ArticlePDF Available

The direct and indirect costs of untreated insomnia in adults in the United States (Provisional record)

Authors:

Abstract

To estimate the direct and indirect cost burden of untreated insomnia among younger adults (age 18-64), and to estimate the direct costs of untreated insomnia for elderly patients (age 65 and over). A retrospective, observational study comparing insomnia patients to matched samples without insomnia. Self-insured, employer sponsored health insurance plans in the U.S. 138,820 younger adults and 75,558 elderly patients with insomnia, plus equal-sized, matched comparison groups. NA. Direct costs included inpatient, outpatient, pharmacy, and emergency room costs for all diseases, for six months before an index date. The index date for insomnia patients was the date of diagnosis with or the onset of prescription treatment for insomnia, some-time during July 1, 1999-June 30, 2003. Non-insomnia patients were assigned the same index dates as the insomnia patients to whom they were matched. Indirect costs included costs related to absenteeism from work and the use of short-term disability programs. Propensity score matching was used to find insomnia and non-insomnia patients who had similar demographics, location, health plan type, comorbidities, and drug use patterns. Regression analyses controlled for factors that were different even after matching was completed. We found that average direct and indirect costs for younger adults with insomnia were about 1,253greaterthanforpatientswithoutinsomnia.Amongtheelderly,directcostswereabout1,253 greater than for patients without insomnia. Among the elderly, direct costs were about 1,143 greater for insomnia patients. Insomnia is associated with a significant economic burden for younger and older patients.
SLEEP, Vol. 30, No. 3, 2007 263
INTRODUCTION
INSOMNIA CAN BE AN ACUTE EXPERIENCE OR A
CHRONIC DISORDER, CHARACTERIZED BY DIFFICULTY
IN FALLING ASLEEP AND/OR REMAINING ASLEEP or by
poor quality of sleep.1 The sleep difficulty is associated with day-
time distress, such as tiredness, negative mood, or difficulty with
memory or concentration. The estimated prevalence of chronic
insomnia in the US is about 10% (about 25 million people), but
prevalence varies a great deal across studies.2 For approximately
20%-25% of chronically affected persons, insomnia appears as
a primary disorder.3, 4 For the majority, insomnia occurs in the
presence of medical and psychiatric conditions, such as depres-
sion,5 anxiety, restless leg syndrome, or painful illnesses, although
the nature of the relationship between insomnia and those condi-
tions has not been established.6,7 For that reason a recent National
Institutes of Health State of the Science conference8 concluded
that the term “comorbid insomnia” was preferable to “secondary
insomnia.”
Risk factors for chronic insomnia include female sex and in-
creasing age, although the latter appears due to the increase in
various illnesses with age, rather than age per se.9 Chronic in-
somnia generally lasts for at least several years6,10 and can be as-
sociated with reduced quality of life11, 12 and an increased risk of a
major depressive disorder.13, 14 Other correlates of insomnia may
include fatigue, reduced physical ability, impaired social perfor-
mance, and higher rates of absenteeism from work, accidents
at work, and presenteeism (i.e., lower productivity while at the
workstation).15
Several studies have examined the health care utilization and
cost burden associated with insomnia. Hatoum and colleagues16
reviewed the experience of 5 American Medical Group Associa-
tion clinics and found that insomnia patients had more emergency
room visits, more calls to the doctor, and more use of over-the-
counter drugs than noninsomnia patients. Health-related quality
of life was also lower for insomnia patients. Similarly, Simon and
VonKorff17 surveyed 1,962 patients at primary health clinics and
conducted face-to-face interviews with a stratified random sub-
sample (n = 373), in order to estimate the prevalence and cost-
burden of chronic insomnia. They found the prevalence of chronic
insomnia was 10%, and they found that chronic insomnia patients
had higher health care costs and significantly greater physical and
social disability than good sleepers. Leger et al.18 also found high-
er rates of absenteeism, more trouble concentrating at work, and
more medical problems (resulting in more physician office visits)
among insomnia patients, compared with good sleepers.
Other studies have attempted to estimate the cost burden of
insomnia from a public health perspective. While methods have
The Direct and Indirect Costs of Untreated Insomnia in Adults in the United States
Ronald J. Ozminkowski, PhD1; Shaohung Wang, PhD2; James K. Walsh, PhD3
1Institute for Health and Productivity Studies, Cornell University, Washington, D.C., Health and Productivity Research, Thomson Medstat, Ann Arbor
MI; 2Thomson Medstat, Cambridge, MA; 3Sleep Medicine and Research Center, Chesterfield, MO
Cost Burden of Untreated Insomnia—Ozminkowski et al
Disclosure Statement
This research was funded by Sepracor, Inc. Dr. Ozminkowski is employed by
Thomson Medstat. Sepracor, Inc. provided the funding directly to Thomson
Medstat for Dr. Ozminkowski’s time in writing this paper. Dr. Ozminkowski
received no direct funds from Sepracor, Inc. Dr. Walsh reported that research
support has been provided to his instution by Evotec, Pfizer, Merck & Co.,
Neurocrine Biosciences, and Cephalon. Dr. Walsh has been a consultant
for Abbott Laboratories, Pfizer, Sanofi-Aventis, Cephalon, Organon, Neuro-
crine Biosciences, Takeda Pharmaceuticals North America, Actelion, Sepra-
cor, Elan, Guilford Pharmaceuticals, Respironics, Merck KGaA, Darmstadt,
King Pharmaceuticals, TransOral Pharmaceuticals, Neurogen Corporation,
GlaxoSmithKline, SleepTech, Eli Lilly and Company, Evotec, and Merck &
Co. Dr. Wang has reported no financial conflicts of interest.
Submitted for publication April 13, 2006
Accepted for publication November 27, 2006
Address correspondence to:Ronald J. Ozminkowski, PhD, Director, Health
and Productivity Research, Thomson Medstat, 777 East Eisenhower Park-
way, 804B, Ann Arbor, Michigan 48108, Tel: (734) 913-3255; Fax: (734) 913-
3338; E-mail: Ron.Ozminkowski@Thomson.com
INSOMNIA
Objectives: To estimate the direct and indirect cost burden of untreated
insomnia among younger adults (age 18 – 64), and to estimate the direct
costs of untreated insomnia for elderly patients (age 65 and over).
Design: A retrospective, observational study comparing insomnia patients
to matched samples without insomnia.
Settings: Self-insured, employer sponsored health insurance plans in the
U.S.
Patients or Participants: 138,820 younger adults and 75,558 elderly pa-
tients with insomnia, plus equal-sized, matched comparison groups.
Interventions: NA
Measurements and Results: Direct costs included inpatient, outpatient,
pharmacy, and emergency room costs for all diseases, for six months be-
fore an index date. The index date for insomnia patients was the date of
diagnosis with or the onset of prescription treatment for insomnia, some-
time during July 1, 1999 – June 30, 2003. Non-insomnia patients were as-
signed the same index dates as the insomnia patients to whom they were
matched. Indirect costs included costs related to absenteeism from work
and the use of short-term disability programs. Propensity score matching
was used to find insomnia and non-insomnia patients who had similar de-
mographics, location, health plan type, comorbidities, and drug use pat-
terns. Regression analyses controlled for factors that were different even
after matching was completed. We found that average direct and indirect
costs for younger adults with insomnia were about $1,253 greater than
for patients without insomnia. Among the elderly, direct costs were about
$1,143 greater for insomnia patients.
Conclusions: Insomnia is associated with a significant economic burden
for younger and older patients.
Keywords: Insomnia, Cost, Burden of illness
Citation: Ozminkowski RJ; Wang S; Walsh JK. The direct and indi-
rect costs of untreated insomnia in adults in the united states. SLEEP
2007;30(3):263-273.
SLEEP, Vol. 30, No. 3, 2007 264
varied, the typical approach has been to estimate the costs of pre-
scription and nonprescription insomnia medications; the costs of
accidents and other work mishaps that are due to insomnia; the
costs of using alcohol to manage sleep problems; and the added
inpatient, outpatient, emergency room, and nursing home costs
associated with insomnia. Reviews of earlier studies by Walsh
and Engelhart19 and Chilcott and Shapiro20 suggest a total cost
of insomnia ranging from about $30 billion to $35 billion in the
United States, in the early to mid-1990s.
To date, most of the studies of the burden of insomnia have
been based on relatively small samples of patients at a small num-
ber of treatment sites. While there are some advantages of using
survey methods in those studies (e.g., one can address a wide ar-
ray of indirect costs via survey), self-reports may be prone to reli-
ability and validity concerns.
With a different set of data sources at hand, we took a different
approach to estimating the insomnia cost burden. We used medi-
cal claims data to investigate direct costs; employer absenteeism
and short term disability program records were used to estimate
indirect costs. We also focused on the costs of untreated insom-
nia.
METHODS
Study Design
Retrospective, observational studies were conducted using data
from July 1, 1999 to June 30, 2003. Direct costs were investi-
gated, using information from medical claims for inpatient, out-
patient, pharmacy, and emergency room services. Expenditures
for these services were transformed to year-2003 metrics, to ad-
just for inflation. Expenditures were then compared for 138,820
patients aged 18 – 64 years who developed insomnia, and for an
equal-sized, matched sample of patients who did not. Expendi-
tures were also compared for 75,558 elderly insomnia patients,
versus an equal-sized, matched sample of elderly patients who did
not have insomnia. To estimate indirect costs (also in year-2003
cost values), absenteeism records and short-term disability pro-
gram records were examined for matched workers who did and
did not develop insomnia. Propensity score methods were used to
conduct the matching processes, as described below. The propen-
sity score analyses were then supplemented by multiple regres-
sion analyses, to control for differences that remained after the
matching was completed.
