Does Publicly Provided Home Care Substitute for Family Care?

Document Sample
Does Publicly Provided Home Care Substitute for Family Care? Powered By Docstoc
					 Does Publicly Provided Home
Care Substitute for Family Care?
      By Liliana E.Pezzin, Peter Kemper, and
                James Reschovsky;
   Journal of Human Resources, Summer 1996, v. 31, n.3.
               Presented by Mark L. Trueman
                    Introduction
• Elderly population in U.S. expected to double by 2030.
  Demand for long term care (LTC) increasing.
• Types of LTC for the poor:
   – In kind, by unpaid “informal” caregivers (IC).
   – Institutional arrangements: e.g., nursing homes (N)-
      thru Medicaid.
   – “Formal” home care (FC)- Subsidy. Limited.
• Concern: subsidies may increase LTC expenditures w/o
  lowering nursing home use.
• Purpose of paper: gain a better understanding of the extent
  to which public subsidy of formal home care substitutes
  for family care.
• Key feature: estimate how living arrangement (LA)
  choices and hours of IC (HIC)/LA differ in presence/
  absence of public subsidy program.                          2
        Decompose the Effects of
            Subsidy (FC)
1. Direct effect (aka “hours effect”): the induced
   marginal change in behavior of informal caregivers :
   •   change in hours of IC to individuals in a given LA,
       weighted by probability of choosing a particular LA.
2. Indirect effect (aka “LA effect”): change in the
   probability of choosing a specific LA:
   •   this change in probability is weighted by E(HIC/ LA).

   *   Note: Policy making may be more effective in countering
       the direct effects- tailor interventions through case
       management guidelines on IC and respite care.
                                                               3
                      Theoretical Model
•   Max U=U(X, L, F; τ) s.t. an “arrangement specific budget constraint”
     – X is a vector of private goods, or a composite commodity
     – L is leisure
     – F is a measure of disabled, elderly person’s functioning; May be produced in:
          • Community setting: Either Separate/ Joint households
               – Requires Compensatory LTC: either HFC or HIC where H is hours
               – F = F(HFC, HIC ; D) where D is the level of disability
          • Institution: Nursing home services (N), exclusively
               – F = F(N; D)
     – τ is a taste parameter- captures family preferences for privacy and
        independence (affects utility of LA options)
•   Step 1: Choose optimal X, L, and F/ on each LA
     –   { X*, L*, F*; H*FC, H*IC } are the cond’l commodity and input demands.
•   Step 2: Compare values, then choose type of care (IC and/or FC; or N) and LA
    (separate/ joint households) which maximizes overall utility, U*.
     – Let j = index of LA options:
          • j = 0, independent living (separate household)
          • j = 1, shared living (joint “intergenerational” household)
          • j = 2, institutional living (nursing home)
                                                                                  4
            Theoretical Model Continued
•   By substitution, we get a set of indirect utility functions, Vj , where:
     –   I = family’s unearned income
     –   PIC = shadow price of informal caregiving time
     –   PFC = price of formal home care
     –   PN = price of nursing care
     –   Px = 1 (nummeraire)
           • V0 = V0(I, PIC, PFC) = U*[X0*, L0*, F0*(HIC,0*, HFC,0*; D); τ]
           • V1 = V1(I, PIC, PFC) = U*[X1*, L1*, F1*(HIC,1*, HFC,1*; D); τ]
           • V2 = V2(I, PN) = U*[X2*, L2*, F2*(N; D); τ]                              (1)
•   Chosen living arrangement will be:
     – j* = argmax[Vj(I, PIC, PFC, PN; t, D)] for j = 0,1, 2
•   Implied (conditional) demands:
     – HIC* = HIC(I, PIC, PFC, PN; D/ j*)                                             (2)
     – HFC* = HFC(I, PIC, PFC, PN; D/ j*)
     – N* = N (I, PIC, PFC, PN; D/ j*)
•   Eq’n (2) implies that a subsidy program that reduces PFC affects family’s conditional
    demand functions but also LA choices.
• əE(HIC)/ əPFC = Σ{ə E(HIC / j)/ əPFC * Prob(j) + [əProb(j)/ əPFC]* E(HIC / j)
                  j

