Transcript 7550_L9_CB2

CB Analysis – 2
Old Wine in New Bottles?
© Allen C. Goodman, 2013
Lots of Questions About CB
• Traditional CB looked at issues like land use.
For example, should we build a dam or not?
• Incremental Benefits?
– Some land, that we had been using, became more
productive.
– Some land, that we couldn’t use before, now could
be used.
• Incremental Costs?
– Costs of the building the dam itself.
– Costs attributed to the dam regarding land that
was currently being used.
– Costs attributed to the dam regarding land that
was brought into use.
Questions About Health-Related CB
• Health-related CB looks at similar issues. For
example, should we give kidney dialysis?
• Incremental Benefits?
– People, who have been ill from kidney disease, are
now less ill, and presumably more productive.
– People, who would have died from kidney disease, will
live longer and be more productive.
• Incremental Costs?
– Starting kidney dialysis is expensive.
– Continuing kidney dialysis involves additional costs.
– Person who lives longer may need a kidney transplant
down the road and that’s REALLY expensive.
Other substantive issues
• What discount rate do we use? It
makes a difference!
• How do we measure willingness-to-pay
(benefits) for improvement?
• Do we include all costs, no matter
what?
• If we count future benefits don’t we also
have to count future costs?
Good example
Manns, Meltzer, Taub, Donaldson, Health
Economics 12, 949-958, 2003, “Illustrating the
impact of including future costs in economic
evaluations: an application to end-stage renal
disease.”
(1) How does high cost of ongoing dialysis affect
cost/QALY?
(2) What impacts do “future costs” have?
(3) Do QALYs represent adequate measures of
benefits for ESRD (end-stage renal disease)?
Gold Standard
• Double Blind.
• Can we double blind a dialysis test?
• Why or why not?
Future Costs for ESRD
Ongoing dialysis 3 times per week and/or
transplantation.
Unrelated medical and non-medical
expenditures.
Both could end up being big!
They looked at a comparison between synthetic
(new) hemodialysers and cellulose (older)
hemodialysers.
Discounted QALYs at 5% per year.
Various Analyses
(1) inclusion of the cost of the dialyser only,
excluding the cost of related medical care;
(2) inclusion of the cost of the dialyser and of
related medical costs, such as dialysis and
transplantation, assuming that all patients are
treated with in-center hemodialysis;
(3) inclusion of the cost of the dialyser and both
related and unrelated medical costs; and
(4) inclusion of the cost of the dialyser and all
related and unrelated medical costs and
nonmedical expenditures.
Items 3 and 4
(3) The cost of unrelated health care (including the annual
cost of non-kidney failure-related medications and the
annual mean cost of non-kidney failure related
hospitalisation) for hemodialysis patients using synthetic
dialysers was estimated from a local study (Table 1).
(4) Non-medical expenditures were estimated by calculating
lifetime total net resource use for patients by adding agespecific estimates of average consumption [28,29] net of
earnings [6,30]. Average consumption by age was
estimated with data from the Canadian Survey of
Household Spending and included the annual
consumption of non-medical and medical goods.
ITEM: In baseline analyses, the estimates considered for
each of the above variables were for 60-year-old men.
Extra Y
Major findings
Extra C
time
Over a 19 year horizon, synthetic dialysers  an extra
0.38 QALYs/patient compared to cellulose. They were
also more expensive.
Looking ONLY at intervention (dialyser only), cost/Q
gained = $5,036.
When Related Costs included (costs of dialysis and
transplantation*), cost/QALY gained = $83,501
When all future costs included, cost/QALY gained =
$121,124!
Magnitude of increases was largely related to high costs
of future dialysis and possible transplantation.
*Includes nursing salary, physician charges,overhead, cost of kidney
failure-related admissions, and cost of erythropoietin.
These are HIGH!!!
• “Critical value” is often taken as $50 000/Q.
• Why do we provide hemodialysis if costs are so
high?
– Cost estimates for future care are critical. How good are
they?
– Are some QALYs different than others?
• W/O dialysis, patients with ESRD WILL DIE!
Possibly a QALY that prevents certain death is more
valuable than one resulting from an improvement in
the probability of survival or of experiencing a higher
quality of life.
• AG: Also, if you’re going to add future costs to the
formula, you’ll have to increase the “critical value.”
Sensitivity and Specificity
• If you get involved in health sector, you read a lot
about these two terms. What do they mean?
• Let’s look at a screening test.
A perfect test yields
No
(1)
Test
No
Yes
True incidence Yes
(2)
Specificity:
Finding “no” when
True = “no”
0
0
Sensitivity:
Finding “yes” when
True = “yes”
Sensitivity and Specificity
• If you get involved in health sector, you read a lot
about these two terms. What do they mean?
• Let’s look at a screening test.
