Medicare Advantage Health Care Data Brief #5 November 2016 Utilization: Hospital Readmissions KEY FINDINGS Trends in Medicare Advantage hospital readmissions Hospital readmission rates in the In this data brief, the Health Care Cost ter a previous hospital stay, e.g., a re- Medicare Advantage population Institute (HCCI) reports on inpatient admission, is considered, by many, to readmission trends in the Medicare be an indicator of the quality of care declined from 2010 through Advantage population, aged 65 years during the initial hospitalization as 2014. and older, for the period 2010 through well as any associated follow-up. More The average decrease in hospital read- 2014. specifically, higher readmission rates mission was 0.8% per year. Medicare Advantage (MA), formerly are associated with lower quality care known as Medicare Part C, is a health and as such hospital readmissions are insurance program available to indi- often reported as a quality measure. Cardiac related admissions, viduals eligible for fee-for-service For example, the Centers for Medicare heart failure and AMI, had the (FFS) coverage. The MA program al- & Medicaid Services (CMS) includes hospital readmission measures in the highest readmission rate and the lows individuals to obtain health in- surance coverage through commercial Hospital Compare database.4 The Af- largest decrease in rate, respec- health plans, in lieu of FFS coverage.1 fordable Care Act also instituted Medi- tively. care payment penalties for hospitals Compared to the under age 65 com- The heart failure readmission rate de- related to a hospital’s readmissions mercially insured and Medicare FFS clined 8.3% from 17.06 to 15.64. The rates.5 populations, the MA population has AMI readmission rate decreased 8.9% been studied less and the cost and uti- In this brief, we report on five read- from 12.41 to 11.30. lization experience of MA enrollees is mission rate measures that are com- not as well-documented. However, the monly used to study the Medicare FFS MA population accounts for a substan- population: 30-day all-cause hospital- odology, a 30-day all-cause readmission in- tial amount of the Medicare eligible wide readmissions and 30-day all- cludes readmissions within 30 days, regard- population and that share is growing. cause readmissions following acute less of whether the reason for the hospitali- Over the past decade, the proportion myocardial infarction (AMI), heart zation is related to the initial hospitalization. of individuals enrolled in MA plans has failure, chronic obstructive pulmonary (See the “Data and methods” section for nearly doubled. In 2006, MA plans disease (COPD), and pneumonia. more details regarding identifying index ad- covered approximately 16% of the For each readmission measure, the missions and readmissions.) Medicare eligible population; as of initial admissions or “index admis- Readmission rates varied across the five 2015, over 31% were enrolled in an sions” were first identified by mem- measures studied MA plan.2 ber. Hospital-wide index admissions Table 1 presents readmission rates for the Limited research on the MA popula- included hospitalizations with any 65 and older MA population in the HCCI da- tion has been due, in part, to limited diagnosis except those with a primary ta. The readmission rates in Table 1 express availability of MA data. The HCCI data diagnosis for psychosis, rehabilitation, the number of all-cause readmissions within set, used here, provides a rich data or medical treatment of cancer. Index 30 days of an index admission for every 100 source for research on the MA popula- admissions for the conduction specific index admissions. For example, in 2014, tion. These data include health insur- measures (AMI, heart failure, COPD, there were 11.71 readmissions for every 100 ance membership and claims from MA and pneumonia) only included hospi- hospital-wide index admissions. In other enrollees in all 50 states and the Dis- talizations with condition specific di- words, the percent of overall hospital admis- trict of Columbia. Over the study peri- agnosis codes. sions followed by a readmission for any od, the HCCI data accounted for ap- For each member with one or more cause within 30 days was 11.71%. The con- proximately 25% of the total MA pop- index hospitalizations, a hospital ad- dition specific measures are interpreted in ulation.3 This data brief is the first mission within 30 days of the index the same way. For example, in 2014 the analysis of the MA population by HCCI admission was identified as a read- number of hospital admissions for an AMI using these data. mission. Consistent with common diagnosis followed by a readmission for any Hospital readmissions methodologies for identifying read- cause within 30 days was 11.30%. missions, including Medicare’s meth- Over the study period, 2010 through 2014, An