Point-Based Estimate

Prevalence is a point-based estimate, so although at the time of the survey they may have Tb, by the time the study started they may have recovered. Including these people would overestimate the effect as there will be more ‘cases’ than in reality. Alternatively, the results could be underestimated as it is likely that more people will have developed Tb since the survey was conducted. Furthermore, using a point-based estimate like prevalence to obtain cases means that the data will only be valid for that time period. This means that the results will be over or underestimated, but also not generalizable to other time periods, which affects external validity of the study.

Yes: Studying the 85,474 persons identified by the prevalence survey
…show more content…
The authors could have compared time exposed to biomass fuels between men and women, then included interaction terms in their statistical analysis: It appears that exposure duration or fuel composition may act as a modifier of biomass fuel use. The authors state that men are more likely to be exposed due to factors such as tobacco use,1 but that women are more likely to be exposed due to smoke whilst cooking.1 Assuming that the amount of exposure is therefore the same, the differences in Tb may be due to the length of exposure or the composition of the biomass fuel. For example, women who cook for one hour may have very different risks of Tb than a man that smoked for an equivalent amount of time. It would have been beneficial to have measured how long participants were exposed to biomass fuels, and to measure fuel …show more content…
Observational design should be done. This ecological study was:15 Inexpensive – cohort studies are expensive Not time-consuming – cohort studies are time-consuming Only assessing a single effect – a cohort study would be no better in this situation Not limited by causality issues – although a cohort has clear temporal patterns, there was enough evidence to conclude causality (question 15), so a cohort study would be no better Not limited to prevalence data – a cross-sectional study only provides a ‘snap-shot’ so is not suited to this study Unlikely to be affected by recall bias and selection bias as data was gathered at the population level – cohort and case-control studies are limited by this, so an ecological study is better than a cohort and case-control study Identifying data at a country and population level – best answers the research question Not complicated – measuring the exposure/outcome relationship can be a limitation, but not for this study Not subject to ecological fallacy as the authors limited their inferences to the population

Related Documents