Comparison of indexes to measure comorbidity burden and predict all-cause mortality in rheumatoid arthritis

Objectives To examine the comorbidity burden in patients with rheumatoid arthritis (RA) patients using a nationwide population-based cohort by assessing the Charlson Comorbidity Index (CCI), Elixhauser Comorbidity Index (ECI), Multimorbidity Index (MMI), and Rheumatic Disease Comorbidity Index (RDCI) scores and to investigate their predictive ability for all-cause mortality. Methods We identified 24,767 RA patients diagnosed between 1998–2008 in Taiwan and followed up until December 31, 2013. The incidence of comorbidities was estimated in three periods (before, during, and after the diagnostic period). The incidence rate ratios were calculated by comparing during vs. before and after vs. before the diagnostic period. One- and 5-year mortality rates were calculated and discriminated by low and high-score groups and modified models for each index. Results The mean score at diagnosis is 0.8 in CCI, 2.8 in ECI, 0.7 in MMI, and 1.3 in RDCI, and annual percentage changes are 11.0%, 11.3%, 9.7%, and 6.8%, respectively. The incidence of any increase in the comorbidity index is significantly higher in the periods of ‘during’ and ‘after’ the RA diagnosis (incidence rate ratios for different indexes: 1.33-2.77). The mortality rate significantly differs between the high and low-score groups measured by each index (adjusted hazard ratios: 2.5-4.3 for different indexes). CCI is slightly better in the prediction of 1- and 5-year mortality rates. Conclusion Comorbidities are common before and after RA diagnosis, and the rate of accumulation accelerates after RA diagnosis. All four comorbidity indexes are useful to measure the temporal changes and to predict mortality.


INTRODUCTION
Comorbidities are conditions that coexist with a specific disease. The importance of determining comorbidity in the routine care of rheumatoid arthritis has been recognized because it may aid in predicting disability, medical cost, survival time, and mortality. (1)(2)(3)(4) Systematic quantification of the comorbidity burden is essential for clinical management of index diseases. Currently, there are many comorbidity indexes developed for various purposes. The simple method is to use the sum of each comorbidity score, whereas the more complicated way is to aggregate and weight specific comorbidities to measure the burden and impact them on the index disease management. (5)(6)(7)(8) The selection of different comorbidity indexes is according to the purpose of the study, type of data available (administrative data or self-report data), and outcome of interest. These tools were decided for research studies, they are not embedded widely in electronic health record systems, and there is not consensus or guidelines about how these should be used in clinical practice at this point.
The Charlson comorbidity index (CCI) and the Elixhauser Comorbidity Index (ECI) are general-purpose comorbidity indexes while the Multimorbidity Index (MMI) and the Rheumatic Disease Comorbidity Index (RDCI) are designed specifically for patients with rheumatic diseases. They are similar in core comorbidity categories but are validated against different patient outcomes. For example, CCI was validated against mortality, hospital stay, functional disability, and healthcare utilization (9)(10)(11)(12). ECI uses 30 comorbidities (17 comorbidities from CCI) to predict hospital stay, cost, and mortality. A further revised weighted version was developed, which predicted mortality with better predictive ability than the original ECI (13)(14)(15) and CCI. (16,17) MMI was explicitly developed for RA patients based on the health-related quality of life (HRQol) (18). The RDCI was developed with 11 comorbid conditions for patients with RA, noninflammatory rheumatic disorders, systemic lupus erythematosus, and fibromyalgia to evaluate their quality of life using self-reported data (19) The RDCI can also be used to predict physical disability and mortality. (20) They are useful for clinicians to measure comorbidity burden and provide various predictive values to help the decision making for patient management.
There are few studies to compare comorbidity indexes in patients with RA. (Supplementary table 3) This study aims to understand the temporal changes of comorbidity indexes with the diagnosis of RA and their comparability in mortality prediction. We conducted the study using the National Health Insurance (NHI) Database, which covers the entire population since 1998; therefore is suitable to delineate the temporal changes of comorbidity indexes and their predictive power for mortality.

Study design
This retrospective study compares the trend, measurement time, and discriminant capacity of different comorbidity indexes in RA patients. This study was approved by the Institutional Review Board at the Chang Gung Memorial Hospital, Taiwan, and was conducted in accordance with the Declaration of Helsinki.
The requirement for signed informed consent was waived since the data presented in this paper were based on the NHI database, which is fully encrypted to prevent confidentiality leak. The data can only be accessed in the dedicated computer center in the NHI administration, the data holder. Therefore, the study team does not hold the original data.

