School of Population Health, The University of Queensland, Herston Road, Herston, Queensland 4006, Australia.
J Clin Epidemiol. 2013 May;66(5):575-81. doi: 10.1016/j.jclinepi.2012.12.002. Epub 2013 Feb 4.
Minimizing attrition is a key objective in longitudinal research, with possible consequences being additional bias and reduced generalizability. Identifying determinants of attrition is essential in determining attrition prevention strategies. The objective of this study was to investigate a number of these determinants, with an emphasis on contactability.
Data were taken from the Passports project, a randomized controlled trial of an intervention to provide postrelease support to ex-prisoners in Queensland, Australia. Measures of contactability included intervention intensity, baseline collaterals, and follow-up telephone calls, with attrition at follow-up being the outcome event. Multivariable modeling was used to assess the independent effects of these measures on attrition.
Attrition was found to be more likely among those who were not contacted between the release and follow-up (adjusted odds ratio [AOR]: 2.93; 95% confidence interval [CI]: 1.87, 4.60), did not provide collaterals (AOR: 2.58; 95% CI: 1.68, 3.97), and received more than four telephone calls (AOR: 2.42; 95% CI: 1.61, 3.63). Evidence of dose-response relationships between attrition and the measures of collaterals and telephone calls was also seen to exist.
These findings have implications for sample size maintenance, especially those involving hard-to-reach populations. Subject to cost constraints and possible diminishing returns, researchers should endeavor to implement a study protocol that facilitates continued contact during follow-up.
减少流失是纵向研究的一个关键目标，其可能的后果是额外的偏差和降低的普遍性。确定流失的决定因素对于确定流失预防策略至关重要。本研究的目的是调查其中的一些决定因素，重点是可接触性。
数据来自护照项目，这是一项在澳大利亚昆士兰州对出狱囚犯提供释放后支持的干预措施的随机对照试验。可接触性的衡量标准包括干预强度、基线抵押品和随访电话，随访时的流失是结局事件。多变量建模用于评估这些措施对流失的独立影响。
发现在释放和随访之间未联系的人（调整后的优势比 [AOR]：2.93；95%置信区间 [CI]：1.87，4.60）、未提供抵押品（AOR：2.58；95% CI：1.68，3.97）和接受超过四次电话的人更有可能流失（AOR：2.42；95% CI：1.61，3.63）。还存在抵押品和电话次数与流失之间存在剂量-反应关系的证据。
这些发现对样本量维持具有影响，特别是对于那些难以接触的人群。受成本限制和可能的收益递减的影响，研究人员应努力实施一项研究协议，以便在随访期间保持持续联系。
