Read for the decision your project needs. Map the study, inspect the
results, judge what they can support, then save the judgement beside
the reference.
Bring: one potentially useful paper and your research question. Make: one structured appraisal and reading note. Remember: critical appraisal is not fault-finding. It is deciding how much weight a study deserves and what its findings can and cannot support.
The central habit
Read for a purpose, not from page one to the end.
First decide why the paper matters to your question. Then read in
passes: orientate, map the methods, inspect the results and only then
weigh the authors’ interpretation.
Before appraising: write the question you need this
paper to answer. You can stop early when it is clearly irrelevant.
By the end you can
identify the study question, design, population and main comparison;
follow participants and denominators through a paper;
read estimates, confidence intervals and absolute numbers together;
use the CASP checklist matched to the study design;
write a note that says what the study can and cannot support.
1 · Orientate
Take a ten-minute first pass.
01
Why this paper?
Write the exact question or claim it might help you address.
02
What was asked?
Find the study aim, not just the topic.
03
What design?
Name the design before judging what it can establish.
04
Who and where?
Identify the source population, setting and analysed sample.
05
What was compared?
Map the exposure or intervention, comparator, outcome and timeframe.
06
What was found?
Locate the main absolute numbers, estimate and uncertainty.
07
What could distort it?
Mark the main risks of bias, confounding, chance and missing data.
08
Why does it matter here?
Decide whether it is trustworthy and relevant enough for your purpose.
Do not begin by highlighting everything.Make a map first. Highlight only the information that answers one of these eight questions, and label why it matters.
Social deficits are potential risk factors for premature mortality. Most research has focused on social deficits measured at single points in time. It remains unclear if the chronicity of loneliness affects its impact on adverse health outcomes. This study assessed the effects of chronic loneliness and isolation in predicting incident functional impairment and all-cause and cause-specific mortality. 1
This longitudinal study used panel data from the English Longitudinal Study of Ageing, including 14 years of follow-up (waves 2–9, in 2004–2018). Social deficits over three waves (4 years) were measured using the UCLA loneliness scale and social isolation index, categorized as not present, fluctuating or chronic. 2–3
[…]
In survival analyses, compared with people who were not lonely/isolated, people experiencing fluctuating loneliness and social isolation had higher risks of all-cause mortality (loneliness HR 1.29, 1.13–1.48; isolation HR 1.15, 1.01–1.31). People with chronic isolation also had higher risks of all-cause mortality (HR 1.27, 1.05–1.55) and cancer-related mortality (sHR 1.69, 1.23–2.31). 4–6
[…]
Over a 14 year follow-up, we found that chronic loneliness and isolation phenotypes were associated with aggravated risks of incident functional impairment and mortality. 7
Do not let one causal-sounding word decide the design.The abstract says “effects”, but this is an observational longitudinal study. Read the methods before deciding whether a causal interpretation is justified.
The timing is stronger than a single snapshot.Loneliness and isolation were measured across three waves over four years, before later outcomes. That improves time ordering, but it does not remove confounding or reverse causation.
Check how the exposure became a category.The full paper defines loneliness using a three-item UCLA scale and isolation using a five-item index, then classifies each as absent, fluctuating or chronic. Thresholds and repeated measurements shape who enters each group.
Different outcomes use different denominators.The complete abstract reports 5,131 participants in the mortality cohort and 4,279 participants free of functional disability at baseline. Never carry one sample size across every result.
Name the comparison and the effect scale.For the social-isolation result, the reference was no social isolation. HR 1.27 describes the relative hazard for chronic isolation, not a 27 percentage-point increase in risk.
Keep the interval and what is missing beside the estimate.The chronic-isolation HR was 1.27 with a 95% CI from 1.05 to 1.55. The abstract gives no absolute mortality risk, and the interval does not address bias.
The authors report an association, not proof of prevention.The full paper identifies residual confounding, self-reported measures and limited power for cause-specific mortality. Those limits belong beside any claim you write.
3 · Trace
Follow the people and every denominator.
SourceWho could have entered?
