National and State Healthcare-associated Infections Standardized Infection Ratio Report
Using Data Reported to the National Healthcare Safety Network
This report presents data from HAI surveillance during calendar year 2011 that was reported either mandatorily or voluntarily to NHSN from facilities across all 50 states, Washington, D.C., and Puerto Rico. Data included in the report use NHSN definitions that have been in place since 2008 for CLABSI (6) and SSI (7) and 2009 for CAUTI (limited to symptomatic urinary tract infection) (8). These definitions differ slightly from those in use as of January 2013. Any data reported from non-acute care hospitals (e.g., long-term care hospitals, rehabilitation hospitals), outpatient dialysis facilities or inpatient dialysis wards, long term care facilities (e.g., skilled nursing facilities), and outpatient surgical settings were excluded from this report. Data include all reports submitted to NHSN as of September 4, 2012, allowing for a 9-month latency period to allow for complete reporting of infection events and denominator data through December 2011.
Similar to previous reports, the HAI data are summarized across all patient care location types and also stratified into three mutually exclusive categories, by state: critical care units (ICUs), wards (for this report, wards also include step-down and specialty care areas [including hematology/oncology and bone marrow transplant]), and neonatal intensive care units (including Level II/III and Level III). Active efforts by CDC and healthcare facilities reporting to NHSN began in 2011 to more accurately categorize long-term care, long-term acute care, and rehabilitation patient care locations that reside within acute care hospitals; these locations were excluded from this report. Future reports will include these patient care locations and reflect more accurate categorization. Summary statistics of reporting characteristics are presented both nationally and by state for each HAI included in the report. Data external to NHSN were used to construct some of these metrics. To approximate the number of acute care hospitals in each state, CDC used a list of all facilities that have been assigned a CMS Certification Number (CCN), adjusted to account for multiple facilities reporting under the same CCN and to include military and Veterans Affairs hospitals. Additionally, CDC consulted with each state health department to confirm the presence of any mandatory requirements for reporting HAI data to NHSN during 2010 and 2011 and to assess whether or not the health department has performed any internal or external validation studies of NHSN data that they have access to. Validation included data quality assessment for implausible values and detection of outlier facilities (e.g., high or low reported number of infections, rates, denominators) along with more detailed evaluation by health department staff with specific facilities and/or audits of medical records. The SSI data included in this report include only the more commonly reported operative procedures and approximates those targeted for process-of-care improvements by the Surgical Care Improvement Project (SCIP), a national project led by CMS and CMS-funded Quality Improvement Organizations (Appendix A). SSI standardized infection ratios (SIRs) are reported for these procedure categories combined, as well as for each specific procedure category. Only deep incisional and organ/space infections at the primary surgical site detected during the index hospital admission or upon readmission to the same hospital are included in the reported SIR data; superficial incisional SSIs and any SSIs identified through post-discharge surveillance were excluded from the SIR but included in the burden estimates (see below).
Summary HAI Data and Calculation of SIRs
The referent period for this report remains January 2006 through December 2008 for CLABSI and SSI and calendar year 2009 for CAUTI, as in previous SIR reports (2,9). The CLABSI and CAUTI SIRs presented in this report represent comparisons of an observed number of HAIs during each reporting period to the predicted number based on the rates of infections among all facilities during the referent period, adjusting for key covariates (10). Although over 40 patient care location types are included in the referent period, facilities have reported from location types not included in the referent period during 2010 and 2011. In such cases, the CLABSI and CAUTI SIRs in this report cannot include data from these newer location types.
The covariates used to predict CLABSIs and CAUTIs included type of patient care location, bed size of the patient care location, and hospital affiliation with a medical school. For NICUs, the pooled mean umbilical catheter-associated BSI rates and CLABSI infection rates within each birth weight category were used to predict the number of device-associated BSIs from reporting facilities, referred to as CLABSIs for this report. Clinical sepsis (without laboratory-confirmed bloodstream infection) was not included in the calculations of CLABSI during either the reporting periods or referent period. CAUTIs from NICUs are not reported to NHSN. For SSI SIRs, specific risk models were constructed that evaluated all available procedure-level risk factors (e.g., duration of surgery, surgical wound class, use of endoscopes, patient age, and patient assessment at time of anesthesiology [ASA score], among others) to predict the risk of deep incisional or organ/space infections identified during admission or readmission to the same hospital (3).
For national and state SIRs, all eligible data were included and the total number of infections predicted was compared to the number of infections reported to NHSN at each level of aggregation. In state-specific reporting of CLABSI SIRs, an SIR is only produced if at least 5 facilities in a state reported any data for the given location category. Facility-specific SIRs were also calculated for each of the summary measures presented nationally. However, if a single facility’s predicted number of infections for a specific HAI type (e.g., CLABSI) was ‹1.0, a facility-specific SIR was not calculated for that HAI. Distributions of facility-specific SIRs in national and state reports were produced only if at least 20 facilities had at least one predicted infection for a given HAI type. Additionally, summary counts of facility-specific SIRs were produced at the national level. The number of facilities that reported significantly fewer infections than what would be predicted and the number of facilities that reported significantly more infections than what would be predicted are shown for each location type and surgical procedure category.
An SIR that has a confidence interval (CI) that includes 1.0 should be interpreted as indicating that the number of HAIs that an entity (e.g., healthcare facility, state) observed and reported to NHSN is no different than if its experience had been the same as the referent population. The CI around the SIR depends on several factors, including the number of facilities reporting data from the relevant patient care location type or surgical procedure, the number of device days or surgical procedures reported, and the types of facilities reporting.
