Unmasking Bias in HEOR: How Hidden Errors Impact RWE

Bias in HEOR or RWE (Real-world evidence ) refers to systematic errors that can distort findings, leading to incorrect conclusions about healthcare interventions, costs, or patient outcomes. Bias can occur at various stages, from study design to data collection, analysis, and interpretation.

Bias can sneak into HEOR studies like uninvited guests at a party! Let’s break down some common biases with fun, relatable examples.

Types of Bias in RWE Study design

Selection Bias

• Happens when the study population is not representative of the target population.
  • What happens? You want to study the cost-effectiveness of gym memberships in improving health, so you only survey people who already go to the gym.
  • Why is this biased? You’re missing the perspective of couch potatoes who might have different health outcomes!
• Clinical Example: A new cancer drug is tested only in younger, healthier patients. The results look great! But when given to older patients with pre-existing conditions, the drug isn’t as effective. The study doesn’t represent the real-world patient population

Confounding Bias

• Arises when an outside factor influences both the exposure and outcome, leading to misleading conclusions.
  • What happens? A study finds that people who drink coffee daily have higher productivity
  • Why is this biased? Is coffee the magic potion? Not necessarily! Maybe coffee drinkers also sleep less, exercise more, or have high-stress jobs—factors that confound the results!
• Clinical Example: Early studies once suggested that alcohol consumption led to lung cancer. But later, researchers realized that many drinkers in the study were also smokers! Smoking was the real cause, not alcohol.

Measurement Bias

• Occurs when data collection methods systematically misclassify exposure or outcomes.
  • What happens? A university surveys students about their study habits, and 90% claim they study 5+ hours a day.
  • Why is this biased? Reality check! Some may over-report their study time to sound responsible (or to impress the professor). Self-reported data isn’t always accurate!
• Clinical Example: A doctor asks patients to self-report their pain on a scale of 1 to 10. Some patients exaggerate, while others downplay their pain. This misclassification can affect treatment decisions and pain management strategies

Publication Bias

• When studies with positive or significant results are more likely to be published than those with negative or neutral findings.
  • What happens? A movie studio only releases glowing reviews of its latest film while hiding the bad ones.
  • Why is this biased? Only favorable results are shared, while negative or neutral ones are hidden. It distorts reality, making the movie (or a treatment in healthcare research) seem better than it really is.
• Clinical Example: A pharmaceutical company only publishes studies where its drug performs well but buries trials with negative or neutral results. This distorts the medical literature, making the drug seem more effective than it really is!

immortal time bias

• Occurs when a period during which an individual cannot experience the outcome (e.g., death) is misclassified, making a treatment or intervention appear more effective than it actually is.
  • What happens? You run a study on how long superheroes live but only include those who are still alive (like Superman and Wonder Woman)
  • Why is this biased? You forgot to account for heroes who didn’t make it (RIP Iron Man 😢). This is immortal time bias, where excluding early deaths makes survival look better than it is.
• Clinical Example: A study on a new heart medication excludes patients who died early from analysis. Since only survivors are analyzed, the drug looks like it improves survival—but in reality, early deaths weren’t counted!

Survival bias

• Arises when a study only includes individuals who have survived a condition or treatment, ignoring those who did not, leading to overly optimistic conclusions about effectiveness or outcomes. Leads to overly positive conclusions because the study ignores those with worse outcomes
  • What happens? You see 1960s cars still running and think, “They made cars better back then!
  • Why is this biased? You’re only seeing the survivors—many broke down and were scrapped! This creates a false belief that all 1960s cars were more durable.
• Clinical Example: A study finds that patients who undergo a high-risk surgery live longer than those who don’t. But it ignores patients who were too sick to qualify for surgery and died earlier, making the procedure seem more beneficial than it really is.

Attrition bias

• Happens when participants drop out of a study, and their outcomes are not accounted for, leading to misleading conclusions
  • What Happens? You run a reality TV show where contestants follow a strict fitness program for six months. At the end, those who stayed lost significant weight! ️‍️
  • What’s Wrong? You ignored the contestants who quit early—maybe they dropped out because the program was too hard, ineffective, or had side effects. By only analyzing those who stuck with it, the program looks more successful than it really is.
• Clinical Example: A study tests a new diabetes drug, but many patients drop out due to side effects. The final analysis only includes those who remained, making the drug seem safer and more effective than it actually is!

Key Takeaways

• Bias can distort healthcare decisions just like it skews everyday scenarios.
• Recognizing bias helps researchers and clinicians make better, evidence-based choices.
• Better study designs, real-world data, and transparency can help minimize bias in HEOR!

To ensure accurate results, researchers must identify, adjust for, and minimize bias wherever possible. After all, good science isn’t just about collecting data—it’s about interpreting it correctly!