Minnesota Managed Care Longitudinal Data Analysis. 3. Results

03/31/2016

3.1. Enrollment Choice Results

3.1.1. Descriptive Results

Who are the MSHO and MSC+ Enrollees?

TABLE 3-1. Sample Description: Characteristics of MSHO and MSC+ Enrollees, 2010-2012
Characteristic All MSC+ MSHO
Age group:
   65-69 (reference) 16.8 27.6 14.4
   70-74 17.0 17.6 16.9
   75-79 15.8 13.4 16.3
   80-84 16.4 14.4 16.9
   85-89 16.1 13.4 16.7
   90+ 18.0 13.6 18.9
Female 72.4 68.4 73.3
Died during year 14.2 12.8 14.4
Disabilities/conditions:
   Cognitive/Mental illnessa 60.1 58.8 60.4
   Neurological disabilityb 48.0 43.3 49.0
   Physical disabilityc 31.5 32.1 31.4
   Sensory disabilityd 8.1 7.8 8.2
   Other medical disability or chronic diseasee 85.6 82.4 86.2
Calendar year:
   2010 (reference) 32.4 31.8 32.5
   2011 34.1 33.1 34.3
   2012 33.5 35.1 33.2
N 121,696 21,935 99,761
NOTE: Numbers shown in table are percentages of enrollees with each characteristic.
  1. Including any diagnosis for Alzheimer's disease or dementia, chronic mental illness, depression, psychosis, or schizophrenia.
  2. Including any diagnosis for neurologic impairment or Parkinson's disease.
  3. Including any diagnosis for physical impairment.
  4. Including any diagnosis for sensory impairment.
  5. Including diagnoses for selected medical disability or chronic diseases such as arthritis, chronic respiratory disease (COPD, asthma, emphysema, bronchitis), congestive heart failure, coronary heart disease, stroke, or diabetes.

MSHO and MSHO-eligible MSC+ enrollees differed on a number of key dimensions over the 2010–2012 period, as shown in Table 3-1. Most notably, MSHO enrollees tended to be older than the MSC+ enrollees. Only 14.4 percent of MSHO enrollees were under age 70 compared to 27.6 percent of the MSC+ enrollees, while 18.9 percent of MSHO enrollees were age 90 or older compared to 13.6 percent of the MSC+ enrollees. Consistent with their higher average age, MSHO enrollees were more likely to have many of the health problems and disabilities examined (up to 5.7 percentage points more likely) and were slightly more likely to have died during the year (14.4 vs. 12.8 percent). MSHO enrollees were somewhat more likely to be female (73.3 vs. 68.4 percent), which may also reflect their older age given women's longer life span on average.

3.1.2. Where do the MSHO and MSC+ Enrollees Live?

As with the overall Minnesota population, MSHO and MSC+ enrollees were more concentrated in urban areas of the state. Comparatively though, a smaller proportion of MSHO enrollees (65.6 percent) were in urban counties, as compared to 69.6 percent of MSC+ enrollees (Table 3-2). Among urban counties only, Hennepin and Ramsey counties comprised 31.1 percent and 16.4 percent of MSHO enrollees, respectively, which was almost half (47.5 percent) of all urban MSHO enrollees. A larger proportion of the MSC+ population in urban areas was in these two counties (51.8 percent). On average, MSHO enrollment was higher than MSC+ enrollment in rural counties--at 34.4 percent vs. 30.4 percent.

TABLE 3-2. Sample Description: Location of MSHO and MSC+ Enrollees, 2010-2012
Urban Counties
MCOUNTY
Urban Counties
COUNTY_NAME
Urban Counties
All (%)
Urban Counties
MSC (%)
Urban Counties
MSHO (%)
Total Urban Total Urban 66.3 69.6 65.6
MN002 Anoka 5.1 4.2 5.3
MN005 Benton 1.3 1.2 1.3
MN007 Blue Earth 1.7 0.8 1.9
MN009 Carlton 1.6 0.6 1.8
MN010 Carver 1.2 1.2 1.2
MN013 Chisago 1.0 1.1 0.9
MN014 Clay 1.8 0.9 2.0
MN019 Dakota 6.0 8.1 5.5
MN020 Dodge 0.6 0.3 0.6
MN023 Fillmore 1.2 1.3 1.2
MN027 Hennepin 31.4 32.4 31.1
MN028 Houston 0.7 0.2 0.8
MN030 Isanti 0.8 0.5 0.9
MN040 Le Sueur 0.8 0.4 0.9
MN048 Mille Lacs 1.1 0.4 1.2
MN052 Nicollet 0.7 0.4 0.7
MN055 Olmsted 3.6 5.4 3.2
MN060 Polk 1.8 1.6 1.8
MN062 Ramsey 16.9 19.4 16.4
MN069 St. Louis 8.0 8.1 7.9
MN070 Scott 1.9 2.3 1.8
MN071 Sherburne 1.1 0.7 1.2
MN072 Sibley 0.6 0.3 0.7
MN073 Stearns 3.5 2.4 3.7
MN079 Wabasha 0.7 0.4 0.8
MN082 Washington 2.9 3.4 2.8
MN086 Wright 2.1 1.8 2.2
Total N, Urban   80,673 15,263 65,410
Rural Counties Rural Counties Rural Counties Rural Counties Rural Counties
Total Rural Total Rural 33.7 30.4 34.4
MN001 Aitkin 1.7 1.0 1.8
MN003 Becker 2.9 3.0 2.9
MN004 Beltrami 3.0 13.6 0.9
MN006 Big Stone 0.8 0.2 0.9
MN008 Brown 2.0 2.7 1.9
MN011 Cass 1.8 1.7 1.9
MN012 Chippewa 1.6 0.6 1.8
MN015 Clearwater 1.2 4.2 0.6
MN016 Cook 0.3 0.1 0.3
MN017 Cottonwood 1.3 0.5 1.5
MN018 Crow Wing 4.0 2.6 4.3
MN021 Douglas 2.9 1.3 3.3
MN022 Faribault 1.4 0.8 1.5
MN024 Freeborn 2.3 6.3 1.5
MN025 Goodhue 2.7 2.1 2.9
MN026 Grant 0.6 0.2 0.6
MN029 Hubbard 1.3 5.8 0.4
MN031 Itasca 2.6 3.4 2.4
MN032 Jackson 0.8 0.6 0.8
MN033 Kanabec 1.2 0.9 1.2
MN034 Kandiyohi 2.8 1.6 3.1
MN035 Kittson 0.6 0.5 0.6
MN036 Koochiching 1.4 1.4 1.4
MN037 Lac Qui Parle 1.1 0.3 1.2
MN038 Lake 0.6 0.7 0.6
MN039 Lake of the Woods 0.4 2.1 0.0
MN041 Lincoln 0.8 0.7 0.8
MN042 Lyon 2.1 0.9 2.4
MN043 McLeod 2.2 1.2 2.4
MN044 Mahnomen 0.6 0.2 0.7
MN045 Marshall 1.2 0.4 1.4
MN046 Martin 2.1 0.6 2.3
MN047 Meeker 1.9 0.9 2.1
MN049 Morrison 3.3 4.6 3.0
MN050 Mower 2.8 5.2 2.3
MN051 Murray 0.9 0.2 1.0
MN053 Nobles 1.8 1.1 1.9
MN054 Norman 1.1 0.9 1.1
MN056 Otter Tail 5.6 2.8 6.2
MN057 Pennington 1.4 0.4 1.6
MN058 Pine 2.3 1.6 2.4
MN059 Pipestone 1.1 0.8 1.2
MN061 Pope 1.4 0.3 1.6
MN063 Red Lake 0.5 0.3 0.6
MN064 Redwood 1.3 1.1 1.3
MN065 Renville 1.0 0.5 1.1
MN066 Rice 2.7 3.7 2.5
MN067 Rock 0.7 0.1 0.8
MN068 Roseau 1.4 0.6 1.6
MN074 Steele 2.0 3.4 1.7
MN075 Stevens 0.8 0.4 0.8
MN076 Swift 1.6 0.8 1.7
MN077 Todd 3.2 2.4 3.3
MN078 Traverse 0.7 0.2 0.8
MN080 Wadena 1.8 0.7 2.0
MN081 Waseca 1.1 1.2 1.1
MN083 Watonwan 0.9 0.4 1.0
MN084 Wilkin 0.7 0.3 0.7
MN085 Winona 2.9 2.5 3.0
MN087 Yellow Medicine 1.2 0.4 1.3
Total N, Rural   41,023 6,672 34,351
Grand Total, Percent   100.0 18.0 82.0
Grand Total , N   121,696 21,935 99,761
NOTE: Numbers shown in table are percentages of enrollees in each county.

