Empirical Specification and Estimation
Even though utilization was similar across the groups, our descriptive results indicated there were significant differences in satisfaction with services and unmet needs between dual working age SSI beneficiaries in Medicare FFS and non-dual OHP beneficiaries, and in some cases, between the Medicare managed care and the Medicare FFS groups. On the other hand, there were few differences in satisfaction and unmet needs between the two totally managed groups: dual eligibles in the Medicare managed care group and the non-dual beneficiaries. However, as we saw in Tables 2 and 3, the Medicare managed care and fee-for-service dual survey respondents varied substantially by socio-demographic status and by some health and functional measures from their non- dual counterparts, characteristics that can affect need for care and satisfaction with care. We used multivariate regression analysis to determine the individual influence of each
factor and to isolate the effects of dual status and Medicare managed care enrollment on utilization, satisfaction and perceived unmeet need for service.
Using logistic regression, we examined three domains: utilization of various services (10 models), including both outpatient and inpatient services, unmet needs and uncovered services (7 models), and various measures of satisfaction with care (12 models).
To isolate the effect of additional Medicare coverage for OHP beneficiaries we used a dummy variable denoting dual eligibility. To look at the additional effect of Medicare managed care, we created a variable set to one for all respondents who, according to the Oregon enrollment files, were enrolled in a Medicare managed care plan as well as in a Medicaid managed care plan at the time of the survey.
Covariates include various demographic factors as well as multiple measures of health and functional status. We include dummy variables for age groups 30-39, 40-49 and 50-65 years old. The youngest group of 19-29 year old respondents is the omitted group. We created dummy variables for gender (set to one for females), race (set to one for whites), employment status (set to one for employed) and education (set to one for high school graduates). We include living arrangement dummy variables for group residence (residents in all types of group housing vs. everybody else) and living alone (vs. married respondents and other respondents who did not live alone), since these arrangements are known to affect service utilization and general perception of well-being, and could indirectly affect satisfaction with services. Since managed care is not uniformly offered throughout all counties, we also incorporated geographic residence
variables: a dummy for residents of urban areas outside Portland and one for rural areas (metropolitan Portland area residents are the reference group omitted).
We included several health and functional status variables. We created two dummy variables for principal disability category merged from the Social Security Administration records (or self-reported, for those whose reason for disability was missing from the Social Security Administration records): one for those disabled due to mental illness and one for those with developmental disabilities. Those with physical disabilities are the reference group omitted. We deleted from the sample the two individuals for whom we had no information about their primary disabling condition. We created separate dummy variables based on self-rated health: poor and fair vs. all other self-reported health ratings. Functional limitations are described by a set of three dummy variables indicating the number of ADL limitations (defined as having any level of difficulty): 1 - 2 ADLs, 3 - 4 ADLs and a maximum of 5 - 6 ADLs, with those without any ADL limitations as the omitted group.
The model for inadequate help with ADLs is only estimated for respondents with at least one ADL limitation; those with 5-6 ADLs are the omitted group. We also created a dummy variable indicating the use of the separately administered community based long–term care services. This group may be more frail, but also has access to additional advocacy on their behalf as their case manager from Senior and Disabled Services Division has direct access to care coordinators at the managed care plans (see Walsh, French and Bentley,1999 and Walsh, Kulas and Khatutsky, 2000 for further information about the links between the acute and long-term care systems).
Satisfaction with mental health services was estimated only for those who had used a mental service in the 3 months preceding the survey.
Multivariate Results Service Utilization
Table 9 presents results for the logistic regressions estimating the odds of using various inpatient and outpatient services for this sample, all of whom were enrolled in Medicaid managed care plans. Neither dual eligibility nor enrollment in a Medicare managed care plan were significant predictors of any of the utilization measures. Demographic and health and functional status measures did affect the odds of using a variety of health services.
Hospital-Based Services (inpatient and emergency room)
The likelihood of a non-psychiatric hospitalization was related to reason for disability, self-rated health and use of long-term care services. Those reporting mental illness (p<0.05) or developmental disability (p<0.10) were about one third less likely than those disabled due to a physical disability to report at least one hospitalization in the last twelve months. Those reporting poor health status were four times more likely than others to report hospitalization (p<0.01), and those reporting fair health status were almost 3 times more likely to be hospitalized (p<0.01). Long-term care users had twice the likelihood of a hospitalization than those who did not receive long-term care services (p<0.01).
