health equity in jamaica measurement and results · 2018. 9. 14. · jamaica results: 2007 table...
TRANSCRIPT
Health Equity in JamaicaHealth Equity in Jamaica
Measurement and Results
E. Scott and K. Theodore (2010)
Measuring Disparities in Health:
The Tools• The tools used are borrowed from the income distribution
literature and applied to health variables.
• are rank-based measures of inequality. That is, they are based on ranking individuals or households by the measure of living standards and then evaluating the cumulative distribution of the health variable across this
2E. Scott and K. Theodore (2010)
cumulative distribution of the health variable across this ranking.
• Advantage: they take account of the full distribution and so are not dependent upon arbitrary dichotomisation or categorisation.
• The Concentration curve and Concentration Index.
• Are akin to the Lorenz curve and Gini Index
In general, the Concentration curve plots the cumulative % of
health variable against the cumulative % of population ranked
by socioeconomic status
L(s)
cumulativeproportion
1
Must be possible to sum
health variable.
Living standards variable only
needs to provide a ranking.
3cumulative proportion of population
proportion
of ill-health
0 1
ranked by socioeconomic status
Curve above the diagonal �
concentration among the poor
Curve below the diagonal �
concentration among the rich
Curve on the diagonal = equality
Graphing the Ccurve: grouped data
1. Rank indvs/HHs by living standards
variable, into quintiles (or deciles)
2. Obtain for each quintile the mean of
variable of interest and # relevant cases
4E. Scott and K. Theodore (2010)
variable of interest and # relevant cases
3. form cumulative % relevant cases and
corresponding cumulative % of total value
of variable of interest; graph concentration
curve using xy chart
Graphing the Ccurve: micro data
• Grouped data produces only an approx. to
the Ccurve
• Better to compute the cumulative
frequencies from individual level data to get
5E. Scott and K. Theodore (2010)
frequencies from individual level data to get
a picture of the complete distribution
Curve versus IndexConcentration curves can be used to identify
• whether socioeconomic inequality in some health sector variable exists and
• whether it is more pronounced at one point in time than another or in one country than another.
6E. Scott and K. Theodore (2010)
• difficult to compare many concentration curves, particularly when they lie close together
• does not give a measure of the magnitude of inequality that can be compared conveniently across many time periods, countries, regions, etc.
The Concentration index • quantifies the degree of socioeconomic-related
inequality in a health variable (Kakwani, Wagstaff, and van Doorslaer 1997; Wagstaff, van Doorslaer, and Paci 1989).
• It has been used, for example, to measure and to compare the degree of socioeconomic-related inequality
7E. Scott and K. Theodore (2010)
compare the degree of socioeconomic-related inequality in
� child mortality (Wagstaff 2000),
� child immunization (Gwatkin et al. 2003),
� child malnutrition (Wagstaff, van Doorslaer, and Watanabe 2003),
� adult health (van Doorslaer et al. 1997),
� health subsidies (O’Donnell et al. 2007),
� health care utilization (van Doorslaer et al. 2006).
50%
75%
100%
cum
. %
of
hea
lth
va
ria
ble
C = 2 x area between 450 line and concentration curve= A/(A+B)
C>0 (<0) if health variable is disproportionately concentrated on rich (poor)
Concentration index defined
8E. Scott and K. Theodore (2010)
0%
25%
50%
0% 25% 50% 75% 100%
cum. % population ranked by income
cum
. %
of
hea
lth
va
ria
ble
rich (poor)
C=0 if distribution in proportionate
C lies in range (-1,1)
C=1 if richest person has all of the health variableC=-1 of poorest person has all of the health variable
A
B( )h
L p
Concentration index defined
• Note that although a proportionate distribution
(i.e. the concentration curve lying on top of the
45o line) is sufficient for the index to take a
value of zero, it is not necessary.
9E. Scott and K. Theodore (2010)
value of zero, it is not necessary.
• The concentration curve could cross the 45o line
and give a concentration index of zero.
• Hence it is always advisable to examine the
index in conjunction with the curve.
Concentration index formulae
( )1
0
1 2 hC L p dp= − ∫
2 1n
If the living standards variable is discrete:
where n is sample size, h the
10E. Scott and K. Theodore (2010)
1
2 11
n
i i
i
C h rn nµ =
= − −∑where n is sample size, h the
health variable, µ its mean and
r the fractional rank by income
( )2
cov ,C h rµ
=
For computation, this is more convenient:
Concentration index defined• The formulae makes it clear that C reflects the
correlation between health variable and the rank in the living standards distribution In fact, it is the covariance between these two variables scaled by 2 divided by the mean of the health variable.
