Development of a Machine Learning Modelling Tool for Predicting HIV Incidence Using Public Health Data from a County in the Southern United States.

BACKGROUND: Recent advancements in Machine Learning (ML) have significantly improved the accuracy of models predicting HIV incidence. These models typically utilize electronic medical records and patient registries. This study aims to broaden the application of these tools by utilizing de-identified public health datasets for notifiable sexually transmitted infections (STIs) from a southern U.S. County known for high HIV incidence rates. The goal is to assess the feasibility and accuracy of ML in predicting HIV incidence, which could potentially inform and enhance public health interventions. METHODS: We analyzed two de-identified public health datasets, spanning January 2010 to December 2021, focusing on notifiable STIs. Our process involved data processing and feature extraction, including sociodemographic factors, STI cases, and social vulnerability index (SVI) metrics. Various ML algorithms were trained and evaluated for predicting HIV incidence, using metrics such as accuracy, precision, recall, and F1 score. RESULTS: The study included 85,224 individuals, with 2,027 (2.37%) newly diagnosed with HIV during the study period. The ML models demonstrated high performance in predicting HIV incidence among males and females. Influential predictive features for males included age at STI diagnosis, previous STI information, provider type, and SVI. For females, they included age, ethnicity, previous STIs information, overall SVI, and race. CONCLUSIONS: The high accuracy of our ML models in predicting HIV incidence highlights the potential of using public health datasets for public health interventions such as tailored HIV testing and prevention. While these findings are promising, further research is needed to translate these models into practical public health applications.

Public Health Response to Clusters of Rapid HIV Transmission Among Hispanic or Latino Gay, Bisexual, and Other Men Who Have Sex with Men - Metropolitan Atlanta, Georgia, 2021-2022.

During February 2021-June 2022, the Georgia Department of Public Health (GDPH) detected five clusters of rapid HIV transmission concentrated among Hispanic or Latino (Hispanic) gay, bisexual, and other men who have sex with men (MSM) in metropolitan Atlanta. The clusters were detected through routine analysis of HIV-1 nucleotide sequence data obtained through public health surveillance (1,2). Beginning in spring 2021, GDPH partnered with health districts with jurisdiction in four metropolitan Atlanta counties (Cobb, DeKalb, Fulton, and Gwinnett) and CDC to investigate factors contributing to HIV spread, epidemiologic characteristics, and transmission patterns. Activities included review of surveillance and partner services interview data,† medical chart reviews, and qualitative interviews with service providers and Hispanic MSM community members. By June 2022, these clusters included 75 persons, including 56% who identified as Hispanic, 96% who reported male sex at birth, 81% who reported male-to-male sexual contact, and 84% of whom resided in the four metropolitan Atlanta counties. Qualitative interviews identified barriers to accessing HIV prevention and care services, including language barriers, immigration- and deportation-related concerns, and cultural norms regarding sexuality-related stigma. GDPH and the health districts expanded coordination, initiated culturally concordant HIV prevention marketing and educational activities, developed partnerships with organizations serving Hispanic communities to enhance outreach and services, and obtained funding for a bilingual patient navigation program with academic partners to provide staff members to help persons overcome barriers and understand the health care system. HIV molecular cluster detection can identify rapid HIV transmission among sexual networks involving ethnic and sexual minority groups, draw attention to the needs of affected populations, and advance health equity through tailored responses that address those needs.

Gay Neighborhoods: Can They Be Identified in a Systematic Way Using Latent Class Analysis?

Identifying gay neighborhoods could help in targeting HIV prevention efforts for men who have sex with men. This study's purpose was to identify gay neighborhoods using latent class analysis (LCA). Data at the ZIP code level were drawn from the American Community Survey, website lists of gay bars and neighborhoods, and the Florida Department of Health HIV surveillance system. A two-class model was selected based on fit. About 9% of the ZIP code data was in class two, which was designated as gay neighborhoods. Cohen's kappa coefficient was used to examine agreement between the classification of ZIP codes from LCA and websites. Fair agreement was found (0.2501). Gay neighborhoods could serve as a place to disseminate information about pre-exposure prophylaxis and other methods for HIV prevention. Improved measures, such as the planned question about same-sex spouses for the 2020 US Census, are needed to identify gay neighborhoods in population-level surveys.

