Bayesian kernel machine regression for count data: modelling the association between social vulnerability and COVID-19 deaths in South Carolina

The COVID-19 pandemic created an unprecedented global health crisis. Recent studies suggest that socially vulnerable communities were disproportionately impacted, although findings are mixed. To quantify social vulnerability in the US, many studies rely on the Social Vulnerability Index (SVI), a county-level measure comprising 15 census variables. Typically, the SVI is modelled in an additive manner, which may obscure non-linear or interactive associations, further contributing to inconsistent findings. As a more robust alternative, we propose a negative binomial Bayesian kernel machine regression (BKMR) model to investigate dynamic associations between social vulnerability and COVID-19 death rates, thus extending BKMR to the count data setting. The model produces a 'vulnerability effect' that quantifies the impact of vulnerability on COVID-19 death rates in each county. The method can also identify the relative importance of various SVI variables and make future predictions as county vulnerability profiles evolve. To capture spatio-temporal heterogeneity, the model incorporates spatial effects, county-level covariates, and smooth temporal functions. For Bayesian computation, we propose a tractable data-augmented Gibbs sampler. We conduct a simulation study to highlight the approach and apply the method to a study of COVID-19 deaths in the US state of South Carolina during the 2021 calendar year.

Preliminary Effects of a Guided Self-Change Intervention on Perceived Risk and Self-Efficacy in University Students Engaging in Cannabis or Alcohol Misuse

Guided Self-Change (GSC) is a Motivational Interviewing (MI)-based early intervention program, infused with Cognitive Behavioral Therapy (CBT), for individuals with substance use problems. In this study, we implemented a 4-session GSC program with the innovative addition of mindfulness-based techniques at a minority-serving institution to reduce substance use and negative consequences among self-referred university students. We investigated processes that may be associated with behavior change, including perceived risk of use and self-efficacy ratings among university students who reported their primary substance of choice was cannabis (n = 18) or alcohol (n = 18). The sample of 36 participants (Mage = 24.4, SDage = 5, range 18-37) mostly identified as female (58.3%), then male (41.7%); 52.8% identified as Hispanic/Latine, 22.2% as Black or African American, and 19.5% as a sexual minority. Among cannabis primary using students, results indicated that the perceived risk of weekly cannabis use, confidence to change, and readiness to change showed statistically significant increases from pre- to post-assessment. Among alcohol primary using students, confidence to change and readiness to change showed statistically significant increases from pre- to post-assessments. All results yielded large effect sizes, which may be inflated due to the small sample size. Findings suggest that over the course of participation in a brief, 4-session targeted GSC program, there were significant increases in perceived risk and self-efficacy among minority university students who engage in primary cannabis or primary alcohol use.

Timeliness of diagnostic evaluation for postmenopausal bleeding: A retrospective cohort study using claims data

BACKGROUND

Postmenopausal bleeding (PMB) is a common gynecologic condition. Although it can be a sign of uterine cancer, most patients have benign etiology. However, research on quality of diagnostic evaluation for PMB has been limited to cancer patients. To extend this research, we examined the timeliness of diagnostic evaluation for PMB among patients with benign conditions.

METHODS

Using the 2008-2019 MarketScan Research Databases, we identified 499176 patients (456741 with commercial insurance and 42435 with Medicaid insurance) who presented with PMB but did not have gynecologic cancer. For each patient, we measured the time from their PMB reporting to the date of their first diagnostic procedure. The association between patient characteristics and time to first diagnostic procedure was examined using Cox proportional hazards models (for the overall sample and then stratified by insurance type).

