Prior
studies of patients with rheumatic diseases admitted to ICU care have been
limited to case series; none included patients admitted after 2010. We examined
a modern cohort of rheumatic disease patients admitted to a tertiary care center
ICU. Given recent changes in the management of rheumatic diseases, we
hypothesized that infection would be a leading cause of mortality compared to mortality
from disease complications [e.g. interstitial lung disease (ILD), pulmonary
hypertension (PH)].

Methods:

We
queried the University of California, San Francisco electronic health record
(EHR) to identify patients with rheumatic disease and an ICU admission between
6/13/2012-6/5/2015. We included patients with an ICD9 code for 1 of 11
rheumatologic diagnoses listed among admission diagnoses, hospital-problem
lists, discharge diagnoses, or billing data for the encounter. We assessed
in-hospital mortality (primary outcome) and other covariates including reason for
ICU admission and immunosuppressive use via chart review of the EHR by 1 author.
ICU encounters following elective surgeries, for the sole purpose of medication
titration, or for the primary reasons of myocardial infarction and arrhythmia
were excluded. The primary outcome was in-hospital mortality. EHR chart review
was performed to confirm diagnoses, determine the cause of ICU admission, and identify
immunosuppressive use.

Results:

239
ICU encounters for 204 patients were identified. Diagnoses are listed in the
Table.  37 (15%) of encounters resulted
from primary rheumatic disease flare, 48 (20%) from disease complications (e.g.
ILD, PH), and 109 (46%) from presumably unrelated medical illnesses. 81 (34%) encounters
were caused or complicated by sepsis. Overall in-hospital mortality for this
cohort was 15%, with higher mortality among patients with a new rheumatic
disease diagnosis made during that same admission (25%) and disease
complications (23%), although these did not reach statistical significance. Additional
results are shown in the Table. Sepsis-related mortality was
not higher in patients receiving significant immunosuppression (10%, p=0.77)
defined as prednisone > 15 mg/day or the use of another cytotoxic agent.

Conclusion:

To
our knowledge, this is the largest study of patients with rheumatic disease admitted
to the ICU.  Compared with prior studies,
this cohort had lower overall mortality (16% vs. 17-55% in prior reports), and
infection was not the major cause of in-hospital death.  Additional analyses are required to control
for potential confounders. In the future, we plan to investigate the
performance of traditional markers of ICU mortality ( e.g.
APACHE scores, vasopressors, ventilation) in this cohort and expand these
methods (automated extraction of structured and unstructured (text) data) to
construct a registry of critically ill rheumatic disease patients to guide
future research.