1. Measuring Health Outcomes
Thitima Kongnakorn
Community of Scholars
October 9, 2002

2. Measuring Health Outcomes
Clinical Decision Analysis
drug choice, specialty care, disease management program
Cost Effectiveness Analysis
economic aspects
Health Technology Assessment
Drug evaluation, screening tests, surgical interventions, medical devices, health promotion technology

3. Terminology Health Status Measure Health Profile
Used generally to refer to all of these measures
Health Profile
A health status measure that is a vector of scores on different dimensions (e.g. SF-12)
Quality of Life Measure
Preference based health status measures

4. Terminology HALYs: Health-Adjusted Life Years
Using a health status measure for health weights
QALYs: Quality-Adjusted Life Years
A type of HALY computed using a HRQOL measure for health weights

5. Evolution of Output Units
Cost per “case” (e.g., $/cancer found)

6. Evolution of Output Units
cost per “case”
cost per life saved ($/life saved)

7. Evolution of Output Units
cost per “case”
cost per life saved
cost per life-year saved ($/LY saved)

8. Evolution of Output Units
cost per “case”
cost per life saved
cost per life-year saved
cost per quality-adjusted life year ($/QALY saved)

9. QALYs = area under this curve
1.0
quality of
life
additional years of life
now
death
QALYs = area under this curve
QALE = average number of QALYs experienced
by a cohort of the same starting age
and quality of life

10. Longer life, and higher quality of life, so QALYs gained is larger.
Suppose: intervention changes life path from this point
1.0
quality of
life
now
death
additional years of life
Ideal outcome:
Longer life, and higher quality of life,
so QALYs gained is larger.

11. QALYs gained Ideal outcome: Longer life, and higher quality of life,
1.0
quality of
life
now
death
additional years of life
Ideal outcome:
Longer life, and higher quality of life,
so QALYs gained is larger.

12. 1.0
quality of
life
additional years of life
now
death
Shorter life, but higher quality...
total QALYs may be greater or smaller

13. QALYs gained
1.0
quality of
life
additional years of life
now
death
QALYs lost
Shorter life, but higher quality...
total QALYs may be greater or smaller

14. Longer life, but lower quality mostly...
1.0
quality of
life
additional years of life
now
death
Longer life, but lower quality mostly...
QALYs may be larger or smaller

15. Disease Specific General Health non-preference preference based Many!
e.g. ?
joint counts
total cholesterol
physical measures
SIP, Rand GHS,
COOP, MOS short
forms EVGFP
Many!
e.g., Roland Scale,
VFQ-25
rating scales
preference based
QWB, HUI,
EQ-5D ?
indexed
patient’s own prefs.
ad hoc
ad hoc

16. Disease-specific measure...
more sensitive to the particular dysfunction
often seem objective
designed to be sensitive to changes from treatment for a specific disease
acceptable to clinicians because focused on aspects of one health condition -- often measure things they strive to change with treatment.

17. But disease-specific measures may miss things
Many people (especially when older) have multiple health conditions
Many treatments have unintended effects (arthritis & hearing)

18. Allows many comparisons:
Why an interest in measures of General Health? (aka “generic measures”)
Allows many comparisons:
across diseases
in people with multiple conditions
across studies
Needed for cost-effectiveness studies

19. Medical Outcomes Study -- “short forms”
Derived from Rand General Health Survey
Originally 250+ questions
Published short forms that are in use:
SF-12
SF-20
SF-36

20. SF-36 & SF-12 8 components, scaled worst=0 to best=100
Physical functioning
Role function (from physical limitation)
Pain
General Health
Vitality
Social functioning
Role function (from emotional limitation)
Mental health

21. New Scaling for SF-36 & SF-12
PCS : physical component scale
MCS: mental component scale proprietary scoring systems that combine the 8 scales into 2.

22. Measuring Health State Utility
Methods that require the subjects to explicitly trade health against something else that they value
Measure of QOL
Use in calculating QALYs

23. Making Choices – Measuring Utility
Life Expectancy
 Time tradeoff
Probability of survival  Standard Gamble

24. Weight for health state
Time Tradeoff (TTO) X
Vary X until Life A ~ Life B
Life A:
Health state
10 yr
Life B:
Excellent health
Weight for health state X = 10

25. TTO
Scaled to be “QALY”-like
Related to choice
Easier to use than SG

26. Problems with TTO
Difficult to apply to “short-term” health states (e.g. radiologic diagnostic tests)
Unrealistic for a patient to visualize himself/herself in an excellent health state and compare to a short-term unpleasant health state

27. For remaining life expectancy
Standard Gamble
Profile 235
For remaining life expectancy
Life A:
Live remaining LE in
excellent health
P %
Life B:
1-P %
Die immediately
Vary P until Life A  Life B,
then P = health state weight

28. Standard Gamble
Method directly from decision theory incorporating attitudes about risk
Has been used with apparent success in many settings
Many report hard to understand
Not representative of decision at hand
Weights often very near 1.0

29. Results from Empirical Data
Questionnaire Based
SF-12 (Generic)
VFQ-25 (Disease-Specific)
Visual Functioning Questionnaire (25 questions)
Utility-Based
TTO (easy to understand)
Standard Gamble (hard to understand)

30. Subjects 66 subjects Age range: 54 – 99, Average Age: 77
25 males, 41 females
4 Groups (classified by visual acuity)
20/20 – 20/40 (n = 31)
20/20 – 20/50 with AMD (n = 14)
20/60 – 20/100 with AMD (n = 9)
Worse than 20/100 with AMD (n = 12)

31. Time Trade-Off
Assume that your current life expectancy is 20 years from now.
  Suppose there is a technology that can return your eyesight to perfectly normal in both eyes. The technology always works but your length of life will be decreased to 10 years. So, would you be willing to give up 10 years of your life to receive this technology and have perfect vision for your remaining years?
  [The question continues by increasing or decreasing length of life with bisection technique until reaching an indifferent point.]

32. Standard Gamble
Suppose there is a technology that can return your eyesight to normal. When it works, patients respond perfectly and have normal vision in both eyes for the rest of their lives. When it doesn’t work, however, the technology fails and patients do not survive (for example, death under anesthesia). Thus, it either restores perfect vision or causes immediate death.
If there is a 50 percent chance of death, will you accept or refuse to take this technology?
The question continues by increasing or decreasing percent chance of death with bisection technique until reaching an indifferent point.

33. Questionnaire-Based Disease-Specific(VFQ-25) vs Generic (SF-12)

34. Utility-Based Time Tradeoff vs Standard Gamble

35. Correlations TTO is significantly correlated with SG, and VFQs
SG is significantly correlated with TTO, and VFQs
Only PCS is significantly correlated with VFQs
VFQs are significantly correlated with TTO, SG, and PCS

36. Conclusions
VFQ-25 is sensitive to measure outcomes for patients with visual impairment
When using generic measure, SF-12, people did not really take their visual impairment into account
People try to avoid “chance of death” rather than “losing remaining years of life”
VFQs, TTO, and SG are significantly correlated

37. Future Steps More literature review Try to link to HCI Any ideas???
Health outcome measurement
Investigate the limitations of each measurement
Try to link to HCI
Any ideas???