Bringing doctors together

or

A free, private network for verified physicians

Leading voices

Dr. Raphael J. Landovitz

HIV/AIDS Medicine · UCLA

View

Dr. Aniruddha Hazra

HIV/AIDS Medicine · UCMC

View

Dr. Shital Patel

HIV/AIDS Medicine · BCM

View

Dr. Monica Gandhi

HIV/AIDS Medicine · UCSF

View

Dr. Ashwin Balagopal

HIV/AIDS Medicine · Johns Hopkins

View

Dr. Ofole Mgbako

HIV/AIDS Medicine · NYU Langone

View

Dr. Joshua Barocas

HIV/AIDS Medicine · CU Anschutz Medical Campus

View

Dr. Chris Beyrer

Global Health · Duke University

View

Dr. Keith Kaye

General ID · RU

View

Dr. Helen Boucher

Transplant/Immunocompromised States · Tufts

View
  • Discover your community

  • Stay current

  • A profile that represents you

  • Find your people

About RoonPrivacyTerms

    Overview

    Association of Cardiologist Clinic Visits With Cardiovascular Primary Prevention Outcomes Among People With HIV From Underrepresented Racial and Ethnic Groups in the Southern United States.

    Mar 7, 2025DOI: 10.1161/jaha.124.038462Journal ArticleAHA

    Written by: Matthew S DurstenfeldC Larry HillRobert M ClareKaren ChiswellGretchen SandersShamea GrayJoseph ViciniKeith MarsoloNwora Lance OkekeEric G MeissnerKevin L ThomasCaryn G MorseGerald S BloomfieldApril C PettitChris T Longenecker

