Quantifying psychiatric comorbidity--lessions from chronic disease epidemiology

Area of Science:

• Epidemiology
• Psychiatric Epidemiology
• Chronic Disease Epidemiology

Background:

• Psychiatric epidemiology commonly uses measures of association (odds/risk ratios) to quantify disorder links.
• Chronic disease epidemiology increasingly employs measures of clustering (multimorbidity coefficients) for comorbidity studies.
• This article critically compares association and clustering measures in comorbidity research.

Purpose of the Study:

• To compare measures of association and clustering for studying comorbidity.
• To evaluate the suitability of different statistical measures for etiological versus descriptive comorbidity research.

Main Methods:

• Narrative review of existing literature.
• Algebraical examples to illustrate statistical concepts.
• Secondary analysis of an existing dataset.
• Pooled analysis of published data.

Main Results:

• Measures of association (odds/risk ratios) can confound clustering with coincidental comorbidity.
• Multimorbidity coefficients offer a pure estimate of etiological clustering.
• Clustering coefficients can analyze multimorbidity of any number of disorders, unlike association measures limited to two.
• Clustering coefficients are less dependent on illness prevalence in study groups.

Conclusions:

• Measures of association are suitable for descriptive comorbidity research, identifying co-occurring disorders and their nosological implications.
• Clustering coefficients are superior for etiological research, investigating the underlying reasons for multiple health problems in individuals.