Predictors of statin adherence in primary care using real-world data

Using real-world data from 3,432 primary care patients, this study found that statin adherence in the first two years after prescription was significantly higher among older, socioeconomically advantaged individuals with more comorbidities and polypharmacy, while being lower among smokers and those prescribed multiple statin types.

The Gist

Imagine your body as a car that needs to run smoothly for decades. Statins are like the high-quality oil you put in the engine to prevent it from seizing up (which, in human terms, prevents heart attacks and strokes). But here's the problem: even if the doctor gives you the best oil, it doesn't do any good if you forget to pour it in, or if you decide to stop pouring it halfway through.

This study is like a giant detective investigation into why some drivers keep their engines oiled perfectly, while others let them run dry. The researchers looked at real-life data from over 3,400 patients in South-West Sydney to figure out who sticks to their medication and who doesn't.

Here is the breakdown of their findings, translated into everyday language:

The Big Picture: Who is the "Good Driver"?

The study found that 72% of patients were "good drivers." They took their statin pills consistently enough to be considered "adherent" (meaning they had the medicine in their system on at least 80% of the days over two years).

But what makes a good driver? The study found several surprising patterns:

1. The "Older & Wiser" Effect (Age)

The Finding: Older people (over 65) were much more likely to take their pills than younger people.
The Analogy: Think of a young driver who thinks, "I'm invincible, I don't need to check my oil yet." An older driver, however, knows that the engine has been running for a long time and knows that skipping a refill could lead to a breakdown. Older patients seem to understand the stakes better and are more disciplined about their "engine maintenance."

2. The "Team Player" Effect (Comorbidities & Polypharmacy)

The Finding: People who were already taking other medicines for other conditions (like diabetes or high blood pressure) or who had multiple health issues were more likely to stick to their statins.
The Analogy: This seems counterintuitive, right? Usually, we think taking 10 pills is a nightmare. But imagine a person who is already used to a complex "pit crew" routine. If you are already taking a pill for your blood sugar and a pill for your blood pressure, adding one more pill for your cholesterol feels like just another stop at the gas station. You are already in the habit of checking your dashboard. The "non-adherent" group was often the "lone wolf" who only had one issue and didn't feel the urgency to build a routine.

3. The "Neighborhood" Factor (Socioeconomics)

The Finding: People living in areas with slightly better resources (but not the richest) were more likely to take their meds.
The Analogy: Think of this like the neighborhood you live in. If you live in a place where the roads are well-paved, the gas stations are open 24/7, and the mechanics speak your language, it's easier to keep your car running. People in areas with fewer resources often face "roadblocks" like cost, language barriers, or just not having a reliable way to get to the pharmacy.

4. The "Confusion" Trap (Too Many Options)

The Finding: Patients who were switched between different types of statins (e.g., from Atorvastatin to Rosuvastatin) were less likely to stick with it.
The Analogy: Imagine you are driving a car, and every week the mechanic swaps your engine for a different brand. One week it's a Ford, the next it's a Toyota. You start to get confused about how to drive it, or maybe the new engine feels weird. When patients are switched between different statin brands or types, it creates a sense of instability. They might think, "Maybe this one isn't working," and stop taking it.

5. The "Habit" Clash (Smoking)

The Finding: Smokers were less likely to take their statins.
The Analogy: This is like trying to fix a leak in your roof while you are still smoking inside the house. It's a clash of behaviors. Smoking is often a sign of a lifestyle that prioritizes immediate gratification over long-term health. If someone is ignoring the "smoke alarm" (their lungs), they might also ignore the "check engine" light (their cholesterol).

The "Missing Data" Mystery

The study also noticed that when doctors didn't write down whether a patient smoked or not, those patients were less likely to take their meds.
The Analogy: This isn't because "not smoking" is bad. It's like a mechanic who doesn't write down the car's mileage. If the mechanic doesn't bother to check the details, it suggests the car isn't being taken seriously. If a doctor doesn't record a patient's smoking status, it might mean the patient isn't engaging deeply with their healthcare team, which leads to them skipping their pills.

The Bottom Line

This study tells us that taking a pill isn't just about willpower; it's about habits, environment, and clarity.

• To help people take their meds: We shouldn't just tell them "it's good for you." We need to help them build a routine (like the "Team Players"), make sure they aren't confused by switching medicines (keep the "engine" consistent), and ensure they have access to the "gas stations" (pharmacies) they need.
• The Takeaway: The best way to keep the heart engine running is to treat the whole driver, not just the engine. If we understand that older people, people with multiple conditions, and people with good support systems are the most likely to succeed, we can design better systems to help everyone else catch up.

Technical Summary

1. Problem Statement

Cardiovascular diseases (CVD) represent a significant burden in Australia, contributing to approximately 12% of the total disease burden and over 57,000 acute coronary events annually. Statins are the cornerstone of primary and secondary CVD prevention. However, adherence to statin therapy is suboptimal, with studies suggesting up to 30% of prescribed patients are non-adherent. Non-adherence is linked to higher morbidity, mortality, and increased healthcare costs.

While various demographic and clinical factors (age, gender, comorbidities, socioeconomic status) have been hypothesized to influence adherence, findings in the literature are often inconsistent. Furthermore, many existing studies rely on administrative claims data (e.g., Pharmaceutical Benefits Scheme) which may not capture the nuances of clinical practice or patient behavior in real-world primary care settings. This study aims to identify robust predictors of statin adherence in a diverse Australian primary care population using Electronic Health Record (EHR) data over a two-year period.

