Abstract
Background
Rheumatoid arthritis (RA) is associated with increased risk of sepsis and higher infection-related mortality compared to the general population. However, the evidence on the prognostic impact of RA in sepsis has been inconclusive. We aimed to estimate the population-level association of RA with short-term mortality in sepsis.
Methods
We used statewide data to identify hospitalizations aged ≥18 years in Texas with sepsis, with and without RA during 2014–2017. Hierarchical logistic models with propensity adjustment (primary model), propensity score matching, and multivariable logistic regression without propensity adjustment were used to estimate the association of RA with short-term mortality among sepsis hospitalizations.
Results
Among 283,025 sepsis hospitalizations, 7,689 (2.7%) had RA. Compared to sepsis hospitalizations without RA, those with RA were older (aged ≥65 years, 63.9% vs. 56.4%) and had higher burden of comorbidities (mean Deyo comorbidity index, 3.2 vs. 2.7). Short-term mortality of sepsis hospitalizations with and without RA was 26.8% vs. 31.4%. Following adjustment for confounders, short-term mortality was lower among RA patients (adjusted odds ratio [aOR], 0.910; 95% confidence interval [CI], 0.856–0.967), with similar findings on alternative models. On sensitivity analyses, short-term mortality was lower in RA patients among sepsis hospitalizations aged ≥65 years and those with septic shock, but not among those admitted to intensive care unit (ICU; aOR, 0.990; 95% CI, 0.909–1.079).
Conclusions
RA was associated, unexpectedly, with lower short-term mortality in septic patients. However, this “protective” association was driven by those patients without ICU admission. Further studies are warranted to confirm these findings and to examine the underlying mechanisms.
Keywords: intensive care unit, mortality, rheumatoid arthritis, sepsis, septic shock
INTRODUCTION
Rheumatoid arthritis (RA) is the most common systemic autoimmune disease [1], estimated to affect over 1.3 million adults in the United States (US) [2]. Although major strides were made in the care of affected patients, following the introduction of disease-modifying antirheumatic drugs, life expectancy remains shortened among patients with RA, with over four-times higher standardized mortality rate compared to the general population [3]. Patients with RA are at an increased risk of infectious complications, often progressing to sepsis, stemming from immune dysfunction due to the disease itself, comorbid conditions, and common use of immunomodulating therapy [4], and sepsis is increasingly a major cause of hospitalization in this population [5].
The diagnosis of sepsis can be challenging in patients with RA since pre-existing immune dysfunction may affect its initial manifestations, and these patients may be at a higher risk of infections with drug- resistant pathogens due to repeated exposure to antibiotics related to recurrent infections [6]. These latter factors can result in delays in the diagnosis of sepsis and initiation of time-sensitive interventions and raised risk of inappropriate initial antimicrobial therapy, thus increasing risk of lethal outcomes [7,8]. Indeed, infection is the most common cause of death in patients with RA, accounting for 34% of all deaths in this population [3], with sepsis considered the major cause of death from infectious diseases [9]. Moreover, patients with RA have over six-times greater risk of infection-related mortality than the general population [10].
The higher risk of infection-related mortality in patients with RA compared to the general population may reflect their increased risk of sepsis, higher death rate in septic patients, or a combination of both. Distinguishing between these potential drivers of increased infection-related mortality in patients with RA is important, as it can guide future interventions. However, while greater risk of sepsis among patients with RA is well-documented, ranging for three- [5] to four-times [11] higher than in the general population, the evidence base on the outcomes of patients with RA in sepsis has been scarce, with inconclusive findings.
In a recent population-based study in the US, hospital mortality was slightly lower among septic patients with RA compared to the general population, but the difference was not statistically significant on adjusted analyses [12]. In a registry-based study from Israel, 30-day mortality was higher among septic patients with RA, but did not differ following adjustment for confounders [13]. In a study from Germany, hospital mortality was 3-fold higher in septic patients with RA, and RA was the most important predictor of death [14]. Finally, in a US-based study of the prognostic impact of autoimmune diseases in sepsis, 30-day mortality was markedly lower among these patients as a group, but although the point estimate showed lower mortality in the subgroup with RA, this latter finding was not statistically significant [15]. The interpretation and external validity of these studies is limited by single-center design [14,15], small cohort size [13-15], restriction to intensive care unit (ICU) patients [13-15], limited adjustment for confounders [12], and lack of data on patient characteristics [15]. Improved understanding of outcomes of RA in sepsis may guide clinicians’ decision-making and future efforts to improve sepsis outcomes. In this study, we examine the association of RA with short-term mortality among hospitalizations with sepsis. We postulated that RA is associated with worse outcomes in sepsis.
MATERIALS AND METHODS
This was a retrospective, population-based cohort study. The study was determined by the Institutional Review Board of Texas Tech University Health Sciences Center to be exempt from formal review because we used a publicly available, de-identified data set, which is not considered human research and where informed consent is not applicable per the US Department of Health and Human Services regulation 45 CFR 46.101(c). The reporting of the study findings followed the STROBE guidelines on reporting observational studies in epidemiology [16].
Data Sources and Study Population
We used the Texas Inpatient Public Use Data File (TIPUDF) to identify the target population. In brief, the TIPUDF is an administrative data set maintained by the Texas Department of State Health Services [17] and includes inpatient discharge data from state-licensed, non-federal hospitals, and captures approximately 97% of all hospital discharges in the state. We linked Social Deprivation Index (SDI) data [18] (see details on the SDI under risk-adjustment covariates below) at the Zip Code Tabulation Area level to the TIPUDF data to derive the SDI for individual hospitalizations.
Our primary cohort consisted of hospitalizations aged ≥18 years with a diagnosis of sepsis during the years 2014–2017. We excluded hospitalizations with missing data on hospital disposition. We identified hospitalizations with sepsis based on the presence of the International Classification of Diseases, 9th and 10th revisions, Clinical modification (ICD-9-CM and ICD-10-CM, respectively) codes for severe sepsis (995.92, R65.20) and septic shock (785.52, R65.21) under the principal or secondary diagnosis fields. This ICD code-based definition of sepsis is aligned with the framework of Sepsis-3 [19] and has been used in contemporary studies of sepsis in administrative data [20-22]. Hospitalizations with ICU admissions were identified based on unit-specific revenue codes for an ICU or a coronary care unit.
