Factors Associated with Lipoprotein(a) Testing Among Multiethnic Individuals

Abstract

Background

Lipoprotein(a) [Lp(a)] is a causal risk factor for atherosclerotic cardiovascular disease (ASCVD) and clinical guidelines recommend incorporating Lp(a) testing in routine care.

Objective

Examine real-world, contemporary clinical testing patterns of Lp(a) among multiethnic populations.

Design

In this nested case–control study, we assessed the prevalence and factors associated with Lp(a) testing within a large Northern Californian health system between 2010 and 2021. Incident density matching was used to select controls matched with a case for a case:control ratio of up to 1:5. Conditional logistic regression was used to assess the relationship between Lp(a) testing, sociodemographic, and clinical characteristics.

Participants

We included individuals aged 18 years or older with ≥ 2 primary care visits during the study period.

Main Measures

Lp(a) testing rates over time and factors associated with testing based on demographic, medical, and healthcare utilization variables.

Key Results

Of the 1,484,410 individuals in the cohort, 14,818 (1.0%) underwent Lp(a) testing. The median Lp(a) level was 35 mg/dL and over a third of individuals had Lp(a) levels > 50 mg/dL. After adjustment, South Asian individuals were three times more likely to have undergone Lp(a) testing, as compared to non-Hispanic White individuals [OR = 3.19, (95% CI = 2.98, 3.41)], while those identified as non-Hispanic Black and Hispanic were significantly less likely to have undergone Lp(a) testing [OR = 0.70, (95% CI = 0.62, 0.80) and 0.64 (95% CI = 0.59, 0.69), respectively]. Those with a history of ASCVD had over twice the odds of undergoing testing [OR = 2.14 (95% CI = 1.99, 2.29)], as did individuals with more frequent primary care visits [OR = 1.99 (95% CI = 1.84, 2.15)].

Conclusions

Lp(a) testing rates in real-world settings are low, with significant disparities by race, ethnicity, and healthcare utilization. Expanding access to Lp(a) testing may help reduce disparities within ASCVD risk assessment and treatment as new targeted therapeutic agents become available.

Supplementary Information

The online version contains supplementary material available at 10.1007/s11606-024-09126-6.

INTRODUCTION

Lipoprotein(a) [Lp(a)] is a causal risk factor for atherosclerotic cardiovascular disease (ASCVD) due to its atherogenic and thrombotic characteristics.¹ Lp(a) can be a powerful tool for personalizing ASCVD risk assessment, and multiple medical societies advise Lp(a) testing in routine clinical practice, particularly as a risk enhancing feature to guide preventive treatment decisions. The European Society of Cardiology (ESC) and European Atherosclerosis Society (EAS), the Canadian Cardiovascular Society (CCS), and the National Lipid Association (NLA) all endorse one-time Lp(a) screening in all individuals in order to identify those with very high levels of Lp(a).^2–4 While major societies in the US do not currently endorse one-time screening, the American College of Cardiology (ACC) and American Heart Association (AHA) task force recommendations from 2018 consider Lp(a) levels ≥ 50 mg/dL or ≥ 125 nmol/L a risk-enhancing factor for ASCVD.⁵

Despite existing guidelines, a gap remains in understanding how these recommendations are implemented into real-world practice. With the growing interest in Lp(a) as a causal ASCVD risk-enhancing factor, there is a need to understand testing patterns across diverse patient populations and to determine which sociodemographic factors predict testing. Prior literature has found that Lp(a) testing occurs infrequently as part of routine care.⁶ Consequently, our study aimed to investigate the frequency of completed Lp(a) testing within a large health system and identify characteristics that associated with testing. We hypothesized that Lp(a) testing would differ based on sociodemographic factors including race, ethnicity, ASCVD status, and frequency of healthcare encounters.

METHODS

The study was approved by the Sutter and Stanford University Institutional Review Board. The dataset analyzed during the current study is not publicly available due to reasonable privacy and security concerns. The underlying EHR data are not easily redistributable to researchers other than those engaged in the Institutional Review Board-approved research collaborations in the current project. The corresponding author may be contacted for access to EHR data for an IRB-approved collaboration.

