Bias in Prognostic Studies

In the evidence-based world every now and again you may come across a paper looking at prognosis. Such papers provide us information on understanding the natural history of a disease or condition, which may then be of use in understanding the nature of your specific patient’s illness (or of those patients who may later relate to the coursework you are teaching now). Knowing something about prognosis allows us to provide information to the patient about predicted outcomes of treatment, and for some conditions (such as, say, the common cold), we know a great deal. Conditions such as back pain are less well understood and therefore rife for discussion in a prognostic paper. But there are several kinds of bias which affect prognostic studies. These include:

Lead Time Bias: this occurs when earlier results related to diagnosis of a condition suggest a treatment effect even though the patient has not, for example, lived any longer. In reviewing studies of prognosis, we need to ensure that our inception cohort had a clearly defined inception point so that they all have the same starting point. Look to see that this is presented in the methods section.

Centripetal Bias: This occurs when a referral center has such a good reputation that it attracts people from outside its geographic location. This then affects the characteristics of the patients that center sees, since, for example, only those who can afford to travel there will be included in an analysis of the patients being seen. Effects or poverty would then be taken out of understanding how this disease progresses.

Popularity Bias: This happens when people with certain specific diagnoses are treated differently than those with other diagnoses or conditions. This may affect the patient flow in a tertiary care center known for its management of that specific condition, affecting the nature of the patients they then see and making them not reflective of the condition at it typically presents in a population.

Referral Bias: This occurs when a select group of patients is referred for study. The problem here is that this leads to that sample not being random or even a reasonable sample of the condition of interest.

Diagnostic Access Bias: If your starting point for conducting a prognostic study requires access to, say, an MRI imaging center, does your community even have one? Does each potential participant in the study have equal access to such a center? Some require referral and some require lengthy waits. This may affect the types of patient seen, making them atypical of the condition of interest.

Diagnostic Suspicion Bias: If the physician can identify the group assignment of a patient, he or she may treat that patient differently.

Expectation Bias: This occurs when a physician second-guesses what is happening in a trial due to prior knowledge he or she may have. The more information about each patient we have, the more our expectations may influence what we do and how we treat the patient, affecting prognosis.

These are sources of error in prognostic studies you should be aware of. Prognostic studies are not common in chiropractic, but are likely to occur in the future.