Variable
Dual Eligible 0.97 0.68 1.01 0.95 0.82 0.94 1.03 0.87 1.05 0.97
Medicare Managed Care 1.14 1.2 1.08 0.75 1.13 0.79 1.24 1.24 1.18 0.87
Demographics
Female 1.04 1.07 1.32 * 1.30 * 1.61 ** 1.59 ** 1.29 # 1.03 1.08 1.12
White 0.82 0.61 0.80 0.78 0.69 * 0.76 1.18 0.95 0.74 * 0.87
Employed 0.72 0.07 * 0.67 # 1.44 # 1.33 1.88 ** 1.20 1.31 2.03 ** 1.23
High School graduate 0.96 1.37 0.95 1.07 1.14 1.36 # 0.91 1.15 1.60 ** 1.04
Urban 0.75 0.63 0.69 * 1.04 0.96 0.98 1.06 0.92 0.92 1.22 Rural 0.73 # 0.67 1.07 0.90 1.14 1.14 1.37 # 0.77 # 0.91 1.19 Group Residence 0.88 1.42 0.92 3.09 ** 1.83 ** 1.46 2.62 ** 1.32 2.02 ** 1.64 Live alone 0.79 1 1.12 1.05 1.21 1.14 1.67 ** 1.10 1.14 1.34 Age 30-39 1.36 0.91 1.06 1.37 # 1.04 0.99 1.19 1.61 * 1.42 # 1.18 Age 40-49 1.21 0.38 # 0.85 1.75 ** 1.34 1.01 0.97 1.55 * 1.26 1.56 Age 50-65 1.63 0.41 # 0.54 * 2.25 ** 1.20 1.17 0.67 # 1.31 0.63 * 2.21 Health Status Mentally Ill 0.68 * 9.91 ** 0.99 1.07 0.60 ** 0.41 ** 9.55 ** 0.98 1.24 1.12 Developmentally Disabled 0.68 # 1.76 0.89 1.46 * 0.57 ** 0.50 ** 2.03 ** 0.60 ** 1.01 0.53
Health Status Poor 4.02 ** 1.12 2.35 ** 1.64 ** 2.33 ** 2.01 ** 1.03 2.20 ** 1.00 2.14
Health Status Fair 2.90 ** 1.04 1.55 ** 1.20 2.05 ** 2.29 ** 0.88 1.58 ** 0.97 1.71
LTC User 2.01 ** 2.22 * 1.29 1.11 1.06 1.11 1.01 1.48 ** 0.78 # 1.18 1-2 ADLs 0.85 1.71 0.95 1.23 1.40 * 1.41 # 1.07 1.16 1.21 1.61 3-4 ADLs 1.03 1.3 1.07 0.99 1.34 1.44 0.87 1.58 ** 1.08 1.92 5-6 ADLs 1.17 0.52 1.14 1.16 1.53 # 1.70 # 0.73 1.79 ** 1.08 2.08 N 1,555 1,612 1,603 1,612 1,592 1,612 1,604 1,612 1,612 1,612 NOTES:
** Significantly different at 0.01 level. * Significantly different at 0.05 level. # Significantly different at 0.10 level.
Emergency Hospital Psych Table 9 Hospital Routine
Physical in Past 3 months Specialist Visit Dentist Visit Physician Visit
Dru in Past 12 months
General Medical
Logistic Regression Results for Utilization of Health Services
Provider Visit 3m
Mental Health Prescrip
Hospital Based Services Outpatient Services
Physician Visit Room
Psychiatric hospitalization was most strongly related to having a mental illness as the stated reason for disability, with this factor increasing the odds of a psychiatric hospitalization almost 10 fold (p<0.01). In contrast, those who were employed had a very small likelihood of a psychiatric hospitalization (0.07, p<0.05). Individuals age 40-65 had less than half the likelihood of psychiatric hospitalization than individuals under 30 years of age (p<0.10). Long-term care use was associated with a twofold increase in the odds of reporting a psychiatric hospitalization (p<0.05), perhaps indicating that some of the more severely mentally ill receive long-term care services.
Demographic and health status characteristics, but not the reason for disability, predicted emergency room use. Women were about one third more likely to report emergency room use than men were (p<0.05). The likelihood of ER use decreased for those who were employed (0.67, p<0.10), those living in urban areas (0.69, p<0.05), and individuals age 50-65 (0.54, p<0.05). Those reporting poor or fair health status had an increased likelihood of an ER visit (2.35 and 1.55, p<0.01). ADL limitations and long- term care use were not significant factors in this model.
Outpatient Services
Utilization of most outpatient services was predicted by a combination of demographic and health status characteristics. The likelihood of reporting a routine physical was higher for women (1.30, p<0.05), the employed (1.44, p<0.10), those in group residences (3.09, p<0.01), and for those age 30 and over. Individuals disabled due to developmental disabilities also had an increased likelihood of having a routine physical (1.46, p<0.05), as did those who reported poor health status (1.64, p<0.01).
