Abstract Objective To develop algorithms for calculating the Rheumatic Diseases Comorbidity Index (RDCI), Charlson–Deyo Index (CDI) and Functional Comorbidity Index (FCI) from the Medical Dictionary for Regulatory Activities (MedDRA), and to assess how these MedDRA-derived indices predict clinical outcomes, utility and health resource utilization (HRU). Methods Two independent researchers linked the preferred terms of the MedDRA classification into the conditions included in the RDCI, the CDI and the FCI. Next, using data from the Norwegian Register-DMARD study (a register of patients with inflammatory joint diseases treated with DMARDs), the explanatory value of these indices was studied in models adjusted for age, gender and DAS28. Model fit statistics were compared in generalized estimating equation (prediction of outcome over time) models using as outcomes: modified HAQ, HAQ, physical and mental component summary of SF-36, SF6D and non-RA related HRU. Results Among 4126 patients with RA [72% female, mean (s.d.) age 56 (14) years], median (interquartile range) of RDCI at baseline was 0.0 (1.0) [range 0–6], CDI 0.0 (0.0) [0–7] and FCI 0.0 (1.0) [0–6]. All the comorbidity indices were associated with each outcome, and differences in their performance were moderate. The RDCI and FCI performed better on clinical outcomes: modified HAQ and HAQ, hospitalization, physical and mental component summary, and SF6D. Any non-RA related HRU was best predicted by RDCI followed by CDI. Conclusion An algorithm is now available to compute three commonly used comorbidity indices from MedDRA classification. Indices performed comparably well in predicting a variety of outcomes, with the CDI performing slightly worse when predicting outcomes reflecting functioning and health. rheumatoid arthritis, comorbidity index, MedDRA, physical function, quality of life, utility, health resource utilization Rheumatology key messages An algorithm to compute common comorbidity indices from the MedDRA classification is provided. Comorbidity indices explored were comparable in predicting function, health-related quality of life, utility and healthcare utilization. The Rheumatic Disease Comorbidity Index was the simplest to compute while performing as well as the other indices. Introduction Comorbidities in rheumatic and musculoskeletal diseases (RMDs) play an important role as they contribute to disability, need for healthcare and ultimately mortality [1, 2]. Comorbidities can co-occur by chance or can be related to (or enhanced by) the biopathology of the underlying rheumatic disease, the drug treatment or the patient’s lifestyle. Clinical cohorts and registers have come to play an important role in gathering evidence in effectiveness and safety of RA management, and increasingly guide clinical practice and policy decisions . In this context, a standardized approach to assess adverse events and comorbidities is of the utmost importance. Medical Dictionary for Regulatory Activities (MedDRA) offers a five-level classification with nearly 70 000 terms to assess comorbidities and adverse events and is being increasingly used as an alternative to a three-level International Classification of Diseases-10 with approximately 17 000 items. The MedDRA is not only used by drug regulatory authorities, but also recommended to be used as the preferred dictionary for the coding of adverse events in European Biological Registries to ensure the homogeneity of data collection across registries [4, 5]. In rheumatology, not only the Norwegian Register of DMARDs (NOR-DMARD), but also other registries, such as the German (RABBIT), the British (BSRBR), the Portuguese (Reuma.pt) and the Spanish (BIOBADASER), use this classification system [6–10]. MedDRA is used by regulatory authorities and in many clinical cohorts and registers worldwide to classify comorbidities and adverse events . MedDRA is a hierarchical classification, encompassing 26 system organ classes (SOCs), corresponding to the highest level of the hierarchy. SOCs are followed by a high level group term (HLGT), a high level term (HLT), a preferred term (PT) and a lowest level term (LLT). For example, a myocardial infarction can be found in three SOCs at the highest level of hierarchy, namely, cardiac disorders, vascular disorders and injury, poisoning and procedural complications. In cardiac disorders, myocardial infarction is placed in the HLGT