    Find these authors on Roon

    LO
    Dr. Lance Okeke
    HIV/AIDS Medicine · Duke University

    Methods

    Anonymized data and materials will be made publicly available in the BioData Catalyst repository by June 2025.As previously described, the PATHWAYS (Pathways to Cardiovascular Disease Prevention and Impact of Specialty Referral in Under‐Represented Racial/Ethnic Minorities With HIV; NCT04025125) study is a multicenter collaborative observational electronic health record (EHR)‐based cohort study focused on PWH who are UREGs who receive clinical care in the southern United States.4 For this study, we used a series of nested cohort studies, as described in more detail in the Statistical Analysis section below.Four academic health centers located in the southern United States that participate in the Stakeholders, Technology, and Research Clinical Research Network were included in this study: Duke Health, the Medical University of South Carolina, Vanderbilt University Medical Center, and Wake Forest Baptist Health. EHR data were harmonized in the National Patient‐Centered Clinical Research Network, PCORnet Common Data Model.5 We included EHR data from January 1, 2014, to December 31, 2020, extracted and harmonized from 2021 to 2022.We included individuals aged 18 to 99 years with evidence of HIV documented in the EHR who were retained in HIV care at a participating site (defined by an HIV viral load laboratory test, a prescription for antiretroviral therapy, and a clinic visit with an HIV provider in the 12 months before their index date) and with race or ethnicity documented as Black or African American, American Indian or Alaska Native, Asian, Multiple Race, or Hispanic. We included those who had elevated ASCVD risk defined as a 10‐year risk ≥5% estimated by the pooled cohort equations or ≥7.5% by the Framingham risk score for those aged ≥40 years, and ≥40% lifetime risk by the pooled cohort equations for those aged <40 years. The index date for entry into the overall cohort was defined as the first clinic visit between 2015 and 2018 with elevated ASCVD risk. We excluded those with prior MACE including prior myocardial infarction or stroke (requiring secondary prevention), heart failure, and atrial fibrillation, or any prior clinic visit with a cardiologist within 1 of the participating centers with a lookback period to 2014. We also excluded those who were documented to have a MACE, heart failure, or atrial fibrillation at the time of their cardiologist visit. The Consolidated Standards of Reporting Trials diagram is included in Figure S1. Patients were followed for up to 5 years, until death, or until they were administratively censored at the end of EHR data availability (December 31, 2020).
    The exposure was an ambulatory clinic visit with a cardiologist, as defined previously on the basis of the provider specialty listed in the common data model and the provider's National Provider Identifier restricted to ambulatory clinic visits and excluding encounters for cardiac diagnostic testing alone,4 that occurred after the index date and before the end of follow‐up. The comparator group consisted of patients with at least 1 ambulatory clinic visit after the index date and before the end of follow‐up with at least 1 noncardiologist clinician. The selection of patients into the exposure and comparator groups is described under Statistical Methods. For the MACE outcome, we further stratified the exposure into cardiology visits for “prevention” and “management” on the basis of the International Classification of Diseases (ICD) codes documented at the cardiology visit. All ICD codes used during a cardiology visit were manually classified by 2 board‐certified cardiologists blinded to clinical data with disagreements resolved through consensus (Data S1 through S3). We classified visits which had ICD codes for cardiovascular diagnoses, abnormal cardiovascular testing, and cardiovascular symptoms (eg, chest pain), as management visits. We classified visits that only had ICD codes for noncardiovascular diagnoses, cardiovascular risk factors, and preoperative assessment as prevention visits.
    Our first 2 primary outcomes were blood pressure control (defined as systolic blood pressure <140 mm Hg and diastolic blood pressure <90 mm Hg) and lipid control defined as LDL cholesterol ≤100 mg/dL (2.6 mmol/L). For blood pressure we averaged available blood pressures within each given calendar month and used the last available month within 2 years of the cardiology visit or equivalent noncardiology visit to define control. For lipids we used the last available lipid panel after the cardiology visit or equivalent noncardiology visit within 2 years.
    The third primary outcome was incidence of MACE over 5 years defined as cardiovascular death, myocardial infarction, acute coronary syndrome, stroke including transient ischemic attack, peripheral arterial disease treated with revascularization with percutaneous or surgical bypass, and coronary artery disease treated with revascularization with percutaneous coronary intervention or coronary artery bypass grafting. Nonfatal outcomes were defined by hospitalization discharge ICD codes for diagnoses and Current Procedural Terminology codes for procedures and were not manually adjudicated due to lack of access to clinical notes. Cardiovascular death was defined on the basis of ICD codes for cardiovascular causes of death recorded in the National Death Index Plus.
    As secondary outcomes, we also considered the change in low‐density lipoprotein (LDL) and change in systolic blood pressure over 2 years as continuous measures assessed longitudinally among those with 2 years of follow‐up available. We also considered heart failure hospitalization, all‐cause death, and alternative definitions of MACE including MACE+heart failure hospitalization and MACE+noncardiovascular death.
    We included comorbid conditions and procedures recorded in the EHR as covariates using ICD codes and the Charlson comorbidity index (Table S1). We also included medication prescriptions for antiretroviral therapy and cardiovascular medications, systolic and diastolic blood pressures, body mass index, and laboratory results including lipid, creatinine (to estimate the glomerular filtration rate on the basis of the 2009 version of the Chronic Kidney Disease Epidemiology Collaboration equation without race), and hemoglobin A1c.
