ACCOUNTING FOR DEPENDENT ERRORS IN PREDICTORS AND TIME-TO-EVENT OUTCOMES USING ELECTRONIC HEALTH RECORDS, VALIDATION SAMPLES, AND MULTIPLE IMPUTATION.

Data from electronic health records (EHR) are prone to errors, which are often correlated across multiple variables. The error structure is further complicated when analysis variables are derived as functions of two or more error-prone variables. Such errors can substantially impact estimates, yet we are unaware of methods that simultaneously account for errors in covariates and time-to-event outcomes. Using EHR data from 4217 patients, the hazard ratio for an AIDS-defining event associated with a 100 cell/mm3 increase in CD4 count at ART initiation was 0.74 (95%CI: 0.68-0.80) using unvalidated data and 0.60 (95%CI: 0.53-0.68) using fully validated data. Our goal is to obtain unbiased and efficient estimates after validating a random subset of records. We propose fitting discrete failure time models to the validated subsample and then multiply imputing values for unvalidated records. We demonstrate how this approach simultaneously addresses dependent errors in predictors, time-to-event outcomes, and inclusion criteria. Using the fully validated dataset as a gold standard, we compare the mean squared error of our estimates with those from the unvalidated dataset and the corresponding subsample-only dataset for various subsample sizes. By incorporating reasonably sized validated subsamples and appropriate imputation models, our approach had improved estimation over both the naive analysis and the analysis using only the validation subsample.

Substance Use, Demographic and Socioeconomic Factors Are Independently Associated With Postpartum HIV Care Engagement in the Southern United States, 1999-2016.

BACKGROUND: Retention in care (RIC) and viral suppression (VS) are associated with reduced HIV transmission and mortality. Studies addressing postpartum engagement in HIV care have been limited by small sample size, short follow-up, and a lack of data from the Southeast United States. METHODS: HIV-positive adult women with ≥1 prenatal visit at the Vanderbilt Obstetrics Comprehensive Care Clinic from 1999 to 2015 were included. Poor RIC was defined as not having ≥2 encounters per year, ≥90 days apart; poor VS was a viral load >200 copies/mL. Modified Poisson regression was used to estimate adjusted relative risks (aRRs) of poor postpartum RIC and VS. RESULTS: Among 248 women over 2070 person-years of follow-up, 37.6% person-years had poor RIC and 50.4% lacked VS. Prenatal substance use was independently associated with poor RIC (aRR, 1.40; 95% confidence interval [CI], 1.08-1.80) and poor VS (aRR, 1.20; 95% CI, 1.04-1.38), and lack of VS at enrollment was associated with poor RIC (aRR, 1.64; 95% CI, 1.15-2.35) and poor VS (aRR, 1.59; 95% CI, 1.30-1.94). Hispanic women were less likely and women with lower educational attainment were more likely to have poor RIC. Women >30 years of age and married women were less likely to have poor VS. CONCLUSIONS: In this population of women in prenatal care at an HIV primary medical home in Tennessee, women with prenatal substance use and a lack of VS at enrollment into prenatal care were at greater risk of poor RIC and lack of VS postpartum. Interventions aimed at improving postpartum engagement in HIV care among these high-risk groups are needed.

The relationship between adverse neighborhood socioeconomic context and HIV continuum of care outcomes in a diverse HIV clinic cohort in the Southern United States.

