Abstract
Objective
Systemic lupus erythematosus (SLE) has among the highest hospital readmission rates among chronic conditions. We sought to identify patient-level, hospital-level, and geographic predictors of 30-day hospital readmissions in SLE.
Methods
Using hospital discharge databases from 5 geographically dispersed states, we performed a study of all-cause SLE readmissions between 2008 and 2009. We evaluated each hospitalization as a possible index event leading up to a readmission, our primary outcome. We accounted for clustering of hospitalizations within patients and within hospitals and adjusted for hospital case-mix. Using multi-level mixed-effects logistic regression, we examined factors associated with 30-day readmissions and calculated risk-standardized hospital-level and state-level readmission rates.
Results
We examined 55,936 hospitalizations among 31,903 patients with SLE. 9,244 (16.5%) hospitalizations resulted in readmission within 30 days. In adjusted analyses, age was inversely related to risk of readmission. Black and Hispanic patients were more likely to be readmitted compared to white patients, as were those with Medicare or Medicaid insurance (versus private insurance). Several lupus clinical characteristics, including lupus nephritis, serositis and thrombocytopenia were associated with readmission. Readmission rates varied significantly between hospitals after accounting for patient-level clustering and hospital case mix. There was also geographic variation, with risk-adjusted readmission rates lower in New York and higher in Florida compared to California.
Conclusions
We found that about 1 in 6 hospitalized patients with SLE were readmitted within 30 days, with higher rates in historically underserved populations. Significant geographic and hospital-level variation in risk-adjusted readmission rates suggests potential for quality improvement.
Between 20–25% of individuals with SLE are hospitalized each year, accounting for over 140,000 hospitalizations in the United States (1–3). Health care for systematic lupus erythematosus (SLE) is complex and low overall quality and disparities have been reported for many care processes (4–8). Hospital readmissions are a potentially important outcome measure given that SLE has the sixth highest readmission rate among all medical conditions in the United States (3). The prevalence of early readmissions following these hospitalizations, a measure increasingly used to understand quality of care, costs of care and care transitions (9), has not been studied in SLE.
Most hospitalizations in SLE are for the treatment of disease manifestations, infections or associated medical co-morbidities (10). Previous studies of hospital utilization in SLE have focused on patient characteristics associated with initial hospitalization or factors associated with mortality (10–13). Identifying both risk factors associated with early readmission and variation in readmission rates for SLE could potentially inform efforts to improve the quality of care during initial hospitalizations as well as during ambulatory care transitions.
In this study, we investigated the 30-day all-cause hospital readmission rate for adults with SLE in a multi-state sample that includes a large proportion of all hospitalizations for the condition in the United States. We sought to identify factors associated with readmission in SLE and to examine variation in risk-adjusted readmission rates at the hospital and state levels. Additionally, we compared hospital variation in readmission rates for SLE to other common chronic conditions.
METHODS
Data Source and Population
We used data from the Health Care Cost and Utilization Project State Inpatient Databases from 2008 and 2009 maintained by the Agency for Healthcare Research and Quality. The State Inpatient Databases include administrative data on all inpatient discharges from acute-care, non-federal facilities, covering approximately 85% of U.S. hospitals. These data include information on patient demographic characteristics, principal and secondary diagnoses defined by the International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9 CM), expected payer, date of hospital admission, length of stay, and disposition upon discharge. Several states also provide unique patient identifiers, allowing linkage of hospitalizations for a single patient, whether at the index institution or a different institution, which makes it possible to calculate hospital and state-level readmission rates in those states.
We aggregated state databases from five geographically dispersed states (New York, Florida, Utah, California, and Washington) and that included all of the variables of interest. We included individuals who were 18 years of age or older with hospitalizations including an ICD-9 CM diagnosis of SLE (710.0). Because our primary focus was hospital readmissions, we excluded hospitalizations: 1) that were hospital transfers 2) that did not result in a discharge to home (i.e. those discharged to other hospitals or to skilled nursing or rehabilitation facilities), 3) that included maternity-related diagnoses, and 4) for which the patient died during the index hospitalization. Finally, we excluded patients whose discharge file did not include a unique identifier, since readmissions could not be accurately determined for these patients (n=3,654 encounters were missing this identifier).
The study protocol was submitted and determined to be exempt from review by the Committee on Human Research at the University of California, San Francisco.
Measures
Outcome Measure
The primary outcome measure was all-cause readmission to any acute- care hospital within 30 days of initial hospital discharge. Some readmissions may be planned or unavoidable – we did not attempt to remove these hospitalizations from our analysis. In order to ensure complete 30-day follow-up, we excluded discharges from the final month of available data, leaving us with a dataset covering January 2008 to November 2009. For discharges from hospitals in Florida and Washington State, we included index admissions from the 1st quarter of 2008 to the 3rd quarter of 2009 because only the quarter of discharge (rather than the month of discharge) was available for these states (but the linkage files still permitted determination that readmissions were within 30 days). For individuals with multiple readmissions during the year, additional hospitalizations after initial 30-day readmission were analyzed as new index hospitalizations. Patients were therefore eligible for multiple 30-day readmissions and within-patient repeated measures were accounted for in our analyses.
