Racial, Ethnic, and Language Differences in Screening Measures for Statin Therapy Following a Major Guideline Change

Abstract

Atherosclerotic cardiovascular disease (ASCVD) disproportionally affects racial and ethnic minority populations. Statin prescribing guidelines changed in 2013 to improve ASCVD prevention. It is unknown whether risk screening for statin eligibility differed across race and ethnicity over this guideline change. We examine racial/ethnic/language differences in screening measure prevalence for period-specific statin consideration using a retrospective cohort design and linked electronic health records from 635 community health centers in 24 U.S. states. Adults 50+ years, without known ASCVD, and ≥1 visit in 2009-2013 and/or 2014-2018 were included, grouped as: Asian, Latino, Black, or White further distinguished by language preference. Outcomes included screening measure prevalence for statin consideration, 2009-2013: low-density lipoprotein (LDL), 2014-2018: pooled cohort equation (PCE) components age, sex, race, systolic blood pressure, total cholesterol, high-density lipoprotein, smoking status. Among patients seen both periods, change in period-specific measure prevalence was assessed. Adjusting for sociodemographic and clinical factors, compared to English-preferring White patients, all other groups were more likely to have LDL documented (2009-2013, n=195,061) and all PCE components documented (2014-2018, n=344,504). Among patients seen in both periods (n=128,621), all groups had lower odds of PCE components versus LDL documented in the measures’ respective period; English-preferring Black adults experienced a greater decline compared to English-preferring White adults (OR 0.81; 95% CI: 0.72-0.91).

Racial/ethnic/language disparities in documented screening measures that guide statin therapy for ASCVD prevention were unaffected by a major guideline change advising this practice. It is important to understand whether the newer guidelines have altered disparate prescribing and morbidity/mortality for this disease.

Introduction

Cardiovascular disease morbidity and mortality disproportionally affects racial and ethnic minority populations.^1-6 Risk factors for cardiovascular disease, including hypertension, obesity, elevated hemoglobin A1c, and hyperlipidemia, are also more prevalent among non-white versus white adults.^7-11 Less clear is whether risk screening for initiating HMG-CoA reductase inhibitors (statins) for primary prevention of atherosclerotic cardiovascular disease (ASCVD) differs across race and ethnicity.

Standards of care for ASCVD shifted in 2013 when the American College of Cardiology and American Heart Association released updated guidelines for risk screening and preventive pharmacotherapy (hereafter referred to as the 2013 guidelines)¹² The new guidelines aimed to improve individual risk stratification by using the pooled cohort equations (PCE). The PCE estimate a 10-year risk of cardiovascular event score based on high-density lipoprotein and total cholesterol levels, systolic blood pressure, age, sex, race, smoking status, presence of diabetes, and use of anti-hypertensive medication for those without current ASCVD. Statins have been first-line pharmacotherapy for ASCVD prevention since before the 2013 guidelines.¹² A major change in the 2013 guidelines, however, recommends use of the PCE 10-year risk score instead of low-density lipoprotein (LDL) to guide the clinical consideration to initiate statins for primary preventive care. This shift from an LDL treatment-target approach to one based on individual risk was expected to result in greater ASCVD prevention broadly, and reduced ASCVD disparities specifically.^12-15

Since the early 2000s, the ASCVD risk screening literature has been predominantly survey-based and has produced mixed results as to which groups, if any, experience disproportionate screening.^4,16-23 Some evidence suggests that socioeconomic status, preferred spoken language, and insurance status may explain screening disparities for some racial/ethnic minority groups.¹⁹ One study utilizing 2009-2013 electronic health records (EHR) data reported no racial/ethnic differences in lipid screening among low-income obese patients, however the overall screening rates were low among this higher risk population.²⁴ Moreover, at least one study examining the year before and after the 2013 guidelines suggests disparities in the use of statins persist.²⁵ To our knowledge there are no studies examining whether racial/ethnic differences exist in the prevalence of EHR data to guide statin initiation. Understanding this clinical-care step in the primary prevention of ASCVD is necessary to inform future interventions aimed to eliminate downstream disparities in cardiovascular disease outcomes.

