Content uploaded by Richard Lipton
Author content
All content in this area was uploaded by Richard Lipton
Content may be subject to copyright.
ORIGINAL CONTRIBUTION
Lost Productive Time and Cost Due to
Common Pain Conditions in the US Workforce
Walter F. Stewart, PhD, MPH
Judith A. Ricci, ScD, MS
Elsbeth Chee, ScD
David Morganstein, MS
Richard Lipton, MD
P
AIN IS A COMMON HUMAN
malady that spares no group
and often impairs function.
Pain occurs in a variety of
forms, including monophasic events
(eg, due to injury), chronic episodic
conditions (eg, migraine headache), and
chronic persistent problems (eg, per-
sistent pain from arthritis). The broad-
based impact of pain, especially among
working-age populations, is likely to
have significant cost implications. A
number of studies have described the
impact of pain disorders (migraine,
1-10
tension-type headache,
11-12
back
pain,
13-18
arthritis,
19-23
and general mus-
culoskeletal disorders)
24-25
on work.
Although these and other pain con-
ditions have a profound impact on the
ability to work, available research has
substantial limitations. Most studies
focus on a single pain disorder and
do not provide composite estimates
across a range of common pain
disorders.
1-13,15-19,21-26
Many studies on the
work-related impact of pain focus on lost
time due to absenteeism
7-11,13-28
; few have
estimated the pain-related impact of re-
duced performance while at work.
1-6,12,29
This limitation is important because in-
creasing evidence indicates that re-
duced work performance due to pain, not
absenteeism, is the dominant cause of lost
productive time.
1-6,12,29-34
Some studies fo-
cus on specific employers
2,7
or other
populations limited in generalizability in
other ways.
4,27,29
Finally, no study has
quantified lost time due to common pain
conditions capturing both absenteeism
and health-related reduced perfor-
mance on days at work in a representa-
tive sample of the US workforce and then
translated those estimates into eco-
nomic terms.
The American Productivity Audit pro-
vides an opportunity to better under-
stand the impact of pain on the US work-
force. The American Productivity Audit
captures a large, representative na-
tional sample of the US workforce and
assesses lost productive time due to
health conditions, with a specific focus
on common pain conditions. Survey re-
spondents report time absent due to pain
overall and due to specific pain condi-
tions and reduced performance while at
work due to pain overall and due to spe-
cific pain conditions. We estimate pain-
Author Affiliations: AdvancePCS Center for Work and
Health, Hunt Valley, Md (Drs Stewart, Ricci, and Chee);
Geisinger Health Systems, Danville, Pa (Dr Stewart);
Statistical Group, Westat, Rockville, Md (Mr Morgan-
stein); and Department of Neurology, Albert Einstein
College of Medicine, Bronx, NY (Dr Lipton).
Corresponding Author and Reprints: Walter F.
Stewart, PhD, MPH, Center for Health Research and
Rural Advocacy, Geisinger Health Systems, 100 N
Academy Ave, Danville, PA 17822 (e-mail:wfstewart
@geisinger.edu).
Context Common pain conditions appear to have an adverse effect on work, but
no comprehensive estimates exist on the amount of productive time lost in the US
workforce due to pain.
Objective To measure lost productive time (absence and reduced performance due
to common pain conditions) during a 2-week period.
Design and Setting Cross-sectional study using survey data from the American Pro-
ductivity Audit (a telephone survey that uses the Work and Health Interview) of work-
ing adults between August 1, 2001, and July 30, 2002.
Participants Random sample of 28902 working adults in the United States.
Main Outcome Measures Lost productive time due to common pain conditions
(arthritis, back, headache, and other musculoskeletal) expressed in hours per worker
per week and calculated in US dollars.
Results Thirteen percent of the total workforce experienced a loss in productive time
during a 2-week period due to a common pain condition. Headache was the most
common (5.4%) pain condition resulting in lost productive time. It was followed by
back pain (3.2%), arthritis pain (2.0%), and other musculoskeletal pain (2.0%). Work-
ers who experienced lost productive time from a pain condition lost a mean (SE) of
4.6 (0.09) h/wk. Workers who had a headache had a mean (SE) loss in productive
time of 3.5 (0.1) h/wk. Workers who reported arthritis or back pain had mean (SE)
lost productive times of 5.2 (0.25) h/wk. Other common pain conditions resulted in a
mean (SE) loss in productive time of 5.5 (0.22) h/wk. Lost productive time from com-
mon pain conditions among active workers costs an estimated $61.2 billion per year.
The majority (76.6%) of the lost productive time was explained by reduced perfor-
mance while at work and not work absence.
Conclusions Pain is an inordinately common and disabling condition in the US work-
force. Most of the pain-related lost productive time occurs while employees are at work
and is in the form of reduced performance.
JAMA. 2003;290:2443-2454 www.jama.com
©2003 American Medical Association. All rights reserved. (Reprinted) JAMA, November 12, 2003—Vol 290, No. 18 2443
at Colorado State University, on November 20, 2006 www.jama.comDownloaded from
related lost productive time and the as-
sociated costs due to headache, back
pain, arthritis, and other musculoskel-
etal pain in the US workforce in aggre-
gate and individually for each pain dis-
order.
METHODS
The American Productivity Audit is a
national survey of the US workforce
35
completed using the Work and Health
Interview (W.F.S., unpublished data,
2003).
36,37
The survey was completed
by IMR, a survey and clinical research
division of AdvancePCS. The Work and
Health Interview captures data on work
absence, reduced performance while at
work, and health-related causes of work
absence and reduced performance. The
study protocol and the informed con-
sent statement were approved by the Es-
sex institutional review board.
Work and Health Interview
The structure, development, and vali-
dation of the Work and Health Inter-
view is described in detail elsewhere
(W.F.S., unpublished data, 2003).
36,37
In brief, the Work and Health Inter-
view, a computer-assisted telephone in-
terview, comprises 8 modules. The first
3 modules capture detailed data on em-
ployment status, usual work time, and
the presence of 22 health conditions.
In particular, specific questions were
asked about headache or pain in the
back, feet, hands, wrists, or other places
in the past 2 weeks, and about arthri-
tis or pain in 1 or more joints in the past
12 months. The question regarding ar-
thritis and joint pain was followed by
a question about the specific location
of the pain and the frequency with
which it occurred in the past 2 weeks.
A job visualization module ensures that
respondents focus on general descrip-
tions of their work before answering
questions about reduced work perfor-
mance due to pain. Questions were
asked about tasks and activities per-
formed at work, the time allotted to
each, and those deemed most impor-
tant. Participants also characterized oc-
cupations in terms of job demand and
job control.
38
Two modules quantify lost produc-
tive time. A missed workday module
quantifies the number of missed work-
days and health-related cause(s). The
module on lost productive time for days
at work asks about missed hours (ie,
partial workdays) and reduced perfor-
mance on days at work not feeling well,
and health-related cause(s). Not feel-
ing well was broadly defined during the
interview as a health condition that ei-
ther comes and goes or as an ongoing
health condition. Validation of the lost
productive time metric have been de-
scribed in detail elsewhere (W.F.S., un-
published data, 2003).
37.38
The respondent ascribed the cause(s)
of work absence(s) and of reduced per-
formance while at work. If an indi-
vidual reported lost productive time (ie,
either time absent from work or re-
duced performance) in the previous 2
weeks and reported having more than
1 pain condition (during the first part
of the interview), they were reminded
of the pain conditions that they had re-
ported. They were then asked to select
the primary reason for their time ab-
sent from work or reduced perfor-
mance while at work. At the end of the
interview, information on salary was ob-
tained.
Household Sampling and Selection
of Household Members
Households were selected as a ran-
dom sample of residences with tele-
phones in the 48 contiguous states and
the District of Columbia. Genesys Sam-
pling Systems (Fort Washington, Pa)
provided a probability sample of resi-
dential telephone numbers and house-
holds were called on different days of
the week and at different times of the
day. Respondents were deemed eli-
gible if they (1) were aged 18 to 65
years; (2) were a permanent member
of the household; (3) responded yes to
the Current Population Survey (CPS)
39
question on employment status: “Last
week, did you do any work for either
pay or profit?”; and (4) were em-
ployed in current job at least 14 days.
A 1 in 10 random sample of adults who
responded no to the CPS question was
also selected to participate if the re-
spondent was aged 18 to 65 years and
a permanent member of the house-
hold.