group coronary artery disorders, the HLT ischaemic coronary artery disorders with six corresponding PTs, namely acute myocardial infarction, coronary no-reflow phenomenon, myocardial infarction, myocardial stunning, post-procedural myocardial infarction and silent myocardial infarction. Finally, the six PTs are further split into 27 LLTs. In total, there are 69 019 terms in the lowest level of the hierarchy. While the detailed information on comorbid conditions can be essential to answer some specific research questions such as prevalence of certain rare adverse events, it is impractical when aiming to understand the overall role of comorbidities in outcome studies of a specific index disease. For such purposes, a composite comorbidity index that condenses the information on a series of co-morbidities into one single score is more useful. Several comorbidity indices have been developed and validated for use in clinical studies [12, 13]. They differ in the type and number of comorbidities included, approaches to weigh the comorbidities into the index, the type of outcome they aim to predict and whether they use self-report or physician assessed comorbidities. The Charlson–Deyo Index (CDI)  and the Functional Comorbidity Index (FCI)  are among the most often used indices, the former being originally developed to predict mortality in cancer patients and the latter to predict functional status. Recently, a new and simple to compute index, the Rheumatic Disease Comorbidity Index (RDCI) , has been proposed in addition to the existing indices for use in patients with RMDs. While all indices have been validated using information retrieved from clinical records, the RDCI can also be assessed by means of a patient reported questionnaire. Available evidence showed that the RDCI performs comparably well to existing indices in predicting functional status and mortality . However, these data come from one single study. Moreover, the comparative performance of these indices in health care utilization and a broad range of outcomes assessing functioning has never been addressed. Despite the fact that these comorbidity indices as well as disease classification systems such as the MedDRA or the International Classification of Diseases are widely used, as far as we know, so far no algorithm has been proposed for the computation of any comorbidity index from MedDRA. The objective of this study was to develop algorithms to calculate the RDCI, the CDI and the FCI from the MedDRA, and to assess how these MedDRA-derived indices predict physical function, physical and mental health, health utility, and health resource utilization (HRU). Methods Data from the NOR-DMARD were used. NOR-DMARD is a register of patients with inflammatory joint diseases treated with DMARDs, both conventional synthetic (csDMARDs) and biologic (bDMARDs). Five rheumatology departments covering a total of 1.7 million inhabitants (approximately one-third of the total population in Norway) have provided patient data since 2000 . Each time a new DMARD has been started, study variables have been collected at treatment initiation, after 3, 6 and 12 months, and then yearly up to treatment termination. By 2012, the NOR-DMARD database comprised 6839 treatment courses in a total of 4126 patients with confirmed RA. The NOR-DMARD has been approved by the Norwegian Data Inspectorate and the Regional Ethics Committee of Eastern Norway. Patients gave written informed consent before participation . No additional approval was required for this study. Comorbidity indices and their computation from MedDRA Table 1 describes the comorbidity indices considered for this study. The RDCI was developed as a self-reported questionnaire to understand the role of comorbidities on functioning in patients with RA, OA, SLE or FM . It contains 11 conditions (currently present and/or present in the past) and the final score is derived by a sum score, with an extra weight for pulmonary and certain cardiovascular conditions. The CDI was developed to predict mortality and comprises 22 currently present conditions, identified from clinical records or databases. Each condition is assigned a weight of 1, 3 or 6, resulting in a final score ranging from 0 to 36 . The FCI aims to predict physical function by assessing the currently present comorbid conditions. The final score is the unweighted sum of 18 comorbidities (range 0–18) . Table 1 