FRINK Linked Data Fragments server

Matches in SemOpenAlex for { <https://semopenalex.org/work/W81893890> ?p ?o ?g. }

• W81893890 abstract "Occupational epidemiology is most commonly used to investigate causes of disease or illness, and to plan and evaluate preventive strategies. However, another important application lies in identifying and characterising predictors of future outcomes so that risks can be better managed. For example, in deciding whether and at what stage a fork lift truck driver who has suffered a single generalised epileptic seizure should be allowed to resume his job, we need to know the probability that he will experience a further sudden loss of consciousness while driving. This will depend on various prognostic indices, some more important than others.Aetiological and prognostic investigations both explore the relation of an outcome (e.g. the occurrence of a disease or some other adverse event) to risk factors, often using multivariate modelling. However, the optimal strategy for multivariate analysis differs according to the purpose of the study.In an aetiological investigation, the aim is to estimate the causal impact of a risk factor, usually in terms of a relative risk or odds ratio. The risk that we would like to establish is that which would be found in a theoretical infinite randomised experiment. To this end, we adjust risk estimates from observational data for known and suspected confounders, these being factors that are statistically associated with the risk factor under investigation, and which independently determine risk of the relevant health outcome. An ideal statistical model would adjust risk estimates for all confounding factors, but not for other variables which are not confounders.In prognostic investigations, causation and confounding are irrelevant. It does not matter why a risk factor is related to an outcome, only the extent to which its occurrence or level is associated with a different probability of the outcome. The best analytical model will include those risk factors which in combination predict individual probabilities of the outcome most efficiently. This entails a trade-off between the increased discrimination of risk when an extra risk factor is incorporated in the statistical model and the added costs that would be entailed in ascertaining or measuring that risk factor if it were to be used in clinical practice. Even if they are significantly related to the outcome of interest, risk factors may not be worth including in a prognostic model if they are expensive to measure or are associated with little difference in the probability that the outcome will occur.The paper by Detaille et al (1) describes a systematic review of prognostic factors for work disability among employees with various chronic diseases. Knowing the important predictors of disability in such workers would be useful when making decisions about their employment, and the review points to several factors that might be useful for such prediction. However, the method of analysis has major limitations, and the discussion of future research needs is misleading.Evidence was synthesised according to a grading system based on the number of relevant studies that were identified, and the proportions of those studies that showed statistically significant positive or negative associations between a risk factor and the outcome under consideration. The authors adopted this approach because they felt it combined “the statistical rigour of meta-analyses to synthesise qualitative findings as well as the flexibility of traditional narrative reviews”. However, unlike a good narrative review, the method did not take into account the potential magnitude and direction of bias in individual studies. Moreover, no consideration was given to the predictive value of risk factors, either alone or in combination. Thus, a risk factor could be classed as a strong predictor of an outcome if at least three relevant studies were found, with more than half of them showing a significant association in the same direction, and the rest finding no significant association in either direction. It would not matter that the probability of the outcome in someone with the risk factor was little different from that in someone without the risk factor, provided the difference was demonstrated in studies of sufficient size that it was statistically significant. Furthermore, if additional small and uninformative studies were published, in which there were no statistically significant associations, the level of evidence could be downgraded, even if two very large and well-conducted investigations had found the risk factor to be a powerful predictor. Given this limitation, little weight can be given to the reported grading of predictors, and risk factors identified in the review cannot be considered more than candidates for further research.In their recommendations for further research, the authors advocate “a combination of strategies” with “more efficient control of confounding”. As already explained, confounding is irrelevant in studies of prognosis. The need is for studies that systematically explore the probabilities (with associated confidence intervals) of poor employment outcome for combinations of risk factors that can reasonably be ascertained in routine clinical practice. For a predictive algorithm to be of practical value, the absolute differences in risk that it discriminates must be sufficiently large, and they must apply to sufficient numbers of workers." @default.
• W81893890 created "2016-06-24" @default.
• W81893890 creator A5044644859 @default.
• W81893890 date "2009-07-01" @default.
• W81893890 modified "2023-09-24" @default.
• W81893890 title "Commentary" @default.
• W81893890 cites W2100163949 @default.
• W81893890 doi "https://doi.org/10.5271/sjweh.1338" @default.
• W81893890 hasPubMedCentralId "https://www.ncbi.nlm.nih.gov/pmc/articles/3630494" @default.
• W81893890 hasPubMedId "https://pubmed.ncbi.nlm.nih.gov/19593895" @default.
• W81893890 hasPublicationYear "2009" @default.
• W81893890 type Work @default.
• W81893890 sameAs 81893890 @default.
• W81893890 citedByCount "3" @default.
• W81893890 countsByYear W818938902012 @default.
• W81893890 crossrefType "journal-article" @default.
• W81893890 hasAuthorship W81893890A5044644859 @default.
• W81893890 hasConcept C105795698 @default.
• W81893890 hasConcept C107130276 @default.
• W81893890 hasConcept C126322002 @default.
• W81893890 hasConcept C156957248 @default.
• W81893890 hasConcept C161584116 @default.
• W81893890 hasConcept C23131810 @default.
• W81893890 hasConcept C2779134260 @default.
• W81893890 hasConcept C33923547 @default.
• W81893890 hasConcept C50440223 @default.
• W81893890 hasConcept C71924100 @default.
• W81893890 hasConcept C77350462 @default.
• W81893890 hasConceptScore W81893890C105795698 @default.
• W81893890 hasConceptScore W81893890C107130276 @default.
• W81893890 hasConceptScore W81893890C126322002 @default.
• W81893890 hasConceptScore W81893890C156957248 @default.
• W81893890 hasConceptScore W81893890C161584116 @default.
• W81893890 hasConceptScore W81893890C23131810 @default.
• W81893890 hasConceptScore W81893890C2779134260 @default.
• W81893890 hasConceptScore W81893890C33923547 @default.
• W81893890 hasConceptScore W81893890C50440223 @default.
• W81893890 hasConceptScore W81893890C71924100 @default.
• W81893890 hasConceptScore W81893890C77350462 @default.
• W81893890 hasLocation W818938901 @default.
• W81893890 hasLocation W818938902 @default.
• W81893890 hasOpenAccess W81893890 @default.
• W81893890 hasPrimaryLocation W818938901 @default.
• W81893890 hasRelatedWork W15971883 @default.
• W81893890 hasRelatedWork W1983038599 @default.
• W81893890 hasRelatedWork W1985035418 @default.
• W81893890 hasRelatedWork W1989993227 @default.
• W81893890 hasRelatedWork W2012727722 @default.
• W81893890 hasRelatedWork W2020511244 @default.
• W81893890 hasRelatedWork W2046991079 @default.
• W81893890 hasRelatedWork W2096258850 @default.
• W81893890 hasRelatedWork W2100551803 @default.
• W81893890 hasRelatedWork W2110799663 @default.
• W81893890 hasRelatedWork W2132278422 @default.
• W81893890 hasRelatedWork W2158762028 @default.
• W81893890 hasRelatedWork W2173929050 @default.
• W81893890 hasRelatedWork W2264733396 @default.
• W81893890 hasRelatedWork W2417380433 @default.
• W81893890 hasRelatedWork W2441670565 @default.
• W81893890 hasRelatedWork W2598476089 @default.
• W81893890 hasRelatedWork W2904086957 @default.
• W81893890 hasRelatedWork W3188662093 @default.
• W81893890 hasRelatedWork W2403126500 @default.
• W81893890 isParatext "false" @default.
• W81893890 isRetracted "false" @default.
• W81893890 magId "81893890" @default.
• W81893890 workType "article" @default.