As would be expected given the wide dispersion of MSHO and MSC+ enrollees across the 87 counties in Minnesota, there were differences in the communities in which MSHO and MSC+ enrollees were living over the 2010-2012 period (Table 3-3). As shown, the counties of MSHO enrollees tended to be very similar to those of MSC+ enrollees, with only very small differences across measures of primary care availability and measures of the living arrangements, marital status, and education of adults 65 and older.

TABLE 3-3. Sample Description: Characteristics of County of Residence for MSHO and MSC+ Enrollees, 2010-2012
County Characteristics All MSC+ MSHO
# PCPs per 1,000 population 0.80 0.85 0.79
% Population 65+ who do not live in community 5.1 4.8 5.2
% Population 65+ who live in community with others 64.9 64.9 64.9
% Population 65+ with college education 21.7 23.0 21.4
% Population 65+ who are married 57.5 57.0 57.6
N 121,696 21,935 99,761
NOTE: Numbers shown in table are mean numbers of percentages.

What Plans are Available to MSHO and MSC+ Enrollees?

Over the 2010-2012 period there were eight different MCOs serving MSHO and/or MSC+ in Minnesota's 87 counties, including Blue Plus, HealthPartners, Itasca Medical Care, Medica, Metropolitan Health Plan, Primewest, South Country Health Alliance, and UCare. On average, the Minnesota counties with at least one MSHO and one MSC+ plan had 2.2 MSHO and 2.2 MSC+ plans over the period, generally operated by the same MCOs. In general, the urban counties had a larger number of plans in both MSHO and MSC+ than did rural counties.

How often do MSHO Enrollees Disenroll from MSHO?

There was very little disenrollment from MSHO over the 2010-2012 period. We estimate that less than 0.4 percent of MSHO enrollees who were in the program in January of any year had subsequently left the program during that same year (excluding those who died).

How often do MSC+ Enrollees Disenroll from MSC+? What are Their Characteristics?

On average, 12.8 percent of MSC+ enrollees at the beginning of any year subsequently switched to MSHO during the same year. Those who switch are slightly younger, are less likely to die, and have a slightly higher prevalence of medical conditions and disabilities (Table 3-4).

TABLE 3-4. Comparison of Consistent MSHO Enrollees and Those Who Switched to MSC+ During the Year, 2010-2012
Characteristics Overall
(%)
MSC+ Only Enrollees
(%)
MSC+ to
MSHO Switchers
(%)
Statistical
Significance
Age Group ***
   65-69 28.0 27.6 30.4  
   70-74 17.7 17.6 18.4  
   75-79 13.6 13.4 15.0  
   80-84 14.4 14.4 14.5  
   85-89 13.1 13.4 11.7  
   90+ 13.1 13.6 10.0  
Sex Group ***
   Female 68.1 68.4 66.1  
   Male 31.9 31.6 33.9  
Died during year ***
   Alive 87.9 87.2 93.0  
   Died 12.1 12.8 7.0  
Disabilities/conditions  
   Cognitive/mental illness 58.9 58.8 59.3  
   Neurological disability 43.6 43.3 45.4 *
   Physical disability 33.0 32.1 39.3 ***
   Sensory disability 8.0 7.8 9.8 ***
   Other medical disability 82.6 82.5 83.3  
N 25,162 21,935 3,227  
(%) (100.0) (87.2) (12.8)  
NOTE: Numbers shown in table are percentages of enrollees with each characteristic.
*/**/*** means statistically significant at p<0.10, 0.05, 0.01, respectively.

Are there differences in MSHO Enrollees and MSC+ Enrollees in Urban and Rural Areas?

The MSHO eligible population in urban areas is similar to the MSHO eligible population in rural areas on many dimensions, as shown in Table 3-5. Of note, however, the urban population tends to be younger than the rural population. For example, 16.1 percent of the MSHO eligible population in urban counties is under age 70 as compared to 11.0 percent in rural counties.

There are many more differences in the characteristics of the counties in which the urban and rural MSHO eligible population is living, with the population in urban counties having access to almost 50 percent more PCPs per capita than the population in rural counties. The population aged 65 and older in urban counties is also more likely to have a college education and less likely to be married than the rural population 65 and older.

TABLE 3-5. Sample Description: Characteristics of MSHO and MSC+ Enrollees and Their County of Residence by Urban/Rural Status, 2010-2012
Characteristics Alla MSC+a MSHOa Alla MSC+a MSHOa
Age group:
   65-69 (reference) 18.4 28.5 16.1 13.4 25.7 11.0
   70-74 18.7 18.6 18.7 13.7 15.3 13.3
   75-79 16.6 13.6 17.4 14.0 13.0 14.2
   80-84 16.1 14.5 16.5 17.0 14.3 17.6
   85-89 14.9 12.9 15.3 18.5 14.4 19.3
   90+ 15.2 12.0 16.0 23.3 17.2 24.5
Female 71.9 68.4 72.7 73.3 68.3 74.3
Died during year 13.0 12.2 13.2 16.5 14.3 16.9
Disabilities/conditions:
   Cognitive/mental illnessc 59.5 59.4 59.5 61.2 57.5 61.9
   Neurological disabilityd 48.4 44.7 49.2 47.2 40.2 48.5
   Physical disabilitye 32.4 33.3 32.2 29.7 29.3 29.8
   Sensory disabilityf 8.7 8.5 8.8 6.8 6.2 7.0
   Other medical disability or chronic diseaseg 84.9 82.0 85.5 86.9 83.5 87.6
Calendar year:
   2010 (reference) 32.2 31.6 32.3 32.8 32.2 33.0
   2011 34.1 33.1 34.3 34.1 33.2 34.3
   2012 33.7 35.3 33.4 33.0 34.6 32.7
County characteristics:
   # PCPs per 1,000 population 0.90 0.94 0.88 0.61 0.63 0.61
   % Population 65+ who do not live in community 5.0 4.8 5.0 5.4 4.9 5.5
   % Population 65+ who live in community with others 64.4 64.5 64.4 65.7 65.8 65.7
   % Population 65+ with college education 25.0 25.9 24.8 15.2 16.5 14.9
   % Population 65+ who are married 55.5 55.2 55.6 61.4 60.9 61.5
N 80,673 15,263 65,410 41,023 6,672 34,351
NOTE: Numbers shown in table are percentages of enrollees with each characteristic.
  1. Urban.
  2. Rural.
  3. Including any diagnosis for Alzheimer's disease or dementia, chronic mental illness, depression, psychosis, or schizophrenia.
  4. Including any diagnosis for neurologic impairment or Parkinson's disease.
  5. Including any diagnosis for physical impairment.
  6. Including any diagnosis for sensory impairment.
  7. Including diagnoses for selected medical disability or chronic diseases such as arthritis, chronic respiratory disease (COPD, asthma, emphysema, bronchitis), congestive heart failure, coronary heart disease, stroke, or diabetes.

Notwithstanding these urban and rural population differences, the differences between the MSHO and MSC+ enrollees within the urban and rural areas tended to mirror the differences for the overall population. For example, MSHO enrollees tend to be older and are more likely to be female than MSC+ enrollees in both urban and rural counties.

3.1.3. Multivariate Results

What Factors are Associated with MSHO Enrollment?

