Trends in prevalence of guideline‐based use of lipid‐lowering therapy in a large health system

Abstract Background The 2013 ACC/AHA (American College of Cardiology/American Heart Association) cholesterol guidelines provided an evidence‐based rationale for the allocation of lipid‐lowering therapy based on risk for atherosclerotic cardiovascular disease (ASCVD). Adoption of these guidelines was initially suboptimal but whether this has improved over time remains unclear. Hypothesis Prevalence of guideline‐based statin therapy will increase over time. Methods Electronic health record data were used to create two cross‐sectional data sets of patients (age 40‐75) served in 2013 and 2017 by a large health system. Data sets included demographics, clinical risk factors, lipid values, diagnostic codes, and active medication orders during each period. Prevalence of indications for statin therapy according to the ACC/AHA guidelines and statin prescriptions were compared between each time period. Results In 2013, of the 219 376 adults, 57.7% of patients met statin eligibility criteria, of which 61.3% were prescribed any statin and 19.0% a high intensity statin. Among those eligible, statin use was highest in those with established ASCVD (83.9%) and lowest in those with elevated ASCVD risk >7.5% (39.3%). In 2017, of the 256 074 adults, 62.3% were statin eligible, of which 62.3% were prescribed a statin and 24.3% a high intensity statin. In 2017, 66.4% of statin eligible men were prescribed a statin compared to 57.4% of statin eligible women (P < 0.001). The use of ezetimibe (3.6% in 2013, 2.4% in 2017) and protein convertase subtilisin/kexin type 9 inhibitors (<0.1% and 0.1%) was infrequent. Conclusion In a large health system, guideline‐based statin use has remained suboptimal. Improved strategies are needed to increase statin utilization in appropriate patients.


| INTRODUCTION
In 2013, the American College of Cardiology (ACC) and American Heart Association (AHA) released guidelines for the treatment of blood cholesterol, recommending that statin therapy be allocated to those at elevated risk of atherosclerotic cardiovascular disease (ASCVD), including patients with established ASCVD, diabetes, or a low-density lipoprotein cholesterol (LDL-C) ≥190 mg/dL, as well as patients with an estimated 10-year ASCVD risk of ≥7.5%. 1 This riskbased approach was a significant shift from the prior adult treatment panel III guidelines, 2 and was reinforced in the recent 2018 ACC/AHA cholesterol guideline update. 3 Subsequent analyses suggested that implementation of the 2013 guidelines would lead to a dramatic increase in statin use across the United States, with US National Health and Nutrition Examination Survey (NHANES) data suggesting that almost of half of US adults aged 40 to 75 years are statin eligible by the 2013 guidelines. 4 The majority of the increase in statin eligibility was seen in older individuals without established ASCVD. 4,5 However, a subsequent analysis suggested that in the years immediately following publication of the 2013 guidelines, statin utilization minimally changed. 6 Whether or not compliance with the 2013 ACC/AHA cholesterol guidelines has improved in more recent years is unclear. Statin therapy can significantly reduce ASCVD risk, including in individuals at relatively low ASCVD risk, 7 and, in contrast to public perception, 8 statins have an excellent safety profile with minimal side effects in blinded trials. 9,10 Improved adoption of the 2013 cholesterol guidelines holds the potential to significantly reduce ASCVD rates at a population level. 11 We therefore aimed to analyze the prevalence of statin eligibility and subsequent rates of statin use in a large Midwestern health system.

| Setting and patients
This retrospective cohort study analyzed data from Allina Health, a large Midwestern health system operating 12 hospitals and over 100 clinics (primary care and specialty clinics within Minnesota and western Wisconsin). Data extracts were generated from Allina's electronic health record (EHR). All hospitals and clinics in the Allina Health system use the EpicCare EHR system (Epic Systems Corporation) which Allina has branded Excellian. This study was approved by the Allina Health Institutional Review Board.
Two cross-sectional data sets were created, the first including patients who received care in 2013 and the second, 4 years after the 2013 guidelines were released (2017). We included adults aged 40 to 75 years at the start of each extract time period (ie, 2013 and 2017) with at least one in-person ambulatory clinic visit during that year including primary care office visits, OB/GYN (obstetrics or gynecology) encounters for patients who were not pregnant, pre-operation visits, and nurse-only visits. Pregnant patients were excluded. Individuals were excluded who had opted out of use of their data for research purposes through the Minnesota Research Authorization process (typically <5% of the patient population). After applying the initial inclusion criteria, we then excluded patients with insufficient data to determine statin eligibility. The initial inclusion criteria identified 316 092 patients in 2013 and 361 514 in 2017. Approximately 30% of each of these cohorts was excluded due to lack of sufficient data to determine statin eligibility, leaving 219 376 patients in 2013 and 256 074 patients in 2017 ( Figure 1).
Patients excluded due to missing data differed from those who were included in the final sample on several characteristics. In the 2013 cohort, the excluded patients were an average of 4 years younger than the included, and had lower prevalence of hypertension (36% in the excluded vs 55% in the included), and <1% had documented diabetes or heart disease compared to 18% and 14% for these conditions, respectively in the included sample. Differences were very similar for the included and excluded in the 2017 cohort. Additionally, statin use was much lower in those excluded, with 9% of those excluded from the 2013 sample and 10% of those excluded from the 2017 sample having an active order for statins (compared to 43% of those included in the 2013 sample and 44% in the 2017 sample).