To fill these gaps, we conducted a 10-year investigation spanning the 2013 guidelines’ release utilizing objective cardiovascular clinical-care and social determinants of health (SDH) measures. Importantly, we conducted this investigation in the primary care setting of community health centers (CHC) - clinics in which racial-, ethnic-, and language-minority persons, especially low-income individuals, receive a disproportionate amount of their care.^26-28 Utilizing 2009-2018 EHR data from a large national network of CHCs, we examined the presence of documented ASCVD risk measures used to initiate consideration of statin pharmacotherapy. For the years 2009-2013, we assessed documentation of LDL, and for 2014-2018, documentation of all components required to use the PCE, each (LDL or PCE) being the driving criteria for statin initiation within the specific observation periods.^12,14,15 We examined these measures in several racial/ethnic groups each disaggregated by language preference. As language may help explain some racial/ethnic disparities in ASCVD risk screening¹⁹, we hypothesized that all non-English preferring groups would have lower prevalence of these data in their EHR when compared with non-Hispanic White English speakers, within both observation periods and adjusted for important confounders. Further, because CHCs may help in reducing racial/ethnic disparities^28-30, we hypothesized equal prevalence of screening data for all English preferring groups regardless of race or ethnicity.

Methods

Data Source.

We utilized patient-level data from EHRs in the Accelerating Data Value Across a National Community Health Center (ADVANCE) Clinical Research Network (CRN), from 635 community health centers in 24 US states within the years 2009-2018. These data were part of a larger study of adults who reach a minimum age of 60 years by study end, thus the sample includes patients 50 years and older at study start.

Population.

Adults age 50 and older with at least one primary care encounter in the ADVANCE CRN 2009-2013 were included in the LDL outcome analysis, while those with an encounter in 2014-2018 were included in the PCE measures analysis. Adults seen in both outcome-specific study periods comprised a subsample analysis evaluating the effect of the 2013 guideline change on documented screenings. We defined primary care encounters as those occurring with a physician (Allopath, Osteopath, or Naturopath), physician assistant, or nurse practitioner utilizing Current Procedural Terminology codes 99201-99205, 99212-99215, 99241-99245, 99386-99387 and 99396-99397.

Exclusion Criteria.

As this study investigated the presence of data to inform primary preventive care for ASCVD, patients with known ASCVD, that is documented fatal/non-fatal myocardial or cerebrovascular infarction, prior to study start were excluded. This criteria affected our exposure groups equally, with 0.1% to 0.2% of eligible patients excluded from each group. Moreover, if the ASCVD event occurred during the study period we utilized that patient’s data up to the date of the event. This censoring was chosen to follow the 2013 guideline as for whom a 10-year ASCVD risk score should be calculated and applied to <2% of the study population. Further, approximately 6% of the eligible LDL and 9% of the PCE measures populations were missing either race, ethnicity, or sex data in the EHR or identified with a race too small in sample size to include in these analyses; these patients were excluded from the study.

Dependent Variable.

The primary outcome is a binary indicator denoting presence/absence of the time-period specific ASCVD risk screening data. For the 2009-2013 period the outcome measure was presence/absence of a documented LDL measurement. For the 2014-2018 period the outcome measure indicated whether all data values required for the PCE were present in the EHR (age, sex, race, systolic blood pressure, total cholesterol, high-density lipoprotein, and smoking status). Of note, the PCE risk score itself is not a data field in the EHR utilized for this study. We did not measure whether patients were actually screened for ASCVD, but whether patients had the necessary data in their EHR for such screening that could be used to inform the consideration of statin therapy by their practitioner.

Independent Variables.

Race, ethnicity, and preferred language defined eight mutually exclusive groups consisting of Latino, non-Hispanic White, non-Hispanic Black, and Asian adults further distinguished by their preference for either English or other language. While we use Latino because it is often preferred in our study population, the actual ethnicity information collected by clinics is Hispanic/non-Hispanic. Due to sample size limitations, anyone identifying as Hispanic regardless of race was grouped in the Latino cohort (ex. <1% of the sample self-identified as Hispanic and with Black race).

