Currently Viewing:
The American Journal of Managed Care February 2012
Nurse-Run, Telephone-Based Outreach to Improve Lipids in People With Diabetes
Henry H. Fischer, MD; Sheri L. Eisert, PhD; Rachel M. Everhart, MS; Michael J. Durfee, MSPH; Susan L. Moore, MSPH; Stanley Soria, RN; Diana I. Stell, RN; Cecilia M. Rice-Peterson, RN, BSN; Thomas D. MacKenzie, MD, MSPH; and Raymond O. Estacio, MD
TO THE EDITORS:
Daniel O. Scharfstein, ScD; Cynthia M. Boyd, MD, MPH; Jennifer L. Wolff, PhD; and Chad Boult, MD, MPH, MBA
A Dementia Care Management Intervention: Which Components Improve Quality?
Joshua Chodosh, MD, MSHS; Marjorie L. Pearson, PhD, MSHS; Karen I. Connor, PhD, RN, MBA; Stefanie D. Vassar, MS; Marwa Kaisey, BS; Martin L. Lee, PhD; and Barbara G. Vickrey, MD, MPH
Currently Reading
Hospital Readmission Rates in Medicare Advantage Plans
Jeff Lemieux, MA; Cary Sennett, MD; Ray Wang, MS; Teresa Mulligan, MHSA; and Jon Bumbaugh, MA
Identifying Patients With Osteoporosis or at Risk for Osteoporotic Fractures
Yong Chen, MD, PhD; Leslie R. Harrold, MD, MPH; Robert A. Yood, MD; Terry S. Field, DSc; and Becky A. Briesacher, PhD
Care by Cell Phone: Text Messaging for Chronic Disease Management
Henry H. Fischer, MD; Susan L. Moore, MSPH; David Ginosar, MD; Arthur J. Davidson, MD, MSPH; Cecilia M. Rice-Peterson, RN, BSN; Michael J. Durfee, MSPH; Thomas D. MacKenzie, MD, MSPH; Raymond O. Estacio, MD; and Andrew W. Steele, MD, MPH, MSc
Systematic Review of the Impact of Worksite Wellness Programs
Karen Chan Osilla, PhD; Kristin Van Busum, MPA; Christopher Schnyer, MPP; Jody Wozar Larkin, BSN, MLIS; Christine Eibner, PhD; and Soeren Mattke, MD, DSc
Adaptation and Psychometric Properties of the PACIC Short Form
Katja Goetz, PhD; Tobias Freund, MD; Jochen Gensichen, MD, MA, MPH; Antje Miksch, MD; Joachim Szecsenyi, MD, MSc; and Jost Steinhaeuser, MD
EHRs in Primary Care Practices: Benefits, Challenges, and Successful Strategies
Debora Goetz Goldberg, PhD, MHA, MBA; Anton J. Kuzel, MD, MHPE; Lisa Bo Feng, MPH; Jonathan P. DeShazo, PhD, MPH; and Linda E. Love, LCSW, MA

Hospital Readmission Rates in Medicare Advantage Plans

Jeff Lemieux, MA; Cary Sennett, MD; Ray Wang, MS; Teresa Mulligan, MHSA; and Jon Bumbaugh, MA
Hospital readmission rates in Medicare Advantage plans are considerably lower than those in Medicare's traditional fee-for-service program, after accounting for differences in risk.
The current risk adjustment system used by CMS for reimbursement of MA plans provides a general measure of patients’ health costs, but it does not directly measure the probability of a readmission. Medicare’s risk-adjustment system is now based more on diagnosis codes for major illnesses (eg, diabetes, heart disease) than demographic indicators (eg, age, sex). However, before the diagnosis-based risk adjustments were phased in, MA plans had an incentive to attempt to enroll healthierthan-average patients within a demographic cohort, and the impact of this incentive may persist.

However, there is less reason to believe that hospitalized patients in MA plans—those who are actually at risk for readmission— are healthier than hospitalized FFS patients. In fact, given the incentives created by capitated reimbursements, MA plans have reason to attempt to treat all but their sickest patients out of the expensive inpatient hospital setting. (We observed in this and prior studies that MA enrollees had lower overall hospitalization rates than FFS patients. This is likely due in part to enrollment of healthier-than-average benefi - ciaries, but it is also due to efforts designed to help at-risk patients avoid preventable hospitalizations in the fi rst place.) Thus, there may not necessarily be a reason to assume that hospitalized MA patients would have lower readmission rates than hospitalized FFS patients. In fact, the opposite assumption—that hospitalized MA patients could be less healthy than hospitalized FFS patients—is not implausible. Nevertheless, our analysis indicates that for the purpose of analyzing readmission rates, there are important differences in the types of admissions that should be taken into account when comparing readmission rates.