Among the MSHO eligible population in the 87 counties that had at least one MSHO and one MSC+ plan during the 2010-2012 period, the probability of MSHO enrollment is higher for women than men and increases with age in the model that controls for individual characteristics (Table 3-6, Model 1). As shown, adults age 70 and older are more likely to enroll in MSHO than adults under age 70 (the reference group), with odds ratios ranging from 1.82 for adults age 70-74 to 2.53 for adults age 90 or older. The odds ratio of 1.82 means a person having the characteristic described (in this case, age 70 or older), is 82 percent more likely to enroll in MSHO than those under age 70. The increase in the odds of enrolling in MSHO as the eligible population ages is consistent with a cumulative effect of outreach and enrollment efforts over time and, as noted above, very little disenrollment from MSHO for those who do enroll. Perhaps also reflecting the influence of time, eligible adults who die during the year are somewhat less likely to be MSHO enrollees after controlling for other factors.

TABLE 3-6. Logistic Regression Results on Enrollment in a MSHO Plan (vs. MSC+), 2010-2012: Overall Sample, Model 1
Characteristic β AOR
Age group (Ref.=65-69):
   70-74 0.600 *** 1.822
   75-79 0.824 *** 2.280
   80-84 0.774 *** 2.169
   85-89 0.826 *** 2.283
   90+ 0.927 *** 2.527
Female 0.147 *** 1.159
Died during year -0.058 * 0.943
Disabilities/conditions:
   Mental illnessa -0.038 * 0.963
   Neurological disabilityb 0.057 ** 1.059
   Physical disabilityc -0.009 0.991
   Sensory disabilityd 0.060 * 1.061
   Other medical disability or chronic diseasee 0.176 *** 1.192
Calendar year (Ref.=2010):
   2011 0.021 1.021
   2012 -0.065 *** 0.937
County characteristics:
   # PCPs per 1,000 population ---
   % Population 65+ who do not live in community ---
   % Population 65+ who live in community with others ---
   % Population 65+ with college education ---
   % Population 65+ who are married ---
   County fixed effects NO
N 121,696
NOTE: Model predicting the probability of being an MSHO enrollee (i.e., MSHO=1, MSC+=0).
  1. Including any diagnosis for Alzheimer's disease or dementia, chronic mental illness, depression, psychosis, or schizophrenia.
  2. Including any diagnosis for neurologic impairment or Parkinson's disease.
  3. Including any diagnosis for physical impairment.
  4. Including any diagnosis for sensory impairment.
  5. Including diagnoses for selected medical disability or chronic diseases such as arthritis, chronic respiratory disease (COPD, asthma, emphysema, bronchitis), congestive heart failure, coronary heart disease, stroke, or diabetes.
  6. Including 87 counties. Reference/omitted category = Hennepin County.
  7. N is smaller in this model because of missing values on county characteristics.

Likelihood Ratio Chi square values: Model 1=2458, Model 2=3311, Model 3=9677, all p<0.0001.
Statistical significance: * p<0.05, ** p<0.01, *** p<0.001.

The association between health and disability status and MSHO enrollment is somewhat mixed. Primarily, adults with other chronic disease/medical disabilities (beyond the specific categories of illness and disability listed in the table) are almost 20 percent more likely to enroll in MSHO than MSC+, all else equal. Otherwise, there was no significant difference in MSHO enrollment for adults with a physical disability, while adults with a sensory disability, neurological disability, or other chronic disease/medical disabilities are somewhat more likely to enroll in MSHO and those with a mental disability or illness are somewhat less likely. MSHO enrollment varied modestly over time, with enrollment somewhat higher in 2011 and lower in 2012 relative to 2010.

Adding controls for county characteristics improves the explanatory power of the model but has little impact on the key findings as the basic relationship between the characteristics of the eligible adults and the likelihood MSHO enrollment remains stable (Table 3-7, Model 2). As with the model with only individual-level covariates, we see increased MSHO enrollment for women and older adults, and a mixed pattern for adults with different health conditions and disabilities after adding controls for county characteristics (Model 2).

Similarly, replacing the area-level covariates with county fixed effects also improves the explanatory power of the model over the base model with little effect on the relationship between the characteristics of the eligible adults and the likelihood of MSHO enrollment (Table 3-8, Model 3). As with the base model and the models with county-level characteristics, the odds of MSHO enrollment increase for women and older adults, while the pattern remains mixed for adults with different health conditions and disabilities, all else equal.

Are there Differences in the Factors Associated with MSHO Enrollment in Urban and Rural Areas?

As was true for the overall population, the probability of MSHO enrollment is higher for women than men and increases with age in both urban counties (Table 3-9, Model 1; Table 3-10, Model 2; Table 3-11, Model 3) and rural counties (Table 3-12, Model 1; Table 3-13, Model 2; Table 3-14, Model 3). As was true for the overall population results in Table 3-6, the addition of county-level characteristics (Model 2) or county fixed effects (Model 3) in Table 3-7 and Table 3-8, although the area-level characteristics improved the overall explanatory power of the model, had little impact on the core findings.

The enrollment analyses identified differences between MSHO and MSC+ enrollees and their communities and highlights the need to control for these characteristics in analyses comparing health care use among those in the MSHO and MSC+ programs. The odds of MSHO enrollment are significantly higher for women, older adults, and adults with certain medical conditions and disabilities, as well as for adults in some Minnesota communities. Controlling for those differences areimportant in determining whether there are significant improvements in health care use for MSHO enrollees relative to similar individuals who remain in MSC+.

TABLE 3-7. Logistic Regression Results on Enrollment in a MSHO Plan (vs. MSC+), 2010-2012: Overall Sample, Model 2
Characteristic β AOR
Age group (Ref.=65-69):
   70-74 0.603 *** 1.827
   75-79 0.817 *** 2.264
   80-84 0.735 *** 2.086
   85-89 0.765 *** 2.148
   90+ 0.848 *** 2.335
Female 0.143 *** 1.154
Died during year -0.070 ** 0.932
Disabilities/conditions:
   Mental illnessa -0.053 ** 0.949
   Neurological disabilityb 0.079 *** 1.083
   Physical disabilityc 0.012 1.012
   Sensory disabilityd 0.077 ** 1.080
   Other medical disability or chronic diseasee 0.164 *** 1.178
Calendar year (Ref.=2010):
   2011 0.100 *** 1.105
   2012 -0.013 0.987
County characteristics:
   # PCPs per 1,000 population -0.155 *** 0.856
   % Population 65+ who do not live in community 0.081 *** 1.085
   % Population 65+ who live in community with others 0.015 *** 1.016
   % Population 65+ with college education -0.019 *** 0.981
   % Population 65+ who are married -0.011 *** 0.989
   County fixed effects NO
N 120,815g
NOTE: Model predicting the probability of being an MSHO enrollee (i.e., MSHO=1, MSC+=0).
  1. Including any diagnosis for Alzheimer's disease or dementia, chronic mental illness, depression, psychosis, or schizophrenia.
  2. Including any diagnosis for neurologic impairment or Parkinson's disease.
  3. Including any diagnosis for physical impairment.
  4. Including any diagnosis for sensory impairment.
  5. Including diagnoses for selected medical disability or chronic diseases such as arthritis, chronic respiratory disease (COPD, asthma, emphysema, bronchitis), congestive heart failure, coronary heart disease, stroke, or diabetes.
  6. Including 87 counties. Reference/omitted category = Hennepin County.
  7. N is smaller in this model because of missing values on county characteristics.

Likelihood Ratio Chi square values: Model 1=2458, Model 2=3311, Model 3=9677, all p<0.0001.
Statistical significance: * p<0.05, ** p<0.01, *** p<0.001.

