased adjusted risk of an adverse infant outcome associated with having a mental health condition overall (aRR: 1.28; 95%CI: 1.23–1.32) and for all racial/ethnic groups and payers. We additionally found increased risk associated with substance use disorder overall (aRR: 1.32; 95%CI: 1.25–1.40) and for White, Black, privately-covered, and Medicaid-covered individuals. There was increased risk associated with six of seven mental health and three of four substance use disorders. Conclusions Given the risk of adverse infant outcomes associated with mental health and substance use disorders across racial/ethnic groups and payers, our findings highlight the critical importance of policies and clinical guidelines that support early identification and treatment of a broad spectrum of mental health and substance use disorders throughout the perinatal period.

Background and Aims Longitudinal change in income is crucial in explaining cardiovascular health inequalities. However, there is limited evidence for cardiovascular disease (CVD) risk associated with income dynamics over time among individuals with type 2 diabetes (T2D). Methods Using a nationally representative sample from the Korean National Health Insurance Service database, 1 528 108 adults aged 30–64 with T2D and no history of CVD were included from 2009 to 2012 (mean follow-up of 7.3 years). Using monthly health insurance premium information, income levels were assessed annually for the baseline year and the four preceding years. Income variability was defined as the intraindividual standard deviation of the percent change in income over 5 years. The primary outcome was a composite event of incident fatal and nonfatal CVD (myocardial infarction, heart failure, and stroke) using insurance claims. Hazard ratios (HRs) and 95% confidence intervals (CIs) were estimated after adjusting for potential confounders. Results High-income variability was associated with increased CVD risk (HRhighest vs. lowest quartile 1.25, 95% CI 1.22–1.27; Ptrend < .001). Individuals who experienced an income decline (4 years ago vs. baseline) had increased CVD risk, which was particularly notable when the income decreased to the lowest level (i.e. Medical Aid beneficiaries), regardless of their initial income status. Sustained low income (i.e. lowest income quartile) over 5 years was associated with increased CVD risk (HRn = 5 years vs. n = 0 years 1.38, 95% CI 1.35–1.41; Ptrend < .0001), whereas sustained high income (i.e. highest income quartile) was associated with decreased CVD risk (HRn = 5 years vs. n = 0 years 0.71, 95% CI 0.70–0.72; Ptrend < .0001). Sensitivity analyses, exploring potential mediators, such as lifestyle-related factors and obesity, supported the main results. Conclusions Higher income variability, income declines, and sustained low income were associated with increased CVD risk. Our findings highlight the need to better understand the mechanisms by which income dynamics impact CVD risk among individuals with T2D.

Background The risk of incident atrial fibrillation (AF) among breast cancer survivors, especially for younger women, and cancer treatment effects on the association remain unclear. This study aimed to investigate the risk of AF among breast cancer survivors and evaluate the association by age group, length of follow-up, and cancer treatment. Methods Using data from the Korean Health Insurance Service database (2010–2017), 113,232 women newly diagnosed with breast cancer (aged ≥ 18 years) without prior AF history who underwent breast cancer surgery were individually matched 1:5 by birth year to a sample female population without cancer (n = 566,160) (mean[SD] follow-up, 5.1[2.1] years). Sub-distribution hazard ratios (sHRs) and 95% confidence intervals (CIs) considering death as a competing risk were estimated, adjusting for sociodemographic factors and cardiovascular/non-cardiovascular comorbidities. Results BCS had a slightly increased AF risk compared to their cancer-free counterparts (sHR 1.06; 95% CI 1.00–1.13), but the association disappeared over time. Younger BCS (age < 40 years) had more than a 2-fold increase in AF risk (sHR 2.79; 95% CI 1.98–3.94), with the association remaining similar over 5 years of follow-up. The increased risk was not observed among older BCS, especially those aged > 65 years. Use of anthracyclines was associated with increased AF risk among BCS (sHR 1.57; 95% CI 1.28–1.92), which was more robust in younger BCS (sHR 1.94; 95% CI 1.40–2.69 in those aged ≤ 50 years). Conclusions Our findings suggest that younger BCS had an elevated risk of incident AF, regardless of the length of follow-up. Use of anthracyclines may be associated with increased mid-to-long-term AF risk among BCS. Graphical Abstract

