Hospitalization and Mortality Due to Infection Among Children and Adolescents With Systemic Lupus Erythematosus in the United States

DISCUSSION

In our analysis of a large, multicenter national dataset of > 8000 youth with cSLE, we assessed frequency and types of infection, risk factors for ICU admission and death, and trends in infection frequency and in-hospital mortality over time. We found that hospitalizations for infection comprised an increasing proportion of total admissions among children and youth with SLE over our study period from 2009 through 2021. Although these infection-related hospitalizations still accounted for only a small proportion of total cSLE inpatient encounters, they contributed the majority of in-hospital mortality events, with over half of all total deaths occurring in hospitalizations with infection. Whereas over time, all-cause mortality rates have decreased among children and youth with cSLE,¹⁵ we found that the mortality rate attributable to infections among those hospitalized in our cohort was sustained over the study period, with an average of 2% per year. The observed mortality rate is somewhat lower than in studies of hospitalized adults with SLE, including Medicaid enrollees, with an inpatient and up-to-30-day postdischarge mortality rate of 6.9%¹ and, in a Spanish national registry, an inpatient mortality rate of 5%.¹⁶ In addition, our observed mortality rate was lower than that reported in children with cSLE enrolled in Medicaid from 2000 to 2006 (4.4% mortality within 30 days of admission for infection).⁸ Although some of the difference may be due to changes in treatment over time, we did not see an increased risk of mortality or ICU admission among publicly insured children in our cohort. It is possible that our lower observed mortality rate compared to the prior pediatric Medicaid study reflects differences in care settings, as Medicaid enrollees in our cohort were treated at the same hospitals as those who were privately insured. Despite lower mortality rates in pediatric vs adult studies, the mortality rate differences observed between children with cSLE and adults with SLE are far less than the expected differences in overall mortality in these age groups in the general population, underscoring the disproportionate effect of infection-related death in children and youth with cSLE. Potential explanations for lack of improvement in mortality among children with cSLE and infection over time in the PHIS dataset include the fact that, despite an overall lower coronavirus disease 2019 (COVID-19) mortality rate compared to adult SLE cohorts, cases of COVID-19–related mortality obscured a general trend toward improvement in infection-related mortality over the preceding decade. It is also possible that high-dose glucocorticoid use has continued to be common in cSLE treatment, and that new targeted immunosuppressants infer additional infection risks, though we were unable to test these hypotheses directly. Alternatively, it is also possible that practice changes over time may have affected our analysis. For example, there may have been a shift to more cSLE care occurring in infusion centers and the outpatient setting in recent years, resulting in relatively fewer uncomplicated cSLE admissions.

As has been previously demonstrated in prior SLE studies among both children and adults, bacterial infections were common reasons for hospitalization and death. Despite the broad implementation of the pneumococcal conjugate vaccination since our study start,^9,10 bacterial pneumonia remained a leading contributor to hospitalization and inpatient mortality in the PHIS SLE cohort, consistent with prior work in a cohort of Medicaid-enrolled children with cSLE.⁸ The persistence of bacterial pneumonia as a leading cause of infection and death among youth with cSLE suggests that further preventive care approaches may be important, including administration of the pneumococcal polysaccharide vaccine as recommended by the American College of Rheumatology and the American Academy of Pediatrics.^17,18 We also found a relatively high frequency of viral and fungal infections, along with high mortality with fungal infections including candidiasis, aspergillosis, and PJP. The higher burden of fungal infections in our cohort compared to prior US studies may be partly due to differences in case ascertainment, as we included a comprehensive list of viral and fungal codes, although fungal infection has also been identified as a leading cause of mortality in patients with cSLE in low-resource settings.¹⁹ We also included codes for systemic candidiasis, which has been omitted in some studies due to concerns about the specificity of ICD-9 coding; however, this was the largest category of fungal infections ending in in-hospital mortality in our cohort.^1,8,20

Among our cohort, PJP had the highest mortality rate of any specific type of infection (21%). Prior studies have reported highly variable rates of PJP among adults with SLE and other rheumatic diseases,^21-25 including one demonstrating low rates of PJP and higher rates of drug adverse events from antibiotic prophylaxis, leading to uncertainty about the value of PJP prophylaxis.²¹ However, our results demonstrate that, although very rare, PJP remains an important contributor to death among children and youth with cSLE, with 5% of children who died having a PJP diagnosis. Although we were not able to assess prescription of or adherence to PJP prophylaxis in our cohort due to the lack of outpatient medication data, our findings suggest that PJP prophylaxis may be an important intervention among highly immunocompromised youth with cSLE.

We observed a markedly lower COVID-19 in-hospital mortality rate (2%) compared to studies of adults with SLE and other rheumatic diseases (≥ 20%).^26-28 While prior data on COVID-19 in children and youth with cSLE are limited, the 2 previous fatalities reported in patients with cSLE diagnosed with COVID-19 in the Childhood Arthritis and Rheumatology Research Alliance and European Alliance of Associations for Rheumatology/Paediatric Rheumatology European Society cohorts were limited to those with new diagnoses of SLE at the time of COVID-19 infection²⁹; these findings may be consistent with a substantially lower risk of mortality from COVID-19 infection alone among children and youth with cSLE compared to adult SLE. We were unable to determine COVID-19 vaccination status; however, the majority of the time frame of our study (discharges ending in 2021) preceded widespread COVID-19 vaccination among children and young adults in the US.