The effect of maternal vitamin C intake on fetal telomere length.

BACKGROUND: A telomere is a nucleoprotein structure that is located at the end of a chromosome. Reduced telomere length manifests as physical ailments such as the increased risk of age-related illnesses. These age-related illnesses include heart disease and failure. Telomere length has been studied extensively in adults; however, limited information exists regarding maternal dietary influences on fetal telomere length. OBJECTIVES: The objective of this study is to investigate the relationship between maternal vitamin C intake and fetal telomere length. METHODS: Data for this analysis were collected as part of a prospective cohort study that recruited pregnant women upon admission into labor and delivery. Umbilical cord serum was collected for 96 maternal-fetal dyads, and DNA analysis was performed using a quantitative polymerase chain reaction. The telomere to single copy gene ratio method was used to determine telomere length, and maternal vitamin C intake was measured using the Dietary History Questionnaire (DHQ). Statistical analysis was conducted using generalized linear modeling-based analyses. RESULTS: The linear model indicates that maternal vitamin C intake (OR = 1.0032, 95%CI: 1.0014-1.0052, p ≤ .05) was positively associated with fetal telomere length. BMI (OR = 1.1096, 95%CI: 1.0619-1.1660, p ≤ .05) had a significant positive association with fetal telomere length while sodium intake was negatively associated with this outcome (OR = 0.9997, 95%CI: 0.9995-0.9998, p ≤ .05). Black ethnicity had a significant negative association with fetal telomere length (OR = 0.0186, 95%CI: 0.0031-0.0824, p ≤ .05). CONCLUSIONS: Our study shows a positive association between maternal vitamin C intake and fetal telomere length. These findings may provide a method of understanding and preventing adult-onset disease and mortality through intrauterine reprograming.

Estimating the size of HIV-negative MSM population that would benefit from pre-exposure prophylaxis in Florida.

PURPOSE: This study aimed to estimate the size of the population of men who have sex with men (MSM) in Florida with high-risk behaviors that would indicate eligibility for pre-exposure prophylaxis (PrEP) use. METHODS: Three methods were used to estimate the MSM population. Estimates from the three methods were averaged, and the number of MSM living with HIV in each zone improvement plan (ZIP) code was subtracted. RESULTS: The average MSM estimate was 1-2184 men (1.5-22.9%) by ZIP code. The size of the MSM population with indications for PrEP use was highest when using estimates of MSM with more than one sex partner in the past year obtained from the National HIV Behavioral Surveillance system and lowest when the MSM estimate was multiplied by 24.7% (percentage of MSM with PrEP indications from other studies). CONCLUSION: Areas with high numbers of MSM with PrEP indications could be targeted with information to reduce HIV acquisition.

Maternal Caffeine Consumption and Racial Disparities in Fetal Telomere Length.

BACKGROUND AND OBJECTIVES: The identification of risk factors for shorter telomere length, especially during fetal development, would be important towards caffeine consumption recommendations for pregnant women on a global scale. The purpose of this study was to evaluate the association between caffeine intake and fetal telomere length as well as racial/ethnic differences in telomere length regardless of maternal caffeine consumption status. METHODS: Caffeine intake was measured using a food frequency questionnaire (FFQ). Three generalized linear models (GLM) were compared based on binary categorical variables of caffeine levels using data mean value of 117.3 mg as cut-off; the World Health Organization (WHO) recommendations of 300 mg; and the American College of Obstetricians and Gynecologists (ACOG) recommendations of 200 mg. The association between caffeine consumption and telomere length (telomere to single-copy [T/S] ratio) was then assessed. RESULTS: Among 57 maternal-fetal dyads, 77.2% reported less than 200 mg of caffeine (ACOG) and 89.5% less than 300 mg (WHO). Both WHO and ACOG models found that caffeine intake was significantly and positively associated with longer telomere length (p<0.05); and sodium (p<0.05). Other" race (p<0.001) and "white" race (p<0.001) were also significantly and positively associated with longer telomere length in the same models. Increasing maternal age shortened telomere length significantly in all models (p<0.001). CONCLUSION AND GLOBAL HEALTH IMPLICATIONS: Caffeine intake, maternal age, and race may be associated with alterations in fetal telomere length. This indicates that caffeine consumption during pregnancy may have long-term implications for fetal development. The racial/ethnic differences in telomere length found in this study warrant larger studies to further confirm these associations.