 

TABLE 3-8. Logistic Regression Results on Enrollment in a MSHO Plan (vs. MSC+), 2010-2012: Overall Sample, Model 3
Characteristic β AOR
Age group (Ref.=65-69):
   70-74 0.643 *** 1.902
   75-79 0.861 *** 2.364
   80-84 0.772 *** 2.163
   85-89 0.798 *** 2.221
   90+ 0.881 *** 2.413
Female 0.153 *** 1.165
Died during year -0.070 ** 0.933
Disabilities/conditions:
   Mental illnessa -0.050 ** 0.951
   Neurological disabilityb 0.050 ** 1.051
   Physical disabilityc -0.007 0.993
   Sensory disabilityd 0.053 1.054
   Other medical disability or chronic diseasee 0.166 *** 1.180
Calendar year (Ref.=2010):
   2011 0.043 * 1.044
   2012 -0.049 * 0.952
County characteristics:
   # PCPs per 1,000 population ---
   % Population 65+ who do not live in community ---
   % Population 65+ who live in community with others ---
   % Population 65+ with college education ---
   % Population 65+ who are married ---
   County fixed effects YESf
N 121,696
NOTE: Model predicting the probability of being an MSHO enrollee (i.e.,MSHO=1, MSC+=0).
  1. Including any diagnosis for Alzheimer's disease or dementia, chronic mental illness, depression, psychosis, or schizophrenia.
  2. Including any diagnosis for neurologic impairment or Parkinson's disease.
  3. Including any diagnosis for physical impairment.
  4. Including any diagnosis for sensory impairment.
  5. Including diagnoses for selected medical disability or chronic diseases such as arthritis, chronic respiratory disease (COPD, asthma, emphysema, bronchitis), congestive heart failure, coronary heart disease, stroke, or diabetes.
  6. Including 87 counties. Reference/omitted category = Hennepin County.
  7. N is smaller in this model because of missing values on county characteristics.

Likelihood Ratio Chi square values: Model 1=2458, Model 2=3311, Model 3=9677, all p<0.0001.
Statistical significance: * p<0.05, ** p<0.01, *** p<0.001.

 

TABLE 3-9. Logistic Regression Results on Enrollment in a MSHO Plan (vs. MSC+), 2010-2012: Urban Sample, Model 1
Characteristic β AOR
Age group (Ref.=65-69):
   70-74 0.564 *** 1.758
   75-79 0.794 *** 2.211
   80-84 0.669 *** 1.952
   85-89 0.699 *** 2.012
   90+ 0.807 *** 2.241
Female 0.143 *** 1.154
Died during year -0.078 ** 0.925
Disabilities/conditions:
   Mental illnessa -0.083 *** 0.921
   Neurological disabilityb 0.054 * 1.055
   Physical disabilityc -0.033 0.968
   Sensory disabilityd 0.039 1.040
   Other medical disability or chronic diseasee 0.178 *** 1.195
Calendar year (Ref.=2010):
   2011 0.023 1.023
   2012 -0.066 ** 0.937
County characteristics:
   # PCPs per 1,000 population ---
   % Population 65+ who do not live in community ---
   % Population 65+ who live in community with others ---
   % Population 65+ with college education ---
   % Population 65+ who are married ---
   County fixed effects NO
N 80,673
NOTE: Model predicting the probability of being an MSHO enrollee.
  1. Including any diagnosis for Alzheimer's disease or dementia, chronic mental illness, depression, psychosis, or schizophrenia.
  2. Including any diagnosis for neurologic impairment or Parkinson's disease.
  3. Including any diagnosis for physical impairment.
  4. Including any diagnosis for sensory impairment.
  5. Including diagnoses for selected medical disability or chronic diseases such as arthritis, chronic respiratory disease (COPD, asthma, emphysema, bronchitis), congestive heart failure, coronary heart disease, stroke, or diabetes.
  6. Including 27 metro (urban) counties. Reference/omitted category = Hennepin County.
  7. N is smaller in this model because of missing values on county characteristics.

Likelihood Ratio Chi square values: Model 1=1388, Model 2=1770, Model 3=2521, all p<0.0001.
Statistical significance: * p<0.05, ** p<0.01, *** p<0.001.

 

TABLE 3-10. Logistic Regression Results on Enrollment in a MSHO Plan (vs. MSC+), 2010-2012: Urban Sample, Model 2
Characteristic β AOR
Age group (Ref.=65-69):
   70-74 0.570 *** 1.768
   75-79 0.796 *** 2.216
   80-84 0.647 *** 1.910
   85-89 0.665 *** 1.944
   90+ 0.762 *** 2.144
Female 0.132 *** 1.141
Died during year -0.085 ** 0.918
Disabilities/conditions:
   Mental illnessa -0.095 *** 0.909
   Neurological disabilityb 0.054 * 1.056
   Physical disabilityc -0.015 0.985
   Sensory disabilityd 0.047 1.048
   Other medical disability or chronic diseasee 0.171 *** 1.187
Calendar year (Ref.=2010):
   2011 0.094 *** 1.099
   2012 0.001 1.001
County characteristics:
   # PCPs per 1,000 population -0.300 *** 0.741
   % Population 65+ who do not live in community 0.024 *** 1.024
   % Population 65+ who live in community with others -0.011 * 0.990
   % Population 65+ with college education -0.014 *** 0.987
   % Population 65+ who are married 0.000 1.000
   County fixed effects NO
N 80,063g
NOTE: Model predicting the probability of being an MSHO enrollee.
  1. Including any diagnosis for Alzheimer's disease or dementia, chronic mental illness, depression, psychosis, or schizophrenia.
  2. Including any diagnosis for neurologic impairment or Parkinson's disease.
  3. Including any diagnosis for physical impairment.
  4. Including any diagnosis for sensory impairment.
  5. Including diagnoses for selected medical disability or chronic diseases such as arthritis, chronic respiratory disease (COPD, asthma, emphysema, bronchitis), congestive heart failure, coronary heart disease, stroke, or diabetes.
  6. Including 27 metro (urban) counties. Reference/omitted category = Hennepin County.
  7. N is smaller in this model because of missing values on county characteristics.

Likelihood Ratio Chi square values: Model 1=1388, Model 2=1770, Model 3=2521, all p<0.0001.
Statistical significance: * p<0.05, ** p<0.01, *** p<0.001.

 

TABLE 3-11. Logistic Regression Results on Enrollment in a MSHO Plan (vs. MSC+), 2010-2012: Urban Sample, Model 3
Characteristic β AOR
Age group (Ref.=65-69):
   70-74 0.582 *** 1.790
   75-79 0.798 *** 2.221
   80-84 0.651 *** 1.918
   85-89 0.662 *** 1.938
   90+ 0.757 *** 2.131
Female 0.140 *** 1.150
Died during year -0.083 ** 0.920
Disabilities/conditions:
   Mental illnessa -0.101 *** 0.904
   Neurological disabilityb 0.047 * 1.048
   Physical disabilityc -0.024 0.977
   Sensory disabilityd 0.045 1.046
   Other medical disability or chronic diseasee 0.171 *** 1.187
Calendar year (Ref.=2010):
   2011 0.025 1.025
   2012 -0.062 ** 0.940
County characteristics:
   # PCPs per 1,000 population ---
   % Population 65+ who do not live in community ---
   % Population 65+ who live in community with others ---
   % Population 65+ with college education ---
   % Population 65+ who are married ---
   County fixed effects YESf
N 80,673
NOTE: Model predicting the probability of being an MSHO enrollee.
  1. Including any diagnosis for Alzheimer's disease or dementia, chronic mental illness, depression, psychosis, or schizophrenia.
  2. Including any diagnosis for neurologic impairment or Parkinson's disease.
  3. Including any diagnosis for physical impairment.
  4. Including any diagnosis for sensory impairment.
  5. Including diagnoses for selected medical disability or chronic diseases such as arthritis, chronic respiratory disease (COPD, asthma, emphysema, bronchitis), congestive heart failure, coronary heart disease, stroke, or diabetes.
  6. Including 27 metro (urban) counties. Reference/omitted category = Hennepin County.
  7. N is smaller in this model because of missing values on county characteristics.

Likelihood Ratio Chi square values: Model 1=1388, Model 2=1770, Model 3=2521, all p<0.0001.
Statistical significance: * p<0.05, ** p<0.01, *** p<0.001.