We observed a large number of cases of herpes zoster infection, despite excluding ICD codes for postherpetic neuralgia and other postinfectious complications. High rates of herpes zoster have previously been demonstrated in adults with SLE.^23,30,31 Greater recognition of the burden of herpes zoster in children and adolescents with cSLE may have implications for preventive care, as current vaccine guidelines suggest herpes zoster vaccination only for immunocompromised adults aged ≥ 19 years.

Consistent with previous literature, we found that youth with cSLE and renal involvement were at increased risk of severe outcomes from infection, including ICU admission and death.³² Those with ESRD had increased risk of both ICU admission and death in our cohort, whereas those with LN were more likely to be admitted to an ICU. It is possible that medications used more frequently to treat LN, such as cyclophosphamide and pulse-dose glucocorticoids, may contribute to the risk of severe outcomes from infection, as these have previously been identified as risk factors.^13,32 However, we were unable to assess the contribution of immunomodulatory regimens to infection outcomes directly since only inpatient medication data were available in PHIS. Dialysis itself may also contribute to infection risk, but we were unable to reliably determine outpatient peritoneal vs hemodialysis regimens given the possibility for use of different modalities in the acute care setting.

We did not observe any significant association between ICU admission or death with type of medical insurance. Compared to the weighted mean, children who were of non-Hispanic White or unknown race and ethnicity had higher rates of ICU admission, whereas non-Hispanic Black children were less likely to be admitted to an ICU. Those who were Asian or of unknown race and ethnicity also had significantly higher mortality rates in our cohort of patients with infections. These findings could represent differences in disease severity or care received, although conclusions are limited by our cohort being defined by inpatient admission, as children of different racial and ethnic groups may have had variable rates of admission for routine care such as medication infusion. As previously demonstrated in PHIS,¹⁴ we observed that youth who were publicly insured, had lower household income, or were Black or Hispanic had higher overall rates of hospitalization. Therefore, our finding of lower rates of ICU admission and in-hospital mortality in some demographic groups could also suggest that youth in these groups were more likely to be admitted for routine care. Interpretation of racial and ethnic group differences in outcomes is also limited by the number of children whose race/ethnicity was recorded as “unknown,” and inability to verify that racial and ethnic identifiers submitted by contributing PHIS hospitals accurately reflected patient self-identification.

Some of the strengths of our approach include the use of a large multicenter dataset, with > 8000 youth with cSLE identified across > 50 hospitals. We demonstrated that the ICD-10 codes for SLE have high validity for physician diagnosis of SLE in pediatric hospitals. We were also able to validate infection codes with pharmacy data, increasing the specificity of infection identification in this study.

Our approach also had several limitations, including the fact that this was a retrospective study using secondary data. As diagnoses were determined by discharge billing codes, we were unable to determine the primary reason for hospitalization and death—including among patients who had codes for multiple infections—and it is possible that noninfectious complications of SLE or other comorbidities may have also contributed to in-hospital mortality. Although this was a large and diverse national cohort, PHIS includes only freestanding children’s hospitals, and most of these participating institutions have highly specialized pediatric ICUs as well as pediatric subspecialty care including pediatric rheumatologists, pediatric nephrologists, and pediatric infectious disease doctors. Therefore, the care received and treatment outcomes may not be generalizable to cSLE more generally, especially in relation to care received in primarily adult or community hospital settings, where an unknown proportion of cSLE care is delivered, or in non-US settings. The PHIS dataset includes only emergency department and inpatient encounters, and therefore the total population of youth with SLE served by these hospitals, including those cared for only in the outpatient setting, is unknown. We were therefore unable to estimate incidence or prevalence of hospitalized infections in the cSLE population. It is possible that some patients identified in our cohort had additional hospitalized encounters in hospitals that do not participate in PHIS, which we were unable to identify. Similarly, we were unable to assess the date of first SLE diagnosis using PHIS codes, as it is possible that initial cSLE diagnosis occurred in the outpatient setting. Although we identified LN and ESRD as significant predictors of ICU admission and in-hospital mortality, we had limited data on ESRD management, including the contribution of different modalities of dialysis to infection risk. The PHIS does not contain outpatient medication prescription information, and therefore we were unable to assess outpatient oral glucocorticoid use or types of immunosuppressants used to treat cSLE as risk factors for infection or poor outcomes among those hospitalized with infection.

In conclusion, we found that infections comprised a small but increasing proportion of hospitalizations among a large national cohort of children and adolescents with cSLE. Hospitalizations with infections accounted for the majority of in-hospital mortality. Although bacterial infections remained common, mortality in patients with fungal and viral infections was also high in our cohort. Children and adolescents with LN and ESRD were at increased risk of poor outcomes. Further research is needed to estimate the incidence and prevalence of hospitalized infections using population-based pediatric SLE cohorts, and to understand modifiable risk factors for infection and associated mortality, including the contribution of immunomodulatory medications, glucocorticoid dosing, and preventive care practices like vaccination and infectious prophylaxis.