Cross-disciplinary research / Health economics / Research

Self-reported Health Outcomes (1) : Summary Statistics

In many health economics and outcomes research studies, patient outcomes are typically measured by three general indicators – morality rate, length of hospital stay, and xx-day re-hospitalization rate with xx varying from 30 to 90 in most cases. From literature discussed previously on health outcomes measurement, we can see a structure along the time dimension from the point of care onward: with the mortality rate being the immediate outcomes, the length-of-stay being the intermediate, and re-hospitalization rate, the long-term outcomes / consequence of care.

These indicators can be considered as objective outcomes generally applicable to all hospitalized patients, especially those with acute conditions. These general outcomes, however, may not be appropriate for all patients at all time. For example, mortality rate may not be immediately relevant to patients with chronic diseases. To measure the morbidity and functional aspects of health, generic health survey such as EQ-5D, SF-36 were used. Besides generic health surveys, condition-specific outcomes sets are being developed by ICHOM – an international organization dedicated to the standardization of health outcomes measurement.

Two major trends in the development of health outcomes measurement stood out to me:

  • The measures of health and health outcomes are growing in dimensions, from one dimensional (time), to multi-dimensional (health surveys); from univariate generic health state (HRQoL) to multivariate condition-specific outcomes sets (see ICHOM).
  • The measures of health and health outcomes are also growing in variety, from objective outcomes (simple statistics) to a growing number of subjective outcomes (time and risk preferences, choice, self-reports on health outcomes, wellbeing, and patient satisfaction, etc).

These growingly sophisticated instruments are designed and developed to capture important changes in every aspect of a patient’s life, making outcome measurement more informative and relevant to all stakeholders including patients, providers, and health authorities as they strive to increase efficiency in health care market and achieve better health through health care quality improvement programs and other reform initiatives.

As more measures and metrics are developed and included in health outcomes research, practice, and policy decisions, it is worth noting that there exist a delicate balance between simplicity and complexity in the choice of outcomes measurement. The reason being that as we increase the complexity by moving away from a few simple, easy-to-interpret, general measures to multiple arrays of specific, sophisticated, and difficult-to-generalize measures, we increase the computational complexity while sacrifice the interpretability and generalizability of these results, making  it more costly to implement and less efficient to communicate and compare between cases and providers. As a result, this could lead to a worrying scenario in which the very purpose of value-based health care is compromised. Economist Michael E. Porter has an excellent NEJM article on this issue which you can read here.

What I would like to focus on here is the measurement of subjective health outcomes using survey data as a case study. A good place to begin research on this subject may be to examine data from the Medicare Health Outcomes Survey (HOS), the first patient-reported outcomes measure used in Medicare managed care in the U.S..

The goal of the Medicare HOS program is to gather valid and reliable clinically meaningful data that have many uses, such as for targeting quality improvement activities and resources; monitoring health plan performance and rewarding top-performing health plans; helping beneficiaries make informed healthcare choices; and advancing the science of functional health outcomes measurement. (see http://www.hosonline.org/)

To have an overview on the dataset, a summary statistics table for the “2012-2014 Cohort 15 Analytic PUF Data File” is shown below. In this public-use data file, baseline data collected in 2012 and its follow-up in 2014 were merged into one table by beneficiaries ID. Please note that I have created my own labels to name each variable to make the code more readable. I will show a few data exploration results in the next post. (Please see more about this dataset at http://www.hosonline.org/en/research-data-files/).