 

TABLE 3-12. Logistic Regression Results on Enrollment in a MSHO Plan (vs. MSC+), 2010-2012: Rural Sample, Model 1
Characteristic β AOR
Age group (Ref.=65-69):
   70-74 0.694 *** 2.002
   75-79 0.896 *** 2.450
   80-84 0.998 *** 2.712
   85-89 1.060 *** 2.886
   90+ 1.112 *** 3.040
Female 0.157 *** 1.170
Died during year -0.026 0.974
Disabilities/conditions:
   Mental illnessa 0.054 1.055
   Neurological disabilityb 0.098 ** 1.103
   Physical disabilityc 0.052 1.053
   Sensory disabilityd 0.155 ** 1.167
   Other medical disability or chronic diseasee 0.175 *** 1.192
Calendar year (Ref.=2010):
   2011 0.018 1.018
   2012 -0.063 0.939
County characteristics:
   # PCPs per 1,000 population ---
   % Population 65+ who do not live in community ---
   % Population 65+ who live in community with others ---
   % Population 65+ with college education ---
   % Population 65+ who are married ---
   County fixed effects NO
N 41,023
NOTE: Model predicting the probability of being an MSHO enrollee.
  1. Including any diagnosis for Alzheimer's disease or dementia, chronic mental illness, depression, psychosis, or schizophrenia.
  2. Including any diagnosis for neurologic impairment or Parkinson's disease.
  3. Including any diagnosis for physical impairment.
  4. Including any diagnosis for sensory impairment.
  5. Including diagnoses for selected medical disability or chronic diseases such as arthritis, chronic respiratory disease (COPD, asthma, emphysema, bronchitis), congestive heart failure, coronary heart disease, stroke, or diabetes.
  6. Including 60 non-metro (rural) counties. Reference/omitted category = Otter Tail County.
  7. N is smaller in this model because of missing values on county characteristics.

Likelihood Ratio Chi square values: Model 1=1116, Model 2=2354, Model 3=7201, all p<0.0001.
Statistical significance: * p<0.05, ** p<0.01, *** p<0.001.

 

TABLE 3-13. Logistic Regression Results on Enrollment in a MSHO Plan (vs. MSC+), 2010-2012: Rural Sample, Model 2
Characteristic β AOR
Age group (Ref.=65-69):
   70-74 0.710 *** 2.033
   75-79 0.909 *** 2.482
   80-84 1.016 *** 2.762
   85-89 1.063 *** 2.895
   90+ 1.096 *** 2.992
Female 0.157 *** 1.170
Died during year -0.023 0.978
Disabilities/conditions:
   Mental illnessa 0.083 ** 1.087
   Neurological disabilityb 0.106 ** 1.112
   Physical disabilityc 0.043 1.043
   Sensory disabilityd 0.161 ** 1.175
   Other medical disability or chronic diseasee 0.166 *** 1.180
Calendar year (Ref.=2010):
   2011 -0.034 0.967
   2012 -0.145 *** 0.865
County characteristics:
   # PCPs per 1,000 population 0.083 1.087
   % Population 65+ who do not live in community 0.209 *** 1.232
   % Population 65+ who live in community with others 0.059 *** 1.061
   % Population 65+ with college education -0.065 *** 0.937
   % Population 65+ who are married 0.022 *** 1.022
   County fixed effects NO
N 40,752g
NOTE: Model predicting the probability of being an MSHO enrollee.
  1. Including any diagnosis for Alzheimer's disease or dementia, chronic mental illness, depression, psychosis, or schizophrenia.
  2. Including any diagnosis for neurologic impairment or Parkinson's disease.
  3. Including any diagnosis for physical impairment.
  4. Including any diagnosis for sensory impairment.
  5. Including diagnoses for selected medical disability or chronic diseases such as arthritis, chronic respiratory disease (COPD, asthma, emphysema, bronchitis), congestive heart failure, coronary heart disease, stroke, or diabetes.
  6. Including 60 non-metro (rural) counties. Reference/omitted category = Otter Tail County.
  7. N is smaller in this model because of missing values on county characteristics.

Likelihood Ratio Chi square values: Model 1=1116, Model 2=2354, Model 3=7201, all p<0.0001.
Statistical significance: * p<0.05, ** p<0.01, *** p<0.001.

 

TABLE 3-14. Logistic Regression Results on Enrollment in a MSHO Plan (vs. MSC+), 2010-2012: Rural Sample, Model 3
Characteristic β AOR
Age group (Ref.=65-69):
   70-74 0.847 *** 2.332
   75-79 1.080 *** 2.945
   80-84 1.145 *** 3.142
   85-89 1.185 *** 3.271
   90+ 1.215 *** 3.369
Female 0.183 *** 1.201
Died during year -0.036 0.964
Disabilities/conditions:
   Mental illnessa 0.095 ** 1.100
   Neurological disabilityb 0.060 1.062
   Physical disabilityc 0.041 1.042
   Sensory disabilityd 0.110 1.116
   Other medical disability or chronic diseasee 0.165 *** 1.180
Calendar year (Ref.=2010):
   2011 0.092 * 1.096
   2012 -0.015 0.985
County characteristics:
   # PCPs per 1,000 population ---
   % Population 65+ who do not live in community ---
   % Population 65+ who live in community with others ---
   % Population 65+ with college education ---
   % Population 65+ who are married ---
   County fixed effects YESf
N 41,023
NOTE: Model predicting the probability of being an MSHO enrollee.
  1. Including any diagnosis for Alzheimer's disease or dementia, chronic mental illness, depression, psychosis, or schizophrenia.
  2. Including any diagnosis for neurologic impairment or Parkinson's disease.
  3. Including any diagnosis for physical impairment.
  4. Including any diagnosis for sensory impairment.
  5. Including diagnoses for selected medical disability or chronic diseases such as arthritis, chronic respiratory disease (COPD, asthma, emphysema, bronchitis), congestive heart failure, coronary heart disease, stroke, or diabetes.
  6. Including 60 non-metro (rural) counties. Reference/omitted category = Otter Tail County.
  7. N is smaller in this model because of missing values on county characteristics.

Likelihood Ratio Chi square values: Model 1=1116, Model 2=2354, Model 3=7201, all p<0.0001.
Statistical significance: * p<0.05, ** p<0.01, *** p<0.001.

3.2. Outcomes Analyses

We present results for facility/institutional measures first, followed by results for community-based care, including physician care, HCBS, assisted living and hospice care.

Hospital Care and Long-Term Nursing Home Care. Looking first at simple differences between the MSHO and MSC+ enrollees, we find that MSHO enrollees were less likely than MSC+ enrollees to have had an inpatient stay in a given year (18.6 percent vs. 27.4 percent) (Table 3-15). Further, among those enrollees with an inpatient stay, MSHO enrollees tended to have somewhat fewer stays than MSC+ enrollees over the year (1.8 vs. 2.5 stays). By contrast, MSHO and MSC+ enrollees were equally likely to have had an outpatient ED visit during a given year, at about 30 percent. However, among enrollees who used the ED, MSHO enrollees had less than two annual visits on average, while MSC+ enrollees had almost three visits per year. In contrast to their lower levels of hospital use, a higher proportion of MSHO enrollees than MSC+ enrollees had a long-term nursing facility stay over the year, as derived from a yes/no indicator of long-term nursing home use in the encounter data (30.6 percent vs. 24.9 percent).