Data sources and patient cohort
The primary data source came from the NHI Database, which contained registration information and original claims data of all beneficiaries of NHI in Taiwan since its establishment in 1995. The NHI is a single-payer mandatory enrolment system. Therefore, the coverage rate is virtually universal (approximately 99%). All entries for an individual are linked by a unique personal identifier assigned to each Taiwanese resident, allowing accurate linkage of records from the registration files and the original claims data. Before release for research, personal identifiers are de-identified to ensure confidentiality. The registry of beneficiaries, one of the registration files, contains details of demographics, residence, kinship relationships, occupation categories, insurance status, and insurance amount of all beneficiaries of NHI. Claims data on all outpatient visits, inpatient care, and pharmacy dispensing were recorded in specific datasets with information such as dates of events, medical diagnoses, medical expenditure, and details of prescriptions, operations, examinations, and procedures. The NHI database is linked to the National Death Registry as well, so to ascertain survival status. The availability of comprehensive medical records from national medical institutions and the focus on chronic comorbidities minimizes the risk of missing comorbidities of interest, but the accuracy of coded diagnoses can be suboptimal. Standard procedures used in administrative claims studies, such as requiring 2 codes at least 30 days apart, were used to improve reliability of the coded diagnoses.
Patients with RA were identified from the catastrophic illness registry, which records patients with specified disorders eligible for copayment waivers. A panel review of the applicant's clinical data is mandatory for the issuance of such benefits; therefore, the case definition is stringent. Patients have to fulfill the American Rheumatism Association 1987 revised classification criteria (21) to be included in the registry and receive waivers. We identified 24,767 RA patients diagnosed between 1998 to 2008 and followed them until December 31, 2013, or patient death. Specific organ involvement is considered the major comorbidities in all indexes, such as lung disease, stroke, cancer, heart failure, diabetes, and coronary heart disease. It is worth noting that the fracture is included in the RDCI, but renal and liver diseases are excluded. Since CCI and ECI both have a category of connective tissue disease, including RA, the score in this category was not counted in the subsequent analysis.

Statistical analysis
We calculated the four comorbidity indexes at diagnosis and annual percentage change of the comorbidity index in RA patients. Using conditional Poisson regression, we subsequently estimated the IRRs and their 95% confidence intervals by comparing the IRs during the diagnostic and post-diagnostic periods with the IRs during the pre-diagnostic period in the same patients. (24) We identified patients with a comorbidity score in the quintile of highest scores of each comorbidity index assessed during RA diagnostic period(the period of 4 months before and after the initial diagnosis). This group of patients was termed 'high-score group' and the cut-off value was 2 for CCI, 3 for ECI, 1 for MMI, and 2 for RDCI. The high-score group included patients in the top 20% who obtained the highest scores in the four comorbidity indexes. The 1year and 5-year mortality rates were calculated for the low and high-score groups for each index. The Cox proportional hazards model was used to calculate hazard ratios (HR) and 95% confidence interval (CI) for death associated with a high comorbidity index. The predictive capability for the 1-year and 5-year mortality rates by the four comorbidity indexes are discriminated by using the Harrell's c-statistics and Akaike information criteria (AIC). We also compared the performance of the four comorbidity indexes to predict 1and 5-year mortality by the Harrel's c-statistic (which is equivalent to area under the receiver-operator curve in the logistic regression). The impact of the comorbidity index was compared by using the weighted Kaplan-Meier survival estimates. The baseline comorbidity indexes are calculated from all comorbidities occurring during RA diagnostic period(the period of 4 months before and after the initial diagnosis). All statistical analyses were carried out by SAS 9.4.

Baseline characteristics of RA patients
Among the 24,767 patients with RA, the mean age±standard deviation at diagnosis was 50.2±15.7 years, and the male was 20.8%. The median (interquartile range) follow-up duration was 8.6 (6.5-10.8) years. The patient characteristics and comorbidity compositions, according to different definitions of four comorbidity indexes, are shown in table 1. The distribution of prevalent comorbidity items is different among different comorbidity indexes.