Retention in care and viral suppression are critical to delaying HIV progression and reducing transmission. Neighborhood socioeconomic context (NSEC) may affect HIV care receipt. We therefore assessed NSEC's impact on retention and viral suppression in a diverse HIV clinical cohort. HIV-positive adults with ≥1 visit at the Vanderbilt Comprehensive Care Clinic and 5-digit ZIP code tabulation area (ZCTA) information between 2008 and 2012 contributed. NSEC z-score indices used neighborhood-level socioeconomic indicators for poverty, education, labor-force participation, proportion of males, median age, and proportion of residents of black race by ZCTA. Retention was defined as ≥2 HIV care visits per calendar year, >90 days apart. Viral suppression was defined as an HIV-1 RNA <200 copies/mL at last measurement per calendar year. Modified Poisson regression was used to estimate risk ratios (RR) and 95% confidence intervals (CI). Among 2272 and 2541 adults included for retention and viral suppression analyses, respectively, median age and CD4 count at enrollment were approximately 38 (1st and 3rd quartile: 30, 44) years and 351 (176, 540) cells/µL, respectively, while 24% were female, and 39% were black. Across 243 ZCTAs, median NSEC z-score was 0.09 (-0.66, 0.48). Overall, 79% of person-time contributed was retained and 74% was virally suppressed. In adjusted models, NSEC was not associated with retention, though being in the 4th vs. 1st NSEC quartile was associated with lack of viral suppression (RR = 0.88; 95% CI: 0.80-0.97). Residing in the most adverse NSEC was associated with lack of viral suppression. Future studies are needed to confirm this finding.

Health Literacy and Demographic Disparities in HIV Care Continuum Outcomes.

Studies evaluating the association between human immunodeficiency virus (HIV) infection continuum of care outcomes [antiretroviral (ART) adherence, retention in care, viral suppression] and health literacy have yielded conflicting results. Moreover, studies from the southern United States, a region of the country disproportionately affected by the HIV epidemic and low health literacy, are lacking. We conducted an observational cohort study among 575 people living with HIV (PLWH) at the Vanderbilt Comprehensive Care Clinic (Nashville, Tennessee). Health literacy was measured using the brief health literacy screen, a short tool which can be administered verbally by trained clinical personnel. Low health literacy was associated with a lack of viral suppression, but not with poor ART adherence or poor retention. Age and racial disparities in continuum of care outcomes persisted after accounting for health literacy, suggesting that factors in addition to health literacy must be addressed in order to improve outcomes for PLWH.

Trends in HIV Continuum of Care Outcomes over Ten Years of Follow-Up at a Large HIV Primary Medical Home in the Southeastern United States.

Longitudinal studies of retention in care (RIC) and viral suppression (VS) in the southeastern United States (US), a region disproportionately affected by HIV infection, are lacking. HIV-infected adults with ≥1 medical visit at the Vanderbilt Comprehensive Care Clinic (Nashville, Tennessee) from 2004 to 2013 were included. RIC was ≥2 (a) laboratory dates [CD4+ counts or HIV-1 viral loads (VLs)] or (b) provider encounters and/or laboratory dates in the year of interest, ≥90 days apart. VS was a VL of <200 copies/ml at last measurement in the year of interest. Modified Poisson regression estimated relative risk (RR) of RIC and VS, adjusting for age, race, sex, HIV transmission risk, and socioeconomic status (SES). Among 4,641 persons, 76.8% achieved RIC and 70.2% achieved VS. RIC and VS increased from 2004 to 2013 (p < .001 each). For lack of RIC, younger patients (RR = 1.2 and RR = 1.1, 18-24 and 25-34 vs. 35-44 year-olds, respectively), Blacks (RR = 1.3 vs. Whites), and injection drug users (IDUs) (RR = 1.2 vs. heterosexual contact [Hetero]) fared worse (p < .05 each); those with male-to-male sexual contact fared better (RR = 0.8 vs. Hetero, p < .05). For lack of VS, younger patients (RR = 1.3 and RR = 1.2, 18-24 and 25-34 vs. 35-44 year olds, respectively), Blacks (RR 1.3 vs. Whites), Females (RR = 1.1 vs. Males), IDUs (RR 1.3 vs. Hetero), and those with low SES (RR = 1.1 vs. not low SES) fared worse (p < .05, each). RIC and VS increased over time, suggesting that efforts to improve outcomes have been effective. However, disparities persist and resources should focus on groups most at risk.

Hormonal Contraception and Risk of Psychiatric and Other Noncommunicable Diseases in HIV-Infected Women.