Covariates.

SDH risk factors^31,32, care utilization, and ASCVD risk factors were included as potential confounders: patient’s sex, age at first study encounter, primary clinic location (urban, rural), insurance status (always, sometimes, never insured), and income reflecting designations relative to 138% of the US federal poverty level (FPL); number of primary care visits and years of contact over the study period; smoking status; and an aggregate high-risk flag indicating if the patient had a diagnosis of diabetes, hypertension, or obesity. We performed sensitivity analyses using separate indicators for each comorbidity as well as stratification by diabetes status. Further, we included a Not Documented category in both income and smoking variables as features of clinical assessment potentially not provided.

Statistical Analysis.

Patient characteristics were described by race/ethnicity/language group and for the overall sample. We considered two main analytic approaches to investigate the impact of the guideline changes on ASCVD risk-measure documentation. First, we modeled the binary indicator denoting the presence/absence of the time-period specific ASCVD screening measure separately for each time-period (2009-2013 and 2014-2018). Generalized estimating equations (GEE) logistic regression was used to analyze both the LDL and PCE measures outcomes within their outcome-specific study periods. We used a robust sandwich variance estimator and assumed exchangeable correlation structure to account for patient clustering within clinics. In each of the period-specific analyses, we report unadjusted prevalence of risk-measure documentation and the GEE estimated adjusted odds ratios (aOR) with 95% confidence intervals (CI) for each race/ethnicity/language group in comparison to the group of largest representation (English-preferring non-Hispanic White).

Next, we subset our overall sample to patients who had visits in both study periods (2009-2013 and 2014-2018) which yielded a cohort of patients that allowed us to estimate change in documentation over time utilizing a difference-in-difference approach. Among individuals eligible for both period-specific measures, we conducted a patient-level analysis assessing group differences in documented screening before and after the 2013 guidelines change. For this subsample analysis, we performed a GEE logistic regression with similar specifications as above and reported three adjusted estimates: 1) predicted probabilities for each outcome measure by group, 2) odds ratios for the within group differences of having PCE measures versus LDL documentation (pre/post guideline change), and 3) odds ratios comparing each group change to the change within the referent English-preferring non-Hispanic White cohort.

Analyses were conducted in Stata v.15 with two-sided testing and set 5% type I error. The Oregon Health & Science University Institutional Review Board approved this study.

Results

Analytic sample sizes for the 2009-2013 LDL and 2014-2018 PCE measures outcomes were 195,061 and 344,504 patients respectively (Table 1 and Appendix Table 1). 128,621 individuals were eligible for both study periods and comprised the sample population used to compare within and across group differences pre/post 2013 guidelines change (Appendix Table 2). Roughly 57% in each of the three time-period samples were female, 95% visited clinics in urban settings, and nearly 60% were always <=138% of the US FPL. Further, English preferring non- Hispanic White adults were more likely always insured while non-English preferring Latinos were more likely never insured. English preferring non-Hispanic White individuals were least likely to be at/below 138% FPL, but also on average had fewer primary care encounters in the study period than all other groups. More English preferring non-Hispanic Black adults were documented smokers and they were also the group most likely to have one or more of the high-risk comorbidities, including diabetes alone (roughly 85% with 1+ comorbidity, and 74% with diabetes).

Table 1.

Patient characteristics for the 2014-2018 observation period for assessing documentation of all pooled cohort equation measures in the electronic health record.