    Data were extracted from the EHRs of the participating sites and harmonized into a single data set. Data included participant‐level data (demographics) as well as billing code (ICD/Current Procedural Terminology) data, vital signs, and laboratory data. Mortality data, including cause of death, were obtained from the National Death Index Plus.
    To select the cardiology group and a noncardiology comparator group, we divided the study period (2015–2018) into eight 6‐month intervals. In each interval, we constructed a cohort study (Figure 1). For each cohort we identified living participants who met the eligibility criteria in that period and had not yet had a MACE event, or a prior cardiology visit. We considered those “exposed” if they had a clinic visit with a cardiologist during that period using the first cardiology visit date as the start of follow‐up time. We classified those without a cardiology encounter during that period as “unexposed” and started their follow‐up time at the last outpatient (noncardiology) visit date in that 6‐month period. Patients who were “unexposed” were eligible for inclusion in the subsequent cohort if they continued to meet eligibility criteria, thus patients could be counted in multiple cohorts. We then pooled the 8 sequential cohort studies into a single analysis, accounting for inclusion of the same patient in multiple cohorts.6 This pooled cohorts design allowed us to ensure that both exposed and unexposed patients were still eligible at the time of entry into the cohort; provided a mechanism to align the start of follow‐up between exposed and unexposed groups; and minimized immortal time bias that could result from assigning patients to groups after the start of follow‐up (eg, if follow‐up had been started before cardiology visit at the time when ASCVD risk was elevated).
    We used a sequential cohort design in which the study period was divided into 86‐mo intervals. Within each interval we constructed a cohort study of eligible patients, identifying patients with a cardiology clinic visit (exposed group), with remaining patients comprising the unexposed group, excluding those with major adverse cardiovascular event or a cardiology visit during the lookback period. For each cohort, follow‐up started at the time of the first cardiology visit in the interval (red circle), or the last noncardiology visit in the interval (blue square), thus minimizing the immortal time from the delay between eligibility and first cardiology encounter. As shown in the bottom panel, patients with a cardiology visit (exposed) were ineligible for subsequent cohorts, but patients without a cardiology visit (unexposed) could be included in the subsequent cohort if they continued to meet eligibility criteria (eg, primary prevention status).
    Baseline data were assessed for each of the 8 cohorts. Vital signs, LDL, and cardiovascular medications were assessed at/before the start of follow‐up within each cohort (up to 2 years prior for LDL and blood pressure and up to 13 months prior for medications). Other characteristics were assessed using the most recent available data up to the end date of the cohort for which that patient was included.
    For missing data on covariates and primary prevention status (no history of MACE), we carried forward the last available data based on the baseline date for each subcohort. Because we used mixed effects models for the longitudinal data, we did not impute missing outcomes (eg, LDL cholesterol).
    To estimate associations between cardiology encounters and subsequent lipid and blood pressure changes, we used hierarchical mixed effects models with random effects for patients, and patients nested within a cohort to account for the potential inclusion of a single patient in multiple cohorts as well as longitudinal measures for a patient within a cohort. For time‐to‐first‐MACE analyses, we used hierarchical Cox proportional hazards models using a 3‐level exposure variable (cardiovascular prevention encounter, cardiology management encounter, and no cardiology encounter) based on the first recorded cardiology encounter, censored participants at the time of first MACE, and used robust sandwich covariance estimates to account for potential contribution of the same participant across cohorts. Adjustment variables included site of care, age, sex, insurance, rural location, social deprivation index, HIV variables (cluster of differentiation 4 [CD4] count, viral suppression, antiretroviral therapy), hepatitis C, Charlson comorbidity index, estimated glomerular filtration rate, body mass index, diabetes, and smoking. For the lipid analyses, we additionally adjusted for baseline LDL cholesterol and lipid‐lowering medication use and the blood pressure analyses for baseline systolic blood pressure and antihypertensives. For MACE, we additionally adjusted for systolic blood pressure, LDL cholesterol, antihypertensive treatment, and lipid therapy. All statistical testing was 2‐sided, and nominal statistical significance was defined as a P value <0.05. No adjustment was made for multiple hypothesis testing. Statistical analyses were conducted using SAS version 9.4 (SAS Institute, Cary, NC).
    We hypothesized that cardiology encounters would be unlikely to have an impact on HIV viral suppression (defined as viral load <200 copies/mL) or change in CD4 counts. We also hypothesized that cardiology encounters would be unlikely to impact hospitalization or death from pneumonia or sepsis or from accident, suicide, or homicide. Therefore, we used falsification end points as negative controls to detect confounding by being unaffected by the exposure of interest but likely reflecting an intrinsic quality that affects the outcomes of interest.7
    The Duke University Health System institutional review board approved this study with a waiver of informed consent as the single institutional review board with the other sites relying on this institutional review board. The study results are reported in accordance with the Strengthening the Reporting of Observational Research guidelines.8