Composition of comorbidity indices Comorbidity index Composition of the index (formula) Original range Rangea Rheumatic diseases comorbidity index (RDCI)b  (2 × (pulmonarynow or pulmonarypast)), + 2 if (MInow or MIpast) or (otherCVnow or otherCVpast) or (CVallnow or CVallpast), plus 1 for each of the following conditions: (hypertensionnow or hypertensionpast), diabetesnow, (fracturenow or fracturepast), (depressionnow or depressionpast), (ulcernow or ulcerpast), (ALLGInow or ALLGIpast) or cancernow 0–9 0–9 Charlson–Deyo Index (CDI)  6 × (metastatic solid tumour + AIDS) + 3 × (severe or moderate liver disease) + 3 × (hemiplegia + renal disease + diabetes with end organ damage + tumour without metastasis + lymphoma + leukaemia) + MI + CHF + PVD + stroke + dementia + COPD + CTD + ulcer + mild liver disease 0–36 0–35 Functional comorbidity index (FCI)  Arthritis + osteoporosis + asthma + COPD + angina + CHF + MI + neurological disease + stroke + PVD + diabetes + upper GI disease + depression + anxiety + visual impairment + hearing impairment + degenerative disc disease + obesity 0–18 0–17 Comorbidity index Composition of the index (formula) Original range Rangea Rheumatic diseases comorbidity index (RDCI)b  (2 × (pulmonarynow or pulmonarypast)), + 2 if (MInow or MIpast) or (otherCVnow or otherCVpast) or (CVallnow or CVallpast), plus 1 for each of the following conditions: (hypertensionnow or hypertensionpast), diabetesnow, (fracturenow or fracturepast), (depressionnow or depressionpast), (ulcernow or ulcerpast), (ALLGInow or ALLGIpast) or cancernow 0–9 0–9 Charlson–Deyo Index (CDI)  6 × (metastatic solid tumour + AIDS) + 3 × (severe or moderate liver disease) + 3 × (hemiplegia + renal disease + diabetes with end organ damage + tumour without metastasis + lymphoma + leukaemia) + MI + CHF + PVD + stroke + dementia + COPD + CTD + ulcer + mild liver disease 0–36 0–35 Functional comorbidity index (FCI)  Arthritis + osteoporosis + asthma + COPD + angina + CHF + MI + neurological disease + stroke + PVD + diabetes + upper GI disease + depression + anxiety + visual impairment + hearing impairment + degenerative disc disease + obesity 0–18 0–17 a Without CTD in CDI and without arthritis in FCI. b In the RDCI, only present conditions were recorded and included in the index. ALLGI: all gastrointestinal; CV: cardiovascular; MI: myocardial infarction; CHF: congestive heart failure; COPD: chronic obstructive pulmonary disease; PVD: peripheral vascular disease. To compute the comorbidity indices from MedDRA, all PTs (second lowest level, which was considered detailed enough for the purpose of the study) were manually linked to the corresponding conditions represented in the three comorbidity indices of interest by two researchers [A.B. (internist/rheumatologist) and S.R. (trainee in rheumatology)] based on consensus. After forward linkage, the conditions in the indices were also searched for in MedDRA to provide reassurance that no term was missed. As the aim was to develop an algorithm to be used by researchers in patients with inflammatory RMDs, ‘Arthritis’ was excluded from the FCI and ‘CTD’ from the CDI. Having been developed as a comorbidity index in RMDs, the RDCI did not include any index disease. Therefore, in this study the score for FCI ranged from 0 to 17 and for CDI from 0 to 35 (Table 1). The link between each of the MedDRA terms and the diseases included in the different comorbidity indices can be found in supplementary Table S1, available at Rheumatology online. Outcomes Physical function was measured by the modified HAQ (mHAQ) (0–3, higher scores are worse) . HAQ was computed from the mHAQ . Physical and mental function were measured by the physical and mental component summary (0–100, the higher the better) of the Short Form 36 (SF-36) . Additionally, a measure of health utility, the SF6D, was computed. SF6D is a societal preference-based single index measure for health, composed of six multi-level dimensions, derived from 11 questions of SF-36 items . The SF-6D includes six health domains: pain, mental health, physical functioning, social functioning, role limitations and vitality; and has 18 000 health states. HRU not related to RA was computed from the number of hospitalizations in a department other than rheumatology in the past 3 months and the number of visits to an outpatient clinic other than rheumatology in the past 3 months. To deal with implausible