EligibleWho met the rules?
RecruitedWho agreed or was enrolled?
MeasuredWho supplied usable data?
AnalysedWho entered this result?
OutcomeHow many events or observations?
Selection
Who is missing?
Compare included and excluded people. Ask whether entry, dropout or complete-case analysis could change the comparison.
Measurement
How was each variable known?
Inspect timing, instruments, thresholds, blinding, record linkage and whether errors may differ between groups. For a survey study, use the Survey Methods Handbook to examine questionnaire development, sampling and fieldwork.
Analysis
Which sample made each number?
Tables may use different denominators. Find the footnotes, missing-value rules and model-specific sample size.
The denominator is part of the result.If it changes, ask why. Never assume that the abstract, table and model used the same participants.
4 · Interpret
Read the result before the authors’ story.
Name the quantity. Is it a risk, prevalence, mean difference, odds ratio, hazard ratio or another estimand?
Find the reference and direction. Which group is the numerator or comparison, and how were categories coded?
Read the absolute numbers. They make scale, frequency and denominators visible.
Read estimate and confidence interval together. Judge magnitude, direction and precision before the p-value.
Inspect the model. Identify the analysis population, adjustment set, functional forms, assumptions and sensitivity analyses.
Separate result from explanation. Ask which alternative explanations, such as bias, confounding, chance or reverse causation, remain plausible.
Overclaimed
“Chronic social isolation causes premature death.”
The observational design cannot by itself establish cause, and the estimate, uncertainty, population and remaining confounding have disappeared.
Defensible
“In Gao and colleagues’ ELSA cohort in England, chronic social isolation was associated with a higher all-cause mortality hazard than no isolation.”
The fully adjusted Cox model gave HR 1.27 (95% CI 1.05 to 1.55). The observational design, self-reported exposure measures and possible residual confounding limit causal interpretation.
CASP means Critical Appraisal Skills Programme.
It provides free checklists matched to different study designs. Each
checklist prompts you to examine trustworthiness, results and
relevance in a structured way. CASP supports your judgement; it
does not replace it or generate a definitive quality score.
Identify the main estimates, absolute numbers, uncertainty, consistency and sensitivity to decisions.
03 · Usefulness
Does it help with my question?
Consider population, setting, outcomes, benefits, harms, feasibility and fit with the wider evidence.
Choose the checklist that matches the study design.Use the cohort checklist for a cohort, the qualitative checklist for a qualitative study, and so on. Record reasons for each judgement; a row of ticks is not an appraisal.
Verified reference, DOI or stable link, and the project question it may answer.
Map
Question + study design
Population, setting, exposure or intervention, comparator, outcome, timeframe and analysis sample.
Extract
One result that matters
Absolute numbers, effect estimate, confidence interval, units, reference group and adjusted variables.
Judge
Strengths + limitations
Specific features that raise or lower confidence; avoid empty labels such as “small sample” without explaining the consequence.
Bound
What it can + cannot support
Write the most defensible claim and the overclaim the evidence does not earn.
Use
Place in my argument
Note the paragraph, comparison or decision it informs, and which paper or question to pursue next.
Attach the note to the reference.Put it in your EndNote or Zotero record, or link it unambiguously. Add tags or collections for project, topic, design and status so you can retrieve it later.
Download · reusable scaffold
Use the paper-appraisal and reading-note template.
Editable Word template · DOCX
One paper, one structured judgement
Work through the first pass, study map, results, CASP-informed appraisal, writing claim and reference-manager action.
prompts tell you exactly what to look for;
the result and its limits stay together;
the final note is ready to save in EndNote or Zotero.
Go deeper · critical reading, appraisal and reporting
Use these authoritative resources when you need a design-specific checklist, a broader critical-reading strategy or help checking what should have been reported.
Nature Mental Health · Open-access paper
Read the complete study behind the example
Compare the selected abstract sentences with the full methods, participant flow, results, discussion and limitations.
Find the guideline matched to the study type. Reporting completeness helps you locate information, but it does not prove that a study was well designed.