Exposure and Outcome
The primary exposure was a diagnosis of RA. Hospitalizations with RA were identified based on the presence of ICD-9-CM and ICD-10-CM codes 714.0, 714.1, 714.2, 714.8x, or 714.9, and M05x or M06x, respectively, in the principal or secondary diagnosis fields [23,24]. The primary outcome was short-term mortality, defined as in-hospital death or discharge to hospice. We have included discharge to hospice since this is an increasingly common end-of-life destination in sepsis. Therefore, focusing only on in-hospital mortality can produce misleading estimates and for this reason this composite outcome is increasingly used in epidemiological studies of sepsis [20].
Risk-Adjustment Covariates
Risk-adjustment covariates were selected a priori based on biological and clinical plausibility and existing literature [25-27] and included patients’ demographics (age, sex, race/ethnicity, primary health insurance), SDI, major comorbidities, including chronic lung disease, congestive heart failure, chronic renal disease, diabetes mellitus, malignancy, and chronic liver disease, as well as the Deyo Comorbidity Index (all based on the Deyo modification of the Charlson Comorbidity Index [28,29]), sites of infection, obesity, tobacco use, hospitals’ teaching status, and year of admission. The SDI represents a composite measure of poverty, level of education, single parenthood, rental living, overcrowding, transportation, level of education at a given area, and high-need demographics [30]. The SDI is expressed on a 1 (least deprived)-100 (most deprived) point scale [18,30]. Severity of illness was characterized using ICD codes for organ dysfunction [31]. Procedure use was identified using ICD-9 and ICD-10 procedure codes for mechanical ventilation, hemodialysis, and blood transfusion. The ICD-9-CM and ICD-10-CM used to identify sites of infection, obesity, tobacco use, organ dysfunction and procedures are detailed in Supplementary Tables 1-4.
Statistical Analysis
We summarized categorical variables as frequencies and percentages, while continuous variables were reported as mean and standard deviation. The chi-square test was used for group comparison involving categorical variables, while the t-test was used for comparison of continuous variables. We used three distinct prespecified analytical approaches to examine the association between a diagnosis of RA and short-term mortality: (1) multilevel multivariable logistic regression with propensity adjustment (our primary analysis), (2) propensity score matching, and (3) multivariable logistic regression without propensity adjustment. Multicollinearity was excluded using variance inflation factors. We report the models’ findings as adjusted odds ratios (aOR) and 95% confidence intervals (95% CI). The choice of these models represents the core analytical approaches in observational studies on the association between studied exposures and specific outcomes: (1) covariate adjustment (e.g., regression-based analyses) and (2) analyses based on patients’ likelihood to have the examined exposure (e.g., propensity score-based analyses). Although we have used a very large dataset, we have employed several different types of models to estimate the association of RA with short-term mortality among patients with sepsis, in order to examine whether the findings of the primary analysis are robust on alternative modeling approaches.
Propensity score calculation
A propensity score is a probability-based measure indicating the propensity of a septic patient to have RA. We estimated the propensity for RA among sepsis hospitalizations using multivariable logistic regression with RA used as the dependent variable. The variables included in the model, based on prior reports [32,33] were age, sex, race/ethnicity, SDI (as deciles), obesity, and tobacco use.
Multilevel multivariable logistic regression with propensity adjustment
We used multilevel multivariable logistic regression to assess the association between RA and short-term mortality in sepsis hospitalizations. The covariates entered into the multivariable model included all those described for risk-adjustment, as well as RA, and with individual hospitals entered as random intercepts to account for clustering of hospitalizations within hospitals. We excluded rheumatological conditions from the Deyo comorbidity index to prevent duplicate analysis of RA. The propensity score was added as an independent variable. Combining regression analysis and propensity scores into a single model, known as double robust estimation [34], allows all study patients to be retained in the analysis, as opposed to traditional 1:1 propensity score matching, which excludes unmatched patients and thus can reduce model’s generalizability. In addition, this method provides more accurate estimation of variance and reduces bias [35].
Propensity score matching
Hospitalizations with RA were matched to those without RA, but who had similar propensity for it. Matching was performed with 1:1 nearest neighbor matching without replacement, using maximal caliper width equal to 20% of the standard deviation of the logit of the propensity score [36]. We examined the balance of each covariate between the matched cohorts with and without RA using standardized differences, without consideration of any outcome variable. A standardized difference <0.1 was considered to represent well-balanced groups on a given covariate [37]. Pre-match versus post-matched data were represented graphically as density plots for propensity scores and as changes of standardized differences (Supplementary Figure 1). Multivariable logistic regression was used following propensity score matching to generate estimates of the association of RA with short-term mortality.
Multivariable logistic regression without propensity adjustment
We used multivariable logistic regression to assess the association between RA and short-term mortality, using the same covariates as described for our primary analysis, including the adjustment of the Deyo comorbidity index.
Sensitivity analyses
We probed the robustness of the observed association between RA and short-term mortality with six additional analyses. The first three analyses were restricted to sepsis hospitalizations aged ≥65 years and those more severely ill, including those admitted to ICU and those with septic shock. The three other analyses addressed the impact of non-random missingness of gender data in the cohort (see below). The primary analysis approach was used for sensitivity analyses. The State of Texas masks gender data of hospitalizations with a diagnosis of HIV (human immunodeficiency virus) infection, and of those with ethanol or drug abuse. Gender data were missing nonrandomly in 8.8% of hospitalizations in our cohort, precluding imputation of missing values. We examined the sensitivity of the association between RA and short-term mortality to missing gender data using missing gender as indicator variable (see Supplementary data file for details). In this article, we present the results of our analyses using data restricted to hospitalizations with gender data.