Study Population

This is a nested case–control study using data from a large Northern California health system (Sutter Health). The organization is a multiple-payer system serving 3.3 million individuals. The racial and ethnic composition of this health system includes 59% non-Hispanic White, 13% Asian, 15% Hispanic, and 5% Black individuals. Eligible individuals were those aged 18 or older who had 2 or more primary care visits between the years 2010 to 2021 to ensure adequate follow-up and complete data. Lp(a) cases were defined as those who had at least 1 Lp(a) laboratory test ordered, with the first ordered date classified as the index date. If an individual had more than 1 Lp(a) test, only the first ordered Lp(a) value was included in the analysis. Lp(a) levels > 50 mg/dL were considered elevated, as per national guidelines.⁵

Incident density matching was performed to select controls since Lp(a) testing was infrequent.^7,8 Control individuals were subject to the same inclusion and exclusion criteria as above. They were matched to cases if they had a clinic visit within 6 months before or after the case’s index date. Up to 5 controls were included for each case, and the index date for control individuals was defined as the index date for the matched case individual. As the intent of this study was to better understand the variables that were associated with Lp(a) testing, controls were not matched on any variables other than visit date.

Covariate Data

Baseline demographic, clinical, and healthcare utilization characteristics were collected on or within 2 years before the index date. Demographic information included age, sex and self-reported race and ethnicity. Individuals reporting Hispanic ethnicity were classified as Hispanic regardless of race. Individuals who reported more than 1 race were classified as non-Hispanic Other. Clinical data included a history of hypertension, diabetes, and ASCVD. These were defined as having 2 or more encounters (outpatient, emergency department, or inpatient discharge diagnosis) with International Classification of Disease (ICD)-9 or ICD-10 code for the relevant diagnosis within 2 years of index Lp(a) date. ASCVD history was defined using the ICD-9 or -10 codes for coronary artery disease, stroke, and peripheral arterial disease. Among individuals without ASCVD, 10-year ASCVD risk was calculated using the Pooled Cohort Equations.⁵ Serum lipid measurements and statin use were defined using the most current measurement within 2 years prior to the Lp(a) testing date. Insurance type was classified as Health Maintenance Organization (HMO), Medicaid/Medi-Cal, Medicare Fee-for-service (FFS) or Medicare HMO, Preferred Provider Organization (PPO) or FFS, or Other, self-insured, or unknown insurance status. Cost of testing and insurance coverage of testing was not able to be reliably captured in our database. Healthcare utilization was measured as having a primary care physician (PCP) visit, My Health Online visit, Telephone encounter, or Cardiology encounter within 1 year prior to the index Lp(a) testing date. The number of visits within each category 1 year prior to the index date were also measured. Statin prescriptions within two years of the index date were classified by intensity (low, moderate, or high) based on current practice guidelines⁵ and can be found in Supplemental Table 1.

Statistical Analysis

Cochran-Armitage testing was performed for Lp(a) testing over time. Chi-square testing was used to compare categorical variables and Wilcoxon two-sample testing was performed for continuous variables if normality did not hold; otherwise t-testing was used. Conditional logistic regression models were used to adjust for the correlation within the matched cases and controls. Our model included age, sex, race, medical comorbidities (hypertension, diabetes, body mass index, and ASCVD history), smoking status, insurance status, and healthcare utilization. Only cases and controls with complete covariate data were included. Odds ratio and associated 95% confidence interval were reported for each predictor from the model. Sensitivity analyses were performed to compare the odds ratio with and without missing values for any of the risk factors in which missingness occurred, including race, ethnicity, BMI, and smoking status.

All analyses were conducted using SAS 9.4 (Cary, NC). A statistically significant result was defined as a p-value < 0.05. The STROBE Statement checklist for case–control studies was used to construct this manuscript.⁹

RESULTS

Of the 1,484,410 individuals who met the inclusion criteria, 14,818 (1.0%) underwent Lp(a) testing between the years 2010 to 2021. Of these, 12,349 individuals had complete data and were compared to 49,835 matched controls (Supplemental Fig. 1). The median Lp(a) level in those who obtained testing was 35 mg/dL (IQR 17–78 mg/dL) and 5,598 (37.8%) of those who underwent testing had Lp(a) levels > 50 mg/dL. Non-Hispanic Black individuals had the highest median Lp(a) of 65 mg/dL. There was a statistically significant increase (p = 0.0001) in the frequency of Lp(a) testing over time (Fig. 1).

Figure 1.

Proportion of unique, active individuals who underwent Lp(a) testing per year.