    Results

    We identified 3972 individuals who met our inclusion and exclusion criteria including 276 (6.9%) who had a cardiology clinic visit (Table 1). Approximately one third of the cohort were women, nearly all identified as Black or African American (95%), and 9% identified their ethnicity as Hispanic or Latinx. There were notable differences between those who did and did not have a cardiology clinic visit. For example, those with a cardiology clinic visit were more likely to be older, women, and have diabetes or have a history of an AIDS diagnosis, suggestive of more advanced HIV disease (lower nadir CD4 count, opportunistic infection, or opportunistic malignancy). They were also more likely to have a higher predicted ASCVD risk by the pooled cohort equations (median 9.0% over 10 years compared with median 6.6%) and higher median Charlson comorbidity index (5 versus 3). There were no differences in the proportion prescribed antiretroviral therapy, CD4 count, or the proportion virally suppressed.
    Participant Characteristics by Cardiology Clinic Visit
    BMI indicates body mass index; BP, blood pressure; CD4, cluster of differentiation 4; HRSA, Health Resources and Services Administration; IQR, interquartile range; LDL, low‐density lipoprotein; and MUSC, Medical University of South Carolina.
    Continuous variables are listed with median (IQR). Number included is listed in brackets. Categorical variables are listed as frequency (percentage).
    Baseline visit is defined as the first cardiology clinic visit or the last noncardiology clinic appointment within each 6‐mo cohort.
    Components are age, myocardial infarction, heart failure, peripheral vascular disease, cerebrovascular event including transient ischemic attack, dementia, chronic obstructive pulmonary disease, connective tissue disease, peptic ulcer disease, liver disease, diabetes, hemiplegia, moderate to severe chronic kidney disease, solid tumor, leukemia, lymphoma, and acquired immunodeficiency syndrome.
    Estimated glomerular filtration rate calculated from creatinine using 2009 version of the Chronic Kidney Disease Epidemiology Collaboration equation without race.
    For descriptive purposes, patients with a cardiology visit were included in the cardiology visit column, and their data are summarized using the cohort of their first cardiology visit. These patients are classified as preventative/management on the basis of diagnoses coded for their first cardiology visit. Remaining patients are included in the no cardiology visit column, and their data are summarized using the first cohort for which they were eligible.
    At baseline, there were no differences in LDL (median 100 versus 97.5 mg/dL), but a higher proportion were prescribed lipid‐lowering therapy at baseline among those with a future cardiology clinic visit (22% versus 12%). The proportion prescribed antihypertensives increased to 24% after a cardiology visit compared with 15% after the equivalent ambulatory visit. Cardiology clinic visits were not associated with subsequent lipid control over 2 years (55% versus 47%, adjusted odds ratio, 2.25 [95% CI, 0.72–7.0]; P=0.16; Figure 2), although with wide confidence intervals that do not exclude a possible benefit. Compared with those without a cardiology clinic visit, there was no significant difference in the change in LDL (3.7 mg/dL increase per year among those with a cardiology visit compared with 0.44 mg/dL increase per year among those without a cardiology encounter, or a between‐group difference of 3.3 mg/dL per year [95% CI, −3 to 10]; P=0.31; Table 2).
    Among people with HIV from underrepresented racial and ethnic groups with elevated cardiovascular risk and without diagnosed cardiovascular disease, cardiology clinic visits were not associated with improved lipid control (adjusted OR, 2.25 [95% CI, 0.72–7.01]; P=0.16), blood pressure control (adjusted OR, 0.78 [95% CI, 0.49–1.24]; P=0.29), or with the falsification end point of viral suppression, defined as either viral load <200 copies/mL or undetectable (adjusted OR, 0.34 [95% CI, 0.08–1.45]; P=0.15). BP indicates blood pressure; LDL‐C, low‐density lipoprotein cholesterol; and OR, odds ratio.
