Understanding bias in healthcare data
Bias refers to a conscious or unconscious inclination towards or against an individual, a group, or an object. Such biases can influence decision-making and skew data in any domain. In the field of healthcare, such biases pose a significant risk to people’s lives. In this blog, we will explore three common types of data bias focusing particularly on examples from healthcare data.
1. Sampling bias
If the sample under study is not representative of the whole population, we encounter a sample bias. Consider a situation where a pharma company is conducting clinical trials to test the efficacy of a vaccine. Younger individuals are more motivated and perceive lower risk in participating in clinical trials. If the trial includes mostly younger adults and excludes older individuals who are more vulnerable to the disease, the vaccine’s efficacy for the elderly may be unknown. You need all sides of a story to avoid sampling bias. Conclusions made from such data can lead to misleading outcomes as the data is age-biased and the study ignored the vaccine response on elderly in the population.
2. Observer Bias
Different people have a tendency to observe things differently due to which data can become biased. Human observation is a critical step in data collection. The initial observations recorded make the foundation for further study. Monitoring blood pressure is a regular practice in healthcare. However, observer bias might happen during manual blood pressure monitoring. Because the pressure meter is so sensitive healthcare workers often get pretty different results. Usually, they will just round up to the nearest whole number to compensate for the marginal error. But if doctors consistently round up and down the blood pressure readings of their patients, health conditions may be messed and any studies involving these patients wouldn’t have precise and accurate data.
3. Confirmation Bias
People see what they want to see. The tendency to search for or interpret information in a way that confirms pre-existing beliefs is a confirmation bias. Someone can be so eager to confirm a gut feeling that they only notice things that support it ignoring all other signals. Let us understand this with an example. Suppose a pharma company is conducting clinical trials to test the efficacy of a vaccine and recording the response of the participants after 30 minutes of vaccination. The healthcare workers are so sure that the vaccination is safe that they tend to ignore any potential side effects that might have occurred after the vaccination and let them unnoticed. Conclusions drawn from such biased perception might be life-threatening.
I think now you are well aware of how biases in data can distort findings and pose a substantial risk to the public. When working with data, make sure to diligently assess and address any potential sources of bias that may affect the data.
