A set of training materials for professionals working in intervention epidemiology, public health microbiology and infection control and hospital hygiene.
Need help with your investigation or report writing? Ask the Expert. Free advice from the professional community.
You can't make decissions on this page's approval status because you have not the owner or an admin on this page's Group.
If an epidemiological study is viewed as a measurement exercise , then we need to consider how much we can trust the measurement (risk, rate, effect) obtained from that study. Can we use it to safely describe (accurately estimate) the association between an exposure (potential causal characteristic ) and a disease/outcome, or to conclude that a risk factor really does cause the disease/outcome in the population in which the study was done?
Bailey et al  define an association as a 'statistical dependence between two or more events, characteristics, or other variables'. According to Rothman , a measure of association compares what happens in two distinct populations (or sub-populations), although these two populations may correspond to one population in different time periods (e.g. before and after an event). Relative measures of association (e.g. relative risk/ risk ratio, rate ratio, odds ratio) estimate the size of an association between exposure and disease/outcome (strength of association), and indicate how much more likely people in the exposed group are to develop the disease/outcome than those in the unexposed group . The presence of an association does not necessarily imply a causal relationship.
Before we can conclude that an observed association between an exposure (risk factor) and outcome (e.g. disease) as measured in our study is causal, and may reflect the true situation in the population, we first need to exclude other possible reasons why we might have obtained that result and be sure that the measurement/ result has been estimated with little error. We need to consider whether the result could be due to systematic error (bias or confounding) or to random error (due to chance). If we consider that the results reflect the true situation in the population, they then need to be interpreted according to causality criteria.
Random error reflects the amount of variability in the data . Assessment of random error aims to distinguish findings (variations of observed values from the true population values) due to chance alone (findings that we cannot readily explain) from findings we could replicate if we repeat the study many times. Precision is the opposite of random error, and an estimate with little random error can be described as precise .
In epidemiological studies, biases are systematic errors that result in incorrect estimates when measuring the effect of exposure on risk of disease/outcome. Any error that results in a systematic deviation from the true association between exposure and outcome is a bias . Validity is the opposite of bias, and an estimate that has little systematic error can be described as valid . Biases may distort the design, execution, analysis and interpretation of studies . Some authors define bias more broadly. Daly's definition - defining bias as any factor or process that tends to produce results or conclusions that differ systematically from the truth - thus includes errors in analytical epidemiology and errors of interpretation .
As described in the following graphic adapted from Rothman , there is a way to distinguish random errors from systematic errors. If we increase the size of a study until it is infinitely large (increase our sample size), random errors (due to chance) can be reduced to zero and corrected for. However, systematic errors (biases) are not affected by increasing the size of the study and will remain.
In this chapter, we will focus on systematic errors (bias). Epidemiologists frequently classify bias into three broad categories: selection bias (bias in the way the study subjects are selected), information bias (bias in the way the study variables are measured), and confounding (described in a specific chapter).
Selection biases in case-control studies include among others: case ascertainment (surveillance) bias, referral bias, diagnostic bias, non-response bias, and survival bias. Selection biases in cohort studies include: healthy worker effect, diagnostic bias, non-response bias, and loss to follow-up.
The term "misclassification" is frequently used to describe information bias, the mechanism of which can be differential or non-differential (random). Misclassifications might be introduced by the observer (interviewer bias, biased follow-up), by the study participants (recall bias, prevarication), or by measurement tools (e.g. questionnaires).
1. Rothman KJ. Epidemiology - An Introduction. New York: Oxford University Press; 2002.
2. Rothman KJ, Greenland S, Lash TL, editors. Modern Epidemiology. 3rd ed. Philadelphia: Lippincott Williams & Wilkins; 2008.
3. Bailey L, Vardulaki K, Langham J, Chandramohan D. Introduction to Epidemiology. Black N, Raine R, editors. London: Open University Press in collaboration with LSHTM; 2006.
4. Sackett DL. Bias in analytic research. J Chronic Dis. 1979;32(1-2):51-63.
5. Daly LE, Bourke GJ. Interpretation and Uses of Medical Statistics. 5th ed. Blackwell Science.
Join the discussion about this article in the forum!
Mike Catchpole posted on 3/30/2010 11:54:18 AM:
The definitions of different types of bias should be reconciled with the definitions included in the ECDC internal lexicon (ie that developed by Laszlo)
Vladimir Prikazsky replied on 6/4/2010 8:36:54 AM:
Current status of the terminology services - answer from Laszlo:
The current version as such is running behind the firewall, within our internal network. However, as it has a machine interface, there could be a method to feed a list in the wiki with proper terms queried from the Term Server and updated regularly. This will need some work to tag our categories with the attribute `is required by fem wiki`. That way the users of the wiki could consult the derived lists in the wiki itself.
For this we have discuss what term sets are needed. Now a systematic audit of all systems / applications is ongoing to check their information objects regarding the core metadata standard implementation. We could join these two tasks in a single work process.
*** I hope this answer is promising. However at present we have to refer to books or other ready to use services.
Arnold Bosman replied on 8/6/2010 12:07:47 PM:
Reading through the chapter again, I was wondering if it could be useful to add the 2x2 tables that we have as examples in the EPIET Powerpoint lecture for each form of bias. What do you think?
Aileen Kitching replied on 9/7/2010 2:03:18 PM:
thanks for the suggestion! and apologies for the delayed reply. I have added the 2x2 tables from the EPIET lectures we had in Menorca (these still need some editing, up/down arrows put in etc), and some other examples also from the lectures, and from Rothman and other books/ articles.
Please let me know if you would like an changes made to the additions to the text/ edits of the text.
All the best,
Aileen Kitching replied on 9/7/2010 2:06:12 PM:
This is a question for Arnold & Vladimir :-).