11E. Scott and K. Theodore (2010)
scaled by 2 divided by the mean of the health variable.
• Note that C depends only on the relationship between the health variable and the rank of the living standards variable and not on the variation in the living standards variable itself. A change in the degree of income inequality need not affect the concentration index measure of income-related health inequality.
Properties of the concentration index
• Strictly, requires ratio scaled, non-negative
variable
• Invariant to multiplication by scalar
• But not to any linear transformation
• So, not appropriate for interval scaled
12E. Scott and K. Theodore (2010)
• So, not appropriate for interval scaled
variable with arbitrary mean
• This can be problematic for measures of
health that are often ordinal
Interpreting the concentration index
• How “bad” is a C of 0.10?
• Koolman & van Doorslaer (2004) 75C = % of health variable that must be (linearly) transferred from richer to poorer half of pop. to arrive at distribution with a C of zero
13E. Scott and K. Theodore (2010)
of pop. to arrive at distribution with a C of zero
– But this ensures equality of health predicted by income rank, and not equality per se
Computing C with grouped data
)(...)()( 1123321221 −− −++−+−= TTTT LpLpLpLpLpLpC
• In the formula, pt is the cumulative
percentage of the sample ranked by
economic status in group t, and L is the
14E. Scott and K. Theodore (2010)
economic status in group t, and Lt is the
corresponding concentration curve ordinate
• C is the sum of the differences of their cross
products from adjacent groups
Estimating C from micro data• Use covariance formula C=2cov(h,r)/µ
– Weights applied in computation of mean, covar and rank
• Equivalently, use regression
22 ih
rσ α β ε
= + +
15E. Scott and K. Theodore (2010)
– OLS estimate of β is the estimate of C
22 ir i i
hrσ α β ε
µ
= + +
Standardization• Seek a more refined description of the relationship between
health and SES
• Want to examine socioeconomic-related inequality in health conditional on other factors, e.g. age/sex
• Standardization necessary in case where these factors are correlated with both health and SES
• Direct standardization � distribution if all SES groups had same age/sex structure
16E. Scott and K. Theodore (2010)
same age/sex structure
• Indirect standardization � corrects distribution by comparing it with that expected given actual age/sex
• Direct standardization requires grouping
• Both methods can be implemented by regression
• Can include other (control) variables to reduce bias in the estimated effects of the confounding variables (age/sex) on health
Standardization
• for health status/morbidity variables we
standardize by age/sex groups
• for health care utilization we standardize for
17E. Scott and K. Theodore (2010)
• for health care utilization we standardize for
need which include demographic and health
status/morbidity variables
Direct standardization
Group (g) specific regression:
i g jg ji kg ki i
j k
y x zα β γ ε= + + +∑ ∑
Standardized health:x sample means
18E. Scott and K. Theodore (2010)
ˆˆ ˆˆ ˆDS DS
i g g jg j k kg
j k
y y x zα β γ= = + +∑ ∑j
x sample means
kgz group-specific
means
Immediately gives standardized distribution of health
across (e.g., income) groups
Indirect standardization
ki j ji ki i
j k
y x zα β γ ε= + + +∑ ∑ yi – health, xji – age/sex,
zki – control vbl. e.g. education
Predicted values from:
ˆˆ ˆˆ Xy x zα β γ= + +∑ ∑ ˆˆ ˆ, ,α β γ are OLS estimates
19E. Scott and K. Theodore (2010)
ˆˆ ˆˆ X
i j ji k k
j k
y x zα β γ= + +∑ ∑ ˆˆ ˆ, ,α β γ are OLS estimates
kz are sample means
Standardized health:
ˆ ˆIS X
i i iy y y y= − +Sample mean is added to ensure
standardized = actual mean
Standardization of the concentration
index
• Can use either method of standardization to
compute the C index for the standardized
distribution