Overview of Analytic Strategy
Our goal was to estimate the average dollar impact of untreated
insomnia on total medical expenditures, absenteeism from work,
and the use of short-term disability program services. With regard
to medical expenditures, the following general equation summa-
rizes the way we estimated cost burden:
(1) Average dollar impact of untreated insomnia on medical expendi-
tures = (Average health care expenditures for sample members who
were diagnosed with, or treated for, insomnia) – (Average health
care expenditures for matched sample members who were not diag-
nosed with, or treated for, insomnia).
For those who were diagnosed with or treated for insomnia,
average medical expenditure was calculated for 6 months before
the diagnosis of insomnia or beginning treatment for it. For those
who did not develop insomnia, a matching calendar period was
used, as noted below.
Similar equations were used to summarize the average dollar
impact of untreated insomnia on absenteeism-related costs and
the costs of short-term disability program use. Therefore:
(2) Average total dollar impact of untreated insomnia = Average impact
on medical expenditures + Average impact on absenteeism costs +
Average impact on short-term disability costs.
Prior to estimating the figures needed for equations (1) and (2),
the following steps were completed to enhance the accuracy of
the analyses.
First, those eventually diagnosed with or treated for insomnia
were statistically matched to those who were not, using propen-
sity score analyses. The propensity score analyses matched the
eventual insomnia patients to the most similar subset of those
who were never diagnosed with or treated for insomnia, based on
their demographics and casemix.
Second, since no matching process can ever be perfect, we
compared demographics and casemix measures after the match-
ing was completed. Two-sided t-tests that were adjusted for dif-
ferences in variances were used to learn whether averages for
continuous measures of demographics or casemix were different.
T-tests for differences in proportions were used to learn whether
there were significant differences in categorical measures, such
as the existence of particular diagnoses or the use of pharmaceu-
ticals of interest. P-values <0.05 were considered statistically sig-
nificant.
Third, multiple regression analyses were used to estimate the
relationship between the eventual diagnosis of or treatment for
insomnia and medical expenditures. These regressions controlled
for any significant demographic or casemix factors that were
found in the second step above.
Fourth, multiple regression analyses were used to estimate the
relationship between eventual diagnosis of or treatment for in-
somnia and the dollar value of lost work time, for the subsets
of sample members who were employed and for whom absen-
teeism or short-term disability program use could be observed.
Separate analyses of absenteeism and short-term disability were
conducted.
Fifth, the results of the regression analyses were input into
equations (1) and (2) above, to estimate the cost burden of un-
treated insomnia. Thus, our cost burden estimates accounted for
measurable differences in demographics and casemix, increasing
the likelihood that any dollar differences between the 2 groups of
patients would be due to untreated insomnia.
Details of our analytic strategy are described below, after not-
ing data contributors and sources and study inclusion criteria.
Data Contributors and Data Sources
Three data sources were used for this study. These include The
Medstat Group MarketScan Commercial Claims and Encounters
Database, the MarketScan Medicare Supplemental and Coordina-
tion of Benefits Database, and the MarketScan Health and Produc-
tivity Management Database. Each is described briefly below.
For younger adults (those age 18 64), direct costs were es-
timated with data from The Medstat Group’s MarketScan Com-
mercial Claims and Encounters (CCAE) Database for 1999 -
2003. Over this period, the CCAE database included data on over
3.2 million enrollees; this accounts for about 3% of all privately
insured lives in the United States.
The enrollees whose data were included in the CCAE files
were those whose employers self-insured for medical care ser-
Cost Burden of Untreated Insomnia—Ozminkowski et al
SLEEP, Vol. 30, No. 3, 2007 265
vices. There were 79 such large employers (those with more than
200 employees in the CCAE database for 1999 2003). These
employers contracted with 150 insurance plans to arrange for
health care services. These plans included commercial insurance
companies, Blue Cross and Blue Shield plans, and third-party ad-
ministrators that have fee-for-service, partially and fully capitated
arrangements. (Data from capitated health plans were not used for
this study, because those data lack service-level payment informa-
tion.)
While the 79 employers did use the services of insurance com-
panies, it is important to note that the employers were “self-in-
sured”; they (and their employees) paid all of the medical costs.
They did not pay premiums to the insurance companies.
The 79 employers who contributed to the Commercial Claims
and Encounters Database are a convenience sample of employ-
ers who do business with The Medstat Group, Inc. of Ann Arbor,
Michigan. The CCAE Database has been the source of informa-
tion used in over 160 peer-reviewed studies published since 1997
(a complete list of publications is available upon request; inqui-
ries can be made at www.medstat.com).
The CCAE database includes information on patient demo-
graphics, diagnosis codes, procedure codes, type and place of
service, service dates, payment information, and other metrics.
Financial, clinical, and demographic variables are standardized to
common definitions.
Direct costs for elderly insomnia patients (≥65 years) were
obtained from The Medstat MarketScan Medicare Supplemen-
tal and Coordination of Benefits Database, for 1999 - 2003. This
Medicare database includes the same types of financial, clinical,
and demographic variables as in the CCAE database, for 545,645
elderly plan members whose medical care was covered in part
by the federal government under the Medicare program, and in
part by their former employers or their spouses’ former employ-
ers. Contributors to the MarketScan Medicare Database include
large employers who supplement Medicare by offering prescrip-
tion drug and other services to retirees. In 2001, about 34% of
Medicare beneficiaries had prescription drug coverage through
their current or former employers.21 By examining the experience
of employees with these benefits, we can learn about the possible
impact of insomnia on their health expenditures.
Indirect costs were obtained from Medstat’s MarketScan
Health and Productivity Management Database. This database
contains employee absence and short-term disability data from 9
large employers in the United States. Since not all employers who
contribute to the MarketScan Commercial Claims and Encoun-
ters Database submit productivity data to Medstat, the Health and
Productivity Management Database is a nonrandom subset of the
CCAE Database. The Health and Productivity Management Da-
tabase contains short-term disability data for more than 300,000
employees. Absence data were available for more than 200,000
employees in 2001; 160,000 employees in 2000; 112,000 employ-
ees in 1999; and 15,000 employees in 1998 and 1997.
Study Sample and Inclusion Criteria
For the studies of direct costs, sample selection began by
searching the CCAE and Medicare databases for all patients who
had one or more claims with a diagnosis of insomnia (ICD-9-
CM diagnosis code = 307.41, 307.42, 780.52), or one or more
claims for an insomnia medication between July 1, 1999 and June
30, 2003. The first observed date when evidence of insomnia was
found in the period is called the sample member’s “index date.”
Insomnia medications and associated doses are provided in
Table 1. Because the doses noted in the table for mirtazepine are
sometimes used to treat depression, patients who took that drug
were excluded from the analysis if they also had a diagnosis of de-
pression or used other drugs to treat depression. All other patients
were retained for analysis if they had a diagnosis of insomnia or
took one of the medications in the dose ranges noted in Table 1
between July 1, 1999–June 30, 2003. All such patients who could
be tracked for at least 6 months before that index date and who
had no evidence of insomnia treatment in the 6-month period be-
fore that date were retained for initial analysis (n = 184,879 for
the younger adult sample and n = 84,015 for the elderly sample).
The initial samples also included 1,051,787 younger adults
and 457,701 elderly patients who had no evidence of insomnia in
1999–2003. These people were used to find subsets of noninsom-
nia patients who could be matched to insomnia patients using the
propensity score analyses described below.
The samples of patients used for the indirect cost (i.e., absen-
teeism and short-term disability) analyses were subsets of the
younger adult samples obtained from the Commercial Claims and
Encounters Database. These included active employees for whom
absenteeism or short-term disability data were available. Absen-
teeism and short-term disability information was obtained from
the MarketScan Health and Productivity Management Database.
Patients were excluded from the indirect cost analysis if they
did not have at least 6 months of eligibility for absenteeism or
short-term disability benefits prior to or after the insomnia index
date. After matching, there were 3,033 employees found to have
insomnia and absenteeism data (i.e., about 1.6% of all insomnia
patients), and 5,028 employees who had insomnia and short-term
disability data (2.7% of all insomnia patients).
Variables Used in the Propensity Score Matching Process
Once the initial samples of insomnia and noninsomnia pa-
tients were selected, the variables needed for the propensity score
matching process were created. The dependent variable for this
process was a binary indicator for having a diagnosis of insomnia
or being treated for insomnia with prescription pharmaceuticals
in the study period (coded as 1 if yes, and 0 if no).
The objective of the propensity score matching process was
to generate samples of patients with and without insomnia who
were comparable to each other, before estimating the cost burden
of untreated insomnia. Comparability of the patient samples was
assessed by considering factors related to the likelihood of hav-
ing insomnia (e.g., demographics, comorbidities, and prescrip-
tion medication use patterns) and other factors that may influence
direct or indirect costs for insomnia in similar patients (e.g., index
year and health plan type). These factors are noted below.
The demographic factors included in the analyses were patient
age and sex. As noted earlier, insomnia is more prevalent among
females, and its prevalence tends to increase with age.
Comorbidities were measured for the first 6 months that pa-
tients were observed during the 1999–2003 study period. For
insomnia patients, this was prior to their insomnia index date.
