Investigating Systematic Misclassification of Central Line–Associated Bloodstream Infection (CLABSI) to Secondary Bloodstream Infection During Health Care–Associated Infection Reporting

Methods

Data from Pennsylvania acute care hospitals reported to the NHSN were used to calculate inpatient care location-specific CLABSI incidence rates per 1000 central line days (CLDs) and secondary BSI incidence rates per 1000 patient days (PTDs) for two 1-year periods (year 1, July 2008-June 2009; year 2, July 2009-June 2010). Inpatient care locations that reported both CLABSI and secondary BSI numerator and denominator data (CLD and PTD) throughout the 2-year period and those that reported ≥100 CLDs during year 1 and year 2 were included. Neonatal intensive care and specialty care area inpatient care locations were excluded from analysis.

The median and distribution of inpatient care location-specific rates were compared to identify changes in reported CLABSI and secondary BSI incidence for years 1 and 2. To determine if the change in secondary BSI rates from year 1 to year 2 differed for inpatient care locations that had either an increase or decrease in their CLABSI rates, they were grouped into 1 of 3 mutually exclusive categories: increase, decrease, or no change in CLABSI rate. The median and distribution of the change in secondary BSI rates for locations with an increase or decrease in their CLABSI rates were then compared with one another. Finally, the authors assessed if there was any correlation between the change in CLABSI rate and change in secondary BSI rate for locations from year 1 to year 2.

The median and distribution of rates were compared using the Wilcoxon rank sum and Kuiper tests, and correlation was assessed using the Spearman’s rank correlation coefficient (r_s) and 95% confidence interval (CI). All tests were 2 sided, and P values <.05 were considered statistically significant.

Results

Data from 975 patient care locations reported by 165 Pennsylvania hospitals were included in the analysis, including 625 (64%) inpatient wards, 264 (27%) critical care units, and 86 (9%) step-down units. From year 1 to year 2, there was a significant decrease in the median CLABSI rate and change in the distribution of CLABSI rates but no change in the median or distribution of secondary BSI rates (Table 1). The 3 major primary infection types responsible for secondary BSIs over the 2 years were surgical site infection (28%), catheter-associated urinary tract infection (16%), and gastrointestinal infection (15%). There was no change in primary infection types reported for secondary BSIs from year 1 to year 2 (χ² P = .624).

OPEN IN VIEWER

Table 1.

CLABSI and Secondary BSI Rates Reported for 975 Inpatient Care Locations by 165 Pennsylvania Hospitals to the National Healthcare Safety Network, July 2008 to June 2010

	Year 1 (July 2008-June 2009)	Year 2 (July 2009-June 2010)	P Value
Median (IQR) CLABSI rate per 1000 central line days	0.64 (0.00-1.87)	0.00 (0.00-1.22)	<.0001 a ; .0002 b
Median (IQR) secondary BSI rate per 1000 patient days	0.10 (0.00-0.31)	0.10 (0.00-0.29)	.9946 a ; .8209 b

Abbreviations: CLABSI, central line–associated bloodstream infection; BSI, bloodstream infection; IQR, interquartile range.

a

Wilcoxon rank sum test.

b

Kuiper test.

From year 1 to year 2, a total of 260 patient care locations had an increase in their reported CLABSI rate, 416 had a decrease, and 299 had no change. The median and distribution of the change in secondary BSI rates (Figure 1) from year 1 to year 2 were no different (Wilcoxon rank sum P = .6438; Kuiper P = .9216) when comparing locations that had an increase with those that had a decrease in CLABSI rates. Furthermore, no correlation was found between the change in CLABSI and change in secondary BSI rates from year 1 to year 2 (r _s = −0.01 [95% CI = −0.08 to 0.05; P = .6480]).

OPEN IN VIEWER

Figure 1.

Distribution of the change in secondary bloodstream infection (BSI) rates per 1000 patient days for locations that had an increase, decrease, or no change in central line–associated bloodstream infection (CLABSI) rate; data are reported for 975 inpatient care locations by 165 Pennsylvania hospitals to the National Healthcare Safety Network, July 2008 to June 2010

Discussion

Mandatory reporting of comprehensive HAI data from Pennsylvania provided a unique opportunity to evaluate 2 full years of both CLABSI and secondary BSI incidence rate data from a large number of hospitals and patient locations. We found that whereas CLABSI rates per 1000 CLD had decreased over the 2-year period, secondary BSI rates per 1000 PTD in the same locations over the same time period remained stable. The median and distribution of the change in secondary BSI rates did not differ for locations that experienced an increase or decrease in their CLABSI rates. Furthermore, we did not find evidence of a negative correlation between the change in CLABSI and secondary BSI rates from year 1 to year 2. That is, we found no evidence of an increase in secondary BSI rates to suggest that concurrent decreases in CLABSI rates were a result of misclassification of CLABSIs as secondary BSIs.

Although there may be inherent biases related to public reporting^7,8 and, thus, value in exploring whether reported trends are real or influenced by artifact, our study suggests that systematic misclassification of CLABSIs as secondary BSIs likely did not play a significant role in the observed decreases in CLABSI incidence nationally. Decreases in CLABSI incidence have been reported over the long term ¹ with contemporaneous, widespread adoption of effective CLABSI prevention measures likely contributing.^9,10

There are several limitations to consider. First, we did not include data from all patient locations in our analysis. Neonatal intensive care units and specialty care areas (eg, solid organ or bone marrow transplant and hematology/oncology) were excluded because of differences in the calculation of CLABSI rates for these locations (eg, stratification by birth weight categories and line type). Second, patient care locations with small numbers of CLDs (<100 per year) were excluded, although their total number of CLABSIs and secondary BSIs reported are unlikely to influence our overall findings. Third, although CLABSI reporting mandates are common, Pennsylvania is unique in that reporting of all HAIs in the NHSN Patient Safety Component is required. Therefore, there may be fewer incentives to misclassify CLABSIs as secondary BSIs in Pennsylvania hospitals. This is likely minimized because although the total number of BSIs is publicly disclosed at the facility level, only CLABSIs are used for facility-to-facility benchmarking. ⁴ Finally, this analysis is based on data grouped at the location level and not at the patient level. No individual patient medical records were reviewed as part of this ecological analysis of case misclassification. As with all ecological study designs, our results are subject to potential bias related to the ecological fallacy. ¹¹

The Pennsylvania Department of Health is actively engaged in validation of HAI data reported to NHSN, including BSIs. This includes manual review of selected patient charts to validate data reported by hospitals to NHSN. Such efforts are an integral part of the process of determining the accuracy and completeness of publicly reported data, in addition to providing opportunities for reporting hospitals to receive education and valuable feedback on the surveillance process. ¹

We found no evidence of concomitant increases in secondary BSI rates among Pennsylvania hospital locations that experienced a decrease in CLABSI rates over the 2-year period, suggesting that intentional CLABSI case misclassification is not widespread in this state. As decreases in CLABSI incidence within Pennsylvania are similar to those reported nationally,^3,4,6 these findings help build confidence that reporting trends in CLABSI incidence nationally are reflective of actual experience. However, additional efforts to validate data reported to NHSN are important to further evaluate the degree of case misclassification and generalize our findings.