Data management was performed using Microsoft Excel (Microsoft, Redmond, WA, USA) and statistical analyses were performed with R 4.0.5 (R Foundation for Statistical Computing, Vienna, Austria). The R code supporting these analyses is provided in the Supplementary Material 1. A two-sided P-value <0.05 was considered statistically significant.
RESULTS
Among 283,025 hospitalizations with sepsis, meeting inclusion criteria, 7,689 (2.7%) had RA. There was 90.7% matching rate by propensity score (corresponding to 6,977 pairs of sepsis hospitalizations with and without RA) (Supplementary Table 5).
Cohort Characteristics
The characteristics of sepsis hospitalizations are described in Table 1. Compared to sepsis hospitalizations without RA, those with RA were older (aged ≥65 years 63.9% vs. 56.4%), less commonly racial/ethnic minority (42.9% vs. 48.1%), had higher burden of comorbidities (mean Deyo comorbidity index, 3.2 vs. 2.7), while having lower mean number of organ dysfunctions (2.5 vs. 2.7) and lower use of mechanical ventilation (30.3% vs. 35.1%). Sepsis hospitalizations with RA and without RA had similar rates of admission to ICU (46.5% vs. 46.0%) and septic shock (59.5% vs. 59.6%).
Table 1.
The characteristics of sepsis hospitalizations with and without rheumatoid arthritis
| Variable | Rheumatoid arthritis (n=7,689) | Non-rheumatoid arthritis (n=275,336) | P-value | |||
|---|---|---|---|---|---|---|
| Age (yr) | <0.001 | |||||
| 18–44 | 439 (5.7) | 30,418 (11.0) | ||||
| 45–64 | 2,338 (30.4) | 89,375 (32.4) | ||||
| ≥65 | 4,912 (63.9) | 155,543 (56.4) | ||||
| Sexa) | ||||||
| Female | 5,536 (75.3) | 124,592 (49.7) | <0.001 | |||
| Race/ethnicity | <0.001 | |||||
| White | 4,392 (57.1) | 142,827 (51.9) | ||||
| Hispanic | 1,965 (25.6) | 73,517 (26.7) | ||||
| Black | 837 (17.9) | 36,203 (13.1) | ||||
| Other | 494 (6.4) | 22,728 (8.3) | ||||
| Health insurance | <0.001 | |||||
| Private | 2,621 (34.1) | 89,843 (32.6) | ||||
| Medicare | 4,315 (56.1) | 136,859 (49.7) | ||||
| Medicaid | 385 (5.0) | 20,701 (7.5) | ||||
| Uninsured | 296 (3.8) | 24,055 (8.7) | ||||
| Other | 65 (0.8) | 3,539 (1.3) | ||||
| Social deprivation index | 59.0±27.5 | 61.0±27.7 | <0.001 | |||
| Deyo comorbidity index | 3.2±2.1 | 2.7±2.4 | <0.001 | |||
| Selected comorbidity | ||||||
| Chronic lung disease | 2,638 (34.3) | 72,935 (26.5) | <0.001 | |||
| Congestive heart failure | 2,445 (31.8) | 85,341 (31.0) | 0.135 | |||
| Renal disease | 2,205 (28.7) | 92,276 (33.5) | <0.001 | |||
| Diabetes | 2,709 (35.2) | 110,363 (40.1) | <0.001 | |||
| Malignancy | 683 (8.9) | 39,819 (10.9) | <0.001 | |||
| Liver disease | 677 (8.8) | 33,314 (12.1) | <0.001 | |||
| Obesity | 1,335 (17.4) | 41,433 (15.0) | <0.001 | |||
| Tobacco use | 1,534 (20.0) | 54,511 (19.8) | 0.664 | |||
| Site of infection | ||||||
| Respiratory | 3,036 (39.5) | 103,100 (37.4) | <0.001 | |||
| Urinary | 2,997 (39.0) | 101,446 (36.8) | <0.001 | |||
| Abdominal | 914 (9.0) | 34,034 (12.4) | <0.001 | |||
| Skin and soft tissue | 815 (10.6) | 27,263 (9.9) | 0.043 | |||
| Device-related | 169 (2.2) | 6,824 (2.5) | 0.096 | |||
| Otherb) | 368 (4.8) | 14,567 (5.3) | 0.053 | |||
| Septic Shock | 4,573 (59.5) | 164,225 (59.6) | 0.860 | |||
| Number of organ dysfunctions | 2.5±1.4 | 2.7±1.5 | <0.001 | |||
| Type of organ dysfunctions | ||||||
| Respiratory | 3,957 (51.5) | 149,553 (54.3) | <0.001 | |||
| Cardiovascular | 4,908 (63.8) | 176,836 (64.2) | 0.471 | |||
| Renal | 4,600 (59.8) | 168,368 (61.1) | 0.021 | |||
| Hepatic | 473 (6.2) | 24,488 (8.9) | <0.001 | |||
| Hematological | 1,374 (17.9) | 57,615 (20.9) | <0.001 | |||
| Neurological | 1,903 (24.7) | 76,518 (27.8) | <0.001 | |||
| Intensive care unit admission | 3,579 (46.5) | 126,579 (46.0) | 0.386 | |||
| Mechanical ventilation | 2,331 (30.3) | 96,535 (35.1) | <0.001 | |||
| Hemodialysis | 640 (8.3) | 35,235 (12.8) | <0.001 | |||
| Blood transfusion | 1,402 (18.2) | 54,746 (19.9) | <0.001 | |||
| Teaching hospital | 2,002 (26.0) | 76,948 (27.9) | <0.001 | |||
| Hospital disposition | ||||||
| In-hospital death | 1,546 (20.1) | 63,891 (23.2) | <0.001 | |||
| Hospice | 518 (6.7) | 22,520 (8.2) | <0.001 | |||
| Home | 2,986 (38.8) | 98,146 (35.6) | <0.001 | |||
| Another acute care hospital | 217 (2.8) | 8,678 (3.2) | 0.049 | |||
| Post-acute care facilityc) | 2,395 (31.1) | 80,018 (29.1) | <0.001 | |||
| Leave against medical advise | 27 (0.4) | 2,083 (0.8) | <0.001 |
Values are presented as number (%) or mean±standard deviation.