outliers (>10 for hospitalizations and >20 outpatient visits in the past 3 months), two scenarios were developed: first (liberal) assuming a data entry mistake (10 was recoded with 1, 20 was recoded with 2, etc.), and second (conservative) recoding the outlier into missing. For analyses, HRU was converted into three dichotomous variables, hospitalization (yes/no), outpatient visit (yes/no) and a combined variable reflecting ‘any HRU’ (yes/no) in the past 3 months. Additionally, data on age, gender and DAS28 assessment at each visit were recorded. Statistical methods The predictive value for outcomes over time was assessed by linear and logistic generalized estimating equations (GEEs) with Gaussian family or binomial logit link function (as appropriate) and exchangeable correlation structure which assumes that the correlation between any pair of measurements on the same individual is the same. This technique is suitable for analysis of longitudinal data, and makes efficient use of all information longitudinally collected along all observations for each patient. Models were adjusted for age, gender and DAS28. As a sensitivity analyses, we have also explored the contribution of erosive disease and disease duration. Explanatory value of each of the comorbidity indices with respect to the outcomes was explored by comparing quasi-likelihood under the independence model criterion (QIC, which is a modification of Akaike information criteria for GEE) . We first examined which of the three indices provided the best model fit: the lower the QIC, the better the model fit. Next, the improvement in the model fit was computed as the difference in QIC between the model not adjusted for comorbidities and the model adjusted for each of the three comorbidity indices, in order to capture the effect of adding the corresponding comorbidity index to the model: the higher the absolute value of the improvement, the better. Of note, absolute value of QIC cannot be meaningfully interpreted or compared across models with different outcomes. Analyses were done on complete cases per outcome, that is, the number of cases was the same across models within the same outcome. Analyses were performed with Stata v. 13 (StataCorp, College Station, TX, USA). Results The correspondence between the MedDRA codes and the categories of each index used for building the three comorbidity indices is presented in supplementary Table S1, availabe at Rheumatology online. Data from 4126 patients with RA were analysed [72% female, mean (s.d.) age 56 (14) years]. Mean (s.d.) baseline DAS28 was 4.9 (1.4), and nearly half (n = 1805, 44%) of patients had disease duration of <1 year at baseline. Median (interquartile range) of RDCI at baseline was 0.0 (1.0) [range 0–6], CDI 0.0 (0.0) [0–7] and FCI 0.0 (1.0) [0–6] (Table 2). The most common comorbidity was pulmonary disease (n = 386, 10%) in RDCI, followed by osteoporosis (n = 284, 7%) in FCI. Table 2 Baseline characteristics of included RA patients (n = 4126) Variable Results Age, mean (s.d.), years 55.6 (13.9) Gender, female, n (%) 2951 (71.6) Education at baseline, n (%) Low 1399 (34.4) Intermediate 1406 (34.5) High (college/university) 1266 (31.1) RF, n (%) 2625 (65.1) Erosive disease, n (%) 1910 (48.0) DAS28, mean (s.d.) 4.9 (1.4) Disease duration from diagnosis <1 year, n (%) 1805 (44) mHAQ, mean (s.d.), 0–3 0.7 (0.5) HAQ recalculated from mHAQ, mean (s.d.), 0–3 1.3 (0.6) Pain VAS, mean (s.d.), 0–100 47.4 (24.4) SF-36 PCS, mean (s.d.), 0–100 31.0 (9.9) SF-36 MCS, mean (s.d.), 0–100 45.9 (11.4) Patients with a hospitalization in a department other than rheumatology in the previous 3 months, n (%) 105 (6.6) Patients with an outpatient visit to a department other than rheumatology in the previous 3 months, n (%) 308 (19.4) Rheumatic Diseases Comorbidity Index, median (IQR), 0–9 0.0 (1.0) Charlson–Deyo Comorbidity Index, median (IQR), 0–35 0.0 (0.0) Functional Comorbidity Index, median (IQR), 0–16 0.0 (1.0) Variable Results Age, mean (s.d.), years 55.6 (13.9) Gender, female, n (%) 2951 (71.6) Education at baseline, n (%) Low 1399 (34.4) Intermediate 1406 (34.5) High (college/university) 1266 (31.1) RF, n (%) 2625 (65.1) Erosive disease, n (%) 1910 (48.0) DAS28, mean (s.d.) 4.9 (1.4) Disease duration from diagnosis <1 year, n (%) 1805 (44) mHAQ, mean (s.d.), 0–3 0.7 (0.5) HAQ recalculated from mHAQ, mean (s.d.), 0–3 1.3 (0.6) Pain VAS, mean (s.d.), 