Therefore, we developed a risk-adjustment measure directly associated with readmission risk among hospitalized patients. The method is straightforward: fi rst, we computed the probability that each admission DRG would be associated with readmissions in the FFS population. Then we compared the distribution of those DRGs in the MA data against this baseline. If the FFS population had more admission codes with higher-than-average readmission rates, then an adjustment could be computed based on the admission distribution and each DRG’s likelihood of readmission. In eAppendix C (available at www.ajmc.com), we discuss this method in detail.

RESULTS AND ANALYSIS

Table 1 shows unadjusted benchmark readmission rates found in the MA plans for the 2006-2008 period alongside the Jencks et al results for 2004 FFS.1 The MA 30-day readmission rates in 2006, 2007, and 2008 were remarkably consistent across each of the years, with an overall 30-day readmission rate of 14.5%, just over 25% lower than the FFS rate of 19.6%. The 60-day and 90-day readmission rates in the MA plans were similarly consistent over the 3-year period and were also about 25% lower than the FFS rates measured in the Jencks et al study. Table 2 shows the same breakouts of readmission rates for medical and surgical index admissions of various types as were shown in the Jencks et al study for FFS.

A secondary goal of this report was to address the comparability of the benchmark MA and FFS readmission rates. We discuss the effects on comparability of the different periods measured, the presence of disabled enrollees under age 65 years and enrollees 90 years and older, the geographic location of enrollees in the samples, and the distributions of DRGs associated with MA and FFS enrollees’ admissions and whether those admissions would naturally tend to have higher or lower readmission rates.

Contemporaneous Comparisons in 2006 to 2008

Our benchmark for MA readmission rates is the 2006-2008 period, a few years after the 2004 Jencks et al calculations for FFS.1 To address the effect on comparability of possible changes in readmission rates over time, we used an alternative readmission rate calculation method and applied it to both the MA and FFS populations during the 2006-2008 period. The alternative calculations for FFS were made available to us by Dr Gerard Anderson of Johns Hopkins University; we then duplicated Anderson’s calculation method for the MA sample. Unlike the Jencks et al method, which tracks readmission from the fourth quarter of the prior year, the Anderson method tracks index admissions occurring in the fi rst 9 months of each year and captures readmissions in the subsequent 30-, 60-, and 90-day periods following each index admission. Like the Jencks et al method, the Anderson method used an all-cause or any-DRG readmission calculation, also excluding admissions for rehabilitation.

Under this alternative method, the FFS 30-day readmission rate for patients 65 years and older in the 2006-2008 period was 18.4%, which compares to the 19.6% rate estimated by Jencks et al for 2004.1 The MA result was 14.4% in 2006- 2008, virtually unchanged from the Jencks-style calculation of 14.5%, presented above, for the same period. The contemporaneous gap between FFS and MA was lowered to 22% by this approach.