    Change in Blood Pressure, LDL Cholesterol, and CD4 Count per Year by Cardiology Clinic Visit
    CD4 indicates cluster of differentiation 4; and LDL, low density lipoprotein.
    Adjustment variables include age, sex, insurance (public, private, Health Resources and Services Administration, other, no insurance), rural location, social deprivation index, HIV variables (CD4 count, viral suppression, antiretroviral therapy), hepatitis C, Charlson comorbidity index, estimated glomerular filtration rate, body mass index, diabetes, smoking, systolic blood pressure, and baseline LDL cholesterol. Blood pressure analyses additionally adjust for antihypertensive treatment and lipid analyses additionally adjust for lipid‐lowering treatment at baseline.
    N=3376 unique patients (244 with cardiology encounter).
    N=2340 (159 with cardiology encounter) due to missing outcome assessments.
    N=3100 unique patients included in this analysis, N=229 with cardiology visit.
    At baseline, there were no differences in systolic blood pressure (median, 130 versus 129 mm Hg), but a higher proportion among those with a future cardiology clinic visit were prescribed antihypertensives at baseline (41% versus 28%). The proportion prescribed antihypertensives increased to 51% after a cardiology visit compared with 29% after the equivalent ambulatory visit. Cardiology clinic visits were not associated with subsequent blood pressure control over 2 years (66% versus 70%; adjusted odds ratio, 0.78 [95% CI, 0.49–1.24]; P=0.29; Figure 2). Compared with those without a cardiology clinic visit, those with a cardiology clinic visit did not subsequently have a lower increase in systolic blood pressure (0.7 mm Hg increase per year for patients with a cardiology visit, compared with an 0.2 mm Hg increase per year for those without a cardiology visit, for a between‐group difference of 0.5 mm Hg per year [95% CI, −0.4 to 1.4]; P=0.27; Table 2).
    Results were similar for our falsification end points of viral suppression (Figure 2) and CD4 count (Table 2) with no difference in the proportion with viral suppression or in the CD4 count over 2 years among those with a cardiology clinic visit compared with those without a cardiology clinic visit.
    Among 3972 included patients, 276 (6.9%) had a cardiology clinic visit, of which 154 (56%) were classified as preventive and 122 (44%) were classified as management encounters. Over a median 5 years of follow‐up, 237 individuals had a MACE event with 206 in the no cardiology clinic visit group and 31 in the cardiology clinic visit group (Figure 3). There were 2.5 incident MACE events per 100 patient‐years among those without a cardiology encounter (95% CI, 2.3–2.8), 7.2 with a prevention cardiology encounter (95% CI, 5.2–9.9), and 5.6 with a management cardiology encounter (95% CI, 3.8–8.4). The unadjusted risk of MACE was 2.5 times higher among those with a prevention cardiology encounter (hazard ratio [HR], 2.5 [95% CI, 1.6–3.9]) and 2.0 times higher for a management cardiology encounter (95% CI, 1.1–3.5) than in patients without a cardiology visit. After adjustment for potential measured confounders, the HRs were only slightly attenuated (adjusted HR, 1.74 [95% CI, 1.1–2.9] for prevention versus none; and adjusted HR, 1.76 [95% CI, 0.9–3.5] for management versus none).
    Kaplan–Meier plots of first MACE by cardiology visit with the number at risk at each time interval below (No Cardiology Clinic Visit; Cardiology Clinic Visit, Prevention; and Cardiology Clinic Visit, Management). MACE events occurred earlier among those with a cardiology clinic visit; those with a management visit had earlier occurrence of first MACE, largely due to higher rates of coronary revascularization (see “Coronary Events” in Figure 4). MACE indicates major adverse cardiovascular event.