TABLE 3-15. Descriptive Statistics on Hospital Care, Long-Term Nursing Home Care, and Community-Based Care by MSHO and MSC+ Enrollees, 2010-2012
Outcome Measure Total
Sample Size
Enrollees
MSHO
Enrollees
MSC+
Difference
Hospital Care
Any hospital stay (%) 121,696 18.6 27.4 -8.9
Number of stays among those with a stay (mean) 24,067 1.8 2.5 -0.7
Any outpatient ED visit (%) 121,696 30.0 30.5 -0.5
Number of visits among those with a visit (mean) 36,541 1.7 2.8 -1.1
Long-Term Care Nursing Home Care
Any long-term nursing facility stay (%) 121,696 30.6 24.9 5.7
Community-Based Care
Any physician visit (%) 121,696 94.6 91.5 3.1
Number of visits among those with a visit (mean) 114,371 12.3 19.0 -6.7
Any PCP visit (%) 121,696 85.2 71.9 13.3
Number of visits among those with a visit (mean) 100,622 7.2 11.5 -4.3
Any specialist visit (%) 121,696 83.7 84.4 -0.7
Number of visits among those with a visit (mean) 101,830 6.5 10.8 -4.3
Any HCBS use (%) 121,696 55.7 52.4 3.3
Any assisted living use (%) 121,696 17.3 16.9 0.4
Any hospice use (among those who died during the year) (%) 17,221 42.7 40.9 1.8
NOTE: This analysis relies on measures of the number of visits that delete extreme outliers, which involves dropping less than 0.05% of the overall sample from all outpatient physician measures.

As noted above, MSHO enrollees tend to be older, are more likely to be female, have more health problems, and are more likely to live in rural areas than MSC+ enrollees, all of which could contribute to differences in health care use. After controlling for the differences between MSHO and MSC+ enrollees, we find significantly lower use of hospital care for both inpatient stays and outpatient ED visits, and no significant difference in long-term nursing home care (derived from an encounter data indicator) for MSHO enrollees relative to similar MSC+ enrollees (Table 3-16). As shown, MSHO enrollees were substantially less likely than similar MSC+ enrollees to have a hospital stay or an outpatient ED visit and, among enrollees who used that care, fewer episodes of care. Controlling for the effects of covariates in all outcomes analyses greatly matters. For example, for hospital stays, the unadjusted odds ratio calculated from the descriptive statistics in Table 3-15 is 0.604 (not shown above) vs. the adjusted odds ratio (AOR) of 0.515 in Table 3-16.

TABLE 3-16. Regression-Based Estimates of Effect of MSHO Relative to MSC+ on Hospital Care, Long-Term Nursing Home Care, and Community-Based Care, 2010-2012
Outcome Measure Sample Size Logit Model
for Any Use:
Odds Ratio
Count Model
for Level of
Use Among Users:
IRR
Hospital Care
Any hospital stay 121,696 0.515 ***  
Number of stays among those with a stay 24,067   0.744 ***
Any outpatient ED visit 121,696 0.938 ***  
Number of visits among those with a visit 36,541   0.616 ***
Long-Term Nursing Home Care
Any long-term nursing facility stay 121,696 1.015  
Community-Based Care
Any physician visit 121,696 1.604 ***  
Number of visits among those with a visit 114,371   0.679 ***
Any PCP visit 121,696 2.705 ***  
Number of visits among those with a visit 100,622   0.633 ***
Any specialist visit 121,696 0.964  
Number of visits among those with a visit 101,830   0.641 ***
Any HCBS use 121,696 1.134 ***  
Any assisted living use 121,696 0.842 ***  
Any hospice use (among those who died during the year) 17,221 1.087 *  
NOTE: This analysis relies on measures of the number of visits that delete extreme outliers, which involves dropping less than 0.05% of the overall sample from outpatient physician measures.
*/**/*** Statistically significant at the 0.10/0.05/0.01 level, two-tailed test.

Community-Based Care. Looking at the simple differences in community-based care between MSHO and MSC+ enrollees, we find that both groups of enrollees were equally likely to have had a specialist visit over the past year (roughly 84 percent), while MSHO enrollees were more likely than MSC+ enrollees to have had a PCP visit (85.2 percent vs. 71.9 percent) and, due to that, somewhat more likely to have had any physician visit over the year (94.6 percent vs. 91.5 percent) (Table 3-15). However, among those who used any physician care over the past year, MSHO enrollees had an average of almost seven fewer visits than MSC+ enrollees, including an average of four fewer PCP visits among those who saw a PCP and four fewer specialist visits among those who saw a specialist (Table 3-15). Along with the higher level of PCP use, MSHO enrollees were more likely to use HCBS (55.7 percent vs. 52.4 percent) and hospice care over the past year (42.7 percent vs. 40.9 percent). There was little difference between the two groups in the use of assisted living, approximately about 17 percent among both MSHO and MSC+ (Table 3-15).

Controlling for the differences between MSHO and MSC+ enrollees, MSHO enrollees are rough 1.6 times as likely as similar MSC+ enrollees to have had any physician visit over the past year, driven by the greater use of PCPs (Table 3-16). MSHO enrollees are almost three times as likely as similar MSC+ enrollees to have had a PCP visit and equally likely to have had a specialist visit over the past year. Further, among those using each type of care examined, MSHO enrollees had fewer visits than similar MSC+ enrollees, with the incidence rate for MSHO enrollees relative to MSC+ enrollees roughly two-thirds for the number of visits to any physician, to PCPs, and to specialists. MSHO enrollees were also significantly more likely to have used HCBS care and less likely to have been in assisted living than similar MSC+ enrollees during the year. MSHO enrollees who died during the year were marginally more likely to have used hospice care than similar MSC+ enrollees. (The incidence rate for MSHO enrollees relative to similar MSC+ enrollees based on negative binomial models that estimate the impact of MSHO enrollment on the full sample and not just users of a service yield a similar pattern, with MSHO enrollees estimated to have fewer visits than similar MSC+ enrollees (data not shown).

TABLE 3-17. Non-MSHO/MSC+ Changes in Minnesota in Overall Levels of Care, 2010-2012
Service Type Likelihood of
Any Use
Amount of
Use if Any
Hospital use -15% -3%
ED use +6% +2%
Long-term nursing home use -14% Not assessed
HCBS use +10 Not assessed
Assisted living use +6% Not assessed
Any type of physician -10% +5%
PCP use -4% -1%
Specialist physician use No change +13%
Hospice use +25% Not assessed

Beyond differences in the impacts of MSHO, the analysis also provides evidence of changes in overall levels of care in the state over time (evident via time dummy variables comparing year 2012 to year 2010). As shown in Table 3-17, between 2010 and 2012, hospital and long-term nursing home use was greatly decreasing, and ED, HCBS, and assisted living facility use was increasing.

FIGURE 3-1. Effect of MSHO Relative to MSC+ on the Probability of Health Care and Long-Term Care Use: Logistic Regression Model Estimates
FIGURE 3-1: The 9 service types estimated using logistic regression are displayed as rows on the vertical axis, and the x-axis is the odds ratio value estimated from the regression. A vertical line at odds ratio value equals 1 is displayed. The confidence intervals for each of the 9 services are graphed as short horizontal lines. Those services with statistically significant differences as shown in Table 3-16 have confidence interval lines crossing the vertical line where the odds ratio equals 1.

We visualize the estimated effects of enrollment in MSHO relative to MSC+, in the form of AORs and their 95 percent confidence intervals from logistic regression analyses, on each of the binary utilization outcomes (Figure 3-1). Statistically significant effects are indicated by AORs whose 95 percent confidence intervals do not encompass the value of 1.00 (as demarcated by a vertical line on the graph).

Similarly, we graph the estimated effects of enrollment in MSHO relative to MSC+, in the form of IRRs and their 95 percent confidence intervals from negative binomial regression analyses, on each of the utilization count outcomes (Figure 3-2). As shown, in all cases the estimated IRRs and their 95 percent confidence intervals are below 1.00, indicating lower levels of utilization (fewer inpatient stays, ED visits and physician visits) by MSHO enrollees than MSC+ enrollees. As an example of interpreting these IRRs, the ratio for hospital stays is 0.744, which means that MSHO enrollees had approximately 25.6 percent (1 minus 0.744 equals 0.256) fewer stays than MSC+ enrollees who had any hospital admissions.

FIGURE 3-2. Effect of MSHO Relative to MSC+ on the Count of Health Care Utilizations: Negative Binomial Regression Model Estimates
FIGURE 3-2: The 5 service types estimated using negative binomial regression are displayed as rows on the vertical axis, and the x-axis is the incidence rate ratio value estimated from the regression. A vertical line at incidence rate ratio value equals 1 is displayed. The confidence intervals for each of the 5 services are graphed as short horizontal lines. Those services with statistically significant differences as shown in Table 3-16 have confidence interval lines crossing the vertical line where the incidence rate ratio equals 1.

