Slide 1Everything so far — effect sizes, confidence intervals, p-values — applies to quantitative research. But what do you do when you encounter a phenomenological study of patient experiences, a grounded theory analysis of health worker decision-making, or an ethnography of hospital culture? The numbers disappear, and different questions take their place.

Slide 2Qualitative research isn't trying to measure how much or how often. It's trying to understand how and why: the meanings people make, the processes they navigate, the contexts that shape their experiences. This requires different methods, and critically appraising those methods requires different questions. So the same instruction holds. Go to the Methods section first.

Slide 3Start with sampling. Qualitative studies aim for informational richness, and not for statistical representativeness. Look for purposive sampling, deliberately selecting participants who can inform the research question. Theoretical sampling, selecting participants based on emerging analysis. Or maximum variation sampling, seeking diverse perspectives. The key question: did the researchers recruit people who could actually illuminate the phenomenon under study?

Slide 4Then sample size, which is where quantitative readers get tripped up. Qualitative studies typically have smaller samples. Sometimes only a couple dozen participants, sometimes fewer. That can reflect the depth of analysis required, so look for evidence that the researchers collected data until they reached saturation, the point where new data stops generating new insights. If a study interviewed only 6 people and doesn't mention saturation, that's a concern.

Slide 5Three more things to check in the Methods. How did researchers actually gather data? In-depth interviews capture individual perspectives but may miss group dynamics. Focus groups reveal social interaction but may suppress minority viewpoints. Observation captures behavior but not internal meaning, and good studies often triangulate. Did they share the interview guide or the observation protocol? That transparency helps you assess whether they asked about what matters. And who conducted the research, and how might their background shape what they saw and heard? A physician interviewing patients about treatment experiences occupies a different position than a community member doing the same. Neither is inherently better, and the relationship matters. Look for reflexivity, evidence that researchers considered how their own perspectives influenced the work. Finally, how did researchers move from raw data to findings? Thematic analysis, grounded theory, phenomenology, framework analysis. The specific approach matters less than whether it's clearly described and consistently applied.

Slide 6Now the findings. Instead of tables of numbers, you'll mostly find themes, categories, or theoretical models illustrated with participant quotes. Are the themes well supported? Each theme should be illustrated with multiple quotes from multiple participants. A theme supported by one quote from one person is an anecdote. Is there thick description? Good qualitative research brings themes to life with rich, contextualized detail, so you understand what participants meant, and not just what they said. And are negative cases addressed? If everyone in a study says the same thing, be suspicious. Real social phenomena are messy, and strong qualitative research identifies and grapples with contradictions, outliers, and complexity.

Slide 7Two more. Is the interpretation grounded? The researchers' interpretation should flow logically from the data they present, and if findings feel like a leap from the quotes provided, the analytic connection may be weak. And are quotes doing real work? Participant quotes should illuminate, and not decorate. Watch for studies that present a theme, offer a quote that merely restates it, then move on. The best qualitative writing uses quotes to reveal something you couldn't have anticipated.

Slide 8Which brings us to the last section of any paper, quantitative or qualitative: the discussion. This is where authors interpret their findings, connect their work to the literature, acknowledge limitations, and make claims about importance. Your job is to evaluate whether their interpretation matches the data. Watch for logical leaps. The most common move is to slip from associated with to leads to without acknowledging the gap. An observational study finding that A is associated with B does not establish that A causes B without a clear strategy for causal inference. A second common overreach is extrapolating far beyond the study population: a study conducted among urban, educated young adults shouldn't draw conclusions about rural, less-educated, older populations without acknowledging the leap. Authors also tend to attribute effects to mechanisms that weren't tested. If an intervention improved outcomes but the hypothesized mechanism was never measured, that mechanism remains speculation. It's ok for authors to speculate, and it needs to be clearly labeled as such.

Slide 9Then the limitations, where you're looking for proportionality. Do the stated limitations match the design weaknesses you noticed while reading the methods? Consider a pilot study of group prenatal care in Malawi and Tanzania, which concluded cautiously. The results should be interpreted with caution because the sample size was small and implementation occurred in only three health facilities. The authors went further, stating that to address these weaknesses, the next step is to seek funding for a large RCT. That's appropriate humility. They named the weaknesses and used them to direct future research. Other discussion sections bury limitations in a brief paragraph at the end, or dismiss them as unlikely to affect conclusions.

Slide 10Experienced readers learn to notice what papers don't say. The chapter lists six patterns; three of them do most of the work. Missing null results: if a study measured multiple outcomes but only reports some, ask what happened to the others. Selective reporting of positive results is common. Buried findings: key results that undermine the main narrative are sometimes mentioned briefly in the discussion. A sentence like, while the intervention showed no effect on our primary outcome, we observed improvements in secondary endpoints, deserves close attention. And changed endpoints: if the abstract emphasizes an outcome that wasn't the pre-registered primary outcome, be alert. Outcome switching is a well-documented problem. None of these patterns proves a paper is wrong. They're signals to read more carefully.

Slide 11So, ten core questions for any study, and they are the whole chapter in one list. One. What is the research question? If you can't state it clearly, you can't evaluate the answer. Two. What type of study is this? Three. Who was studied and how were they selected, and is that population relevant to you? Four. What comparison was made?

Slide 12Five. How were outcomes measured? Valid, reliable, blinded? Six. What is the effect size and precision? Not just p-values, but actual magnitude and confidence intervals. Seven. What could have gone wrong? Confounding, selection bias, measurement bias, attrition. Eight. Are results consistent across subgroups and sensitivity analyses, or do findings depend heavily on specific assumptions?

Slide 13Nine. Do conclusions match the data, and are the authors appropriately cautious? And ten. Does this apply to my context? Now, these six videos have covered most of that list, and they have not covered all of it. The chapter carries the established checklists and the reporting guidelines, the reference tables on effect sizes, the number needed to treat, and the questions to ask about external validity. Everything left out of these videos is in the chapter, and most of it is easier to read than to watch.

Slide 14I want to close with a note about humility. Every study we evaluate was conducted by researchers who worked hard, faced constraints, and made difficult tradeoffs. Critical appraisal is about understanding what we can learn from imperfect evidence. In global health, resources are limited and contexts are complex, and the alternative to imperfect evidence is often no evidence at all. Decisions must still be made. The goal of critical appraisal is calibrated confidence: knowing when to act on findings, when to wait for more data, and when to proceed with caution. The papers you read represent our collective effort to understand how to improve health. Read them critically, and also read them generously, recognizing the genuine contributions even imperfect studies make. At the same time, don't shy away from providing honest feedback. Some grant proposals should not be funded, some manuscripts not published. Whenever possible, pair critical appraisal with constructive feedback intended to move the idea forward.