Individuals who had Lp(a) tested were older (mean age 54.3 vs. 51.6 years) and more likely to be male (51.9% vs 40.7%). Compared to those without Lp(a) testing, those with Lp(a) testing were more likely to be Asian (26.6% vs. 19.2%) and less likely to be non-Hispanic Black or Hispanic (2.3% vs 3.5% and 6.1% vs. 11.9%, respectively). Among disaggregated Asian subgroups, South Asian individuals represented 67.8% of individuals who underwent Lp(a) testing compared to 31.7% of untested Asian individuals. Additional demographic and clinical characteristics differences by Lp(a) testing status are shown in Table 1.

Table 1.

Demographic and Clinical Characteristics of Those with and Without Lipoprotein(a) Testing. Controls were Matched with a Case if a Visit Date was within ± 6 Months of the Index Lp(a) for a Case:Control Ratio of up to 1:5^*

Characteristic	Lp(a) tested N = 14,818	Control N = 74,090	p-value
Age, mean (SD), y	54.3 ± 13.8	51.6 ± 17.8	< .0001
18–39	2,178 (14.8%)	20,932 (28.3%)
40–64	8,976 (60.9%)	34,264 (46.2%)
65–79	3,165 (21.4%)	13,933 (18.8%)
80 +	499 (3.4%)	4,961 (6.7%)
Sex, No., %			< .0001
Female	7,121 (48.1%)	43,901 (59.3%)
Male	7,697 (51.9%)	30,189 (40.7%)
Racea, No., %			< .0001
Non-Hispanic White	8,078 (54.5%)	40,458 (54.6%)
Non-Hispanic Black	345 (2.3%)	2,622 (3.5%)
Hispanic	900 (6.1%)	8,794 (11.9%)
Non-Hispanic Asian	3,939 (26.6%)	14,225 (19.2%)
Non-Hispanic Other	427 (2.9%)	2,422 (3.3%)
Asian group, No., %			< .001
South Asian	2,670 (67.8%)	4,509 (31.7%)
Other Asian group	1,269 (32.2%)	9,716 (68.3%)
History of Hypertensionb	4,112 (27.8%)	19,351 (26.1%)	< .0001
Type 2 Diabetesc	1,399 (9.4%)	6,790 (9.2%)	0.29
ASCVD Category Ever Diagnosedd	2,598 (17.5%)	6,372 (8.6%)	< .0001
ASCVD Category			< .0001
Myocardial Infarction Only OR CAD only	1,521 (10.3%)	2,912 (3.9%)
Cerebrovascular Disease only	419 (2.8%)	1,409 (1.9%)
Peripheral Vascular Disease only	402 (2.7%)	1,431 (1.9%)
Polyvascular Disease	256 (1.7%)	620 (0.8%)
10-year ASCVD Riske among those not diagnosed with ASCVD			< .0001
< 5%	6,459 (52.9%)	19,744 (29.2%)
5- < 7.5%	971 (7.9%)	3,014 (4.5%)
7.5 to 20%	1,600 (13.1%)	6,847 (10.1%)
> 20%	697 (5.7%)	5,155 (7.6%)
Mean value	6.3%	9.7%
BMI (mean, SD)	27.5 ± 5.6	27.8 ± 6.3	< .0001
BMI < 25	5,153 (34.8%)	24,135 (32.6%)
BMI 25–30	5,502 (37.1%)	22,589 (30.5%)
BMI > 30	3,572 (24.1%)	19,520 (26.3%)
Index Lp(a)f, median (IQR), mg/dL	35 (17, 78)	-	-
Lipid Panel (mg/dL)
Total cholesterol, mean ± SD	202.8 ± 51.0	186.2 ± 39.8	< .0001
Baseline HDL, mean ± SD	55.9 ± 18.2	56.7 ± 17.9	< .0001
Baseline LDL, mean ± SD	120.1 ± 50.3	105.9 ± 36.0	< .0001
Baseline Triglycerides, mean ± SD	126.5 ± 95.4	121.4 ± 84.7	< .0001
LDL ≥ 190 mg/dLg	743 (5.1%)	540 (1.3%)	< .0001
Smoking Statush			< .0001
Current Smoker	471 (3.2%)	4,088 (5.5%)
Ever Smoker	3,401 (23.0%)	16,070 (21.7%)
Never Smoker	10,142 (68.4%)	45,653 (61.6%)
Statin Usei
No	11,407 (77.0%)	61,227 (82.6%)	< .0001
Low Intensity	331 (2.2%)	1,717 (2.3%)
Moderate Intensity	1,983 (13.4%)	8,263 (11.2%)
High Intensity	1,097 (7.4%)	2,883 (3.9%)
Primary Insurancej			< .0001
HMO	2,473 (16.7%)	11,990 (16.2%)
Medicaid/Medi-Cal	153 (1.0%)	1,676 (2.3%)
Medicare FFS + Medicare HMO	3,333 (22.5%)	17,433 (23.5%)
PPO/FFS	8,237 (55.6%)	36,752 (49.6%)
Other/Self/Unknown	622 (4.2%)	6,239 (8.4%)
Health care utilizationk, mean ± SD
PCP visits, mean ± SD	2.54 ± 2.50	2.34 ± 2.65	< .0001
MHO encounters, mean ± SD	5.41 ± 9.46	3.40 ± 6.79	< .0001
Telephone encounters, mean ± SD	5.19 ± 7.76	4.51 ± 7.38	< .0001
Cardiology encounters, mean ± SD	1.17 ± 3.89	0.57 ± 2.78	< .0001