Population-based methods for estimating the number of men who have sex with men: a systematic review.

The objective of this systematic review was to summarise population-based methods (i.e. methods that used representative data from populations) for estimating the population size of men who have sex with men (MSM), a high-risk group for HIV and other sexually transmissible infections (STIs). Studies using population-based methods to estimate the number or percentage of MSM or gay men were included. Twenty-eight studies met the inclusion criteria. Seven studies used surveillance data, 18 studies used survey data, and six studies used census data. Sixteen studies were conducted in the US, five were conducted in European countries, two were conducted in Canada, three were conducted in Australia, one was conducted in Israel, and one was conducted in Kenya. MSM accounted for 0.03-6.5% of men among all studies, and ranged from 3.8% to 6.4% in the US, from 7000 to 39100 in Canada, from 0.03% to 6.5% in European countries, and from 127947 to 182624 in Australia. Studies using surveillance data obtained the highest estimates of the MSM population size, whereas those using survey data obtained the lowest estimates. Studies also estimated the MSM population size by dimensions of sexual orientation. In studies examining these dimensions, fewer people identified as MSM than reported experience with or attraction to other men. Selection bias, differences in recall periods and sampling, or stigma could affect the estimate. It is important to have an estimate of the number of MSM to calculate disease rates, plan HIV and STI prevention efforts, and to allocate resources for this group.

Acceptance of Opt-Out HIV Screening in Outpatient Settings in the United States: A Systematic Review and Meta-Analysis.

OBJECTIVES: In the United States, about 15% of persons living with HIV infection do not know they are infected. Opt-out HIV screening aims to normalize HIV testing by performing an HIV test during routine medical care unless the patient declines. The primary objective of this systematic review and meta-analysis was to assess the acceptance of opt-out HIV screening in outpatient settings in the United States. METHODS: We searched in PubMed and CINAHL (Cumulative Index to Nursing and Allied Health Literature) for studies published from January 1, 2006, through December 31, 2018, of opt-out HIV screening in outpatient settings. We collected data from selected studies and calculated for each study (1) the percentage of persons who were offered HIV testing, (2) the percentage of persons who accepted the test, and (3) the percentage of new HIV diagnoses among persons tested. We also collected information on the reasons given by patients for opting out. The meta-analysis used a random-effects model to estimate the average percentages of HIV testing offered, HIV testing accepted, and new HIV diagnoses. RESULTS: We initially identified 6986 studies; the final analysis comprised 14 studies. Among the 8 studies that reported the size of the study population eligible for HIV screening, 71.4% (95% confidence interval [CI], 53.9%-89.0%) of the population was offered an HIV test on an opt-out basis. The test was accepted by 58.7% (95% CI, 47.2%-70.2%) of persons offered the test. Among 9 studies that reported data on new HIV diagnoses, 0.18% (95% CI, 0.08%-0.26%) of the persons tested had a new HIV diagnosis. Patients' most frequently cited reasons for refusal of HIV screening were that they perceived a low risk of having HIV or had previously been tested. CONCLUSIONS: The rates of offering and accepting an HIV test on an opt-out basis could be improved by addressing health system and patient-related factors. Setting a working target for these rates would be useful for measuring the success of opt-out HIV screening programs.

Racial/Ethnic Disparities in Failure to Initiate HIV Care: Role of HIV Testing Site, Individual Factors, and Neighborhood Factors, Florida, 2014-2015.