RESULTS

Overall, 54.3% of patients received a diagnostic procedure on the same day when they reported PMB and 86.6% received a diagnostic procedure within 12 months after reporting PMB. These percentages were 39.4% and 77.1%, respectively, for Medicaid patients, compared to 55.7% and 87.4%, respectively, for commercially insured patients (p<0.001 for both). Medicaid patients had an 18% lower rate of receiving a diagnostic procedure at any given time point than commercially insured patients (adjusted hazard ratio = 0.82, 95% CI: 0.81-0.83). Meanwhile, older age and non-gynecologic comorbidities were associated with a lower rate whereas concomitant gynecologic conditions and recent use of preventive care were associated with a higher rate of receiving diagnostic procedures. Analysis stratified by insurance type identified additional risk factors for delayed diagnostic procedures (e.g., non-metropolitan versus metropolitan location for commercially insured patients and Black versus White race for Medicaid patients).

CONCLUSION

A sizable proportion of patients did not receive prompt diagnostic evaluation for PMB. Both clinical and non-clinical factors could affect timeliness of evaluation.

Burden of unmet health-related social needs in an academic adult primary care practice in San Francisco California

OBJECTIVES

Unmet health-related social needs can influence health outcomes and increase healthcare utilization. There is growing interest in integrating social needs care into healthcare delivery. We conducted an assessment of health-related social needs in an academic adult primary care practice in San Francisco, California.

METHODS

We recruited a random convenience sample of adult English-, Chinese- or Spanish-speaking patients from clinic waiting rooms at the study sites to complete a self-administered, anonymous survey. We used the Accountable Health Communities Health-Related Social Needs Screening Tool for these domains: housing instability, food insecurity, transportation problems, utility help needs, interpersonal safety, financial strain, and family/community support. We conducted univariate and multivariate analyses adjusting for age, sex and survey language.

RESULTS

679 patients completed the survey. Respondents were 57% female and mean age of 58 ± 18 years old. 54% of patients had at least one unmet health-related social need. The most prevalent health-related social needs were financial strain (35%), at least one issue with housing conditions (27%), and food insecurity (23%). Respondents completing the survey in Spanish had significantly higher odds of reporting food insecurity (AOR 3.97, 95%CI 1.86, 8.46), transportation problems (AOR 3.13, 95%CI 1.32, 7.43), and need for support with activities of daily living (AOR 4.58, 95%CI 2.04, 10.25) than respondents completing the survey in English.

CONCLUSIONS

The burden of unmet health-related social needs was considerable in this adult primary care practice. These findings can support a case for integrating health-related social need screening and social care in the delivery of primary care in the United States to advance health equity.

Investigating the association between educational attainment and allostatic load with risk of cancer mortality among African American women

BACKGROUND

African American (AA) women navigate the world with multiple intersecting marginalized identities. Accordingly, AA women have higher cumulative stress burden or allostatic load (AL) compared to other women. Studies suggest that AA women with a college degree or higher have lower AL than AA women with less than a high school diploma. We examined the joint effect of educational attainment and AL status with long-term risk of cancer mortality, and whether education moderated the association between AL and cancer mortality.

METHODS

We performed a retrospective analysis among 4,677 AA women within the National Health and Nutrition Examination Survey (NHANES) from 1988 to 2010 with follow-up data through December 31, 2019. We fit weighted Cox proportional hazards models to estimate adjusted hazard ratios (aHRs) of cancer death between educational attainment/AL (adjusted for age, income, and smoking status).

RESULTS

AA women with less than a high school diploma living with high AL had nearly a 3-fold increased risk (unadjusted HR: 2.98; 95%C CI: 1.24-7.15) of cancer death compared to AA college graduates living with low AL. However, after adjusting for age, this effect attenuated (age-adjusted HR: 1.11; 95% CI: 0.45-2.74). AA women with high AL had 2.3-fold increased risk of cancer death (fully adjusted HR: 2.26; 95% CI: 1.10-4.57) when compared to AA with low AL, specifically among women with high school diploma or equivalent and without history of cancer.