• If want to standardized for the total correlation
20E. Scott and K. Theodore (2010)
• If want to standardized for the total correlation
with demographic confounding variables (x),
then can do in one-step
• OLS estimate of β2 is indirectly standardized
concentration index 2
2 22 ir i j ji i
j
hr xσ α β δ υ
µ
= + + +
∑
Equity in health care
• Policies are usually geared towards an equitable distribution of health care
• In health care, most attention given to Horizontal Equity “equal treament for equal medical need irrespective of income, race, etc”
• Specifying a vertical equity norm
21E. Scott and K. Theodore (2010)
• Specifying a vertical equity norm
– assume that, on average, differential utilisation across different levels of need is appropriate
• Given this, assess whether there is equal utilisation for the same level of need
• Methods to measure and explain horizontal inequity in health care utilisation
(1) Identifying HI through the need-
standardized distribution of health care• Care is unequally distributed by income
• But so is need
• To assess whether care is equitably distributed:
• Either compare actual distribution of care (by
income) with distribution of need
22
income) with distribution of need
• Or assess (in)equality in the need-
standardized distribution of care
• Income-related distribution of:
• Actual use describes inequality
• Need-standardized use describes inequity
(2) Measuring HI through
DecompositionFor any linear model:
the concentration index for y can be written as:
( )β µ γ µ µ= + +∑ ∑( / ) / /C x C z C GC
α β γ ε= + + +∑ ∑i j ji k ki ij k
y x z
23
The index of horizontal inequity (HI) can be obtained
directly from this decomposition:
That is, inequity is total inequality minus need-related
inequality
( ) εβ µ γ µ µ= + +∑ ∑( / ) / /j j j k k kj kC x C z C GC
( )/jj j j
HI C x Cβ µ∑= −
(2) Explaining HI through
DecompositionCan decompose total income-related inequality in
observed use into:
(a) “acceptable”, or need-induced, inequality
(b) inequity, or non-need related inequality, due to
24E. Scott and K. Theodore (2010)
(b) inequity, or non-need related inequality, due to
i. direct contribution of income
ii. contribution of other, non-need variables (e.g.
education, health insurance, location)
iii. contribution of residuals (unexplained inequality)
HI = C - (a) = (i) + (ii) + (iii)
More Details
• “Analyzing Health Equity Using Household
Survey Data: A Guide to Techniques and
their Implementation” 2008. Owen
25E. Scott and K. Theodore (2010)
their Implementation” 2008. Owen
O’Donnell, Eddy van Doorslaer, Adam
Wagstaff and Magnus Lindelow. The World
Bank, Washington DC.
Jamaica Data
• SLC 2007 & 2004
• Ranking variable: adult equivalent per
capita household consumption
26E. Scott and K. Theodore (2010)
capita household consumption
• Sample of individuals aged 18+
• Health Status: P(ill/injured) & Duration
• Utilization: P(HPvisit) & #HPvisit,
P(hospitalized) & Days hospitalized.
Jamaica DataStandardizing variables
• Age-gender categories: male and female dummy variables for age categories 18-34, 35-44, 45-64, 65-74, 75+. Reference category = male 18-34 years.
• Self-assessed health (SAH): dummy variables for very good, good, fair, poor, very poor. Reference category =
27E. Scott and K. Theodore (2010)
good, good, fair, poor, very poor. Reference category = very good
• Activity/physical limitation: dummy variables for moderate, severe, none. Reference category = no limitation
• Chronic disease: dummy variable for the presence of a chronic problem, defined here as illness of more than 4 weeks duration.
Jamaica Data
Control (non-standardizing) variables include
dummies for
• education
• economic activity status
28E. Scott and K. Theodore (2010)
• economic activity status
• urban-rural location
• marital status
• private insurance coverage
• household size.