Independent Variables
Individual demographic characteristics included age, sex, and race/ethnicity (white, Black, Asian/Pacific Islander, Hispanic/Latino, or other). Socioeconomic status was categorized into quartiles for each state based on the median income of individual patient ZIP codes. Primary expected payer was categorized as Medicaid, Medicare, private, uninsured (defined as self-pay or no charge), or other. Because up to two expected payers can be coded for a hospital stay in HCUP, we selected the primary payer.
We used non-overlapping variables from validated indexes to characterize disease status, including variables comprising the Charlson comorbidity index (calculated from discharge diagnoses of all available encounters) (14), and variables comprising a lupus-specific mortality index developed by Ward that consists of disease features associated with high hospital mortality in SLE. Ward’s index includes common manifestations of SLE, including glomerulonephritis, chronic renal failure, pericarditis, pleuritis, psychosis, seizures, hemolytic anemia, and thrombocytopenia (15). Because of the high rate of infections in SLE related to both immunosuppression and the underlying disease, we also included a validated measure of infection, including both bacterial and opportunistic infections in our analysis (16). Length-of-stay was calculated for the index admission preceding the readmission as an additional proxy of disease severity. We also characterized causes of readmission using Clinical Classifications Software (CCS) for ICD-9-CM for descriptive purposes.
We included information from the American Hospital Association’s annual survey to determine each hospital’s size, teaching status and location (urban versus rural).
Analysis
Characteristics associated with 30-day readmissions
We examined the characteristics of initial hospitalizations among those with and without subsequent 30-day readmissions using Student’s t-tests for continuous measures and Chi-squared tests for categorical measures.
Factors associated with 30-day readmissions
We used a multilevel mixed-effects logistic model to examine factors associated with readmissions within 30 days among patients with SLE. We chose the generalized linear mixed model (GLMM) over a marginal model, such as generalized estimating equations (GEE), because we specifically wanted to obtain subject-specific estimates of the covariates, as opposed to the population averages produced by a GEE model.
We counted each hospitalization as a possible index event leading up to a readmission. Admissions were nested within patients, who were in turn nested within hospitals. In order to account for repeated observations for patients, a random intercept was estimated for each patient’s series of admissions, which represents that patient’s effect on risk of readmission. A random intercept was also estimated for each hospital, controlling for its case-mix. All other covariates were treated as fixed in order to obtain global (i.e., across all patients) estimates for the effects of these covariates on readmission risk. In our model, the fixed effects quantify the global effects of socioeconomic and demographic status, disease severity and comorbidity burden, among other covariates. The model parameter estimates were approximated by Gauss-Hermite quadrature, with 15 quadrature points.
Risk-standardized readmission rates (RSRRs)
In order to calculate risk-standardized hospital-level readmission rates, we used a multilevel mixed-effects logistic regression model to obtain the best linear unbiased predictions (BLUPs) of the risk-standardized odds ratio, OR, of readmission to each hospital. The risk-standardized readmission rate for each hospital iRSRR,i, was obtained by scaling the crude grand mean, p0, by the BLUP of the hospital’s OR as follows:
RSRRi = p0*ORi/{1+p0*(ORi-1)}
where ORi = log σi and σi denotes the estimated random effect for hospital i.
To compare readmission rates among hospitals, we constructed a sample of hospitals with ≥25 individuals with SLE admitted over the study period (n=486 hospitals). This minimum number of admissions aligns with the Centers for Medicare and Medicaid Services’ requirement of at least 25 cases for a specific condition over a three-year period to report a hospital’s excess readmission ratio. Outlier hospitals were those having 95% confidence intervals for their risk-standardized readmission rates excluding the benchmark rate among these hospitals. All variables were checked for non-linearity, and we tested for interactions between age, race/ethnicity and primary payer.
These methods are analogous to those used by the Centers for Medicare and Medicaid Services to calculate per-hospital risk-standardized readmission rates for a variety of conditions (17). The method has the advantage of shrinking rate estimates for individual hospitals towards the grand mean, depending on sample size. Hospitals with fewer cases experienced greater shrinkage of their risk-standardized readmission rates towards the grand mean compared to those with more cases. This has the effect of producing more reliable estimates of risk-standardized readmission rates, particularly for hospitals with smaller numbers of admissions.
In additional analysis, we compared the risk-standardized readmission rate for SLE with those for other common chronic conditions to see if hospitals with higher readmission rates for SLE also had higher rates for these conditions. We obtained 2009 (the first year for which these data were available) per-hospital 30-day risk-standardized readmission rate data from Medicare (through hospitalcompare.gov) for the following conditions: pneumonia, acute myocardial infarction, and chronic heart failure, and combined these with our data, resulting in three sets of pairs of risk-standardized readmission rates for SLE and each of the three conditions for each hospital. We then estimated correlation coefficients and performed linear regression analyses for each pair of risk-standardized readmission rates across all three conditions.