	Group, %
	Non-Hispanic White		Non-Hispanic Black		Latino		Asian
Characteristic	Prefers English	Other Language	Prefers English	Other Language	Prefers English	Other Language	Prefers English	Other Language
Patients	N=159188	N=8602	N=52086	N=6445	N=29692	N=75035	N=6298	N=7158
Female	52.9%	58.1%	53.8%	67.1%	55.2%	61.0%	58.3%	59.2%
Age at First Visit	median=61, range:55-73	median=62, range:55-73	median=61, range:55-73	median=62, range:55-73	median=61, range:55-73	median=61, range:55-73	median=62, range:55-73	median=63, range:55-73
Insurance Status
Always Insured	76.6%	68.8%	64.2%	63.3%	70.8%	60.6%	71.0%	78.2%
Sometimes Insured	10.6%	9.9%	19.2%	19.1%	14.1%	18.4%	13.8%	9.6%
Never Insured	12.8%	21.2%	16.6%	17.6%	15.2%	21.0%	15.2%	12.2%
Federal Poverty Level
Always > 138%	17.3%	8.9%	9.1%	5.2%	13.8%	7.3%	9.8%	4.8%
Always <= 138%	42.0%	56.5%	70.6%	73.1%	62.8%	71.9%	60.3%	67.5%
Above & Below 138%	8.2%	4.7%	7.0%	4.1%	7.2%	7.0%	6.3%	5.0%
Not Documented	32.4%	30.0%	13.3%	17.5%	16.2%	13.7%	23.6%	22.7%
Smoking Status
Non-Smoker	71.5%	79.9%	67.2%	88.5%	74.9%	85.4%	86.7%	86.6%
Smoker	21.6%	11.3%	25.2%	6.1%	14.8%	7.7%	8.7%	10.4%
Not Documented	6.9%	8.8%	7.6%	5.5%	10.2%	6.9%	4.6%	3.0%
High-Risk Indicators
1+ of the following:	67.8%	69.9%	84.7%	83.3%	77.0%	81.6%	72.3%	69.3%
Diabetes Mellitus	51.6%	52.9%	74.2%	67.8%	58.6%	63.3%	63.3%	60.6%
Hypertension	19.3%	24.5%	34.1%	36.5%	32.0%	38.0%	35.2%	32.1%
Obesity	44.6%	45.3%	51.2%	48.3%	50.0%	52.2%	21.6%	17.0%
Total Primary Care Visits
1	20.3%	24.6%	18.0%	13.9%	17.9%	15.3%	16.5%	13.3%
2-5	34.2%	35.0%	32.2%	32.1%	34.7%	32.3%	34.3%	30.8%
6+	45.5%	40.4%	49.8%	53.9%	47.4%	52.4%	49.2%	55.9%
Years of Primary Care Contact	median= 2, range:1-5	median= 1, range:1-5	median= 2, range:1-5	median= 2, range:1-5	median=2, range:1-5	median= 2, range:1-5	median=2, range:1-5	median= 2, range:1-5
Primary Clinic in Urban Area	93.3%	97.0%	97.9%	100.0%	97.4%	98.6%	99.0%	99.8%

These data were representative of 635 clinics spanning 24 US states: Alaska, California, Colorado, Florida, Georgia, Hawaii, Indiana, Kansas, Massachusetts, Maryland, Minnesota, Missouri, Montana, North Carolina, New Mexico, Nevada, New York, Ohio, Oregon, Rhode Island, Texas, Utah, Washington, and Wisconsin.

Low-density lipoprotein documented, 2009-2013

Among patients with a visit in 2009-2013, the unadjusted prevalence of LDL documentation in the EHR ranged from 56% for English preferring non-Hispanic White adults to 75% for non-English preferring Asian adults (Appendix Table 3). Results from adjusted GEE logistic modeling (Figure 1) show that compared to English preferring non-Hispanic White patients, all groups had higher odds of LDL documentation. Appendix Table 4 contains values corresponding to Figure 1.

Figure 1. Adjusted odds of having an LDL measure documented in the EHR, 2009-2013.

EHR = electronic health record; GEE = generalized estimating equations; LDL = low-density lipoprotein

Note: The reference group, English-preferring non-Hispanic White adults, is represented by the vertical dotted line at 1.00 on the x-axis. The dots represent the adjusted odds ratio point estimates and the horizontal lines represent 95% confidence intervals for those estimates. Estimates derived using GEE logistic regression adjusted for patient's sex, age, insurance status, %federal poverty level, smoking status, high-risk status (high-risk if any of diabetes, hypertension, or obesity diagnoses), number of primary care encounters, years of clinic utilization, and urban-rural status with clustering on patient's primary clinic.