Data on sex was missing for 333 (4.3%) hospitalizations with rheumatoid arthritis and for 24,639 (8.9%) hospitalizations without rheumatoid arthritis; the percent figures for sex in each column refer to that column's denominator for sex;
Other sites of infection are genital, blood, endocarditis, central nervous system, and bone and joint;
Post-acute care facilities include long-term acute care hospitals, inpatient rehabilitation, skilled nursing facilities, and nursing homes.
The Association of RA with Short-term Mortality
Short-term mortality among sepsis hospitalizations with RA was lower than in those without RA (26.8% vs. 31.4%, respectively; P<0.001). Similarly, short-term mortality was lower in septic patients with RA than those without RA among those aged ≥65 years (31.5% vs. 36.4%; P<0.001), ICU admissions (30.8% vs. 34.9%; P<0.001), and those with septic shock (34.1% vs. 40.1%; P<0.001). Following adjustment for confounders, RA was associated with 9% lower risk of short-term mortality (aOR, 0.910; 95% CI, 0.856–0.967). This latter finding was consistent in the alternative models (Table 2).
Table 2.
Odds ratios for short-term mortality associated with rheumatoid arthritis among sepsis hospitalizations
| Model | aOR (95% CI) | P-value |
|---|---|---|
| Multilevel multivariable logistic regression with propensity adjustment | 0.910 (0.856–0.967) | 0.007 |
| Alternative analysis | ||
| Propensity score-matched sample | 0.808 (0.719–0.907) | <0.001 |
| Multivariable logistic regression without propensity adjustment | 0.901 (0.849–0.956) | <0.001 |
aOR: adjusted odds ratio; CI: confidence interval.
RA remained associated with lower short-term mortality on sensitivity analyses among sepsis hospitalizations aged ≥65 years (aOR, 0.874; 95% CI, 0.816–0.937) and those with septic shock (aOR, 0.897; 95% CI, 0.834–0.964). However, RA was not associated with short-term mortality among those admitted to ICU (aOR, 0.990; 95% CI, 0.909–1.079). Similar findings were obtained in the modeling approach for missing gender for each of the three models for the whole cohort, as well as for sepsis hospitalizations aged ≥65 years, ICU admissions, and those with septic shock (Supplementary Tables 6-11).
DISCUSSION
Key Findings
In this population-based study, we found that short-term mortality was markedly lower among septic patients with RA compared to those without RA. This favorable prognostic association of RA in sepsis remained robust following adjustment for confounders across three different modeling methods. The lower risk of short-term mortality in RA patients with sepsis remained among older patients and those with septic shock, but RA was not associated with mortality among septic patients admitted to ICU.
Relationship to Prior Studies
Our finding of a ”protective” association of RA in sepsis was unexpected. Septic patients with RA in our cohort were older and had higher comorbidity burden that those without RA. Moreover, as noted earlier, patients with RA may face higher risk of delays in recognition of sepsis and initiation of time-sensitive interventions, while a potentially increased risk of infection with multidrug resistant pathogens may further increase their risk of death, compared the general population. However, we found that short-term mortality is lower in RA patients with sepsis, which remained following adjustment for confounders for the whole cohort and, notably, among the more severely ill subgroups of older patients and those with septic shock. The findings of our study contrast those from a recent population-based report from the US by Colbert and colleagues [12]. In that study, although hospital morality was lower among septic patients with RA compared to the general population (24.5% vs. 26.5%, respectively), RA was not associated with mortality following adjustment for confounders (aOR, 0.91; 95% CI, 0.82–1.01]) [12]. However, lack of adjustment for severity of illness in that study [12] may have affected their findings. Our findings suggest that, when compared to the general population, the higher risk of infection-related mortality in patients with RA [10] reflects the increased risk of sepsis among the latter, rather than higher case fatality of septic patients.
On the other hand, although crude short-term mortality was lower among septic RA patients admitted to ICU compared to those without RA (30.8% vs. 34.9%), the lack of independent association of RA with short-term mortality on adjusted analyses is consistent with two recent ICU-based studies from the US [15] and Israel [13]. However, our findings on the subgroup of ICU admissions differed from the findings by Krasselt et al. [14] from Germany, who found that septic ICU admissions with RA (n=49) had nearly two-fold higher rate of septic shock than those without RA (65.3% vs. 29.4%), with nearly three times higher hospital mortality that those without RA (42.9% vs. 15.7%, respectively), and over 32 times higher odds of death among those with RA on adjusted analyses. The difference between the later study and our findings may reflect a single-center design, small number of patients with RA, difference in case mix, and ICU triage and care processes.
The factors contributing to the absence of independent association of RA with short-term mortality among ICU admissions with sepsis are unclear and our administrative dataset does not include information to directly examine the potential causes. Several possibilities for this observation may be considered. First, the timing of ICU admission and thus pre-ICU care may have differed between sepsis hospitalizations with and without RA, possibly leading to elimination of their differences of adjusted short-term outcomes. Second, a recent multicenter study by Anesi et al. [38], using instrumental variable-based analysis, showed that among septic patients presenting to the emergency department with high acuity but no need for organ support interventions, initial admission to ICU rather than to ward was associated with increased risk of hospital mortality on adjusted analyses. The investigators hypothesized that initial triage to ICU rather than to ward in these septic patients may have exposed the former to greater use of harmful intervention (e.g., excess intravenous fluid resuscitation), higher frequency of complications, or different end-of-life care processes [38]. While not directly reflecting our cohort, we speculate that these types of ICU-level exposures of septic patients with and without RA may have had greater adverse effect on the former and have eliminated their differences of adjusted short-term outcomes. Last, although we have performed an extensive adjustment for confounders among ICU admissions with and without RA, there may have been residual differences in patient characteristics between these groups that could not be captured in our dataset and may have contributed to the observed lack of association RA with short-term mortality among ICU admissions with sepsis. Additional investigations using more granular data are warranted to explore potential mechanisms for the observed prognostic associations of septic patients with RA receiving ICU care. In the interim, our findings show that care escalation to ICU among septic patients with RA does not indicate worse short-term outcome compared to those in the general population. This finding can inform clinicians’ decision-making and early discussions with patients and their surrogates about goals of care.