In the original text of my chapter 'biases in epi studies' , it says that random error will be dealt with in a separate chapter. I realise that not all chapters are finished yet, but was wondering which chapter will deal with random error, confidence intervals, p-values etc?
Derval Igoe replied on 9/22/2010 10:29:50 AM:
This chapter is well written, accurate and clear overall. I have made some editing suggestions. (My machine crashes each time I am in the information bias section and try to edit, so I havent edited from there on - not sure if this is a local, or global issue)
The definition of bias could perhaps be introduced earlier. There is another broader definition used by Daly which defines bias as "any factor or process that tends to produce results or conclusions that differ systematically from the truth" (Interpretation and uses of medical statistics, Leslie E Daly, Geoffrey J Bourke, 5th edition, 2007). Perhaps this could be used, and then lead on to the more focused definition given later
Also, has what we mean by an association been defined elsewhere in the text? If not, suggest that this is discussed too.
I have suggested including outcome as well as disease wherever its mentioned in the first section, as we are not only interested in diseases, rather outcomes too.
In the non-response bias section, I think that it would be helpful to have some text to explain the scenarios more fully
In the preventing non-response bias section one additional method to achieve high response rates (as well as incentives) is to make it easy to contribute, eg by using questionnaires that are not too long, and dont take too much time to complete
Arnold Bosman replied on 9/27/2010 4:23:14 PM:
These sound like good points to me. There may be a challenge in finding the right balance between the amount of text per screen, and keeping the overview of the topic.
Still, it is in my view worth to try and put these concepts in the chapter. Derval, would you care to try an attempt in editing? This will also allow us to test how that part of our process will work out once the WIKI is open to the public. We will definitely get users who start modifying the texts, and that is exactly what we want.
So if you feel inspired, please go ahead, and we can check out how that works for the chief-chapter-editor :)
Arnold Bosman replied on 9/27/2010 4:26:13 PM:
sorry for the tardy reply, and we have already touched upon the solution in our last teleconference. There will not soon be a 'random error' chapter, as it seems to have been described sufficiently for the time being in the lines within the Bias chapter.
However, the chapter on 'P-value' would in my view indeed be a useful next priority.
Perhaps an idea for the current facilitators in Menorca? :)
Aileen Kitching replied on 10/5/2010 12:23:28 AM:
Many thanks Derval for the review of the chapter, and Arnold for additional comments!
Derval, I was also having the same problem with the Information Bias page - it kept crashing every time I tried to edit it - it didn't seem to be able to be fixed, so I have written the page again, and if you would like to take another look again, that would be great!
I have incorporated some of your other suggestions already in case you want to review those:
- I have added Leslie Daly's definition of bias to the existing paragraph, and added a definition of an association in the 2nd paragraph in that page (feel free to adapt it if you would like)
- disease/outcome is now throughout the text
- I added your suggestion about the questionnaire to the 'preventing non-response bias' section
I would be happy for you to try some editing, as Arnold suggested, and if you feel inspired :-)
Lisa Lazareck replied on 10/5/2010 12:40:10 PM:
Aileen + Derval,
Great work! Please do try editing with Internet Explorer...in case the browser freezing/crashing problem is stemming from the use of Firefox.
Another message to follow shortly, Aileen, about your table formatting query.
Aileen Kitching replied on 10/5/2010 4:00:27 PM:
Many thanks for your reply!
I have been using Internet Explorer all the time as a browser - I mentioned this also at the last telco. I really don't think it;s a browser issue, else why would it only happen on one page?? and not on others while editing.
I did wonder whether it could be related to the amount of editing on the page? e.g. if many changes are made to a page, could that be causing the problem?
if so, I would be concerned that this would cause issues when the wiki goes live......
I would be grateful if you & Martin could take a look at this. Perhaps you could also check with the others whether any of them are having problems with editing formatting, etc. as I also find that sometimes I have changed things & this is not reflected in the View page afterwards.
Derval Igoe replied on 10/5/2010 4:40:58 PM:
Hi Aileen and Lisa
Sorry about this, but its crashing again. The first page worked, I made a couple of suggestions, but cant access it from there on. I am using internet explorer
Aileen Kitching replied on 10/5/2010 5:06:45 PM:
There is definitely a problem with the "selection bias and case-control studies" page (it kept crashing on me too), but the others (e.g. Info Bias, Preventing Bias etc) were ok for me earlier, if you would like to try those?
all the best,
Lisa Lazareck replied on 10/6/2010 6:43:36 PM:
Really sorry to read about your troubles - and Martin has indeed been looking into this..a response should follow ASAP tomorrow - and if this hinders your meeting the Friday (8th) deadline, then by all means, we will extend it for you.
Thank you and more to come,
Theodore Lytras replied on 3/20/2015 9:50:04 PM:
The plot showing the random vs systematic error is erroneous. The measure of random error is the standard error, which (regardless of outcome type) is proportional to the inverse square root of sample size. Therefore the plot should actually look like this:
Arnold Bosman replied on 3/20/2015 10:02:55 PM:
Thanks so much Theodore, this is very good. Feel free to modify the article. That is why this is a Wiki: to be improved where needed.
Arnold Bosman replied on 3/20/2015 10:06:08 PM: By the way, I believe Rothman shoUld have a Femwiki account :)
Theodore Lytras replied on 3/21/2015 1:01:21 PM:
OK, replaced the plot with a modified version. I guess even Rothman can make a slight mistake now and then! :)
sbpmebxu replied on 7/29/2015 7:20:20 PM: 1
sbpmebxu replied on 7/29/2015 7:59:24 PM: 1
sbpmebxu replied on 7/29/2015 8:31:58 PM: 1
You need to be logged in to post comments.
You can log in here. You can register here if you haven't done so yet.