Pooled Cohort Equations measures documented, 2014-2018

Among patients with a visit from 2014-2018, the unadjusted prevalence of having all required PCE measures documented in the EHR ranged from 57% for English preferring non-Hispanic White adults to 78% for non-English preferring Asian adults (Appendix Table 3). Similar to the period prior to the guideline changes, all race/ethnicity/language groups had higher odds of documentation compared to English preferring non-Hispanic White patients (Figure 2). Of note, 92% of individuals who did not have all the required measures to use the PCE were missing the lipid measurements (HDL and total cholesterol). The next most common missing value was smoking status (18%) obtained during history taking or visit rooming process. Appendix Table 4 contains values corresponding to Figure 2.

Figure 2. Adjusted odds of having all PCE measures documented in the EHR, 2014-2018.

EHR = electronic health record; GEE = generalized estimating equations; PCE = pooled cohort equations

Note: The reference group, English-preferring non-Hispanic White adults, is represented by the vertical dotted line at 1.00 on the x-axis. The dots represent the adjusted odds ratio point estimates and the horizontal lines represent 95% confidence intervals for those estimates. Estimates derived using GEE logistic regression adjusted for patient's sex, age, insurance status, %federal poverty level, smoking status, high-risk status (high-risk if any of diabetes, hypertension, or obesity diagnoses), number of primary care encounters, years of clinic utilization, and urban-rural status with clustering on patient's primary clinic.

Cohort study of LDL and PCE measures documented, 2009-2018

Among patients with visits in both the pre- and post-guideline change periods, lower odds of complete risk measure documentation were observed for all racial/ethnic/language groups in the period following the 2013 guidelines change as compared to the period prior to the change (Table 2). The magnitude of decline in risk measure documentation was greater for English-preferring non-Hispanic Black individuals than for the referent group of English-preferring non-Hispanic White patients (relative difference in change from pre- to post: aOR 0.81; 95% CI: 0.72-0.91). These two groups were the least likely to have all PCE measures documented in their EHR (55% and 56%, respectively). The other racial/ethnic/language groups saw similar declines in documentation odds compared to English-preferring non-Hispanic White patients.

Table 2.

Changes in documentation of atherosclerotic cardiovascular disease risk screening used for statin consideration pre/post 2013 guideline change.

	Atherosclerotic Cardiovascular Disease (ASCVD) Documented Risk Screening
Race-Ethnicity / Language Preference	Low-density Lipoprotein (2009-2013)	PCE Measuresb (2014-2018)
Non-Hispanic White / English
Adjusted probabilitya	62.8%	55.5%
PCE Measures vs. LDL, aOR (95% CI)		0.68 (0.62, 0.75)
Non-Hispanic White / non-English
Adjusted probabilitya	67.5%	58.1%
PCE Measures vs. LDL, aOR (95% CI)		0.59 (0.48, 0.73)
Non-Hispanic Black / English
Adjusted probabilitya	66.3%	55.4%
PCE Measures vs. LDL, aOR (95% CI)		0.55 (0.49, 0.61)
Non-Hispanic Black / non-English
Adjusted probabilitya	66.4%	60.2%
PCE Measures vs. LDL, aOR (95% CI)		0.71 (0.54, 0.94)
Latino / English
Adjusted probabilitya	64.7%	58.8%
PCE Measures vs. LDL, aOR (95% CI)		0.72 (0.56, 0.92)
Latino / non-English
Adjusted probabilitya	70.9%	61.3%
PCE Measures vs. LDL, aOR (95% CI)		0.57 (0.47, 0.69)
Asian / English
Adjusted probabilitya	69.7%	62.3%
PCE Measures vs. LDL, aOR (95% CI)		0.65 (0.55, 0.76)
Asian / non-English
Adjusted probabilitya	74.5%	63.5%
PCE Measures vs. LDL, aOR (95% CI)		0.51 (0.36, 0.71)

