Constrained Innovation in Managing Care for High-Risk Seniors in Medicare + Choice Risk Plans. D. Analysis Methods

01/01/2002

The analysis of case study information was based on the site visit reports and follow-up discussions with key staff at the four MCOs. Using the VIP and I-CAN frameworks, we looked for ways in which these organizations had attempted to meet the diverse and often extensive needs of high-risk seniors. In particular, we compared methods used to identify high-risk seniors and to then coordinate and manage their care.

Much of the survey data analysis is based on descriptive statistics (averages and cross-tabulations). In general, we present weighted means in order to provide as close a measure as possible for the target populations being described. When comparing the three organizations whose members were surveyed, we use regression analysis to control for the underlying differences in the characteristics of the beneficiaries enrolled in the three MCOs. These regressions are not weighted but do control for factors that reflect the probability of selection into the survey and survey nonresponse. In particular, the following control variables were used for most regressions:

  • Plan (Kaiser Colorado, Aspen, Keystone East)
  • Survey Subgroup (care management, advanced age, hip, stroke)
  • Respondent Type (sample member, proxy, representative proxy)
  • Gender (male, female)
  • Age (age 65-74, age 75-84, age>84)
  • Race (white, black, other)
  • Education (no high school diploma, high school graduate, at least some college)
  • Income (less than $10,000, $10 to $20,000, more than $20,000)
  • Self-Reported Health Status (excellent, good, fair, poor)
  • Medicaid (whether sample member reported having Medicaid coverage)
  • Marital Status (whether married)
  • Residential Status (whether sample member lives alone)
  • Chronic Conditions (2 or fewer conditions, 3 or 4 conditions, 5 or more conditions)
  • Dementia (whether the sample member has Alzheimer’s disease or other dementia)
  • ADL Limitations (no limitations, limited in 1 or 2 activities, limited in 3 or more activities)

Means for these control variables, and their variation among the three MCOs included in the survey, are presented in Table II.6. The race categories used in the regressions (Table II.6) differ from those presented in Table II.4, because we were unable to control for ethnicity as a result of the small number of Hispanics. Ethnicity is therefore ignored as a control variable, and Hispanics are classified into their corresponding race category. Similarly, the regressions use only 3 categories to describe the number of chronic conditions. We combined seniors who reported no chronic conditions with those who reported fewer than two such conditions, because the group with no conditions was fairly small. Finally, we control for the presence of dementia, because we believe that that controlling for the number of chronic conditions alone will not capture the effect of relatively high rates of dementia in the Aspen and Kaiser samples, and the absence of an explicit control for this measure could lead to omitted variables bias. In the analysis of the hip fracture and stroke sample (see Chapter VI), the sample was too small to permit us to control for all the variables included in this list. We therefore developed a slightly smaller set of control variables that are listed in Appendix Table A.1.

When one of these control variables was missing for a sample member, we imputed the mean for the full sample. In addition, when a variable was missing for more than five percent of a sample, we added an extra control variable that indicated whether or not we had imputed for each sample member. This extra control variable enables us to control for any characteristics that are systematically related to whether the variable was missing for a sample member. We never imputed values for any of the variables used as outcomes in the regression analysis.