The mechanisms underlying the lower risk of short-term mortality associated with RA among septic patients are unclear. Dysregulated immune response to infection, including imbalance of pro- and anti-inflammatory cytokines [39] is considered a key driver of sepsis manifestations [40], and the aberrant cytokine signatures among patients with autoimmune diseases, including RA, include some of those involved in the pathophysiology of sepsis [41,42]. It has been hypothesized that baseline cytokine imbalance in patients with autoimmune disease can affect their response to infection and subsequent sepsis [15].
Several recent studies have shown that the risk of death in sepsis may be lower in patients with autoimmune diseases, but the findings were inconsistent. In a population-based study of inflammatory bowel disease, Crohn disease was associated with lower risk-adjusted in-hospital mortality among septic patients (aOR, 0.78; 95% CI, 0.63–0.97), while ulcerative colitis was associated with increased risk of death [12]. In the recent single-center study by Sheth et al. [15], noted earlier, autoimmune diseases were associated as a group with lower risk-adjusted 30-day mortality (aOR, 0.73; 95% CI, 0.57–0.93), though the favorable prognostic association was generally not statistically significant for individual diseases. However, when analyses in that study were stratified by levels of cytokine expression of individual autoimmune diseases, those with overexpression of interleukin (IL)-12 and interferon-gamma (IFN-γ) were associated, as a group, with lower risk of death, with similar trends among those diseases with underexpression of IL-4 and IL-10 [15]. Similar pattern of cytokine expression was reported in patients with RA, with both IL-12 [43] and IFN-γ [43,44] being overexpressed, while IL-4 and IL-10 are underexpressed [42-44]. The “protective” prognostic implications of immune dysregulation pre-sepsis are supported by reports of sepsis-related impairment of IL-12 [45] and IFN-γ [46] production, with preclinical studies showing that therapies that increase the expression of both can improve sepsis survival [47]. In addition, the immunosuppressive effect of sepsis is enhanced by increased expression of IL-4 and IL-10 [46], both of which were shown to reduce expression of pro-inflammatory cytokines. It has been postulated that chronic overexpression of select cytokines may improve sepsis survival of affected patients due to their better ability to overcome the immune dysfunction on sepsis [15]. If correct, this hypothesis may explain the findings of the present study.
However, notwithstanding the compelling findings reported by Sheth et al. [15] and the cytokine profiles reported in patients with RA, there have not been, to our knowledge, studies directly comparing the immune responses of septic patients with and without RA to date. Thus, additional investigations are warranted to confirm our observations and to compare specific sepsis-related domains of immune response in patients with and without RA, in order to guide future efforts to improve survival in sepsis. Recent reports characterizing sepsis endotypes through mechanistic signatures of gene expression of immune effectors to predict sepsis severity [48] may inform such studies.
Strengths and Limitations
Our study has relevant strengths and limitations. In terms of strengths, the present study evaluates a relatively little-examined and important research question. We used a high-quality, all- payer data set of consecutive hospitalizations, thus allowing transcending variation in practice patterns and case mix. We adhered closely to reporting guidelines and used statistical methods to limit confounding and enhance trustworthiness in measures of association.
However, our study has several important limitations. First, data were analyzed retrospectively and we used an administrative data set. Second, sepsis was identified by using ICD codes, similar to other epidemiological studies using administrative data [22], and not from clinical records, an inherent limitation of administrative data. Third, although the ICD codes for RA in the present report were used in prior epidemiological studies [23,24], misclassification may have occurred between groups. However, misclassifying RA hospitalizations would have likely reduced differences in outcomes between sepsis hospitalizations with and without RA. Such blurring of outcome differences would have resulted in possible underestimation of the magnitude of the better outcomes observed among the former. Fourth, the TIPUDF data does not provide information on the duration of RA, its activity level, or details on immunomodulating therapy. In addition, data on the processes of care (e.g., adherence to sepsis bundles, timeliness of therapeutic interventions, hospital admission, and admission to ICU) is not included in the TIPUDF data set and it is possible that these have been different in septic patients with and without RA. Thus, residual confounding in our models cannot be excluded. However, as outlined earlier, septic patients with RA may face greater delays in recognition of sepsis and use of time-sensitive interventions, as well as timely triage, compared to those without RA, which would have led to worse outcomes among RA patients with sepsis. Finally, it is unclear whether our findings can be generalized to other regions and countries.
In conclusion, RA was associated with lower short-term mortality among septic patients. This “protective” prognostic association was consistent across alternative modeling approaches, and among the subsets of older patients and those with septic shock, but not among septic patients admitted to ICU. Future studies are needed to determine the mechanisms underlying these observations, in order to inform efforts to improve sepsis outcomes.
KEY MESSAGES
▪ Patients with rheumatoid arthritis had lower risk of short-term mortality than those without rheumatoid arthritis among septic patients.
▪ This finding suggests that the increased risk of infection-related mortality in patients with rheumatoid arthritis compared to the general population is due to their increased risk of sepsis, rather than higher case fatality among septic patients.
Footnotes
CONFLICT OF INTEREST
No potential conflict of interest relevant to this article was reported.
FUNDING
None.
ACKNOWLEDGMENTS
None.
AUTHOR CONTRIBUTIONS
Conceptualization: LO. Data curation: all authors. Formal analysis: all authors. Methodology: all authors. Project administration: LO. Visualization: all authors. Writing–original draft: LO. Writing–review & editing: all authors.
SUPPLEMENTARY MATERIALS
Supplementary materials can be found via https://doi.org/10.4266/acc.2022.00787.