	Comparisons to Non-Hispanic White / English Cohort of Change in Screening Documentation, aOR (95% CI)
Non-Hispanic White / non-English	0.87 (0.71, 1.08)
Non-Hispanic Black / English	0.81 (0.72, 0.91)
Non-Hispanic Black / non-English	1.05 (0.78, 1.40)
Latino / English	1.06 (0.82, 1.39)
Latino / non-English	0.84 (0.68, 1.04)
Asian / English	0.96 (0.81, 1.13)
Asian / non-English	0.75 (0.54, 1.05)

These data were representative of 635 clinics spanning 24 US states: Alaska, California, Colorado, Florida, Georgia, Hawaii, Indiana, Kansas, Massachusetts, Maryland, Minnesota, Missouri, Montana, North Carolina, New Mexico, Nevada, New York, Ohio, Oregon, Rhode Island, Texas, Utah, Washington, and Wisconsin.

^aGeneralized estimating equation logistic regression adjusted for patient's sex, age, insurance status, %federal poverty level, smoking status, high-risk status (high-risk if any of diabetes, hypertension, or obesity diagnoses), number of primary care encounters, years of clinic utilization, and urban-rural status with clustering on patient's primary clinic. Boldface indicates statistical significance (p<0.05).

^bPresence of all required PCE measures for ASCVD risk score calculation. Required measures include age, sex, race, systolic blood pressure, total cholesterol, high-density lipoprotein, and smoking status. Additional measures used in PCE include indicators for diabetes and hypertension diagnoses and for prescribed antihypertensive medication.

aOR, adjusted odds ratio; CI, confidence interval; LDL, low-density lipoprotein; PCE, Pooled Cohort Equations

Discussion

Assessing for ASCVD risk is vital for mitigating the morbidity and mortality associated with this disease. We investigated racial, ethnic, and language differences in ASCVD risk-screening measures used in the consideration of statin pharmacotherapy. Utilizing 10 years of objective clinical and SDH data, this study provides a novel assessment of this clinical care. We observed patterns contrary to our expectations that elicit areas for improving ASCVD primary prevention efforts.

Patient preferred spoken language had a reverse association with screening measure documentation from our hypothesis. Asian, Black, Latino, and White individuals with preference for a non-English language were more likely to have ASCVD screening components documented than English-preferring non-Hispanic White adults. That language barriers may explain screening disparities for some racial/ethnic minority groups may not generalize to the CHC population.¹⁹ CHCs are mandated to provide language services. These services may naturally extend the care team for these patients. Bilingual staff, interpreters, and community health workers could provide more opportunities for additional patient intake outside face-to-face time with the primary care practitioner, as well as more patient education.

Low-income non-Hispanic White English-speaking patients may have limitations as a comparison group. Prior investigations in this population have shown poorer care-quality outcomes for this group compared to Spanish-preferring Latino patients.³³ This study however presents a disparity for this population relative to Asian, Black, and Latino groups regardless of preferred language. While our modelling attempts to account for group differences, unmeasured confounding may have contributed to these findings. First, non-Hispanic White adults preferring the English language had the lowest number of CHC primary care visits during the study period. This could indicate low overall utilization of services by this group or may reflect greater utilization of care delivered by specialists or outside the CHC network. The insurance and FPL status measures we observed within this comparator group support having the means for increased specialist and/or outside care. Second, non-Hispanic White English speakers were among the least likely to be at high-risk because of diabetes, hypertension, or obesity. These differences along with family history may have guided clinical decision-making for further laboratory screening (e.g., the most commonly missing PCE component was lipid panel measurement). Disaggregation of this population beyond language preference, for example, by country of origin is a direction for future work.

Among CHC users, the 2013 guidelines had no effect on racial/ethnic/language disparities in documented ASCVD risk screening measures used for statin consideration. Neither did the guideline roll-out improve the prevalence of adequate risk factor documentation in this population. To the contrary, in our cohort sample of patients seeking care both before and after the 2013 guidelines, adequate risk factor documentation worsened after the guideline change. The PCE suggested in the 2013 guidelines have been shown to overestimate risk^34-36; perhaps clinicians have sought to balance this tendency by relying less on the PCE in clinical practice to drive clinical decision-making. If this is true, we would expect the observed patterns to continue beyond the end of the study period. Investigations comparing the effects of the 2013 guidelines on CHC and non-CHC users are necessary to fully assess the guidelines’ impact on health equity and intention to improve ASCVD outcomes. However, this guideline change, in this population, did not increase the collection of data necessary to consider statin initiation.

