Cohort Studies - Science method

Hi Dear Izzeldin,

Thank you so much for your answer. I have the protocol from Resecker in which they mentioned:

"This study will be a comparison between before and after implementation of a program using data collected through a retrospective chart review.

The program was implemented at their clinic site in 2023.

The study will compare data collected 1-year prior to (2022), the transitional period of implementation (2023) and 1-year post (2024) implementation."

so they considered looking at 3 time periods

1. Pre implementation (1 year)

2. Implementation (1 year) or “transition”

3. Post implementation (1 year).

I was thinking of suggesting them to exclude the transitional period, as the goal is to compare some outcome variables pre and post-implementation

I would appreciate it if you have any insight on this.

Bahar

Dear Alwielland Q. Bello ,

Postoperative pulmonary complications (PPCs) are significant causes of postoperative morbidity and mortality. This study presents the utilization of machine learning for predicting PPCs and aims to identify the important features of the prediction models. This study used a retrospective cohort design and collected data from two hospitals. The dataset included perioperative variables such as patient characteristics, preexisting diseases, and intraoperative factors.

Regards,

Shafagat

A cohort study is an observational research design where a group of individuals (cohort) is followed over a period of time to observe the occurrence of certain outcomes, such as disease development. In a cohort study, participants are classified based on their exposure to a specific factor, such as lifestyle habits, environmental conditions, or medical interventions. There are two types of cohort studies: prospective, where participants are followed from the present into the future, and retrospective, where past data is used to establish exposure and outcome relationships. To conduct a cohort study, researchers define the groups based on exposure status, dividing participants into "exposed" and "non-exposed" groups. The study then tracks the incidence of the outcome of interest (e.g., disease) in both groups over time. The effect of exposure is measured by comparing the outcome rates between these groups using statistical measures such as relative risk. Confounding variables, which may influence the relationship between exposure and outcome, are controlled through techniques like stratification or multivariate analysis. This approach helps establish the association between exposures and outcomes while maintaining a temporal relationship.

The sample size for a cohort study depends on some factors and there is not a one-size-fits-all answer. Because sample size in studies is very much influenced by the study's objective, design, dropout rate, alpha level and statistical consideration. However, tools like PASS and R Packages can calculate the required sample size based on specified factors.

Thank you

Subjects are recruited from one centre only, and outcome observed over future time. There may also be a comparison group to assess a cause effect relationship from some putative causal factor. The subjects are classified into exposure classes before the follow uo period starts. The comparison group may be internal or external.

Such a study could be considered a retrospective cohort study.The main principles of epidemiologic design apply to c;linical studies.

hi

thank you for replying my question. well.. actually it wasnt a real question. i have just created my account and I was exploring how can I add question to here. so I tried to create question .

sincerely ,

Amine ( not Amina :) )

I think cross sectional study is the best

Dear Hassan Abushukair I hope this reaches you well. Did you ever find out the answer to your question? I am trying to do the same thing, however with no luck. Do you have any adapted NOS scale for single group retrospective cohort studies?

I found a direct response to this question here. I hope it helps: https://www.vehjir.com/2023/01/is-it-possible-to-use-newcastle-ottawa.html

A cohort study is a type of comparative study because it compares two or more groups of individuals based on their exposure to a particular risk factor or intervention. In a cohort study, a group of individuals who are exposed to a particular risk factor or intervention (the "exposed" or "intervention" group) is compared to a group of individuals who are not exposed to that risk factor or intervention (the "unexposed" or "control" group). The goal of a cohort study is to determine if there is an association between the exposure and an outcome of interest.

A non-comparative study, on the other hand, does not involve the comparison of two or more groups of individuals. Instead, it simply describes the characteristics or outcomes of a single group of individuals. One example of a non-comparative study design is a case series, which is a type of study that involves the observation and documentation of a series of cases of a particular health condition or disease. A case series is considered as a non-comparative study since it doesn't have a control group.