2. Methodology

• Data Source: The study utilized the Electronic Practice Based Research Network (ePBRN) Linked Dataset, derived from 11 general practice facilities in the Fairfield and Wollondilly regions of South-West Sydney. The dataset covers approximately 249,345 patients with visits recorded between 2012 and 2019.
• Study Population & Selection Criteria:
  • Inclusion: Patients aged ≥18 years initiating statin therapy (index date), with no prior statin history, and at least three GP visits over a two-year follow-up period.
  • Exclusion: Patients >80 years, those with missing gender/birth year, those who died during the observation period, those with CVD events before or during the 2-year window, and those with a Proportion of Days Covered (PDC) >130% (to avoid misclassification from stockpiling).
  • Final Cohort: 3,432 patients (57,227 visits) met the criteria.
• Outcome Measure:
  • Adherence: Calculated using a modified Proportion of Days Covered (PDC) formula over the first two years post-index date.
  • Threshold: Patients with PDC ≥ 80% were classified as "adherent"; those < 80% were "non-adherent."
  • Calculation: The numerator utilized data on repeats, quantity, dose, and frequency from the second-to-last prescription, assuming continuity until the last visit. The denominator was the total days between the index date and the last visit.
• Predictors (Exposures):
  • Socio-demographic: Age, gender, ethnicity (ATSI status), SEIFA IRSAD (socioeconomic index), employment status, smoking status.
  • Clinical/Health-related: Statin intensity (low/moderate/high), statin type, number of statin types prescribed, changes in intensity, polypharmacy count (anti-diabetic, anti-hypertensive, anticoagulant, others), and comorbidity count (diabetes, hypertension, respiratory disease, cancer, etc.).
• Statistical Analysis:
  • Descriptive: Summarized characteristics by adherence status.
  • Multivariate Logistic Regression: A backwards stepwise approach was used to identify significant predictors (p < 0.05). Adjusted Odds Ratios (AOR) with 95% Confidence Intervals (CI) were calculated.
  • Validation: A bootstrap simulation (1,000 iterations) confirmed variable selection.
  • Sensitivity Analysis: Performed using the Akaike Information Criterion (AIC) stepwise model selection to ensure model parsimony and robustness.
  • Software: All analyses were conducted using R.

3. Key Contributions

• Real-World Evidence (RWE): Unlike studies relying on dispensing claims, this study leverages granular EHR data from general practices, capturing prescription details, visit reasons, and clinical context directly from the point of care.
• Comprehensive Predictor Analysis: The study simultaneously evaluates a wide range of factors, including the often-overlooked impact of "number of statin types" and specific socioeconomic indices (SEIFA) within a culturally diverse Australian cohort.
• Methodological Rigor: The application of bootstrap simulations and AIC-based sensitivity analysis strengthens the reliability of the identified predictors, addressing potential overfitting in the regression model.

4. Results

• Adherence Rates:
  • Mean PDC: 91.6% (±22.2%).
  • Adherent Population: 72.0% of patients (2,471/3,432) achieved PDC ≥ 80%.
• Significant Predictors (Multivariate Analysis):
  • Positive Associations (Increased Adherence):
    • Age: Patients >65 years had significantly higher adherence (AOR 1.7; 95% CI 1.4–2.0).
    • Socioeconomic Status: Higher SEIFA IRSAD categories (specifically category 3 vs. 1) were associated with better adherence (AOR 1.8; 95% CI 1.2–2.6).
    • Polypharmacy: Taking 3 or more additional medications was positively associated with adherence (AOR 1.8 for 3 drugs; 95% CI 1.3–2.5).
    • Comorbidities: Presence of comorbidities (e.g., diabetes, hypertension) increased adherence odds (AOR 1.4; 95% CI 1.1–1.7).
  • Negative Associations (Decreased Adherence):
    • Number of Statin Types: Patients prescribed 2 different statin types had lower adherence (AOR 0.6; 95% CI 0.5–0.9).
    • Smoking Status: Unrecorded smoking status was negatively associated with adherence (AOR 0.7; 95% CI 0.6–0.9), interpreted as a marker for missing clinical engagement rather than a biological effect.
• Non-Significant Factors: Gender, specific statin type (e.g., atorvastatin vs. rosuvastatin), and statin intensity did not show statistically significant associations in the adjusted model.
• Model Performance: The logistic regression model showed modest discriminative ability (AUC = 0.6465), suggesting that while identified predictors explain some variance, other unmeasured factors influence adherence. The sensitivity analysis confirmed the stability of the main predictors.

5. Significance and Implications

• Paradox of Multimorbidity: Contrary to the common assumption that complex regimens (polypharmacy) reduce adherence, this study found a positive association. This suggests that patients with multiple conditions may be more vigilant due to higher perceived risk and closer clinical monitoring, or that integrated care plans for multimorbid patients are effective.
• Impact of Regimen Complexity: The negative association with the "number of statin types" highlights that switching medications or managing multiple statin options can disrupt adherence, likely due to confusion or adverse effects.
• Socioeconomic Nuance: The finding that higher SEIFA categories correlate with better adherence reinforces the role of social determinants of health, though the specific relationship varied by category, indicating a complex interplay between local service delivery and population characteristics.
• Clinical Recommendations:
  • Clinicians should prioritize consistent statin regimens to avoid the adherence drop associated with switching types.
  • Tailored interventions are needed for younger patients (<65) and those with unrecorded smoking status, who appear at higher risk of non-adherence.
  • Integrated care for multimorbid patients should be maintained and strengthened, as these patients currently demonstrate good adherence.
• Limitations: The study acknowledges potential overestimation of adherence due to EHR prescription data (which may reflect stockpiling rather than actual consumption) and selection bias (excluding patients with early CVD events). The generalizability is limited to the South-West Sydney demographic.

In conclusion, this study provides critical insights into the real-world drivers of statin adherence, emphasizing that adherence is driven by an aging, multimorbid population engaged in polypharmacy, but is threatened by regimen instability (multiple statin types) and specific behavioral markers (smoking status).