Article Text

THU0364 Deriving a Comorbidity Index form the Meddra Classification: Performance of Rheumatic Disease Comorbidity index, Charlson-Deyo Index and Functional Comorbidity Index Among Patients with RA in Nor-Dmard Cohort
  1. P. Putrik1,2,
  2. S. Ramiro3,
  3. E. Lie4,
  4. A.P. Keszei5,
  5. T.K. Kvien4,
  6. T. Uhlig4,
  7. A. Boonen2
  1. 1Health Promotion, Maastricht University, CAPHRI
  2. 2Rheumatology, Maastricht University, CAPHRI, MUMC, Maastricht
  3. 3Rheumatology, LUMC, Leiden, Netherlands
  4. 4Rheumatology, Diakonhjemmet Hospital, Oslo, Norway
  5. 5Medical Informatics, Uniklinik RWTH Aachen University, Aachen, Germany


Background Comorbidities have an important impact on outcomes in chronic diseases. A new and simple to compute index, the Rheumatic Disease Comorbidity Index1 (RDCI), has recently been proposed in addition to existing indices. Evidence on its performance in relation to functional status and health-related quality of life compared to other known indices is scarce.

Medical Dictionary for Regulatory Activities (MedDRA) is a clinically validated international medical terminology dictionary used by regulatory authorities and by researchers to code comorbidities and adverse events. However, no simple algorithm is as yet available to convert MedDRA classification into existing comorbidity indices.

Objectives To develop algorithms to calculate the RDCI, Charlson-Deyo index (CDI) and Functional comorbidity index (FCI)2 from MedDRA, and to test in patients with RA how MedDRA-derived indices predict function (HAQ) and health-related quality of life (based on SF-36).

Methods First, two researchers coded the conditions listed in MedDRA classification into the categories of each index. Next, using data from patients with RA from the Norwegian NOR-DMARD study (2000-2012), we computed and tested predictive values of the RDCI, CDI, and FCI for physical function (HAQ) and Physical and Mental Component Summary measures (PCS and MCS) from SF-36). Outcomes (HAQ, MCS and PCS) were modeled at baseline and over time. Two models were constructed for each outcome: a bare model (with age and gender) and a clinical model (including also DAS28). Generalised estimating equations (GEE) (outcome over time) and linear regression models (outcome at baseline) were fitted and model fit measures (the quasi likelihood under the independence model criterion (QIC) for GEE and R-square for linear regression) were compared. We examined which of the three indices provided the best model fit to draw conclusions about the comparative performance of the indices: the lower the QIC or the higher the R-square, the better the model fit.

Results Data from 4,080 patients were analysed (28.4% male, mean age 56 years, mean DAS28 at baseline 4.9). RDCI (mean 0.6, range 0-6) and FCI (mean 0.40, range 0-6) performed comparably well in predicting the three outcomes considered. CDI (mean 0.24, range 0-7) performed worst on all outcomes HAQ, SF-36 PCS and MCS. Of note, the comorbidities had almost no influence on SF-36 MCS (Table).

Conclusions We have shown that the MedDRA classification, which is widely used in registries and clinical trials (also in rheumatology), can be used to compute currently used comorbidity indices. The new RDCI performed comparably well with FCI on both HAQ and the SF-36 (both physical and mental components). CDI performed worst on all outcomes explored, but it needs to be reminded the CDI was developed to predict mortality and not functioning.


  1. England 2014 Arthritis Care & Research

  2. Groll 2005 J Clin Epid

Disclosure of Interest None declared

Statistics from

Request permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.