Seeing a physician in the last 12 months for any reason, was also related to a variety of demographic and health status measures, although health status played more of a role here than in the likelihood of a routine physical. Female gender, employment and having graduated high school all increased the odds of a physician visit (1.59, p<0.01; 1.88, p<0.01; and 1.36, p<0.10). However, there was no impact of urban residence, type of residence or living arrangement or age. Being disabled due to a mental illness or developmental disability reduced the odds of a physician visit by about half (0.41, p<0.01; 0.55, p<0.01). General health ratings of poor and fair doubled the likelihood of a physician visit (2.01, p<0.01; and 2.29, p<0.01). ADL impairment approached significance in increasing likelihood of a physician visit in the last 12 months.
The factors affecting the likelihood of a physician visit in the last three months were similar to those related to a physician visit in the last year. However, being white significantly decreased the likelihood of a physician visit in the last three months by almost a third, and employment status was not statistically significant in this model. Group residence almost doubled the likelihood of having a recent physician visit (1.83, p<0.01).
The likelihood of reporting an outpatient mental health visit in the last three months was a function of demographic characteristics and reason for disability, but is one area in which the other health status characteristics were not significant. Women and those living in rural areas had about a one third increased likelihood of a mental health visit, and the oldest respondents, age 50-65, had about a one third lower likelihood of a mental health visit, all significant only at the ten percent level. Group home residence almost tripled the likelihood of a mental health visit (2.62, p<0.01), and living alone also
increased the likelihood of a mental health visit significantly (1.67, p<0.01). Not surprisingly, disability due to a mental illness greatly increased the odds of a mental health visit (9.55, p<0.01). Those disabled due to a developmental disability also had an increased likelihood of a mental health visit (2.03, p<0.01).
Specialty care utilization was to some extent related to demographic variables, but was most strongly affected by the health status variables. Rural residence decreased the likelihood of seeing a specialist (significant at the 10 percent level), and those ages 30-49 had an increased likelihood of seeing a specialist (significant at the 5 percent level). Poor or fair health status, long-term care use, and 3 or more ADL impairments all increased the odds of seeing a specialist by 50, and sometimes more than 100%, significant at the 1 percent level.
In contrast, dental care utilization was most related to demographic characteristics. While being white slightly decreased the likelihood of a dental visit (0.74, p<0.05), being employed or a high school graduate greatly increased the likelihood of a dental visit in the last year (2.03, p<0.01, and 1.60, p<0.01, respectively). Group residents also increased the likelihood of a dental visit two fold (2.02, p<0.01). Compared to respondents under age thirty, those age 30-39 had an increased likelihood (1.42, p<0.10), and those age 50-64 had a decreased likelihood of receiving dental care (0.63, p<0.05).
Prescription drug utilization was associated with group residence (1.64, p<0.10), and age 50-64 (2.21, p<0.01), as well as reason for disability and other health status variables. Those with developmental disabilities had a decreased likelihood of
prescription drug use (0.53, p<0.05), while poor and fair health status, and increasing ADL impairment all increased the likelihood of prescription drug use in the last year.
Access and Satisfaction with General Medical Services.
We estimated a total of 12 models of aspects of access to and satisfaction with services, covering such domains as overall satisfaction with services, complaints against OHP, and satisfaction with individual features of health services. In many of these models, our key dependent variables were significant: dual eligibility significantly increased satisfaction compared to non-dual status, while being a dual eligible in a Medicare managed care plan significantly decreased satisfaction compared to duals in fee-for-service and non-duals combined. The findings for three general satisfaction measures are reported in Table 10, and for nine specific aspects of care in Table 10. In this section, rather than go through each model in detail, we highlight the significant variables across subgroups of these models in each table.