Serial Comparison of SIRs
Progress in preventing HAIs was evaluated by comparing the SIRs between 2010 and 2011. To fairly compare CLABSI and CAUTI SIRs, the patient care location rules used in this report (e.g., removal of data from long-term acute care and rehabilitation facilities) were applied to 2010 data and 2010 SIRs were recalculated. This evaluation was first accomplished by comparing the SIRs between each of the two sequential reporting periods for all data reported from all facilities. A second (sensitivity) analysis was then performed by restricting the facilities included to only those that reported during both 2010 and 2011, referred to as the change in SIR for continuously reporting facilities. A conditional binomial test was performed to assess for statistically significant changes in the pairs of sequential SIRs for each level of aggregation (two-sided P-value less than or equal to 0.05). Because this report uses all data reported to NHSN before September 4, 2012, calculations of 2010 and 2011 SIRs will differ slightly from reports using datasets created earlier in time, including those created by individual state health departments for public reporting.
National Disease Burden Estimates
The calculation of national estimates of the number of CLABSIs in hospitalized critical care patients involved several data sources and steps. CMS Hospital Cost Reports from 1990 through 2009 were used to obtain patient-days specifically occurring in critical care units in all Medicare-certified US hospitals (11), stratified by major hospital types: small (‹200 beds) teaching, medium (201-500 beds) teaching, large (›500 beds) teaching, small non-teaching, medium non-teaching, and large non-teaching. Because Federal hospitals do not file Hospital Cost Reports with CMS, we inflated patient-day estimates by between 5% and 10%, based on a weighted estimate of the annual ratio of all patient-days to non-Federal patient-days reported to the American Hospital Association (12). Based on historic secular trends we used linear regression to project critical care patient-days to 2011 (2009 was the most recent complete data year) and to generate standard errors around annual patient-day estimates for these six acute care hospital types.
To apply overall critical care CLABSI rates to these denominators, we constructed a negative binomial model for each hospital type based on data reported to NHSN from critical care units and generated estimated critical care CLABSI rates (per 1,000 patient-days) for 2011. To address differences between the types of hospitals reporting to NHSN and all hospitals nationally, an average of the six predicted CLABSI rates was calculated for 2011, weighted by the estimated number of national critical care patient-days occurring in each of the six hospital types (i.e., the rates were standardized to the estimated national distribution of critical care patient-days by hospital type).
The total number of CLABSIs in 2011 was calculated by applying estimated CLABSI rates to the estimated number of critical care patient-days nationally for 2011. We used Monte Carlo simulation to quantify the uncertainty around these estimates. Input distributions were created using predicted values and standard errors from the linear models (patient-days and federal inflation factor) and negative binomial models (CLABSI rates) described above. We sampled values from each of the input distributions in 10,000 simulation cycles and used the sampled values to calculate a CLABSI estimate for each cycle. We calculated 95% credible intervals based on the 2.5th and 97.5th percentiles of all output distributions. Analyses were conducted using SAS version 9.1 (©2002-2010, SAS Institute Inc., Cary, NC) and @Risk for Excel version 5.7 (©2010, Palisade Corp., Ithaca, NY).
Estimating SSIs for the U.S. in 2011 was performed using the procedure-specific crude infection rates for both deep incisional and organ/space infections as well as superficial infections and included infections detected after discharge among the SCIP procedures. These rates were extrapolated to the entire United States using estimates of the total number of procedures performed from the 2010 National Inpatient Sample (NIS), and adjusted to account for federal facilities performing procedures but not represented in the NIS.
Attributable Medicare Reimbursement for CLABSIs
Confirmed CLABSI cases from eight states reporting to NHSN were linked to claim records in the Medicare Provider Analysis and Review (MedPAR) database using hospital admission date, date of birth, sex, and facility. For both data sources, we limited the population to those over the age of 64 with a valid date of admission from January 2008 through December 2009, a valid date of birth, sex, and facility. In the MedPAR file, patients were also limited to those who aged into the cohort with or without end stage renal disease, enrolled in Medicare Part A and B throughout their eligibility, and never enrolled in a Medicare Advantage (HMO) program. Facility locations between NHSN and MedPAR were linked using an algorithm that matched data from the NHSN facility file and the CMS Cost Reports from 2004-2009. To link, first, the frequency of combinations of admission date, date of birth, sex, and facility was determined for each data source. If a particular combination occurred more than once in either data source, those observations would no longer be considered for linking. Once each data source contained a unique set of records based on those combinations of variables, the two data sources were linked through those variables. Only exact matches were included.
Using this linked dataset, we performed a retrospective cohort study comparing hospitalized patients who had a CLABSI to patients who did not. The primary outcome was Medicare reimbursement for the hospitalization. Frequency matching and multivariate regression were employed to control for potential confounders. For this analysis, five non-CLABSI control stays were selected such that the frequency of primary ICD-9-CM procedure category, which we found to be a valid predictor of length of stay, and ICU care were similar between CLABSI stays and non-CLABSI stays. The reimbursement attributable to CLABSI was estimated as the difference in medians between exposed and unexposed using multivariate median regression. Multivariate models included terms for age, race, sex, morbidity score, number of secondary procedures prior to infection, CMS wage index, CMS case mix index, facility bed size, teaching status, and number of critical care beds. Presence of an ICD-9-CM procedure code for insertion of a central line was an additional term in the CLABSI model.