Background and Aims: Longitudinal change in income is crucial in explaining cardiovascular health inequalities, but there is limited evidence for cardiovascular disease (CVD) risk associated with income dynamics over time among individuals with type 2 diabetes (T2D). Methods: Using a nationally representative sample from the Korean Health Insurance Service database, 1,528,108 adults with T2D aged 30-64 years and no history of CVD were enrolled between 2009-2012 (mean follow-up of 7.3 years). Using monthly health insurance premiums information, income levels were assessed annually for 4 years before the baseline year. Income variability was defined as the intraindividual standard deviation of the percent change in income across 5 years. The primary outcome was a composite event of incident fatal and nonfatal CVD (myocardial infarction, stroke, and heart failure) using insurance claims. Hazard ratios (HRs) and 95% confidence intervals (CIs) were estimated after adjusting for potential confounders. Results: Sustained low income (i.e., lowest income quartile) over 5 years was associated with increased CVD risk (HRn=5years vs. n=0years 1.38, 95% CI 1.35-1.41; Ptrend<0.0001), whereas sustained high income (i.e., highest income quartile) was associated with decreased CVD risk (HR 0.71n=5years vs. n=0years 95% CI 0.70-0.72; Ptrend<0.0001). High-income variability was associated with increased CVD risk (HRhighest vs. lowest quartile 1.25, 95% CI 1.22-1.27; Ptrend<0.001). Individuals who experienced an income decline across 5 years leading up to baseline had increased CVD risk, particularly in a decrease to the lowest income level (i.e., Medical Aids beneficiaries), regardless of initial income status. Sensitivity analyses, including potential mediators, such as lifestyle-related factors and obesity, supported the results. Conclusions: Among non-elderly Korean adults with T2D, sustained low income, higher income variability, and income declines were associated with increased CVD risk. Our findings highlight the need to understand better the mechanisms by which income dynamics impact CVD risk among individuals with T2D.

Objective: Disparities in COVID-19 vaccine coverage among college students are not well studied. We analyzed the extent of COVID-19 vaccination among all students at public institutions of higher education in Arkansas in spring and fall 2021 and examined factors associated with vaccination rates. Methods: We matched student enrollment data for the spring and fall 2021 semesters separately with immunization registry data to determine the COVID-19 vaccination status of students in each semester. We used multivariable logistic regression to determine variables associated with being fully vaccinated. Results: Overall, by the end of the fall 2021 semester, 49.1% of students were fully vaccinated, compared with 52.3% of students at the end of the spring 2021 semester. However, we observed important differences between student groups. Students who were in rural areas (vs urban areas), freshman (vs all other class years), at a 2-year college (vs 4-year college), non-Hispanic Black (vs Hispanic or non-Hispanic White), or in a nonhealth major (vs health major) were less likely to be fully vaccinated. Conclusion: Given the disparities in COVID-19 vaccine coverage among college students, continued efforts are needed in this young, but important, population segment to address students' confidence in vaccines and to make vaccines more available and accessible on and near campuses. Novel approaches for tailored messaging and interventions to facilitate vaccine uptake may be needed for groups such as non-Hispanic Black students and students in rural areas who are less likely, compared with their counterparts, to accept COVID-19 vaccines.