Overall Satisfaction
Table 10 describes results for overall satisfaction with care, with the dependent variable equal to one in each model representing a negative rating or dissatisfaction. As seen from the odds ratios presented, dual eligibility, i.e., having Medicare coverage as well as OHP, decreased the odds of being dissatisfied with the general quality of services almost 50 percent, and decreased the probability of respondents filing or wanting to file a complaint against OHP by more than 30 percent. Although not significant in the other
Variable
Dual Eligible 0.57 # 0.68 * 0.74
Medicare Managed Care 1.72 # 1.85 ** 1.77 *
Demographics
Female 1.02 1.25 1.24
White 0.97 0.92 0.97
Employed 1.16 1.00 0.89
High School graduate 0.86 1.07 1.34 #
Urban 1.00 1.08 0.89 Rural 0.93 1.24 0.95 Group Residence 1.05 0.74 0.79 Live alone 1.38 0.62 ** 0.89 Age 30-39 0.95 0.74 0.98 Age 40-49 0.70 0.75 0.91 Age 50-65 0.44 * 0.52 * 0.43 ** Health Status Mentally Ill 1.51 # 1.63 ** 1.37 # Developmentally Disabled 0.71 0.73 0.89
Health Status Poor 2.43 ** 2.31 ** 2.54 **
Health Status Fair 2.22 ** 1.66 ** 1.45 *
LTC User 0.72 1.07 0.90 1-2 ADLs 1.01 1.44 # 1.39 # 3-4 ADLs 1.23 1.59 * 1.77 * 5-6 ADLs 2.11 ** 2.32 ** 2.22 ** N 1,612 1,609 1,514 NOTES:
* Significantly different at 0.05 level. # Significantly different at 0.10 level. ** Significantly different at 0.01 level.
OUTPUT: P2169ahm and P2SUR183
Dissatisfied with Overall Quality of Care
Very/ Somewhat Hard to Get Care Table 10
Logistic Regressions for Overall Satisfaction with Services Complaint/ Wanted to Complain
model in this table, the direction of the odds ratios is the same, i.e., dual eligibles were less likely to be dissatisfied compared to the non-dual eligible group.
However, the odds ratios for dual eligibles who were also Medicare HMO enrollees, go in the opposite direction and cancel out the relative satisfaction associated with dual eligibility. Compared to the dual eligibles in fee-for-service and non-duals together, the Medicare managed care group was 70 percent more likely to be dissatisfied with overall quality of care (p<0.10), 85 percent more likely to have filled or wanted to file a complaint (p<0.01), and 77 percent more likely to report that it was somewhat or very hard to get care (p<0.05). We also calculated the net effect of being in the Medicare managed care group compared to the non-dual eligible group for each of these models3. The odds ratios for the Medicare managed care group compared to the non-dual group were 0.98 for overall dissatisfaction, 1.26 for complaining or wanting to complain, and 1.31 for reporting it was hard to get their care needs met. In other words, satisfaction for Medicare managed care enrollees is at the same level or lower than for non-dual eligibles in OHP plans.
Demographic characteristics had less impact on satisfaction than we observed in the previous table reporting utilization. Only age was associated with dissatisfaction with overall care, with those age 50-64 reporting less dissatisfaction compared to the omitted group, age 19-29 (0.44, p<0.05). The likelihood of having lodged a formal complaint or considered complaining, was lower among those who reported living alone (0.62, p<0.01), and for the older age group (significant for those age 50-64, 0.52, p<0.05). High school graduates were somewhat more likely to state it was hard to get the care they
needed (1.34, p<0.10), while those who were in the oldest group were significantly less likely to say it was hard to get the care they needed (0.43, p<0.01).
Disability due to mental illness increased the likelihood of dissatisfaction in these models. This variable was associated with increased dissatisfaction with overall quality of care (1.51, p<0.10), the complaint variable (1.63, p<0.01), and difficulty getting care needs met (1.37, p<0.10). There were no significant relationships with satisfaction in these models for those disabled due to developmental disabilities, compared to the omitted group of persons disabled due to physical conditions.
Dissatisfaction was most strongly affected in these models by health status and increased ADL impairment. Being in poor or fair health significantly increased the likelihood of being dissatisfied in all of these measures, with those in poor health status being 2-3 times more likely to be dissatisfied than other respondents. Those with 5-6 ADL impairments were twice as likely as those without ADL impairments to be dissatisfied with the overall quality of care, to have complained or thought about complaining, and to report that it was hard to get their care needs met.