admission to a hospital shortly af- www.healthcostinstitute.org 1 the average hospital-wide, AMI, and There was noticeable variability be- hospitals. Hospital-wide readmissions pneumonia readmission rates were sim- tween measures in both the annual index admissions were not limited to ilar, 12 readmissions per 100 index ad- changes in readmission rates in a given specific conditions but did not include missions. The average COPD and heart year as well as over time for a given admissions for medical treatment of failure readmission rates, however, measure. For example, from 2012 to cancer or admissions with a primary were higher, 13.72 and 16.64, respec- 2013, the largest decreases in readmis- rehabilitation or psychiatric diagnosis. tively. sion rates were in the hospital-wide and The index admissions for the condition Although there was variation in all of AMI measures, both decreased 3.4%. specific measures were identified by the readmission rates over time, the Over the same period, the heart failure condition-specific ICD-9 codes on the magnitude of the variation differed by readmission rate only decreased 0.8% inpatient claims. measure. The pneumonia readmission and the COPD rate increased 4.8%. The There were two other exclusion criteria rate had the smallest difference between following year, COPD had the largest applied to hospitalizations identified as the highest and lowest rate, 0.67. The decrease, 6.3%; the decrease in the possible index admissions for both the difference for the hospital-wide meas- heart failure readmission rate was 5.7%; both the hospital-wide and condition ure was 0.85. This implies a difference and pneumonia had the smallest de- specific. First, index admissions exclud- of less than 1 readmission per 100 index crease of 0.5%. ed hospitalizations with a discharge sta- admissions. However, AMI, heart failure, Conclusion tus code of discharged against medical and COPD had differences between the The results presented in this brief show advice or death. Second, if a patient was highest and lowest readmission rates of that for a large, national MA population, admitted to one hospital then trans- 1.57, 1.55, and 1.25, respectively. Com- ages 65 and older for the years 2010 ferred to another hospital only the hos- pared to the hospital-wide rate, there through 2014, 30-day all-cause hospital pitalization at the second hospital is were almost twice as many readmis- readmission rates declined. The declines counted as an index admission. sions for these three conditions, on av- were observed for both hospital-wide After identifying index admissions, all erage. readmissions and for four common con- hospitalizations within 30 days of an Readmission rates decreased from dition specific types of hospitalizations index admission for that member were 2010 through 2014 (AMI, heart failure, COPD, and pneumo- identified as possible readmissions. The Figure 1 depicts the trends in readmis- nia). The trends are generally consistent readmission rates were calculated as the sion rates over time. Over the five-year with trends reported elsewhere for total readmissions divided by the total study period, all five readmission rates Medicare and commercially insured index admissions for the entire popula- declined. The largest overall decrease populations.6 Furthermore, there was tion. The ratio of readmission to index was among AMI readmissions (-8.9%). substantial variation across the admissions was multiplied by 100 in Heart failure readmission rates declined measures, both in the readmission rates order to report a rate in terms of the 8.3% over the same period. Although all and in the changes in rates over time. number of readmissions per 100 index of the rates declined overall, there were Data and Methods admissions.8 increases in four of the five admission The analysis sample was limited to indi- Consistent with the CMS/Yale Center for rates in 2011 relative to 2010: hospital- viduals aged 65 years and older enrolled Outcomes Research methodology, both wide, AMI, heart failure, and COPD. Ad- in a MA plan in the HCCI national claims the hospital-wide and condition specific ditionally, the change in the COPD read- database. Inclusion in the study sample measures only count the first hospitali- mission rate varied more than the other also required at least one inpatient hos- zation within 30 days of an index admis- readmission rates with increases in pitalization with three or more months sion as a readmission. However, a read- 2011 and 2013. of MA plan membership prior to a hospi- mission could also be counted as an in- Table 2 presents the annual changes in talization and at least 30 days of mem- dex admission in the hospital-wide readmission rates for each measure and bership following a hospitalization. The measure. For the condition specific the average of the annual changes. Both