Abbreviations: ASCVD = Atherosclerotic Cardiovascular Disease, BMI = Body Mass Index (reported as kg/m²), CAD = Coronary Artery Disease, CVD = Cardiovascular Disease, FFS = Fee-for-service, HDL = High-density Lipoprotein, HMO = Health Maintenance Organization, LDL = Low-density Lipoprotein, Lp(a) = Lipoprotein(a), MHO = My Health Online, PPO = Preferred Provider Organization, PCP = Primary Care Physician, SD = Standard Deviation

^*Missing baseline covariate data reported in Table S1

^aIndividuals reporting Hispanic ethnicity were classified as Hispanic regardless of race. Individuals who reported more than 1 race were classified as non-Hispanic Other

^bDefined as having 2 or more encounters (outpatient, emergency department, or inpatient discharge diagnosis) with ICD-9 or ICD-10 code for hypertension within 2 years of index Lp(a) date

^cDefined as having 2 or more encounters (outpatient, emergency department, or inpatient discharge diagnosis) with ICD-9 or ICD-10 code for type 2 diabetes within 2 years of index Lp(a) date

^dDefined as having 2 or more encounters (outpatient, emergency department, or inpatient discharge diagnosis) with ICD-9 or ICD-10 code for ASCVD within 2 years of index Lp(a) date

^eCalculated using the Revised Pooled Cohort Equation

^fSerum lipid measurements were assigned using the most current measurement within 2 years prior to the Lp(a) date

^gProportions calculated using total individuals with baseline LDL measurement in each group as denominator

^hDefined as self-reported tobacco use history within 2 years the index Lp(a) date

ⁱDefined as statin prescription within 2 years of the index Lp(a) date

^jDefined as the primary insurance during the same year as index Lp(a) date, or the year prior if information not available

^kDefined as visits or encounters within 1 year prior of index Lp(a) date

Of those who underwent Lp(a) testing, there was a higher proportion of individuals with diagnosed ASCVD (17.5% vs. 8.6%). In those without diagnosed ASCVD and with available data for PCE estimation, those who underwent Lp(a) testing generally had a lower 10-year ASCVD predicted risk, with an estimated risk of 6.3% in the Lp(a) testing group vs. 9.7% in the control group. Lipid and PCE data were more likely to be complete in cases rather than controls (differences in missing data between cases and controlled are presented in Supplementary Tables 2 and 3).

Individuals who underwent Lp(a) testing had a similar mean number of PCP visits (2.5 vs. 2.3), yet they had more online portal encounters (5.4 vs. 4.1) and telephone encounters (5.2 vs. 4.5). Those who underwent testing had higher rates of either moderate- (13.4% vs. 11.2%) or high-intensity (7.4% vs. 3.9%) statin use. A higher proportion of individuals who had Lp(a) checked had at least 1 cardiology visit (15.6% vs. 6.9%) or had a prescription for a statin (23.0% vs. 17.4%).