CONCLUSIONS

Our findings suggest that high allostatic load is associated with a higher risk of cancer mortality among AA women with lower educational attainment, while no such association was observed among AA women with higher educational attainment. Thus, educational attainment plays a modifying role in the relationship between allostatic load and the risk of cancer death for AA women. Higher education can bring several benefits, including improved access to medical care and enhanced medical literacy, which in turn may help mitigate the adverse impact of AL and the heightened risk of cancer mortality among AA women.

Investigating the Joint Effect of Allostatic Load among Lesbian, Gay, and Bisexual Adults with Risk of Cancer Mortality

Sexual minorities (SM) have higher chronic physiologic stress as indicated by allostatic load (AL), which may be explained in part by consistent experiences of discriminatory practices. This is one of the first studies to examine the joint effects of SM status and AL on the association with long-term risk for cancer death. Retrospective analyses were conducted on 12,470 participants using National Health and Nutrition Examination Survey (NHANES) from years 2001 through 2010 linked with the National Death Index through December 31, 2019. Cox proportional hazards models estimated adjusted hazard ratios (aHRs) of cancer deaths between groups of SM (those reporting as gay, lesbian, bisexual, or having same-sex sexual partners) status and AL. SM adults living with high AL (n = 326) had a 2-fold increased risk of cancer death (aHR: 2.55, 95% CI: 1.40-4.65) when compared to straight/heterosexual adults living with low AL (n = 6674). Among those living with high AL, SM (n = 326) had a 2-fold increased risk of cancer death (aHR: 2.26, 95% CI: 1.33-3.84) when compared to straight/heterosexual adults with high AL (n = 4957). SM with high AL have an increased risk of cancer mortality. These findings highlight important implications for promoting a focused agenda on cancer prevention with strategies that reduce chronic stress for SM adults.

Disparities in access to robotic technology and perioperative outcomes among patients treated with radical prostatectomy

BACKGROUND

Most radical prostatectomies are completed with robotic assistance. While studies have previously evaluated perioperative outcomes of robot-assisted radical prostatectomy (RARP), this study investigates disparities in access and clinical outcomes of RARP.

STUDY DESIGN

The National Cancer Database (NCDB) was used to identify patients who received radical prostatectomy for cancer between 2010 and 2017 with outcomes through 2018. RARP was compared to open radical prostatectomy (ORP). Odds of receiving RARP were evaluated while adjusting for covariates. Overall survival was evaluated using a propensity-score matched cohort.

RESULTS

Overall, 354 752 patients were included with 297 676 (83.9%) receiving RARP. Patients who were non-Hispanic Black (82.8%) or Hispanic (81.3%) had lower rates of RARP than non-Hispanic White (84.0%) or Asian patients (87.7%, p < 0.001). Medicaid or uninsured patients were less likely to receive RARP (75.5%) compared to patients with Medicare or private insurance (84.4%, p < 0.001). Medicaid or uninsured status was associated with decreased odds of RARP in adjusted multivariable analysis (OR 0.61, 95% CI 0.49-0.76). RARP was associated with decreased perioperative mortality and improved overall survival compared to ORP.

CONCLUSION

Patients who were underinsured were less likely to receive RARP. Improved access to RARP may lead to decreased disparities in perioperative outcomes for prostate cancer.

Racial Differences in 25-Hydroxy Vitamin D and Self-Reported Pain Severity in a Sample of Individuals Living with Non-Specific Chronic Low Back Pain

Introduction

Considerable evidence suggests that there are significant ethnic/racial differences in the experience of pain among individuals suffering from chronic musculoskeletal conditions. Additionally, low levels of vitamin D have been associated with pain severity. Further, vitamin D deficiency is more prevalent in Non-Hispanic Black (NHB) individuals compared to Non-Hispanic Whites (NHW).

Objective

The aim of this study was to investigate the associations among race, pain severity, and serum levels of vitamin D in a sample of patients with chronic low back pain (cLBP).

