European Journal of Epidemiology

, Volume 20, Issue 8, pp 657–662 | Cite as

Methods to Account for Attrition in Longitudinal Data: Do They Work? A Simulation Study

  • Vicki L. Kristman
  • Michael Manno
  • Pierre Côté
Methods

Abstract

Attrition threatens the internal validity of cohort studies. Epidemiologists use various imputation and weighting methods to limit bias due to attrition. However, the ability of these methods to correct for attrition bias has not been tested. We simulated a cohort of 300 subjects using 500 computer replications to determine whether regression imputation, individual weighting, or multiple imputation is useful to reduce attrition bias. We compared these results to a complete subject analysis. Our logistic regression model included a binary exposure and two confounders. We generated 10, 25, and 40% attrition through three missing data mechanisms: missing completely at random (MCAR), missing at random (MAR) and missing not at random (MNAR), and used four covariance matrices to vary attrition. We compared true and estimated mean odds ratios (ORs), standard deviations (SDs), and coverage. With data MCAR and MAR for all attrition rates, the complete subject analysis produced results at least as valid as those from the imputation and weighting methods. With data MNAR, no method provided unbiased estimates of the OR at attrition rates of 25 or 40%. When observations are not MAR or MCAR, imputation and weighting methods may not effectively reduce attrition bias.

Keywords

Bias (epidemiology) Cohort studies Computer simulation Epidemiologic methods Logistic models 

Abbreviations

CI

confidence interval

MAR

missing at random

MCAR

missing completely at random

MNAR

missing not at random

MTBI

mild traumatic brain injury

OR

odds ratio

SD

standard deviation

Preview

Unable to display preview. Download preview PDF.

Unable to display preview. Download preview PDF.