No
(1)
Test
No
Yes
True incidence Yes
(2)
Specificity:
Finding “no” when
True = “no”
False Positive
False Negative
Sensitivity:
Finding “yes” when
True = “yes”
The PERFECT Screen
• Sensitivity = 100%
• Specificity = 100%
• Problem
– Screens aren’t perfect. Increased
sensitivity usually comes and the cost of
decreased specificity, and vice versa.
Sensitivity and Specificity
• Specificity = (True -) / (True - plus False +) -- Column
1
• Sensitivity = (True +) / (True + plus False -) -Column 2
No
(1)
Test
No
Yes
True incidence
Specificity
False Positive
Yes
(2)
False Negative
Sensitivity
Example; N = 400
• True incidence: 250 + ; 150 - ; Test: 250 + ; 150 • Specificity = (True -) / (True - plus False +) = 100/150 =
0.67
• Sensitivity = (True +) / (True + plus False -) = 200/250 =
0.80
True incidence Yes
No
(1)
Test
No
Yes
(2)
100
50
50
200
Sensitivity and Specificity
• True positives (sensitivity) give us benefits.
• Greater specificity implies fewer false positives, reducing
unnecessary treatment expenses.
• False negatives have no impact in net benefit equation, because
what happens to them is assumed to be no different than what
would have occurred w/o screening.
True incidence
No
Test
No
Yes
Specificity
False Positive
Yes
False Negative
Sensitivity
Educated Guesses
Louise Russell
Becoming VERY
dated, but it asks
some important
questions!
Screening
• Must balance the benefits of screening against the
costs.
• She looks at:
– cervical cancer
– prostate cancer
– high blood cholesterol
• We’ll concentrate on cervical cancer, because the
effectiveness of screening and of treatment are both
well-established.
• Pap test -- Scrape cells from the cervix onto a glass
slide. Smear is sent to a lab, where it is examined for
abnormal cells.
•
•
•
•
There’s now a
Screeningvaccine, although
it does not negate
need for
Pap
Pap tests detects precursors tothe
cervical
cancer
testsabnormalities
as well as the cancer itself. Early
can be followed up and treated with relatively
simple outpatient procedures.
Screening and early treatment can reduce
incidence of cervical cancer by over 90%.
STANDARD ADVICE -- Used to be to get a Pap
smear each year.
This has changed, even since 1980, when
American Cancer Society recommended that if
the first 2 annual tests were negative, subsequent
tests could be given every 3 years.
So, what are the problems?
Cervical cancer is sometimes missed -- possibly as
often as 20% of the time.
1.
2.
3.
Scraping may not include abnormal cells, EVEN if they
are present.
Even if they are there, technician may miss them.
Cancer may develop between tests.
To counter risk that a cancer will be missed, women
can take test MORE OFTEN. Take the 20% miss
rate.
If 100 women with cervical abnormalities are
screened, 20/100 will be missed by the first test.
If screened again, (20/100) * (100 - 80) = 4 will be
missed again.
So after 2 tests, you’ve gotten 96/100 with cervical
cancer – missed 4.
False Positives
• Just as you are likely to find those who have
cancer, which is good, you are also likely to
diagnose some as having cancer …
WHO DON’T HAVE IT
“Common conditions like inflammation or injury
to the cervix produce tissues so similar in
appearance to dysplastic tissue that they can
lead to a mistaken diagnosis when the sample is
examined.”
• Good estimate of the FALSE positive rate is about
5%.
• PROBLEM -- the more often you examine, the
more likely that you’ll get a false positive
Key aspects
• False positives may occur only 5% of the time,
BUT
• ALL WOMEN MAY EXPERIENCE THEM
• Without screening
– chance of developing invasive cervical cancer is 2.5%
over the lifetime.
– Chance of dying from it is 1.2%
– Thus, at most, 2 or 3 of every 100 women will eventually
develop cervical cancer that could be missed by the test
WHILE
– ALL 100 ARE VULNERABLE TO THE POSSIBILITY OF A
FALSE POSITIVE RESULT
So, with false positives:
• Chance = 0.05 + 0.05*(1 - 0.05) + 0.05*(1-0.05)2 + …
• If you work that out, the chance of having at least
1 false positive is 40% over 10 tests.
• What can happen??
ANSWER>
Unnecessary treatment
Stress, worry
WASTED treatment because the condition may
have never developed into a cancer
Benefits
• Screening every 4 years   average woman’s
life expectancy by 94 days.
• Screening every 3 years   average woman’s
life expectancy by 95.5 days. Not much in the
way of incremental benefits.
• Tangible gains appear from movement from 0 to 1
test for women who had had NO screening (older
and Hispanic women in her study).
Costs
• (At the time) about $75 for visit and test.
• Treatment costs of $300 to $1,300.
