Background: Numerous quality metrics for heart failure care now exist based on process (from clinical guidelines) or outcome (readmission and mortality). It is unclear, however, how valid these measures are at serving as indicators of hospital quality for patients with heart failure (HF). We sought to determine validity by comparing convergent validity (correlation among measures) and reliability (correlation over time).
Methods: We gathered data on several HF performance measures from the Veterans Affairs Health Care System from 2008-2013. Process data are collected through chart review of a random sample of facilities. There are between 114 and 119 facilities that have at least one patient that is a treatment candidate each year. There are a mean 72 patients per facility per year included. For each facility and each year, we determined, among ideal candidates, use of ACE-inhibitor (ACEi), any beta-blocker (BB), recommended beta-blocker (rBB) (carvedilol, metoprolol succinate, bisoprolol), mineralocorticoid receptor antagonist (MRA), hydralazine/isosorbide dinitrate in African Americans, as well as rates of 30-day readmission, 30-day mortality, and 1-year mortality. We measured validity in 2 ways: determining correlations between measures, assuming that if a measure is associated with overall “quality” that it should be correlated with the other measures, and year-to-year reliability, assuming that if a measure has a good signal/noise ratio then there should be a strong correlation from one year to the next.
Results: Data were available for 55,735 hospitalizations for HF from 2008-2013. In determining correlations between performance measures, we found a statistically significant correlation between ACEi and rBB (R=0.25, p<0.001), ACE inhibitors and any BB (R=0.20, p<0.001), ACEi and 30-day mortality (R=-0.08, p<0.05), ACEi and 1-year mortality (R=-0.12, p<0.01), rBB and MRA (R=0.12, p<0.01), rBB and 30-day mortality (R=-0.11, p<0.01), rBB and 1-year mortality (R=-0.18, p<0.001), any BB and MRA (R=0.12, p<0.01), any BB and 30-day mortality (-0.08, p<0.05), and any BB and 1-year mortality (R=-0.13, p<0.001). Additionally, both rBB and any BB showed statistically significant year-to-year reliability (R=0.57, p<0.0001 and R=0.48, p<0.0001 respectively). MRA, 30-day mortality, and 1-year mortality also had similar reliability (R=0.50, R=0.42, R=0.41, respectively, p<0.0001), but not ACEi (R=0.04, p=0.71). Readmission within 30 days and hydralazine/nitrate were not reliable year-to-year nor did they correlate with any of the above performance measures.
Conclusion: BB, ACEi and MRA use rates showed evidence of validity as measures of HF quality through correlation with each other, mortality and/or year to year reliability. In contrast, 30-day readmission was poorly correlated with process of care and mortality and had poor reliability.