International Classification of Diseases, ninth and tenth revisions, clinical modification (ICD-9-CM and ICD-10-CM) codes used to identify sites of infection
International Classification of Diseases, ninth and tenth revisions, clinical modification (ICD-9-CM and ICD-10-CM) codes used to identify organ dysfunctions
International Classification of Diseases, Ninth and Tenth Revisions, Clinical Modification (ICD-9-CM and ICD-10-CM) codes used to identify procedures
International Classification of Diseases, Ninth and Tenth Revisions, Clinical Modification (ICD-9-CM and ICD-10-CM) codes used to identify selected comorbidities
The characteristics and outcome of propensity score-matched sepsis hospitalizations with and without rheumatoid arthritis
Multilevel multivariable logistic regression with propensity adjustment for the association of rheumatoid arthritis with short-term mortality for the whole cohort: alternative modeling for the impact of missing gender data
Propensity matched cohort for the association of rheumatoid arthritis with short-term mortality for the whole cohort: alternative modeling for the impact of missing gender data
Multivariable logistic regression without propensity adjustment for the association of rheumatoid arthritis with short-term mortality for the whole cohort: alternative modeling for the impact of missing gender data
Multilevel multivariable logistic regression with propensity adjustment for the association of rheumatoid arthritis with short-term mortality among ICU admissions: alternative modeling for the impact of missing gender data
Multilevel multivariable logistic regression with propensity adjustment for the association of rheumatoid arthritis with short-term mortality among hospitalizations with septic shock: alternative modeling for the impact of missing gender data
Multilevel multivariable logistic regression with propensity adjustment for the association of rheumatoid arthritis with short-term mortality among hospitalizations aged ≥65 years: alternative modeling for the impact of missing gender data
Propensity scores and standardized differences of covariates for comparisons of sepsis hospitalizations with rheumatoid arthritis and those without rheumatoid arthritis. (A) Density plots of the distribution of propensity scores for sepsis hospitalizations with (RA) and without rheumatoid arthritis (Non-RA) before and after matching. (B) Absolute standardized differences between sepsis hospitalizations with and without rheumatoid arthritis for the whole cohort (before matching) and matched data.
The R code used for the study analyses.
REFERENCES
- 1.Helmick CG, Felson DT, Lawrence RC, Gabriel S, Hirsch R, Kwoh CK, et al. Estimates of the prevalence of arthritis and other rheumatic conditions in the United States: part I. Arthritis Rheum. 2008;58:15–25. doi: 10.1002/art.23177. [DOI] [PubMed] [Google Scholar]
- 2.Hunter TM, Boytsov NN, Zhang X, Schroeder K, Michaud K, Araujo AB. Prevalence of rheumatoid arthritis in the United States adult population in healthcare claims databases, 2004-2014. Rheumatol Int. 2017;37:1551–7. doi: 10.1007/s00296-017-3726-1. [DOI] [PubMed] [Google Scholar]
- 3.Mok CC, Kwok CL, Ho LY, Chan PT, Yip SF. Life expectancy, standardized mortality ratios, and causes of death in six rheumatic diseases in Hong Kong, China. Arthritis Rheum. 2011;63:1182–9. doi: 10.1002/art.30277. [DOI] [PubMed] [Google Scholar]
- 4.Jani M, Barton A, Hyrich K. Prediction of infection risk in rheumatoid arthritis patients treated with biologics: are we any closer to risk stratification? Curr Opin Rheumatol. 2019;31:285–92. doi: 10.1097/BOR.0000000000000598. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Jinno S, Lu N, Jafarzadeh SR, Dubreuil M. Trends in hospitalizations for serious infections in patients with rheumatoid arthritis in the US between 1993 and 2013. Arthritis Care Res (Hoboken) 2018;70:652–8. doi: 10.1002/acr.23328. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Subesinghe S, Rutherford AI, Byng-Maddick R, Leanne Hyrich K, Benjamin Galloway J. Recurrent serious infections in patients with rheumatoid arthritis: results from the British Society for Rheumatology Biologics Register. Rheumatology (Oxford) 2018;57:651–5. doi: 10.1093/rheumatology/kex469. [DOI] [PubMed] [Google Scholar]
- 7.Filbin MR, Lynch J, Gillingham TD, Thorsen JE, Pasakarnis CL, Nepal S, et al. Presenting symptoms independently predict mortality in septic shock: importance of a previously unmeasured confounder. Crit Care Med. 2018;46:1592–9. doi: 10.1097/CCM.0000000000003260. [DOI] [PubMed] [Google Scholar]
- 8.Kumar A, Ellis P, Arabi Y, Roberts D, Light B, Parrillo JE, et al. Initiation of inappropriate antimicrobial therapy results in a fivefold reduction of survival in human septic shock. Chest. 2009;136:1237–48. doi: 10.1378/chest.09-0087. [DOI] [PubMed] [Google Scholar]
- 9.Rudd KE, Johnson SC, Agesa KM, Shackelford KA, Tsoi D, Kievlan DR, et al. Global, regional, and national sepsis incidence and mortality, 1990-2017: analysis for the Global Burden of Disease Study. Lancet. 2020;395:200–11. doi: 10.1016/S0140-6736(19)32989-7. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Sihvonen S, Korpela M, Laippala P, Mustonen J, Pasternack A. Death rates and causes of death in patients with rheumatoid arthritis: a population-based study. Scand J Rheumatol. 2004;33:221–7. doi: 10.1080/03009740410005845. [DOI] [PubMed] [Google Scholar]
- 11.Franklin J, Lunt M, Bunn D, Symmons D, Silman A. Risk and predictors of infection leading to hospitalisation in a large primary-care-derived cohort of patients with inflammatory polyarthritis. Ann Rheum Dis. 2007;66:308–12. doi: 10.1136/ard.2006.057265. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Colbert JF, Schmidt EP, Faubel S, Ginde AA. Severe sepsis outcomes among hospitalizations with inflammatory bowel disease. Shock. 2017;47:128–31. doi: 10.1097/SHK.0000000000000742. [DOI] [PubMed] [Google Scholar]