If more than 1 eligible adult was a
member of the household, we selected
the person whose next birthday would
occur closest to the day of the inter-
view. This procedure approximates a
probability-based selection method
without the need to enumerate all eli-
gible members of the household.
40
Ver-
bal informed consent was obtained be-
fore initiating the interview. Once an
interview was completed, the inter-
viewer requested to speak with the next
eligible member of the household who
would have a birthday. Up to 2 eli-
gible members per household were in-
terviewed to optimize the efficiency of
the sampling strategy.
41
Data Collection and Benchmarking
Data collection began on August 1,
2001, and continued for 1 year. Ap-
proximately 2500 interviews were com-
pleted each month. The sample in-
cluded individuals who worked for pay
or profit in the past 7 days (ie, occu-
pation-eligible) and a 10% random
sample of individuals who did not work
for pay or profit in the past 7 days (ie,
occupation ineligible). Details on the
participation are described else-
where.
35,42
A total of 33996 respon-
dents agreed to participate in the sur-
vey (ie, gave a complete or partial
interview) and 30523 completed the
full interview. Of this number, 28902
(92.2%) were occupation-eligible. Over-
all participation was estimated at
66.2%.
35,42
A 2-step weighting method was used
to account for selective participation.
One weight was applied to individual
participants as the inverse of the num-
ber of telephone lines available for
incoming calls to account for the
unequal probability of selecting house-
holds. Second, a population-weight-
ing adjustment accounted for selec-
tion bias due to incomplete coverage of
the US population and to ensure that
estimates of certain sample demo-
graphic subgroup totals conformed to
LOSS IN PRODUCTIVE TIME DUE TO PAIN
2444 JAMA, November 12, 2003—Vol 290, No. 18 (Reprinted) ©2003 American Medical Association. All rights reserved.
at Colorado State University, on November 20, 2006 www.jama.comDownloaded from
known values for these totals. The CPS
was used as the external reference da-
tabase because it provided high preci-
sion estimates on a nationally repre-
sentative sample of the US workforce.
A raking method was used for popula-
tion weighting adjustment, benchmark-
ing to 4 variables (age, sex, region, and
a composite of work status and num-
ber of missed work hours) common to
both the American Productivity Audit
and the CPS. Raking used an iterative
proportional fitting procedure to en-
sure that the weights assigned to indi-
vidual respondents lead to marginal dis-
tributions on auxiliary variables that
were equivalent in the American Pro-
ductivity Audit and the CPS.
43
Wesvar
statistical software (Version 4, We-
stat, Rockville, Md) was used to per-
form the raking adjustments.
Analysis
Analysis was restricted to the 28 902 oc-
cupation-eligible respondents who com-
pleted the interview. Analyses were first
completed to describe variation in
health-related lost productive time
among workers by selected character-
istics. The method for estimating lost
productive time from Work and Health
Interview data is described in detail else-
where.
37,42
Lost productive time for a per-
sonal health reason was the sum of hours
per week absent from work for a health-
related reason (absenteeism) and the
hour equivalent of health-related re-
duced performance on days at work
(“presenteeism”). Absenteeism in-
cluded missed workdays and reduced
work hours on days at work during the
recall period. Reduced performance at
work was quantified based on re-
sponses to 6 questions.
For 5 of the 6 questions, respon-
dents were asked how often, on aver-
age during the recall period, they lost
concentration, repeated a job, worked
more slowly than usual, felt fatigued at
work, and did nothing at work on days
when they were at work not feeling
well. Response options were all of the
time, most of the time, half of the time,
some of the time, and none of the time.
A sixth question asked respondents
about the average amount of time it
took them to start working after arriv-
ing at work on days not feeling well dur-
ing the recall period. The aggregate
measure of reduced performance was
then derived in 4 steps: (1) convert the
categorical response options for 5 of the
6 questions into percentages as fol-
lows: all of the time (100%), most of
the time (75%), half of the time (50%),
some of the time (25%), and none of
the time (0%); (2) average the re-
sponses to the 5 categorical behavior
questions to yield the average percent-
age of lost productive work time and
multiply this percentage by the num-
ber of hours worked per day to yield its
hour equivalent; (3) add the hours of
lost productive work time to the re-
ported average amount of time it took
to start working after arriving at work;
and (4) divide by the number of weeks
per recall period for the hours per week
of lost productive time on days at work.
Respondents attributed the cause of
their health-related lost productive time
to specific health conditions. In this
study, we targeted the common pain
conditions that affect both men and
women in the workplace. These in-
cluded arthritis, back pain, headache,
and other musculoskeletal pain. We did
not include a number of less common
conditions in our estimate of pain-
related lost productive time (eg, pain
associated with cancer or cancer treat-
ment, diabetic neuropathy) or condi-
tions that do not affect both men and
women (eg, menstrual pain).
Lost labor costs were estimated by
translating hours of lost productive time
into lost dollars using self-reported an-
nual salary or wages. Lost dollars were
calculated by multiplying lost hours by
hourly earnings. Data were first sum-
marized to describe the percentage of
workers with lost productive time (ie,
absenteeism and health-related re-
duced performance on days at work)
due to pain in the previous 2 weeks. We
estimated the percentage of all work-
ers with pain-related lost productive
time in the previous 2 weeks, and the
percentage with 2 h/wk or more of pain-
related lost productive time in the pre-
vious 2 weeks. Estimates were derived
for any pain and separately for each of
the 4 common pain condition catego-
ries (ie, headache, back pain, arthritis,
musculoskeletal pain).
We describe variation in the propor-
tion of all individuals in a defined group
who reported 2 h/wk or more of lost pro-
ductive time due to pain. Two or more
hours was selected as a meaningful
threshold for lost productive time. In this
population-level analysis, crude propor-
tions were derived. To determine if 2
h/wk or more of pain-related lost pro-
ductive time varied by demographic and
other features, we analyzed data using
a generalized linear model framework
(SAS Proc GENMOD). The log of the ex-
pectation of each binary response vari-
able (ie, ⱖ2 h/wk of pain-related lost
productive time vs no time) was mod-
eled as a linear function of the explana-
tory variable. Log link was used so that
parameters could be interpreted as
prevalence ratios (ie, proportion with ⱖ2
h/wk of pain-related lost productive time
in one group divided by the same mea-
sure in the reference group) rather than
as odds ratios. We also restricted analy-
sis to those who reported an episode of
pain-related lost productive time in the
previous 2 weeks and described varia-
tion in mean lost productive time by co-
variates among these individuals. Varia-
tion in lost productive time was modeled
using linear regression (SAS Proc GLM).
Variation in lost productive time was
evaluated in relation to a number of co-
variates considered to be relevant to
employers and policy makers. These fac-
tors included sex, age (18-29, 30-39, 40-
49, or 50-65 years), race (white, black,
or other), education (⬍high school
diploma, high school diploma or
GED, some college or associate degree,
bachelor degree, or graduate degree),
annual salary (⬍$10000, $10 000-
$19999, $20000-$29999, $30000-
$39999, $40000-$49999, or ⱖ$50000),
type of occupation (white collar or blue
collar), composite job-demand and job-
control category (high demand-high
control, high demand-low control, low
demand-high control, or low demand-
low control) based on Karasek et al,
38
LOSS IN PRODUCTIVE TIME DUE TO PAIN
©2003 American Medical Association. All rights reserved. (Reprinted) JAMA, November 12, 2003—Vol 290, No. 18 2445
at Colorado State University, on November 20, 2006 www.jama.comDownloaded from
duration of time at job (⬍0.5, 0.5-0.9,
1-1.9, 2-2.9, 3-4.9, 5-9.9, 10-19.9, or ⱖ20
years), month of interview (January-
February, March-April, May-June, July-
August, September-October, or
November-December), health insur-
ance (insured or not insured), and num-
ber of nonpain health conditions re-
ported in previous 2 weeks (0-1, 2, 3,
or ⱖ4).
Nonpain health conditions were de-
fined as all health conditions reported
in the previous 2 weeks excluding the
pain conditions targeted by this re-
search (ie, headache, arthritis, back pain,
or other musculoskeletal pain). We also
included geographic region (North-
east, South, Midwest, or West) as a
broad-based surrogate for possible so-
ciocultural differences in views on work.