BACKGROUND: Hormonal contraception use is common among human immunodeficiency virus (HIV)-infected women. Risk of psychiatric and other noninfectious complications of hormonal contraception use has not been described in this population. METHODS: We performed a retrospective cohort study of HIV-infected women receiving care in Tennessee from 1998 to 2008 to examine the risks of incident psychiatric and other noncommunicable diseases (NCDs), including cardiovascular, hepatic, renal, and malignant diseases, and hormonal contraception use, including depot medroxyprogesterone acetate (DMPA) and combined estrogen- and progestin-containing hormonal contraceptives. We used marginal structural models with inverse probability weights to account for time-varying confounders associated with hormonal contraception use. RESULTS: Of the 392 women included, 94 (24%) used hormonal contraception during the study period. Baseline psychiatric disease was similar between women who received and did not receive hormonal contraception. There were 69 incident psychiatric diagnoses and 72 NCDs. Only time-varying DMPA use was associated with increased risk of psychiatric disease (adjusted odds ratio [aOR] 3.70; 95% confidence interval [95% CI] 1.32-10.4) and mood disorders, specifically (aOR 4.70 [1.87-11.8]). Time-varying and cumulative combined hormonal contraception use were not statistically associated with other NCDs (aOR 1.64, 95% CI 0.64-4.12 and aOR 1.16, 95% CI 0.86-1.56, respectively). However, risk of incident NCDs was increased with cumulative DMPA exposure (per year exposure aOR 1.45, 95% CI 1.01-2.08). CONCLUSIONS: Among HIV-infected women, DMPA was associated with risk of incident psychiatric diseases, particularly mood disorders, during periods of use. Cumulative DMPA exposure was also associated with risk of other NCDs. However, combined estrogen and progestin-containing hormonal contraception use was not statistically associated with risk of any NCDs.

HIV and cancer: a comparative retrospective study of Brazilian and U.S. clinical cohorts.

BACKGROUND: With successful antiretroviral therapy, non-communicable diseases, including malignancies, are increasingly contributing to morbidity and mortality among HIV-infected persons. The epidemiology of AIDS-defining cancers (ADCs) and non-AIDS-defining cancers (NADCs) in HIV-infected populations in Brazil has not been well described. It is not known if cancer trends in HIV-infected populations in Brazil are similar to those of other countries where antiretroviral therapy is also widely available. METHODS: We performed a retrospective analysis of clinical cohorts at Instituto Nacional de Infectologia Evandro Chagas (INI) in Rio de Janeiro and Vanderbilt Comprehensive Care Clinic (VCCC) in Nashville from 1998 to 2010. We used Poisson regression and standardized incidence ratios (SIRs) to examine incidence trends. Clinical and demographic predictors of ADCs and NADCs were examined using Cox proportional hazards models. RESULTS: This study included 2,925 patients at INI and 3,927 patients at VCCC. There were 57 ADCs at INI (65% Kaposi sarcoma), 47 at VCCC (40% Kaposi sarcoma), 45 NADCs at INI, and 82 at VCCC. From 1998 to 2004, incidence of ADCs remained statistically unchanged at both sites. From 2005 to 2010, ADC incidence decreased in both cohorts (INI incidence rate ratio per year = 0.74, p < 0.01; VCCC = 0.75, p < 0.01). Overall Kaposi sarcoma incidence was greater at INI than VCCC (3.0 vs. 1.2 cases per 1,000 person-years, p < 0.01). Incidence of NADCs remained constant throughout the study period (overall INI incidence 3.6 per 1,000 person-years and VCCC incidence 5.3 per 1,000 person-years). Compared to general populations, overall risk of NADCs was increased at both sites (INI SIR = 1.4 [95% CI 1.1-1.9] and VCCC SIR = 1.3 [1.0-1.7]). After non-melanoma skin cancers, the most frequent NADCs were anal cancer at INI (n = 7) and lung cancer at VCCC (n = 11). In multivariate models, risk of ADC was associated with male sex and immunosuppression. Risk of NADC was associated with increased age. CONCLUSIONS: In both cohorts, ADCs have decreased over time, though incidence of KS was higher at INI than VCCC. Rates of NADCs remained constant over time at both sites.