    The risk of our falsification end points, which we hypothesized would not be impacted by a cardiologist clinic visit, were similarly higher among those with a cardiology clinic visit. The event rates per 100 patient‐years for hospitalization or death from accident, homicide, or suicide were 2.2 (95% CI, 2.1–2.4), 6.8 (95% CI, 4.8–9.6), and 4.0 (95% CI, 2.5–6.4) for those without a cardiology clinic visit, those with a preventive cardiology clinic visit, and those with a management cardiology clinic visit, respectively. Compared with those without a cardiology clinic visit, the adjusted HRs were 2.6 (95% CI, 1.7–3.9) for preventive and 1.8 (95% CI, 1.1–3.2) for management, compared with patients without a cardiology visit. The results for hospitalization or death from pneumonia or sepsis were nearly identical with event rates of 2.0 (95% CI, 1.9–2.1), 7.3 (95% CI, 5.2–10.2), and 3.7 (95% CI, 2.3–6.1), and adjusted hazard ratios of 2.6 (95% CI, 1.9–3.6) and 2.1 (95% CI, 1.4–3.4), respectively. Results were similar in sensitivity analyses including noncardiovascular death and heart failure in the MACE outcome (Table S2).
    In a descriptive analysis, we examined the cumulative incidence of events within 5 years (Table 3). Compared with those with a cardiology visit, the overall incident event rates were lower among those without a cardiology clinic visit with differences in the event types that occurred (Figure 4). Heart failure hospitalization had a similar cumulative incidence to noncardiovascular death and MACE among those with a cardiology clinic visit, regardless of whether it was a management or prevention visit but was less common among those without a cardiology clinic visit. Noncardiovascular death was highest among those with a cardiology prevention clinic visit and revascularization was more common among those with a cardiology management clinic visit.
    Incident Events by Cardiology Visit
    CD4 indicates cluster of differentiation 4; HF, heart failure; and MACE, major adverse cardiovascular event.
    MACE definition includes myocardial infarction, acute coronary syndrome, angina, percutaneous coronary intervention, coronary artery bypass graft surgery, ischemic stroke, transient ischemic attack, peripheral arterial disease with revascularization, and cardiovascular death.
    N=3972 unique patients included in unadjusted models, N=276 with cardiology visit. Unadjusted results account for patients in multiple cohorts but does not include demographic or clinical variables.
    N=3152 unique patients included in adjusted models, N=227 with cardiology visit. Adjusted results account for patients in multiple cohorts as well as controlling for age, sex, insurance, rurality, Social Deprivation Index, HIV variables (CD4 count, viral suppression, antiretroviral therapy use), hepatitis C, Charlson comorbidity index, estimated glomerular filtration rate, body mass index, systolic blood pressure, diabetes, smoking, low‐density lipoprotein cholesterol, lipid treatment.
    Bar charts of cumulative incidence of MACE events, HF hospitalization, and all‐cause death over 5 y of follow‐up (bars represent events per 100 patients over 5 y and lines represent 95% CI). These events are not mutually exclusive; a single patient could have a coronary event, cerebrovascular event, and heart failure hospitalization within 5 y, for example. Coronary event included myocardial infarction, hospitalization for acute coronary syndrome, percutaneous coronary intervention, and coronary artery bypass graft surgery; cerebrovascular event included stroke and transient ischemic attack, and peripheral vascular event includes peripheral vascular disease requiring revascularization. All outcomes were more common among individuals with a cardiology encounter. HF indicates heart failure; and MACE, major adverse cardiovascular event.

