[ evidence in practice ]

STEVEN J. KAMPER, PhD1

Bias: Linking Evidence

Downloaded from www.jospt.org at The University of British Columbia Library on August 1, 2018. For personal use only. No other uses without permission.

With Practice
J Orthop Sports Phys Ther 2018;48(8):667-668. doi:10.2519/jospt.2018.0703

I
f you ask an epidemiologist for a definition of bias, you will get let in study 2, which means the estimate
something like, “A systematic deviation of the sample parameter of treatment effectiveness may be biased.
estimate from the population value.” Unless you are also an The randomization in study 1 would over-
come such risk of selection bias.
epidemiologist, this is unlikely to be helpful. Broken down into
Performance Bias  Often, the people in-
more accessible language, it would look accuracy of a diagnostic test could lead a volved in a study, particularly therapists
something like, “A difference [deviation] clinician to weight the results of that test and participants, will have a “horse in the
Copyright © 2018 Journal of Orthopaedic & Sports Physical Therapy®. All rights reserved.

in a particular direction [systematic] be- more highly than is warranted. Informa- race,” that is, a belief that one or another
tween the results of a study [parameter tion from a study with a biased estimate treatment is better and therefore a desire
estimate] and what happens in real life of prognosis could lead to inaccurate ad- for the study results to fall a certain way.
[population value].” vice and expectations. For example, a physical therapist might
Put another way, bias means that the hold a strong belief that exercise is neces-
information in front of us gives an incor- Some Types of Bias sary for these patients, and should not be
rect impression of what is really going on. There are many, many different types of replaced by a simple information book-
Note that bias is different from er- bias. A few that apply to research (and let. Similarly, a patient might be quite
ror. Random error occurs because the sometimes to clinical experience) are disappointed if he or she receives just a
Journal of Orthopaedic & Sports Physical Therapy®

way we measure things—whether by below. booklet instead of an exercise program.

observation, questionnaires, interviews, Selection Bias Consider 2 studies, (1) If this is the case, then the physical thera-
measurement tools, or images—is never a randomized controlled trial in which pist might not deliver the 2 interventions
perfectly accurate (FIGURE). Later Evi- patients with back pain are randomly al- with equal enthusiasm and confidence,
dence in Practice articles on measure- located to receive either an exercise inter- and participants might not read and
ment will discuss this issue. vention or an information booklet, and follow the information in the booklet.
Any information can be biased, (2) a comparative study in which patients Therapist blinding and patient blinding
whether it comes from research, clini- with low back pain receive either exercise overcome performance bias, but blinding
cal experience, professional courses, col- or the booklet, based on the choice of a is often difficult or impossible in physical
leagues/peers, the media, etc. The first treating clinician. Estimates of compara- therapy trials. The risk of performance
Evidence in Practice article in the series tive effectiveness can be derived from ei- bias can be reduced by ensuring that the
explained some of the biases related to ther study by looking at the difference in
information from clinical experience. outcome between the exercise group and
Bias tends to overestimate effects. For the booklet group. It is not difficult to
example, treatments appear more effec- imagine that the treating clinician would
tive than they will be in practice, diagnos- choose the exercise intervention in study
tic tests appear more accurate than they 2 for the patients the clinician thought
Low bias
really are, and prognostic factors more would benefit most from exercise; after all, High bias
High error
Low error
strongly predict outcomes. In the clinic, that’s what would happen in practice. But
it is easy to see how overestimation of the there are likely to be important differences Truth Information
effectiveness of a treatment could influ- between the people who get the exercise
FIGURE. Bias and error.
ence management. Overestimation of the intervention and those who get the book-

School of Public Health, University of Sydney, Camperdown, Australia; Centre for Pain, Health and Lifestyle, Australia. t Copyright ©2018 Journal of Orthopaedic & Sports Physical
1

Therapy®

journal of orthopaedic & sports physical therapy | volume 48 | number 8 | august 2018 | 667

 [ evidence in practice ]
control intervention appears as credible bias, researchers attempt to reduce the high follow-up rate (greater than 85%)
and worthwhile as the index intervention. risk of detection bias by ensuring that and conducting intention-to-treat analy-
Detection Bias The beliefs that the the intervention and control treatments sis are ways of reducing the risk of attri-
Downloaded from www.jospt.org at The University of British Columbia Library on August 1, 2018. For personal use only. No other uses without permission.

treating clinicians and participants have are equally credible. tion bias.
about the treatments can also result in Attrition Bias No matter how well a
bias when it comes to measuring out- study is run, there will be participants Summary
comes. If a treating clinician is collect- who drop out and others who do not All studies are at some risk of bias. The
ing outcome measures and knows which complete all the follow-up measures. idea of assessing risk of bias is to provide
treatment a participant received, then he The fundamental issue is that we do not a basis for giving more or less weight to
or she may subconsciously interpret or know what happened to the people who that piece of information: if a study is at
score outcomes in a way that favors the did not complete all the outcome mea- higher risk of bias, then we should be less
intervention group. If a participant feels sures. We cannot just assume that, on confident about the findings. The same
discouraged about being in the control average, they are the same as the people goes for information from clinical expe-
group, then he or she may form a more who stayed in the study. The likelihood rience. When risk of bias is low, we think
negative appraisal of his or her symp- that dropouts and missing data introduce the information is closer to “the truth”;
toms, or express disappointment with bias into a study depends on a number when bias is high, we think the opposite.
scores on self-reported outcomes. Detec- of factors. These include the proportion Clinical practice involves collecting rel-
tion bias is overcome by blinding the out- of participants who completed the study, evant information from various sources
come assessors. However, this is difficult the balance of dropouts between groups, and boiling it down to make a decision.
Copyright © 2018 Journal of Orthopaedic & Sports Physical Therapy®. All rights reserved.

for patient-reported outcomes, as, es- the comparability of dropouts and com- In doing so, you should pay more atten-
sentially, the patient is also the outcome pleters, and the way that the statistical tion to information that is less likely to
assessor. As in the case for performance analyses were performed. Ensuring a be biased. t
Journal of Orthopaedic & Sports Physical Therapy®

GO GREEN By Opting Out of the Print Journal

JOSPT subscribers and APTA members of the Orthopaedic and Sports
Physical Therapy Sections can help the environment by “opting out” of
receiving JOSPT in print each month as follows. If you are:

• A JOSPT subscriber: Email your request to [email protected] or call

the JOSPT office toll-free at 1-877-766-3450 and provide your name
and subscriber number.

• APTA Orthopaedic or Sports Section member: Go to

http://www.apta.org/, log in, and select My Profile. Next click
on Email Management/GoGreen. Toward the bottom of the list, you
will find the Publications options and may opt out of receiving
the print JOSPT. Please save this preference.

Subscribers and members alike will continue to have access to JOSPT

online and can retrieve current and archived issues anytime and anywhere
you have Internet access.

668 | august 2018 | volume 48 | number 8 | journal of orthopaedic & sports physical therapy