After adjustment, older individuals and men were more likely to have undergone Lp(a) testing (Fig. 2). Those with a history of ASCVD had twice the odds of have undergone testing as compared with those without ASCVD [OR = 2.14 (95% CI = 1.99, 2.29)]. Those with a BMI > 30 were less likely to have undergone Lp(a) testing. South Asian individuals were three times more likely to have undergone Lp(a) testing as compared with non-Hispanic White individuals [OR = 3.19, (95% CI = 2.98, 3.41)], while those identified as non-Hispanic Black or Hispanic were significantly less likely to have undergone Lp(a) testing, [OR = 0.70, (95% CI = 0.62, 0.80) and 0.64 (95% CI = 0.59, 0.69), respectively]. As compared to individuals with PPO or FFS insurance, those with Medicaid/Medi-Cal and Medicare FFS or HMO insurance were less likely to have undergone Lp(a) testing [OR = 0.55 (95% CI = 0.45, 0.66) and 0.87 (0.79, 0.95), respectively]. Patients with more frequent primary care and cardiology visits were more likely to have undergone Lp(a) testing (Fig. 2). A sensitivity analysis including missing data yielded qualitatively similar results (Supplemental Table 4).

Figure 2.

Adjusted odds ratio (OR) of having had lipoprotein(a) tested vs. control using logistic regression with 95% confidence interval. Sample size n = 62,184*. Abbreviations: ASCVD = Atherosclerotic Cardiovascular Disease; BMI = body mass index, DM = Diabetes Mellitus, FFS = Fee-for-service, HMO = Health Maintenance Organization, PCP = Primary Care Physician, PPO = Preferred Provider Organization. *Model includes age, sex, race and ethnicity, hypertension, type 2 diabetes mellitus, atherosclerotic cardiovascular disease diagnosis, BMI, smoking status, primary insurance, primary care physician visit within 1 year prior to index date, and cardiology visit within 1 year prior to index date. Excludes cases (and relevant controls) with missing data in race, BMI, smoking status, and primary insurance.

DISCUSSION

In this study, we leveraged a diverse cohort of 1.5 million individuals within an integrated health system to understand real-world practice patterns and factors associated with Lp(a) testing over a decade. Overall, we found that Lp(a) testing was low, with 1% of adults undergoing testing. Of those who underwent testing, over a third of individuals had Lp(a) levels > 50 mg/dL. There were important differences in testing based on sex, race and ethnicity, ASCVD risk level, insurance status, and health care utilization. Those who underwent Lp(a) testing were more likely to be South Asian, less likely to be non-Hispanic Black or Hispanic, less likely to be a woman, and more likely to have higher healthcare utilization.

The low rate of Lp(a) testing within our study population suggests there are still significant gaps in widespread testing of Lp(a) as recommended by some cardiovascular societies. Our findings are consistent with other studies examining rates of Lp(a) testing, which have found rates of Lp(a) testing below 1% in general populations and those at high-risk for ASCVD.^11–13 While Lp(a) testing is generally low, we found the overall Lp(a) testing rate increased over time. This likely reflects recent multi-society guideline recommendations for routine testing of Lp(a) among those at increased risk for ASCVD, as well as increased awareness of Lp(a) as a potent risk factor for cardiovascular disease.^2,14 The slightly higher rate of Lp(a) testing observed in our study as compared with others may be attributable to the more recent years captured in our analysis that reflect incorporation of these guideline recommendations into clinical practice.

It is noteworthy that among those tested, the median Lp(a) value was 35 mg/dL and more than one-third of individuals had a value elevated greater than 50 mg/dL, with variation by race, ethnicity, and ASCVD risk level. One analysis examining rates of Lp(a) testing within an international cohort of individuals with established ASCVD found a median Lp(a) of 18.0 mg/dL.¹⁵ The higher median level in our study might reflect local practice patterns with providers targeting groups with the highest probability of having an elevated Lp(a) — e.g., those at very high risk due to prior ASCVD events and those at high risk due to family history or race and ethnicity, but without other traditional risk factors such as smoking, hypertension, or diabetes. While Lp(a) ≥ 50 mg/dL is generally considered elevated and a risk enhancing feature, recent guidelines favor interpreting Lp(a) levels as a continuum of risk.^4,5

We found significant variation in Lp(a) testing by race and ethnicity. Compared to non-Hispanic White individuals, South Asian individuals were 3 times more likely to be tested, whereas non-Hispanic Black and Hispanic individuals were 30% and 40% less likely to be tested, respectively. The increased testing in South Asian individuals might be explained by the growing recognition of higher Lp(a) testing in this population.^16,17 Additionally, increased testing in South Asian individuals may have been influenced by the presence of a focused South Asian clinic. Although South Asian specific cardiovascular clinics are growing,¹⁸ the observations may not be broadly generalizable.