    Discussion

    Among a cohort of PWH from UREGs without known cardiovascular disease across 4 health systems in the southern United States, few patients had a “primary prevention” cardiology clinic visit before a major cardiovascular event or diagnosis; cardiology clinic visits were not associated with improvements in blood pressure or lipid control. Even accounting for higher baseline risk, those with a cardiology clinic visit had a higher subsequent risk of MACE, heart failure hospitalization, and death as well as our falsification end points than those without a cardiology clinic visit regardless of whether the first cardiology encounter was classified as a prevention or management visit.
    To our knowledge, there are no prior studies establishing whether cardiology encounters result in better lipid control and blood pressure control (primordial prevention) or prevent future cardiovascular events (primary prevention). Compared with people without HIV, PWH have worse cardiovascular health as operationalized by the American Heart Association's Life's Essential 8.9 Earlier control of risk factors or optimization of cardiovascular health is associated with longer morbidity‐free survival among PWH.10 Although the American Society for Preventive Cardiology has established clinical practice guidelines for primordial and primary prevention including the role of the preventive cardiologist to promote health equity,11 data are lacking on the association between cardiology clinic visits and cardiovascular outcomes among patients at risk for cardiovascular disease.
    In this study, we found that a cardiologist encounter alone is not strongly associated with achievement of lipid control, defined as a LDL cholesterol <100 mg/dL, or blood pressure control, defined as <140/90 mm Hg. Those with a cardiology clinic visit within our study had higher baseline rates of antihypertensive and lipid‐lowering prescriptions such that there were no differences in LDL or blood pressure at baseline. We expected that cardiologists would be more aggressive about prescribing lipid‐lowering therapy, and that statin prescriptions would be a specific mechanism by which a single cardiology clinic visit could prevent cardiovascular events. Unexpectedly, the proportion prescribed lipid‐lowering therapy was nearly identical after the cardiology encounter, while there was a modest increase in antihypertensive prescriptions, with no significant improvements in mean LDL or blood pressure or the proportion controlled, although the wide confidence intervals for LDL do not exclude a possible clinically significant benefit. Perhaps cardiologists were reluctant to prescribe statins due to concerns about drug–drug interactions with older antiretroviral therapy regimens, did not think statins were appropriate without clinical trial data to support their use among PWH, or did not feel that it was their role to prescribe them for primary prevention. Our qualitative research in the same population suggests that individual and structural stigma may impact the effectiveness of cardiology specialty care.12
    It is important to note this study was conducted before completion of the REPRIEVE (Randomized Trial to Prevent Vascular Events in HIV), which demonstrated that pitavastatin reduced cardiovascular events among PWH at elevated cardiovascular risk compared with placebo,13 so it may be that cardiologists (as well as HIV and primary care physicians) would now be more aggressive about initiation of lipid‐lowering therapy. Guidelines have now shifted toward considering statins for all PWH aged >40 years and recommending them for all with elevated cardiovascular risk.3, 14
    Our findings that cardiology clinic visits were associated with more than double the risk of MACEs, even controlling for measured confounders, were unexpected. It is possible that referring providers are identifying and selectively referring those at higher risk of cardiovascular disease and overall death in ways that are not adequately captured in the EHR and thus could not be adequately accounted for in our adjusted models. A hypothetical patient vignette that may help explain how our findings diverged from our hypothesis is a patient with worsening chronic stable angina being managed by a primary care physician who subsequently sees a cardiologist and is referred for revascularization; such a patient would have a shorter time to a revascularization event in the cohort in which they saw a cardiologist. However, similar findings for all‐cause death and most strikingly for our falsification end points suggest that the patients seen by cardiologists are at higher overall baseline risk. Another alternative explanation is that the cardiologist clinic visits could increase the risk of MACEs, particularly via referral for procedures that have risk, but we would not expect the cardiology encounter would increase the risk of our falsification end points, so this explanation is less likely.
    Our study highlights how there is equipoise regarding interventions and strategies to promote cardiovascular health and equity. It is challenging to adequately answer these causal questions with observational EHR‐based research. One recent randomized clinical trial, EXTRA‐CVD (A Nurse‐Led Intervention to Extend the HIV Treatment Cascade for Cardiovascular Disease Prevention), of an implementation strategy for a nurse‐led care coordination program that includes an EHR tool, home blood pressure monitoring, and an evidence‐based treatment algorithm, improved blood pressure and lipid control over 12 months, something that we were not able to demonstrate in this observational study.15 Several features of EXTRA‐CVD that may have contributed to its success were that the intervention was a multidisciplinary, nurse‐led approach with longitudinal relationships aided by integrated EHR tools. Finally, by randomizing individuals, EXTRA‐CVD was able to ensure greater exchangeability between those who did and did not get the intervention to assess the causal impact of the intervention.
    This is an observational comparative effectiveness study with limitations inherent to the study design. The use of EHRs is a reasonable approach to study the question of whether cardiology clinic visits (which are easily measurable in the EHR) are associated with process and outcome measures, including blood pressure and lipids, which are often measured within the EHR, but it still comes with serious drawbacks. For example, capturing the factors that lead to a cardiology referral without access to the clinician and patient interaction (or even the documentation of the referral in the EHR) makes the risk of confounding by indication challenging to overcome, and even our qualitative work in this patient population only captures the perspectives of those who completed their cardiology referral.12 Nonetheless, we tried to stratify our exposure variable by the diagnoses coded by the cardiologist, which may have been misclassified and depends on the coding strategy of the cardiologists and was likely subject to misclassification. We did not have data on referral to cardiology to identify time from referral to cardiology clinic visit, to classify those who were referred but did not see a cardiologist as exposed, or to identify those who had MACE events or died while waiting to see a cardiologist. In our previous work in this cohort, we observed a long delay between eligibility (onset of elevated cardiovascular risk) and cardiology encounter (≈2 years),4 so we had to develop an analytic strategy to prevent immortal time bias in our outcomes analysis. Our use of a sequential cohort design and using equivalent noncardiology ambulatory encounters partially addresses the issues of immortal time bias but does not address the concern of those who had MACE events occurring between cardiology referral, which was unmeasured, and scheduled cardiology clinic visit. Because the factors leading to cardiology encounters were not well captured, we could not accurately develop a model to predict propensity for cardiology encounters for a matching‐based strategy to maximize exchangeability between groups. Because our study population is limited to PWH who identify as UREGs who live in the South, these findings may not be externally generalizable to all PWH, people without HIV, or people in other regions of the country where discrimination or other structural factors attenuate the efficacy of a cardiology clinic visit in ways specific to the included patient population.