Methods

All study participants (n = 155) self-identified their race/ethnicity as either NHB or NHW. Blood samples were collected to assess circulating levels of serum 25- hydroxy vitamin D. Vitamin D levels were categorized as optimal (≥20 ng/mL), insufficient (12-19 ng/mL) or deficient (<12 ng/mL). Participants then self-reported their pain severity using the Brief Pain Inventory - Short Form.

Results

Results showed that a greater proportion of NHB versus NHW participants were categorized as Vitamin D deficient (χ 2 (2, N = 155) = 16.79, p < 0.001). An analysis of covariance (ANCOVA) revealed that NHBs reported significantly greater pain severity relative to NHWs (F(1150) = 6.45) p = 0.012. Further, self-reported pain severity significantly differed according to Vitamin D clinical categories (F(2150) = 4.19, p = 0.013). Participants with deficient vitamin D reported significantly greater pain severity in comparison to participants with optimal vitamin D (F(1101) = 7.28, p = 0.008).

Conclusion

The findings suggest that Vitamin D deficiency may be linked to greater pain severity in a sample of individuals with cLBP, especially for those who identify as NHB.

Attitudes Toward HIV-Positive Status Disclosure Among U=U-Aware Sexual and Gender Minority Individuals in the USA: a Consensual Qualitative Research Approach

Introduction

The U=U (i.e., undetectable equals untransmittable) campaign is founded upon biomedical advancements that have positioned HIV as a manageable condition with effectively zero risk of transmission. In spite of these developments, attitudes of sexual and gender minority populations regarding the necessity of seropositive status disclosure remain unexamined.

Methods

The current study analyzed qualitative data regarding the necessity of seropositive status disclosure from 62 sexual minority men as well as transgender and gender non-conforming individuals who have sex with men from 2020 to 2021.

Results

The majority of participants believed disclosure to be necessary and invoked several social and structural factors that informed their attitudes. Participants cited HIV criminalization laws, the ethics of non-disclosure, and disclosure as a means of educating sex partners when appraising the necessity of disclosure. Participants also presented concerns regarding U=U efficacy and HIV stigma.

Conclusions

Findings indicate that the disclosure of seropositive status to sex partners is still important to U=U-aware sexual and gender minority individuals. The majority of the study sample, irrespective of HIV status, believed seropositive status disclosure was necessary in advance of sex.

Policy Implications

Findings suggest opportunities for public health messaging to remediate concerns about U=U efficacy, combat misinformation, and clarify out-of-date information on HIV criminalization.

Utilization of machine learning methods for predicting surgical outcomes after total knee arthroplasty

Background

Predictive models could help clinicians identify risk factors that cause adverse events after total knee arthroplasty (TKA), allowing for appropriate preoperative preventive interventions and allocation of resources.

Methods

The National Inpatient Sample datasets from 2010–2014 were used to build Logistic Regression (LR), Gradient Boosting Method (GBM), Random Forest (RF), and Artificial Neural Network (ANN) predictive models for three clinically relevant outcomes after TKA—disposition at discharge, any post-surgical complications, and blood transfusion. Model performance was evaluated using the Brier scores as calibration measures, and area under the ROC curve (AUC) and F1 scores as discrimination measures.

Results

GBM-based predictive models were observed to have better calibration and discrimination than the other models; thus, indicating comparatively better overall performance. The Brier scores for GBM models predicting the outcomes under investigation ranged from 0.09–0.14, AUCs ranged from 79–87%, and F1-scores ranged from 41–73%. Variable importance analysis for GBM models revealed that admission month, patient location, and patient’s income level were significant predictors for all the outcomes. Additionally, any post-surgical complications and blood transfusions were significantly predicted by deficiency anemias, and discharge disposition by length of stay and age groups. Notably, any post-surgical complications were also significantly predicted by the patient undergoing blood transfusion.

Conclusions

The predictive abilities of the ML models were successfully demonstrated using data from the National Inpatient Sample (NIS), indicating a wide range of clinical applications for obtaining accurate prognoses of complications following orthopedic surgical procedures.

