Multivariate Analysis of Patterns of Informal and Formal Caregiving among Privately Insured and Non-Privately Insured Disabled Elders Living in the Community. A. Use of Long-Term Care Services

The descriptive analysis showed that the privately insured disabled population receives an average of two hours more per day of ADL and IADL assistance than does the disabled population without private insurance -- 65 hours per week compared to 51 hours (DHHS & RWJ, 1999). Much of this is attributable to the receipt of formal services. Second, while about 63% of all ADL assistance to the privately insured is provided by formal caregivers, the figure is only 30% for the non-privately insured. Third, the non-insured population is much less likely to use formal caregivers for IADLs; in fact, informal caregivers provide 80% of IADL assistance to the non-privately insured, whereas only 45% of IADL assistance is provided informally to those with private insurance.

To test whether or not the presence of LTC insurance has an effect on total service use even after controlling for other variables, we employed the sample selection technique outlined in Appendix 1. Table 1 summarizes the results for three separate equations. The first is a probit analysis for insurance ownership. It is based on the entire sample and estimates the probability that an individual will have private LTC insurance. The coefficients derived from this equation are used to calculate the Inverse Mills Ratio (IMR). The IMR is entered as a regressor in the second equation. This equation estimates the weekly hours of care exclusively for the Insured Panel. By including the IMR in this equation, we are controlling for selection effects on service utilization that are associated with being a LTC insurance policyholder. The third and final equation is estimated from the combined sample -- insured and non-insured -- and includes a dummy variable for having insurance. Again, the dependent variable is the total number of weekly hours of ADL and IADL assistance received by the disabled individuals.7

As shown in the results of the second equation, the coefficient on the IMR was not found to be significant. This indicates that selection bias is not present in the model.8, 9

TABLE 1. Probit and Regression Model for Expected Number of Weekly Hours of ADL and IADL Care Received
(standard errors in parentheses)
Variable Mean Value
(Total Sample)
Having/
Not-Having
Insurance
(n=1,805)
Insured Only
Total Weekly
(n=467)
Total Sample
Total Weekly
(n=1,822)
Age 81 -.0134***
(.004)
-.390
(.264)
.464***
(.122)
Gender
(0=Male)
.68 .05683
(.0707)
-4.929
(3.57)
-2.278
(1.969)
Living Alone
(0=No)
.26 .38512***
(.079)
-15.113***
(4.43)
-6.863
(2.255)
Education
(0=less than high school)
.30 .7793***
(.072)
----- -----
Having Children nearby
(0=No)
.64 -.01984
(.069)
----- -----
(0=<3)
.59 ----- 25.118***
(3.82)
26.891***
(2.156)
(0=<6)
.69 ----- 7.525
(4.79)
25.782***
(2.289)
Self-assessed health
(0=excellent, good or fair)
.32 ----- 2.330
(3.53)
7.331***
(2.026)
Cognitive Impairment
(0=No)
.54 ----- 8.971***
(3.34)
5.015***
(1.898)
Presence of Informal Caregiver
(0=No)
.86 -.3339***
(.089)
7.237***
(4.57)
6.811***
(2.767)
LTC Insurance
(0=None)
.30 ----- ----- 10.856***
(3.636)
Income \$22,659 .0354***
(.002)
-2.91
(.233)
.370***
(.097)
LTC Insurance Income \$11,4421 ----- ----- -.362***
(.115)
Medicaid as a payer
(0=No)
.04 ----- ----- 13.462***
(4.698)
(0=No)
.18 ----- 4.993
(3.71)
10.771***
(2.424)
Inverse Mills ratio .40 ----- 9.786
(8.26)
----
Constant ----- -1.078***
(.361)
68.822***
(22.262)
-35.376***
(10.270)
Adjusted R2 ----- ----- 20.5% 31.7%
Significance Level * p=.10; ** p=.05; *** p=.01.
1. This includes 70% of the sample with a value of \$0 for the non-privately insured.

Consider first, the characteristics associated with having insurance. As shown, age, living arrangement, education level, income and the presence of an informal caregiver are all significantly related to having LTC insurance. On the other hand, older age and the presence of an informal caregiver are negatively associated with having insurance.

That older age is negatively related to insurance purchase may reflect the fact that LTC insurance premiums are level funded but age-rated. This means that the premiums that individuals face increase substantially at older ages and therefore the probability of having a policy may decline as the policy price increases. The negative relationship between the presence of an informal caregiver and LTC insurance policy ownership is also not surprising. The absence of an available informal caregiver would be a primary motivation for purchasing LTC insurance. The insurance is designed to bring formal (paid) caregivers into the home. Individuals who want to remain in their homes and who would need to rely solely on formal caregivers if they became disabled because they lack informal support should find LTC insurance attractive. Similarly, individuals who live alone are more likely to have the insurance than are individuals who live with a spouse or children. Finally, gender and the presence of children living within 25 miles were not significant predictors of insurance status.

