Analysing research papers

What to mention when stating a paper’s hypothesis (cohort study)

Exposures are associated/ contribute to risk of development of outcome

What to mention when describing the base population

age, location, time period (for prospective),any restrictions to participation?

When defining an exposure and outcome mention (cohort study)

Exposure, and how it was measured, particularly objective measures and how levels are quantified

Outcome, how it was measured or diagnosed, any scans or criteria used? Any interviews conducted?

What to mention when asked if relevant confounders have been identified? (cohort study)

Main ones- age, gender, ethnicity, socioeconomic status, education, comorbidities

Some to consider- genetic risk, stressful life events, substance use (alcohol, smoking etc), comorbidities, age of onset

When asked about bias, what should be mentioned

Are biases mentioned and minimised, or is it not clear from the paper

What possible improvements can be applied to prospective cohort studies

Using objective measures of exposure/ outcome, through medical reports, census data

Extend follow up

Control confounders that were not identified

When using multiple cross sectional studies, why does methodology need to be standardised

If methodologies are not the same the apparent difference in behaviours may be related to the difference in methodology between the two waves, not due to exposure/ changes over time

What needs to be considered what mentioning samples in cross sectional studies (3)

Is the sample representative of the target population: limits generalisability,

If the sample sizes are relatively small there is less statistical power to find an effect.

It would be better if the study controlled for more demographic variables in case there was a systematic difference between the two samples and so the variable could confound the results (eg parental socioeconomic status)

What are alternative explanations for cross sectional findings

Potential confounders, changes in environment, if qualitative data was used it can limit inferences we can draw from the study

What are issues of qualitative data in cross sectional studies

IF qualitative data was only collected in one survey we are limited in the inferences we can draw from this data. Furthermore the very nature of qualitative data means it cannot be analysed using statistical methods, making hypothesis testing difficult

What increases the validity of assessments in cross sectional studies

* pilot studies to assess the validity and reliability of the measures

* widely applied and used elsewhere with good results (however mainly in western countries)

* use of lay people and not psychiatrists (feasibility/ cost perspective).

What improvements can be made to the design of a cross sectional study

Measuring and controlling for confounds

Ensure validity of assessments, use pilot studies for reliability of measure

Adding incentives

What is important to remember about cases and controls

Cases = diagnosis

controls= no diagnosis, BUT BOTH can be assessed on the exposure/risk factor

Same inclusion/ exclusion

What do you need to look out for when seeing if information bias applies to a case control study

modifying a broadly inclusive scale when they could have used a validated well-established trauma questionnaire instead of adapting

If the researchers seem to be overly inclusive when establishing disease status. For example, they included people with dysthymic disorder.

What to mention when commenting on the unit of analysis in an ecological study

the unit itself and the n(sample size)

What are some other sources of selection bias in ecological studies

As the exposures rely on census data, selection bias could occur from people not completing the census, eg homeless people, migrants

the number of suicides/outcome needs to be divided by size of population to determine the rate.

Why is weighting imoprtant in an ecological study

A highly populated county should have more weight in the analysis. If there was an unweighted analysis, then very small counties that could be outliers would disproportionately affect the results.

When looking for confound in an ecological study, what is important to know?

Does the study look at changes over time. it may be more important to possible confounding variables that are likely to change over time. (so not demographic characteristics eg gender, ethnicity, but other factors like age, substance misuse, development of mental health issues)