Talk:Cohort study

Untitled
I don't think this should be merged with cohort (statistics) because I think there is value in defining cohort separately from the concept of the cohort study, which is a specific medical term.

Excellent idea - very much linked

Paragraph 1: Most cohort studies do not follow children from birth due to the length of time it would take for disease endpoints to occur. Heart disease and cancer are largely diseases of later life. Most cohort studies recruit participants who are over 30.

A distinction should be made between cohorts and birth cohorts. While the participants of a birth cohort are recruited on the basis of their year of birth, a cohort study generally recuits on the basis of an age range and perhaps other factors such as geographic location or occuption. For example, the Framingham Heart study recruited men and women aged 30-60 years living in Framingham. The Nurses health study recruited women nurses aged 30-55. The EPIC-Norfolk study recuited men and women aged 45-74 living in Norfolk,UK.

Paragraph 2: Rare outcomes are studied with cohort studies. However, the cohort must be sufficiently large and the investigator willing to wait long enough for sufficient disease endpoints to occur.

Some questions can only be answered using a prospective cohort study design. For example, a case control design would not be suitable to study the link between diet and cancer. The main objection would be the possiblity of 'recall bias'. The advantage of a cohort study design is that participants are asked about their lifestyle and diet before any of them become ill. When people become ill, their diet often changes and it is hard to remember what one used to eat.

Randomized controlled trials have the great weaknesses that the hypothesis is fixed at the outset. Cohort studies however are much more flexible in that many exposures and many outcomes can be studied and many hypotheses tested.

RNL

Technical detail
"Whereas prospective cohorts should be summarized with the relative risk, retrospective cohorts should be summarized with the odds ratio."

I don't think this statement is necessarily correct. A case-control study (synonym: retrospective study) should be summarized with an odds ratio. A historical cohort study (synonym: retrospective cohort study) can indeed be summarized with a relative risk, as far as I know. Despite the fact that it is "retrospective", it is still forward-looking (i.e., longitudinal). Does anyone agree with this? Also, in the text, it should be referred to as a "historical cohort study" first, not a "retrospective cohort". This usage is supported by John M. Last in the Dictionary of Epidemiology. What do you guys think? —Preceding unsigned comment added by 131.111.223.101 (talk) 14:45, 26 May 2008 (UTC)

Rothman, Greenland, and Lash (Modern Epidemiology, 3rd Edition) have a substantive discussion on these distinctions. A retrospective cohort is not confined to summary by odds ratio. The phrase "historical cohort study" doesn't particularly improve on "retrospective cohort". 66.156.15.189 (talk) 22:06, 9 April 2009 (UTC)

Removed: "Whereas prospective cohorts should be summarized with the relative risk, retrospective cohorts should be summarized with the odds ratio. " as discussed above. CaseInPoint (talk) 18:25, 26 July 2010 (UTC)

Merger proposal

 * JA: Oppose merge. A large portion, if not the lion's share, of cohort studies are carried out in qualitative research settings, at least at the exploratory stages, and though one may hope to integrate qualitative and quantitative sources of knowledge over time, there is no need to force the issue by means of such mergers and mushing togethers.  Jon Awbrey 13:12, 9 February 2006 (UTC)


 * Oppose. Cohort (statistics) is little more than a dicdef. If anything cohort study needs to be widened in scope to reflect a wider range of research methods areas (i.e. more than just med/epidemiology). BrendanH 22:30, 7 March 2006 (UTC)


 * support dont agree that cohort studies are carried out non-quantitatively....there is only one topic here, even tho both articles are poorly written Anlace 04:08, 23 March 2006 (UTC)


 * I don't agree that there is only one topic. There is the demographic concept of cohort, and the research design category of cohort study. Indeed, the more I look at it the more it seems that many medical "cohort studies" do not study cohorts, but follow population samples prospectively. If the articles develop, they should develop in distinct directions, one conceptual and one methodological. JA is of course wrong about most cohort studies being qualitative, but that is a red herring. BrendanH 09:07, 23 March 2006 (UTC)

I have reverted the merge between cohort (statistics) and cohort study as this was not clearly supported. They are different concepts and should develop differently. Cohort (statistics) should develop in terms of the demographic definition of the concept, and cohort study should develop in a research methods direction. The latter needs to be tidied up somewhat, as it currently has too much of a medicine slant and the cohort studies are used rather more widely than that. BrendanH 15:55, 6 April 2006 (UTC)


 * Oppose for the reasons mentioned by BrendanH.  BakerStMD  T&#124;C 17:38, 29 December 2014 (UTC)

Another merger proposal
The article panel study says that such a study is also called a cohort study. Either needs merger or clearer distinction. Even if there is a distinction there is surely enough in common for a merger.Cutler (talk) 17:27, 11 December 2007 (UTC)
 * I agreed with this merger proposal and due to the two months of inactivity on the suggestion, went ahead and did the merge. --jwandersTalk 00:08, 9 February 2008 (UTC)

This was an error as cohort and panel studies are not the same. They really need to be unmerged because it is quite misleading. — Preceding unsigned comment added by 89.185.247.13 (talk) 17:52, 28 November 2014 (UTC)


 * I had never heard of a panel study, but this definition seems to suggest that they are different from cohort studies.

http://www.oxfordbibliographies.com/view/document/obo-9780199756384/obo-9780199756384-0108.xml BakerStMD T&#124;C 17:37, 29 December 2014 (UTC)