Jamaica Results: 2007Table A.2: Quintile Distributions of Actual and Standardized Illness and Duration:
Jamaica 2007
Injured or IllInjured or Ill
(stndrdized)Days ill
Days ill
(stndrdized)
Quintiles of
adult eqv
29E. Scott and K. Theodore (2010)
adult eqv
consumption
1 0.1706 0.1638 9.7192 9.3202
2 0.1582 0.1605 7.4138 6.9792
3 0.1491 0.1529 5.9726 5.9985
4 0.1504 0.1562 6.1222 6.5100
5 0.1493 0.1547 3.4921 3.8157
Total 0.1555 0.1575 6.5193 6.4924
CI / SCI -0.0255 -0.0119 -0.1815 -0.1535
Jamaica Results: 2007Concentration Curves Injured or Ill (actual and standardized)
60
70
80
90
100C
um
ula
tive %
Outc
om
e
30E. Scott and K. Theodore (2010)
0
10
20
30
40
50
60
0 2 4 6 8 10
Population Consumption Deciles
Cum
ula
tive %
Outc
om
e
Inj/Ill (S)
Equality
Inj/Ill
Jamaica Results: 2007Concentration Curves Days of Illness (actual and standardized)
70
80
90
100C
um
ula
tive %
Outc
om
e
31E. Scott and K. Theodore (2010)
0
10
20
30
40
50
60
0 2 4 6 8 10
Population Consumption Deciles
Cum
ula
tive %
Outc
om
e
Equality
Days Ill (S)
Days Ill
Jamaica Results: 2007
Probability
HP Visit
Probability HP
Visit
(stndrdized)
Number of
Visits
Number of Visits
(stndrdized)
Quintiles of
adult eqv
32E. Scott and K. Theodore (2010)
adult eqv
consumption
1 0.0945 0.0734 0.1146 0.0823
2 0.0954 0.0945 0.1323 0.1300
3 0.1028 0.1020 0.1575 0.1572
4 0.1095 0.1246 0.1608 0.1820
5 0.1023 0.1168 0.1420 0.1654
Total 0.1009 0.1024 0.1410 0.1433
CI / HIWV 0.0294 0.1000 0.0468 0.1236
Jamaica Results: 2007Concentration Curve HP Visit (actual and standardized)
80
100
Cum
ula
tive %
Outc
om
e
33E. Scott and K. Theodore (2010)
0
20
40
60
0 2 4 6 8 10
Population Consumption Deciles
Cum
ula
tive %
Outc
om
e
HPvisit (S)
Equality
HPvisit
Jamaica Results: 2007Concentration Curve Number Visits to HP (actual and
standardized)
80
100
34E. Scott and K. Theodore (2010)
0
20
40
60
0 2 4 6 8 10
Population Consumption Deciles
Cum
ula
tive %
Outc
om
e
# visits to GP
# visits to GP (S)
Equality
Jamaica Results: 2007
P(Hospitalized)P(Hospitalized)
(standardized)
Days
Hospital
Days Hospital
(standardized)
Quintiles of
adult eqv
35E. Scott and K. Theodore (2010)
adult eqv
consumption
1 0.0041 0.0033 0.0199 0.0160
2 0.0049 0.0047 0.0248 0.0241
3 0.0049 0.0048 0.0320 0.0302
4 0.0079 0.0083 0.0363 0.0327
5 0.0091 0.0092 0.0739 0.0777
Total 0.0062 0.0061 0.0374 0.0364
CI / HIWV 0.1803 0.2253 0.2842 0.3340
Jamaica Results: 2007
Concentration Curves Hospitalization (actual and standardized)
70
80
90
100
36E. Scott and K. Theodore (2010)
0
10
20
30
40
50
60
70
0 2 4 6 8 10
Population Consumption Deciles
Cum
ulative Pct.
Hospitalization (P) (S)
Equality
Hospitalization (P)
Jamaica Results: 2007
Concentration Curves Nights Hospitalized (actual and standardized)
70
80
90
100
37E. Scott and K. Theodore (2010)
0
10
20
30
40
50
60
70
0 2 4 6 8 10
Population Consumption Deciles
Cum
ula
tive P
ct.
Equality
# Nights Hospitalized
# Nights Hospitalized (S)
2004-2007
Health Status
Prob(Injured or Ill)Days Ill
2004 2007 2004 2007
38E. Scott and K. Theodore (2010)
2004 2007 2004 2007
Concentration
index, CI0.0244 -0.0255 -0.0154 -0.1815
Standardized CI 0.0161 -0.0119 -0.0267 -0.1535
2004-2007
Utilization: Visits
Prob(GP Visit) Number GP Visits
39E. Scott and K. Theodore (2010)
2004 2007 2004 2007
Concentration index, CI 0.0368 0.0294 0.0355 0.0468
Standardized CI (HIWV) 0.0691 0.1000 0.0691 0.1236
Inequity* 0.0696 0.0941 0.0743 0.1207
2004-2007
Utilization: Hospitalization
Prob(Hospitalized) Days Hospitalized
40E. Scott and K. Theodore (2010)
2004 2007 2004 2007
Concentration index, CI 0.0812 0.1803 0.3326 0.2842
Standardized CI (HIWV) 0.1091 0.2253 0.2877 0.3340
Inequity* 0.1427 0.1834 0.2510 0.2227
“Hefty Price for Free Health Care”
• Jamaica Gleaner, April 10, 2010
• Revenue foregone: J$4b (2 years)
41E. Scott and K. Theodore (2010)
• Revenue foregone: J$4b (2 years)
• Increased utilization 16-23%
• long wait for treatment
• unavailability of several prescription drugs