Occupations coded according to the
1998 Standard Occupational Classifica-
tion System (Bureau of Labor Statis-
tics, US Department of Labor) were di-
chotomized as white collar or blue collar
according to US Office of Personal Man-
agement definitions.
44
White collar jobs
included professional, administrative, or
support-type occupations; blue collar
jobs included trade or labor occupa-
tions.
44
Imputation procedures for missing
values in benchmarking and weight-
ing variables and annual salary are de-
scribed elsewhere.
42
SAS statistical soft-
ware was used for all analysis (Version
8.2, SAS Institute Inc, Cary, NC).
RESULTS
A profile of participants is described
elsewhere
35
and available on request.
Fifty-six percent of participants were
women. Respondents were equally dis-
tributed across 4 age groups (18-29, 30-
39, 40-49, and 50-65 years), a major-
ity were white (77.0%), 67% were
formally educated beyond high school,
83% were working more than 30 h/wk,
and 51% earned less than $40000 per
year. The most common occupational
category was office or administrative
support (16.4%), which was followed
by sales (9.3%), and education/training/
library occupation (7.6%). Benchmark-
ing (ie, reweighting in reference to the
CPS) resulted in several significant dis-
tributional changes. Compared with the
participation sample, reweighting pri-
marily influenced the percentage dis-
tribution by sex, age (ie, more adults
aged 18-29 years and fewer adults aged
40-49 years), and geographic region.
For the latter, weighting was in-
creased for underrepresentation in the
West and decreased for overrepresen-
tation in the South.
A total of 52.7% of the workforce re-
ported having headache, back pain, ar-
thritis, or other musculoskeletal pain
in the past 2 weeks. Overall, 12.7% of
the workforce lost productive time in
a 2-week period due to a common pain
condition; 7.2% lost 2 h/wk or more of
work. Headache was the most com-
mon pain condition resulting in lost
productive time, affecting 5.4% (2.7%
with ⱖ 2 h/wk) of the workforce
(T
ABLE 1), which was followed by back
pain (3.2%), arthritis (2.0%), and other
musculoskeletal pain (2.0%). Among
those who lost productive time due to
a pain condition, an average of 4.6 h/wk
was lost (Table 1). The mean lost pro-
ductive time was lowest for headache
(3.5 h/wk) and highest for other mus-
culoskeletal pain (5.5 h/wk). Absence
days were uncommon. A total of 1.1%
of the workforce was absent from work
1 or more days per week from 1 of the
4 pain conditions; 0.12% were absent
2 d/wk or more. Headache and back
pain were dominant causes of missed
days of work. Overall, lost productive
time due to health-related reduced per-
formance on days at work accounted for
4 times more lost time than absentee-
ism. The ratio of lost productive time
due to health-related reduced perfor-
mance on days at work compared with
absenteeism varied among categories of
pain disorders: headache, 4.5 h/wk; ar-
thritis, 6.5 h/wk; back pain, 2.9 h/wk;
and other musculoskeletal pain, 3.6
h/wk.
Variation in Lost Productive Time
Factors that determine variation in lost
productive time were examined in the
total workforce (estimated percentage
who lost ⱖ2 h/wk of productive time
due to pain) and among the sub-
sample of the workforce with some
pain-related lost productive time (es-
timated mean lost productive time in
hours per week) (T
ABLE 2 and
T
ABLE 3). For the first method, the
prevalence measure is a composite of
Table 1. US Workforce With Lost Productive Time Due to Common Pain Conditions in the
Previous 2 Weeks
*
Type of Pain
Headache Arthritis Back Other† Any
Total lost productive time
⬎0 h/wk due to pain in past 2 wk‡§ 5.43 2.03 3.20 2.02 12.68
ⱖ2 h/wk due to pain in past 2 wk‡§ 2.72 1.23 1.97 1.32 7.24
Hours per worker per week,
mean (SE)§㛳
3.51 (0.1) 5.19 (0.25) 5.28 (0.25) 5.47 (0.22) 4.56 (0.09)
Missed workdays‡¶
ⱖ1 d/wk due to pain in past 2 wk 0.39 0.11 0.39 0.23 1.12
ⱖ2 d/wk due to pain in past 2 wk 0.02 0.01 0.06 0.03 0.12
Absent, hours per worker per week,
mean (SE)㛳
0.64 (0.05) 0.69 (0.12) 1.35 (0.16) 1.20 (0.13) 0.92 (0.05)
Reduced performance at work due to
pain, hours per worker per week,
mean (SE)㛳
2.87 (0.09) 4.50 (0.26) 3.93 (0.19) 4.27 (0.22) 3.64 (0.09)
*
Values expressed as percentages unless otherwise indicated. Estimates benchmarked to the Current Population Sur-
vey.
†Includes unspecified musculoskeletal pain.
‡Denominator includes 28 902 occupation-eligible participants.
§Total lost productive time includes periods when employees were either absent or present (health-related reduced
performance due to pain).
㛳Denominator includes occupation-eligible participants with more than 0 hours of lost productive time attributed to
pain in the previous 2 weeks: headache (n = 1688), arthritis (n = 608), back (n = 947), musculoskeletal (n = 611), or
any pain (n = 3830).
¶Refers to full days of work missed due to pain.
LOSS IN PRODUCTIVE TIME DUE TO PAIN
2446 JAMA, November 12, 2003—Vol 290, No. 18 (Reprinted) ©2003 American Medical Association. All rights reserved.
at Colorado State University, on November 20, 2006 www.jama.comDownloaded from
the prevalence of the pain condition in
the population in general and the fre-
quency of pain episodes that actually
have an impact on work function.
No difference in the proportion of the
workforce losing 2 h/wk or more due
to pain was observed by sex, age, re-
gion of residence, type of occupation
(ie, blue collar or white collar), dura-
tion of time in job, month of inter-
view, or health insurance status after ad-
justed prevalence ratios were examined
(Table 2 and Table 3). On the other
hand, blacks exhibited a 20% excess
prevalence of 2 h/wk or more of pain-
related lost productive time compared
with whites (P =.002), and a small dif-
ference was observed by annual salary
(ie, overall P =.002; prevalence ratio
[PR] for lowest income group com-
pared with the highest income group
was 0.70 [95% confidence interval {CI},
0.57-0.85]). The largest differences
were observed by education (P⬍.001)
in which the prevalence of 2 h/wk or
more of pain-related lost productive
time was inversely related to level of
educational attainment. Compared with
those with a high school diploma, the
PR was 1.24 (95% CI, 1.05-1.46) for
those without a high school degree and
0.71 (95% CI, 0.59-0.85) for those with
a graduate degree. Job demand-
control category also significantly im-
proved the fit of the model (P⬍.001).
