Healthcare utilization and costs associated with S. aureus and P. aeruginosa pneumonia in the intensive care unit: a retrospective observational cohort study in a US claims database

This is the largest US claims database study of healthcare costs and outcomes for
ICU patients with a diagnosis of S. aureus or P. aeruginosa pneumonia. Our findings highlight the comprehensive economic consequences attributed
to S. aureus and P. aeruginosa pneumonia and can permit policy makers, payers, and healthcare providers to assess
the effect of prevention or therapeutic efforts on the cost and morbidity of these
ICU infections.

In our study, ICU patients with pneumonia had substantially higher healthcare costs
during the index admission: ?$213,000 for P. aeruginosa pneumonia and??$146,000 for with S. aureus pneumonia versus?$33,000 for patients without pneumonia. Increased utilization continued
after index hospitalization discharge, with total healthcare costs through 90 days
post discharge of??$17,000 for patients with S. aureus pneumonia and??$22,000 for patients with P. aeruginosa pneumonia versus??$10,000 for patients without pneumonia. Patients with S. aureus or P. aeruginosa pneumonia had estimated incremental index hospitalization costs of $100,000-$160,000
and total healthcare costs of $107,000-$167,000 versus ICU patients without pneumonia.
In previous studies of the general US inpatient population, the mean cost of hospital
care in patients with hospital-acquired pneumonia was $72,000 versus $46,400 to $65,292
for patients without pneumonia 2], 22]. A German study reported an excess mean cost of DM29,610 (equivalent to $16,824 in
2001) for hospital-acquired pneumonia in ICU patients compared with ICU patients without
pneumonia 23]. Previous US studies in VAP patients estimated that the incremental costs were $39,000
to $100,000 for VAP patients versus ICU patients without VAP, which is consistent
with our observations 24], 25].

Our findings showed that S. aureus pneumonia and, especially, P. aeruginosa pneumonia had prolonged hospitalizations (48 days longer for P. aeruginosa and 30 days longer for S. aureus) and ICU stays (approximately 15 days for those with P. aeruginosa pneumonia, 7 days for those with S. aureus pneumonia, compared with 1 day for those without pneumonia). Mortality was also substantially
higher: 20 % of those with P. aeruginosa pneumonia and 16 % of those with S. aureus pneumonia died during the index hospitalization versus approximately 3 % of control
patients, and this trend continued after discharge. The findings of longer ICU lengths
of stay and longer hospital lengths of stay are consistent with prior studies 2], 22]–24].

We found significantly higher rehospitalization rates in patients with S. aureus and P. aeruginosa pneumonia versus controls. Readmission after discharge can be costly and problematic
for the patient, hospitals, and payers. It was estimated that hospital readmissions
cost Medicare $17.5 billion alone annually in the United States 26]. Financial penalties are now incurred for hospitals with excess 30-day readmission
in the United States. Therefore, preventing S. aureus and P. aeruginosa pneumonias could reduce the substantial costs of and healthcare utilization associated
with rehospitalization in patients with these infections, resulting in benefits for
patients, payers, and healthcare providers.

ICU patients with S. aureus or P. aeruginosa pneumonia were older, more likely to be male, and in poorer general health than those
who did not have pneumonia. Most patients (62 %) were mechanically ventilated; this
is likely to be an underestimate, because mechanical ventilation is not uniformly
recorded in a claims database. These mechanically-ventilated patients represent both
those with VAP and those with non-ventilator-associated pneumonia that required ventilator
support. The demographic skew to older, male patients has been previously reported
in patients with VAP; however, in some studies, patients with VAP were younger than
were those without VAP, which may reflect the contribution of trauma patients to the
demographics 24], 27].

The findings regarding the impact of comorbid conditions on the hospitalization costs
of S. aureus or P. aeruginosa pneumonia are unique to the present study. The results indicate that while the presence
of comorbid conditions increases the likelihood that an ICU patient will develop S. aureus or P. aeruginosa pneumonia, the presence of comorbidities does not inflate the costs of hospital care
for the patient. A number of comorbid conditions were associated with lower, rather
than higher, hospitalization costs for S. aureus or P. aeruginosa pneumonia. Exploratory analysis demonstrated that the lower costs are not solely
due to higher mortality or higher rates of Medicare coverage among patients with the
comorbid conditions. It is unknown why these inverse associations exist.

Claims data represent an excellent starting point for the examination of health outcomes,
treatment patterns, healthcare resource utilization, and costs. However, our analyses
have several limitations. This study was retrospective and used ICD-9 codes to include
and exclude subjects from specific cohorts, a method associated with known limitations
28]. Because the included codes were selected to be specific rather than sensitive, the
numbers of patients identified cannot be used to establish incidence rates. It is
possible that other pneumonia codes, including the code for VAP, included some episodes
of P. aeruginosa or S. aureus pneumonia. Comorbidities were identified by their presence in coded diagnoses in
the 12 months before the index hospitalization; this approach is likely very sensitive
and intended to capture as many comorbid conditions as possible but likely has limited
specificity due to the criteria of only requiring one diagnosis code. Diagnoses made
during the index hospitalization were not included, because the goal of the current
analysis was to assess the information available to clinicians upon admission; this
may have contributed to underdiagnosis of co-morbid conditions. Future studies may
want to examine the association between costs and both pre-existing and new comorbidities
to explore the impact of how comorbidities are defined. Additionally, separating patients
into those hospitalized for a life-threatening pneumonia from those who acquired pneumonia
in the hospital would be useful in assessing prevention strategies, however claims
database analyses do not permit separation into these groups. Furthermore, the database
only included patients who had commercial healthcare insurance and may not be representative
of the overall US population; patients covered solely by Medicare or Medicaid were
excluded. However, the claims database comprised patients aged ?18 years in a managed
care setting and included substantial numbers of subjects of Medicare age; moreover,
results may be generalizable to similar patient groups. Despite these limitations,
our results provide accurate assessments of resource utilizations and costs in these
patients due to appropriate adjustments in the statistical models.