A case series is a type of observational study that describes a group of individual cases that have a similar diagnosis or condition. In a case series, the cases are described in terms of their demographic characteristics, symptoms, treatment, and outcomes. The main goal of a case series is to describe the clinical features and the natural course of a disease or condition, and it is a way to identify cases that are rare, unusual or unique to better understand the disease and its impact.

Thank you! All the answers were very helpful!

Thanks Andrew Paul McKenzie Pegman

Hi,

Your question's last statement itself is the answer. You can refer to STROBE guidelines for the components of such studies:

Pilot study needs to be done to get Mean and SD

And then you can use the below link

Statulator – Simple | Free |

David Eugene Booth thank you

Good afternoon..No cohort with no comparison group

Studies on fibromyalgia involve the cognitive aspects

The Ottawa Hospital has resources that provide concise explanation on using the Newcastle-Ottawa Scale (http://www.ohri.ca/programs/clinical_epidemiology/oxford.asp). In here, I think that the coding manual they shared (http://www.ohri.ca/programs/clinical_epidemiology/nos_manual.pdf) addresses your question succintly. These can be found easily through Google searching.

If you find the explanation wanting, please read on confounders/confounding in research from any epidemiology/research design textbook.

Hello Malihe Khoramdad

Have you used the regression model? For example, logistic regression model?

Maybe not useful for the original question anymore, but still good to know: a very detailed exposition on how to perform a matched cohort analysis is present in Kleinbaum's Logistic Regression 3rd edition. It gives the specifics on how to do it in SPSS using the Cox regression module in the Appendix. I used it some years ago, and it works (obviously). Hope this helps.

you can share your study protocol via your Researchgate or other academic and scientific pages or try the FDA system for registry of cross-sectional, case-control, and retrospective study designs (the system gives you a code like the RCT registry code but it's acceptable/valid for all journals and reviewers). However, as you know, there is no obligation for early protocol registration of such studies (before running the study). At this time, clinical trials and systematic reviews/meta-analyses are the most important studies there is sensitivity and obligation on their prior protocol registration. So, if you don't register your retrospective study protocol (in prior) it's ok! But if you will register it, it's a valuable point for your study. And the last but not least; if you will register or share your protocol before running the study, you should determine your primary and secondary aims and statistical analysis. So you should follow the pre-determined rules and aims (as you have registered in your protocol) in your manuscript/article. Also, if you want to change your way different from your registered protocol, you should justify it for reviewers of your manuscript.

Good luck!

How can I use NOS for assessing a nested case-control study? Should I assess all the primary cohort studies?

Plos One, BMJ and trial

Thanks all of you for your kind answers. I got it because of your support. Thank you guys ,have a blessed life

You can modify the PICO model into PECO model (E stands for exposure) while developing your research question for observational cohort design. Here C is considered as the 'unexposed' group. For example, we want to check whether 'attending a health education session on HPV vaccination' is linked to the 'outcome (eg correct knowledge about need and timing of HPV vaccination)' among a 'population of mothers of 8 year old girls'. Here we use PECO in following manner to develop our research question:

P - mothers of 8 year old girls

E - 'attending a health education session on HPV vaccination'

C - not attending this health education session (though invited)

O - correct knowledge about need and timing of HPV vaccination (captured by using a HPV vaccination knowledge scale)

Research Question: Does attendance in a health education session on HPV vaccination improve the correct knowledge about need and timing of HPV vaccination among mothers of 8 year old girls (as compared to not attending this health education session)?