Comorbidities were measured in terms of severity, number, and
type. To control for differences in severity of the comorbidities,
we used the Charlson Comorbidity Index, which estimated the
Cost Burden of Untreated Insomnia—Ozminkowski et al
SLEEP, Vol. 30, No. 3, 2007 266
likelihood of death or serious disability in the coming year, on
the basis of diagnosis codes for up to 18 different diseases that
were observed in the data. Values of the index may range from
zero to 28, with the number of points for each disease depending
on its prognosis for death or major disability. Higher values are
associated with higher probabilities of these outcomes. Charlson
Comorbidity Index values below 2.0 suggest low odds of death or
major disability for most patients; values between 2 and 6 suggest
moderate risk, and values above 6.0 indicate high risk.22 (A recent
study of the predictive ability of the index with regard to health
care expenditures is described in a paper by Farley et al.23)
To control for the number of comorbidities, we included the
number of unique ICD-9-CM diagnosis codes that each person
had (at the 3-digit level), and the number of psychiatric diagnostic
groups that any mental health problems fell into. Psychiatric diag-
nostic groupings were developed by Ashcraft et al. as an efficient
way of accounting for the types of mental health problems that
people may have.24
To control for specific types of comorbidities, we used the
MarketScan data to find the most prevalent comorbidities among
insomnia patients. Since there were hundreds of possible comor-
bidities, we arbitrarily focused on the ones that were most costly
or prevalent. The comorbidities were characterized in terms of
whether they were primary diagnoses (coded first on a claim), or
secondary. The most expensive primary diagnoses (those 11 or 12
(depending on age group) that accounted for more than half of the
total expenditures of the sample) were included in the analyses,
and the most prevalent secondary diagnoses (those that were most
often listed in claims for sample members) were also included.
The detailed list of comorbidities used in the matching processes
is available upon request. Examples include angina, diabetes, low
back problems, severe osteoarthritis, hypertension, various forms
of cancer, and other back or joint problems.
The analyses also controlled for the types of medications that
sample members used. These measures included binary indica-
tors to account for most of the pharmacy expenditures incurred
by insomnia patients. These were drugs for all diseases except
insomnia, and were measured for the first 6 months when patients
were observed in the study period. This was done to control for
the cost-impact of medications used to treat comorbidies. Con-
trolling for drug use also helped find patients with conditions that
may not be recorded with diagnosis codes on medical claims be-
cause of stigma or other reasons, and to find patients with chronic
conditions that may have been diagnosed >6 months prior to the
index date. For example, depression (a common comorbidity with
insomnia)25 was not among the most costly or common comor-
bidities when diagnosis codes were reviewed, but antidepressant
drugs were the most costly drugs taken by these patients, so many
patients with depression were found in the search for medications.
The detailed list of pharmaceuticals that accounted for most of
the drug expenditures among sample members is available upon
request; examples are antidepressants, gastrointestinal drugs, an-
tihyperlipidemic drugs, analgesics/antipyretics, and nonsteroidal
anti-inflammatories, opiate agonists, and medications for heart
disease
When controlling for location, the objective was to balance the
samples in terms of where patients lived, using indicators for US
geographic census region of residence and urban versus rural lo-
cation. The 4 US census regions included the Northeastern, North
Central, Southern, and Western regions of the country. Urban (vs
rural) location was measured on the basis of residence in a Metro-
politan Statistical Area, as designated by the US Census. In gen-
eral, health care expenditures tend to be higher in urban areas26
and in northern and eastern census regions.27
Next, it is unknown whether insomnia prevalence differs by
plan type, but one may surmise that the availability of sleep spe-
cialists and pharmacotherapy choices are associated with plan
type, and plan type is well known to influence health care utiliza-
tion.28 We controlled for plan type by using indicators for mem-
bership in indemnity plans, preferred provider organizations, or
point of service plans.
The index year is the first year the patient was observed to have
insomnia (for insomnia patients) or the first year he or she was
observed in the data base (for noninsomnia patients) during the
study period. Index years ranged from 1999–2003. Medical ex-
penditures generally increase over time for all patients. Thus, the
objective here was to account for differences in the distributions
of insomnia and noninsomnia patients according to index year.
All of the above variables were observable for all patients;
missing data were not problematic.
Conducting the Propensity Score Matching Process
The conventional application of propensity score analysis is to
use important variables to balance 2 samples of interest29 (in our
case, patients with and without eventual diagnosis for or treat-
ment of insomnia). This is typically done via logistic regression
analyses designed to predict the probability that each observa-
tion belongs to one of the two types of samples. For example, we
know that only 184,879 of our younger adult sample members
were diagnosed with insomnia or treated for it, but all sample
members (even the 1,051,787 patients without evidence of in-
somnia) had an underlying probability of having insomnia. If
we have reason to believe these underlying probabilities depend
on the demographic, case mix, location, and other factors noted
above, then we can estimate the underlying probability of having
insomnia for each sample member, using information about these
variables. Matching insomnia and noninsomnia patients on these
probabilities helps minimize their differences on these variables.
All of the variables listed above were used as predictors of in-
somnia in logistic regression analyses. (All of the variables were
entered into the regression at the same time; no stepwise proce-
dures were used.) These analyses yielded a predicted probability
that each patient would eventually be diagnosed with or treated
for insomnia. By matching on these predicted probabilities (and
thereby excluding any insomnia or noninsomnia patients who
could not be matched), many of the differences between insom-
nia and noninsomnia patients were minimized. This yielded sets
of insomnia and noninsomnia sample members who were com-
parable.
Once the matches between patients with and without insomnia
were made, each patient with insomnia was assigned an index
date. The index date was defined as the calendar date of the first
inpatient or outpatient medical claim showing a diagnosis of in-
somnia, or the date of the first prescription for an insomnia medi-
cation, during the study period. Each patient without insomnia
was assigned the same index date as the one associated with the
insomnia patient to whom he or she was matched. To estimate
the costs of untreated insomnia, each patient was then tracked
for 6 months before his or her index date. This assured that in-
Cost Burden of Untreated Insomnia—Ozminkowski et al
SLEEP, Vol. 30, No. 3, 2007 267
somnia and noninsomnia patients were followed for exactly the
same calendar periods when direct and indirect costs of untreated
insomnia were estimated.
Outcome Variables and Statistical Analyses
If the propensity score matching process had been perfect, one
could simply compare the costs of patients with and without even-
tual diagnosis of or treatment for insomnia, using t-tests. How-
ever, no matching process is ever perfect.30 Our propensity score
matching process worked reasonably well, but many variables
were still significantly different between samples after matching
was conducted (details are available upon request. Even though
the statistical power associated with large sample sizes may have
been the cause of many of these significant differences, we used
multiple regression analyses to adjust for the cost-impact of vari-
ables that remained significantly different for insomnia versus
noninsomnia patients after the matching process was completed.
For the analyses of direct medical expenditures, exponential
conditional regression models were used to control for these re-
maining differences. (More information about exponential condi-
tional regression models can be found in Mullahy.31) The results
obtained from the regressions produced more accurate estimates
of the average direct medical costs of untreated insomnia.
For the analyses of indirect (i.e., absenteeism and short-term
disability) costs, 2-part regression models were used to control for
remaining differences. Separate 2-part modeling processes were
used for absenteeism and short-term disability, because sample
sizes differed for these metrics, due to the differences in data
availability from the Health and Productivity Management data
contributors.
A 2-part regression process was used to study absenteeism and
short-term disability benefit costs because not all employees used
these benefits. Two-part statistical models have been designed for
situations like this, where there are large percentages of non-us-
ers of a benefit.31 The first step of each 2-part model included a
logistic regression designed to estimate the impact of untreated
insomnia on the probability of using any absenteeism (or disabil-
ity) benefits. The second step of each 2-part model included an ex-
ponential regression designed to estimate the impact of untreated
insomnia on the magnitude of absenteeism (or short-term disabil-
ity) costs, if any such benefits were used. Each part of each regres-
sion model controlled for those factors that remained significantly
different between eventual insomnia and noninsomnia patients,
after the matching process was completed. Thus, the results of the
2-part modeling processes yielded more accurate estimates of the
impact of untreated insomnia on indirect (absenteeism and short-
term disability benefit) costs.
Sensitivity Analyses
To provide some context for interpreting the results of our main
analyses described above, we also conducted some sensitivity
analyses. Two types of sensitivity analyses were conducted. The
first sensitivity analysis involved removing from the sample those
observations whose medical care expenditures were abnormally
low or high during the 6-month observation period. This was
done to assure that the range of medical expenditures would be
the same, for eventual insomnia and noninsomnia patients. Such
leveling has been suggested by Heckman et al.32 when propensity
score analyses are used, to see whether a small number of outlier
observations would have a large impact on the results.
The second type of sensitivity analysis was conducted with the
entire sample (including outliers), to address the arbitrary choice
of using a 6-month period for our main analyses. The obvious as-
sumption here is that patients had insomnia for at least 6 months
prior to its diagnosis or the onset of pharmaceutical therapy to
treat it. This seems reasonable, since most insomnia patients re-
port sleeping problems for more than one year.6,10 Nevertheless,
we conducted sensitivity analyses to see how cost burden would
vary if other lengths of time were considered, ranging from 1 to 5
months prior to the diagnosis of or treatment for insomnia.