References

  1. 1.
    Pennefather, PM, Tin, W, Clarke, MP, Dutton, J, Fritz, S, Hey, EN 1999Bias due to incomplete follow up in a cohort studyBr J Ophthalmol83643645PubMedGoogle Scholar
  2. 2.
    Touloumi, G, Pocock, SJ, Babiker, AG, Darbyshire, JH 2002Impact of missing data due to selective dropouts in cohort studies and clinical trialsEpidemiology13347355CrossRefPubMedGoogle Scholar
  3. 3.
    Greenland, S 1977Response and follow-up bias in cohort studiesAm J Epidemiol106184187PubMedGoogle Scholar
  4. 4.
    Schafer, JL, Graham, JW 2002Missing data: Our view of the state of the artPsychol Methods7147177CrossRefPubMedGoogle Scholar
  5. 5.
    Engels, JM, Diehr, P 2003Imputation of missing longitudinal data: A comparison of methodsJ Clin Epidemiol56968976CrossRefPubMedGoogle Scholar
  6. 6.
    Heath, AC, Madden, PA, Martin, NG 1998Assessing the effects of cooperation bias and attrition in behavioral genetic research using data-weightingBehav Genet28415427CrossRefPubMedGoogle Scholar
  7. 7.
    Landrum, MB, Becker, MP 2001A multiple imputation strategy for incomplete longitudinal dataStat Med2027412760CrossRefPubMedGoogle Scholar
  8. 8.
    Musil, CM, Warner, CB, Yobas, PK, Jones, SL 2002A comparison of imputation techniques for handling missing dataWest J Nurs Res24815829CrossRefPubMedGoogle Scholar
  9. 9.
    Crawford, SL, Tennstedt, SL, McKinlay, JB 1995A comparison of analytic methods for non-random missingness of outcome dataJ Clin Epidemiol48209219CrossRefPubMedGoogle Scholar
  10. 10.
    Schafer, JL 1999Multiple imputation: A primerStat Methods Med Res8315CrossRefPubMedGoogle Scholar
  11. 11.
    Barzi, F, Woodward, M 2004Imputations of missing values in practice: Results from imputations of serum cholesterol in 28 cohort studiesAm J Epidemiol1603445CrossRefPubMedGoogle Scholar
  12. 12.
    Rubin, DB 1996Multiple imputation after 18+ yearsJ Am Stat Assoc91473489Google Scholar
  13. 13.
    Little, RJA, Rubin, DB 1987Statistical Analysis with Missing DataJohn Wiley & SonsNew YorkGoogle Scholar
  14. 14.
    Bisgard, KM, Folsom, AR, Hong, CP, Sellers, TA 1994Mortality and cancer rates in nonrespondents to a prospective study of older women: 5-year follow-upAm J Epidemiol1399901000PubMedGoogle Scholar
  15. 15.
    McLean, RR, Hannan, MT, Epstein, BE,  et al. 2000Elderly cohort study subjects unable to return for follow-up have lower bone mass than those who can returnAm J Epidemiol151689692PubMedGoogle Scholar
  16. 16.
    Kristman, V, Manno, M, Côté, P 2004Loss to follow-up in cohort studies: How much is too much?Eur J Epidemiol19751760CrossRefPubMedGoogle Scholar
  17. 17.
    Twisk, J, Vente, W 2002Attrition in longitudinal studies: How to deal with missing dataJ Clin Epidemiol55329337CrossRefPubMedGoogle Scholar
  18. 18.
    Rao, RS, Sigurdson, AJ, Doody, MM, Graubard, BI 2005An application of a weighting method to adjust for nonresponse in standardized incidence ratio analysis of cohort studiesAnn Epidemiol15129136CrossRefPubMedGoogle Scholar
  19. 19.
    Makela, P 2003Impact of correcting for nonresponse by weighting on estimates of alcohol consumptionJ Stud Alcohol64589596PubMedGoogle Scholar
  20. 20.
    Gerberich, SG, Church, TR, McGovern, PM,  et al. 2004An epidemiological study of the magnitude and consequences of work related violence: The Minnesota Nurses’ StudyOccup Environ Med61495503CrossRefPubMedGoogle Scholar
  21. 21.
    Hopman, WM, Berger, C, Joseph, L,  et al. 2004Stability of normative data for the SF-36Can J Public Health95387391PubMedGoogle Scholar
  22. 22.
    Mishra, GD, Dobson, AJ 2004Multiple imputation for body mass index: Lessons from the Australian Longitudinal Study on Women’s HealthStatist Med2330773087CrossRefGoogle Scholar
  23. 23.
    SAS Publishing. SAS/STAT Software: Changes and Enhancements, Release 8.1. Cary, NC: SAS Institute, 2000Google Scholar
  24. 24.
    Elashoff, JD, Elashoff, RM 1974Two-sample problems for a dichotomous variable with missing dataAppl Stat232634Google Scholar
  25. 25.
    SAS Publishing. SAS/STAT User’s Guide, Version 8, Volumes 1, 2 and 3. Cary, NC: SAS Institute, 1999Google Scholar
  26. 26.
    Kleijnen, JPC 1979

    The role of statistical methodology in simulation

    Zeigler, BP eds. Methodology in Systems Modelling and SimulationNorth-HollandAmsterdam
    Google Scholar
  27. 27.
    Youk, AO, Stone, RA, Marsh, GM 2004A method for imputing missing data in longitudinal studiesAnn Epidemiol14354361CrossRefPubMedGoogle Scholar
  28. 28.
    Taylor, JM, Cooper, KL, Wei, JT, Sarma, AV, Raghunathan, TE, Heeringa, SG 2002Use of multiple imputation to correct for nonresponse bias in a survey of urologic symptoms among African-American menAm J Epidemiol156774782CrossRefPubMedGoogle Scholar
  29. 29.
    Little, RJA 1995Modeling the drop-out mechanism in repeated-measures studiesJ Am Stat Assoc9011121121MathSciNetGoogle Scholar
  30. 30.
    Maldonado, G, Greenland, S 1997The importance of critically interpreting simulation studiesEpidemiology8453456PubMedGoogle Scholar

Copyright information

© Springer 2005

Authors and Affiliations

  • Vicki L. Kristman
    • 1
    • 3
  • Michael Manno
    • 2
    • 3
  • Pierre Côté
    • 1
    • 3
  1. 1.Institute for Work & HealthTorontoCanada
  2. 2.Samuel Lunenfeld Research InstituteMount Sinai HospitalTorontoCanada
  3. 3.Department of Public Health SciencesUniversity of TorontoTorontoCanada

Personalised recommendations