Daniel Wright

sorry for the vague term, I used simple term to explain to someone and forgot to edit it in this post (and English is not my first language, sorry!). I want to find the correlation (or prove there is a relationship) between my variables. My hypothesis is that when Y1 is normal, visuospatial tests are normal and as we go to displaced and ruptured, the visuospatial tests are getting worse (probably linearly). After, if we find a correlation between the integrity of the tract and visuospatial skills, I want to show that if we ´´repair’´ the tract during surgery, the visuospatial skills are reversible and the patient will score better on the test. I have some background in stats, but not in biostats and the statistician I contacted isn’t either…

David Eugene Booth

exactly, in the litterature those 3 tracks have some unclear link to visuospatial abilities, so that’s why we chose them. They should be independent (meaning that when only y1 is rupture for example, there is a decrease in visuospatial score, same for y2 and y3 taken alone). But, if y1 and y2 are ruptured, I want to show that there is even worse score

It is confounded by time. The patients coming in may have been systematically different (particularly if done during COVID). So, short answer is NONE.

retrospective cohort study,as you went back to study relationship between exposure and outcome

In the equations to estimate sample size, you are required to have prevalence of the outcome in exposed and unexposed group; which you can get from any related study from similar setting else where. Or you assume OR of 2. But in either case you need to have an estimate of the prevalence of the exposure in the population to decide on unexposed:exposed ratio.

Thanks to Proloy, Robert and Vasileios for your answers!

You could include the extended cohort as it is the result of your article screening process. However, you should properly mention it in your result, as you explaining the characteristics of each included study. It won't (supposedly) affect the desired outcome nor giving another bias, as long as you are certain all the samples are subjected to similar conditions or interventions (similar starting point). If it were indeed different, you might want to try subset analysis.

Una muestra es representativa si representa más del 10% de la población estudiada.. Operó para ser confiable en un 95% y a un nivel de significación de 0.05 debes hallar la proporción del evento. P=1915/8592

P= 22%

Por tanto necesitas una muestra igual o mayor a 256 para tener una muestra en el rango del nivel de confianza necesitas

Provided there was no temporal effect on the confounders and no significant environmental change between the two time periods (eg Covid, heatwave,fire, flood ,war) I would say you could regard it as retrospective cohort study.

.There is a one year follow up period for the exposed and unexposed group but it is a different calendar year for each.

By using all the datasets from the single cohort study haphazardly, you will problematically duplicate or perhaps triplicate the statistical weight contribution of that study in your systematic review. It is your responsibility to assess which of the three analyses - the cohort, case-control #1 or case-control #2 - is uniquely relevant for your review. It cannot be all three. The more important component of the term "meta-analysis" is "analysis" - don't just pool data just for the sake of it.

You can, and I notice that this practice was quite prevalent in the past. Consider the following points:

1. Making a single publication (your cohort study together with a systematic review & meta-analysis [SRMA]) will make your work more loaded with content. However, I foresee that aiming for a single publication will take more time compared to publishing first the cohort study.

2. An SRMA is not simply about pooling common outcomes and data across included studies - it is a distinct endeavor in itself, with rationale, methods, results, discussions (with limitations) and conclusions. I fear that in your attempt to generate a single work from the cohort study and the SRMA, you will inadvertently put more weight in the cohort study and "relegate" the SRMA as if it is a minor part of the discussion. This is problematic. You have to be as detailed in your cohort study reporting as you did your subsequent SRMA. In doing so, it may be difficult for future readers to digest/understand your work as they are basically dealing with two studies in one paper.

3. If you are faculty member, it would be more advantageous for your promotion/tenure if you publish the cohort study and the SRMA separately - consider the weight of 2 publications versus 1 publication only. While in a lot of cases this will be considered "devious," I think that this will be acceptable in this context as an SRMA is a legitimate, standalone research endeavor on its own.

Do you think only matching for a few covariates can eliminate confounding bias? If not, besides the exact matching, you can consider propensity score matching. Once you achieve the covariate balance between the treated and control group, then run the outcome model (e.g., linear regression for your case). Since there are multiple matched controls for each treated subject, you should account for it. You can use the robust standard error, but the better approach could be bootstrapping.

Hi,

Bayley Scales of Infant and Toddler Development, however, requires country-specific standardization.