Individuals with high control jobs (ie,
high demand-high control and low de-
mand-high control) were 30% to 40%
more likely to have lost 2 h/wk or more
Table 2. Prevalence of Pain-Related Lost Productive Time in the Total Sample by Demographics
*
Demographic
Pain-Related Lost Productive Time in Previous 2 wk
ⱖ2 h/wk
⬎0 h/wk
No. of
Participants
(N = 28 902)
Crude
Prevalence, %
Adjusted
PR (95% CI)†
No. of
Occupational-Eligible
Participants
(n = 3830)
Crude
Mean (SE),
h/wk
Adjusted
Mean (SE),
h/wk†
Sex
Men 12 701 6.48 0.92 (0.83-1.01) 1351 5.5 (0.2) 5.6 (0.4)
Women 16 201 8.82 1.00 2479 4.6 (0.2) 4.5 (0.3)
Age, y
18-29 6453 8.63 1.00 887 4.6 (0.2) 4.9 (0.4)
30-39 7043 8.40 1.00 (0.89-1.13) 983 5.2 (0.2) 5.3 (0.4)
40-49 8416 7.82 0.96 (0.85-1.09) 1150 5.1 (0.2) 5.2 (0.4)
50-65 6990 6.36 0.85 (0.74-0.98) 810 4.7 (0.2) 4.8 (0.4)
Race
White 22 246 7.60 1.00 2984 4.7 (0.1) 4.5 (0.3)
Black 2579 9.58 1.22 (1.08-1.39) 357 6.3 (0.4) 6.0 (0.4)
Other 2720 7.64 0.98 (0.85-1.13) 327 5.0 (0.4) 4.7 (0.5)
Not stated 1357 7.70 NA 162 5.8 (0.7) NA
Education
⬍12th grade; no diploma 1517 11.01 1.24 (1.05-1.46) 250 6.7 (0.6) 6.4 (0.5)
High school graduate or GED 8134 8.07 1.00 1128 5.1 (0.2) 5.2 (0.3)
Some college or associate degree 8561 8.78 1.04 (0.94-1.15) 1249 4.9 (0.2) 4.9 (0.4)
Bachelor degree 6439 6.53 0.81 (0.71-0.92) 741 4.4 (0.2) 4.6 (0.4)
Graduate degree 3139 5.24 0.71 (0.59-0.85) 323 4.0 (0.3) 4.2 (0.5)
Not stated 1112 8.31 NA 139 5.7 (0.7) NA
Annual salary, $
⬍10 000 2416 7.19 0.70 (0.57-0.85) 342 4.2 (0.3) 4.2 (0.5)
10 000-19 999 4171 9.13 0.89 (0.77-1.04) 643 4.7 (0.3) 4.9 (0.4)
20 000-29 999 5392 9.02 0.96 (0.84-1.11) 808 4.9 (0.2) 4.9 (0.4)
30 000-39 999 5113 8.03 0.98 (0.86-1.13) 695 5.2 (0.3) 5.4 (0.4)
40 000-49 999 3613 7.62 1.04 (0.90-1.21) 461 5.0 (0.3) 5.1 (0.4)
ⱖ50 000 6991 6.23 1.00 741 5.1 (0.3) 5.8 (0.4)
Not stated 1206 7.40 NA 140 5.6 (0.7) NA
Region of residence
Northeast 5438 7.37 1.00 (0.87-1.14) 716 4.4 (0.2) 4.6 (0.4)
South 10 544 8.17 1.03 (0.92-1.16) 1020 4.7 (0.2) 5.2 (0.3)
Midwest 7735 7.44 0.98 (0.86-1.11) 1419 5.2 (0.2) 4.8 (0.4)
West 5185 7.97 1.00 675 5.3 (0.3) 5.4 (0.4)
Abbreviations: CI, confidence interval; GED, General Education Development (test); PR, prevalence ratio.
*
Estimates are not benchmarked to the Current Population Survey.
†Adjusted for all other covariates included in this table and in Table 3. NA indicates PR (95% CI) was not calculated.
LOSS IN PRODUCTIVE TIME DUE TO PAIN
©2003 American Medical Association. All rights reserved. (Reprinted) JAMA, November 12, 2003—Vol 290, No. 18 2447
at Colorado State University, on November 20, 2006 www.jama.comDownloaded from
of lost productive time due to pain com-
pared with those with low demand-
low control jobs (Table 3). Finally, a
strong statistically significant gradient
(P =.01) was observed in relation to the
number of other health conditions re-
ported to occur in the previous 2 weeks
(ie, responses to specific questions
about health conditions). Compared
with those with 4 or more pain condi-
tions, the proportion losing 2 h/wk or
more of pain-related lost productive
time was significantly lower among
those with 3 pain conditions (PR, 0.76;
95% CI, 0.68-0.84), 2 (PR, 0.56; 95%
CI, 0.50-0.63), and 1 or no (PR, 0.35;
95% CI, 0.31-0.40) other pain condi-
tion (Table 3).
Among those with some productive
time lost due to pain, no difference in
mean lost productive time per week was
observed by age, region of residence, type
of occupation (ie, white collar or blue
collar), duration of time in job, month
of interview, or health insurance status
(Table 2 and Table 3). Crude mean (SE)
lost productive time was significantly
higher (P⬍.001) for males (5.5 [0.2]
h/wk) than females (4.6 [0.2] h/wk) and
for blacks (adjusted mean [SE], 6.0 [0.4];
Table 3. Prevalence of Pain-Related Lost Productive Time in the Total Sample by Employment and Health Characteristics
*
Characteristic
Pain-Related Lost Productive Time in Previous 2 wk
ⱖ2 h/wk
⬎0 h/wk
No. of
Participants
(N = 28 902)
Crude
Prevalence, %
Adjusted
PR (95% CI)†
No. of
Occupational-Eligible
Participants
(n = 3830)
Crude
Mean (SE),
h/wk
Adjusted
Mean (SE),
h/wk†
Type of occupation
White collar 18 162 7.66 1.03 (0.80-1.33) 2413 4.4 (0.1) 4.6 (0.3)
Blue collar 9312 8.35 1.07 (0.83-1.39) 1271 5.9 (0.2) 5.5 (0.3)
Undetermined 1428 5.72 1.00 146 4.8 (0.6) 5.1 (0.7)
Job demand/control
High/high 10 039 10.91 1.29 (1.11-1.51) 1805 5.3 (0.2) 5.5 (0.3)
High/low 8364 7.14 0.92 (0.78-1.08) 1133 4.1 (0.2) 4.6 (0.4)
Low/high 2778 13.76 1.44 (1.21-1.71) 573 5.9 (0.3) 5.8 (0.4)
Low/low 2189 8.14 1.00 315 4.0 (0.3) 4.4 (0.5)
Not stated 5532 0.02 NA 4 3.1 (2.4) NA
Duration of time in job, y
⬍0.5 3569 8.09 0.97 (0.79-1.19) 501 4.5 (0.3) 4.9 (0.4)
0.5-0.9 3185 8.94 1.11 (0.91-1.34) 486 5.0 (0.3) 5.4 (0.4)
1-1.9 3805 8.55 1.07 (0.88-1.29) 520 4.9 (0.3) 5.1 (0.4)
2-2.9 2520 8.20 0.99 (0.81-1.22) 344 4.9 (0.3) 5.0 (0.5)
3-4.9 3351 7.81 1.01 (0.84-1.23) 451 5.1 (0.3) 5.2 (0.4)
5-9.9 4573 7.59 1.00 (0.84-1.20) 594 5.3 (0.3) 5.2 (0.4)
10-19.9 4871 6.98 0.94 (0.78-1.12) 595 4.9 (0.3) 4.9 (0.4)
ⱖ20 2818 6.74 1.00 325 4.7 (0.4) 4.6 (0.5)
Not stated 210 3.37 NA 14 3.1 (1.2) NA
Month of interview
January-February 6820 7.08 0.93 (0.82-1.04) 818 5.0 (0.2) 5.3 (0.4)
March-April 4757 7.79 1.00 (0.88-1.13) 633 5.0 (0.3) 5.3 (0.4)
May-June 4371 6.93 0.89 (0.78-1.02) 520 5.0 (0.3) 5.2 (0.4)
July-August 1987 7.38 0.98 (0.81-1.18) 274 4.3 (0.4) 4.4 (0.5)
September-October 3762 9.53 1.08 (0.95-1.22) 585 4.9 (0.3) 4.9 (0.4)
November-December 7205 8.18 1.00 1000 5.0 (0.2) 5.2 (0.4)
No. of pain conditions reported
in previous 2 wk
0-1 13 881 3.11 0.35 (0.31-0.40) 969 3.2 (0.2) 3.3 (0.4)
2 5050 8.23 0.56 (0.50-0.63) 827 4.0 (0.2) 4.3 (0.4)
3 3977 11.52 0.76 (0.68-0.84) 753 5.1 (0.3) 5.4 (0.4)
ⱖ4 5994 15.88 1.00 1281 6.7 (0.2) 7.2 (0.4)
Health insurance
Yes 24 592 7.63 1.00 3232 4.9 (0.1) 5.2 (0.3)
No 2889 9.47 1.07 (0.92-1.23) 434 5.3 (0.4) 4.9 (0.4)
Not stated 1421 7.21 NA 164 5.2 (0.6) NA
Abbreviations: CI, confidence interval; PR, prevalence ratio.
*
Estimates are not benchmarked to the Current Population Survey.
†Adjusted for all other covariates included in this table and in Table 2. NA indicates PR (95% CI) was not calculated.
LOSS IN PRODUCTIVE TIME DUE TO PAIN
2448 JAMA, November 12, 2003—Vol 290, No. 18 (Reprinted) ©2003 American Medical Association. All rights reserved.
at Colorado State University, on November 20, 2006 www.jama.comDownloaded from
P⬍.001) compared with whites (ad-
justed mean [SE], 4.5 [0.3]) (Table 2).
A statistically significant gradient
(P⬍.01) of increasing mean lost produc-
tive time was observed in relation to de-
creasing education. In contrast, a more
modest, but statistically significant
(P =.049) gradient of increasing mean
lost productive time was observed in re-
lation to increasing salary (Table 2).