| Assessment of statin use and eligibility
The use of lipid-lowering medications was defined by any medication order (indicating a prescription) within a lipid lowering medication category in the EHR from a non-hospital visit with a start date prior to the end of the study period and no end date prior to the start of the study period. This broad definition capturing "any active order" during the study period was used for the primary analysis. As a sensitivity analysis, we also created a secondary measure that looked at active orders only at the time of the last visit during the study period. Medication orders for a lipid lowering medication were divided into two 10-year CVD risk ≥7.5% (with an LDL-C 70-189 mg/dL) according to the Pooled Cohort Equations calculator, which was created in conjunction with the guidelines. 1 For individuals meeting one of these four criteria, a moderate to high intensity statin is a Class I recommendation for individuals age 40 to 75 years. For individuals with diabetes or a 10-year CVD risk >7.5%, a risk-based discussion is recommended prior to consideration of statin therapy. For individuals with a 10-year CVD risk 5% to 7.5%, a risk-based discussion is recommended and a moderate intensity statin can be considered (Class IIa recommendation). Individuals with a 10-year CVD risk <5% are recommended to not undergo pharmacologic treatment to lower their cholesterol.
Patients were classified into statin eligibility categories in a stepwise fashion, first selecting individuals with a diagnosis of CVD, then individuals without known CVD but with a diagnosis of diabetes, then individuals without CVD or diabetes but with an LDL >190 mg/dL, and finally, individuals not meeting the three previous criteria but with a 10-year CVD risk ≥7.5% according to the Pooled Cohort Equations calculator. 1 For individuals with an identified race other than white or African American (ie, Asian/Pacific Islander, American Indian), the model developed for white populations was applied. The remaining individuals were stratified into two categories based on their 10-year CVD risk (5%-7.5% and <5%).
As a sensitivity analysis, we reanalyzed statin eligibility using estimated off treatment LDL-C levels for individuals on statin therapy.
We assumed a 30% increase in LDL-C for individuals on a low or moderate intensity statin and a 50% increase in LDL-C for individuals on a high intensity statin. Individuals were then reclassified into eligibility categories using the same stepwise fashion described above.

| Statistical analysis
Patient characteristics of the final sample were compared to those who were excluded due to missing data (primarily missing lipid panel) for each time period using chi-square for categorical variables and   Table 1. The cohort in 2017 was slightly older with a higher prevalence of coronary heart disease, kidney disease, and diabetes.
In 2013, 57.7% of patients met statin eligibility criteria, with the highest prevalence of statin eligibility due to ASCVD risk ≥7.5% (25%), followed by 16.5% with a clinical ASCVD diagnosis, 13.1% with F I G U R E 1 Study sample selection process for 2013 and 2017 data sets. ASCVD, atherosclerotic cardiovascular disease diabetes, and 3.1% with an LDL-C ≥190 (  were prescribed a statin compared to 57.4% of statin eligible women (P < .001). Women were less likely to receive treatment within each of the statin eligibility categories (Table 3).
A sensitivity analysis categorizing statin eligibility using estimated off-treatment LDL-C levels for those on statin therapy, assuming a 30% and 50% increase in LDL-C for a moderate and high intensity sta-

| DISCUSSION
In an analysis of approximately one-quarter of a million adults from a large Midwestern healthcare system, we found that the prevalence of    The trial reported no difference in muscle-related adverse events during the blinded portion of the trial but during the non-blinded portion of the trial, muscle-related adverse events in those on statin therapy were reported at a frequency 41% higher (P = 0.006) than those not on statin therapy. While these data demonstrate the significant contribution of the nocebo effect to statin intolerance and subsequent lack of appropriate statin use, other factors may be even more important.
The aforementioned PALM registry data also demonstrated that approximately 60% of statin eligible adults that were not on statin therapy reported never even being offered a statin by their provider. 8 Cost may be a potential limiting factor in statin use but nearly all statins are currently generic. From 2013 to 2017, the high intensity statins atorvastatin and rosuvastatin became generic, which is likely a significant driving factor for the increase use of high intensity from 2013 to 2017 seen in our study.

| Strengths and limitations
Our analysis has several strengths, including a contemporary analysis of statin utilization, including the specific statin and dosage, in a large sample of men and women from urban, suburban, and rural populations. Use of EHR data allowed identification of ASCVD diagnoses as well as other ASCVD risk factors, allowing for calculation of 10-year ASCVD risk and placement of patients into appropriate categories of statin eligibility. It also has potential limitations. Statin use is known to vary by geographic location, 16 making the generalizability of these findings to populations outside the Midwestern United States less reliable. A lack of historical lipid values likely led to misclassification of some individuals with LDL-C ≥190 mg/dL who were identified and treated prior to our assessment. However, as indicated in our sensitivity analysis, this likely affected 2% of the patient population. Additionally, due to missing data, we were unable to identify statin eligibility for approximately 30% of the patients within this healthcare system and we could only assess statin orders provided within the health system studied.

| CONCLUSION
In conclusion, we found that over one-third of statin eligible adults in a large Midwestern healthcare system are not currently prescribed a statin, with suboptimal use of high intensity statin therapy, ezetimibe, and PCSK9 inhibitors as well. Despite sound evidence supporting the recommendations in the ACC/AHA cholesterol guidelines, use of statin therapy remains suboptimal and strategies to improve guideline adherence hold the potential to have a significant impact on population rates of ASCVD events.