Table 3-18 contains summary-level estimates of the effect of MSHO enrollment (compared to MSC+) for the overall sample and separately for those enrollees in urban vs. rural areas (no other covariates shown) using Model 3.

Appendix Tables A-1 through A-9 in the Appendix contain complete Logistic regression output results for the nine settings examined, presenting the results on the three models listed in the Methods section (individual-level only, individual plus specific area-level covariates, and individual-level plus county-fixed effects).

Tables A-10 through A-14 in the Appendix contain complete negative binomial regression output results for five settings for which we were able to create count measures, again presenting the results on the three models listed in the Methods section.

TABLE 3-18. Estimated Effect of MSHO Relative to MSC+ on Health Care and Long-Term Care Use by Overall, and for Urban and Rural Subgroups, 2010-2012
Outcome Overall Urban Rural
Any use (logit models) β AOR β AOR β AOR
Any hospital stay -0.663 0.515 *** -0.722 0.486 *** -0.517 0.597 ***
Any outpatient ED visit -0.064 0.938 *** -0.094 0.910 *** -0.005 0.995
Any nursing facility stay 0.015 1.015 0.011 1.011 0.031 1.031
Any physician visit 0.473 1.604 *** 0.551 1.736 *** 0.340 1.405 ***
Any PCP visit 0.995 2.705 *** 1.245 3.474 *** 0.476 1.610 ***
Any specialist visit -0.036 0.964 -0.079 0.924 *** 0.065 1.067
Any HCBS use 0.126 1.134 *** 0.120 1.127 *** 0.150 1.161 ***
Any assisted living use -0.171 0.842 *** -0.251 0.778 *** 0.022 1.022
Any hospice use 0.083 1.087 * 0.032 1.033 0.217 1.243 **
Count of visits, among users only (negative binomial models) β IRR β IRR β IRR
Number of hospital stays -0.296 0.744 *** -0.277 0.758 *** -0.341 0.711 ***
Number of outpatient ED visits -0.484 0.616 *** -0.551 0.577 *** -0.335 0.715 ***
Number of physician visits -0.387 0.679 *** -0.401 0.670 *** -0.346 0.708 ***
Number of PCP visits -0.458 0.633 *** -0.468 0.626 *** -0.424 0.655 ***
Number of specialist visits -0.446 0.641 *** -0.473 0.623 *** -0.354 0.702 ***
NOTE: */**/*** Statistically significant at the 0.10/0.05/0.01 level, two-tailed test.

3.3. Minimum Data Set Results Comparing Level of Function for New Nursing Home Admissions

In Appendix Table A-15, we specify the definition and coding of select MDS measures, including demographics, physical functioning as measured by ADLs, and cognitive impairments, separately for years before 2010 (based on MDS 2.0) and years after 2010 (based on MDS 3.0). New nursing home admissions were identified in MDS data using the methods described in Section 2.3.5 and include both Medicare-reimbursed (short stay) and Medicaid-reimbursed (long-stay) admissions.

In Appendix Table A-16 we present select characteristics of newly admitted nursing home residents, by MSHO vs. MSC+ and further stratified by gender and age group, for 2011-2012. Overall, there is little or minimal difference between newly admitted nursing home residents from the MSHO group compared with those from the MSC+ group (last two data columns). For example, upon admission, 90.0 percent of newly admitted residents in the MSC+ group were totally dependent or required extensive assistance with performing at least one ADLs, compared to 92.7 percent in the MSHO group. The prevalence of severe cognitive impairment at admission was 24.3 percent for the MSC+ group and 25.2 percent for the MSHO group. A slightly higher percentage of new nursing home admissions in the MSC+ group (83.7 percent) were admitted from an acute care hospital than in the MSHO group (79.6 percent); conversely, a lower percentage in the MSC+ group (9.2 percent) were admitted from the community than in the MSHO group (13.3 percent). There is a notable racial difference among new nursing home admissions between the MSC+ and MSHO groups, with a higher percentage of newly admitted residents being minorities (non-White) in the MSC+ group (15.5 percent) than in the MSHO group (9.6 percent). (The proportion of minorities in the MSC+ program is slightly larger than in the MSHO program.)

A similar pattern was seen in 2008-2009, as shown in Appendix Table A-17. It is noteworthy that there was a higher percentage of minorities among new nursing home admissions in the MSC+ group than in the MSHO group: 6.6 percent vs. 3.2 percent in 2008-2009 (15.5 percent vs. 9.6 percent in 2011-2012). In addition, the rate of new nursing home admissions was by and large comparable between MSC+ and MSHO enrollees and it decreased in both groups from 2008-2009 (13.7 percent vs. 15.8 percent) to 2011-2012 (11.1 percent vs. 11.8 percent).

Appendix Table A-17 also reveals sharp differences in all the select characteristics between newly admitted nursing home residents among Medicare-only beneficiaries compared to MSHO or MSC+ dual eligibles. New nursing home admits from the Medicare-only group were notably healthier, as indicated by a much lower prevalence of short-term memory problems (46.1 percent for Medicare-only vs. 62.5 percent for MSC+ and 65.4 percent for MSHO), long-term memory problems (17.1 percent vs. 30.4 percent and 35.1 percent), and total dependence on others or need for extensive assistance with the five ADLs (e.g., 11.5 percent vs. 20.6 percent and 21.6 percent, with respect to eating). Moreover, new admits from the Medicare-only group were more likely to be married (41.1 percent vs. 20.0 percent for MSC+ and 18.4 percent for MSHO) and admitted from an acute care hospital (89.4 percent vs. 85.3 percent and 84.3 percent) but were much less likely to be non-Whites (0.9 percent vs. 6.6 percent and 3.2 percent). Overall, the rate of new nursing home admissions is much lower in the Medicare-only group (4.9 percent) than in the MSC+ (13.7 percent) or MSHO group (15.8 percent). Since Medicare does not cover long-term nursing facility care, we suspect that most (if not all) of the new admits from the Medicare-only group entered the facility for short-term rehabilitation under the Medicare SNF care benefit. Thus, the observed differences in their functional impairment levels as compared with the duals at the time of nursing home admission would be as expected.

3.4. Comparison of MSHO and Medicare-Only Beneficiaries

The differences between the two populations are substantial (Table 3-19). The Medicare-only beneficiaries were considerably younger, healthier, less likely to died, and had less service use on the range of outcome measures analyzed. MSHO enrollees are more similar to MSC+ enrollees than they are to Medicare-only beneficiaries.