Background: The mid- and long-term risk of incident atrial fibrillation (AF) among breast cancer survivors, especially for younger women, and cancer treatment effects on the association remain unclear. We aimed to investigate the risk of AF among breast cancer survivors and evaluate the association by age group, length of follow-up, and cancer treatment. Methods: Using data from the Korean Health Insurance Service database between January 2010 and December 2017, 113,232 women newly diagnosed with breast cancer (aged 18 years or older) without prior AF history who underwent breast cancer surgery were individually matched 1:5 by birth year to a sample female population without cancer (n=566,160). Sub-distribution hazard ratios (sHRs) and 95% confidence intervals (CIs) considering death as a competing risk were estimated, adjusting for sociodemographic factors and cardiovascular/non-cardiovascular comorbidities. Results: During follow-up (mean [SD] follow-up, 5.1 [2.1] years), AF developed in 1,166 (1.0%) breast cancer surgery survivors at least 1 year after enrollment. Overall, breast cancer survivors had a slightly increased AF risk compared to their cancer-free counterparts (sHR 1.06; 95% CI 1.00-1.13), but the association disappeared over time. Younger breast cancer survivors (age<40 years) had more than a 2-fold increase in AF risk (sHR 2.79; 95% CI 1.98-3.94), with the association remaining similar over 5 years of follow-up. The increased risk was not observed among older breast cancer survivors, especially those aged>65 years. Breast cancer survivors who received anthracyclines had an increased risk of AF compared to those without the exposure (sHR 1.57; 95% CI 1.28-1.92) over the entire course of follow-up. The association between anthracyclines and AF incidence was also more robust in younger breast cancer survivors (sHR 1.94; 95% CI 1.40-2.69 in those aged 50 years or younger). Sensitivity analyses, including further adjustments for obesity and lifestyle factors, supported the results. Conclusions: Our findings suggest that younger breast cancer survivors had an elevated risk of incident AF, regardless of the length of follow-up. The use of anthracyclines may increase the mid-to-long-term AF risk among breast cancer surgery survivors.

Objective: Assess university students' SARS-CoV-2 antibody seroprevalence and mitigation behaviors over time. Participants: Randomly selected college students (N = 344) in a predominantly rural Southern state. Methods: Participants provided blood samples and completed self-administered questionnaires at three timepoints over the academic year. Adjusted odds ratios and 95% confidence intervals were estimated from logistic regression analyses. Results: SARS-CoV-2 antibody seroprevalence was 18.2% in September 2020, 13.1% in December, and 45.5% in March 2021 (21% for those with no vaccination history). SARS-CoV-2 antibody seroprevalence was associated with large social gatherings, staying local during the summer break, symptoms of fatigue or rhinitis, Greek affiliation, attending Greek events, employment, and using social media as the primary COVID-19 information source. In March 2021, seroprevalence was associated with receiving at least one dose of a COVID-19 vaccination. Conclusion: SARS-CoV-2 seroprevalence was higher in this population of college students than previous studies. Results can assist leaders in making informed decisions as new variants threaten college campuses.

Only a few studies and reports assessing the natural history and symptomatology for COVID-19 by gender have been reported in literature to date. Thus, the objective of this study was to examine patterns in symptomology of COVID-19 by gender among a diverse adult population in Arkansas. Data on COVID-19 symptoms was collected at day of testing, 7th day and 14th day among participants at UAMS mobile testing units throughout the state of Arkansas. Diagnosis for SARS-CoV-2 infection was confirmed via nasopharyngeal swab and RT-PCR methods. Data analysis was conducted using Chi-square test and Poisson regression to assess the differences in characteristics by gender. A total of 60,648 community members and patients of Arkansas received RT-PCR testing. Among adults testing positive, we observed a statistically significant difference for fever (p < 0.001) and chills (p = 0.04). Males were more likely to report having a fever (22.6% vs. 17.1%; p < 0.001) and chills (14.9% vs. 12.6%; p = 0.04) compared to females. Among adults testing negative, females were more likely to report each symptom than males. To conclude, we observed a greater prevalence of certain symptoms such as fever and chills among men testing positive for COVID-19, compared to women during the time of testing. These differences elucidate the important issue of rapidly emerging health disparities during the COVID-19 pandemic.

Advancements in cancer diagnosis and treatment have resulted in improvements in survivor outcomes; however, cancer survivors are more likely to experience adverse employment outcomes such as job loss, reduced working hours, and early retirement. The purpose of this study was to examine employment disparities among cancer survivors. Our study collected data from 29,136 cancer survivors (ages 18–65) between 2015 and 2021 using electronic health records (EHR) and linked to cancer registry data. Of those with employment information (n = 7296), differences in employment status were explored by race, ethnicity, sex, geography, marital status, education, age, and cancer site. Of the patients with employment status available, 61% were employed, 28% were not employed, 9% were disabled, 2% were retired. Logistic regression results revealed adjusted effects: a positive association between employment and marriage, while racial and ethnic minority adults, rurality, and certain age categories were less likely to be employed. Unadjusted results showed a positive association between employment and education. These results contribute to an emerging body of literature showing adverse employment outcomes for cancer survivors.