Other notable differences were ob-
served for job demand/control (P⬍.001)
and number of other health conditions
(P⬍.001) in a manner that mirrored the
PRs. Individuals in high-control jobs had
higher mean pain-related lost produc-
tive time per week than those in low-
control jobs. Mean pain-related lost pro-
ductive time per week was directly
related to number of other reported
health conditions (Table 3).
Sex- and Age-Specific Occurrence
by Pain Condition
Although we did not find substantial
overall differences by sex and age in the
proportion of the workforce affected by
significant pain episodes (ie, ⱖ2 h/wk
of pain-related lost productive time),
differences were observed for specific
pain-related conditions. In males and
females, headache was dominant at a
younger age, peaking between ages 25
and 29 years and declining thereafter
(F
IGURE). The proportion of the work-
force with 2 h/wk or more of headache-
related lost productive time was ap-
proximately 2 times higher in females
than males. In contrast, the preva-
lence of 2 h/wk or more of lost produc-
tive time due to arthritis pain in-
creased with increasing age in both
males and females.
Cost of Lost Productive Time
in the US Workforce
The percentage distribution of lost pro-
ductive time (in hours) and lost pro-
ductive time costs (in dollars) are sum-
marized by demographic (T
ABLE 4)and
other factors (TABLE 5). We estimated
the cost of total lost productive time at-
tributed to common pain conditions in
the US workforce in hours and dollars
because of the influence of salary on
cost estimates. Differences between the
lost productive time distributions ex-
pressed in hours and in dollars are ex-
plained by variation in the average
hourly cost of labor by various sub-
groups. For example, individuals with
an annual salary of $50000 or more ac-
count for only 22% of the lost produc-
tive time in hours but 42% of the lost
productive time cost in dollars. These
estimates are limited to workers ac-
tively engaged in work and amount to
$61.2 billion per year (T
ABLE 6). A total
of 76.6% of this cost occurs while em-
ployees are at work and is explained by
health-related reduced performance.
The pain-related reduced perfor-
mance on days at work component of
the lost productive time cost varies
somewhat by condition with a low of
69.7% for back pain and a high of 84.4%
for arthritis.
COMMENT
Overall, the estimated $61.2 billion per
year in pain-related lost productive time
in our study accounts for 27% of the total
estimated work-related cost of pain con-
ditions in the US workforce.
35
Lost pro-
ductive time varied to some degree in the
workforce. First, little or no variation was
observed by age. In large part, the lack
of differences by age was due to the
counterbalancing effects of different pain
conditions. Headache, common at
younger ages (ie, 18-34 years), rapidly
declines in prevalence thereafter. In con-
trast, the other 3 pain conditions are ei-
ther more common with increasing age
(eg, arthritis) or peak at a later age than
headache (eg, back pain).
The relatively strong inverse relation-
ship with education that we found in our
study may be explained by several fac-
tors. First, for some conditions like mi-
graine,
45
prevalence is inversely related
to education. Potentially hazardous work
conditions, physically demanding work,
or other risk factors may be more com-
mon among those with a lower educa-
tion level and lead to restricted activity
days, an established finding for back
pain.
17
Second, access to medical care
and, more generally, health literacy are
known to vary by education level
46
and
influence access to treatment and qual-
ity of care.
47
Third, our finding could be
confounded by factors associated with
common pain disorders as well as with
lost productive time. Depression is one
such confounder that may be particu-
larly important because it is strongly and
inversely related to education in the
workforce
42
and often co-occurs with a
number of pain conditions including
back pain
42
and migraine.
48
Our method of estimating the im-
pact of pain on work productivity dif-
fers in several respects from previous
studies. First, our focus is on individu-
als who experienced a recent episode of
pain that impaired their ability to work.
Most other studies have captured more
general information (eg, frequency of
episodes and average effect) over longer
recall periods to estimate the effect of
pain. We used a 2-week period to en-
sure accurate recall of episodes of health
problems that impair work function. In
a previous study,
37
we showed that re-
call of health-related lost productive time
was underestimated with a 4-week re-
call period and possibly overestimated
due to telescoping with a 1-week recall
period.
Second, in our study, pain was de-
noted as a primary cause of lost pro-
ductive time only if the respondent
made the specific attribution. In pre-
vious studies, those with a specific pain
condition are identified first, and sub-
sequently, the impact of the condition
on work is assessed. This method is
prone to overestimation bias in 2 ways:
respondents report lost productive time
for a specific cause, which may result
in their overattributing lost produc-
tive time to that cause; or all work lost
during a recall interval when the pain
condition was present is attributed to
that condition. In contrast, our method
may be prone to underestimate lost pro-
ductive time due to selected condi-
tions deemed socially undesirable as a
cause for missing work (eg, headache).
Third, we did not use diagnostic
questions to identify individuals with
different pain conditions. We used re-
call prompts at the beginning of the in-
terview by asking whether specific
LOSS IN PRODUCTIVE TIME DUE TO PAIN
©2003 American Medical Association. All rights reserved. (Reprinted) JAMA, November 12, 2003—Vol 290, No. 18 2449
at Colorado State University, on November 20, 2006 www.jama.comDownloaded from
Figure. Respondents With 2 h/wk or More of Lost Productive Time Due to Pain Condition by Age
7
3
2
1
4
5
6
0
7
3
2
1
4
5
6
0
Age Group, y
Percentage
18-24 40-4425-29 30-34 35-39 60-6545-49 50-54 55-59
Age Group, y
18-24 40-4425-29 30-34 35-39 60-6545-49 50-54 55-59
Headache
Men Women
Arthritis
7
3
2
1
4
5
6
0
7
3
2
1
4
5
6
0
Age Group, y
Percentage
18-24 40-4425-29 30-34 35-39 60-6545-49 50-54 55-59
Age Group, y
18-24 40-4425-29 30-34 35-39 60-6545-49 50-54 55-59
Back Pain
3.5
1.5
1.0
0.5
2.0
2.5
3.0
0
3.5
1.5
1.0
0.5
2.0
2.5
3.0
0
Age Group, y
Percentage
18-24 40-4425-29 30-34 35-39 60-6545-49 50-54 55-59
Age Group, y
18-24 40-4425-29 30-34 35-39 60-6545-49 50-54 55-59
Total No. of
Participants 1661 17761292 1467 1677 5981703 1420 953
3.5
1.5
1.0
0.5
2.0
2.5
3.0
0
7
3
2
1
4
5
6
0
Age Group, y
Percentage
18-24 40-4425-29 30-34 35-39 60-6545-49 50-54 55-59
Age Group, y
1896
18-24
2415
40-44
1562
25-29
1803
30-34
2063
35-39
783
60-65
2224
45-49
1924
50-54
1268
55-59
Other Pain
LOSS IN PRODUCTIVE TIME DUE TO PAIN
2450 JAMA, November 12, 2003—Vol 290, No. 18 (Reprinted) ©2003 American Medical Association. All rights reserved.
at Colorado State University, on November 20, 2006 www.jama.comDownloaded from
health conditions occurred in the pre-
vious 2 weeks or 12 months (ie, for
chronic diseases). Later in the inter-
view, health-related lost productive time
was assessed and the respondent was
asked to attribute a cause. In general,
our method is likely to be accurate in
identifying episodes of lost produc-
tive time in which pain is the cause. It
is also likely to be accurate for condi-
tions such as headache in which a di-
agnosis is not required. On the other
hand, respondents may understate the
extent to which back pain or other mus-
culoskeletal pain is caused by arthritis
unless they have received a medical di-
agnosis for arthritis.
No study to date provides informa-
tion on the work impact of all com-
mon pain conditions. A few studies
have quantified lost productive time and
projected these costs to the US work-
force. Hu et al,
5
using data from sev-
eral population-based studies, esti-
mated the work-related costs from
migraine in the United States to be $13
billion. Schwartz et al
28
concluded that
the lost work time from tension-type
headache was similar to the costs from
migraine headache and that health-
related reduced performance at work,
which was roughly measured, ac-
counted for 70% of the overall work-
related lost productive time from all
headaches. Together, these studies sug-
gest that the total lost productive time
cost from headache is $26 billion com-
pared with our estimate of $19.6 bil-
lion. The differences we have cited be-
tween our methods and the more
traditional methods used in these 2 pre-
vious studies may account for our lower
cost estimate. In addition, as migraine
and tension-type headache are comor-
bid, estimates of migraine alone or ten-
sion-type headache alone may inad-
vertently capture lost productive time
due to the other disorder.