0–100 47.4 (24.4) SF-36 PCS, mean (s.d.), 0–100 31.0 (9.9) SF-36 MCS, mean (s.d.), 0–100 45.9 (11.4) Patients with a hospitalization in a department other than rheumatology in the previous 3 months, n (%) 105 (6.6) Patients with an outpatient visit to a department other than rheumatology in the previous 3 months, n (%) 308 (19.4) Rheumatic Diseases Comorbidity Index, median (IQR), 0–9 0.0 (1.0) Charlson–Deyo Comorbidity Index, median (IQR), 0–35 0.0 (0.0) Functional Comorbidity Index, median (IQR), 0–16 0.0 (1.0) IQR: interquartile range; MCS: mental component summary; mHAQ: modified HAQ; PCS: physical component summary; VAS: visual analogue scale. All three comorbidity indices improved the model fit compared with the model unadjusted for comorbidities, and overall, the differences in their performance in prediction of the selected outcomes were small to moderate. The FCI and RDCI consistently performed better than CDI on the clinical outcomes mHAQ (QIC improvement compared with model not adjusted for comorbidities −60 and −46 vs −24, respectively), HAQ (QIC improvement −84 and −72 vs −42), physical component summary (QIC improvement −48 744 and −49 394 vs −27 237) mental component summary (QIC improvement −28 586 and −22 682 vs −18 377) and SF6D (−4 and −3 vs −1). Results on HRU were mixed, and any HRU was best predicted by RDCI followed by CDI (Table 3 and Fig. 1; supplementary Fig. S1, available at Rheumatology online). Further details of the GEE models are presented in supplementary Table S2, available at Rheumatology online. Table 3 Comparison of model fit statistics between models with the three comorbidity indices Model mHAQ HAQ PCS MCS SF6D Any HRU Outpatient visit Hospitalization Number of visits (n) 25 325 21 379 21 577 21 577 22 222 17 312 17 294 17 304 QIC Base modela 4133 5574 1 781 625 2 542 363 311 19 883 18 043 10 412 RDCI 4087 5502 1 732 231 2 519 681 308 19 566 17 806 10 249 CDI 4109 5532 1 754 388 2 523 986 310 19 568 17 797 10 272 FCI 4073 5490 1 732 881 2 513 777 307 19 575 17 824 10 235 Improvement in QIC (QIC adjusted – QIC base model) RDCI −46 −72 −49 394 −22 682 −3 −317 −237 −163 CDI −24 −42 −27 237 −18 377 −1 −315 −246 −140 FCI −60 −84 −48 744 −28 586 −4 −308 −219 −177 Model mHAQ HAQ PCS MCS SF6D Any HRU Outpatient visit Hospitalization Number of visits (n) 25 325 21 379 21 577 21 577 22 222 17 312 17 294 17 304 QIC Base modela 4133 5574 1 781 625 2 542 363 311 19 883 18 043 10 412 RDCI 4087 5502 1 732 231 2 519 681 308 19 566 17 806 10 249 CDI 4109 5532 1 754 388 2 523 986 310 19 568 17 797 10 272 FCI 4073 5490 1 732 881 2 513 777 307 19 575 17 824 10 235 Improvement in QIC (QIC adjusted – QIC base model) RDCI −46 −72 −49 394 −22 682 −3 −317 −237 −163 CDI −24 −42 −27 237 −18 377 −1 −315 −246 −140 FCI −60 −84 −48 744 −28 586 −4 −308 −219 −177 Results from linear and logistic generalized estimating equations (GEE). Note on interpretation: a lower value of QIC indicates a better fit and a higher explanatory value of the model. Model with the best fit, that is, lowest QIC as the largest QIC improvement (absolute value), in each group is highlighted in bold. aModel not adjusted for comorbidities. CDI: Charlson–Deyo Index; FCI: Functional Comorbidity Index; HRU: health resource utilization; mHAQ: modified HAQ; PCS/MCS: physical/mental component summary of SF-36; QIC: quasi-likelihood under the independence model criterion; RDCI: Rheumatic Disease Comorbidity Index; SF6D: short form six dimensions. Fig. 1 View largeDownload slide Comparison of improvement of model fit statistics between models with the three comorbidity indices The comparison is done with the model not adjusted for comorbidities. Note that model fit plotted here corresponds to the improvement in QIC (compared with base model not adjusted for comorbidities), a measure of model fit used for analysis over time (GEE). The higher the (absolute) improvement, the better the fit. CDI: Charlson–Deyo Index; FCI: Functional Comorbidity Index; HRU: health resource utilization; QIC: quasi-likelihood under the independence model criterion; RDCI: Rheumatic Disease Comorbidity Index; SF6D: short form six dimensions. Fig. 1 View largeDownload slide Comparison of improvement of model fit statistics between models with the three comorbidity indices The comparison is done with the model not adjusted for comorbidities. Note that model fit plotted here corresponds to the improvement in