- 13.Barrett O, Abramovich E, Dreiher J, Novack V, Abu-Shakra M. Short- and long-term mortality due to sepsis in patients with rheumatoid arthritis. Rheumatol Int. 2017;37:1021–6. doi: 10.1007/s00296-017-3694-5. [DOI] [PubMed] [Google Scholar]
- 14.Krasselt M, Baerwald C, Petros S, Seifert O. Mortality of sepsis in patients with rheumatoid arthritis: a single-center retrospective analysis and comparison with a control group. J Intensive Care Med. 2021;36:766–74. doi: 10.1177/0885066620917588. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Sheth M, Benedum CM, Celi LA, Mark RG, Markuzon N. The association between autoimmune disease and 30-day mortality among sepsis ICU patients: a cohort study. Crit Care. 2019;23:93. doi: 10.1186/s13054-019-2357-1. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.von Elm E, Altman DG, Egger M, Pocock SJ, Gøtzsche PC, Vandenbroucke JP, et al. The Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) statement: guidelines for reporting observational studies. PLoS Med. 2007;4:e296. doi: 10.1371/journal.pmed.0040296. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Texas Department of State Health Services, Center for Health Statistics . Austin (TX): Texas Department of State Health Services, Center for Health Statistics; 2018. Texas inpatient public use data file (PUDF) [Internet] [cited 2002 Aug 13]. Available from: http://www.dshs.state.tx.us/thcic/hospitals/Inpatientpudf.shtm. [Google Scholar]
- 18.Robert Graham Center . Washington (DC): Robert Graham Center; 2021. Social deprivation index [Internet] [cited 2002 Aug 13]. Available from https://www.graham-center.org/maps-data-tools/social-deprivation-index.html. [Google Scholar]
- 19.Singer M, Deutschman CS, Seymour CW, Shankar-Hari M, Annane D, Bauer M, et al. The third international consensus definitions for sepsis and septic shock (Sepsis-3) JAMA. 2016;315:801–10. doi: 10.1001/jama.2016.0287. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Rhee C, Dantes R, Epstein L, Murphy DJ, Seymour CW, Iwashyna TJ, et al. Incidence and trends of sepsis in US hospitals using clinical vs claims data, 2009-2014. JAMA. 2017;318:1241–9. doi: 10.1001/jama.2017.13836. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.Prescott HC, Cope TM, Gesten FC, Ledneva TA, Friedrich ME, Iwashyna TJ, et al. Reporting of sepsis cases for performance measurement versus for reimbursement in New York state. Crit Care Med. 2018;46:666–73. doi: 10.1097/CCM.0000000000003005. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22.Buchman TG, Simpson SQ, Sciarretta KL, Finne KP, Sowers N, Collier M, et al. Sepsis among Medicare beneficiaries: 1. The burdens of sepsis, 2012-2018. Crit Care Med. 2020;48:276–88. doi: 10.1097/CCM.0000000000004224. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23.Kannayiram SS, Guraya A, Muojieje CC, Nair KM, Idolor ON, Odion J, et al. An analysis of rheumatoid arthritis hospitalizations. Cureus. 2020;12:e12344. doi: 10.7759/cureus.12344. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24.Edigin E, Ojemolon PE, Eseaton PO, Shaka H, Akuna E, Asemota IR, et al. Rheumatoid arthritis patients have better outcomes than non-rheumatoid arthritis patients when hospitalized for ischemic stroke: analysis of the national inpatient sample. J Clin Rheumatol. 2022;28:e13–7. doi: 10.1097/RHU.0000000000001563. [DOI] [PubMed] [Google Scholar]
- 25.Sakr Y, Jaschinski U, Wittebole X, Szakmany T, Lipman J, Ñamendys-Silva SA, et al. Sepsis in intensive care unit patients: worldwide data from the intensive care over nations audit. Open Forum Infect Dis. 2018;5:ofy313. doi: 10.1093/ofid/ofy313. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26.Kumar G, Kumar N, Taneja A, Kaleekal T, Tarima S, McGinley E, et al. Nationwide trends of severe sepsis in the 21st century (2000-2007) Chest. 2011;140:1223–31. doi: 10.1378/chest.11-0352. [DOI] [PubMed] [Google Scholar]
- 27.Angus DC, Linde-Zwirble WT, Lidicker J, Clermont G, Carcillo J, Pinsky MR. Epidemiology of severe sepsis in the United States: analysis of incidence, outcome, and associated costs of care. Crit Care Med. 2001;29:1303–10. doi: 10.1097/00003246-200107000-00002. [DOI] [PubMed] [Google Scholar]
- 28.Quan H, Sundararajan V, Halfon P, Fong A, Burnand B, Luthi JC, et al. Coding algorithms for defining comorbidities in ICD-9-CM and ICD-10 administrative data. Med Care. 2005;43:1130–9. doi: 10.1097/01.mlr.0000182534.19832.83. [DOI] [PubMed] [Google Scholar]
- 29.Quan H, Li B, Couris CM, Fushimi K, Graham P, Hider P, et al. Updating and validating the Charlson comorbidity index and score for risk adjustment in hospital discharge abstracts using data from 6 countries. Am J Epidemiol. 2011;173:676–82. doi: 10.1093/aje/kwq433. [DOI] [PubMed] [Google Scholar]
- 30.Butler DC, Petterson S, Phillips RL, Bazemore AW. Measures of social deprivation that predict health care access and need within a rational area of primary care service delivery. Health Serv Res. 2013;48(2 Pt 1):539–59. doi: 10.1111/j.1475-6773.2012.01449.x. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 31.Martin GS, Mannino DM, Eaton S, Moss M. The epidemiology of sepsis in the United States from 1979 through 2000. N Engl J Med. 2003;348:1546–54. doi: 10.1056/NEJMoa022139. [DOI] [PubMed] [Google Scholar]
- 32.Mohammed A, Alshamarri T, Adeyeye T, Lazariu V, McNutt LA, Carpenter DO. A comparison of risk factors for osteo- and rheumatoid arthritis using NHANES data. Prev Med Rep. 2020;20:101242. doi: 10.1016/j.pmedr.2020.101242. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 33.Xu Y, Wu Q. Prevalence trend and disparities in rheumatoid arthritis among US adults, 2005-2018. J Clin Med. 2021;10:3289. doi: 10.3390/jcm10153289. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 34.Funk MJ, Westreich D, Wiesen C, Stürmer T, Brookhart MA, Davidian M. Doubly robust estimation of causal effects. Am J Epidemiol. 2011;173:761–7. doi: 10.1093/aje/kwq439. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 35.Li L, Greene T. A weighting analogue to pair matching in propensity score analysis. Int J Biostat. 2013;9:215–34. doi: 10.1515/ijb-2012-0030. [DOI] [PubMed] [Google Scholar]