Data were analyzed using Stata/SE 12.0 (StataCorp LP, College Station, TX) and R version 2.14.2 (R Foundation for Statistical Computing, Vienna, Austria).
RESULTS
In 2008 and 2009, there were 55,936 hospitalizations among 31,903 individuals with SLE in the 5 states examined. Data from 810 hospitals were included in the analysis. Readmissions within 30 days occurred following 9,244 (16.5%) hospitalizations. These readmissions occurred among 4,916 unique patients. Characteristics of SLE hospitalizations with and without record linkages are provided in Supplemental Appendix 1.
Characteristics of patients with initial hospitalizations and with readmissions within 30 days are listed in Table 1. Readmitted individuals were predominantly young (82.6% <65 years), female (89.2%), and racially/ethnically diverse (54.9% from racial/ethnic minority groups). A majority had Medicare or Medicaid as the primary payer (63.4%). Readmitted patients had a high prevalence of severe lupus manifestations. The mean Charlson comorbidity index for readmitted patients was higher than for those without a readmission (4.53, SD 3.22 compared to 3.59, SD 2.97). Similarly, the mean Ward index was also higher (3.83, SD 4.04 compared to 2.89, SD 3.82). A detailed list of the primary diagnoses associated with readmission within 30 days using Clinical Classifications Software (CCS) for ICD-9-CM can be found in the Supplemental Appendix 2.
Table 1.
Characteristics of adults hospitalized for systemic lupus erythematosus in California, Florida, New York, Utah, and Washington State in 2008 and 2009.
| Admissions, total number | 55,936 | ||
|---|---|---|---|
| Readmissions within 30 days, total number (%) | 9,244 (16.5%) | ||
| Unique patients, total number | 31,903 | ||
| Demographic characteristics of unique patients | |||
| Not readmitted (n=26,987) |
Readmitted (n=4,916) |
p-value | |
| Age at first encounter | |||
| 18-30 years | 2,560 (9.5%) | 802 (16.3%) | |
| 31-45 years | 6,392 (23.7%) | 1,406 (28.6%) | |
| 46-64 years | 11,466 (42.5%) | 1,855 (37.7%) | <0.001 |
| ≥65 years | 6,569 (24.3%) | 853 (17.4%) | |
| Female | 24,129 (89.4%) | 4,385 (89.2%) | 0.677 |
| Race | |||
| White | 15,533 (57.6%) | 2,215 (45.1%) | |
| Black | 5,361 (19.9%) | 1,347 (27.4%) | |
| Asian | 957 (3.5%) | 238 (4.8%) | <0.001 |
| Hispanic | 4,271 (15.8%) | 947 (19.3%) | |
| Other | 865 (3.2%) | 169 (3.4%) | |
| Median income of patient ZIP code, state quartiles | |||
| Fourth | 5,839 (21.6%) | 893 (18.2%) | |
| Third | 6,695 (24.8%) | 1,204 (24.5%) | |
| Second | 6,793 (25.2%) | 1,228 (25.0%) | <0.001 |
| First (lowest) | 7,660 (28.4%) | 1,591 (32.4%) | |
| Primary payer | |||
| Public | 15,012 (58.3%) | 3,114 (63.4%) | |
| Private | 10,025 (37.1%) | 1,456 (29.6%) | <0.001 |
| Uninsured | 1,090 (4.0%) | 197 (4.0%) | |
| Other/unknown | 860 (3.2%) | 149 (3.0%) | |
| Clinical characteristics of hospitalizations | |||
|
Not a readmission (n=46,692) |
Readmission (n=9,244) |
||
| Charlson Comorbidity Index (mean ± S.D.) | 3.59±2.97 | 4.53±3.22 | <0.001 |
| Ward Index (mean ± S.D.) | 2.89±3.82 | 3.83±4.04 | <0.001 |
| Clinical conditions* | 8,822 (18.9%) | 2,817 (30.5%) | <0.001 |
| Nephritis | |||
| Chronic renal failure | 8,899 (19.1%) | 3,174 (34.3%) | <0.001 |
| Autoimmune hemolytic anemia | 446 (1.0%) | 111 (1.2%) | 0.034 |
| Thrombocytopenia | 2,281 (4.9%) | 665 (7.2%) | <0.001 |
| Pericarditis | 761 (1.6%) | 208 (2.3%) | <0.001 |
| Pleuritis | 1,822 (3.9%) | 475 (5.1%) | <0.001 |
| Seizure | 3,811 (8.2%) | 1,095 (11.8%) | <0.001 |
| Psychosis | 3,787 (8.1%) | 776 (8.4%) | 0.373 |
| Cancer | 1,872 (4.0%) | 411 (4.4%) | 0.056 |
| Congestive heart failure | 5,594 (12.0%) | 1,586 (17.2) | <0.001 |
| Myocardial infarction | 1,243 (2.7%) | 226 (2.4%) | 0.247 |
| Cerebrovascular accident | 960 (2.1%) | 126 (1.4%) | <0.001 |
| Diabetes | 8,348 (17.9%) | 1,713 (18.5%) | 0.140 |
| Peripheral vascular disease | 4,151 (8.9%) | 931 (10.1%) | <0.001 |
| Liver disease | 1,445 (3.1%) | 297 (3.2%) | 0.572 |
| Infection | 6,454 (13.8%) | 1,551 (16.8%) | <0.001 |