 Descriptive statistics


Statistic N Mean St. Dev. Min Max
AGE 296,320 2.2 0.7 1 3
RACE 255,121 1.3 0.6 1 3
GEND 285,169 1.6 0.5 1 2
MARI 264,436 1.5 0.5 1 2
EDU 262,838 2.2 0.8 1 3
BMI 257,286 1.3 0.5 1 2
health.0 278,059 3.1 1.0 1 5
activ.0 274,534 2.1 0.8 1 3
climb.0 267,392 2.0 0.8 1 3
plim.0 272,936 2.5 1.3 1 5
klim.0 267,077 2.5 1.4 1 5
elim.0 272,872 1.9 1.2 1 5
care.0 265,854 1.8 1.2 1 5
pain.0 272,189 2.5 1.3 1 5
calm.0 271,602 2.8 1.4 1 6
engy.0 270,200 3.5 1.5 1 6
feel.0 268,855 4.8 1.4 1 6
social.0 271,714 3.9 1.3 1 5
physical.0 272,092 3.2 0.9 1 5
mental.0 268,460 2.9 0.9 1 5
bath.0 270,034 1.2 0.5 1 3
dress.0 269,966 1.2 0.5 1 3
eat.0 269,393 1.1 0.3 1 3
getio.0 269,461 1.3 0.5 1 3
walk.0 269,312 1.4 0.6 1 3
uset.0 269,324 1.1 0.4 1 3
phys.days.0 260,984 8.1 11.9 0 98
ment.days.0 262,390 5.6 10.3 0 90
bad.days.0 240,568 6.7 11.2 0 90
cpain.act.0 266,266 4.5 1.0 1 5
cpain.rest.0 266,898 4.7 0.8 1 5
short.lying.0 266,286 4.5 0.9 1 5
short.sit.0 265,641 4.6 0.8 1 5
short.walk.0 264,661 4.1 1.3 1 5
short.climb.0 262,970 3.8 1.4 1 5
feet.numb.0 267,325 4.1 1.3 1 5
feet.tingle.0 266,935 4.2 1.2 1 5
feet.temp.0 265,336 4.5 1.1 1 5
feet.wound.0 266,495 4.8 0.7 1 5
arth.pain.0 264,773 2.9 1.4 1 5
vision.0 266,870 1.1 0.3 1 2
hear.0 260,801 1.2 0.4 1 2
hbp.0 267,433 1.3 0.5 1 2
cad.0 263,718 1.9 0.3 1 2
chf.0 264,926 1.9 0.3 1 2
ami.0 265,539 1.9 0.3 1 2
other.heart.0 265,046 1.8 0.4 1 2
stroke.0 266,266 1.9 0.3 1 2
copd.0 266,223 1.8 0.4 1 2
bowel.0 264,786 1.9 0.2 1 2
arth.low.0 265,572 1.6 0.5 1 2
arth.up.0 265,237 1.6 0.5 1 2
bone.0 264,181 1.8 0.4 1 2
sciat.0 264,545 1.7 0.4 1 2
diabetes.0 266,723 1.7 0.4 1 2
cancer.0 267,046 1.9 0.4 1 2
colon.0 117,256 2.0 0.1 1 2
lung.0 116,225 2.0 0.1 1 2
breast.0 115,693 2.0 0.2 1 2
prostate.0 111,780 1.9 0.2 1 2
back.0 266,523 3.7 1.4 1 5
dep.2wk.0 265,951 1.7 0.5 1 2
dep.much.0 265,819 1.8 0.4 1 2
dep.most.0 264,187 1.8 0.4 1 2
dep.1wk.0 265,364 1.6 0.9 1 4
health.comp.0 266,353 2.9 1.1 1 5
smoke.0 266,849 2.8 0.6 1 4
uleak.0 262,804 1.6 0.5 1 2
uleak.level.0 135,191 2.2 0.7 1 3
uleak.visit.0 119,907 1.6 0.5 1 2
uleak.intv.0 118,713 1.7 0.4 1 2
pa.visit.0 259,960 1.5 0.6 1 3
pa.intv.0 255,932 1.5 0.5 1 2
fall.visit.0 263,469 1.8 0.5 1 3
fall.0 266,257 1.7 0.4 1 2
balance.0 265,094 1.6 0.5 1 2
fall.intv.0 259,384 1.7 0.5 1 3
bone.test.0 262,101 1.5 0.5 1 2
who.0 285,176 1.2 0.4 1 4
PCNT 296,320 89.8 25.8 1.4 100.0
BLANK 296,320 1.1 0.3 1 4
health.1 120,369 3.0 1.0 1 5
activ.1 119,261 2.2 0.8 1 3
climb.1 116,215 2.0 0.8 1 3
plim.1 118,693 2.5 1.3 1 5
klim.1 115,616 2.5 1.3 1 5
elim.1 118,912 1.8 1.2 1 5
care.1 115,605 1.8 1.1 1 5
pain.1 119,007 2.4 1.3 1 5
calm.1 118,899 2.6 1.3 1 6
engy.1 118,383 3.4 1.4 1 6
feel.1 117,945 4.9 1.3 1 6
social.1 118,944 4.0 1.2 1 5
physical.1 119,080 3.2 0.9 1 5
mental.1 117,830 2.9 0.8 1 5
bath.1 118,710 1.2 0.5 1 3
dress.1 118,641 1.1 0.4 1 3
eat.1 118,491 1.1 0.3 1 3
getio.1 118,411 1.2 0.5 1 3
walk.1 118,152 1.4 0.5 1 3
uset.1 118,470 1.1 0.4 1 3
meal 118,460 1.3 0.7 1 3
money 118,360 1.2 0.5 1 3
meds 118,269 1.1 0.4 1 3
phys.days.1 113,901 7.2 10.8 0 96
ment.days.1 113,985 4.6 9.0 0 95
bad.days.1 106,946 5.6 10.0 0 96
vision.1 118,209 1.9 0.3 1 2
hear.1 117,991 1.9 0.3 1 2
think.1 117,798 1.8 0.4 1 2
do.1 117,761 1.8 0.4 1 2
memo.1 116,161 4.2 1.0 1 5
hbp.1 118,004 1.3 0.5 1 2
cad.1 116,700 1.9 0.3 1 2
chf.1 117,076 1.9 0.3 1 2
ami.1 117,061 1.9 0.3 1 2
other.heart.1 117,052 1.8 0.4 1 2
stroke.1 117,505 1.9 0.3 1 2
copd.1 117,513 1.8 0.4 1 2
bowel.1 117,158 1.9 0.2 1 2
arth.low.1 117,304 1.6 0.5 1 2
arth.up.1 117,073 1.6 0.5 1 2
bone.1 116,827 1.8 0.4 1 2
sciat.1 116,896 1.8 0.4 1 2
diabetes.1 117,670 1.7 0.4 1 2
depress.1 117,009 1.8 0.4 1 2
cancer.1 117,010 1.8 0.4 1 2
colon.1 57,316 2.0 0.1 1 2
lung.1 56,699 2.0 0.1 1 2
breast.1 56,730 2.0 0.2 1 2
prostate.1 54,869 1.9 0.2 1 2
other.cancer.1 56,696 1.9 0.2 1 2
pain.living.1 117,204 2.2 1.2 1 5
pain.social.1 117,141 1.8 1.1 1 5
pain.scale.1 115,827 3.6 2.6 1 10
lowint.1 116,026 1.5 0.9 1 4
feel.1.1 115,612 1.4 0.8 1 4
health.comp.1 117,845 2.7 1.1 1 5
smoke.1 117,996 2.8 0.5 1 4
uleak.1 116,546 1.6 0.5 1 2
uleak.level.1 60,494 2.2 0.7 1 3
uleak.visit.1 53,643 1.6 0.5 1 2
uleak.intv.1 53,746 1.7 0.4 1 2
pa.visit.1 115,391 1.5 0.6 1 3
pa.intv.1 113,615 1.5 0.5 1 2
fall.visit.1 116,725 1.8 0.5 1 3
fall.1 117,808 1.8 0.4 1 2
balance.1 117,521 1.6 0.5 1 2
fall.intv.1 115,466 1.7 0.5 1 3
bone.test.1 116,373 1.5 0.5 1 2
who.1 121,161 1.1 0.4 1 4
PCNT.1 174,402 67.2 45.2 0.0 100.0
LANG 174,402 1.6 0.9 1 4
REGI 296,320 5.4 2.6 1 10
ELIG 296,320 1.4 0.5 1 2
RESP 296,320 2.5 1.5 1 5