TABLE II.6. Variability of Sample Characteristics Across MCOs
(Percentage and Their Standard Errors)
  Total MCO
  Aspen     Kaiser Colorado     Keystone East  
Sample Indicatorsa
   Care management 30.9 (0.1) 21.3 (0.0) 9.5 (0.1) 38.1 (0.1)
   Advanced age   68.8 (0.1)   74.2 (0.1) 91.5 (0.1) 61.6 (0.1)
   Hip fracture 1.6 (0.0) 2.1 (0.0) 1.7 (0.1) 1.5 (0.0)
   Stroke 4.4 (0.0) 3.8 (0.1) 1.0 (0.0) 5.5 (0.0)
Plan
   Aspen 6.8 (0.0)   100.0 (0.0)   0.0 (0.0) 0.0 (0.0)
   Kaiser Colorado 21.1 (0.1) 0.0 (0.0) 100.0 (0.0) 0.0 (0.0)
   Keystone East 72.1 (0.1) 0.0 (0.0) 0.0 (0.0) 100.0 (0.0)
Respondent Type
   Sample member 64.0 (1.6) 58.0 (2.6) 66.9 (3.0) 63.7 (2.1)
   Proxy 22.5 (1.5) 24.7 (2.3) 22.9 (2.7) 22.2 (1.8)
   Representative proxy 13.5 (1.2) 17.2 (2.0) 10.2 (1.9) 14.1 (1.6)
Age at Time of Interview
   65 to 74 15.8 (0.6) 4.5 (0.2) 2.8 (0.3) 20.7 (0.9)
   75 to 84 14.2 (0.6) 13.9 (0.3) 5.7 (0.3) 16.7 (0.9)
   85 or older 70.0 (0.2) 81.6 (0.2) 91.8 (0.1) 62.5 (0.3)
Gender
   Male 33.6 (1.6) 26.3 (2.2) 27.5 (2.8) 36.1 (2.0)
   Female 66.4 (1.6) 73.7 (2.2) 72.5 (2.8) 63.9 (2.0)
Raceb
   White 83.8 (1.3) 95.7 (1.1) 92.8 (1.6) 80.1 (1.7)
   Black 12.1 (1.1) 1.7 (0.7) 1.6 (0.8) 16.1 (1.6)
   Other 4.2 (0.7) 2.6 (0.9) 5.7 (1.5) 4.9 (2.2)
Education
   Did not complete high school   46.0 (1.7) 41.7 (2.6) 37.8 (3.0) 48.9 (2.2)
   High school graduate 33.9 (1.6) 32.6 (2.4) 27.2 (2.8) 36.1 (2.1)
   At least some college 20.1 (1.2) 25.7 (2.3) 34.9 (3.0) 15.0 (1.5)
Total Household Incomee
   Less than $10,000 35.3** (1.9) 35.5** (2.9) 31.7** (3.3) 36.4** (2.4)
   $10,000 to less than $20,000 40.1** (1.9) 39.9** (2.9) 31.3** (3.3) 42.8** (2.4)
   $20,000 or more 24.6** (1.5) 24.6** (2.4) 37.0** (3.4) 20.8** (1.9)
Health Assessment
   Excellent 30.2 (1.6) 24.6 (2.3) 31.7 (3.0) 30.3 (2.0)
   Good 32.7 (1.5) 37.3 (2.2) 30.6 (2.7) 32.9 (1.8)
   Fair 25.7 (1.5) 26.1 (2.3) 25.9 (2.8) 25.6 (1.9)
   Poor 11.4 (1.1) 11.9 (1.6) 11.8 (2.0) 11.2 (1.4)
Has Medicaide 19.6* (1.4) 27.9* (2.4) 18.2* (2.5) 19.2* (1.8)
Married 30.3 (1.4) 21.8 (1.9) 25.2 (2.7) 32.5 (1.8)
Lives Alone 41.5 (1.7) 59.8 (2.8) 58.6 (3.2) 35.1 (2.2)
ADL Limitationsc
   Two or fewer 91.1 (1.0) 92.5 (1.6) 90.4 (2.0) 91.1 (1.2)
   Three to five 8.9 (1.0) 7.5 (1.6) 9.6 (2.0) 8.9 (1.2)
Number of Chronic Conditionsd
   Two or fewer 49.8 (1.7) 48.8 (2.6) 42.5 (3.1) 52.1 (2.1)
   Three or four 35.6 (1.6) 34.4 (2.4) 41.1 (3.1) 34.0 (2.0)
   Five or more 14.6 (1.1) 16.7 (1.9) 16.4 (2.3) 13.9 (1.4)
Alzheimer's or Other Dementia 9.9 (1.0) 16.8 (2.1) 12.5 (2.1) 8.5 (1.2)
SOURCE: Telephone survey of 1,657 high-risk seniors from three managed care organizations, conducted between March and December 1999 by MPR.
NOTE: Values are percentages, with standard errors in parentheses.
  1. Percentages will sum to more than 100, because some seniors are in multiple sample frames.
  2. Race categories are different from those in Table II.4, because we were unable to control for ethnicity as a result of the small number of Hispanics. Ethnicity is therefore ignored as a control variable, and Hispanics are classified into their corresponding race category.
  3. ADL limitations involve the need of help or supervision with the five activities of daily living: bathing, eating, dressing, transferring, and toileting. These questions were asked of the 1,399 community residents only.
  4. People were asked whether they had been diagnosed with any of the following chronic conditions: arteriosclerosis, hypertension, heart attack, other heart disease, previous stroke, depression, cancer, diabetes, arthritis, asthma, previous hip fracture, or Alzheimer’s or other dementia.
  5. Nonresponse was high for the income question. In general, fewer than 80 percent of respondents answered these questions. Nonresponse was also something of a problem for the Medicaid question, where between 95 and 80 percent of respondents answered it.

* 5 to 20 percent nonresponse.
** Over 20 percent nonresponse.

View full report

Preview
Download

"constrai.pdf" (pdf, 32.58Kb)

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®