Probably you can look at the cross-references from the references given below

Wu W, Qu G, Wang L, Tang X, Sun YH. Meta-analysis of the mental health status of left-behind children in China. J Paediatr Child Health. 2019 Mar;55(3):260-270. doi: 10.1111/jpc.14349

Cook F, Conway LJ, Giallo R, Gartland D, Sciberras E, Brown S. Infant sleep and child mental health: a longitudinal investigation. Arch Dis Child. 2020 Jul;105(7):655-660. doi: 10.1136/archdischild-2019-318014

Wozney L, Radomski AD, Newton AS. The Gobbledygook in Online Parent-Focused Information about Child and Adolescent Mental Health. Health Commun. 2018 Jun;33(6):710-715. doi: 10.1080/10410236.2017.1306475.

Stewart SL, Hamza CA. The Child and Youth Mental Health Assessment (ChYMH): An examination of the psychometric properties of an integrated assessment developed for clinically referred children and youth. BMC Health Serv Res. 2017 Jan 26;17(1):82. doi: 10.1186/s12913-016

:

Fatemeh Torabi Prevalence and cumulative incidence to measure risk ratio

Best to use a tool for prevalence studies as you are extracting prevalence information and the design is not really relevant

A tool of interest would be

Hi Elisabeth. Take a look at EUPATI guidance (other valuable resources available at the website): https://eupati.eu/resources/patient-engagement-roadmap/

Hope this helps.

Another point to consider is if the two groups are matched for HDL-C and LDL-C levels (these being significantly different or not). If they are matched, I would be much less worried about further normalisations. However, it also depends upon the question you are trying to address.

Well, the cohort study is highly suggested in covid pendamic. however, the covid-19 is no more a cohort situation it is done for observation and per and post-test by training and other medication.

no. a cohort study can be done can be done for all exposures.

a simple power /sample size calculator is given at

Cases and controls are often not matched for life status. In mortality studies controls tend to be live whole cases have died. Live controls have more information about past exposures than relatives of people who have died. For example if work with asbestos brakes is an exposure of interest, relative are far less likely to know about occ changes than are actual cases or controls.

In short, retrospective cohort studies allow researchers to examine the risk associated with the exposures and express them as relative risk (RR), while case-control studies only allow researchers to express the association between exposures and outcomes using odds ratio (OR). Unlike RR, OR cannot be interpreted directly and thus yield lower quality to explain causations underlying the observed effects. I agree with Babak Saravi for the rest of the explanations.

As Babak Saravi and Shuaibu Suleiman Adeiza have already pointed pout: Usually Q and I² are used when assessing the heterogeneity. However, one has to be cautious when intepreting them. I'll address the problems below.

The problem with Q is that it depends on the number of studies so it's hard to interpret it without having additional information. Researchers often report a p-value for the Q statistic. A small p-value could mean that there is heterogeneity. However, it is well known that p-values depend on the sample size (in this case number of studies). Thus, when analyzing many studies one can get a small p-value even if there is only marginal heterogeneity. In contrast, when only a few studies are available one could get a large p-value even if there is a considerable amount of heterogeneity.

The problem with I² is that it only indicates the RELATIVE amount of true heterogeneity (as compared to the error variance). Babak Saravi cited the threshold of 75%, which is often interpreted as high heterogeneity. However, it is not recommended to evaluate the heterogeneity solely using the I² value. Why? I'll illustrate it with a short example.

Summing up, please do not rely solely on Q and I² when assessing heterogeneity. It can lead to wrong conclusions. Consider interpreting the value of true heterogeneity (tau) too.

Longitudinal and cohort studies follow the same group of individuals over time. They allow researchers to compare outcomes according to exposure to suspected risk factors by creating a hypothesis to study. The strength of a cohort study is that it is the best way to determine the causes of a disease, condition or issue because it allows researchers to follow one group of people over time and allows researchers to study multiple outcomes at one time.

Numerous limitations exist in conducting longitudinal/cohort studies. They are costly to perform, a large sample is needed for rare outcomes (e.g. mortality), and loss to follow-up is common. Given the inherent time frame required of this approach, it can be very challenging to employ such a study in conflict and post-conflict settings where conditions can change rapidly.