You can read the data file with the following code in R:

[code language=”r”]
setwd("/Users/your_name/your_working-dir")
time <- system.time(C15A <- read.fwf(
"C15A_PUF.TXT",
widths = c(9, rep(1, 26), 2,2,2,
rep(1, 50), 3, 2, 5, 1,
rep(1, 23), 2,2,2,
rep(1, 27), 2,
rep(1, 16), 3, 2, 5,
1, 3, 2, 1, 1)))
[/code]

And the first a few columns of the data frame can be views at:

ID AGE RACE GEND MARI EDU BMI health.0 activ.0 climb.0
1 A15000001 3 1 2 2 1 1 3 2 3
2 A15000002 3 1 1 1 3 1 4 2 2
3 A15000003 2 1 2 2 3 2 3 3 3
4 A15000004 3 1 1 1 3 1 3 3 2
5 A15000005 2 4 2 1
6 A15000006 3 1 2 2 1 1 2 2 2
7 A15000007 3 1 2 2 2 1 3 3 3
8 A15000008 2 1 2 1 3 1 4 2 1
9 A15000009 2 1 2 1 2 1 4 2 3
10 A15000010 2 1 2 3 2 4 1 1

Also, for more info on the measurement of subjective well-being, please see

SAGE_Meeting_Dec2012_StoneA

Ph.D. in Economics with interests in Life Science, Behavioral Science, Health Economics Evaluation and Health Technology Assessment. Executive MBA student at The University of Chicago Booth School of Business in Hong Kong, graduating in 2020.

1 Comment

Leave a Reply

This site uses Akismet to reduce spam. Learn how your comment data is processed.