Medicare Health Outcomes Survey

Survey Results


The Medicare Health Outcomes Survey (HOS) is an assessment of a health plan's ability to maintain or improve the physical and mental health functioning of its Medicare beneficiaries over a two-year period of time, using the best available science in functional status and health outcomes measurement. The survey is used as a way of measuring how the care provided by Medicare Advantage (MA) Plans is affecting the functional status of their enrollees.

The first cohort of baseline data was collected in 1998. Beginning in 2000, both a baseline cohort and a follow up cohort were collected. Most recently in 2007, Cohort 10 Baseline and Cohort 8 Follow Up data were collected. For further information, please refer to the survey administration timeline. The following table depicts general status information, including response rates, for the baseline and follow up cohorts administered to date.

Medicare HOS Survey Status Information1
COHORT DATE
FIELDED
POPULATION REPORTING UNITS (PLANS)2 SAMPLE SIZE INVALID SURVEYS3 COMPLETED SURVEYS4 RESPONSE RATE5
Cohort 1 Baseline May 1998 M+CO 287 279,135 7,196 167,096 61.4%
Cohort 1 Follow Up April 2000 M+CO 188 89,332 1,203 75,593 85.8%
Cohort 2 Baseline March 1999 M+CO
PACE
EverCare
312
20
6
301,184
4,225
5,015
8,829
435
655
194,378
1,752
1,226
66.5%
46.2%
28.1%
Cohort 2 Follow Up June 2001 M+CO 160 88,468 1,539 73,015 84.0%
Cohort 3 Baseline April 2000 M+CO
PACE
306
20
298,883
3,831
7,662
564
208,655
1,672
71.6%
51.2%
Cohort 3 Follow Up May 2002 M+CO 146 87,091 1,523 65,992 77.1%
Cohort 4 Baseline May 2001 M+CO
PACE
197
17
190,523
3,943
6,041
362
126,255
1,321
68.4%
36.9%
Cohort 4 Follow Up May 2003 M+CO 152 71,549 951 55,480 78.6%
Cohort 5 Baseline April 2002 M+CO
PACE
177
20
171,504
4,834
5,963
710
106,168
1,385
64.1%
33.6%
Cohort 5 Follow Up May 2004 MA 153 74,562 1,265 58,417 80.0%
Cohort 6 Baseline April 2003 M+CO 163 161,409 5,579 100,669 64.6%
Cohort 6 Follow Up May 2005 MA 154 69,865 854 56,075 81.3%
Cohort 7 Baseline March 2004 MA 161 159,311 6,777 99,649 65.3%
Cohort 7 Follow Up May 2006 MA 151 69,873 1,077 57,105 83.0%
Cohort 8 Baseline April 2005 MA 164 162,902 6,966 101,588 65.1%
Cohort 8 Follow Up May 2007 MA TBD TBD TBD TBD TBD
Cohort 9 Baseline April 2006 MA 203 188,515 7,725 120,708 66.8%
Cohort 10 Baseline April 2007 MA TBD TBD TBD TBD TBD


UPDATED August 2007

1 Over the course of time, the refinement of the HOS measure has resulted in changes to the definition of a "completed survey." The information in the table provides a presentation of how a completed survey and response rate were calculated at the time of reporting for each of the baseline and follow up cohorts. Given that the definition of a completed survey has evolved over time, this table should not be utilized for response rate comparisons across the cohorts. If you are interested in comparative response rate information, please contact the Medicare HOS Information and Technical Support Line.

2 For the baseline cohorts, the reporting units represent the individual plans sampled for the survey. However for the follow up cohorts, the reporting units have been adjusted to accommodate selected plan consolidations and service area reductions at the time of performance measurement reporting (which typically occurs in the year subsequent to the collection of the follow up data).

3 For the baseline cohorts, the invalid beneficiaries meet one of the following criteria: deceased; not enrolled in the MA Plan; have an incorrect address and phone number; or have a language barrier. For the follow up cohorts, the invalid beneficiaries meet one of the following criteria: not enrolled in the MA Plan; have an incorrect address and phone number; or have a language barrier.

4 In general, for Cohorts 1, 2, and 3 Baseline a completed survey is defined as a survey that has at least 80% of the questions answered. However, beginning with Cohort 1 Follow Up and Cohort 4 Baseline, a completed survey is defined as a survey that can be used to calculate physical or mental health summary scores.

5 Please note, invalid beneficiaries are removed from the denominator in the response rate calculations. In other words, Response Rate = [Number of Respondents/(Total Sample - Invalids)].