| Length of Stay, median (IQR) | 3 (2-6) | 4 (2-7) | <0.001 |
| Community and hospital characteristics | |||
| Patient geographical location | |||
| Urban | 41,748 (89.4%) | 8,438 (91.3%) | <0.001 |
| Rural | 4,944 (10.6%) | 806 (8.7%) | |
| Hospital teaching status | |||
| Non-teaching | 33,878 (72.6%) | 6,405 (69.3%) | |
| Teaching | 12,814 (27.4%) | 2,839 (30.7%) | <0.001 |
| State | |||
| California | 17,774 (38.1%) | 3,624 (39.2%) | |
| Florida | 15,183 (32.5%) | 3,536 (38.3%) | <0.001 |
| New York | 11,589 (24.8%) | 1,786 (19.3%) | |
| Utah | 753 (1.6%) | 85 (0.9%) | |
| Washington state | 1,393 (3.0%) | 213 (2.3%) | |
Clinical conditions are components of the Charlson-Deyo comorbidity index, Ward lupus mortality index and also validated definitions for infection.
IQR=Interquartile range.
The adjusted odds of 30-day readmissions are presented in Table 2; all variables listed were entered in the multilevel mixed-effects logistic model. Age had a striking inverse relationship with risk of 30-day readmission (Figure 1; OR 0.98 per year, 95% CI 0.98-0.98). Black and Hispanic patients were significantly more likely to be readmitted (OR 1.18, 95% CI 1.09-1.28, and OR 1.12, 95% CI 1.02-1.22, respectively). Similarly, compared to those having private insurance, individuals with Medicare or Medicaid as the primary payer were substantially more likely to be readmitted (OR 1.57, 95% CI 1.45-1.69 and OR 1.53, 95% CI 1.40-1.67, respectively). Those living in rural locations were less likely to be readmitted (OR 0.86, 95% CI 0.77-0.95).
Table 2.
Odds of 30-day readmission for adults with systemic lupus erythematosus in California, Florida, New York, Utah, and Washington State in 2008 and 2009.
| OR (95% C.I.) | p-value for 95% C.I. |
|
|---|---|---|
| Demographic characteristics | ||
| Age, per year | 0.98 (0.98-0.98) | <0.001 |
| Sex | ||
| Female | reference | - |
| Male | 0.93 (0.85-1.03) | 0.160 |
| Race | ||
| White | reference | - |
| Black | 1.18 (1.09-1.28) | <0.001 |
| Hispanic | 1.12 (1.02-1.22) | 0.009 |
| Asian | 1.13 (0.97-1.31) | 0.110 |
| Other | 1.14 (1.01-1.39) | 0.042 |
| Income quartile | ||
| Fourth | reference | - |
| Third | 1.08 (0.99-1.18) | 0.097 |
| Second | 1.01 (0.92-1.11) | 0.872 |
| First (lowest) | 1.06 (0.97-1.16) | 0.230 |
| Primary payer | ||
| Private insurance | reference | - |
| Medicare | 1.57 (1.45-1.69) | <0.001 |
| Medicaid | 1.53 (1.40-1.67) | <0.001 |
| Uninsured | 1.20 (1.01-1.42) | 0.040 |
| Other | 1.18 (1.01-1.38) | 0.036 |
| Clinical characteristics of hospitalizations | ||
| Clinical conditions | ||
| Nephritis | 1.25 (1.16-1.34) | <0.001 |
| Chronic renal failure | 1.61 (1.50-1.73) | <0.001 |
| Autoimmune hemolytic anemia | 1.06 (0.82-1.37) | 0.661 |
| Thrombocytopenia | 1.17 (1.04-1.30) | 0.006 |
| Pericarditis | 1.08 (0.89-1.30) | 0.428 |
| Pleuritis | 1.13 (0.99-1.27) | 0.064 |
| Seizure | 1.17 (1.07-1.28) | <0.001 |
| Psychosis | 1.11 (1.01-1.23) | 0.027 |
| Cancer | 1.64 (1.43-1.88) | <0.001 |
| Congestive heart failure | 1.40 (1.29-1.52) | <0.001 |
| Myocardial infarction | 1.07 (0.90-1.27) | 0.439 |
| Cerebrovascular accident | 0.78 (0.63-0.96) | 0.019 |
| Diabetes | 1.10 (1.02-1.18) | 0.011 |
| Peripheral vascular disease | 1.19 (1.09-1.30) | <0.001 |
| Liver disease | 0.96 (0.82-1.12) | 0.538 |
| Infection | 0.99 (0.92-1.07) | 0.847 |
| Length of stay, per day | 1.01 (1.00-1.01) | <0.001 |
| Community and hospital characteristics at index visit | ||
| Patient geographical location | ||
| Urban | reference | - |
| Rural | 0.86 (0.77-0.95) | 0.005 |
| Hospital teaching status | ||
| Non-teaching | reference | - |
| Teaching | 0.94 (0.85-1.03) | 0.185 |
| State | ||
| California | reference | - |
| Florida | 1.20 (1.11-1.32) | <0.001 |
| New York | 0.77 (0.70-0.85) | <0.001 |
| Utah | 0.76 (0.57-1.02) | 0.065 |
| Washington | 0.91 (0.75-1.12) | 0.339 |
Clinical conditions are components of the Charlson-Deyo comorbidity index, Ward lupus mortality index and also validated definitions for infection.
