Avoiding time-dependent bias in estimating the attributable cost of healthcare-associated methicillin-resistant Staphylococcus aureus infections
Previous estimates of the impact of healthcare-associated infection (HAI) on healthcare costs have been inflated due to time-dependent bias. In this analysis, we demonstrate the magnitude of this inflation by estimating the healthcare costs attributable to an HAI due to methicillin resistant Staphylococcus aureus (MRSA) using a unique dataset in the Department of Veterans Affairs (VA) that allowed us to distinguish between costs that occurred before and after an HAI.
Data from the VA's activity-based accounting system provides a separate observation for every calendar month during a patient's inpatient stay. In our "improved analysis", we used this data to construct multivariable generalized linear models to compare the healthcare costs occurring after the 1st calendar month in patients with an MRSA HAI on the 1st day of a calendar month with those without an MRSA HAI. We compared these results to those from a "conventional analysis" which considered healthcare costs across the patients' entire inpatient stay in those with and without an HAI. Both analyses are described in Figure 1.
Our study cohort consisted of patients with an inpatient admission lasting longer than 48 hours within the VA between 10/1/2007 and 11/30/2010. MRSA HAIs, identified from microbiology reports in the VA electronic medical record, were defined as positive clinical cultures for MRSA between 48 hours after admission and discharge.
The cohort consisted of 121,520 patients (0.08% of whom had an MRSA HAI on the 1st day of a calendar month) in our improved analysis and 386,794 patients (1.04% of whom had an MRSA HAI) in our conventional analysis. In our improved analysis, estimates of the increase in inpatient costs due to MRSA HAI were $12,272 (p<0.001) and $23,733 (p<0.001) for variable and total costs, respectively. In our conventional analysis, estimates of the increase in inpatient costs due to MRSA HAI were more than 40% greater ($18,003 (p<0.001) and $33,885 (p<0.001) for variable and total costs, respectively).
This is the first study to account for time-dependent bias in the estimation of incremental per-patient healthcare costs attributable to HAI. We found that failure to account for this bias can lead to substantially inflated estimates.
R. E. Nelson,
C. F. Liu, None
M. Samore, None
M. Evans, None
N. Graves, None
B. Lee, None
M. Rubin, None