Lastly, while collectively all racial/ethnic/language groups appear to have reduced screenings since the guideline implementation, our findings suggest English-preferring Black individuals may have experienced the worst impact. In addition to having the greatest drop in screening prevalence compared to the English-preferring non-Hispanic White group, they showed the lowest prevalence of PCE measures (55%) among all groups. Overall, Black individuals in our study had more high-risk indicators for ASCVD, however we controlled for these in our modelling and also separately conducted a sensitivity analysis by diabetes status (yes/no) and found no differences in the direction or magnitude of our estimates. It is possible that enough of these individuals within this CHC population were already on statin therapy to justify the differences in screening; however, evidence demonstrates Black adults are under-prescribed statin therapy compared with other race/ethnicity groups in non-CHC settings.^37-40 Assessing guideline-concordant pharmacotherapy within CHCs in all of these racial/ethnic groups is an important direction for future consideration.

Limitations.

First, most non-English preferring Latino adults in our study preferred the Spanish language, but there was substantial heterogeneity in languages preferred by the other groups (Black, White, Asian). Consequently, inadequate sample sizes prompted our designation of English/non-English groups. While language differences in different race/ethnicity groups may not be comparable, English proficiency as a social factor for understanding differences in care quality is well documented.^41-45 Proficiency may reflect degree of acculturation, immigrant status, or nativity, factors on which future investigations can expand. Second, we analyzed screening measures captured in the EHR which does not offer a complete assessment of clinical screening. It is possible lab results from outside sources were reviewed during visits and would have only been captured with a chart-review outside the scope of this study. Similarly, we could not ascertain whether patients left the network during the study period. However, there is evidence that patients who utilize this network generally do not leave, and further, CHC patients often lack financial resources to receive care outside the network limiting any effect from this potential lack of capture.⁴⁶-⁴⁸ It is also possible that the PCE components were not obtained nor used for the purpose of ASCVD screening (risk score calculation), though this would be expected to be similar across cohorts. Still, given evaluation of individual risk scores was outside the scope of this study future work in this area is needed. Third, while the data that informed this study came from a large network of CHCs in 24 states, the findings may not generalize to the entire US. However, our findings may generalize to CHCs nationwide since the OCHIN network members share characteristics of the roughly 1400 CHCs in the US.⁴⁹ Fourth, our primary exposure is a component of the PCE prompting a selection bias which may result in overestimating the prevalence of PCE screening data in the EHR for people most likely to not report race, ethnicity, or sex data. Lastly, race in particular, as a social construct is a non-homogenous measure. Yet we use race in scientific analyses to attempt to understand and improve upon the experience of people suffering disadvantage through society’s structure. Our work highlights the need to continue deconstructing all heterogeneous groups to gain a better understanding of how we may improve health outcomes for all individuals.

Conclusion.

Among patients utilizing CHCs for primary care, racial/ethnic and language minority groups were more likely than English-preferring non-Hispanic White individuals to have guideline concordant measures for statin consideration documented in their EHR. The 2013 guidelines did not alter these differences. However, screening components for all groups decreased after guideline implementation and this decrease was most significant for English-preferring non-Hispanic Black adults. Future research can assess the impact of the 2013 guidelines on ASCVD pharmacotherapy and morbidity/mortality as well as further investigate within racial/ethnic group differences.

Highlights.

• Race, ethnicity, language associated with statin screening measures documentation

• Electronic health record data for cardiovascular screening best for minority groups

• Black adults’ cardiovascular screening data less robust after 2013 guideline change

Non-standard Abbreviations and Acronyms

ASCVD

atherosclerotic cardiovascular disease

ADVANCE

Accelerating Data Value Across a National Community Health Center Network

CHC

community health center

GEE

generalized estimating equations

EHR

electronic health record

OCHIN

not an abbreviation

PCE

pooled cohort equations