Data presented are adjusted results from a multilevel mixed-effects logistic model in which all of the listed variables were entered.
Figure 1.
Relationship between age and probability of one or more readmissions within 30 days of hospital discharge for systemic lupus erythematosus patients in 2008–09. Smoothing methods were used to depict probability of readmission for each age point. The 95% confidence intervals for these estimates are represented by the light blue region.
Clinical variables associated with readmission are listed in Table 2. Severe manifestations of SLE, including renal, hematological, and neurological diagnoses were associated with more 30-day readmissions, as were co-morbid conditions known to have higher prevalence in SLE, including cardiovascular disease, diabetes and cancer.
Risk-adjusted readmission rates varied between states. We found significantly lower 30-day readmission rates in New York (OR 0.77, 95% CI 0.70-0.85) and significantly higher rates in Florida (OR 1.20, 95% CI 1.11-1.32) compared to California in our adjusted analysis.
Figure 2 presents the point estimates of risk-standardized 30-day hospital readmission rates for hospitals with moderate or high volumes of SLE admissions per year (n=486). To classify outlier hospitals, we identified those having confidence intervals for their risk-standardized readmission rates excluding the benchmark rate, which was the average readmission rate (16.8%) for this subset of hospitals having extant identifiers. At the 95% confidence level depicted in Figure 2 (corresponding to two standard deviations around the mean), we identified 19 outlier hospitals compared to the benchmark, all of which had higher than expected readmission rates.
Figure 2.
Risk-standardized 30-day hospital readmission rates and 95% (i.e., 2 SD) confidence intervals are presented for 486 hospitals in California, Florida, New York, Utah, and Washington State in 2008 and 2009. Hospitals with more than 25 systemic lupus erythematosus readmissions are depicted. Each point represents a hospital, and the error bars represent the confidence interval. Low-performing outliers (n=19) are hospitals in which the confidence intervals of the risk-adjusted readmission rate falls above the mean rate of 16.8%, which is depicted with a horizontal line.
When we compared the risk-standardized readmission rates for SLE with those for three common chronic conditions to see if hospitals with higher readmission rates for SLE also had higher rates for these other conditions, we did not find statistically significant correlations (for heart failure, rho=0.038 and p=0.330, for anterior myocardial infarction, rho=0.001 and p=0.892, and for pneumonia, rho=0.047 and p=0.229). Our linear regression analyses for the per-hospital risk-standardized readmission rates across all three conditions also did not reveal a statistically significant relationship.
We did not find any interactions between age, race/ethnicity, and primary payer.
Discussion
In this population-based study of 55,936 hospitalizations for SLE, we found that about 1 in 6 patients returned to the hospital within 30 days. Because we used a multi-state, multi-payer sample, our results provide the first large-scale depiction of readmissions in SLE. We found that readmissions occurred most frequently among young, racial/ethnic minorities, those with publicly funded health insurance, and those with specific SLE manifestations, such as renal disease, neurological disease, and thrombocytopenia. After adjusting for differential case-mix based on these characteristics as well as clustering of hospitalizations within patients, we identified statistically significant variation in readmission rates for SLE both among hospitals and among the five states examined. In addition, we found that hospitals with higher readmission rates for SLE did not have higher readmission rates for other common chronic conditions (heart failure, myocardial infarction, pneumonia), justifying the importance of examining condition-specific readmissions for SLE.
Although 30-day hospital readmissions are increasingly tied to financial quality programs in the U.S. health care system, the purpose of our study was not to examine readmissions as a performance measure. Instead, we were interested in understanding the utility of readmissions in SLE in making inferences about utilization and quality, particularly in identifying groups at risk for readmission and investigating variations across the population. This distinction is important since it may be impractical to subject hospitals to performance measures for each chronic condition separately. However, for the purposes of population health management or quality improvement, examining readmissions for individual chronic conditions might inform local or regional efforts to organize and improve chronic disease care, particularly for severe conditions such as SLE.
