DNR Orders Known at the Time of Admission Can Improve Hospital Mortality Ratings

Abstract

Research ObjectiveThe Center for Medicare and Medicaid Services (CMS) publishes measures of risk‐adjusted hospital mortality rates that are meant to guide consumers and to contribute to Value‐Based Payment (VBP) quality ratings used to redistribute up to $2 billion in Medicare payments annually. A hospital’s actual to expected (A/E) mortality ratio accounts for 25% of its VPB rating. Current risk adjustment for mortality rates includes previously coded comorbidities and complications (CCs) but does not include indicators of higher mortality risk such as Do Not Resuscitate (DNR) orders present on admission (POA). We examined whether hospital mortality ratings could be improved by allowing DNR orders that are POA to identify a high‐risk group of patients among the 3 CMS mortality cohorts: acute myocardial infarction (AMI), heart failure (HF), and pneumonia (PN).Study DesignWe created a risk‐adjusted ranking of hospitals with and without admissions that had a DNR code POA using the All‐Patient Refined Diagnosis Risk Group Risk of Mortality (APR DRG RoM) at admission for the VBP cohorts. We computed the ratio of deaths to admissions for each DRG and ROM level for the 3 VBP cohorts as well as for all hospital admissions for the year. These ratios provided an expected rate of mortality for every admission through indirect rate standardization. For the 3 mortality rate cohorts, we calculated independent A/E mortality ratios for each hospital using existing CMS cohort definitions. We then repeated the calculation excluding admissions where DNR was reported as POA. We ranked the hospitals by A/E mortality rates and examined the change in decile ranks with and without DNR POA cases.Population StudiedWe used data for calendar year 2016 (CY2016) from the CMS Fee‐for‐Service (FFS) claims file for diagnosis codes and DNR status and used the Medicare Master Beneficiary Summary File (MBSF) to identify all deaths within 30 days of hospital admission. Among a total hospitalized population of 8.21 million with a death rate of 8.2%, there were 1.032 million admissions from the 3 VBP cohorts, of whom 14.2% had a DNR order POA and 11.3% died.Principal FindingsUnsurprisingly, the presence of a DNR order POA was associated with a much higher risk of 30‐day mortality. Overall, death rate for the mortality cohorts with DNR POA was 29.7%, compared to 8.2% without. For the entire hospitalized population, death rate was 30.3% with a DNR order POA, and 5.8% without. With the removal of DNR POA patients, hospital A/E mortality rankings changed substantially. 70% of hospitals moved up or down by at least one decile, and 34% moved up or down by 2 or more deciles.ConclusionsThe DNR POA order identifies patients at higher risk of death and can thereby alter VBP hospital mortality rate rankings and the resulting payment redistributions.Implications for Policy or PracticeIf the purpose of scrutinizing mortality rates is to identify preventable deaths, then continuing to include patients whose goals of care include DNR status can distort the risk‐adjustment process. The CMS mortality model should include adjustments for patients with DNR POA orders or remove them from the analysis entirely.