RESULTS
Insomnia Medication Use
Table 1 lists the percentages of patients who used each medica-
tion for insomnia in our analyses. For younger adults and elderly
patients, zolpidem tartrate was used most often, by roughly 39%
of patients. Amitriptyline was used by about 21% of both sam-
ples, and temazepam was used by about 6% of the younger adults
and 13% of the elderly. Trazodone was used by about 10% of the
elderly and 12% of the younger patients. Other drugs were used
less frequently. These percentages pertain to the first insomnia
medication observed. About 8% of the younger insomnia patients
and 4% of the elderly patients were diagnosed with insomnia but
used no prescription therapy involving any of the study drugs.
Matching Process
To save space, we do not report the detailed results from the
logistic regression analysis that was used to match younger adults
eventually diagnosed with or treated for insomnia with those who
were not. Similarly, we do not report the detailed results for the
Table 1—Percent of sample members prescribed medications be-
lieved to be used to treat insomnia, and percent of sample members
diagnosed with insomnia but not prescribed a study medication to
treat it
Age 18–64 years Age ≥65 years
Drug Type and Dosage Number Percent Number Percent
Patient had insomnia diagnosis 11,280 8.13% 3,077 4.07%
but no study drug
Zolpidem tartrate -- 5 mg 55,179 39.75% 29,273 38.74%
- 20 mg
Zaleplon -- 5 mg - 20 mg 8,670 6.25% 3,697 4.89%
Temazepam/Temaz/Razepam 8,940 6.44% 9,461 12.52%
-- 15 mg - 30 mg
Trazodone -- < 150 mg/day 16,011 11.53% 7,214 9.55%
Triazolam -- 0.125 mg - 0.5 mg 4,288 3.09% 1,511 2.00%
Flurazepam -- 15 mg - 30 mg 1,985 1.43% 1,221 1.62%
Estazolam -- 1 mg or 2 mg 503 0.36% 438 0.58%
Quazepam -- 7.5 mg - 15 mg 48 0.03% 59 0.08%
Amitriptyline -- 10 - 100 mg 29,542 21.28% 15,562 20.60%
Mirtazepine -- 15 - 30 mg 2,374 1.71% 4,045 5.35%
Total: 138,820 100.00% 75,558 100.00%
Mirtazepine patients could not have a depression diagnosis or antide-
pressant study period.
Sources: MarketScan® Commercial Claims and Encounter Database,
and Medicare Supplemental Insurance and Coordination of Benefits
Database, 1999–2003
Cost Burden of Untreated Insomnia—Ozminkowski et al
SLEEP, Vol. 30, No. 3, 2007 268
matching processes used for the elderly sample. All of these re-
sults are available upon request. In both cases, nearly every de-
mographic, location, plan type, index year, comorbidity, and drug
use measure had a statistically significant impact on the odds of
having insomnia. Among the younger sample members, the odds
of having insomnia were significantly higher for females, and
increased by about 0.08% per year of age. The odds of having
insomnia were also higher for those in Northeast, North Central,
and Western US census regions (compared to those living in the
South). The odds of having insomnia were lower for those in ur-
ban areas, and for those in point of service and preferred provider
organization health plans. The odds of having insomnia also var-
ied by the year of entry into the study, but this is an anomaly
related to the fact that data contributors varied by year.
In general, higher Charlson Comorbidity Index values and
higher numbers of physical or mental health problems were asso-
ciated with a higher likelihood of having insomnia. Finally, most
but not all of the comorbidity and pharmacy-use measures were
associated with higher odds of having insomnia.
Among Medicare beneficiaries, increasing age was associated
with a lower probability of insomnia, but the probability of in-
somnia declined only about 1% per year after age 65. The impact
of census region and urban location was about the same in this
sample as in younger sample members, but members of preferred
provider organizations were more likely than traditional fee-for-
service members to have insomnia. In contrast to the younger
adult sample, Medicare beneficiaries were more likely to have
insomnia if they entered the study in earlier years.
As with the younger sample, most of the comorbidity and drug
use variables influenced the likelihood of having insomnia among
Medicare beneficiaries. The lists of most costly or most prevalent
comorbidity measures differed somewhat though, as one would
expect.
Sample Characteristics After Matching
Table 2 shows the demographic, location, plan type, index
year, and some of the clinical metrics which describe the younger
adult and elderly samples after matching. Many of these means
and percentages were significantly different between those even-
tually diagnosed with or treated for insomnia and those who were
not. However, statistical significance is due primarily to the large
sample sizes used in the analyses. A close inspection indicates
very little difference in the magnitude of the characteristics mea-
sured in insomnia and noninsomnia patients. Thus, the matching
seems to have worked well.
Regression Results
Tables 3 - 6 present the results of the regression analyses. Ta-
bles 3 and 4 focus on analyses of direct medical costs (inpatient,
outpatient, emergency room, and pharmacy expenditures) during
the 6 months prior to the index date. For younger adults and elder-
ly patients alike, these 2 tables show that eventual insomnia pa-
tients had significantly higher expenditures (P < 0.001 for both),
after matching was completed and regression analyses were used
to control for differences in age, location, plan type, index year,
Table 2—Demographic and clinical characteristics, after matching, for those eventually diagnosed with or treated for insomnia, and those who
were not, by age
Age 18–64 years Age ≥65 years
Parameter Eventual Insomnia Control P-value Eventual Control P-value*
Patients ( n = 138,820 ) Insomnia Patients ( n = 75,558 )
( n = 138,820 ) ( n = 75,558 )
N/Mean ( %/S.D.) N/Mean ( %/S.D.) N/Mean ( %/S.D.) N/Mean ( %/S.D.)
Mean age 47.07 (11.13) 47.58 (12.27) 0.00 75.19 ( 7.01) 75.31 ( 6.86) 0.00
Female 85350 (61.48%) 85067 (61.28%) 0.27 46163 (61.10%) 46447 (61.47%) 0.13
Region
Northeast 11432 ( 8.24%) 12853 ( 9.26%) 0.00 10426 (13.80%) 10470 (13.86%) 0.74
North Central 44623 (32.14%) 46073 (33.19%) 0.00 29476 (39.01%) 29254 (38.72%) 0.24
South 66247 (47.72%) 63436 (45.70%) 0.00 27251 (36.07%) 27302 (36.13%) 0.78
West 16518 (11.90%) 16458 (11.86%) 0.72 8405 (11.12%) 8532 (11.29%) 0.30
Resided in urban area 99861 (71.94%) 99652 (71.79%) 0.38 58279 (77.13%) 58022 (76.79%) 0.12
Insurance Plan Types
Indemnity 42688 (30.75%) 42117 (30.34%) 0.02 53231 (70.45%) 52911 (70.03%) 0.07
Point of service 32271 (23.25%) 34381 (24.77%) 0.00 1499 ( 1.98%) 1503 ( 1.99%) 0.94
Preferred Provider Organization 63861 (46.00%) 62322 (44.89%) 0.00 20828 (27.57%) 21144 (27.98%) 0.07
Index Year
1999 10673 (7.69%) 11146 (8.03%) 0.00 7819 (10.35%) 7841 (10.38%) 0.85
2000 22024 (15.87%) 21974 (15.83%) 0.79 13816 (18.29%) 13766 (18.22%) 0.74
2001 33913 (24.43%) 33800 (24.35%) 0.62 22456 (29.72%) 22071 (29.21%) 0.03
2002 43502 (31.34%) 43089 (31.04%) 0.09 22281 (29.49%) 22394 (29.64%) 0.52
2003 28708 (20.68%) 28811 (20.75%) 0.63 9186 (12.16%) 9486 (12.55%) 0.02
Baseline Clinical Characterisitcs
Charlson Comorbidity Index 0.34 ( 0.89) 0.35 ( 0.88) 0.00 1.09 ( 1.61) 1.07 ( 1.56) 0.06
Number of psychiatric diagnosis groups 0.12 ( 0.38) 0.10 ( 0.34) 0.00 0.08 ( 0.31) 0.06 ( 0.28) 0.00
Number of unique 3-digit ICD-9 codes 3.84 ( 3.17) 3.95 ( 3.08) 0.00 6.24 ( 4.56) 6.18 ( 4.46) 0.01
*Two-sided t-tests of differences between insomnia cohort and matched control were used.
Sources: 1999-2003 MarketScan® Commercial Claims and Encounter and Medicare Supplemental Insurance and Coordination of Benefits Data-
bases
Cost Burden of Untreated Insomnia—Ozminkowski et al
SLEEP, Vol. 30, No. 3, 2007 269
and comorbidity patterns.
Table 3 also presents estimates of average medical expendi-
tures that were obtained from the regression, first for 138,820
younger adult patients eventually diagnosed with or treated for
insomnia and the 138,820 noninsomnia patients to whom they
were matched. (The process of estimating average expenditures
from exponential regression results has been described in detail
by Mullahy31). After the matching process was completed, and
after further controls for the variables just mentioned were ap-
plied via the exponential regression, younger adult patients even-
tually diagnosed with or treated for insomnia were found to incur
an average of $4,755 in medical expenditures, while those never
diagnosed with or treated for insomnia had average medical ex-
penses of $3,831 (2003 dollars). The $924 difference in average
direct medical expenditures is our estimate of the direct medical
costs of untreated insomnia, for patients who were under age 65
(P < 0.001).
A similar pattern was found for elderly patients, but the direct
costs were much higher for both elderly groups, as one would
expect. Also, the difference in direct costs between eventual in-
somnia and noninsomnia patients was larger. Table 4 shows that
Table 3—Regression Analyses of Direct Costs, Age 18–64
138,820 patients eventually diagnosed with or treated for insomnia;
138,820 matched controls.