Overall Effect =                  Direct Effect +                       Indirect Effect     (3)   5
       Channeling Experiment & Data
• National test of expanded public financing of home care, 1982-1985.
• Aim: Test whether a managed system of home and community based services
  could be a cost effective alternative to institutionalization.
• 5 communities/ 6,236 eligible applicants
• individuals channeled into one of 3 groups, which had case managers (CMs)
   – Group 1/ “Basic”: CMs determine needs/ services under existing system
      (i.e., limited grants, $, to finance home care services)
   – Group 2/ “Financial”: Direct provision of home care subject to a CM’s
      authorization & cost limits >>>> substantial increase in use of HFC ($$$)!
   – Group 3/ Control Group
• Screening interviews to establish eligibility (disability & unmet need) & follow-
  up interviews to collect data (service use, LA, # of “visiting”/ “resident” hrs.).
• Average age: 79, most w/ multiple functional limitations.
• Average monthly income: < $ 530.
• At 1-Year follow-up interview: 28% died; 15% nonrespondents; 57% analyzed.
                                                                              6
     Living Arrangements : Unmarried Persons                    Living Arrangements: Married Persons


                                                                      11%
          19%


                            40%       Independent Living        19%                                 Independent Living
                                      Shared Living                                                 Shared Living
                                      Institution                                                   Institution

                                                                                       70%
         41%




     Informal Care: Avg. # of "Visiting" Hours                  Informal Care: Avg. # of "Resident" Hrs.
        (Living Arrangement: Independent)                            (Living Arrangement: Shared)
                                                           30

                                                           29
10
8                                                          28

6                                                          27
4
                                                           26
2
                                                           25
0
                                                                   Unmarried Persons         Married Persons
        Unmarried Persons         Married Persons

                                                                                                               7
                        Empirical Model (1)
      • Eq’n (2) counterparts of LA choice & HIC:
           – Vijt* = ßj0 + ΣßjgTig,t + ΣßjmAim,t-1 + ΣßjkZik,t-1 + εijt                        (4)
                             g            m                k
           – ln HICijt = γj0 + Σ γjgTig,t + Σ γjkZik,t-1 + μijt
                               g            k
                                                                                               (5)
                      – Vijt* = latent variable, value to family i choosing jth LA, at time t,
                 • Where:
                      – ln HICijt = ln of IC hours in family i, choosing jth LA, at time t,
“HFC: subsidy”        – t is at 1- year follow up evaluation; t-1 is at initial screening,
                      – T is a (1 X g) vector of dummies for treatment status (“basic”, “financial”)
                      – Z is a (1 X k) vector of variables proxying remaining elements which affect family’s
                        utility & cond’l demand functions: (I,PIC, PN, D). Preexperimental measures of family’s
                        economic, demographic, & health status, prior service use, and “site” dummies.
                      – A is a (1 X m) vector capturing family’s transactions costs and τ (prior LA)
                      – Assumes εijt and μijt are distributed BVN (0,0,σε2, σμ2,ρ) where ρ is the correlation
                        between LA choices and hours of informal care (HIC).
                 • ß and γ are the vectors of coefficients to be estimated in the model.
                 • Vijt* is an indicator of which of the “j” LA alternatives is chosen.
                 • An elderly person will be observed in a particular LA, Vijt = 1 iff
                   Vit* falls into a particular interval αj-1 < Vit* < αj .
                 • [αj] is a vector of (J +1= 2+1=3) thresholds in the latent variable index, where α 0=0.
                                                                                                          8
               Empirical Model (2)
• Model is operationalized by assuming that LA choices
  can be ordered in hierarchy corresponding to higher
  levels of assistance:
   – Independent, shared, institutional living
• Eq’n (4) reduces to an ordered probit model.
   – Elderly person will attempt to live independently as long as
     possible, until a threshold is reached; then switches.
• Probability of any level Vit* is chosen is given by:
   – Prob [Vijt = 1] = Φ[ (αj –β'Yi)/ σε] - Φ[ (αj-1 –β'Yi)/ σε]           (6)
       • Φ(•) represents a cdf,
       • Y is a matrix of all nonstochastic explanatory variables in eq’n (4),
       • Assumes and σε = 1.