As in other chronic conditions, we found that patient sociodemographic characteristics were an important predictor of readmissions in SLE. Historically underserved populations were at highest risk of 30-day readmissions, including Black and Hispanic patients and those with publicly funded insurance, even after adjusting for a variety of patient, disease, and health care system factors. This adds to a growing literature revealing striking disparities in SLE; racial/ethnic minorities and those with low socioeconomic status have a higher prevalence of the disease (18), significantly greater disease-related organ damage, and higher mortality (19–22). The finding that age was inversely related to the risk of readmission is also noteworthy. SLE differs from most chronic conditions examined to date, where the risk of hospital readmissions has been found to increase with age (9, 23). This difference may at least in part reflect the greater severity of SLE in younger individuals, who are more likely to have life-threatening organ manifestations (24, 25). More research is needed, but this paradoxical relationship of age with readmissions illustrates that investigating condition-specific factors in SLE may be important in appropriately identifying patients at risk for readmission. Further research is required to understand whether interventions to improve care transitions for at-risk populations can reduce readmissions and ultimately improve health outcomes.
We found that severe manifestations of SLE, such as renal disease, thrombocytopenia, serositis, and seizures were associated with higher readmission rates. SLE-related comorbidities, including cardiovascular disease, diabetes and malignancy, were also associated with more readmissions (26, 27). While more severe cases of many chronic illnesses have higher readmission rates, these findings suggest that SLE patients with these clinical characteristics also warrant more careful attention, especially when coordinating health care transitions and specialty care follow-up.
After adjusting for differential case-mix based on these demographic and clinical characteristics, we found variation in readmission rates between the hospitals examined. In other chronic conditions, variation in hospital readmission rates has signaled opportunities for quality improvement, particularly in discharge planning, post-hospitalization follow-up, adherence, or coordination of care with providers in the ambulatory setting (28–30). Patients with SLE may be particularly vulnerable to problems in these areas given the counter-intuitive severity of symptoms of younger patients, the complexity of the disease and the limited experience of many care coordinators and clinicians with its management.
Finally, we found further evidence of significant variation in readmission rates when we aggregated data at the state-level, which allowed us to include the 40% of hospitals in our sample that had relatively low volumes of SLE admissions. Both high-performing and low-performing states were identified. Our study does not address the reasons underlying this state-level variation. However, it is interesting to note that the state with significantly lower risk-adjusted readmission rates for SLE, New York, also has significant clinical expertise in SLE, with a high concentration of dedicated SLE centers (31). The impact of how SLE care is organized and its influence on outcomes such as readmissions will require further study.
Taken together, our findings regarding both risk factors and variability in 30-day readmissions suggest that readmissions may be an important outcome measure in SLE. First, our data suggest that it is possible to identify both demographic and clinical risk factors for readmission in SLE. For clinicians caring for patients with SLE in both the hospital and ambulatory settings, information on these risk factors can help identify patients that need particularly careful attention to care transitions and specialty care follow-up. Second, although there is much interest in reducing the striking racial/ethnic and socioeconomic disparities in SLE, there are currently few tools available for this purpose. The relatively low prevalence and complexity of the disease have posed challenges in developing broadly applicable outcomes to track potential health disparities and begin to investigate their causes or to design interventions. Since up to a quarter of all patients with SLE are hospitalized each year, hospitalizations might serve as a critical opportunity for study and intervention. Finally, although our study does not address the reasons for variation in readmission rates between hospitals and states, the presence of unexplained variation after careful risk-adjustment suggests that there is room for quality improvement. Further work to identify care processes that can reduce readmission rates for this complex disease is needed.
Our study has limitations. Findings based on the five states examined may not be nationally representative. However, our study covers a substantial number of admissions among individuals with SLE in the United States. Although administrative definitions of SLE used in our study have moderate or good sensitivity and specificity (32), some patients may have been miscoded as having SLE, potentially decreasing the precision of our estimates. Several states in our study did not provide information on whether admissions were planned or acute, and therefore some planned rehospitalizations may have been misclassified as unplanned readmissions. Finally, because we used administrative data, clinical information on hospitalizations was limited, and we could not assess the preventability of readmissions.
In conclusion, we found that nearly 1 in 6 hospitalized patients with SLE were readmitted within 30 days. Significant geographic and hospital-level variation in risk-adjusted readmission rates suggests potential for quality improvement.
Supplementary Material
Acknowledgments
Grant Support: Research reported in this manuscript was supported by the National Institute of Arthritis and Musculoskeletal and Skin diseases under award numbers K23 AR060259 and P60 AR053308. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health. Additional support was provided by the UCSF Comparative Effectiveness Large Dataset Analysis Core (CELDAC), and the Rosalind Russell Medical Research Center for Arthritis. Dr. Ward was supported by the National Institutes of Arthritis and Musculoskeletal and Skin Diseases, National Institutes of Health.