Delayed initiation of human immunodeficiency virus (HIV) care affects disease progression. To determine the role of HIV testing site and neighborhood- and individual-level factors in racial/ethnic disparities in initiation of care, we examined Florida population-based HIV/AIDS surveillance system records. We performed multilevel Poisson regression to calculate adjusted prevalence ratios (APR) for non-initiation of care by race/ethnicity adjusting for HIV testing site type and individual- and neighborhood-level characteristics. Of 8,913 people diagnosed with HIV during 2014-2015 in the final dataset, 18.3% were not in care within three months of diagnosis. The APR for non-initiation of care for non-Hispanic Blacks relative to non-Hispanic Whites was 1.57 (95% confidence interval [CI] 1.38-1.78) and for those tested in plasma/donation centers relative to outpatient clinics was 2.45 (95% CI 2.19-2.74). Testing site and individual variables contribute to racial/ethnic disparities in non-initiation of HIV care. Linkage procedures, particularly at plasma/blood donation centers, warrant improvement.

Using multilevel models to evaluate the influence of contextual factors on HIV/AIDS, sexually transmitted infections, and risky sexual behavior in sub-Saharan Africa: a systematic review.

PURPOSE: To describe the use of multilevel models (MLMs) in evaluating the influence of contextual factors on HIV/AIDS, sexually transmitted infections (STIs), and risky sexual behavior (RSB) in sub-Saharan Africa. METHODS: Ten databases were searched through May 29, 2016. Two reviewers completed screening and full-text review. Studies examining the influence of contextual factors on HIV/AIDS, STIs, and RSB and using MLMs for analysis were included. The Quality Assessment Tool for Quantitative Studies was used to evaluate study quality. RESULTS: A total of 118 studies met inclusion criteria. Seventy-four studies focused on HIV/AIDS-related topics; 46 focused on RSB. No studies related to STIs other than HIV/AIDS met the eligibility criteria. Of five studies examining HIV serostatus and community socioeconomic factors, three found an association between poverty and measures of inequality and increased HIV prevalence. Among studies examining RSB, associations were found with numerous contextual factors, including poverty, education, and gender norms. CONCLUSIONS: Studies using MLMs indicate that several contextual factors, including community measures of socioeconomic status and educational attainment, are associated with a number of outcomes related to HIV/AIDS and RSB. Future studies using MLMs should focus on contextual-level interventions to strengthen the evidence base for causality.

Role of Gay Neighborhood Status and Other Neighborhood Factors in Racial/Ethnic Disparities in Retention in Care and Viral Load Suppression Among Men Who Have Sex with Men, Florida, 2015.

This study's objective was to examine the role of gay neighborhood residence and other neighborhood factors in racial/ethnic disparities in retention in HIV care and viral load suppression during 2015. Florida residents diagnosed 2000-2014 with HIV infection and with transmission mode of men who have sex with men (MSM) were included in multi-level logistic regression models. Of 29,156 MSM, 29.4% were not retained and 34.2% were not virally suppressed. Non-Hispanic Blacks (NHB) had a higher likelihood of not being retained (adjusted prevalence ratio [aPR] 1.31, 95% confidence interval [CI] 1.24-1.38, p value < 0.0001) and not being virally suppressed (aPR 1.82, 95% CI 1.67-1.98, p value < 0.0001) compared with non-Hispanic Whites. Among NHBs, rural residence was protective for both outcomes. Although gay neighborhood residence was not associated with either outcome, the role of other neighborhood factors suggests that individual and neighborhood barriers to HIV care and treatment should be addressed among MSM.

Retention in HIV Care and Viral Suppression: Individual- and Neighborhood-Level Predictors of Racial/Ethnic Differences, Florida, 2015.

