METHODOLOGY FOR THE ENDOCRINOLOGIST: Basic aspects of confounding adjustment

Dekkers, Olaf M.

doi:10.1530/eje-20-0075

Cited by 8 publications

(2 citation statements)

References 16 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…However, while our results seem to support the face validity of the concept that early continuity of care by psychiatrists and GPs is ‘causing’ a reduction of the psychiatric re‐hospitalisation rate, the statistical association identified is open to many different interpretations. The reason is that in retrospective observational studies confounding by unmeasured variables is always possible (Groenwold & Dekkers, 2020 ). Causal inference approaches, as the target trial approach (Hernán, 2021 ), could allow for stronger causal claims, but would rely on more detailed information on potential confounders, which is not available in our data set.…”

Section: Discussionmentioning

confidence: 99%

Are early post‐discharge physician contacts associated with 30‐day psychiatric re‐hospitalisation? A nationwide claims data based retrospective cohort study in Austria free of immortal time bias

Katschnig,

Straßmayr,

Endel

et al. 2023

Int J Methods Psych Res

View full text Add to dashboard Cite

ObjectivesCost containment and quality of care considerations have increased research interest in the potential preventability of early re‐hospitalisations. Various registry‐based retrospective cohort studies on psychiatric re‐hospitalisation have focused on the role of early post‐discharge service contacts, but either did not consider their time‐dependent nature (‘immortal time bias’) or evaded the issue by analysing late re‐hospitalisations. The present study takes care of the immortal time bias in studying early psychiatric re‐hospitalisations.MethodsIn a retrospective cohort study using nationwide electronic claims data in Austria, 10,689 adults discharged from acute psychiatric inpatient wards were followed up for 30 days. Cox regression analyses were performed with post‐discharge psychiatric and general practitioner contacts as time‐dependent covariates and time to first psychiatric re‐hospitalisation as outcome.ResultsPost‐discharge ambulatory physician contacts were significantly associated with a decreased psychiatric re‐hospitalisation rate (hazard ratio 0.77 [95% CI 0.69; 0.87], p < 0.0001), with similarly strong contributions to this association by general practitioners and psychiatrists.ConclusionsDespite avoiding the immortal time bias and controlling for several confounders, we suggest to be cautious with a causal interpretation of the identified association, since potentially relevant confounders, such as disease severity, were unavailable in our claims data base.

show abstract

Section: Discussionmentioning

confidence: 99%

Are early post‐discharge physician contacts associated with 30‐day psychiatric re‐hospitalisation? A nationwide claims data based retrospective cohort study in Austria free of immortal time bias

Katschnig,

Straßmayr,

Endel

et al. 2023

Int J Methods Psych Res

View full text Add to dashboard Cite

show abstract

“…This shows that content knowledge should guide the statistical model, not the other way round. Linear regression can be expanded to adjust for confounding in a multivariable model (1).…”

Section: Regressionmentioning

confidence: 99%

Correlation or regression, that’s the question

Cessie

Dekkers

2021

European Journal of Endocrinology

Self Cite

View full text Add to dashboard Cite

There are different ways to quantify the relation between two or more continuous variables. Some researchers use correlation coefficients; others will apply regression methods such as linear regression. In this paper we show that the choice between correlation and regression is not purely a statistical one, but largely depends on the research aims. Importantly, one should always inspect the data before using either of the two methods.

show abstract

Is it a risk factor, a predictor, or even both? The multiple faces of multivariable regression analysis

Groenwold

Dekkers

2023

European Journal of Endocrinology

View full text Add to dashboard Cite

The medical research literature is abundant with regression analyses that include multiple covariates, so-called multivariable regression models. Despite their common application, the interpretation of their results is not always clear or claimed interpretations are not justified. To outline the distinctions between different interpretations, we describe several possible research objectives for which a multivariable regression analysis might be an appropriate way of analyzing the data. In addition, we describe caveats in the interpretation of results of multivariable regression analysis.

show abstract

METHODOLOGY FOR THE ENDOCRINOLOGIST: Basic aspects of confounding adjustment

Cited by 8 publications

References 16 publications

Are early post‐discharge physician contacts associated with 30‐day psychiatric re‐hospitalisation? A nationwide claims data based retrospective cohort study in Austria free of immortal time bias

Are early post‐discharge physician contacts associated with 30‐day psychiatric re‐hospitalisation? A nationwide claims data based retrospective cohort study in Austria free of immortal time bias

Correlation or regression, that’s the question

Is it a risk factor, a predictor, or even both? The multiple faces of multivariable regression analysis

Contact Info

Product

Resources

About