Cronbach's alpha score above 0.7 is recommended.

Why you don't try Matlab

Dear Saikarthik Jayakumar,

another usefull online tools are OpenEpi and powerandsamplesize.com. Both provide specific calculation each study design and provide basic knownledge about power and sample size calculation itselfes and about how to use the tools.

You can reach them via the following links:

Good luck and kind regards!

Hi Haneen,

Have a look here - perhaps helpful in your search:

http://www.p3gobservatory.org/

Good luck :)

I still confused with the answer , if we tried to do cohort case control study .... witch equation we use to calculate sample number and controls number ?

As the samples are acquired from the same patient across time, the ideal thing would be to perform a paired analysis (repeated measures ANOVA or a non-parametric equivalent). However, the levels of the factor time (the timepoints) are required for every sample group, which are absent in controls). Therefore, I'd move to a combination of pairwise comparisons (t test or Mann-Whitney, depending on data normality), at the expense of sacrifizing the pvalue of interactions. However, post-hoc corrections as FDR may penalize less the false positives as variables in your glicomics data are considered in individual pairwise comparisons, while the correction initially should correct for all variables tested in the study.

Hope this helps!!!

Any validated tool can be used but I personally believe that Newcastle Ottawa Scale is quite handy for observational studies.

i think the best option in this case is the Cochrane Collaboration tool for assessing risk of bias for RCTs and Risk Of Bias In Non-randomized Studies – of Interventions (ROBINS-I) for non-randomized studies

Many thanks for the helpful feedback!!!

Dear Samuel,

Hello,

First of all, I would like to thank you for the sake of this interesting discussion!

Although many of important criteria have been addressed by the friends participated in this discussion, I would like to add that novelty of the idea, good review of the literature, accurate research method, clarity and presentation of the work, and usefulness of the article contribution can also taken into account.

All the best,

Ali

The tool you use will depend on the study designs. Different study designs require different tools. Some existing tools (i.e. GRADE) gain popularity not because they are better than others. "Today" it is GRADE, "tomorrow" it will be another popular tool. Some tools are more subjective than others. I prefer more objective tools. One easy and recommended tool for longitudinal designs is NOS: Newcastle-Ottawa Quality Assessment Form for Cohort Studies. Another tool is the RoB (Cochrane risk of bias tool) which is more for randomised or controlled studies.

It's not logic to have 10 not applicable questions,

Anyway,

Yes>>1

No and NA>> 0

0-4 points= poor

5-9= fair

10-14= good

QA is summarised as fair, good, fair as a column in characteristics table

Joe Stanek, Paul Yarnold

Thank you both for your immediate reply and providing possible ways for my analysis. Your opinions really helps.

I might need some time to have a thorough understanding of these methods, but will try to figure out which fit the study best.

I will update here when I completed it.

Much thanks again!

if you take this approach, alternative methods would be to restict study to comparable exposure durations or to use regression modelling with length of exposure as a covariate.

It depends on the outcome/disease of the study.

In the past (and sometimes still) clinical researchers sometimes used the term "case control study" to mean comparing outcome in an exposed group to that in a "control" (ie un exposed group) ,cross sectionally. That is why in modern epidemiology (ie since about 1985) the term case-referent study is now preferred .

Hello dr Adane,,,

Thanks for this good question. I recommend to check the following brief pdf

Difference between case cohort and nested case control

The quality assessment is usually based on the reported data. So, if the author reported that they calculated the sample size or they mentioned the CI for OR, RR, or MD, you can consider that as they calculated the sample size. you can find that mentioned in the study methodology

If you did not find any evidence, so it is unclear and you cannot consider it as 1

According to the global publishing standard, the least statistically possible number is 100, less than this number is not possible as a result

Would this be more accurately termed a 'case study' if there is just one participant?