QIC (compared with base model not adjusted for comorbidities), a measure of model fit used for analysis over time (GEE). The higher the (absolute) improvement, the better the fit. CDI: Charlson–Deyo Index; FCI: Functional Comorbidity Index; HRU: health resource utilization; QIC: quasi-likelihood under the independence model criterion; RDCI: Rheumatic Disease Comorbidity Index; SF6D: short form six dimensions. Finally, conservative and liberal scenarios for health care utilization did not yield different results (results from conservative scenario are not shown). Adding erosive disease or disease duration to the models did not affect the contribution of the comorbidity indices to the improvement of the model fit. Discussion In this study, the MedDRA coding was linked to the comorbidities represented in the RDCI, FCI and CDI in order to provide an algorithm to compute these often used comorbidity indices. Further, the RDCI, FCI and CDI performed similarly in predicting physical function, the physical and mental health, utility and HRU. Some differences were observed in how the three indices predicted outcomes of interest with especially better association for the RDCI and FCI when physical function or physical and mental health were the outcome. When any HRU (hospitalization or outpatient visit) was the outcome, FCI had inferior performance. It should be borne in mind that the CDI was developed to predict mortality and although RMDs result in higher mortality, they still have a higher impact on morbidity and therefore on the outcomes studied in these analyses. MedDRA is a dynamic and regularly updated classification. To our knowledge, this study is the first one to offer an algorithm for computing several comorbidity indices for RMD field from this classification (the table of correspondence between all MedDRA codes and categories of each of the three comorbidity indices is presented in supplementary Table S1, available at Rheumatology online). Further, few studies have explored the comparative performance of comorbidity indices, and the available studies looked at function and mortality as outcomes [15, 17, 25, 26]. Our results are consistent with the existing literature and show that FCI is best to predict physical function [15, 17]. To our knowledge, this is the first study in RA that included non-RA related HRU as an outcome. We observed that FCI was somewhat superior in predicting hospitalization, while CDI followed by the RDCI had best prediction value for an outpatient visit. The current study does not reveal a strong preference for any of the outcome measures considered. While the choice of the comorbidity index is defined primarily by the purpose of the study (mental or physical functioning, mortality), type of data available (e.g. administrative records or self-report) and outcome of interest, previous studies have emphasized that the recently developed RDCI includes only a limited set of comorbidities (n = 11), which makes it simpler to compute from patient-reported data or to retrieve from chart review when MedDRA registration is not available for automatic computation . Our study has several limitations. First, we were not able to include mortality as an outcome; however, this outcome is one of the most studied and our results would be most likely only a validation of known evidence. Next, data on HRU was dichotomized, which did not allow studying costs of care received as an outcome, which would have been additionally informative. However, this is the first study that attempted to include this important outcome in a comparative study. Next, MedDRA classification codes were reviewed by two researchers with frequent discussions to reach a consensus, and therefore Cohen’s κ coefficient could not be calculated. In order to ensure correctness of the developed algorithm, after assigning each of the MedDRA terms to the conditions from the indices, the diseases included in each of the comorbidity indices linked to MedDRA codes were crosschecked for accuracy and consistency and were additionally searched in MedDRA terms. Only a few inconsistencies were detected and corrected with these additional checks, which supports the robustness of this method. Last but not least, validation of MedDRA-derived comorbidity indices by comparing with other ways of deriving comorbidities (e.g. from patient questionnaires) was not possible as relevant data were