Independent Variable Parameter Standard Wald
Estimate Error P-value
Intercept 6.35 0.01 0.00
Eventual Insomnia Patient 0.22 0.00 0.00
Age 0.01 0.00 0.00
Northeast 0.06 0.01 0.00
North Central 0.03 0.01 0.00
South 0.08 0.01 0.00
Point of Service Plan Type -0.09 0.01 0.00
Preferred Provider Organization -0.02 0.01 0.00
Plan Type
Index Year = 1999 -0.12 0.01 0.00
Baseline Clinical Characteristics
Charlson Comorbidity Index 0.46 0.00 0.00
Number of psychiatric diagnosis groups 0.30 0.01 0.00
Comorbidities and Drug Use Measures
Angina pectoris, chronic maintenance 0.80 0.02 0.00
Diabetes Mellitus, chronic maintenance -0.44 0.01 0.00
Mechanical low back disorder 0.31 0.01 0.00
Renal failure 1.05 0.05 0.00
Preventive health encounters 0.21 0.01 0.00
Essential hypertension, chronic 0.15 0.01 0.00
maintenance
Disease of ears, nose or throat or 0.11 0.01 0.00
mastoid process, not elsewhere
classified
Symptoms involving respiratory 1.22 0.02 0.00
system and other chest symptoms
General symptoms 0.90 0.02 0.00
Disorders of lipoid metabolism 0.17 0.02 0.00
Other and unspecified disorders of back 0.70 0.02 0.00
Special investigations and examinations 0.64 0.02 0.00
Special screening for malignant 0.02 0.02 0.34
neoplasms
Psychother, antidepressants 0.12 0.01 0.00
Antihyperlipidemic drugs, not 0.24 0.01 0.00
elsewhere classified
Analgesics/antipyritics, nonsteroidals/ 0.18 0.01 0.00
anti-inflammatories
Unclassified agents, not elsewhere 0.28 0.01 0.00
classified
Analgesics/antipyritics, opiate agonists 0.94 0.01 0.00
Antihistamines & combinations, not 0.15 0.01 0.00
elsewhere classified
Predicted 6-Month Expenditures
Patients eventually diagnosed with or $4,755
treated for insomnia
Matched comparison group $3,831
Difference $924 (P < 0.01)*
*P-value comes from Wald chi-squared test of regression coefficient
for Eventual Insomnia Patient variable
Source: MarketScan© Research Databases: 1999-2003.
Table 4—Regression analyses of direct costs, Age ≥65 years
75,558 patients eventually diagnosed with or treated for insomnia;
75,558 matched controls.
Independent Variable Parameter Standard Wald
Estimate Error P-value
Intercept 7.70 0.03 0.00
Eventual insomnia patient 0.22 0.01 0.00
Age 0.00 0.00 0.16
Index Year = 2001 -0.09 0.01 0.00
Index Year = 2003 0.14 0.01 0.00
Number of psychiatric diagnosis groups 0.28 0.01 0.00
Comorbidities and Drug Use Measures
Diabetes mellitus, chronic maintenance 0.30 0.01 0.00
Renal failure 1.14 0.02 0.00
Essential hypertension, chronic 0.09 0.01 0.00
maintenance
Cancer of lungs, bronchi, or mediastinum 1.06 0.03 0.00
Cataract 0.21 0.01 0.00
Diseases and disorders of skin & 0.01 0.01 0.09
subcutaneous tissues, not elsewhere
classified
Symptoms involving respiratory system 0.82 0.01 0.00
and other chest symptoms
Essential hypertension 0.65 0.01 0.00
Other forms of chronic ischemic heart 0.88 0.01 0.00
disease
Cardiac dysrhythmias 0.81 0.02 0.00
General symptoms 0.66 0.02 0.00
Gastrointestinal drug miscellaneous, 0.29 0.01 0.00
not elsewhere classified
Antihyperlipidemic drugs, not elsewhere 0.09 0.01 0.00
classified
Analgesics/antipyretics, nonsteroidals/ 0.11 0.01 0.00
anti-inflamitories
Unclassified agents, not elsewhere 0.18 0.01 0.00
classified
Cardiac, calcium channel 0.05 0.01 0.00
Predicted 6-Month Expenditures
Patients eventually diagnosed with or $5,790
treated for insomnia
Matched comparison group $4,648
Difference $1,143 (P < 0.01)*
*P-value comes from Wald chi-squared test of regression coefficient
for Eventual Insomnia Patient variable
Source: MarketScan© Research Databases: 1999-2003.
Cost Burden of Untreated Insomnia—Ozminkowski et al
SLEEP, Vol. 30, No. 3, 2007 270
elderly patients who were diagnosed with or treated for insomnia
had adjusted expenditures of about $5,790, while those never di-
agnosed with or treated for insomnia had expenditures averaging
$4,647. The difference of $1,143 is the estimated direct cost of
untreated insomnia for elderly patients.
Table 5 presents the results obtained from the 2-part regression
models used to analyze absenteeism costs among working adults.
Prior to conducting any regression analyses, absenteeism costs
were measured for each patient by counting all days absent in the
6-month period prior to the index date, and multiplying the num-
ber of days by $240 – the estimated value of a day’s wages and
benefits. The same $240 per day multiplier was used for all em-
ployees, regardless of whether insomnia was diagnosed or treated.
The $240 value of a lost workday is a compromise based on the
$193.20 value suggested by the Bureau of Labor Statistics for all
US companies in 2002 and the $344 per day value that pertains to
very large companies like the ones who contributed to the Health
and Productivity Management database, as found in a benchmark-
ing study conducted by Goetzel et al.33
Table 5 shows that absenteeism costs were significantly in-
fluenced by census region, plan type, urban/rural location, index
year, and some of the comorbidity and prescription drug use pat-
terns. Absenteeism costs were also significantly higher for patients
eventually diagnosed with or treated for insomnia, after control-
ling for these factors. Average absenteeism costs were $3,042 for
patients eventually diagnosed with or treated for insomnia and
$2,637 for other patients, a difference of $405.
Table 6 presents the result obtained from the 2-part regression
model used to analyze short-term disability program costs. The
likelihood of using any short-term disability program services in
the 6 months prior to the index date was significantly higher for
patients eventually diagnosed with or treated for insomnia. How-
ever, utilization of short-term disability services was lower during
that 6-month period for these eventual insomnia patients. Overall,
total short-term disability expenditures were $86 lower for pa-
tients who were diagnosed with or treated for insomnia, on aver-
age ($310 for eventual insomnia patients, and $396 for patients
never diagnosed with or treated for insomnia, P <0.0001).
Sensitivity Analyses
These cost burden estimates were not heavily influenced by a
small number of outlier cases who had very large costs. In analy-
ses of those under age 65 (not shown here), we dropped outliers,
in an effort to equalize the ranges of medical expenditures, absen-
teeism costs, and short-term disability benefit costs for eventual
insomnia and noninsomnia patients in the 6 months prior to the in-
dex date. This further leveled the playing field before the cost bur-
den was estimated. Since the ranges of expenditures were already
close after the matching process was conducted, only 19 sample
members were dropped. The only cost burden estimate that was
affected was for medical expenditures, which were $771 higher
for those eventually diagnosed with or treated for insomnia, com-
pared to the the $924 estimate found prior to dropping outliers
(details are available upon request). Dropping outliers (only 49
sample members) from the Medicare direct cost analyses changed
Table 5—Two-part regression model of absence payments
3,033 patients eventually diagnosed with or treated for insomnia; 4,058 matched controls.
Two-Part Model Part I: Logistic Regression Part 2: Exponential Cost Model
Dependent Variable Any absence from work Dollar value of absence, when it occurred
in 6 Months prior to index date within 6 months prior to index date
Independent Variable Odds Ratio Chi-square Parameter Standard Wald
P-value Estimate Error P-value
Intercept <0.01 8.17 0.04 0.00
Eventual Insomnia Patient 0.88 0.06 0.14 0.02 0.00
Female 1.11 0.12 -0.02 0.02 0.38
Northeast Census Region 0.76 0.02 -0.10 0.04 0.01
North Central Census Region 0.67 <0.01 -0.23 0.02 0.00
West Census Region 0.77 0.01 -0.22 0.03 0.00
Resided in Urban Area 0.47 <0.01 -0.29 0.03 0.00
Point of Service Plan Type 0.65 0.00 0.02 0.03 0.46
Preferred Provider Organization Plan Type 0.38 <0.01 0.58 0.04 0.00
Index Year = 1999 0.78 0.04 0.01 0.04 0.90
Index Year = 2000 15.12 <0.01 0.04 0.03 0.14
Index Year = 2001 0.92 0.28 0.01 0.03 0.77
Comorbidities and Drug Use Measures
Preventive health encounters 1.24 0.03 0.03 0.03 0.23
Disorders of lipoid metabolism 0.67 0.01 0.04 0.05 0.45
Gastrointestinal drug miscellaneous, not elsewhere classified 1.05 0.68 0.14 0.03 0.00
Analgesics/antipyritics, nonsteroidals/anti-inflammatories 1.11 0.24 0.10 0.02 0.00
Unclassified agents, not elsewhere classified 0.98 0.88 0.15 0.04 0.00
Analgesics/antipyretics, opiate agonists 1.60 <0.01 0.17 0.02 0.00
Predicted 6-Month Expenditures
Patients eventually diagnosed with or treated for insomnia $3,042
Matched comparison group $2,637
Difference $405 (P < 0.0001)
*P-value comes from Wald chi-squared test of regression coefficient for Eventual Insomnia Patient variable
Source: MarketScan© Research Databases: 1999-2002.
