Cost-Effectiveness and Cost-Benefit Analysis

N287E Spring 2006Joanne Spetz31 May 2006

Optimal planning involves comparing marginal benefit and marginal cost

If a central health planner needs to pick the “right” level of production, how will she choose?

Social marginal benefit = Social marginal

cost

We can’t always see the “margin”

We can make choices with discrete projects

Benefit of project A >Benefit of project B

This is the purview of cost-effectiveness and cost-benefit analysis

Good CEA/CBA requires good measurement

How to measure benefits?

What are the benefits? Extended life Better quality of life

Reduced morbidity

Measurement options for benefits

Number of disease cases avertedNumber of years of life gained But what about differences in quality

of life?

Dollar value of saved life, reduced medical costs

Quality-adjusted life years

QALYs are a common way to weight for quality of life Each year of life is weighted by the

expected quality

Quality-adjusted life years

How do you determine QALYs? Clinical experts Experimental data Surveys

There is usually not measurement of individual preferences

Example of a QALY calculation

70-year-old man20 year life span10 years of perfect health10 years of 50% quality health(10 x 1 QALY/yr)+(10 x 0.5 QALY/yr) =

15QALYs

Some thoughts about QALYs

Social health can be the sum of population QALYsTrade-offs and comparisons can be made across people

What about creating monetary measures of benefits?

Option 1: Cost of illness Direct cost of medical care resources Indirect costs

Morbidity cost: wages lost due to inability to work and value of housekeeping

Mortality cost: present value of future earnings

Future earnings are “discounted” and summed

Problems with cost of illness

Men valued more than women (due to higher earnings)Ethnic groups valued differentlyChildren valued less than adultsWhat about quality of life?What is the value of non-market work?

What about creating monetary measures of benefits?

Option 2: Willingness to pay Choices that individuals make reveal

information about how they value life Information about value of life comes

from Surveys Data on consumer behavior

Problems with willingness to pay (and one benefit)

Surveys People can say anything Responses vary with income People misestimate risk

Consumer behavior Consumers might not have accurate risk

information Do you count the risk of injury?

WTP more directly addresses the concept of marginal utility

A few more thoughts on financial value of life

Willingness to pay usually values life more highly than discounted future earningsMost studies value life between $1 and $6 millionIn order to make policy decisions, a value of life must be made explicitly or implicitly

What is cost-benefit analysis?

Cost-benefit analysis compares costs and benefits, with benefits measured monetarilyNet benefit = (Bt-Ct)/(1+r)t If net benefit > 0 then do the project

Or, net benefit = (Bt/(1+r)t) (Ct/(1+r)t)

If net benefit > 1 then do the project

Measurements of costs

What is the viewpoint of the analysis Often “social perspective” Viewpoint of a particular agency The patient

What are the comparisons? Two or more treatment programs?

What categories of costs to include? How to handle capital expenditures How to discount for future costs

It is important to not double-count costs and benefits

Example: building a stadium “Jobs created”

But, we pay wages for the jobs! The payments for wages are a cost

“Property values will go up and businesses will get more income” The higher property values reflect the

increased income

To make a CBA-based decision…

Projects can be ranked by net benefitsInformal judgements can be made after ranking Income distribution can be considered Age distribution can be considered

What is cost-effectiveness analysis

CBA is disliked by many health professionalsCost-effectiveness analysis determines the cost of a certain Number of cases of disease prevented Number of QALYs obtained Other non-monetary measurement of

benefits

A classic example of CEA

What do we gain from the sixth stool guaiac? New England Journal of Medicine 1975 293: 226-228

Background on sixth stool guaiac

Six sequential tests for occult bloodIf any test is positive, a barium enema is done

The decision tree is:Test 1

Etc.

Test 2

Test 3

Test 4

neg

neg

neg

neg

pos

pos

pos

pos

enema

enema

enema

enema

What are the detection rates?

~72 people of 10,000 have colon cancer P(detection) = 91.66% P(false positive) = 36.51%

The first test detects 0.9166 x 72 = 65.9952 casesThe second test detects 99.3% of cases .993 = (.9166+.9166(1-.9166))

What about costs?

$4 for the initial test$1 for each additional test$100 for the barium enema

A detection grid for 10,000 people, Test 1

Cancer?

