Premium
Analysis strategies for longitudinal attachment loss data
Author(s) -
Beck James D.,
Elter John R.
Publication year - 2000
Publication title -
community dentistry and oral epidemiology
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 1.061
H-Index - 101
eISSN - 1600-0528
pISSN - 0301-5661
DOI - 10.1034/j.1600-0528.2000.280101.x
Subject(s) - medicine , poisson regression , poisson distribution , statistics , logistic regression , clinical attachment loss , tooth loss , longitudinal data , longitudinal study , econometrics , linear regression , regression analysis , epidemiology , principal component analysis , statistical model , regression , random effects model , dentistry , periodontal disease , data mining , mathematics , pathology , environmental health , computer science , meta analysis , population , oral health
– The purpose of this invited review is to describe and discuss methods currently in use to quantify the progression of attachment loss in epidemiological studies of periodontal disease, and to make recommendations for specific analytic methods based upon the particular design of the study and structure of the data. The review concentrates on the definition of incident attachment loss (ALOSS) and its component parts; measurement issues including thresholds and regression to the mean; methods of accounting for longitudinal change, including changes in means, changes in proportions of affected sites, incidence density, the effect of tooth loss and reversals, and repeated events; statistical models of longitudinal change, including the incorporation of the time element, use of linear, logistic or Poisson regression or survival analysis, and statistical tests; site vs person level of analysis, including statistical adjustment for correlated data; the strengths and limitations of ALOSS data. Examples from the Piedmont 65+ Dental Study are used to illustrate specific concepts. We conclude that incidence density is the preferred methodology to use for periodontal studies with more than one period of follow‐up and that the use of studies not employing methods for dealing with complex samples, correlated data, and repeated measures does not take advantage of our current understanding of the site‐ and person‐level variables important in periodontal disease and may generate biased results.