Cost Burden of Untreated Insomnia—Ozminkowski et al
SLEEP, Vol. 30, No. 3, 2007 271
the estimated direct cost of untreated insomnia among the elderly
to $1,128 on average, compared with $1,143 found when the en-
tire sample was used (details are available upon request).
The results reported above are based upon the assumption that
insomnia existed for 6 months prior to diagnosis or treatment.
Without survey data, this cannot be verified. To address this is-
sue, we estimated the direct cost burden by month, for periods
ranging from 1 month to 5 months before the index date. The
details are available upon request, but all analyses showed statis-
tically significant and higher direct costs for patients eventually
diagnosed with or treated for insomnia. These costs estimates for
untreated insomnia ranged from $677 (for a 1-month analysis) to
$800 (for the 3-month analysis) on average, for younger adults.
For Medicare beneficiaries, the average direct costs of untreated
insomnia ranged from $994 (for a 1-month analysis) to $1,369
(for a 3-month analysis)
DISCUSSION
The objective of this study was to estimate the 6-month cost
burden of untreated insomnia, by focusing on differences in direct
and indirect costs between patients eventually diagnosed with or
treated for insomnia and similar patients who were not.
Why focus on untreated insomnia? Why not estimate cost
burden for insomnia the way that cost burden is often estimated
for other diseases (i.e., by focusing on the cost of treatment?)
We thought it would be more informative to focus on the period
prior to diagnosis or treatment because insomnia often goes un-
treated.6,10 Moreover, the costs of treating insomnia are generally
quite low and therefore of limited financial consequence. In the
MarketScan Commercial Claims and Encounter data we used,
treatment costs rarely exceeded $200 in the year after the index
date. Thus, the medical claims data suggest that, unlike many
other conditions, insomnia is not an expensive condition to treat
when it occurs (details are available upon request). We therefore
assumed that a more complete understanding of its burden of ill-
ness can be gained by estimating cost differences between similar
insomnia and noninsomnia patients, shortly before diagnosis or
treatment begins. Others may wish to focus on the posttreatment
period, comparing the costs of treatment, however small, with
alternative medications.
Next, given our focus on the prediagnosis or pretreatment pe-
riod, how can we infer that the added costs we observed were
related to insomnia? The answer to this question lies in the meth-
ods we used to control for differences between those eventually
diagnosed with or treated for insomnia, and those who were not.
The propensity score analyses, and the subsequent regression
analyses, ruled out the impact of demographics, plan type, loca-
tion, year of entry into the study, comorbidities, and the use of
various pharmaceuticals, as reasons for observing the cost differ-
ences we found. By leveling the field in terms of these factors, it
is more likely that the added costs for those who eventually were
Table 6—Two Part Regression Model of Short-Term Disability Payments
5,028 patients eventually diagnosed with or treated for insomnia; 6,635 matched controls.
Two-Part Model Part I: Logistic Regression Part 2: Exponential Cost Model
Dependent Variable Short-term disability program use Dollar value of short-term disability
in months prior to index date 6 program within 6 months prior
to index date
Independent Variable Odds Ratio Chi-square Parameter Standard Wald
P-value Estimate Error P-value
Intercept <0.01 8.42 0.19 0.00
Eventual Insomnia Patient 1.71 <0.01 -0.25 0.07 0.00
Age 1.01 0.13 0.00 0.00 0.62
Female 1.18 0.02 0.13 0.07 0.04
Northeast 1.25 0.05 0.23 0.10 0.03
North Central 1.32 0.00 -0.01 0.07 0.88
West 0.71 0.00 0.11 0.10 0.28
Point of Service Plan Type 0.78 0.01 0.13 0.09 0.16
Preferred Provider Organization Plan Type 0.84 0.16 -0.10 0.12 0.37
Index Year = 1999 0.94 0.65 0.05 0.12 0.70
Index Year = 2000 0.93 0.32 0.06 0.07 0.36
Index Year = 2003 0.82 0.52 -0.17 0.29 0.54
Number of Psychiatric Diagnosis Groups 1.46 <0.01 0.08 0.06 0.17
Comorbidities and Drug Use Measures
Preventive health encounters 1.33 0.00 -0.11 0.08 0.16
General symptoms 1.28 0.15 0.20 0.15 0.17
Other symptoms involving abdomen and pelvis 2.75 <0.01 -0.04 0.14 0.77
Psychother, antidepressants 0.67 0.00 -0.05 0.11 0.64
Analgesics/antipyretics, nonsteroidals/anti-inflammatories 1.98 <0.01 0.10 0.07 0.18
Antihistamines & combinations, not elsewhere classified 0.91 0.33 -0.09 0.09 0.29
Predicted 6-Month Expenditures
Patients eventually diagnosed with or treated for insomnia $310.4
Matched comparison group $396.8
Difference -$86.4 (P < 0.01)*
*P-value comes from Wald chi-squared test of regression coefficient for Eventual Insomnia Patient variable
Source: MarketScan© Research Databases: 1999-2002.
Cost Burden of Untreated Insomnia—Ozminkowski et al
SLEEP, Vol. 30, No. 3, 2007 272
diagnosed with or treated for insomnia were due to that disorder,
not to other factors.
After matching and regression-based adjustments were made,
we found that direct medical expenditures were $924 higher for
younger patients eventually diagnosed with or treated for insom-
nia, compared with those who were not. Direct medical expen-
ditures were also $1,143 higher for elderly patients eventually
diagnosed with or treated for insomnia, compared with elderly
patients who were not.
These estimates are comparable to earlier estimates of health
care costs of insomnia from a much smaller health maintenance
organization patient sample17 and also similar to the health care
costs of depressive and anxiety disorders.34 We also found dif-
ferences in indirect costs. Specifically, absenteeism costs were
$405 higher for those eventually diagnosed with or treated for
insomnia, but short-term disability costs were $86 lower for those
patients, on average.
Other studies have also addressed the impact of insomnia on
absenteeism. Two recent studies were conducted by Leger et al.35
and Godet-Cayre36 in France. Both studies used the same sample,
but analyzed the data differently. Leger et al. focused on the re-
lationship between insomnia and days of work lost, by job type.
Godet-Cayre focused on the added costs of treated insomnia for
employers and the added costs to the national health care sys-
tem. Both found that absenteeism costs or absent days were about
twice as high for insomnia patients as for good sleepers. We also
found higher costs for those eventually diagnosed with or treated
for insomnia. Our $405 average cost increase for absenteeism,
coupled with our $86 decrease in short-term disability benefits,
amounts to about 1.3 more days’ lost work that may be due to
untreated insomnia. Godet-Cayre found that insomnia was associ-
ated with an additional 3.4 days’ lost work in their sample.
What do our results mean for the typical patient and employer?
In our database, the average insomnia patient paid about 20% of
total medical expenditures out of his or her pocket (the employer
paid the rest). As noted in Table 3, the average medical cost bur-
den of untreated insomnia for those under age 65 was $924. With
a 20% patient share, $184 would have been paid by the patient,
and the other $740 would have been paid by the employer.
The $740 employer share is equivalent to about 3 days’ wages
and benefits (i.e., $740 / $240 value of a day’s wages and benefits
= 3.1 days). Adding the cost of absenteeism and short-term dis-
ability program use (which are also self-insured by the employer),
another $319 would be paid by the employer. The $319 estimate
is the difference between the $405 untreated insomnia cost burden
for absenteeism noted in Table 5, and the $86 cost savings from
lower short-term disability program use by untreated insomnia
patients, also shown in Table 6. This $319 net cost increase is
equivalent to roughly 1.3 days’ wages and benefits, bringing the
total employer’s cost to 4.4 days’ wages and benefits per untreat-
ed insomnia patient. We do not yet know how much of this cost
could be avoided by successful treatment, but would guess that
most employers would consider this cost burden to be important.
The analyses conducted for this project were limited by the
following factors.
First, the burden of insomnia we estimated was financial. We
were unable to estimate the impact of insomnia on psychosocial
functioning, accident rates, or productivity while at the worksta-
tion (presenteeism). Thus, our burden estimates may be conserva-
tive.
Second, absenteeism and disability data were not available for
all sample members, and the number of sample members with ab-
senteeism data was sparse in the earlier years of the study, prior to
2001. This is because the initial request for absenteeism data was
made in 2001, and many employers did not retain data for previ-
ous years. This means that analyses of absenteeism data may not
be generalizable beyond our sample, and cost burden estimates
may vary in other settings.
Third, it has been noted in the literature that many insomnia
patients do not seek medical care to treat that disorder.37 Thus,
some members of the comparison group may have had undiag-
nosed or untreated insomnia. This may also lead to a conservative
estimate of cost burden.
Fourth, without medical records, it is impossible to verify that
all of the insomnia drugs we considered were indeed taken for
that purpose. For example, low-dose amitriptyline may be used
for pain. We expect the majority of uses to be for insomnia, but
if we are incorrect then some bias may have resulted in our find-
ings.