The objective of this study was to estimate racial/ethnic differences in retention in HIV care and viral suppression and to identify related individual and neighborhood determinants. Florida HIV surveillance records of cases aged ≥13 years diagnosed during the years 2000-2014 were analyzed. Retention in care was defined as evidence of ≥2 or more laboratory tests, receipts of prescription, or clinical visits at least 3 months apart during 2015. Viral load suppression was defined as a viral load of <200 copies/mL for the last test in 2015. Multi-level logistic regressions were used to estimate adjusted odds ratios (AORs). Of 65,735 cases, 33.3% were not retained in care, and 40.1% were not virally suppressed. After controlling for individual and neighborhood factors, blacks were at increased odds of nonretention in HIV care [AOR 1.29, 95% confidence interval (CI) 1.23-1.35] and nonviral suppression (AOR 1.55, 95% CI 1.48-1.63) compared with whites. Black and Latino males compared with their female counterparts had higher odds of nonretention and nonviral suppression. Compared with their US-born counterparts, foreign-born blacks and whites, but not Latinos, had higher odds of nonretention and nonviral suppression. Blacks and whites in urban compared with rural areas had higher odds of both outcomes. Disparities in retention in care and viral suppression persist and are not accounted for by differences in age, sex, transmission mode, AIDS diagnosis, neighborhood socioeconomic status, rural/urban residence, or neighborhood racial composition. Further, predictors of poor retention in care and viral suppression appear to differ by race/ethnicity.

Disparity in Retention in Care and Viral Suppression for Black Caribbean-Born Immigrants Living with HIV in Florida.

(1) The study aim was to assess disparities in non-retention in HIV care and non-viral suppression among non-Hispanic Black Caribbean immigrants living with HIV in Florida. (2) We analyzed cases involving individuals, aged ≥13, who met CDC HIV case definition during 2000-2014. Chi square test was used to evaluate differences in non-retention and non-viral suppression by country of origin/race/ethnicity. Multilevel logistic regressions with three referent groups [US-born Blacks, Hispanics, and non-Hispanic Whites (NHWs)] were used to estimate adjusted odds ratios (aOR). (3) Caribbean-born Blacks were less likely to be retained in care or be virally suppressed than US-born Blacks, Hispanics, and NHWs. Bahamians, Haitians, and Trinidadians and Tobagonians had increased odds of non-retention (aOR 3.13, 95% confidence interval [CI] 2.40 -4.10; aOR 1.52, 95% CI 1.40-1.66; aOR 2.30, 95% CI 1.38-3.83), and non-viral suppression (aOR 3.23, 95% CI 2.48-4.21; aOR 1.82, 95% CI 1.68-1.98; aOR 1.76, 95% CI 1.06-2.90) compared with NHWs. (4) Caribbean-born Blacks living with HIV infection are less likely than other racial/ethnic groups to be retained in care and/or achieve viral suppression. Further research is urgently needed to determine social, cultural, and biological factors that contribute to this disparity.

Black-White and Country of Birth Disparities in Retention in HIV Care and Viral Suppression among Latinos with HIV in Florida, 2015.

The study's purpose was to identify HIV, Black-White race, and birth country disparities in retention in HIV care and HIV viral load (VL) suppression among Latinos, in 2015. Florida's surveillance data for Latinos diagnosed with HIV (2000-2014) were merged with American Community Survey data. Multi-level (random effects) models were used to estimate adjusted odds ratios (aOR) for non-retention in care and non-viral load suppression. Blacks and Whites experienced similar odds of non-retention in care. Racial differences in VL suppression disappeared after controlling for neighborhood factors. Compared to U.S.-born Latinos, those born in Mexico (retention aOR 2.00, 95% CI 1.70-2.36; VL 1.85, 95% CI 1.57-2.17) and Central America (retention aOR 1.33, 95% CI 1.16-1.53; VL 1.28, 95% CI 1.12-2.47) were at an increased risk after controlling for individual and neighborhood factors. Among Central Americans, those born in Guatemala (retention aOR 2.39, 95% CI 1.80-3.18; VL 2.20, 95% CI 1.66-2.92) and Honduras (retention aOR 1.39, 95% CI 1.13-1.72; VL 1.42, 95% CI 1.16-1.74) experienced the largest disparities, when compared to U.S.-born Latinos. Disparities in care and treatment exist within the Latino population. Cultural and other factors, unique to Latino Black-White racial and birth country subgroups, should be further studied and considered for intervention.