Footnotes
No financial or other disclosures.
References
- 1.Chakravarty EF, Bush TM, Manzi S, Clarke AE, Ward MM. Prevalence of adult systemic lupus erythematosus in California and Pennsylvania in 2000: estimates obtained using hospitalization data. Arthritis and rheumatism. 2007;56(6):2092–2094. doi: 10.1002/art.22641. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2.Panopalis P, Yazdany J, Gillis JZ, Julian L, Trupin L, Hersh AO, et al. Health care costs and costs associated with changes in work productivity among persons with systemic lupus erythematosus. Arthritis and rheumatism. 2008;59(12):1788–1795. doi: 10.1002/art.24063. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Elixhauser A, Steiner C. Healthcare Cost and Utilization Project (HCUP) Statistical Briefs. >Rockville (MD): 2006. Readmissions to U.S. Hospitals by Diagnosis, 2010: Statistical Brief #153. [Google Scholar]
- 4.Yazdany J, Tonner C, Trupin L, Panopalis P, Gillis JZ, Hersh AO, et al. Provision of preventive health care in systemic lupus erythematosus: data from a large observational cohort study. Arthritis research & therapy. 12(3):R84. doi: 10.1186/ar3011. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Schmajuk G, Yelin E, Chakravarty E, Nelson LM, Panopolis P, Yazdany J. Osteoporosis screening, prevention, and treatment in systemic lupus erythematosus: application of the systemic lupus erythematosus quality indicators. Arthritis care & research. 62(7):993–1001. doi: 10.1002/acr.20150. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Demas KL, Keenan BT, Solomon DH, Yazdany J, Costenbader KH. Osteoporosis and cardiovascular disease care in systemic lupus erythematosus according to new quality indicators. Seminars in arthritis and rheumatism. 40(3):193–200. doi: 10.1016/j.semarthrit.2010.01.001. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Yazdany J, Trupin L, Tonner C, Dudley RA, Zell J, Panopalis P, et al. Quality of care in systemic lupus erythematosus: application of quality measures to understand gaps in care. Journal of general internal medicine. 27(10):1326–1333. doi: 10.1007/s11606-012-2071-z. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Yazdany J, Feldman CH, Liu J, Ward MM, Fischer MA, Costenbader KH. Quality of care for incident lupus nephritis among medicaid beneficiaries in the united states. Arthritis care & research. 2013 doi: 10.1002/acr.22182. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Dharmarajan K, Hsieh AF, Lin Z, Bueno H, Ross JS, Horwitz LI, et al. Diagnoses and timing of 30-day readmissions after hospitalization for heart failure, acute myocardial infarction, or pneumonia. Jama. 309(4):355–363. doi: 10.1001/jama.2012.216476. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Thorburn CM, Ward MM. Hospitalizations for coronary artery disease among patients with systemic lupus erythematosus. Arthritis and rheumatism. 2003;48(9):2519–2523. doi: 10.1002/art.11241. [DOI] [PubMed] [Google Scholar]
- 11.Ward MM. Hospital experience and mortality in patients with systemic lupus erythematosus: which patients benefit most from treatment at highly experienced hospitals? The Journal of rheumatology. 2002;29(6):1198–1206. [PubMed] [Google Scholar]
- 12.Krishnan E. Hospitalization and mortality of patients with systemic lupus erythematosus. The Journal of rheumatology. 2006;33(9):1770–1774. [PubMed] [Google Scholar]
- 13.Edwards CJ, Lian TY, Badsha H, Teh CL, Arden N, Chng HH. Hospitalization of individuals with systemic lupus erythematosus: characteristics and predictors of outcome. Lupus. 2003;12(9):672–676. doi: 10.1191/0961203303lu452oa. [DOI] [PubMed] [Google Scholar]
- 14.Klabunde CN, Potosky AL, Legler JM, Warren JL. Development of a comorbidity index using physician claims data. Journal of clinical epidemiology. 2000;53(12):1258–1267. doi: 10.1016/s0895-4356(00)00256-0. [DOI] [PubMed] [Google Scholar]
- 15.Ward MM. Development and testing of a systemic lupus-specific risk adjustment index for in-hospital mortality. The Journal of rheumatology. 2000;27(6):1408–1413. [PubMed] [Google Scholar]
- 16.Schneeweiss S, Robicsek A, Scranton R, Zuckerman D, Solomon DH. Veteran's affairs hospital discharge databases coded serious bacterial infections accurately. Journal of clinical epidemiology. 2007;60(4):397–409. doi: 10.1016/j.jclinepi.2006.07.011. [DOI] [PubMed] [Google Scholar]