Finally, other drugs may be used for insomnia that we did not
consider. For example, other benzodiazepines (e.g., clonazepam,
alprazolam, lorazepam) or antipsychotics were not considered,
even though they may be used for insomnia in some cases.38
Anxiolytic drugs may be used for insomnia, as may quetiapine,
hydroxyzine, and diphenhydramine,39 but we did not include pa-
tients who took these drugs in our analyses because dose cannot
be used to distinguish between use for insomnia or other condi-
tions. Results may have differed if patients with these drugs were
included.
Acknowledging these limitations, we also note some advantag-
es to the analyses we conducted. Specifically, the matching and
regression processes accounted for 44 demographic and casemix
variables that might differ between insomnia and noninsomnia
patients. We also used more recent econometric techniques (i.e.,
2-part exponential cost regression models) to account for the
natural skew in cost data, without the need for logarithmic trans-
formations in the estimation process. As a result, we were able to
account for a large number of comorbidities while producing a set
of reliable estimates for the cost burden of insomnia. These esti-
mates will complement the estimates produced in prior research.
Finally, for readers who are outside the United States and less
familiar with the US health system, some context may be added
by noting that our data come from large employers who, along
with their employees, paid for the health care services received by
sample members. Payments to health care providers came direct-
ly out of company or personal funds; they were not paid by insur-
ance premiums. Outside insurance companies were used only as
vendors to provide administrative services; the employers did not
pay the insurance companies premiums for health care coverage.
In the late 1990s (most recent data available), about two-thirds of
those under age 65 in the US were covered by employer-spon-
sored, self-insured plans.40
Many of these employers also offered self-insured, paid sick
leave, but usually for only a few days per year. Generally, full-
time employees who were sick >5 consecutive days were also
eligible for short-term disability program benefits, which paid
the employee for 60% to 70% of lost wages while not at work.
Employees who had sick leave and short-term disability benefits
generally had higher incomes than average.31
To the extent that medical coverage and the availability and
Cost Burden of Untreated Insomnia—Ozminkowski et al
SLEEP, Vol. 30, No. 3, 2007 273
use of paid sick leave and disability program services are different
in the US than abroad, the results noted here may not generalize
well beyond US borders. Others may wish to investigate the cost
burden of untreated insomnia outside the United States.
ACKNOWLEDGEMENTS
This work was funded by Sepracor, Inc., under contract to
Thomson Medstat, Ann Arbor, Michigan. No recommendations
about pharmacotherapy are made for either on-label or off-label
use.
REFERENCES
1. American Academy of Sleep Medicine. International classification
of sleep disorders, 2nd ed.: diagnostic and coding manual.Westches-
ter, IL: American Academy of Sleep Medicine, 2005.
2. Ohayon MM. Epidemiology of insomnia: what we know and what
we still need to learn. Sleep Med Rev 2002;6:97-111.
3. Ohayon MM. Prevalence of DSM-IV diagnostic criteria of insom-
nia: distinguishing insomnia related to mental disorders from sleep
disorders. J Psychiat Res 1997;31:333-46.
4. Buysse DJ, Reynolds CF, Kupfer DJ, et al. Clinical diagnoses in
216 insomnia patients using the international classification of sleep
disorder (ICSD), DSM-IV, and ICD-10 categories: a report from the
APA/NIMH DSM-IV field trial. Sleep 1994;17:630-7.
5. Roth T. The relationship between psychiatric diseases and insomnia.
Int J Clin Pract 2001(Suppl 116):3-8.
6. Katz DA, McHorney CA. Clinical correlates of insomnia in patients
with chronic illness. Arch Intern Med 1998;158:1099-1107.
7. Benca RM. Consequences of insomnia and its therapies. J Clin Psy-
chiat 2001;62(Suppl 10):33-8.
8. National Institutes of Health State of the Science Conference State-
ment on Manifestations and Management of Chronic Insomnia in
Adults, June 13-15, 2005. Sleep 2005;28:1049-57.
9. Foley DJ, Monjan AA, Brown SL, Simonsick EM, Wallace RB,
Blazer DG. Sleep complaints among elderly persons: an epidemio-
logic study of three communities. Sleep 1995;18:425-32.
10. Mallon I, Broman JE, Hetta J. Relationship between insomnia, de-
pression, and mortality : a 12-year follow-up of older adults in the
community. Int Psychogeriatr 2000;12:295-306.
11. Katz DA, McHorney CA. The relationship between insomnia and
health-related quality of life in patients with chronic illness. J Fam
Pract 2002;51:229-235.
12. Zammit GK, Weiner J, Damata N, et al. Quality of life in people
with insomnia. Sleep 1999;22(Suppl 2):S379-S385.
13. Ford DE, Kamerow DB. Epidemiologic study of sleep disturbances
and psychiatric disorders: an opportunity for prevention? JAMA
1989;262:1479-84.
14. Chang PP, Ford DE, Mead LA, et al. Insomnia in young men and
subsequent depression: the Johns Hopkins Precursors Study. Am J
Epidemiol 1997;146:105-14.
15. Walsh JK. Clinical and socioeconomic correlates of insomnia. J Clin
Psychiatry 2004;65(Suppl 8):13-19.
16. Hatoum HT, Dong SX, Darnia CM, Wong JM, Mendelson WR. In-
somnia, health-related quality of life, and healthcare resource con-
sumption: a study of managed care enrollees. Pharmacoeconomics
1998;14:629-47.
17. Simon GE, VonKorff M. Prevalence, burden, and treatment of in-
somnia in primary care. Am J Psychiatry 1997;154:1417-23.
18. Leger D, Guilleminault C, Bader G, Levey E, Palliard M. Medical
and socio-professional impact of insomnia. Sleep 2002;25:625-9.
19. Walsh JK, Engelhardt CL. The direct economic costs of insomnia in
the United States for 1995. Sleep 1999;22(Suppl 2):S386-S393.
20. Chilcott LA, Shapiro CM. The socioeconomic impact of insomnia:
an overview. Pharmacoeconomics 1996;10(Suppl 1):1-14.
21. Laschober M. Trends in Medicare supplemental insurance and
prescription drug benefits, 1996 –2001. Data update prepared for
the Henry J. Kaiser Family Foundation. Bearing Point, Inc., June
2004.
22. Charlson M, Pompei P, Ales A, MacKenzie R. A new method of
classifying prognostic comorbidity in longitudinal studies: develop-
ment and validation. J Chron Dis 1987;40:373-83.
23. Farley JF, Harley CR, Devine JW. A comparison of comorbidity
measurements to predict healthcare expenditures. Am J Manag Care
2006;12:110-19.
24. Ashcraft ML, Fries BE, Nerenz DR, et al. A psychiatric patient clas-
sification system: an alternative to diagnosis-related groups. Med
Care 1989;27:543-57.
25. Brunello N, Armitage R, Feinberg I, et al. Depression and sleep dis-
orders: Clinical relevance, economic burden, and pharmacological
treatment. Neuropsychobiology 2000;42:107-19.
26. Miller ME, Sulvetta MB, Englert E. Service mix in the hospital
outpatient department: implications for Medicare payment reform.
Health Serv Res 1995;30:59-78.
27. Fisher ES, Wennberg JE, Stukel TA, et al. Associations among
hospital capacity, utilization, and mortality of US Medicare benefi-
ciaries, controlling for sociodemographic factors. Health Serv Res
2000;34:1351-62.
28. Luft HA, ed. HMOs and the elderly. Ann Arbor, MI: Health Admin-
istration Press, 1994.
29. Drake C, Fisher L. Prognostic models and the propensity score. Int
J Epidemiol 1995;24:183-7.
30. McCaffrey D, Ridgeway G, Moral A. Propensity score estimation
with boosted regression for evaluating causal effects in observa-
tional studies. Psychol Methods 2004;9:403-25.
31. Mullahy J. Much ado about two: reconsidering retransformation
and the two-part model in health econometrics. J Health Econ
1998;17:247-81.
32 . Heckman JJ, Hidehiko I, Prieti T. Matching as an econometric eval-
uation estimator: evidence from a job training program. Rev Econ
Stud 1997;64:605-54.
33. Goetzel RZ, Guindon A, Turshen I, Ozminkowski R. Health and
productivity management: establishing key performance measures,
benchmarks, and best practices. J Occup Environ Med 2001;43:10-
17.
34. Simon G, Ormel J, VonKorff M, Barlow W. Health care costs as-
sociated with depressive and anxiety disorders in primary care. Am
J Psychiatry 1995;152:352-7.
35. Leger D, Massuel MA, Mellaine A, and the SISYPHE Study Group.
Professional correlates of insomnia. Sleep 2006;29:171-8.
36. Godet-Cayre V, Pelletier-Fluery N, Le Valliant M, Dinet J, Mas-
sual MA, Leger D. Insomnia and absenteeism at work.Who pays the
cost? Sleep 2006;29:179-84.
37. Leger D. Public health and insomnia: economic impact. Sleep
2000;(Suppl 3):S69-S76.
38. Walsh JK. Drugs used to treat insomnia in 2002: regulatory-based
rather than evidence-based medicine. Sleep 2004;27:1441-2.
39. Morelock RJ, Tan M, Mitchell DY. Patient characteristics and pat-
terns of drug use for sleep complaints in the United States: Analysis
of National Ambulatory Medical Care Survey data, 1997 – 2002.
Clin Ther 2006;28:1044-53.
40. Fronstin P. Features of employment-based health plans. EBRI Issue
Brief 1998;201:1-21.