- 36.Austin PC. Optimal caliper widths for propensity-score matching when estimating differences in means and differences in proportions in observational studies. Pharm Stat. 2011;10:150–61. doi: 10.1002/pst.433. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 37.Austin PC. Using the Standardized difference to compare the prevalence of a binary variable between two groups in observational research. Commun Stat Simul Comput. 2009;38:1228–34. [Google Scholar]
- 38.Anesi GL, Liu VX, Chowdhury M, Small DS, Wang W, Delgado MK, et al. Association of ICU admission and outcomes in sepsis and acute respiratory failure. Am J Respir Crit Care Med. 2022;205:520–8. doi: 10.1164/rccm.202106-1350OC. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 39.Hotchkiss RS, Monneret G, Payen D. Sepsis-induced immunosuppression: from cellular dysfunctions to immunotherapy. Nat Rev Immunol. 2013;13:862–74. doi: 10.1038/nri3552. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 40.Schulte W, Bernhagen J, Bucala R. Cytokines in sepsis: potent immunoregulators and potential therapeutic targets: an updated view. Mediators Inflamm. 2013;2013:165974. doi: 10.1155/2013/165974. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 41.Davidson A, Diamond B. Autoimmune diseases. N Engl J Med. 2001;345:340–50. doi: 10.1056/NEJM200108023450506. [DOI] [PubMed] [Google Scholar]
- 42.Santamaria P. Cytokines and chemokines in autoimmune disease: an overview. Adv Exp Med Biol. 2003;520:1–7. doi: 10.1007/978-1-4615-0171-8_1. [DOI] [PubMed] [Google Scholar]
- 43.Kim W, Min S, Cho M, Youn J, Min J, Lee S, et al. The role of IL-12 in inflammatory activity of patients with rheumatoid arthritis (RA) Clin Exp Immunol. 2000;119:175–81. doi: 10.1046/j.1365-2249.2000.01095.x. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 44.Nalbant S, Birlik AM. In: New developments in rheumatoid arthritis. Sakkas LI, editor. London: IntechOpen; 2017. Cytokines in rheumatoid arthritis (RA) pp. 55–63. [Google Scholar]
- 45.Reddy RC, Chen GH, Newstead MW, Moore T, Zeng X, Tateda K, et al. Alveolar macrophage deactivation in murine septic peritonitis: role of interleukin 10. Infect Immun. 2001;69:1394–401. doi: 10.1128/IAI.69.3.1394-1401.2001. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 46.Reddy RC, Chen GH, Tekchandani PK, Standiford TJ. Sepsis-induced immunosuppression: from bad to worse. Immunol Res. 2001;24:273–87. doi: 10.1385/IR:24:3:273. [DOI] [PubMed] [Google Scholar]
- 47.Ono S, Ueno C, Aosasa S, Tsujimoto H, Seki S, Mochizuki H. Severe sepsis induces deficient interferon-gamma and interleukin-12 production, but interleukin-12 therapy improves survival in peritonitis. Am J Surg. 2001;182:491–7. doi: 10.1016/s0002-9610(01)00754-1. [DOI] [PubMed] [Google Scholar]
- 48.Baghela A, Pena OM, Lee AH, Baquir B, Falsafi R, An A, et al. Predicting sepsis severity at first clinical presentation: the role of endotypes and mechanistic signatures. EBioMedicine. 2022;75:103776. doi: 10.1016/j.ebiom.2021.103776. [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
International Classification of Diseases, ninth and tenth revisions, clinical modification (ICD-9-CM and ICD-10-CM) codes used to identify sites of infection
International Classification of Diseases, ninth and tenth revisions, clinical modification (ICD-9-CM and ICD-10-CM) codes used to identify organ dysfunctions
International Classification of Diseases, Ninth and Tenth Revisions, Clinical Modification (ICD-9-CM and ICD-10-CM) codes used to identify procedures
International Classification of Diseases, Ninth and Tenth Revisions, Clinical Modification (ICD-9-CM and ICD-10-CM) codes used to identify selected comorbidities
The characteristics and outcome of propensity score-matched sepsis hospitalizations with and without rheumatoid arthritis
Multilevel multivariable logistic regression with propensity adjustment for the association of rheumatoid arthritis with short-term mortality for the whole cohort: alternative modeling for the impact of missing gender data
Propensity matched cohort for the association of rheumatoid arthritis with short-term mortality for the whole cohort: alternative modeling for the impact of missing gender data
Multivariable logistic regression without propensity adjustment for the association of rheumatoid arthritis with short-term mortality for the whole cohort: alternative modeling for the impact of missing gender data
Multilevel multivariable logistic regression with propensity adjustment for the association of rheumatoid arthritis with short-term mortality among ICU admissions: alternative modeling for the impact of missing gender data
Multilevel multivariable logistic regression with propensity adjustment for the association of rheumatoid arthritis with short-term mortality among hospitalizations with septic shock: alternative modeling for the impact of missing gender data
Multilevel multivariable logistic regression with propensity adjustment for the association of rheumatoid arthritis with short-term mortality among hospitalizations aged ≥65 years: alternative modeling for the impact of missing gender data
Propensity scores and standardized differences of covariates for comparisons of sepsis hospitalizations with rheumatoid arthritis and those without rheumatoid arthritis. (A) Density plots of the distribution of propensity scores for sepsis hospitalizations with (RA) and without rheumatoid arthritis (Non-RA) before and after matching. (B) Absolute standardized differences between sepsis hospitalizations with and without rheumatoid arthritis for the whole cohort (before matching) and matched data.
The R code used for the study analyses.