What is Ascertainment Bias & How To Prevent It

In the research world, it is vital to ensure that findings are as accurate and unbiased as possible. While several types of cognitive biases can affect the validity of research, one of the more subtle yet impactful biases is ascertainment bias. This can sometimes be confused with explicit bias or actor-observer bias, but it has its distinct characteristics.

Ascertainment Bias Example

Imagine researchers who want to study the effects of exercise on heart health. They decided to gather data by interviewing members at a local gym about their exercise habits and health.

Bias Introduced

1. People who regularly go to the gym might already be more health-conscious than the general population, leading to a sample that is not representative. They might also have an affinity bias towards a particular type of exercise or health regimen.
2. Those with severe health conditions or limited mobility may not go to the gym, so they are excluded from the sample, creating a ceiling effect where only a certain level of health is captured.
3. Gym-goers might also be more inclined to provide socially desirable responses about their exercise habits and health, given the setting. This can also be seen as a type of bias for action, where individuals act in a way that aligns with the expected norm of their environment.

Outcome of the Bias

If the researchers then conclude that regular exercise leads to significantly better heart health based solely on this data, they could be misled. They missed out on interviewing a significant portion of the population who might not exercise and have varying heart health levels. This skews the results and makes the effects of exercise seem more beneficial than they might be in a more representative sample.

What is Ascertainment Bias?

Ascertainment bias is a type of selection bias that arises when there is a non-random selection or non-representation of subjects or events for observation, which can lead to results that are not representative of the entire population. This typically means that the subjects or events that are observed or included in a study are not representative of the broader group from which they were selected.

A common example of ascertainment bias is in the context of genetic studies. If researchers only study families with a high prevalence of a certain disease, they might overestimate the genetic contribution to that disease because they didn’t account for unaffected families.

Ascertainment bias can lead to skewed results and misleading conclusions if not properly accounted for. It is a reminder that in research, both the inclusion and exclusion criteria for subjects or events can have a profound impact on the results.

Causes of Ascertainment Bias

Here are several causes of ascertainment bias:

Diagnostic Susceptibility

If individuals with certain characteristics or exposures are more likely to be diagnosed than others, it can result in ascertainment bias. For example, wealthier patients might undergo more health screenings and thus have diseases detected earlier or more frequently than less affluent individuals.

Selective Enrolment

Studies that enrol participants based on certain criteria or exposures might not be representative of the general population, which can introduce bias.

Loss to Follow-Up

In cohort studies, if participants drop out or are otherwise lost to follow-up in a non-random manner related to both the exposure and outcome, it can lead to biased results.

Screening Programs

Diseases that are routinely screened for (e.g., mammograms for breast cancer) might have higher reported incidences than diseases without such programs.

Awareness and Media Attention

Conditions that receive significant media attention might be more frequently diagnosed, simply because both doctors and patients are more aware of them.

Population-Based vs. Hospital-Based Studies

Ascertainment can differ greatly depending on where you’re looking. Hospital-based studies might over-represent more severe cases, while population-based studies can capture a broader range of disease severity.

Differential Recording of Information

If data about certain exposures or outcomes are recorded more diligently or consistently for one group over another, it can result in bias.

Referral Bias

In speciality clinics, the patients seen may not be representative of all people with that condition. This can skew the perception of the prevalence and severity of certain conditions.

Survival Bias

If only individuals who survive a condition long enough to be part of a study are included, it can bias results toward a less severe picture of the disease or condition.

Protopathic Bias

This occurs when the treatment for an early symptom of a disease appears to be associated with the disease itself because the disease has not yet been diagnosed.

Confirmation Bias in Diagnosis

If physicians have a hypothesis about what might be causing a patient’s symptoms, they might order tests that confirm that hypothesis and neglect other possible diagnoses.

Ascertainment Bias Examples

Here are some real-world examples to illustrate how this can play out:

ADHD Studies in Speciality Clinics

Studies conducted in speciality clinics might find a higher rate of Attention Deficit Hyperactivity Disorder (ADHD) symptoms or co-morbidities because individuals who go to such clinics often have more severe symptoms or have already been pre-selected because of a suspicion of ADHD. This would not be representative of ADHD in the general population.

Breast Cancer in Families

If researchers were to study the incidence of breast cancer among women who have a family member attending a breast cancer clinic, they might find a higher rate of breast cancer in that group, but this would be because the family already has a known risk.

Rare Disease Diagnoses

For rare diseases, if a study sample consists of individuals from speciality clinics or patient support groups, it might not represent the variety and full spectrum of the disease in the broader population.

Studies on Smoking and Lung Cancer in Hospitals

If a study was conducted by looking at hospitalised patients, it might find a very high association between smoking and lung cancer because a high proportion of people in the hospital with lung issues might be smokers. However, this would miss many smokers who have not yet developed severe health issues and are not hospitalised, thus potentially exaggerating the immediate risk.

Effects of Alcohol on Health

If researchers only studied the health outcomes of people who go to alcohol treatment centres, they might find a higher rate of alcohol-related health problems. This would miss moderate or light drinkers in the general population, potentially painting a bleaker picture of alcohol’s effects on health than what might be the case for moderate drinking.

Age-Related Studies

If a study on a particular condition (say, Alzheimer’s disease) only includes elderly individuals from a nursing home, it might miss milder cases or younger individuals with early-onset forms of the disease, skewing the perceived age of onset and severity.

Sudden Cardiac Death in Athletes

Media coverage might make it seem like sudden cardiac deaths are more common in athletes, leading to potential ascertainment bias in studies if they only focus on cases reported in the media.

Causes of Ascertainment Bias

Here are several causes of ascertainment bias:

Diagnostic Susceptibility

If individuals with certain characteristics or exposures are more likely to be diagnosed than others, it can result in ascertainment bias. For example, wealthier patients might undergo more health screenings and thus have diseases detected earlier or more frequently than less affluent individuals.