Performance Measurement Analytic Sample

After the administration of each follow up cohort, cohort specific performance measurement results are calculated. Seniors (age 65 or older) that had a calculatable physical component summary score or mental component summary score at baseline are eligible for performance measurement. However, some of these seniors belong to plans that went out of business or discontinued offering managed care between the baseline and follow up samples. Therefore, the Performance Measurement Analytic Sample is limited to those seniors who had calculatable physical or mental health summary scores at baseline and were still enrolled in the same participating MA Plan at the time of follow up. Seniors who become deceased between baseline and follow up are classified as "dead" for purposes of performance measurement. Additionally, a certain number of seniors will voluntarily disenroll from their plans between baseline and follow up. These seniors are classified as "voluntarily disenrolled" for purposes of performance measurement. Of the seniors sampled at the time of follow up, a certain percentage will be determined to be invalid for inclusion in the sample. These invalid seniors are: not enrolled in the MA Plan; have an incorrect address and phone number; or have a language barrier. Of the seniors eligible for inclusion in follow up, those that do not return a completed survey are designated as "non-respondents" and those that do return a completed survey are referred to as "respondents." The following table depicts the distribution of the Performance Measurement Analytic Sample presented in all performance measurement reports to date.

Medicare HOS Performance Measurement Distribution of Sample
COHORT YEARS REPORTING UNITS ANALYTIC SAMPLE VOLUNTARILY DISENROLLED DEAD FOLLOW UP INVALID SURVEYS FOLLOW UP NON-RESPONDENTS FOLLOW UP RESPONDENTS
Cohort 1 1998-2000 188 122,444 31,772 8,047 1,203 11,531 71,094
Cohort 2 1999-2001 160 124,835 33,226 8,930 1,179 12,758 68,742
Cohort 3 2000-2002 146 122,317 36,503 8,987 790 16,459 59,578
Cohort 4 2001-2003 152 95,565 24,589 6,998 392 12,950 50,636
Cohort 5 2002-2004 153 92,434 18,603 6,993 781 12,733 53,324
Cohort 6 2003-2005 154 90,154 13,550 6,739 854 12,936 56,075
Cohort 7 2004-2006 151 91,453 14,564 7,016 1,077 11,691 57,105

UPDATED August 2007

Calculating Performance Measurement Results

A performance measurement data set is created by merging a cohort's baseline and follow up data. Additionally, death information is incorporated into the performance measurement data set for those baseline respondents who died between baseline and follow up. The HOS performance measurement results are computed using a rigorous case mix/risk adjustment model.

There are six main categories of actual health outcomes used in the performance measurement analysis:
  1. alive and physical health better;
  2. alive and physical health the same;
  3. dead or physical health worse;
  4. mental health better;
  5. mental health the same; and
  6. mental health worse.

Each beneficiary is classified into one of the three physical health categories and one of the three mental health categories. In calculating expected outcomes, separate case mix models are warranted for death (which requires extensive case mix control), and for physical component summary scores and mental component summary scores (which do not require much case mix control). A series of eight different death models, three different physical health models, and three different mental health models are used, since all beneficiaries do not have data for all of the independent variables that could be used to calculate an expected score. In other words, each expected outcome for a beneficiary is derived from the best fit model, which is based on those variables for which the beneficiary has data. One model is used for each beneficiary, and there are no predictions made with missing data.

Beneficiary level results are aggregated to derive the plan, state, and HOS national percent better, same, and worse than expected values. Outliers are those plans that performed significantly better (i.e., better than expected) or significantly worse (i.e., worse than expected) when compared to the national average. The national average is based on all plans that participated in performance measurement. Plans can be outliers on a measure of physical health (which is based on death and the physical component summary score), or on a measure of mental health (which is based on the mental component summary score).

Analyses of the two-year performance measurement data have demonstrated that at the national level there is significant variation among Medicare health plans with respect to both physical and mental health outcomes. Research has identified difference in outcomes among specific groups of beneficiaries and potential opportunities to improve care. The following table depicts the overall performance measurement results by cohort.

Medicare HOS Performance Measurement Results
COHORT YEARS TOTAL NUMBER OF REPORTING UNITS MENTAL HEALTH BETTER THAN EXPECTED MENTAL HEALTH WORSE THAN EXPECTED PHYSICAL HEALTH BETTER THAN EXPECTED PHYSICAL HEALTH WORSE THAN EXPECTED
Cohort 1 1998-2000 188 plans 13 plans 15 plans None None
Cohort 2 1999-2001 160 plans 8 plans 5 plans 9 plans 5 plans
Cohort 3 2000-2002 146 plans 15 plans 4 plans 20 plans 1 plan
Cohort 4 2001-2003 152 plans None None 22 plans 1 plan
Cohort 5 2002-2004 153 plans 27 plans 3 plans 21 plans None
Cohort 6 2003-2005 154 plans 18 plans 2 plans None None
Cohort 7 2004-2006 151 plans 5 plans 7 plans 2 plans 13 plans

UPDATED August 2007

A summary of the most recently available performance measurement results is available for download in a sample performance measurement report (PDF, 1.6 MB) . Included in the sample report is an overview of the methodology and design followed for sampling, data collection, and the cleaning, scoring, and analysis of the HOS data files.


Select this link for a print-friendly version of the current page