Consistent with other stud-
ies,
5,6,9,26,27,30-34
reduced performance
while at work was the dominant source
(ie, 80% of the lost productive time and
76.6% of the lost productive time-
related cost) of pain-related lost pro-
ductive time in the US workforce. In
previous studies of headache disor-
ders, including daily diary studies of mi-
graine, health-related reduced perfor-
mance on days at work is a more
significant cause of lost work time than
absenteeism.
3
In addition, our previ-
ous analysis
35
indicates that on any
given day relatively few workers are ab-
sent from work. Given that pain con-
ditions are highly prevalent in the work-
force and that work performance is
impaired in a substantial minority of
workers with common conditions, it is
not surprising that a majority of the
pain-related lost productive time we ob-
served results from reduced perfor-
mance while at work. Nonetheless, em-
ployers may challenge the validity of
this finding. Few employers docu-
ment health-related lost productive time
while at work, making it largely invis-
ible and, as a consequence, intangible
and subject to doubt.
Respondents were asked to at-
tribute their lost productive time to a
primary condition. However, pain con-
ditions often co-occur. While we were
not able to identify the extent to which
different pain conditions co-occurred,
we did examine this question in an in-
Table 4. Total Annual Lost Productive Time of US Workers Due to Pain by Demographics
Demographic
Lost Productive Time
Cost Equivalent of
Lost Productive Time
Percentage
*
Hours (SE)† Percentage
*
US $ (SE)‡
US workforce 100 3839.8 (100.3) 100.0 61.2 (2.2)
Sex
Men 48.1 1847.9 (85.8) 52.3 32.0 (1.9)
Women 51.9 1991.9 (66.2) 47.7 29.2 (1.4)
Age, y
18-29 26.2 1007.4 (48.4) 21.1 12.9 (1.0)
30-39 26.3 1010.3 (60.0) 26.5 16.2 (1.2)
40-49 26.8 1028.3 (61.1) 30.0 18.3 (1.5)
50-65 20.7 793.7 (52.0) 22.5 13.8 (1.0)
Race
White 78.0 2994.3 (97.2) 81.2 49.7 (2.2)
Black 11.6 444.5 (39.6) 9.1 5.6 (0.6)
Other 9.6 368.2 (29.0) 9.0 5.5 (0.6)
Not stated 0.8 32.8 (10.4) 0.8 0.5 (0.1)
Education
⬍12th grade; no diploma 8.8 337.4 (34.6) 5.5 3.4 (0.4)
High school graduate or GED 33.2 1277.3 (74.9) 28.4 17.4 (1.5)
Some college or associate degree 34.4 1321.5 (57.7) 34.7 21.3 (1.6)
Bachelor degree 17.0 652.1 (40.6) 22.1 13.5 (1.0)
Graduate degree 6.5 248.9 (27.4) 9.2 5.7 (0.6)
Not stated 0.1 2.5 (1.1) 0.1 0.03 (0.01)
Annual salary, $
⬍10 000 7.6 292.9 (29.9) 1.7 1.0 (0.1)
10 000-19 999 15.7 603.7 (36.1) 7.5 4.6 (0.2)
20 000-29 999 21.9 839.9 (40.0) 14.9 9.1 (0.4)
30 000-39 999 20.6 790.4 (46.4) 18.0 11.0 (0.6)
40 000-49 999 12.7 489.2 (40.9) 15.7 9.6 (1.2)
ⱖ50 000 21.5 823.8 (63.0) 42.2 25.8 (2.0)
Region of residence
Northeast 17.4 670.2 (50.1) 19.0 11.6 (1.1)
South 36.6 1404.3 (73.3) 31.9 19.5 (1.1)
Midwest 23.4 898.6 (50.0) 22.0 13.5 (1.1)
West 22.6 866.6 (50.3) 27.1 16.6 (1.6)
Abbreviation: GED, General Education Development (test).
*
May not equal 100 due to rounding.
†In millions.
‡In billions (2002 US $).
LOSS IN PRODUCTIVE TIME DUE TO PAIN
©2003 American Medical Association. All rights reserved. (Reprinted) JAMA, November 12, 2003—Vol 290, No. 18 2451
at Colorado State University, on November 20, 2006 www.jama.comDownloaded from
dependent national survey of 12416 in-
dividuals. In this survey (conducted in
2002), the interview and sampling
methods were the same as those as de-
scribed herein with one exception. We
asked separate questions about pain
conditions occurring in the last 2 weeks.
Of the 32% of respondents reporting 1
of the 4 pain conditions in the past 2
weeks, 16.4% reported only 1 condi-
tion (5.0% headache, 4.6% arthritis,
4.3% back pain, 2.5% musculoskeletal),
9.6% reported 2 pain conditions, 4.5%
reported 3 conditions, and 1.5% re-
ported all 4. The most common co-
occurring conditions were arthritis and
back pain (4.7%), headache and back
pain (4.5%), and arthritis and muscu-
loskeletal pain (4.3%). The least com-
mon was headache and musculoskel-
etal pain (2.4%).
Our estimate of $61.2 billion per year
in pain-related lost productive time does
not include costs from 4 other causes.
First, we did not include lost produc-
tive time costs associated with dental
pain, cancer pain, gastrointestinal pain,
neuropathy, or pain associated with
menstruation. Second, we do not ac-
count for pain-induced disability that
leads to continuous absence of 1 week
or more. Third, we did not consider sec-
ondary costs from other factors such as
the hiring and training of replacement
workers or the institutional effect
among coworkers.
49
Taking these other
factors into consideration could in-
crease, decrease, or have no net effect
on health-related lost productive time
cost estimates. Fourth, we may be prone
to underestimating current lost pro-
ductive time among those with persis-
tent pain problems (eg, chronic daily
headache). To the extent that these
workers remain employed, they may ad-
just both their performance and per-
ception of their performance over time.
The latter, a form of perceptual accom-
modation, makes it difficult to accu-
rately ascertain the impact of a chronic
pain condition on work in the recent
past through self-report.
An accurate estimate of work im-
pact would require that individuals
compare their recent work perfor-
Table 5. Total Annual Lost Productive Time of US Workers Due to Pain by Employment and
Health Characteristics
Characteristic
Lost Productive Time
Cost Equivalent of
Lost Productive Time
Percentage
*
Hours (SE)† Percentage
*
US $ (SE)‡
Type of occupation
White collar 53.3 2047.7 (57.7) 57.5 35.2 (1.6)
Blue collar 43.6 1675.2 (80.8) 39.9 24.4 (1.6)
Undetermined 3.1 117.0 (24.8) 2.6 1.6 (0.3)
Job demand/control
High/high 50.5 1941.9 (77.5) 57.2 35.0 (1.9)
High/low 25.3 972.5 (49.1) 24.2 14.8 (0.8)
Low/high 17.0 651.2 (46.6) 13.0 7.9 (0.7)
Low/low 7.1 272.3 (22.3) 5.6 3.4 (0.3)
Missing 0.1 2.0 (1.5) 0.03 0.02 (0.01)
Duration of time in job, y
⬍0.5 13.1 504.7 (36.5) 8.7 5.3 (0.4)
0.5-0.9 13.4 512.9 (41.3) 11.2 6.8 (0.8)
1-1.9 14.5 555.1 (43.5) 13.7 8.4 (1.1)
2-2.9 8.7 332.0 (27.2) 7.1 4.3 (0.4)
3-4.9 11.7 450.2 (39.0) 11.6 7.1 (0.7)
5-9.9 15.8 607.3 (39.9) 18.7 11.5 (1.0)
10-19.9 14.7 566.1 (45.2) 17.6 10.8 (1.3)
ⱖ20 7.9 304.3 (34.7) 11.4 7.0 (0.9)
Missing 0.2 7.4 (2.9) 0.1 0.1 (0.02)
Month of interview
January-February 21.1 809.2 (46.0) 21.2 12.9 (1.2)
March-April 17.1 655.5 (51.7) 15.6 9.5 (0.8)
May-June 14.9 572.1 (43.7) 13.6 8.3 (0.7)
July-August 6.8 261.3 (38.8) 7.4 4.5 (0.8)
September-October 15.2 585.8 (47.7) 15.9 9.7 (1.2)
November-December 24.9 955.8 (43.7) 26.5 16.2 (1.1)
No. of health conditions reported
in previous 2 wk
0-1 19.0 728.5 (55.7) 20.9 12.8 (1.1)
2 18.1 695.6 (51.8) 17.9 10.9 (0.9)
3 21.2 814.2 (63.7) 21.1 12.9 (1.2)
ⱖ4 41.7 1601.5 (78.4) 40.2 24.6 (1.7)
Health insurance
Yes 85.7 3290.4 (82.7) 91.7 56.1 (2.0)
No 13.4 516.0 (47.7) 7.6 4.6 (0.5)
Not stated 0.9 33.4 (10.9) 0.7 0.4 (0.1)
*
May not equal 100 due to rounding.