not available in NOR-DMARDs. This was thus a priori left beyond the scope of this paper, while it remains an important point for future research. In summary, this study provided an algorithm to compute three commonly used comorbidity indices from the MedDRA classification. We have found that the RDCI, FCI and CDI perform comparably well in predicting physical function, physical and mental health, health utility and HRU, with the CDI having inferior performance on outcomes reflecting functioning and health compared with FCI and RDCI. All indices are suitable to be used in patients with RA, with RDCI being the shortest to compute and still performing as well as the others. Funding: No specific funding was received from any bodies in the public, commercial or not-for-profit sectors to carry out the work described in this manuscript. Disclosure statement: The authors have declared no conflicts of interest. Supplementary data Supplementary data are available at Rheumatology online. References 1 Radner H, Smolen JS, Aletaha D. Impact of comorbidity on physical function in patients with rheumatoid arthritis. Ann Rheum Dis 2010; 69: 536– 41. http://dx.doi.org/10.1136/ard.2009.118430 Google Scholar CrossRef Search ADS PubMed 2 Radner H, Yoshida K, Smolen JS, Solomon DH. Multimorbidity and rheumatic conditions—enhancing the concept of comorbidity. Nat Rev Rheumatol 2014; 10: 252– 6. Google Scholar CrossRef Search ADS PubMed 3 Zavada J, Dixon WG, Askling J; EULAR Study group on Longitudinal Observational Registers and Drug Studies. Launch of a checklist for reporting longitudinal observational drug studies in rheumatology: a EULAR extension of STROBE guidelines based on experience from biologics registries. Ann Rheum Dis 2014; 73: 628. http://dx.doi.org/10.1136/annrheumdis-2013-204102 Google Scholar CrossRef Search ADS PubMed 4 Dixon WG, Carmona L, Finckh A et al. EULAR points to consider when establishing, analysing and reporting safety data of biologics registers in rheumatology. Ann Rheum Dis 2010; 69: 1596– 602. http://dx.doi.org/10.1136/ard.2009.125526 Google Scholar CrossRef Search ADS PubMed 5 Zink A, Askling J, Dixon WG et al. European biologicals registers: methodology, selected results and perspectives. Ann Rheum Dis 2009; 68: 1240– 6. http://dx.doi.org/10.1136/ard.2008.091926 Google Scholar CrossRef Search ADS PubMed 6 Canhao H, Faustino A, Martins F, Fonseca JE, Rheumatic Diseases Portuguese Register Board Coordination, Portuguese Society of Rheumatology. Reuma.pt – the rheumatic diseases portuguese register. Acta Reumatol Port 2011; 36: 45– 56. Google Scholar PubMed 7 Hetland ML, Unkerskov J, Ravn T et al. Routine database registration of biological therapy increases the reporting of adverse events twentyfold in clinical practice. First results from the Danish Database (DANBIO). Scand J Rheumatol 2005; 34: 40– 4. Google Scholar CrossRef Search ADS PubMed 8 Carmona L, de la Vega M, Ranza R, Casado G et al. BIOBADASER, BIOBADAMERICA, and BIOBADADERM: safety registers sharing commonalities across diseases and countries. Clin Exp Rheumatol 2014; 32(5 Suppl 85): S-163– 7. 9 Zink A, Listing J, Kary S et al. Treatment continuation in patients receiving biological agents or conventional DMARD therapy. Ann Rheum Dis 2005; 64: 1274– 9. http://dx.doi.org/10.1136/ard.2004.031476 Google Scholar CrossRef Search ADS PubMed 10 Silman A, Symmons D, Scott DG, Griffiths I. British Society for Rheumatology Biologics Register. Ann Rheum Dis 2003; 62(Suppl 2): ii28– 9. Google Scholar PubMed 11 MedDRA. Medical Dictionary for Regulatory Activities. 2016. http://www.meddra.org/ (April 2015, date last accessed). 12 Guelfucci F, Bessou A, Aballea S. Review of comorbidity measures to predict economic outcomes in real-life database studies. Value Health 2015; 18: A692– 3. Google Scholar CrossRef Search ADS PubMed 13 Yurkovich M, Avina-Zubieta JA, Thomas J, Gorenchtein M, Lacaille D. A systematic review identifies valid comorbidity indices derived from administrative health data. J Clin Epidemiol 2015; 68: 3– 14. Google Scholar CrossRef Search ADS PubMed 14 Charlson ME, Pompei P, Ales KL, MacKenzie CR. A new method of classifying prognostic comorbidity in longitudinal studies: development and validation. J Chronic Dis 1987; 40: 373– 83. http://dx.doi.org/10.1016/0021-9681(87)90171-8 Google Scholar CrossRef Search ADS PubMed 15 Groll DL, To T, Bombardier C, Wright JG. The development