These alternative calculations were done only for patients 65 years and older, to address possible objections that differences in the numbers of disabled patients younger than 65 years could affect the comparisons. Because there are 3 moving parts in the alternative calculation—the time period (2006-2008 for FFS), the readmission calculation method (first 3 quarters of the year with 3 months of run-out), and the population (no patients younger than 65 years)—we were not able to precisely compute the degree of change in the FFS results attributable to each. The lowering of the FFS readmission rate by this alternative Anderson-method computation for 2006-2008 was likely due in part to the exclusion of disabled patients younger than 65 years and partly due to a lower readmission rate in FFS over time. Based on 2008 data, we estimated that FFS readmission rates for the under-65 population were approximately 20% higher than those for FFS patients 65 years and older. Moreover, roughly 20% of FFS enrollees were under age 65 years in the 2006-2008 period, and these enrollees, because of their disabilities, were more likely to be hospitalized and thus at risk for readmission. (For Medicare elderly versus disabled readmission rates, see Wier et al.14) Therefore, the exclusion of the under-65 population should reduce the computed FFS readmission rate. Finally, we computed readmission rates for FFS for each year during the 2004-2008 period (using yet another method for technical reasons). For this calculation, we used the same-quarter readmission rate for each year from 2004 to 2008. This is simply the number of admissions in a calendar quarter minus 1 (for the index admission) divided by the number of admissions in the quarter. According to the industry studies of readmission rates referred to earlier in this report, the same-quarter readmission rate is a reasonable proxy for a 30-day readmission rate, at least for comparative purposes. By this measure, the FFS readmission rate was about 0.6 percentage points lower in 2008 than in 2004 (22.9% in 2004; 22.6% in 2005; 22.4% in 2006; 22.3% in 2007; and 22.3% in 2008). Thus, the change in time periods could account for nearly half of the lower FFS readmission rate found by this alternative calculation. There was no discernible time trend in readmission rates in the MedAssurant data from 2006 to 2008, using either the Jencks et al method or the alternative Anderson-method calculations.

However, the goal was simply to provide a reasonable contemporaneous comparison between FFS and MA results, and the alternative calculations provided a platform for additional adjustments for comparability. A new report by the Dartmouth Atlas Project found only small changes in various readmission rates among Medicare FFS patients during the 2004-2009 period.15

Adjustment for Risk of Readmission

Table 3 provides a simple illustration of a risk-adjustment method based on indexes of readmission risk for version 24 and version 26 DRG distributions. (DRG defi nitions can change annually, sometimes substantially. The version 26 Medicare Severity DRGs [MS-DRGs] are much more detailed than the version 24 DRGs. Version 25 DRGs were not used due to overlap with their use in the MA data set.) In general, the DRG-based method of risk adjustment lowers the percent reduction in readmission rates in MA from about 22% (unadjusted, contemporaneous 2006-2008 comparisons excluding patients under age 65 years) to a range of about 13% to 20%, depending on the DRG version used. (In theory, comparisons based on the expanded version 26 MS-DRGs should produce a richer measure of readmission risk, because they incorporate the severity of a hospitalization to a greater degree than the version 24 DRGs. Moreover, the version 24 distributions had some unusual results. In the version 24 DRGs, MA plans in the MedAssurant sample coded complicating conditions much less frequently than FFS for many disease categories. For example, the coding for simple pneumonia DRGs showed complication rates in most of the FFS cases but very few of the MA cases. We believe these coding differences are too extreme to be explained by actual patient complication rates, and such differences did not appear in the version 26 data. Therefore the risk adjustment based on the version 24 DRGs should be used with caution, and may represent an overly large adjustment.) Of course, it is possible that some inherent risks of readmission are not captured in this method, even using the more detailed version 26 DRGs, which include expanded information about the presence of complicating conditions. However, it is doubtful that alternative methods of risk adjustment would completely erase a 22% difference in unadjusted readmission rates.

Impact of Age Distribution

Table 4 shows the demographic distribution of patients with a hospitalization—those enrollees at risk for readmission— in the 2006-2008 MedAssurant MA sample and in FFS (from the Medicare 5% sample). The MA sample had a slightly higher share of patients in the 70- to 79-year age range than the FFS sample, and a slightly lower share of patients over age 85 years. However, based on a simple test—excluding patients 90 years or older, who are more common in FFS than MA—we believe that explicitly adjusting for age within the population 65 years and older is not likely to affect the comparisons. Excluding patients 90 years and older dropped the FFS readmission rate from 18.4% to 18.2% and dropped the MA readmission rate from 14.4% to 14.3%. Thus, the MA readmission rate remained virtually unchanged at 22% lower than FFS. Moreover, when we excluded patients over age 89 years, the risk adjustment indicated by the DRG distributions was slightly reduced. The net result for risk-adjusted FFS- to MA-comparisons was minimal. Table 5 summarizes the results of these comparisons and shows the corresponding industry data computed from state-based hospital discharge data sets. In general, the industry studies from these particular states were consistent with our results from the MedAssurant data set.

DISCUSSION

 
Copyright AJMC 2006-2020 Clinical Care Targeted Communications Group, LLC. All Rights Reserved.
x
Welcome the the new and improved AJMC.com, the premier managed market network. Tell us about yourself so that we can serve you better.
Sign Up