†In millions.
‡In billions (2002 US $).
Table 6. Total Cost of Lost Productive Time Due to Common Pain Conditions in the US
Workforce
Total
Type of Pain, Cost (SE), $
*
Headache Arthritis Back Other†
Total productive time lost 61.3 (2.2) 19.6 (1.0) 10.3 (0.7) 19.8 (1.7) 11.6 (0.9)
Absenteeism 14.4 (1.5) 4.2 (0.6) 1.6 (0.4) 6.0 (1.3) 2.6 (0.3)
At work but work impaired
due to pain
46.9 (1.8) 15.4 (0.7) 8.7 (0.6) 13.8 (1.1) 9.0 (0.8)
*
In billions (2002 US $).
†Includes unspecified musculoskeletal pain.
LOSS IN PRODUCTIVE TIME DUE TO PAIN
2452 JAMA, November 12, 2003—Vol 290, No. 18 (Reprinted) ©2003 American Medical Association. All rights reserved.
at Colorado State University, on November 20, 2006 www.jama.comDownloaded from
mance with that before the onset of the
chronic pain condition. Analyses are
under way using other data (ie, qual-
ity of life, changes in work perfor-
mance since onset of a chronic ill-
ness) to examine the potential
underestimation of lost productive time
linked to accommodation from chronic
pain and other conditions. Finally, in
estimating lost productive time costs,
we have assumed that there is a mon-
etary equivalence between an hour of
work absence and reduced perfor-
mance. The validity of this assump-
tion is likely to vary by work setting,
position, percentage-reduced perfor-
mance, degree of interdependence and
exchangeability of workers, and other
factors.
Our estimates of lost productive time
due to pain should be interpreted in
light of the possible beneficial and ad-
verse effects of current pain treat-
ment. Unfortunately, constraints on in-
terview time did not allow us to collect
detailed treatment data. This issue is
currently being addressed in a supple-
mental study. The opportunity for em-
ployers can be defined by the gap be-
tween lost productive time due to pain
given current use of treatments and lost
productive time due to pain given op-
timal use of treatments. The magni-
tude of this gap is difficult to quantify.
Certainly, its aggregate economic bur-
den, as estimated in this study, is enor-
mous, but we cannot state how much
of the burden can be mitigated. Na-
tional survey data that provide de-
tailed data on use of treatments are lim-
ited. Of the common pain conditions,
sufficient details have only been re-
ported on migraine headaches. Recent
data indicate that only 41% of individu-
als who have migraine headaches in the
US population ever receive any pre-
scription drug for migraine.
50
Only
29% report that satisfaction with treat-
ment is moderate, especially among
those who are often disabled by their
episodes.
51
Randomized trials demon-
strate that optimal therapy for mi-
graine dramatically reduces headache-
related disability time in comparison
with usual care.
52,53
This study pro-
vides a measure of the scope of the
problem, but the benefits of optimal in-
tervention will have to be assessed sepa-
rately for each condition.
In conclusion, pain is costly to em-
ployers. Our estimate of the cost of pain
to the US workforce must be translated
into a form that is relevant to employ-
ers. To this end, we have modeled lost
productive time for employers using a
direct adjustment procedure. Stratum-
and condition-specific estimates of
prevalence and mean lost productive
time per week are estimated from the
American Productivity Audit and ap-
plied to the age and sex distribution of
the employer’s workforce. Lost produc-
tive time (in hours) is translated to dol-
lars using age- and sex-specific wage
data. This is a first step to provide em-
ployers with a more concrete under-
standing of the costs they face from
health conditions in their workforce and
to begin to consider how health care dol-
lars can be more effectively targeted to
population-specific needs. Helping em-
ployers understand the cost of health-
related lost productive time may en-
courage them to make more effective use
of the health care dollars they invest in
their workforce. As the primary pur-
chaser of health care, employers are well
positioned to demand programs that re-
duce the impact of common treatable
pain conditions in the workplace.
Author Contributions: Dr Stewart, as principal inves-
tigator, had full access to all of the data in this study
and takes responsibility for the integrity of the data
and the accuracy of the data analysis.
Study concept and design: Stewart, Ricci, Morgan-
stein, Lipton.
Acquisition of data: Stewart, Ricci.
Analysis and interpretation of data: Stewart, Ricci,
Chee, Morganstein, Lipton.
Drafting of the manuscript: Stewart, Ricci, Chee.
Critical revision of the manuscript for important in-
tellectual content: Stewart, Ricci, Morganstein, Lipton.
Statistical expertise: Stewart, Ricci, Chee, Morganstein.
Obtained funding: Stewart.
Administrative, technical, or material support: Ricci,
Chee.
Study supervision: Stewart, Ricci.
Funding/Support: This work was supported by
AdvancePCS.
Role of the Sponsor: AdvancePCS provided finan-
cial support for the American Productivity Audit and
for coauthors who are staff of IMR to complete the
article. The sponsor also paid Dr Stewart to oversee
the analysis and development of the article.
Acknowledgment: We thank Sofia Chaudhry, MPH,
and Carol Leotta, PhD, at AdvancePCS for their con-
tributions to this research.
REFERENCES
1. Stewart WF, Lipton RB, Simon D. Work-related dis-
ability: results from the American Migraine Study.
Cephalalgia. 1996;16:231-238.
2. Michel P, Dartigues JF, Londoulsi A, Henry P. Loss
of productivity and quality of life in migraine suffer-
ers among French workers: results from the GAZEL
cohort. Headache. 1997;37:71-78.
3. Von Korff M, Stewart WF, Simon D, Lipton RB. Mi-
graine and reduced work performance: a population-
based diary study. Neurology. 1998;50:1741-1745.
4. Dasbach EJ, Carides GW, Gerth WC, Santanello NC,
Pigeon JG, Kramer MS. Work and productivity loss in
the rizatriptan multiple attack study. Cephalalgia. 2000;
20:830-834.
5. Hu XH, Markson LE, Lipton RB, Stewart WF, Berger
ML. Burden of migraine in the United States: disabil-
ity and economic costs. Arch Intern Med. 1999;159:
813-818.
6. Gerth WC, Carides GW, Dasbach EJ, Visser WH,
Santanello NC. The multinational impact of migraine
symptoms on healthcare utilisation and work loss. Phar-
macoeconomics. 2001;19:197-206.
7. Schulman EA, Cady RK, Henry D, et al. Effective-
ness of sumatriptan in reducing productivity loss due
to migraine: results of a randomized, double-blind pla-
cebo-controlled clinical trial. Mayo Clin Proc. 2000;
75:782-789.
8. McCunney RJ. Health and productivity: a role for
occupational health professionals. J Occup Environ
Med. 2001;43:30-35.
9. Stang PE, Osterhaus JT. Impact of migraine in the
United States: data from the national health inter-
view survey. Headache. 1993;33:29-35.
10. Stang P, Cady R, Batenhorst A, Hoffman L. Work-
place productivity: a review of the impact of mi-
graine and its treatment. Pharmacoeconomics. 2001;
19:231-244.
11. Rasmussen BK, Jensen R, Olesen J. Impact of head-
ache on sickness absence and utilisation of medical ser-
vices: a Danish population study. J Epidemiol Com-
munity Health. 1992;46:443-446.
12. Schwartz BS, Stewart WF, Simon D, Lipton RB.
Epidemiology of tension-type headache. JAMA. 1998;
279:381-383.
13. Hildebrandt VH. Back pain in the working popu-
lation: prevalence rates in Dutch trades and profes-
sions. Ergonomics. 1995;38:1283-1298.