Selective Enrolment

Studies that enrol participants based on certain criteria or exposures might not be representative of the general population, which can introduce bias.

Loss to Follow-Up

In cohort studies, if participants drop out or are otherwise lost to follow-up in a non-random manner related to both the exposure and outcome, it can lead to biased results.

Screening Programs

Diseases that are routinely screened for (e.g., mammograms for breast cancer) might have higher reported incidences than diseases without such programs.

Awareness and Media Attention

Conditions that receive significant media attention might be more frequently diagnosed, simply because both doctors and patients are more aware of them.

Population-Based vs. Hospital-Based Studies

Ascertainment can differ greatly depending on where you’re looking. Hospital-based studies might over-represent more severe cases, while population-based studies can capture a broader range of disease severity.

Differential Recording of Information

If data about certain exposures or outcomes are recorded more diligently or consistently for one group over another, it can result in bias.

Referral Bias

In speciality clinics, the patients seen may not be representative of all people with that condition. This can skew the perception of the prevalence and severity of certain conditions.

Survival Bias

If only individuals who survive a condition long enough to be part of a study are included, it can bias results toward a less severe picture of the disease or condition.

Protopathic Bias

This occurs when the treatment for an early symptom of a disease appears to be associated with the disease itself because the disease has not yet been diagnosed.

Confirmation Bias in Diagnosis

If physicians have a hypothesis about what might be causing a patient’s symptoms, they might order tests that confirm that hypothesis and neglect other possible diagnoses.

Ascertainment Bias Examples

Here are some real-world examples to illustrate how this can play out:

ADHD Studies in Speciality Clinics

Studies conducted in speciality clinics might find a higher rate of Attention Deficit Hyperactivity Disorder (ADHD) symptoms or co-morbidities because individuals who go to such clinics often have more severe symptoms or have already been pre-selected because of a suspicion of ADHD. This would not be representative of ADHD in the general population.

Breast Cancer in Families

If researchers were to study the incidence of breast cancer among women who have a family member attending a breast cancer clinic, they might find a higher rate of breast cancer in that group, but this would be because the family already has a known risk.

Rare Disease Diagnoses

For rare diseases, if a study sample consists of individuals from speciality clinics or patient support groups, it might not represent the variety and full spectrum of the disease in the broader population.

Studies on Smoking and Lung Cancer in Hospitals

If a study was conducted by looking at hospitalised patients, it might find a very high association between smoking and lung cancer because a high proportion of people in the hospital with lung issues might be smokers. However, this would miss many smokers who have not yet developed severe health issues and are not hospitalised, thus potentially exaggerating the immediate risk.

Effects of Alcohol on Health

If researchers only studied the health outcomes of people who go to alcohol treatment centres, they might find a higher rate of alcohol-related health problems. This would miss moderate or light drinkers in the general population, potentially painting a bleaker picture of alcohol’s effects on health than what might be the case for moderate drinking.

Age-Related Studies

If a study on a particular condition (say, Alzheimer’s disease) only includes elderly individuals from a nursing home, it might miss milder cases or younger individuals with early-onset forms of the disease, skewing the perceived age of onset and severity.

Sudden Cardiac Death in Athletes

Media coverage might make it seem like sudden cardiac deaths are more common in athletes, leading to potential ascertainment bias in studies if they only focus on cases reported in the media.

Autism Spectrum Disorder (ASD) in Advanced Parental Age

Some studies have shown an association between advanced parental age and an increased risk of ASD in children. However, if researchers were to mainly include children diagnosed at speciality clinics that cater to older parents, the association might appear stronger than it is in the broader population.

Some studies have shown an association between advanced parental age and an increased risk of ASD in children. However, if researchers were to mainly include children diagnosed at speciality clinics that cater to older parents, the association might appear stronger than it is in the broader population.

How to Avoid Ascertainment Bias

Avoiding ascertainment bias is crucial for maintaining the integrity of research and ensuring that findings are accurate and representative. Here is how to avoid it:

• Before data collection begins, ensure you have a clear and comprehensive study design. Refer to a scholarly source or a primary source to base your foundational understanding.
• Whenever possible, use random sampling to select participants. This reduces the chance that a particular subgroup will be over- or under-represented.
• If feasible, blind both the researchers and the participants. In a double-blind study, neither the participant nor the researcher knows who is receiving the treatment or intervention, which can prevent bias in data collection.
• Use multiple data sources or methods to validate your findings. This could include both primary and secondary sources. Cross-referencing can also prevent publication bias, where only certain types of results get published.
• Ensure that data collection methods are exhaustive and that you’re capturing all necessary information, not just the most accessible or obvious.
• In longitudinal studies, be cautious of differential dropout rates among study groups, which can introduce bias.
• Regularly train data collectors and researchers on the importance of unbiased data collection.
• When collecting data, use standardised questionnaires or tools that have been validated for your population of interest.
• Always consider populations or groups that might be under-represented due to factors like socio-economic status, race, gender, or location. Make efforts to include them in your study.
• Conduct pilot studies to test your methods and identify potential sources of ascertainment bias.
• When publishing your results, be transparent about your methods and any potential sources of bias. Being transparent can also help avoid the Pygmalion effect, where researchers might inadvertently influence outcomes based on their own expectations.
• Continuously review and refine your methods throughout the research process to identify and address potential biases.
• Before publishing, subject your findings to peer review. Feedback from colleagues can help identify overlooked biases.
• Establish mechanisms for participants or data collectors to provide feedback on the process, which might highlight unintended sources of bias.
• Engage external experts or stakeholders who can provide an outsider’s perspective on potential sources of bias in your study.

Frequently Asked Questions