of a comorbidity index with physical function as the outcome. J Clin Epidemiol 2005; 58: 595– 602. http://dx.doi.org/10.1016/j.jclinepi.2004.10.018 Google Scholar CrossRef Search ADS PubMed 16 Wolfe F, Michaud K, Li T, Katz RS. Chronic conditions and health problems in rheumatic diseases: comparisons with rheumatoid arthritis, noninflammatory rheumatic disorders, systemic lupus erythematosus, and fibromyalgia. J Rheumatol 2010; 37: 305– 15. http://dx.doi.org/10.3899/jrheum.090781 Google Scholar CrossRef Search ADS PubMed 17 England BR, Sayles H, Mikuls TR, Johnson DS, Michaud K. Validation of the rheumatic disease comorbidity index. Arthritis Care Res 2015; 67: 865– 72. Google Scholar CrossRef Search ADS 18 Kvien TK, Heiberg MS, Lie E et al. A Norwegian DMARD register: prescriptions of DMARDs and biological agents to patients with inflammatory rheumatic diseases. Clin Exp Rheumatol 2005; 23(5 Suppl 39): S188– 94. 19 Kvamme MK, Lie E, Kvien TK, Kristiansen IS. Two-year direct and indirect costs for patients with inflammatory rheumatic joint diseases: data from real-life follow-up of patients in the NOR-DMARD registry. Rheumatology 2012; 51: 1618– 27. http://dx.doi.org/10.1093/rheumatology/kes074 Google Scholar CrossRef Search ADS PubMed 20 Pincus T, Summey JA, Soraci SAJr, Wallston KA, Hummon NP. Assessment of patient satisfaction in activities of daily living using a modified Stanford Health Assessment Questionnaire. Arthritis Rheum 1983; 26: 1346– 53. Google Scholar CrossRef Search ADS PubMed 21 Anderson J, Sayles H, Curtis JR, Wolfe F, Michaud K. Converting modified health assessment questionnaire (HAQ), multidimensional HAQ, and HAQII scores into original HAQ scores using models developed with a large cohort of rheumatoid arthritis patients. Arthritis Care Res 2010; 62: 1481– 8. Google Scholar CrossRef Search ADS 22 Ware JEJr, Sherbourne CD. The MOS 36-item short-form health survey (SF-36). I. Conceptual framework and item selection. Med Care 1992; 30: 473– 83. Google Scholar CrossRef Search ADS PubMed 23 Brazier J, Roberts J, Deverill M. The estimation of a preference-based measure of health from the SF-36. J Health Econ 2002; 21: 271– 92. http://dx.doi.org/10.1016/S0167-6296(01)00130-8 Google Scholar CrossRef Search ADS PubMed 24 Pan W. Akaike's information criterion in generalized estimating equations. Biometrics 2001; 57: 120– 5. http://dx.doi.org/10.1111/j.0006-341X.2001.00120.x Google Scholar CrossRef Search ADS PubMed 25 Southern DA, Quan H, Ghali WA. Comparison of the Elixhauser and Charlson/Deyo methods of comorbidity measurement in administrative data. Med Care 2004; 42: 355– 60. http://dx.doi.org/10.1097/01.mlr.0000118861.56848.ee Google Scholar CrossRef Search ADS PubMed 26 Lieffers JR, Baracos VE, Winget M, Fassbender K. A comparison of Charlson and Elixhauser comorbidity measures to predict colorectal cancer survival using administrative health data. Cancer 2011; 117: 1957– 65. Google Scholar CrossRef Search ADS PubMed © The Author 2017. Published by Oxford University Press on behalf of the British Society for Rheumatology. All rights reserved. For Permissions, please email: email@example.com
Rheumatology – Oxford University Press
Published: Mar 1, 2018
It’s your single place to instantly
discover and read the research
that matters to you.
Enjoy affordable access to
over 18 million articles from more than
15,000 peer-reviewed journals.
All for just $49/month
Query the DeepDyve database, plus search all of PubMed and Google Scholar seamlessly
Save any article or search result from DeepDyve, PubMed, and Google Scholar... all in one place.
Get unlimited, online access to over 18 million full-text articles from more than 15,000 scientific journals.
Read from thousands of the leading scholarly journals from SpringerNature, Elsevier, Wiley-Blackwell, Oxford University Press and more.
All the latest content is available, no embargo periods.
“Hi guys, I cannot tell you how much I love this resource. Incredible. I really believe you've hit the nail on the head with this site in regards to solving the research-purchase issue.”Daniel C.
“Whoa! It’s like Spotify but for academic articles.”@Phil_Robichaud
“I must say, @deepdyve is a fabulous solution to the independent researcher's problem of #access to #information.”@deepthiw
“My last article couldn't be possible without the platform @deepdyve that makes journal papers cheaper.”@JoseServera