    Conclusions

    In conclusion, our study demonstrates that for PWH who identify as members of UREGs, a single cardiology clinic visit is not strongly associated with improved lipid or blood pressure control. Those who have a cardiology clinic visit have higher risk of adverse outcomes including MACE, death, and noncardiovascular outcomes. Whether prospective referral of PWH at elevated cardiovascular risk would result in optimization of cardiovascular health and prevention of cardiovascular events cannot be determined from our observational study and is an important area for further research, but our findings suggest that a single cardiology encounter may not have a large effect on cardiovascular health in this population.

    Sources of Funding

    Research reported in this publication was supported by the National Institute on Minority Health and Health Disparities (R01MD013493, PI: Bloomfield); the National Institutes of Mental Health (R01MH113438, PI: Pettit); the National Institute of General Medical Sciences (P20GM130457, PI: Meissner); the National Insitute of Allergy and Infectious Diseases to the Tennessee Center for AIDS Research (P30AI110527, PI: Pettit); and powered by PCORnet. PCORnet has been developed with funding from the Patient‐Centered Outcomes Research Institute. Dr. Durstenfeld and Dr. Okeke are funded by the National Heart, Lung, and Blood Institute grants K23HL172699 and K23HL137611, respectively. Its contents are solely the responsibility of the authors and do not necessarily represent the official views of the National Institutes of Health or views of organizations participating in, collaborating with, or funding PCORnet or of the Patient‐Centered Outcomes Research Institute.

    Disclosures

    Dr Longenecker is on the advisory board for Theratechnologies outside the scope of this work. Dr Durstenfeld received consulting fees from Merck outside the scope of this work. The remaining authors have no disclosures to report.

    Supporting information

    Data S1
    Tables S1–S2
    Figure S1
    LODr. Lance Okeke
    Leading Voice

    Dr. Lance Okeke

    Infectious Disease · HIV & AIDS Medicine
    Duke University
    nwora