BOTTOM LINE
Early detection saves money, but the saving
only partially offsets the costs of the tests
and of treating false positives and
conditions that would have regressed on
their own.
Recommendations – American Cancer Society –
2012
Go to link for
• All women should start screening at age 21.
No longer is
comments.
screening recommended three years after starting vaginal
intercourse.
• Women aged 21 to 29 should get a Pap test (conventional
or liquid-based) every three years. The statement
specifically recommends against annual Pap testing.
• For women 30 and over, Pap tests should be done every
three years. The guidelines recommend against annual or
more frequent Pap testing for this age group.
• Combining the Pap test with HPV testing every three to five
years is the preferred strategy for women aged 30 and
older.
• Screening is not recommended for women 65 or older who
have had three or more normal Pap tests in a row and no
abnormal Pap test results in the past 10 years, or who have
had two or more negative HPV tests in the past 10 years.
http://www.cnn.com/2012/03/14/health/brawley-cervical-cancer-screenings/index.html
Update – Does this still
occur?
• I gave “grand rounds” at Indiana
University Medical School in
February 2006 – it was a variation on
the screening lecture.
• One faculty member allowed that
they had much better tests now, and
that the screening problems that I
had alluded to were not problems
any more.
Yes, it does!
Letter from Fred Stehman MD, host and
Clarence E. Ehrlich Professor and Chair
“As I promised I have included a couple of
articles for your information/files”
“The first is from the American Cancer Society,
their current guidelines for Pap test frequency.
These were updated in 2002 but are a
continuation of the 1980s guidelines.” [1]
[1] Smith, Robert A. Cokkinides, Vilma, Eyre, Harmon J.,
“American Cancer Society Guidelines for the Early
Detection of Cancer, 2006,” CA Cancer J Clin 2006; 56: 1125.
• “The 2nd article from the American Journal of
Obstetrics and Gynecology in 2004 indicates that
in a closed system at Kaiser, half of the positive
Pap smears were false-positives! This is an
alarming figure in light of your statement that
everything below this line is bad.” [2]
• “The 3rd paper, also from AJOG, shows that
American obstetrician-gynecologists are perfectly
capable of ignoring guidelines and continuing to
screen too often and too long. Incentives matter!
While we don’t get paid a lot for doing a Pap smear,
we do get paid some, and it looks like my
colleagues will continue to recommend more of our
own services.”[3]
[2] Insinga, Ralph P., Glass, Andrew G. and Rush, Brenda B.,
“Diagnoses and outcomes in cervical cancer screening: A
population-based study,” AJOG (2004) 191, 105-13.
[3] Saint, Mona, Gildengorin, Ginny, and Sawaya, George F.,
“Current cervical neoplasia screening practices of
obstetrician/gynecologists in the US,” AJOG (2005) 192,
414-21.
Proposed
Framework
Magnify
to 200%
So … why screen?
• Some conditions are asymptomatic. For
example, kidney mass, prostate abnormality,
breast cancer.
• Some risky behaviors (like smoking) are
identifiably risky behaviors.
• Some people have personal or family history
of illness.
Some discussion
http://www.scpr.org/programs/airtalk/201
3/05/14/31786/the-medical-sciencebehind-angelina-jolie-s-decisi/
Supplementary Material
Guaiac Test
Not covered in
class – Students
not responsible
for it
• Neuhauser/Lewicki looked at test w/ 6 sequential stool tests
for occult blood. N = 278 people.
• 24 had positive results (any one of 6 tests were positive)
• 254 had negative results (all 6 tests were negative)
• Upon further assessment only 2 of 24 positives were found to
have cancer. None of the 254 (-) were tested further.
• For 2 diseased cases, 11 of 12 (0.917) were positive.
• For the 22 non-diseased cases 46 of 126 (6 are unaccounted
for) were positive (36.5% false positive). 80 were negative.
• For 254 non-diseased cases, 1524 were negative.
Guaiac Test (2)
Not covered
in class
• Idea. You’re seeking true positives. You do more tests to
find more true positives, BUT at a cost.
• N-L assumed that 72 persons out of a population of 10,000
who are screened will get colonic cancer.
• On first test, then, of the 72, you’ll get 0.917 (fraction) of
them, or about 66.
• On second test, you’ll get 0.917*(72 - 66) = 5.5
• GIVING 71.5 of the 72, after 2 tests !!!
• Incremental impact gets
.
very small
• Incremental cost gets VERY LARGE.
Not covered
in class
Table 24.3
Table 4.3 - Neuhauser/Lewicki
Detection
No. No. I/G
1
2
3
4
5
6
65.95
71.44
71.90
71.94
71.94
71.94
65.95
5.50
0.46
0.04
0.00
0.00
Costs
Total Inc.