Twitter Fingers and Echo Chambers: Exploring Expressions and Experiences of Online Racism Using Twitter

Social media sites, such as Twitter, represent a growing setting in which racism and related stress may manifest. The aims of this exploratory qualitative study were to (1) understand the essence of Twitter users' lived experience with and response to content about race and racism on the platform, and (2) explore their perceptions of how discussions about race and racism on Twitter may impact health and well-being. We conducted six focus groups and four interviews with adult Twitter users (n = 27) from Berkeley, California, and Greenville, South Carolina. We managed the data with NVivo and conducted an interpretative phenomenological analysis to identify themes. Participants described Twitter content as displaying both overt and subtle expressions of racism, particularly for Black and Latinx people, and serving as an echo chamber where similar viewpoints are amplified. Participants described how Twitter users may feel emboldened to type offensive tweets based on the perception of anonymity, and that these tweets were sometimes met with community disapproval used to provide a collective calibration to restore the social norms of the online space. Participants perceived harmful mental, emotional, and physical health impacts of exposure to racist content on Twitter. Our participants responded to harmful race-related content through blocking users and following others in order to curate their Twitter feeds, actively engaging in addressing content, and reducing Twitter use. Twitter users reported witnessing racism on the platform and have found ways to protect their mental health and cope with discussions of race and racism in this social media environment.

Gender-sexuality alliance membership and activities: associations with students' comfort, confidence and awareness regarding substance use resources

Belonging to a school Gender and Sexuality Alliance (GSA) is associated with lower substance use among LGBTQ+ youth. However, it is unknown whether GSA participation facilitates access to resources for substance use concerns. Using longitudinal data from 38 Massachusetts high schools, we compared sources of support for substance use concerns listed by GSA members (n = 361) and nonmembers (n = 1539). Subsequently, we tested whether GSA membership was associated with comfort, confidence and awareness regarding substance use resources in school and the community. Finally, we assessed whether specific GSA activities and discussions (e.g. social support) were associated with these outcomes. Among students with recent substance use, GSA membership was associated with greater comfort, confidence and awareness regarding school-based substance use resources in the spring semester, adjusted for fall semester levels and non-GSA club involvement. Furthermore, students in GSAs where members reported more advocacy and social support activities reported higher levels of comfort, confidence and awareness regarding community-based substance use resources. These results indicate that among students using alcohol or nicotine products, GSA members may be more receptive to school-based substance use prevention efforts. Furthermore, GSA-based social support and activism experiences may promote access to community-based substance use resources.

Neighborhood disadvantage and chronic disease management

OBJECTIVE

To assess the relationship between a composite measure of neighborhood disadvantage, the Area Deprivation Index (ADI), and control of blood pressure, diabetes, and cholesterol in the Medicare Advantage (MA) population.

DATA SOURCES

Secondary analysis of 2013 Medicare Healthcare Effectiveness Data and Information Set, Medicare enrollment data, and a neighborhood disadvantage indicator.

STUDY DESIGN

We tested the association of neighborhood disadvantage with intermediate health outcomes. Generalized estimating equations were used to adjust for geographic and individual factors including region, sex, race/ethnicity, dual eligibility, disability, and rurality.

DATA COLLECTION

Data were linked by ZIP+4, representing compact geographic areas that can be linked to Census block groups.

PRINCIPAL FINDINGS

Compared with enrollees residing in the least disadvantaged neighborhoods, enrollees in the most disadvantaged neighborhoods were 5 percentage points (P < 0.05) less likely to have controlled blood pressure, 6.9 percentage points (P < 0.05) less likely to have controlled diabetes, and 9.9 percentage points (P < 0.05) less likely to have controlled cholesterol. Adjustment attenuated this relationship, but the association remained.

CONCLUSIONS

The ADI is a strong, independent predictor of diabetes and cholesterol control, a moderate predictor of blood pressure control, and could be used to track neighborhood-level disparities and to target disparities-focused interventions in the MA population.