Estimating the optimal CD4 count for HIV-infected persons to start antiretroviral therapy.

BACKGROUND: Optimal timing of antiretroviral therapy in HIV-infected persons is unclear, although 2 recent large observational studies have improved our understanding of the best CD4 threshold for initiation. These studies compared the effect of starting HAART on mortality and mortality/AIDS between strata defined using broad ranges of CD4 counts. We sought to expand this understanding using a novel statistical approach proposed by Robins et al. METHODS: Using observational data from 1034 antiretroviral-naive HIV-infected patients from Nashville, Tennessee, we directly estimated the optimal CD4 count for initiation of HAART to maximize patient health 6, 12, 24, and 36 months after the first instance of CD4 falling below 750. We measured health using 2 outcome metrics, one based on CD4 counts at the end of follow-up and the other based on a published quality-of-life scale; both metrics incorporated death, AIDS-defining events, serious non-AIDS events, and CD4 at the end of follow-up, if asymptomatic. RESULTS: The CD4-based metric estimated that to maximize health 6, 12, 24, and 36 months after study entry, HAART should be initiated within 3 months of CD4 first dropping below 495 (95% confidence interval [CI] = 468-522), 554 (459-750), 489 (427-750), and 509 (460-750), respectively. The quality-of-life-based metric produced CD4 initiation threshold estimates of 337 (95% CI = 201-442), 354 (288-386), 358 (294-750), and 475 (287-750) for the same time points. CONCLUSIONS: Our results support early initiation of antiretroviral therapy, although the criterion for starting therapy depends on the choice of health outcome.

Postpartum discontinuation of antiretroviral therapy and risk of maternal AIDS-defining events, non-AIDS-defining events, and mortality among a cohort of HIV-1-infected women in the United States.

This retrospective cohort study of HIV-infected women receiving highly active antiretroviral therapy (HAART) while pregnant assessed the effect of postpartum HAART discontinuation on maternal AIDS-defining events (ADEs), non-AIDS-defining events (non-ADEs), and death 1997-2008 in Nashville, Tennessee. Cox proportional hazards models compared rates of ADE or all-cause death and non-ADE or all-cause death, and competing risks analyses compared rates of ADE or ADE-related death and non-ADE or non-ADE-related death across the groups. There were two groups: women who stopped HAART postpartum (discontinuation, n = 54) and women who continued HAART postpartum (continuation, n = 69). Fifty percent were African American, 40% had prior non-HAART antiretroviral therapy (ART) use, and 38% had a history of illicit drug use. Median age was 27.5 years, baseline CD4(%) was 532 (34%) and CD4 nadir was 332 cells/mm(3), baseline and peak HIV-1 RNA were 2.6 and 4.32 log(10) copies per milliliter, respectively. Women in the continuation group were older, had lower baseline CD4, CD4%, and CD4 nadir, and had higher peak HIV-1 RNA. In multivariable proportional hazards models, the hazard ratios [95% confidence interval (CI)] of ADE or all-cause death and non-ADE or all-cause death were lower in the continuation group, but not statistically significantly: 0.50 (0.12, 2.12; p = 0.35) and 0.69 (0.24, 1.95; p = 0.48), respectively. The results were similar in competing risks analyses. Despite having characteristics associated with worse prognosis, women who continued HAART postpartum had lower hazard ratio point estimates for ADEs or death and non-ADEs or death than women who discontinued HAART. Larger studies with longer follow-up are indicated to assess this association.

Increased detectability of plasma HIV-1 RNA after introduction of a new assay and altered specimen-processing procedures.

After changes to assay and specimen-processing methods, plasma human immunodeficiency virus type 1 (HIV-1) RNA was frequently detectable in patients who previously had well-suppressed HIV-1 RNA levels. This artifact is attributable to shipping frozen plasma in primary plasma preparation tubes and is not caused by the HIV-1 RNA detection assay; it can be avoided by shipping plasma in a secondary tube.