Prior work from our group and others highlighted significant variability in Lp(a) levels by race and ethnicity.^10,19–21 Black individuals have higher values of Lp(a) on average, whereas Asian and Hispanic individuals have lower Lp(a) values.¹⁰ While there are differences in Lp(a) values between racial and ethnic groups, race- and ethnicity-based normalized values are not typically incorporated into clinical practice. Although there is growing awareness of the higher Lp(a) values in Black individuals, this has not translated to increased testing as seen in our study. More work is needed to understand the underlying determinants of differential Lp(a) testing by race and ethnicity as there appears to be systematic undertesting of historically marginalized populations. Significant opportunities exist to enhance testing among diverse individuals, potentially through increasing outreach to diverse populations and expanding awareness among providers.

Unsurprisingly, individuals who underwent Lp(a) testing received more regular and frequent care, including cholesterol testing, primary care visits, and cardiology visits. In addition, those with Lp(a) testing had higher mean telephone and online patient portal encounters. This aligns with other studies demonstrating those with elevated Lp(a) have higher rates of healthcare utilization and spending.²² As Lp(a) has gained more attention in public media, there is likely some component of patient-driven demand to obtain testing.²³ Medicaid and Medicare beneficiaries were less likely to have undergone Lp(a) testing as compared with private insurance, suggesting potential insurance coverage disparities as Lp(a) is not routinely covered by Medicare.²⁴ Expanding insurance coverage for Lp(a) testing may be one method to remove barriers to testing. Individuals with Lp(a) testing had more cardiologist visits and were more likely to be prescribed statins, supporting previous findings associating Lp(a) testing with lipid-lowering treatment initiation.²⁵ Similar patterns of increased lipid lowering therapy initiation and intensification have been documented in other health systems after Lp(a) testing.^2–4 This suggests that even in the absence of targeted therapies, elevated Lp(a) serves as a useful risk enhancer to help optimize and intensify preventive strategies.

This study should be interpreted in the context of several limitations. First, we were unable to examine how family history of ASCVD or history of premature ASCVD may have influenced testing for Lp(a) as this was not a variable that was routinely captured in our study. Second, our study population was largely insured and cared for in a large healthcare system in Northern California, and we did not have data regarding insurance coverage of testing and cost. Programs that encourage Lp(a) testing, such as the South Asian Heart Center, may have also influenced the higher rate of testing among South Asian individuals. Therefore, the results are not generalizable to other populations. Third, we were unable to capture whether individuals identified as controls (i.e. without Lp(a) testing) in our study population may have obtained Lp(a) testing through other healthcare systems. Fourth, there were missing data, which could affect the interpretation of results. However, separate sensitivity analyses including missing data showed similar results.

In conclusion, this study examined the real-world clinical testing patterns of Lp(a) over a decade within a large, diverse health system. While testing rates are increasing, only a small proportion of the overall population underwent testing and there are significant disparities among testing within groups that have been historically underserved for ASCVD risk prevention. The presence of Lp(a) testing, alone, is associated with increased access to healthcare. Expanding access to Lp(a) testing across diverse populations can improve ASCVD risk assessment and treatment, particularly as targeted therapeutics become available.

Supplementary Information

Below is the link to the electronic supplementary material.

Funding

Dr. Rodriguez was funded by grants from the NIH National Heart, Lung, and Blood Institute (R01HL168188; R01HL167974; R01HL169345), the American Heart Association/Harold Amos Faculty Development program, and the Doris Duke Foundation (Grant #2022051). Dr. Palaniappan reported receiving grants from the National Institutes of Health/National Heart, Lung, and Blood Institute (K24 HL150476). The funders had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.

Data Availability

The dataset analyzed during the current study is not publicly available due to reasonable privacy and security concerns. The underlying EHR data are not easily redistributable to researchers other than those engaged in the Institutional Review Board-approved research collaborations in the current project. The corresponding author may be contacted for access to EHR data for an IRB-approved collaboration.

Declarations:

Conflict of Interest:

Dr. Rodriguez reports equity from Carta Healthcare and HealthPals, and consulting fees from HealthPals, Novartis, NovoNordisk, Esperion Therapeutics, Movano Health, Kento Health, Inclusive Health, Edwards, Arrowhead Pharmaceuticals, HeartFlow, iRhythm, Amgen, and Cleerly Health outside the submitted work. The remaining authors report no relevant disclosures or competing interests.