 * I agree cohort and panel studies are not the same. Actually, panel study looks more like a synonym of longitudinal study to me, so I have updated both pages to make this more consistent. 7804j (talk) 16:01, 6 November 2016 (UTC)

Examples
I removed the first paragraph of the examples section. It sounds more like a rant in a comment stream than an encyclopedia entry, as well as being confusing and incorrect. A cohort study can certainly refer to a study in which groups of individuals are followed and repeatedly followed up on (which may or may not create the conditions for a longitudinal analysis). In the language of the previous author, the event which the cohorts share may be, for example, smoking or not smoking at the baseline interview. The study comparing age categories the author provides as an example is simply a specific type of cohort study which is, at least in medicine, very much less common than the type where cohorts are defined or selected according to baseline covariates.

The classical example of a cohort study is one where a group of smokers and a group of nonsmokers are selected with the restriction that no one in the study have lung cancer at baseline. The smokers and nonsmokers are then followed (i.e. data is collected at multiple time points). For example, the researchers might be interested to know if any of the smokers quit smoking, and when. They might also be interested in time dependent covariates, such as whether or not a participant has ever had a heart attack (may change from no to yes over the course of the study) or the number of hospitalizations the individual has undergone (may increase over the course of the study). For a survival analysis, the primary endpoint of the study would be whether, and at what time, the participant died from lung cancer. One then tests for an association between smoking and lung cancer, possibly adjusting for covariates collected at baseline or during followup. To simply compare the prevalence of lung cancer in the two groups, one would follow participants only to a certain time point, and compare the lifetime prevalence of lung cancer in each group at the end of the study.

One can also use the data for a longitudinal analysis by looking at differences in the time pattern of hospitalizations between the smokers and non-smokers. — Preceding unsigned comment added by 209.94.133.165 (talk) 21:14, 19 August 2011 (UTC)

Hypothesis testing
"... to obtain evidence to try to refute the existence of a suspected association between cause and effect; failure to refute a hypothesis strengthens confidence in it."

I think that this statement has got it all "backwards". It is the null-hypothesis (H0) that is to be refuted, not the hypothesis (H1). If the evidence is strong enough that one can refute H0, then the hypothesis (H1) is strengthened. If the evidence is weak, one is left with H0, which does not introduce anything new. Failure to refute something does not strengthen any confidence in anything, it just indicates that the data or evidence is weak. To rewrite the quoted sentence, it could be something like:

"... to obtain evidence to try to refute the NON-existence (H0) of a suspected association between cause and effect; ability to refute the null-hypothesis (H0) strengtens confidence in the hypothesis (H1); failure to refute the null-hypothesis does not strengthen any hypothesis."

Or it could be rewritten from scratch, independent of the original wording.

Here is a comment to the original wording: "suspected association between cause and effect". By definition there is always an association between cause end effect; if not, they would not be cause and effect. This should be rewritten. Gbwi (talk) 13:19, 16 December 2011 (UTC)

External links modified
Hello fellow Wikipedians,

I have just modified 4 external links on Cohort study. Please take a moment to review my edit. If you have any questions, or need the bot to ignore the links, or the page altogether, please visit this simple FaQ for additional information. I made the following changes:
 * Added archive https://web.archive.org/web/20060310202137/http://clio.stanford.edu:7080/cocoon/cliomods/trailmaps/design/design/prospectiveCohort/index.html to http://clio.stanford.edu:7080/cocoon/cliomods/trailmaps/design/design/prospectiveCohort/index.html
 * Added archive https://web.archive.org/web/20060310202147/http://clio.stanford.edu:7080/cocoon/cliomods/trailmaps/design/design/retrospectiveCohort/index.html to http://clio.stanford.edu:7080/cocoon/cliomods/trailmaps/design/design/retrospectiveCohort/index.html
 * Added archive https://web.archive.org/web/20110809135039/http://www.vet.cornell.edu/hospital/imaging/tutorial/index.html to http://www.vet.cornell.edu/hospital/imaging/tutorial/index.html
 * Added archive https://web.archive.org/web/20100414204016/http://www.esds.ac.uk/longitudinal/resources/international.asp to http://www.esds.ac.uk/longitudinal/resources/international.asp

When you have finished reviewing my changes, you may follow the instructions on the template below to fix any issues with the URLs.

Cheers.— InternetArchiveBot  (Report bug) 07:48, 10 August 2017 (UTC)

Citation 1 is very weak
Citation 1 is a very weak citation - it links to Academic Stack Exchange and there's not much monitoring of the quality of answers given to questions people ask. In fact, I'd consider it no better than an uncited statement people make in wikipedia. There should be better citations out there. — Preceding unsigned comment added by 140.142.105.93 (talk) 20:39, 13 April 2018 (UTC)

Case-cohort design
This article needs a definition of case-cohort design within Variations (definition here); writing this down as a place holder. Feel free to add the section as I won't likely have the time soonish. 70.79.138.139 (talk) 22:38, 16 February 2020 (UTC)

treatment variable contributes to the incidence of the disease,
The use if the word 'contibutes' misleading suggests an element of causality. Rather, 'treatment variable is 'associated' with the incidence of disease' would be more a accurate phrasing. 2600:1700:F0C0:9EE0:28F7:8714:F16A:3EC8 (talk) 19:18, 24 July 2022 (UTC)