Four comorbidity index increased after RA was diagnosed
The mean score at diagnosis was 0.8 in CCI, 2.8 in ECI, 0.7 in MMI, and 1.3 in RDCI. The annual percentage changes of CCI, ECI, MMI, and RDCI after the RA diagnosis were 11.0%, 11.3%, 9.7%, and 6.8%, respectively ( Figure 1). The incidence rates and incidence rate ratios for the accumulation of any disease in the comorbidity index before, during, and after the diagnostic period in patients with newly diagnosed RA were demonstrated in   The one-year mortality risk in the high-score group was two to four times higher than that in the low-score group. (adjusted HR=2. 5-4.3) The five-year mortality risk in the high-score group was two times higher than that in the low-score group (adjusted HR=2.5-4.3)

The predictive ability for mortality among the four comorbidity indexes
The comorbidity demonstrated a significant impact on disease-specific survival by CCI, ECI, MMI, and RDCI in the high-score group than in the low-score group, as shown in the weighted Kaplan-Meier survival curve (Supplementary Figure 1). All of the log-rank test's p-values were <0.001. The discrimination for 1and 5-year survival in the four comorbidity indexes were all acceptable (Table 4). CCI showed a slightly better predictive ability than the other three comorbidity indexes. The Harrell's c-statistic and AIC of the 1- year mortality and 5-year mortality is highest in CCI. Supplementary Figure 2 shows the comparison of the ROC curve of the four comorbidity indexes in the prediction of 1-year and 5-year mortality. The base model included age, sex, income quartile, urbanization, and occupation groups. The Harrell's cstatistics indicates the prediction models, which are as follows: 0.5 (as well as chance), 0.7-0.8 (acceptable), 0.8-0.9 (excellent), and 0.9-1 (outstanding prediction). The AIC statistics was calculated, and a small AIC indicates the better predictive ability of the model.

DISCUSSION
This study demonstrates that all four commonly used comorbidity indexes are useful to track patient outcomes in terms of overall disease burden and mortality, in addition to their original validated outcomes.
The accumulation of comorbidity burden was higher after RA diagnosis. A higher comorbidity burden in RA patients at diagnosis is associated with increased mortality for RA patients. The short-and long-term mortality prediction performed comparably well among the four indexes, and CCI seems slightly better. This study documents the importance of comorbidity assessment, and all four commonly used indexes are all good for research outcome assessment in RA patients.

The factors that cause comorbidity indexes of RA increase with time
The comorbidities accumulated faster during and after the diagnostic periods in RA patients. The possible explanation for this finding is that a broad range of physical examination, laboratory studies, and examinations are performed before prescribing disease-modifying anti-rheumatic drugs (DMARDs) to avoid any side effects; therefore, many diseases or comorbidities are usually found during the diagnostic period.
The reasons why comorbidities accumulate with time may be as followed: First, the inflammatory process and the side effect of a drug can cause disease-associated organ damage. Second, the increase in the risk of infection and malignancy is considered due to the disease process, disordered immune surveillance, and administration of immunosuppressant drugs. (28,29) Third, the increased risk for cardiovascular diseases and osteoporosis is due to the inflammatory process and chronic glucocorticoid use. (30)(31)(32)(33) These indexes can only increase over time as patients don't lose chronic conditions and influences the incresaed comorbidity indexes after rheumatoid arthritis was daignosed. Comorbidity increased with time, and therefore a comprehensive assessment for comorbidity in RA patients since the initial diagnosis of RA is very important for patient management.

A high comorbidity index predicts a high mortality rate
Patients with RA who have a higher comorbidity index score early in their disease course have a significantly higher 1-and 5-years mortality risk.  (39) We should screen and monitor the patient's blood pressure because poorly controlled hypertension will affect the mortality of RA patients.

Comparison of the four comorbidity indexes in RA relevant to mortality
There is still no consensus on the most optimal comorbidity index for RA to predict mortality. Mortality is one of the validated outcomes in the first study of CCI, ECI, and RDCI (40). MMI is later proved that it can be applied in mortality in community-dwelling participants in the United States. In summary, in a large, Taiwan community-based cohort, we found that people with RA had a high burden of comorbidities and accumulated with time. This was found in global high comorbidity indexes scores and specifically in the high incidence of hypertension, ulcer, and diabetes mellitus. In this population-based cohort study, the four comorbidity indexes are good tools for predicting all-cause mortality among RA patients. Patients with higher comorbidity indexes score thus had a higher one-year and five-year mortality rate. Screening and treating comorbidities should be valued and practiced in rheumatologist's daily work.
Furthermore, developing adequate strategies to decrease the burden of these comorbidities among RA is suggested. Close surveillance of risk factors and aggressive management of potentially reversible risk factors are necessary to increase the survival of affected patients.