                                                                           9
                     Empirical Model (3)
• Use a 2-step estimation procedure that provides estimates of
  the effects of Channeling on LA choices and IC provision.
   – Step 1: £ = Π Π [Φ(αj –β'Yi) - Φ(αj-1 –β'Yi)]Vijt  (7)
                 i  j
   – Step 2: Estimate the effect on HIC / LA choices by applying OLS to
     a new version of Eq’n (5):
      • E(ln HICit/ j) = γj0 + Σ γjgTig,t + Σ γjkZik,t-1 + bjλij + μijt* where,
                                g           k
           – HIC represents either “visiting” or “resident” care hrs,                  (8)
           – bjλij is a selectivity bias correction due to nonrandom selection of LA choice.
• Model is estimated separately for each person’s marital
  status at time of follow-up, t.
   – Married/ unmarried have substantially different endowments of
     potential care. Presence of spouse affects role of others/incentives.
   – We expect unmarried to be more responsive to intervention.
                                                                                      10
        Results: LA Choices (1)
• Rows 1 and 2 coefficients     Table 2
  represent Channeling’s        Maximum Likelihood Ordered Probit Estimates
                                of the LA Choice Models
  impact on Vit*, after
  controlling for differences   Variables                Unmarried   Married
  in preexperimental LA.
                                Basic Intervention        (0.053)    0.090
• Financial intervention
  increased probability of      Financial Intervention    (0.215)    (0.008)
  being in more independent
  LA.                           Threshold (a1)            1.348      0.773

• Married folks have lower      Log likelihood             5,176     1,964
  threshold value by which
  they will switch LA.          Sample Size                4,421     1,905

                                                                             11
           Results: LA Choices (2)
• Table 3 shows             Table 3
                            LA Predicted Probabilities
  predicted probabilities                                   Basic Intervention Financial Intervention
  of each LA choice in      Variables             Treatment Control Difference Treatment Control Difference
  presence and absence      Unmarried
  of each interventions      Independent living     0.407     0.390     0.017     0.450    0.379         0.071
  (from equation (6)).       Shared living          0.399     0.403    (0.004)    0.391    0.415        (0.024)
                             Institution            0.194     0.206    (0.012)    0.158    0.205        (0.047)
• Financial interventions
  effect on LA choices of   Married
  unmarried individuals      Independent living     0.651     0.681    (0.030)    0.729    0.726         0.003
  was substantial.           Shared living          0.212     0.199     0.013     0.178    0.180        (0.002)
                             Institution            0.137     0.120     0.017     0.093    0.094        (0.001)

                                                                                                   12
                 Results: HIC/LA choice (1)
                                           Table 4
                                           OLS Estimates of HIC/ LA Choice

• Table 4 presents estimated                                       Visiting Hrs. Resident Hrs.


  experimental impacts on HIC,             UNMARRIED
                                           Independent living
  adjusted for LA choices: Eq’n (8).         Basic treatment           (0.127)       N/A
                                             Financial treatment       (0.203)       N/A
• Overall, there is no evidence that         LA correction ( lambda)   (1.127)       N/A
                                             Adjusted R-sqd             0.115        N/A
  the intervention had a significant         Sample size                1,029        N/A
  impact on conditional hours of care,     Shared living
                                             Basic treatment           0.050        (0.077)
  net of its impact thru LA.                 Financial treatment       0.009        (0.104)
                                             LA correction ( lambda)   0.460        (0.515)
• Strong & significant effect of the         Adjusted R-sqd            0.097        0.147
                                             Sample size               1,059         271
  LA correction term on both visiting      MARRIED
  & resident hours for the unmarried       Independent living
                                             Basic treatment           (0.042)      0.009
  sample.                                    Financial treatment       (0.109)      0.182
                                             LA correction ( lambda)   (0.832)      0.575
• Authors assert that it is important to     Adjusted R-sqd             0.072       0.090
                                             Sample size                 680         173
  control for the endogeneity of the       Shared living
                                             Basic treatment           0.248        0.207
  LA choices when analyzing use of           Financial treatment       0.178        0.591
  LTC.                                       LA correction ( lambda)   0.868        0.824
                                             Adjusted R-sqd            0.043        0.153
                                             Sample size                187           52