- 17.Dharmarajan K, Hsieh AF, Lin Z, Bueno H, Ross JS, Horwitz LI, et al. Diagnoses and timing of 30-day readmissions after hospitalization for heart failure, acute myocardial infarction, or pneumonia. JAMA. 2013;309(4):355–363. doi: 10.1001/jama.2012.216476. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.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 and rheumatism. 2008;58(1):15–25. doi: 10.1002/art.23177. [DOI] [PubMed] [Google Scholar]
- 19.Alarcon GS, McGwin G, Jr, Bastian HM, Roseman J, Lisse J, Fessler BJ, et al. Systemic lupus erythematosus in three ethnic groups. VII [correction of VIII]. Predictors of early mortality in the LUMINA cohort. LUMINA Study Group. Arthritis and rheumatism. 2001;45(2):191–202. doi: 10.1002/1529-0131(200104)45:2<191::AID-ANR173>3.0.CO;2-2. [DOI] [PubMed] [Google Scholar]
- 20.Demas KL, Costenbader KH. Disparities in lupus care and outcomes. Current opinion in rheumatology. 2009;21(2):102–109. doi: 10.1097/BOR.0b013e328323daad. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.Feldman CH, Hiraki LT, Liu J, Fischer MA, Solomon DH, Alarcon GS, et al. Epidemiology and sociodemographics of systemic lupus erythematosus and lupus nephritis among U.S. adults with medicaid coverage, 2000–2004. Arthritis and rheumatism. doi: 10.1002/art.37795. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22.Alarcon GS, McGwin G, Jr, Bartolucci AA, Roseman J, Lisse J, Fessler BJ, et al. Systemic lupus erythematosus in three ethnic groups. IX. Differences in damage accrual. Arthritis and rheumatism. 2001;44(12):2797–2806. doi: 10.1002/1529-0131(200112)44:12<2797::aid-art467>3.0.co;2-9. [DOI] [PubMed] [Google Scholar]
- 23.Podulka J, Barrett M, Jiang HJ, Steiner C. 30-Day Readmissions following Hospitalizations for Chronic vs. Acute Conditions, 2008: Statistical Brief #127. 2006 [PubMed] [Google Scholar]
- 24.Hersh AO, Trupin L, Yazdany J, Panopalis P, Julian L, Katz P, et al. Childhood-onset disease as a predictor of mortality in an adult cohort of patients with systemic lupus erythematosus. Arthritis care & research. 62(8):1152–1159. doi: 10.1002/acr.20179. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25.Levy M, Montes de Oca M, Claude-Babron M. Unfavorable outcomes (end-stage renal failure/death) in childhood onset systemic lupus erythematosus. A multicenter study in Paris and its environs. Clinical and experimental rheumatology. 1994;12(Suppl 10):S63–S68. [PubMed] [Google Scholar]
- 26.Bernatsky S, Boivin JF, Joseph L, Rajan R, Zoma A, Manzi S, et al. An international cohort study of cancer in systemic lupus erythematosus. Arthritis and rheumatism. 2005;52(5):1481–1490. doi: 10.1002/art.21029. [DOI] [PubMed] [Google Scholar]
- 27.Manzi S, Meilahn EN, Rairie JE, Conte CG, Medsger TA, Jr, Jansen-McWilliams L, et al. Age-specific incidence rates of myocardial infarction and angina in women with systemic lupus erythematosus: comparison with the Framingham Study. American journal of epidemiology. 1997;145(5):408–415. doi: 10.1093/oxfordjournals.aje.a009122. [DOI] [PubMed] [Google Scholar]
- 28.Vinson JM, Rich MW, Sperry JC, Shah AS, McNamara T. Early readmission of elderly patients with congestive heart failure. Journal of the American Geriatrics Society. 1990;38(12):1290–1295. doi: 10.1111/j.1532-5415.1990.tb03450.x. [DOI] [PubMed] [Google Scholar]
- 29.Jencks SF, Williams MV, Coleman EA. Rehospitalizations among patients in the Medicare fee-for-service program. The New England journal of medicine. 2009;360(14):1418–1428. doi: 10.1056/NEJMsa0803563. [DOI] [PubMed] [Google Scholar]
- 30.Coleman EA, Parry C, Chalmers S, Min SJ. The care transitions intervention: results of a randomized controlled trial. Archives of internal medicine. 2006;166(17):1822–1828. doi: 10.1001/archinte.166.17.1822. [DOI] [PubMed] [Google Scholar]
- 31.Isenberg DA, Ramsey-Goldman R, Gladman D, Hanly JG. The Systemic Lupus International Collaborating Clinics (SLICC) group - it was 20 years ago today. Lupus. 20(13):1426–1432. doi: 10.1177/0961203311421501. [DOI] [PubMed] [Google Scholar]
- 32.Moores KG, Sathe NA. A systematic review of validated methods for identifying systemic lupus erythematosus (SLE) using administrative or claims data. Vaccine. 2013;31(Suppl 10):K62–K73. doi: 10.1016/j.vaccine.2013.06.104. [DOI] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.