14. Dartigues JF, Michel P, Lindoulsi A, Dubroca B,
Henry P. Comparative view of the socioeconomic im-
pact of migraine versus low back pain. Cephalalgia.
1998;18(suppl 21):26-29.
15. Rizzo JA, Abbott TA III, Berger ML. The labor pro-
ductivity effects of chronic backache in the United
States. Med Care. 1998;36:1471-1488.
16. Guo HR, Tanaka S, Halperin WE, Cameron LL.
Back pain prevalence in US industry and estimates of
lost workdays. Am J Public Health. 1999;89:1029-
1035.
17. Cole DC, Ibrahim SA, Shannon HS, Scott F, Eyles
J. Work correlates of back problems and activity re-
striction due to musculoskeletal disorders in the Ca-
nadian national population health survey (NPHS)
1994-5 data. Occup Environ Med. 2001;58:728-
734.
18. Fautrel B, Guillemin F. Cost of illness studies in
rheumatic disease. Curr Opin Rheumatol. 2002;14:
121-126.
19. Yelin E. Arthritis: the cumulative impact of a com-
mon chronic condition. Arthritis Rheum. 1992;35:
489-497.
20. Greenberg PE, Finkelstein SN, Berndt ER. Calcu-
lating the workplace cost of chronic disease. Busi-
ness Health. 1995;13:27-30.
21. Yelin E, Callahan LF. The economic cost and so-
cial and psychological impact of musculoskeletal con-
ditions: National Arthritis Data Work Groups. Arthri-
tis Rheum. 1995;38:1351-1362.
LOSS IN PRODUCTIVE TIME DUE TO PAIN
©2003 American Medical Association. All rights reserved. (Reprinted) JAMA, November 12, 2003—Vol 290, No. 18 2453
at Colorado State University, on November 20, 2006 www.jama.comDownloaded from
22. Coyte PC, Asche CV, Croxford R, Chan B. The
economic cost of musculoskeletal disorders in Canada.
Arthritis Care Res. 1998;11:315-325.
23. Merkesdal S, Ruof J, Schoffski O, Bernitt K, Zeidler
H, Mau W. Indirect medical costs in early rheuma-
toid arthritis: composition of and changes in indirect
costs within the first three years of disease. Arthritis
Rheum. 2001;44:528-534.
24. Lee P. The economic impact of musculoskeletal
disorders. Qual Life Res. 1994;3(suppl 1):S85-S91.
25. Yelin EH. Musculoskeletal conditions and em-
ployment. Arthritis Care Res. 1995;8:311-317.
26. Cull RE, Wells NEJ, Miocevich ML. The economic
cost of migraine. Br J Med Econ. 1992;2:103-115.
27. Osterhaus JT, Guterman DL, Piachetka JR. Health-
care resource and lost labour costs of migraine head-
ache in the US. Pharmacoeconomics. 1992;2:67-76.
28. Schwartz BS, Stewart WF, Lipton RB. Lost work-
days and decreased work effectiveness associated with
headache in the work place. J Occup Environ Med.
1997;39:320-327.
29. Lerner D, Reed JI, Massarotti E, Wester LM, Burke
TA. The Work Limitations Questionnaire’s validity and
reliability among patients with osteoarthritis. J Clin Epi-
demiol. 2002;55:197-208.
30. Cockburn IM, Bailit HL, Berndt ER, Finkelstein SN.
When antihistamines go to work. Business Health.
1999;17:49-50.
31. Crystal-Peters J, Crown WH, Goetzel RZ, Schutt
DC. The cost of productivity losses associated with al-
lergic rhinitis. Am J Manag Care. 2000;6:373-378.
32. Greenberg PE, Stiglin LE, Finkelstein SN, Berndt
ER. The economic burden of depression in 1990. J Clin
Psychiatry. 1993;54:405-418.
33. Henke CJ, Levin TR, Henning JM, Potter LP. Work
loss costs due to peptic ulcer disease and gastro-
esophageal reflux disease in a health maintenance or-
ganization. Am J Gastroenterol. 2000;95:788-792.
34. Kessler RC, Almeida DM, Berglund P, Stang P. Pol-
len and mold exposure impairs the work perfor-
mance of employees with allergic rhinitis. Ann Al-
lergy Asthma Immunol. 2001;87:289-295.
35. Stewart WF, Ricci JA, Chee E, Morganstein D. Lost
productive work time costs from health conditions in
the US: results from the American Productivity Audit.
J Occup Environ Med. In press.
36. Ricci JA, Stewart WF, Leotta C, Chee E. A com-
parison of six phone interviews designed to measure
health-related lost productive work time. Value Health.
2001;4:460.
37. Stewart WF, Ricci JA, Leotta C. Health-related lost
productive time: Recall interval and bias in cost esti-
mates. J Occup Environ Med. In press.
38. Karasek R, Gordon G, Pietrokovsky C, Frese M,
Pieper C. Job Content Instrument: Questionnaire and
User’s Guide, Version 1.1. Los Angeles: University of
Southern California; 1986.
39. Bureau of Labor Statistics, Bureau of the Census.
Current Population Survey. Available at: http://www
.bls.census.gov/cps/cpsmain.htm. Accessibility veri-
fied September 30, 2003.
40. Groves RM, Lyberg LE. An overview of non-
response issues in telephone surveys. In: Groves RM,
Biemer PB, Lyberg LE, et al, eds. Telephone Survey
Methodology. London, England: John Wiley & Sons
Ltd; 1988:208.
41. Clark RG, Steel DG. The effect of using household
as a sampling unit. Int Stat Rev. 2002;70:289-314.
42. Stewart WF, Ricci JA, Chee E, Hahn SR, Morgan-
stein D. Cost of lost productive work time among US
workers with depression. JAMA. 2003;289:3135-
3144.
43. Brick JM, Kalton G. Handling missing data in sur-
vey research. Stat Methods Med Res. 1996;5:215-
238.
44. US Office of Personal Management. Federal Ci-
vilian Workforce Statistics: Occupations of Federal
White-Collar and Blue-Collar Workers, as of Septem-
ber 39, 1999. Available at: http://www.opm.gov
/feddata/ocwcbc99.pdf. Accessibility verified Sep-
tember 30, 2003.
45. Stewart WF, Lipton RB, Celentano DD, Reed ML.
Prevalence of migraine headache in the United States:
relation to age, income, race, and other sociodemo-
graphic factors. JAMA. 1992;267:64-69.
46. Gazmararian JA, Baker DW, Williams MV, et al.
Health literacy among Medicare enrollees in a man-
aged care organization. JAMA. 1999;281:545-551.
47. Ad Hoc Committee on Health Literacy for the
Council on Scientific Affairs, American Medical Asso-
ciation. Health literacy: report of the Council on Sci-
entific Affairs. JAMA. 1999;281:552-557.
48. Lipton RB, Hamelsky SW, Kolodner K, Steiner TJ,
Stewart WF. Migraine, quality of life and depression:
a population-based case-control study. Neurology.
2000;55:629-635.
49. Berger ML, Murray JF, Xu J, Pauly M. Alterna-
tive valuations of work loss and productivity. J Oc-
cup Environ Med. 2001;43:18-24.
50. Lipton RB, Diamond S, Diamond M, Reed M,
Stewart WF. Migraine diagnosis and treatment: re-
sults from the American Migraine Study II. Head-
ache. 2001;41:638-645.
51. Stewart W, Lipton R. Need for care and percep-
tions of MIDAS among headache sufferers. CNS Drugs.
2002;16(suppl 1):5-11.
52. Cady RC, Ryan R, Jhingran P, O’Quinn S, Pait DG.
Sumatriptan injection reduces productivity loss dur-
ing a migraine attack: results of a double-blind, placebo-
controlled trial. Arch Intern Med. 1998;158:1013-
1018.
53. Dasbach EJ, Carides GW, Gerth WC, et al. Work
and productivity loss in the rizatriptan multiple attack
study. Cephalalgia. 2000;20:830-834.
Knowledge is not the highest of the intellectual goods.
Of higher value is understanding and, beyond that,
wisdom.
—Mortimer Adler (1902-2001)
LOSS IN PRODUCTIVE TIME DUE TO PAIN
2454 JAMA, November 12, 2003—Vol 290, No. 18 (Reprinted) ©2003 American Medical Association. All rights reserved.
at Colorado State University, on November 20, 2006 www.jama.comDownloaded from
