Methods To Estimate the Baseline 2010 PFP National Hospital-Acquired Condition Rate

Prepared by Noel Eldridge.

This document describes the methods AHRQ used to estimate the national rate of hospital-acquired conditions (HACs) for the Partnership for Patients (PFP) program. The estimate includes a wide variety of adverse events, including the nine HACs selected for special focus as part of the PFP, as well as several other HACs. Collectively, 28 specific measures are used: 14 measures to generate rates for the 9 specific PFP HACs and 14 measures to generate a summary rate for "all other" HACs. Of the 28 measures, 21 are from the AHRQ/CMS Medicare Patient Safety Monitoring System (MPSMS), 6 are from the AHRQ Patient Safety Indicators (PSIs), and 1 is based on National Healthcare Safety Network (NHSN) data from CDC. The measures are shown in Table 1 below.

Table 1. Measures Used To Estimate the National Hospital-Acquired Condition Rate

Acronyms: Central Line-Associated Blood Stream Infection (CLABSI), Medicare Patient Safety Monitoring System (MPSMS), Patient Safety Indicators (PSIs), National Healthcare Safety Network (NHSN).

Data from these 28 measures are combined to generate a summary national HAC rate that can be calculated annually, and that uses data from the AHRQ Healthcare Cost and Utilization Project (HCUP) on the annual number of hospital inpatient discharges as the denominator for the rate calculation. The details of how the national HAC rate is calculated for PFP are summarized below.

1. The national rate of PFP HACs is calculated using the following variables:
   1. The rate of 21 MPSMS adverse events in the MPSMS sample provided by the CMS Inpatient Quality Reporting (IQR) sample of charts. The IQR sample for 2010 (the baseline year for the PFP national HAC rate) included approximately 34,000 charts from four patient groups: those included in the Surgical Care Improvement Project (SCIP) (major surgery) sample, and those with a principal diagnosis of acute myocardial infarction (AMI), congestive heart failure (CHF), or pneumonia. The IQR sample is limited to patients 18 years old and over. The rates of the 21 MPSMS adverse events are expressed as a rate for the whole sample population, rather than as a rate for the subpopulation that has the opportunity to experience the adverse event. For example, the CLABSI rate, like all the rates, uses all patients in the denominator, not just the patients who had a central line inserted during their hospital stay. To ensure that the estimated rate will not be affected by variation across years in the proportion of patients from each of the four groups (that is, SCIP, AMI, CHF, and pneumonia), for each of the 21 MPSMS HACs the estimate is a weighted average of the HAC rate for each group, where each group is given a 25% weight.
   2. The national number of adverse events captured by PSIs 18 and 19 on obstetric injury, and the national number of four other PSIs (numbers 6, 9, 11, and 15) included among the "all other" HACs.
   3. The national number of selected surgical site infections computed in a special calculation for the PFP by CDC. This calculation is based on NHSN data for 17 specific operations: the 12 "SCIP" operations and 5 others.
   4. The data described above for MPSMS are unlike the NHSN and PSI data in two ways. First, they originate as a rate, and second, they are an estimate for four specific subpopulations of hospital patients rather than all inpatients. In order to estimate MPSMS rates for all patients, rather than simply those with one of the four diagnoses included in the 2010–2012 IQR sample, we use data from the 2005-2006 MPSMS sample. Using the 2005–2006 MPSMS data, we estimate, for each of the 21 HACs included in the MPSMS analysis, the ratio of the rate of HACs for all patients in the sample to the rate of HACs for patients with the four diagnoses. As shown in Table 2, these ratios vary from .28 to 1.08, with 14 of 21 ratios being between .4 and .8 and the arithmetic mean of the ratios being .64. The fact that the ratios are, with one exception, all below 1.0 makes intuitive sense: it is credible that the rate of harm is greater for patients who have had major surgery, AMI, CHF, or pneumonia than it is for other patients.
   5. In order to estimate the rate of adverse events for each of the 21 HACs for all patients for which the MPSMS data are used, we multiply the adverse event rate for each of the 21 HACs for patients with one of the four conditions included in the 2010–2012 MPSMS sample as estimated in (a) above by the ratio of adverse events for all patients to adverse events for patients with one of the four diagnoses. Then, in order to estimate the number of adverse events for each of these 21 HACs, we multiply the estimated rate for all patients by the estimated total number of hospital discharges, where the estimate for the total number of hospital discharges comes from HCUP.
2. The estimated total count of annual HACs in the PFP national HAC rate is calculated as the sum of the total number of HACs for the 21 MPSMS HACs in (e) above, plus the estimated number of PSI HACs in (b), plus the number of NHSN-estimated HACs in (c).
3. The PFP national HAC rate is the total HAC count in (2) above divided by the number of annual hospital inpatient discharges of patients at least 18 years old, as provided by HCUP. Table 3 provides a summary of the data that shows how the 28 rates from all the measures are combined into the PFP national HAC rate.
4. As of April 2014, this method has been repeated for 2011, and for 2012 using preliminary data for the number of annual hospital discharges. To be able to compare the estimated total number of HACs in 2011 and 2012 to 2010 without having that comparison be affected by changes in the total number of discharges, for 2011 and 2012, the data were normalized to generate national counts based on 32,750,000 discharges, which was the total number of discharges in 2010 for inpatients 18 years old and over. This method will be used again to finalize estimates for 2012, 2013, and 2014 as the data become available.

Table 2. Summary of Ratio Calculations

Table 3. 2010 Estimated Final HAC Data (Finalized May 2012)

Ideally, we would have estimated the adverse event rates for the 21 MPSMS HACs using data from patients with all diagnoses, rather than being restricted to the four diagnoses that were included in the 2010 IQR sample. Because the 2010 IQR sample did not include patients with all diagnoses, we used the 2005 and 2006 MPSMS data, as described above, to estimate the ratio of adverse events for all diagnoses to adverse events for the four diagnoses. However, the 2005 and 2006 sample was restricted to Medicare patients age 65 and over, and it is possible that the ratios we estimated with those data are not correct for all patients. It makes sense to think, for example, that because many of the adverse events we measure are less common among obstetric patients than among other patients, the ratio for all patients would be lower than the ratio for Medicare patients 65 years old and over. It is also possible that the ratio has changed from 2005–2006 to the present.

New MPSMS data are now available based on an expanded 2012-2013 IQR sample that includes charts from patients with diagnoses other than those in the four diagnosis groups present in the 2010–2011 IQR sample. Use of these data may enable the development of a new method that could provide an improved estimate of the PFP national HAC rate. Information from these 2012 and 2013 charts is currently being analyzed to ensure the reliability of the MPSMS rates and other data present in charts from these additional IQR samples. Preliminary data and calculations based on data that include the new IQR samples suggest that a method that includes data from these additional charts would produce somewhat lower national HAC rates than have been computed to date, but not necessarily change the annual trends in the HAC rates.