The informed consent process may over select individuals who are depressed and for some reasons are at high risk for dementia (they may be more eager to participate and to continue follow up) and thus would potentially inflate the association; sort of selection bias. Also if outcome data collectors are not blinded to exposure status, there may be information bias.

On the other hand, depressed individuals may die earlier and thus would not survive long enough to get dementia, which would reduce the association.

Manzoor Hussain well done.

Unfortunately most of the victims of unhealthy food environment are deprived people who do not have many healthy options. The most plausible approach could be changing the food environment. The systems approach is suggested to address this issue whereby everyone involved i.e. food industry, policy makers, consumers, practitioners etc. are brought together to improve food environment. It is a hard approach but promising. Health education alone is not sufficient to make a dietary change for instance. unhealthy food environment is a wicked problem and challenging.

Slight adjustment to Ms. Cowdhuri's reply. First where she says multiple logistic regression she means multinomial logistic regression. Second if the DV is ordinal you want ordinal logistic regression. Useful set of notes is attached. Best, David Booth

This an old list of criteria to judge the quality of research. The list might be useful to you

1. Is the problem stated clearly?

2. Does the problem have a solution the way it is formulated?

3. Is it clear whether this is a normative, or a survey or a study to test hypothesis?

4. Is the literature of previous studies adequately reviewed or taken into account?

5. Are the essential concepts necessary to understanding the problem defined?

6. Is the context of the problem described so that what is included or excluded is readily apparent

7. Is the importance of the problem sufficiently developed so that the judgment can be about the suitability of the method?

8. Are the consequences of the possible findings pointed out?

1. Was the design of the study planned and evaluated beforehand?

2. Does the design take into account all the pertinent aspects of study subjects, materials, environment, variables, measurements, and observations "statistical methods"?

3. Is the design clearly presented?

4. Were alternative designs considered and reasons for their rejection given?

5. Are the compromises made with an ideal design described?

6. Is it possible from the design to answer the questions posed?

7. Can the design efficiently solve the problem in terms of money, subjects and time?

1. Is the sample adequately described?

2. Of what population is the sample representative?

3. Is the sample an appropriate one for the purpose of the study?

4. Were the subjects selected according to the design of the study and with regard to the statistical method to be used?

5. Is the size of the sample adequate?

6. Was the sample collected for the purpose of this study or for other purpose?

7. Are the methods of sampling described?

1. What controls were exercised through sampling?

2. What controls were exercised by selection of setting? By natural habitat?

3. What controls were exercised by experimental manipulation?

4. Were the conditions the same for all subjects or were adjustments made?

5. If controls were changed, were results analyzed with respect to the altered conditions?

6. Are there any important controls missing from the study?

7. Were the usual controls for a study of this type absent, and if so, was the absence justified?

8. Could any additional controls have been included that would have increased the efficiency of the study?

Measurements

1. Are the techniques of measurement adequately described?

2. From the information presented, could the measurement be repeated by another investigator?

3. Are the measurements suitable for the problem?

1. Are the methods of recording and the treatment of data described?

2. Are the statistical procedures described and are they suitable?

3. Are the significant tests described and are they suitable?

1. Are the results or data adequately presented so the reader may verify the author's statements about them?

2. Are estimates of error provided?

3. Have the essential relationships posed by the problem been analyzed and tested for significance?

4. Are the results clearly reported in tables and graphs so that others may use the data or reproduce the results?

1. Does the author draw conclusions about the major problem of the study?

2. Are the conclusions clearly supported by the data?

3. Are important reservations or qualifications pointed out?

4. Are artifacts or spurious relations pointed out?

5. Has the author overlooked important aspects of the results?

6. Are necessary modifications of theory, current interpretations of data or practice pointed out?

7. Are the results interpreted in relation to other published information and is their significance for related fields pointed out?

8. Are the methods used in the study critically reviewed in the light of the obtained results?

9. Are the interpretations, implications for future research, and development of new methods appropriate for the present study, or do they reflect an overestimation or underestimation of the significance of the study?