Dissatisfaction with Aspects of Care and Individual Services
Table 11 presents results from logistic regressions estimating satisfaction with individual general medical services, and access to special equipment, home health aides and rehabilitation services. Dual eligibility is significant in 3 of these models, and associated with decreased likelihood of being dissatisfied, while dual eligibles enrolled in a Medicare HMO were significantly more likely to report dissatisfaction in almost every
Variable
Dual Eligible 0.72 # 0.89 0.77 0.89 0.69 * 0.79 0.61 * 0.79 0.92
Medicare Managed Care 1.87 ** 1.50 # 1.10 1.92 ** 2.05 ** 1.96 ** 2.14 ** 1.48 1.09
Demographics
Female 1.05 1.16 1.07 1.14 0.99 0.88 1.32 # 1.27 1.06
White 0.76 0.79 0.85 1.21 0.80 0.73 # 0.85 0.71 0.80
Employed 0.99 1.10 1.47 # 1.10 1.20 1.00 1.28 0.84 0.77
High School graduate 0.82 1.04 0.89 0.74 * 0.96 0.86 1.33 # 0.98 1.09
Urban 0.76 1.04 0.96 1.17 0.96 0.95 1.03 0.85 1.01 Rural 1.12 1.23 0.80 1.17 1.24 1.01 1.38 # 1.13 1.27 Group Residence 0.76 1.02 0.78 1.18 0.93 0.80 1.30 1.54 0.78 Live alone 1.34 # 1.19 1.31 # 1.04 1.25 1.21 1.05 1.15 0.87 Age 30-39 0.95 1.07 0.72 0.88 0.95 0.93 0.85 0.61 0.97 Age 40-49 0.75 0.93 0.86 0.85 0.97 0.98 0.79 0.56 # 0.93 Age 50-65 0.48 ** 0.55 * 0.47 0.50 ** 0.55 * 0.59 # 0.45 ** 0.32 ** 0.65 Health Status Mentally Ill 1.15 1.18 1.27 1.27 1.69 ** 1.70 ** 1.12 1.64 * 1.44 # Developmentally Disabled 0.86 0.75 1.05 0.80 1.15 1.03 0.99 0.81 1.43
Health Status Poor 2.90 ** 2.12 ** 1.71 ** 2.17 ** 1.94 ** 1.87 ** 2.21 ** 1.96 ** 1.51 #
Health Status Fair 2.41 ** 1.62 ** 1.63 ** 1.91 ** 1.91 ** 1.84 ** 2.19 ** 2.20 ** 1.36
LTC User 0.87 0.89 1.01 0.85 0.91 0.73 0.82 0.74 0.84 1-2 ADLs 0.79 0.76 1.14 0.98 1.15 1.14 0.76 1.21 1.48 # 3-4 ADLs 0.92 0.93 0.82 0.84 1.18 1.19 1.23 1.32 2.09 ** 5-6 ADLs 1.56 # 1.61 * 1.39 1.28 1.52 # 1.58 # 1.76 * 1.64 # 2.75 ** N 1,562 1,220 1,375 1,583 1,408 1,577 1,113 669 809 NOTES:
** Significantly different at 0.01 level. * Significantly different at 0.05 level. # Significantly different at 0.10 level.
Home Care Availabiliy of
Availability of Rehabilitation
Services Easy to See a Dr Care After Information Hours by Phone Time with Dr.Availability ofSpecialist Questions Answer
Availability of Special Equipment
Table 11
Logistic Regressions for Satisfaction with Individual Services
Dissatisfation with Regular Servieces
model compared to all other respondents (duals in fee-for-service and non-duals combined).
Survey respondents with dual status were almost 30 percent less likely to be dissatisfied with the availability of physicians (0,72, p<0.10) and access to specialists (0.69, p<0.05), as well as 40 percent less likely to criticize the availability of special equipment (0.61, p<0.05). Though not significant, the direction of the odds ratios in the other models is consistent with these examples. The effect of enrollment in a Medicare HMO simultaneous with OHP managed care was strongly associated with dissatisfaction in many of these models, often about twice as likely to be dissatisfied, significant at the one percent level. Respondents with their Medicare benefits managed by an HMO were significantly more likely to be dissatisfied with availability of physician services (1.87, p<0.01), access to care after hours (1.50, p<0.10), the time spent by a doctor with them on each appointment (1.92, p<0.01), access to specialists (2.05, p<0.01), their doctors’ ability to answer their questions (1.96, p<0.01), and access to special equipment (2.14, p<0.01). The increase in dissatisfaction for the Medicare managed care group is sustained even in comparison to the non-duals alone, based on our calculations of the net effect. Compared to non-duals, the Medicare managed care enrollees were roughly thirty percent more likely to file or want to file a complaint or report that it is hard for them to get care.
Few demographic variables had any impact in these models, although the oldest group consistently reported less dissatisfaction compared to the other groups. Reason for disability was only significant in a few of the models, with those who were disabled due to mental illness more likely to report dissatisfaction with access to specialists, their
doctors’ ability to provide answers to their questions, and access to home health aides. Self-rated health was highly significant in all of these models, with those reporting poor or fair health generally twice as likely to report dissatisfaction in each model, significant at the one percent level. Receipt of community long-term care services did not affect satisfaction at all. Those reporting difficulty with all or most ADLs (5-6 ADLs) had an increased probability (significant at the 10 percent level) of being dissatisfied with