TABLE 3-19. Comparison of Medicare-Only and Dual Eligible Beneficiaries on Selected Characteristics, 2010-2012
Characteristic Medicare-Only Dual Eligibles
Total
Dual Eligibles
MSC+
Dual Eligibles
MSHO
N % or
Mean
N % or
Mean
N % or
Mean
N % or
Mean
Age:
   65-69 1,843,622 30.5 121,696 16.8 21,935 27.6 99,761 14.4
   70-74 1,843,622 23.3 121,696 17.0 21,935 17.6 99,761 16.9
   75-79 1,843,622 18.1 121,696 15.8 21,935 13.4 99,761 16.3
   80-84 1,843,622 14.2 121,696 16.4 21,935 14.4 99,761 16.9
   85-89 1,843,622 9.0 121,696 16.1 21,935 13.4 99,761 16.7
   90+ 1,843,622 4.8 121,696 18.0 21,935 13.6 99,761 18.9
Female 1,843,622 45.4 121,696 72.4 21,935 68.4 99,761 73.3
Died during year 1,843,622 3.5 121,696 14.2 21,935 12.8 99,761 14.4
Disabilities/conditions:
   Mental illnessa 1,843,622 15.0 121,696 60.1 21,935 58.8 99,761 60.4
   Neurological disabilityb 1,843,622 7.2 121,696 48.0 21,935 43.3 99,761 49.0
   Physical disabilityc 1,843,622 16.0 121,696 31.5 21,935 32.1 99,761 31.4
   Sensory disabilityd 1,843,622 3.5 121,696 8.1 21,935 7.8 99,761 8.2
   Other medical disability or chronic diseasee 1,843,622 45.0 121,696 85.6 21,935 82.5 99,761 86.2
Service utilization:
   Any hospitalization 1,843,622 13.4 121,696 20.2 21,935 27.4 99,761 18.6
      # Hospitalizations, if any 246,356 1.40 24,067 1.96 5,621 2.46 18,446 1.80
   Any outpatient ED visit 1,843,622 11.6 121,696 30.1 21,935 30.5 99,761 30.0
      # ED visits, if any 214,311 1.43 36,541 1.94 6,639 2.85 29,902 1.74
   Any PCP visit 1,843,622 39.0 121,696 82.8 21,935 71.9 99,761 85.2
      # PCP visits, if any 717,250 4.74 100,622 7.88 15,684 11.50 84,938 7.21
   Any specialist visit 1,843,622 44.0 121,696 83.8 21,935 84.4 99,761 83.7
      # Specialist visits, if any 809,740 5.07 101,830 7.28 18,426 10.78 83,404 6.51
   Any physician visit 1,843,622 51.8 121,696 94.0 21,935 91.5 99,761 94.6
      # Physician visits, if any 953,784 7.90 114,371 13.44 20,051 19.02 94,320 12.25
   Any hospice use 1,843,622 2.0 121,696 7.9 21,935 6.7 99,761 8.1
      Any hospice, among deceased only 64,255 46.4 17,221 42.4 2,813 40.9 14,408 42.7
Calendar year:
   2010 1,843,622 32.6 121,696 32.4 21,935 31.8 99,761 32.5
   2011 1,843,622 33.2 121,696 34.1 21,935 33.1 99,761 34.3
   2012 1,843,622 34.2 121,696 33.5 21,935 35.1 99,761 33.2
County characteristics:
   Rural county 1,843,622 31.9 121,696 33.7 21,935 30.4 99,761 34.4
      # PCPs per 1,000 population 1,842,801 0.80 120,815 0.80 21,692 0.85 99,123 0.79
      % Population 65+ who do not live in community 1,842,801 4.9 120,815 5.1 21,692 4.8 99,123 5.2
      % Population 65+ who live in community with others 1,842,801 65.5 120,815 64.9 21,692 64.9 99,123 64.9
      % Population 65+ with college education 1,842,801 21.6 120,815 21.7 21,692 23.0 99,123 21.4
      % Population 65+ who are married 1,842,801 58.1 120,815 57.5 21,692 57.0 99,123 57.6
  1. Including any diagnosis for Alzheimer's disease or dementia, chronic mental illness, depression, psychosis, or schizophrenia.
  2. Including any diagnosis for neurologic impairment or Parkinson's disease.
  3. Including any diagnosis for physical impairment.
  4. Including any diagnosis for sensory impairment.
  5. Including diagnoses for selected medical disability or chronic diseases such as arthritis, chronic respiratory disease (COPD, asthma, emphysema, bronchitis), congestive heart failure, coronary heart disease, stroke, or diabetes.

3.5. Assessing the Potential Effect of Differences in Unobserved Characteristics on the Estimates of MSHO Impacts

The multivariate analyses presented above compares the health care and long-term care outcomes of MSHO and MSC+ enrollees, controlling for differences in the observed characteristics of the two groups of enrollees. To the extent there are differences in unmeasured characteristics of the enrollees that affect both their MSHO enrollment and their health care and long-term care outcomes the estimated impacts of MSHO will be biased. These omitted variables could include, for example, additional components of health and disability status, such as severity of chronic conditions and frailty, and family circumstances, such as marital status, living arrangements, and availability of informal care givers.

We use a method developed by Oster (2015) to assess the sensitivity of impact estimates to potential omitted variable bias. Oster's method draws on the explanatory power of the regression model (as measured by R2) based on the observed characteristics and the maximum potential explanatory power of the regression model, to predict the maximum potential impact of omitted variables on the estimated effect under the assumption that the selection on the observed characteristics is proportional to the selection on the unobserved characteristics.4 If the estimated effects would remain consistent in the face of the high levels of potential omitted variable bias we can be more confident in the reliability of the findings reported above.

Table 3-20 summarizes our application of Oster's method using her Stata command, psacalc. In applying the method, we consider three scenarios for the maximum R2 possible in a study of health care and long-term care outcomes: 0.30, 0.50 and 0.70. As would be expected, the higher the potential R2 for the analysis the greater the potential effects of omitted variables on the estimate of the effect of MSHO. For example, we estimate a 10.2 percentage point reduction in the probability of an inpatient stay due to MSHO. Oster's bounding method finds that that reduction would likely be even larger if we were able to control for omitted variables, ranging from 12.8 percentage points if the maximum R2 for the model were 0.30-18.2 percentage points if the maximum R2 were 0.70. In all three R2 scenarios, the potential effect of omitted variable bias on the MSHO impact estimate would be to produce a stronger negative effect. This suggests that the reduction in inpatient use reported above would likely hold up in the face of omitted variable bias and, in fact, our analysis may underestimate the impact of MSHO on inpatient stays relative to MSC+ for otherwise similar individuals based on unobserved characteristics.

TABLE 3-20. Assessment of Potential Impact of Omitted Variable Bias to Change Estimated MSHO Impacts on Utilization
Outcome Estimate of
MSHO Impact
Estimates of Potential MSHO Impact After
Accounting for Possible Omitted Variable Bias
Under Alternative Assumptions About Maximum Possible R2
Potential Implication
of Omitted
Variable Bias
Max R2
=0.3
Max R2
=0.5
Max R2
=0.7
Any inpatient stay -0.102 *** -0.128 -0.155 -0.182 Stronger effect
Any ED visit -0.012 *** -0.031 -0.049 -0.066 Stronger effect
Any nursing facility use 0.002 -0.001 -0.040 -0.079 Negative effect
Any HCBS use 0.029 *** 0.015 0.003 -0.009 Eliminate positive effect
Any assisted living use -0.021 *** -0.091 -0.154 -0.218 Stronger effect
Any PCP visit 0.140 *** 0.149 0.160 0.172 Stronger effect
Any specialist visit -0.004 0.000 0.005 0.010 Little change
Any physician visit 0.027 *** 0.019 0.012 0.005 Eliminate positive effect
Number of inpatient stays -0.255 *** -0.159 -0.075 0.009 Eliminate negative effect
Number of ED visits -0.468 *** -0.460 -0.448 -0.436 Little change
Number of PCP visits -0.445 *** -0.445 -0.445 -0.445 Little change
Number of specialist visits -0.388 *** -0.355 -0.299 -0.243 Reduce negative effect
Number of overall physician visits -0.310 *** -0.299 -0.267 -0.235 Reduce negative effect
SOURCE: RTI analysis of 2010-2012 Minnesota claims and encounter data.
*/**/*** Statistically significant at the 0.10/0.05/0.01 level, two-tailed test.

Overall, the findings in Table 3-20 suggest that the impacts of MSHO on the key health care and long-term care outcomes are robust to alternative assumptions about the extent of omitted variable bias, with the potential impacts likely even larger if we were able to control for those variables. This would include the potential for even greater reductions in inpatient stays, ED use, and assisted living, and greater increases in visits to PCPs under MSHO relative to MSC+.

In some cases the potential omitted variable bias would have little effect--any specialist visit, number of ED visits, and number of PCP visits. In a few cases it could potentially change the direction of the estimate of the impact of MSHO on the outcome. Controlling for the unobserved characteristics could potentially eliminate the effect of MSHO on any physician visit, reduce the negative effect of MSHO on the number of specialist visits, and reduce the negative effect of MSHO on the number of overall physician visits.

View full report

Preview
Download

"MNmclda.pdf" (pdf, 2.63Mb)

Note: Documents in PDF format require the Adobe Acrobat Reader®. If you experience problems with PDF documents, please download the latest version of the Reader®