The second equation shows the relationship between explanatory variables and hours of care for the Insured Panel only. A primary purpose was to test whether or not the Inverse Mills Ratio (IMR) was statistically significant. Out of 11 variables tested, four were found to be statistically significant and the adjusted R2 was 20.5%. The coefficient on the IMR was not statistically significant, thus indicating that there are no selection effects present.

The third equation is used to estimate the equation for care hours for the combined sample. Thirteen variables were found to be significant predictors of the number of hours of ADL and IADL care received. These variables are discussed below.

Socio-Demographic Characteristics

Age is positively associated with use of formal and informal long-term care services, even when controlling for disability and cognitive status. This variable probably captures additional factors related to frailty and individuals' perceived sense of need for supervision and assistance in performing ADLs and IADLs. When accounting for other variables, living alone was negatively associated with the number of hours of care received. In fact, those who live alone use about seven fewer hours of care compared to those living with another person.10

The relationship between income and care received is more complicated. Because a great deal of LTC is provided informally, the price of care faced by individuals can vary greatly, depending on the availability of informal supports. For most "normal" goods, increased levels of income lead to greater levels of demand for that good. Income is positively related to the number of hours of care received. However, when interacted with insurance status, the effect of income is moderated, as indicated by the negative coefficient on the interaction term. This is what one would expect: having insurance reduces or even eliminates the effect of income variation on the demand for care at the time that need is experienced. In this sample, the effect of income on consumption of services is overwhelmed by the presence of private LTC insurance.

Health Characteristics

Not surprisingly, functional disability status is related to the amount of weekly care received. Individuals with more than three ADL limitations receive about 27 more hours of care per week than do those with fewer limitations. Individuals with significant IADL impairments also receive more weekly hours of care than do those with fewer limitations -- about 26 hours. Although not shown in the table, the beta coefficients on disability variables suggest that they have the greatest impact on the amount of care received. It is also interesting to note that both types of disabilities are associated with greater use of personal care; when there is significant disability in either of these broad categories of function, a great deal of human assistance is required.

Individuals who are cognitively impaired also receive more care -- about five more hours per week -- than do those who are cognitively intact. Finally, self-assessment of health status is related to the amount of care received. Individuals, who view their health as poor, are also more likely to use more weekly assistance -- about seven hours per week.

We also tested eight diagnoses to determine which, if any, were related to use of care in the community.11 With the exception of stroke, none were significantly related to the hours of care received. If an individual had suffered a stroke, they were likely to receive about 11 more hours of care per week (compared to those who had not had a stroke).

Informal Support Availability

Individuals can receive care from two sources: formal service providers or informal supports. When individuals do have an available informal caregiver they receive more care, about seven more hours per week. This is true even when controlling for insurance status. This suggests that both the informal and formal service systems continue to contribute hours of care to disabled elders, even among those who have the financial wherewithal or insurance to purchase formal care only. Clearly, formal and informal care are not perfect substitutes for each other.

Insurance Variables

Two types of insurance have been included in the model: (1) private LTC insurance and (2) public insurance, Medicaid. The latter can be viewed as a public insurance program with a very large deductible. Economic theory suggests that in the presence of insurance, the marginal cost of the insured good is lower, thus increasing quantity demanded of the good. A primary motivation for purchase of LTC insurance is to ensure access to formal services. Medicaid also guarantees continued consumption of formal nursing home care as well as home and community-based care, although on a somewhat more restrictive basis.

Our analysis indicates that the price-quantity relationship is borne out: even when controlling for the other variables, those who have private LTC insurance or public insurance (Medicaid) use more care. In fact, everything else held constant, individuals with LTC insurance receive about ten more hours of care per week, and those receiving Medicaid receive 13 more hours per week than do individuals without any insurance.12, 13

The positive coefficient on the insurance variable, as well as on the informal caregiver variable, suggest that even though there may be some degree of substitution between formal and informal care, the net effect of insurance on total care received is still positive. Put another way, the decline in informal caregiving that may result when insurance benefits are paid is more than offset by the increase in formal care provided. This is true for both the private and the public insurance systems.

Summary

Clearly, individuals with private LTC insurance differ in relevant respects from those without insurance. Nevertheless, we did not find selection effects on service utilization among the insurance sample. This may reflect the fact that insurance companies are effective in the underwriting process and are able to screen out individuals who have a proclivity to service use, given certain revealed characteristics at the time that they apply for insurance.

Having private LTC insurance does increase the amount of care received by disabled individuals, and the same can be said for having public LTC insurance. The presence of insurance increases weekly care by between 10 and 13 hours. Benefits also appear to be distributed rationally: individuals with greater levels of ADL and IADL limitations receive more weekly care. Additionally, the insurance effect substantially eliminates the effect of income variation on the demand for hours of care. Finally, the net effect of having an informal caregiver is more total care received. The additional care financed by insurance does not lead to an overall reduction in the total care received by individuals. Even in cases where informal caregiving may decline, this is more than offset by increases in formal caregiving.

View full report

"multanal.pdf" (pdf, 597.33Kb)