YES NO

POS 66 (91.66%)

309 (36.51%)

375 people

NEG 6 (1-.9166)

9619 9625 people

72 people 9928 people Go to Test

2

Test

resu

lt

Costs from first test

$4 x 10,000 = $40,000 for initial test$100 x 375 = $37,500 for barium enemasTotal = $77,500

A detection grid for Test 2 – 9625 people

Cancer?

YES NO

POS 5.5 299 304.50 people

NEG 0.5 9320 9320.5 people

6 people 9619 people Go to Test

3

Test

resu

lt

Cost from Test 2

$1 x 9625 for repeat test$100 x 304.5 for barium enemasTotal = $40,075

A detection grid for Test 3 – 9320.5 people

Cancer?

YES NO

POS .458 290 290.458 people

NEG .042 9030 9030.042 people

0.5 people

9320 people Go to Test

4

Test

resu

lt

Cost from Test 3

$1 x 9320.5 for repeat test$100 x 290.458 for barium enemasTotal = $38,366.30

How many cases are detected?

True positive False positive

# tests

% # cases % # cases

1 91.6667

65.9469

36.5079

309.1652

2 99.3056

71.4424

59.6876

505.4606

3 99.9421

71.9003

74.4048

630.0926

4 99.9952

71.9385

83.7491

709.2240

5 99.9996

71.9417

89.6819

759.4661

6 99.9999

71.9420

93.4489

791.3660

Cost analysis

# tests

# cases detected

Incre-ment

Total cost

Incre-ment

Marg cost/ marg gain

Average cost

1 65.9469

65.9469

$77,511 $77,511

$1,175 $1,175

2 71.4424

5.4956 $107,690

$30,179

$5,492 $1,507

3 71.9003

0.4580 $130,199

$22,509

$49,150 $1,810

4 71.9385

0.0382 $148,116

$17,917

$469,534 $2,059

5 71.9417

0.0032 $143,141

$15,024

$4,724,695

$2,268

6 71.9420

0.0003 $176,331

$13,190

$47,107,214

$2,451

Comparison

What if the barium enema was done for all patients?Total cost for 10,000 people would be $1,000,000Average cost per case detected = $13,900Marginal cost = $13,900

Sensitivity analysis

It is important to consider the assumptions made in the analysis Should some assumptions be

changed?

New technologies or information could affect conclusions

Sensitivity analysis in guaiac paper

Assume: Protocol detects only 60% of cases

per screening OR Population prevalence is lower,

11/10,000

Sensitivity analysis of costs

# tests Marg cost baseline

Marginal cost –

60% sensitive

Marginal cost –

low prevalence

1 $1,175 $1,743 $7,152

2 $5,492 $1,816 $35,505

3 $49,150 $3,353 $321,991

4 $469,534 $6,584 $3,078,108

5 $4,724,695

$13,696 $30,912,409

6 $47,107,214

$29,941 $325,476,150

What is an acceptable cost-effectiveness ratio?

There is no clear answer!Even if the ratio seems reasonable, we still might not be able to afford the cost We might not be able to invest today

for future gains

Most studies compare their cost-effectiveness ratios to those of well-accepted treatments/screenings

Can you apply this to nurse staffing?

Needleman & Buerhaus, Health Affairs 2006Rothberg, Abraham, et al., Medical Care 2005

Needleman & Buerhaus

Estimated costs of: More total nursing staff Substituting RNs for other staff

Benefits were cost savings in: Shorter length of stay Fewer adverse events Fewer deaths

Needleman & Buerhaus

Results Substituting RNs for other staff

produces net cost reduction Increasing total nursing hours

improves outcomes but also increases costs about 1.5% more

Rothberg et al.

Staffing comparison: 8:1 ratio vs. 4:1 ratio

Benefits: Cost savings from reduced LOS Lower patient mortality

Method: Statistical analysis with random variation in effects of staffing. Outcome = costs per life year saved

Rothberg et al.

Results: 1:8 was least expensive, highest mortality Mortality improved and costs rose and

nursing ratio became richer Incremental cost-effectiveness was $136,000

(95% CI $53,000-402,000) per life saved.

Sensitivity analysis: Sensitive to the effects of ratios on mortality Throughout the ranges tested, ICER <=

$449,000 per life saved.