Duration analysis in longitudinal studies with intermittent observation times and losses to followup

2013

Stat Biosci

SummaryObservational cohort studies of individuals with chronic disease provide information on rates of disease progression, the effect of fixed and time-varying risk factors, and the extent of heterogeneity in the course of disease. Analysis of this information is often facilitated by the use of multistate models with intensity functions governing transition between disease states. We discuss modeling and analysis issues for such models when individuals are observed intermittently. Frameworks for dealing with heterogeneity and measurement error are discussed including random effect models, finite mixture models, and hidden Markov models. Cohorts are often defined by convenience and ways of addressing outcome-dependent sampling or observation of individuals are also discussed. Data on progression of joint damage in psoriatic arthritis and retinopathy in diabetes are analysed to illustrate these issues and related methodology.

Section: Outcome-dependent Inspection Times or Loss To Followupmentioning

confidence: 99%

Section: Outcome-dependent Inspection Times or Loss To Followupmentioning

confidence: 99%

Statistical Issues in Modeling Chronic Disease in Cohort Studies

Cook

2013

Stat Biosci

“…If R i ( t ij ) is conditionally independent of observed life and covariate history

{\overset{D}{falsemml-overline}}_{i} (t)

for t > t i , j − 1 given

{\overset{D}{falsemml-overline}}_{i} (t_{i MathClass-punc, j MathClass-bin- 1})

, then we can use the (partial) likelihoods and for estimation, with k i = max j ( R i ( t ij ) = 1). If this ‘sequential missing at random’ (SMAR) condition does not hold but there exists a vector

x_{i}^{c} (t_{i MathClass-punc, j MathClass-bin- 1})

of observed variables such that R i ( t ij ) is conditionally independent of

{{\overset{D}{falsemml-overline}}_{i} (t) MathClass-punc, t MathClass-rel> t_{i MathClass-punc, j MathClass-bin- 1}}

given

x_{i}^{c} (t_{i MathClass-punc, j MathClass-bin- 1})

and

{\overset{D}{falsemml-overline}}_{i} (t_{i MathClass-punc, j MathClass-bin- 1})

, then we can use inverse probability of censoring weights to adjust the log likelihoods or estimating functions on the basis of or .…”

Section: Some Technical Issuesmentioning

confidence: 99%

“…If R i .t ij / is conditionally independent of observed life and covariate history D i .t / for t > t i;j 1 given D i .t i;j 1 /, then we can use the (partial) likelihoods (4) and (5) for estimation, with k i D max j .R i .t ij / D 1/. If this 'sequential missing at random' (SMAR) condition [50] does not hold but there exists a vector x c i .t i;j 1 / of observed variables such that R i .t ij / is conditionally independent of fD i .t /; t > t i;j 1 g given x c i .t i;j 1 / and D i .t i;j 1 /, then we can use inverse probability of censoring weights to adjust the log likelihoods or estimating functions on the basis of (4) or (5) [1,51].…”

Section: Losses To Follow-upmentioning

confidence: 99%

See 1 more Smart Citation

Armitage Lecture 2011: the design and analysis of life history studies

2013

Statistics in Medicine

SummaryLife history studies collect information on events and other outcomes during people's lifetimes. For example, these may be related to childhood development, education, fertility, health, or employment. Such longitudinal studies have constraints on the selection of study members, the duration and frequency of follow-up, and the accuracy and completeness of information obtained. These constraints, along with factors associated with the definition and measurement of certain outcomes, affect our ability to understand, model, and analyze life history processes. My objective here is to discuss and illustrate some issues associated with the design and analysis of life history studies.

Estimation of finite population duration distributions from longitudinal survey panels with intermittent followup

Hajducek

2013

Lifetime Data Anal

Self Cite

SummaryWe consider survival or duration times associated with spells (sojourns in some state) or events experienced by individuals in a population over a specified time period. Duration distributions can be estimated from data recorded during followup of panel members in longitudinal surveys, but adjustments for the sample design, population structure and losses to followup are typically required. We provided weighted Kaplan-Meier estimates that allow for these features and, in particular, adjust for dependent loss to followup through the use of inverse probability of censoring weights.