Background

Attrition occurs once participants leave during a study. It virtually always wake up to some extent.

Different rates of loss come follow-up in the exposure groups, or losses of different varieties of participants, whether at comparable or various frequencies, may adjust the characteristics of the groups, regardless of whether of the exposure or intervention. Losses may be affected by such factors as unsatisfactory therapy efficacy or intolerable adverse events.

You are watching: Which of the following is not a valid rule of thumb on risk control strategy selection?

When entrants leave, it might not be well-known whether they proceed or discontinue an intervention; there might be no data top top outcomes for these entrants after that time.

Systematic differences in between people who leave the study and those who continue can introduce prejudice into a study’s outcomes – this is attrition bias. However, the results might not have to be biased, in spite of different drop-out rates in the groups. We discuss listed below how to assess the impact of different amounts of attrition.

In part cases, those who leave a examine are likely to be different from those who continue. For instance, in an intervention study the diet in civilization with depression, those with more severe depression might find that harder come adhere come the diet regimen and also therefore more likely to leaving the study.

Example

A study of psychosocial factors amongst patients with cardiac conditions showed the those who completely completed the study differed in clinical and psychosocial functions from those that dropped out before the research ended. Together differential attrition could have biased the study’s results.

A trial investigate the high quality of life amongst patients randomised to aggressive treatment of renal cancer had high prices of attrition owing to toxicity, disease progression, and deaths (64% in the manage group; 70% in the intervention group). Analysis of those tho in the trial proved no distinction in the quality of life. The affect of attrition bias, however, said that even with equal drop-outs in both groups a biased calculation occurred.

Impact

A systematic evaluation assessed the reporting, extent, and also handling the loss to follow-up and also its potential impact, on treatment results in randomised controlled trials published in the 5 top medical journals, The authors calculated the portion of trials in i beg your pardon the family member risk would no longer be far-reaching when attendees loss to follow-up varied. In 160 trials, with an mean loss to follow-up of 6%, and also assuming various event prices in the intervention groups relative come the control groups, between 0% and also 33% of trials were no much longer significant.


*

Potential influence on estimated treatment results of information lost come follow-up in randomised controlled trials (LOST-IT): a organized review. BMJ 2012;344:e2809.


Preventive steps

Techniques for avoiding losses follow-up include ensuring great communication in between study staff and also participants, availability to clinics, effective communication channels, incentives to continue, and ensuring that the study is of relevance to the participants.

However, for numerous studies, complete follow up is unlikely. In together cases, the factors for attrition must be very closely considered. After ~ the study has actually been completed, a variety of analysis methods deserve to be used to mitigate the influence of attrition bias.

Intention come treat analysis: since anything the happens ~ randomisation can impact the possibility that a study participant has the result of interest, it is necessary that all patients (even those that fail to take their medication or by chance or intentionally receive the not correct treatment) are analysed in the groups to which they to be allocated.

It is essential that us not just look because that the term ‘intention-to-treat analysis’ in the methods but additionally look at the results to ensure the the analysis was in reality done.

Methods because that dealing with lacking data include last monitoring (or baseline value) carried forward, combined models, imputation and also sensitivity evaluation using ‘worst case’ scenarios (assuming that those through no info all gained worse) and also ‘best case’ scenarios (assuming that all gained better). Analysing data just from participants continuing to be in the examine is called complete case analysis.

A ascendancy of thumb says that 20% poses severe threats come validity. If this is useful, that is essential to note that even little proportions the patients lost to follow-up can cause significant bias. One method to determine whether losses come follow-up have the right to seriously influence results is to i think a worst-case scenario because that the outcomes in those with absent data and look to watch if the results would change. If this technique doesn’t readjust the study’s conclusions, the loss to follow-up is most likely not a threat to the study’s validity.

Regardless of the mechanisms provided to obtain estimates of outcome data, the reasons that participants leaving the study have to be very closely considered: if human being leave for reasons unrelated to the exposure (treatment) or the result this may have tiny or no affect on the results.


Cite as

Catalogue of predisposition Collaboration, Bankhead C, Aronson JK, Nunan D. Attrition bias. In: Catalogue Of predisposition 2017. Https://sirhenryjones-museums.org/biases/attrition-bias/

Related biases


Cite as

Catalogue of prejudice Collaboration, Bankhead C, Aronson JK, Nunan D. Attrition bias. In: Catalogue Of prejudice 2017. Https://sirhenryjones-museums.org/biases/attrition-bias/


The James Lind LibraryThe must compare like-with-like in treatment comparisons

Related biases

selection bias


All biases

pick a biasAdmission rate biasAll"s well literary works biasAllocation biasApprehension biasAscertainment biasAttrition biasAvailability biasBiases that rhetoricCentripetal biasChronological biasCollider biasCompliance biasConfirmation biasConfoundingConfounding by indicationData-dredging biasDetection biasDiagnostic accessibility biasDiagnostic skepticism biasDifferential reference biasHawthorne effectHot stuff biasHypothetical biasImmortal time biasIncorporation biasIndustry Sponsorship biasInformation biasInformed visibility biasInsensitive measure up biasLack that blindingLanguage biasLead time biasMimicry biasMisclassification biasNon-contemporaneous control biasNon-response biasNovelty biasObserver biasOne-sided referral biasOutcome reporting biasPartial referral biasPerception biasPerformance biasPopularity biasPositive outcomes biasPrevalence-incidence (Neyman) biasPrevious opinion biasPublication biasRecall biasReferral filter biasReporting biasesSelection biasSpectrum biasSpin biasStarting time biasSubstitution game biasUnacceptability biasUnacceptable condition biasUnmasking (detection signal) biasVerification biasVolunteer biasWrong sample size bias


Sources

Akl AE et al. Potential influence on estimated treatment results of info lost to follow-up in randomised regulated trials (LOST-IT): methodical review. BMJ 2012; 344: e2809.

Bell M et al. Differential dropout and bias in randomised regulated trials: when it matters and also when it might not. BMJ 2013; 346: e8668.

Damen N, et al. Cardiac patients that completed a longitudinal psychosocial study had a various clinical and psychosocial baseline profile 보다 patients that dropped the end prematurely. Eur J Prev Cardiol. 2015; 22(2): 196-9.

Hewitt CE et al. Trial attrition examine group. Assessing the impact of attrition in randomized regulated trials. J Clin Epidemiol. 2010; 63(11): 1264-70.

Porta M et al. Editors. A dictionary of epidemiology. 6th edition. Brand-new York: Oxford college Press: 2014.

See more: Was Sonny Bono Ski Accident Helmet On When He Died? Kennedy, Bono Deaths Put Spotlight On Helmet

Zethof D et al. Attrition analysed in five waves that a longitudinal yearly survey of smokers: result from the ITC Netherlands survey. Eur J Public health 2016; 26(4): 693-9.