                                                                                       13
           Results: HIC/LA choice (2)

• The coefficients in    Table 5
                         Predicted Hours of Visiting & Resident Informal Care by LA Choice- Unmarried Sample
  the previous table
  were used to find                                  Basic Intervention      Financial Intervention
  the predicted hours                      Treatment Control Difference      Treatment Control Difference
  in table 5, based on
  Eq’n (8).              Visiting hours
                            Independent       8.6      10.3       1.7           8.1       11.9       3.8
• Greater reductions                                             (2.3)                              (2.5)
  in informal care are
  observed for those       Shared living      5.6      5.1        0.5           6.1       5.5        0.6
  receiving more                                                 (0.8)                              (0.9)
                         Resident hours
  generous financial
                           Shared living     37.5      41.5       4.0          35.0       42.9       7.9
  intervention.
                                                                                                   14
             Results: Overall, Direct, &
            Indirect Program Effects (1)
•   Table 6 presents the results of a simulation: an experimental analog of Eq’n (3):
əE(HIC)/ əPFC = Σ{ə E(HIC / j)/ əPFC * Prob(j) + [əProb(j)/ əPFC]* E(HIC / j)}
                 j
Overall Effect =     Direct (or Hours) Effect               +    Indirect (LA) Effect
• For each individual:
     – calculate predicted probability of choosing each LA, P-hat, assuming treatment, then control
       status; average of these predictions: E[Prob (j)].
     – calculate predicted hours of visiting and resident care, HIC-hat, each would receive had he/she
       chosen each LA, according to the estimated hours equation; average of these predictions:
       E(HIC / j) in the presence and absence of each treatment

•   Overall program impact is given by: (1/N) Σ Σ (PTifHTICif             – PCifHCICif) (9)
                                                         j i
•   LA effect is given by (1/N) Σ    (PTif –   PC        C
                                                    if) H ICif                          (10)
                                   i

•   Hours effect is given by (1/N) Σ (HTif – HCif) PCICif                               (11)
                                      i

                                                                                               15
Table 6
Decomposition of Overall Treatment Effect into Direct (Hours) Effects
and Indirect (LA) Effects- Unmarried Sample
                             Visiting     Resident
                             Hours        Hours      Total
BASIC INTERVENTION
Hours effect
  Independent living               0.4        0.0        0.4
  Shared living                    0.2        1.6        1.4
  Total                            0.2        1.6        1.8
Living arrangement effect
  Independent living               0.1        0.0        0.1
  Shared living                    0.0        0.2        0.2
  Total                            0.1        0.2        0.1
Overall effect                     0.1        1.7        1.9

FINANCIAL INTERVENTION
Hours effect
  Independent living              1.0         0.0       1.0
  Shared living                   0.2         2.9       2.7
  Total                           0.8         2.9       3.7
Living arrangement effect
  Independent living              0.6         0.0       0.6
  Shared living                   0.1         0.9       1.0
  Total                           0.5         0.9       0.4
Overall effect                    0.3         3.8       4.1


                                                                        16
                              Conclusions
•   Public home care provision results in:
      – only small reductions in overall amount of care provided by informal caregivers to
         unmarried persons;
      – No reductions for married persons.
•   Implication: benefits of such programs will flow primarily to disabled elderly recipients
    rather than to informal caregivers.
•   Decomposition suggests that the direct effect on hours of care assuming no change in
    LA, is likely to dominate any overall effect of subsidized care.
•   Policymakers concerned with potential substitution should focus their attention on
    specific measures designed to minimize caregiver’s behavioral responses rather than on
    discouraging effects due to LA choices.
•   Channeling financial intervention had both sizable and statistically significant effects on
    LA decisions of unmarried persons.
      – Increased probability of living independently by 7.1% relative to control group.
      – Increase was associated with corresponding significant reduction in probabilities of
         living with others (2.4%) and of living in a nursing home (4.7%)
•   The more generous intervention appears to have enabled unmarried elderly persons with
    disabilities to live more independently.

                                                                                             17
           Questions Raised
• Are the benefits from increased
  independence and associated quality of life
  sufficient to justify the cost of expanded
  public home care coverage for this group?
• If so, can the targeting of benefits to
  unmarried persons be justified on equity
  grounds?

                                                18

				
DOCUMENT INFO
Shared By:
Categories:
Tags:
Stats:
views:11
posted:8/30/2012
language:Unknown
pages:18