77511
107690
130199
148116
163141
176331
Mgl.
Ave.
77511
1175 1175
30179
5492 1507
22509
49146 1811
17917 470262 2059
15025 4695313 2268
13190 43966667 2451
Guaiac Test (3)
Not covered
in class
• Major point is the distinction between mgl. and average.
• Test-specific numbers depend on independence across
patients (believable) and across tests (less so).
• One useful way is to look at first test, and then at battery of
the next 5. Even this gives you a marginal cost of
$16,000+.
• Suppose you had followed everyone, and ultimately found
that 1 person w/ negatives had gotten cancer. Sensitivity
goes from 11 of 12 (0.917) to 11 of 18 (0.611).
• Also, some of the big numbers depend on very small
denominators.
Which costs do we
Not covered
class
lookinat?
• Consider a world with only 2 diseases D1 and D2. Half get
D1, half get D2.
• At age 60, individuals get D1 or D2.
• They face a 5% chance of dying from D1 or a 5% chance
of dying from D2, or 10% in all.
• Mortality risk can be reduced (that year) by 100% with
either treatment T1 (for D1) or T2 (for D2).
• After 10 years, they die suddenly and unavoidably.
• So (spreadsheet):
Treatment
T1
T2
Treatment Cost
5000
5200  Model Input
Which costs do we
Treatment
Treatment Cost
T1
5000
Not covered
class
lookinat?
T2
5200
 Model Input
Evaluating the cost-effectiveness ratio.
Effectiveness is the reduction in the probability of death *
incremental lifespan. For each treatment, it is
0.05 * 10 years = 0.5 years.
Effectiveness
0.5 years
0.5 years
This will be the
denominator
Not covered
T2 (D2)-related
future
in class
If only T1 (D1),
costs are included
Treatment
T1
Treatment Cost
5000
Effectiveness
0.5 years
Palliative treatment
per year *10
9000
These costs are conditional.
Unconditional costs
Pal * 0.95 * 10
8550
T2
5200  Model Input
0.5 years  Model Input
1000
950
If only T1, T2-related future
Not covered
classincluded
costsinare
Treatment
T1
T2
Treatment Cost
5000
5200  Model Input
Effectiveness
0.5 years
0.5 years  Model Input
Palliative treatment
per year *10
9000
1000
These costs are conditional.
Unconditional costs
Pal * 0.95 * 10
8550
950
Total Exp. Costs
13550
6150
Must compare to what would occur with no intervention.
Not covered
in class
If NO Intervention?
• Expected future costs would be the palliative treatment that
they would be having if they lived past age 60 and had
either D1 or D2. These costs equal:
0.9 (prob. of getting past age 60) * pal costs/yr. * 10 years
Pal * 0.9 * 10
8100
900
So, incremental costs are:
Total Exp. Costs
13550
6150
less
8100
900
Incremental Costs
5450 >
5250
Cost/yr. improv.
10900
10500
Treatment 2 is cheaper than Treatment 1.
Suppose we include
Not covered
in class
UNRELATED
COSTS
• Here’s the idea
– If a person gets palliative treatment for both D1 and
D2, or $10,000. Some of the costs must be unrelated
to the treatment, because the person can get
intervention for D1 OR D2 -- not both.
Treatment
Treatment Cost
10,000 * 0.95 * 10
Total Exp. Costs
T1
5000
9500
14500
T2
5200  Model Input
9500
14700
Not covered
in class
If NO Intervention?
• Expected future costs would be the palliative treatment that
they would be having if they lived past age 60 and had care
for both D1 and D2. These costs equal:
0.9 (prob. of getting past age 60) * pal costs/yr. * 10 years
Pal * 0.9 * 10
9000 !
9000 !
So, incremental costs are:
Total Exp. Costs
less
Incremental Costs
14500
14700
9000
9000
5500 <
5700
Cost/yr. improvement 11000
11400
Treatment 1 is cheaper than Treatment 2!
Why?
Not covered
in class
• Including the “unrelated” future costs tilts
the balance against intervening in the
relatively less expensive disease (because
you are adding LARGE additional
expenses), and:
• Tilts the balance toward intervening in the
relatively more expensive disease -- you’re
adding additional expenses.
small
Issues!
Not covered
in class
• Consider interventions. What are the future
costs and benefits related to heart disease
interventions?
… seatbelt interventions?
Most of the future costs may be unrelated to
the particular intervention, but what about
costs (like stroke or disease) which may be
related to side effects of intervention?
What to include
Not covered
in class
• Weinstein-Manning suggest that you use either ALL the
costs or NONE of them.
• Meltzer uses a human capital approach that says that
individuals whose lives are extended should be charged for
being more costly,
but credited for being more productive. This suggests that
it may be economically more beneficial to save the lives of
more productive members of society.