analysis was limited to admissions to measures, the sets of index admissions the AMI and heart failure rates had aver- short-term general acute care hospitals. and readmissions were mutually exclu- age annual decreases of more than sive. Additionally, hospital-wide read- The readmission rate reported was 2.0%. The average annual changes in missions excluded hospitalizations for based on methodology developed by hospital-wide and COPD readmission planned procedures and same-day read- CMS and the Yale Center for Outcomes rates were less than 1.0%. The pneumo- missions to the same hospital for the Research.7 The calculation required first nia readmission rate was the only rate same condition; AMI readmissions ex- identifying the set of all initial hospitali- that decreased each year; however, the cluded planned readmissions for revas- zations, i.e. the index admissions, for all average annual decrease in the pneumo- cularization procedures. MA members. Index admissions were nia readmission rate was only 1.3%. limited to hospitalizations with an over- Limitations night stay at general acute care inpatient Although, this report provides new sta- www.healthcostinstitute.org 2 tistics on the utilization of health care 4. Centers for Medicare and Medicaid within the MA population, the results Services. “Hospital Compare,” Medi- may not be generalizable to all MA plans care.gov. Web. or members. The HCCI data set is a con- 5. Centers for Medicare and Medicaid venience sample from three insurers Services. “Readmissions Reduction and may not reflect the utilization of the Program (HRRP),” CMS.gov. Web. full MA population or the trends of other 2016. MA payers. Additionally, the scope of 6. Fingar, K., Washington R. “Trends in this brief was intentionally limited to a Hospital Readmissions for Four small set of inpatient hospital related High-Volume Conditions, 2009- measures. There are many other 2013,” HCUP Statistical Brief #196. measures of health care services utiliza- Agency for Healthcare Research and tion and quality, which may provide Quality, Rockville, MD, November additional insight into the utilization 2015. patterns of the MA population. 7. The detailed methodologies are Finally, there were likely numerous fac- available from the Centers for Medi- tors influencing readmission rates dur- care and Medicaid Services. ing the study period but no causal infer- “Measure Methodology,” CMS.gov. ences can be drawn from the results Web. 2016. presented in this brief. Moreover, the analyses did not attempt to identify any 8. The MA readmission rates reported causes underlying the observed rates or in this brief are not directly compa- trends. Moreover, the analyses did not rable to all publically available read- evaluate the impact of any particular mission rates. Many readmission federal, state, or insurer policies or initi- rates, including some reported by atives intended to reduce readmissions. CMS, are risk-adjusted. The rates reported in this brief were not risk Endnotes: adjusted. 1. Centers for Medicare and Medicaid Services. “Medicare Advantage Plans,” Medicare.gov. Web. 2. Jacobson G. et al. “Medicare Ad- vantage 2015 Spotlight: Enrollment Market Update,” Washington DC: Henry J. Kaiser Family Foundation, June 2010. 3. Average HCCI annual membership was calculated as the sum of mem- ber months in year divided by 12. Authors Copyright 2016 Eric Barrette and Katharine McGraves-Lloyd Health Care Cost Institute, Inc. Unless explicitly noted, the content of this report is licensed Contact under a Creative Commons Attribution Non-Commercial No Eric Barrette Derivatives 4.0 License ebarrette@healthcostinstitute.org 571-257-1584 Health Care Cost Institute, Inc. 1100 G Street NW, Suite 600 Washington, DC 20005 202-803-5200 Acknowledgements This HCCI research product was independently initiated by HCCI and is part of the HCCI research agenda. www.healthcostinstitute.org 3 Table 1. Thirty-day, all-cause hospital readmissions per 100 index admissions Hospital-wide AMI Heart failure COPD Pneumonia 2010 12.09 12.41 17.06 13.63 12.58 2011 12.56 12.87 17.19 14.40 12.48 2012 12.41 12.02 16.73 13.40 12.30 2013 11.99 11.61 16.58 14.04 11.97 2014 11.71 11.30 15.64 13.15 11.91 5 year average 12.15 12.04 16.64 13.72 12.25 Source: HCCI, 2016. www.healthcostinstitute.org 4 Figure 1. Thirty-day, all cause hospital readmissions per 100 index admissions Source: HCCI, 2016. www.healthcostinstitute.org 5 Table 2. Annual percentage change in thirty-day, all-cause hospital readmissions per 100 index admissions Hospital-wide AMI Heart failure COPD Pneumonia 2010 to 2011 3.9% 3.7% 0.8% 5.6% -0.7% 2011 to 2012 -1.2% -6.6% -2.7% -6.9% -1.4% 2012 to 2013 -3.4% -3.4% -0.8% 4.8% -2.7% 2013 to 2014 -2.3% -2.7% -5.7% -6.3% -0.5% Average annual percentage -0.8% -2.3% -2.1% -0.7% -1.3% change Source: HCCI, 2016 www.healthcostinstitute.org 6