Cost Burden of Untreated Insomnia—Ozminkowski et al
... Disturbances affecting the duration or quality of sleep occur in nearly half of older adults. 1,2 These disturbances are associated with cognitive decline, depression, disability, institutionalization, and high healthcare costs, [3][4][5][6][7] calling for their accurate identification and treatment. Unfortunately, self-reported sleep may not accurately reflect habitual sleep duration or quality, leading to missed treatment opportunities or overtreatment with high-risk medications. ...
... 9. I felt very confident using the sleep headband. 4 (3,5) 10. I needed to learn a lot of things before I could get going with the sleep headband.* ...
... I needed to learn a lot of things before I could get going with the sleep headband.* 5 (4,5) 11. I think the sleep headband could help people. ...
Preprint
Study Objectives: In older adults, self-reported sleep measures may be inaccurate, but polysomnography (PSG) is burdensome. We assessed the performance of an electroencephalography-measuring headband (HB) or actigraphy (ACT) compared with PSG in older adults with sleep disturbances. Methods: Sixty-three adults aged ≥60 years who reported symptoms of insomnia and/or daytime sleepiness ≥once/week completed a week-long, home-based protocol during which they wore the HB for seven nights, an actigraph for seven days and nights, and completed a one-night level II unattended PSG. For the current analysis, we compared total sleep time (TST) and wake after sleep onset (WASO) from all three devices on the PSG night. We calculated absolute differences and intraclass correlation coefficients (ICCs) for TST and WASO between HB and ACT, respectively, vs. PSG. We also evaluated the performance of the HB among subgroups of the poorest sleepers according to the presence of sleep apnea, insomnia, poor sleep quality, and periodic limb movements of sleep. Feasibility of the HB was assessed by measures of adherence (i.e., ability to use the HB over seven nights) and usability (i.e., ratings of items from the WEarable Acceptability Range [WEAR] scale). Results: The average age was 72.8 [standard deviation 6.6] years, 63.5% were female, and 63.5% identified as non-Hispanic White. On PSG, averages for TST and WASO were 370.1 [93] and 88.9 [63] minutes, respectively. For the HB vs. PSG, mean differences and ICCs were -11.9 minutes and 0.83 [0.74, 0.89] for TST; and -15.5 minutes and 0.65 [0.48, 0.77] for WASO. For ACT vs. PSG, mean differences for TST and WASO were larger, and ICCs showed lower levels of agreement. The HB performed well among the poorest sleepers, with ICCs >0.65 for TST and WASO. On average, participants wore the HB for 6.5 [0.8] nights, and usability was rated highly. Conclusions: The HB demonstrated good agreement with PSG, outperforming ACT, including among the poorest sleepers. Devices like the HB might provide feasible measures of sleep that are more accurate than ACT and enhance the management of sleep health in older adults with sleep disturbances. Future research should focus on further validation of these devices in habitual sleep environments.
... In addition, there is a large literature connecting sleep problems and insomnia disorder among employees to a range of work performance factors and work-related costs borne by employers. Many studies have found insomnia to be associated with increased work absenteeism and presenteeism [37] [38] [39] [40] [41] [42] [43][44] [45] . Insomnia among employees Qeios, CC-BY 4.0 · Article, Qeios ID: 4YS33S.2 ...
... · https://doi.org/10.32388/4YS33S. 2 2/29 is also associated with workplace accidents [46] [47] . The healthcare and workplace costs associated with insomnia combine to create a significant economic burden for businesses [16] [38][40] [48] and for society in general [49] . ...
Preprint
Full-text available
Sleep problems were examined in archival data from 18,646 users of a commercial service that provided online health risk screening, educational resources, and self-guided computerized therapy lessons for insomnia and other mental health disorders. The sample was split between college students and working adults and represents a growing modern segment of adults who voluntarily seek out digital support for common behavioral issues. The goals were to explore the prevalence and possible correlates of insomnia among this unique sample. The cognitive behavioral-based therapy from this service has evidence of its clinical effectiveness and value to users in past research. Results revealed that 36% of all users were at risk for a clinical insomnia disorder. The severity of insomnia was significantly (all _p _< .001) associated with the severity of depression (_r_ = .65; 43% clinical); anxiety (_r_ = .54; 40% clinical); stress (_r_ = .54; 25% clinical); social phobia (_r_ = .34; 27% clinical); and general health status (_r_ = -.26; 15% clinical). Younger age was weakly associated with insomnia (_r_ = -.14; avg. 32 years; range 18-83), while both gender (_r_ = -.05; 76% female) and race (_r_ = .00; 81% White) were unrelated to insomnia. Insomnia was associated with lower work performance and greater work absenteeism (_r_ = -.30; _r_ = .17, respectively). The conclusions are that insomnia was commonly experienced, often comorbid with other common mental health conditions, and linked to work performance problems. Thus, online self-help health services should screen for multiple disorders, including insomnia, rather than focusing on specific disorders.
... In addition, there is a large literature connecting sleep problems and insomnia disorder among employees to a range of work performance factors and work-related costs borne by employers. Many studies have found insomnia to be associated with increased work absenteeism and presenteeism [37] [38] [39] [40] [41] [42] [43] [44] [45] . Insomnia among employees Qeios, is also associated with workplace accidents [46] [47] . ...
... Insomnia among employees Qeios, is also associated with workplace accidents [46] [47] . The healthcare and workplace costs associated with insomnia combine to create a significant economic burden for businesses [16] [38][40] [48] and for society in general [49] . ...
Preprint
Full-text available
Sleep problems were examined in archival data from 18,646 users of a commercial service that provided online health risk screening, educational resources, and self-guided computerized therapy lessons for insomnia and other mental health disorders. The sample was split between college students and working adults and represents a growing modern segment of adults who voluntarily seek out digital support for common behavioral issues. Goals were to explore the prevalence and possible correlates of insomnia among this unique sample. The cognitive behavioral-based therapy from this service has evidence of its clinical effectiveness and value to users in past research. Results revealed 36% of all users were at risk for clinical insomnia disorder. Severity of insomnia was significantly (all _p _<.001) associated with severity of depression (_r_ =.65; 43% clinical); anxiety (_r_ =.54; 40% clinical); stress (_r_ =.54; 25% clinical); social phobia (_r_ =.34; 27% clinical); and general health status (_r_ = -.26; 15% clinical). Younger age was weakly associated with insomnia (_r_ = -.14; avg. 32 years; range 18-83) while both gender (_r_ = -.05; 76% female) and race (_r_ =.00; 81% White) were unrelated to insomnia. Insomnia was associated with lower work performance and greater work absenteeism (_r_ = -.30; _r_ =.17, respectively). The conclusions are that insomnia was commonly experienced, often comorbid with other common mental health conditions, and linked to work performance problems. Thus, online self-help health services should screen for multiple disorders, including insomnia, rather than focusing on specific disorders.
... Insomnia increases the risk of detrimental health outcomes such as cardiovascular disease, psychiatric disorders, obesity, diabetes, cognitive impairments (e.g., Alzheimer's disease), chronic pain, and hyperlipidaemia (6-8). Estimates show 60% higher healthcare costs among patients with untreated insomnia compared to healthy controls (9). The high prevalence rates combined with the considerable individual, economic and healthcare burden of untreated insomnia emphasize the importance of readily available and effective interventions. ...
Preprint
Full-text available
Background: Insomnia is a prevalent disorder in the general population. However, access to the treatment of choice, cognitive-behavioural therapy for insomnia (CBT-I), is very limited. Accessibility increases with group delivery and the aim of this trial was to investigate the effectiveness of a low-threshold group-delivered CBT-I implemented in primary care. Methods: This pragmatic randomized controlled trial was conducted across 26 Healthy Life Centres in Norwegian primary care. Inclusion criteria were age ≥ 18 years and insomnia symptoms with a significant impact, defined as a score of ≥ 12 on the Insomnia Severity Index (ISI). Eligible participants were randomized to group-delivered CBT-I or to a waiting list according to a 2:1 ratio using a computerized centre-stratified allocation sequence. The intervention consisted of four two-hour sessions during a 4-week period, with group sizes ranging from 5 to 15 participants. The primary outcome was self-reported insomnia severity at 3-months follow-up, measured by the ISI. Participants were not blinded to group assignment, but those responsible for analyses and interpretation of the results were blinded. Results : A total of 308 participants were enrolled and randomly allocated to group-delivered CBT-I (n=181) or waiting list (n=127). In all, 126 (99.2%) participants on the waiting list and 179 (98.9%) receiving group-delivered CBT-I completed the ISI at baseline, while 115 (90.6%) participants on the waiting list and 153 (84.5%) receiving group-delivered CBT-I completed the ISI at 3-months follow-up. The mean difference in ISI score between groups at 3-months follow-up was -3.4 (95% CI: -4.3 to -2.5), favoring the group-delivered CBT-I. Of those receiving group-delivered CBT-I, 51 (33.6%) reported a clinically relevant improvement in insomnia severity (i.e., ≥8 points on ISI), compared to 15 (13.2%) on the waiting list (OR 3.4, 95% CI: 1.7 to 6.8). No major adverse events were reported by the participants. Conclusions : Group-delivered CBT-I reduced insomnia severity at 3-months follow-up for adults with insomnia. These findings suggest that group-delivered CBT-I offered at Healthy Life Centres is a suitable low-threshold treatment for insomnia in primary care. Trial registration : The study was retrospectively registered in the ISRCTN registry on 17/11/2022 (ISRCTN16185698).
... Poor sleep health is devastating for people and societies, damaging health, social functioning, and costing billions annually Ozminkowski et al., 2007). As common targets of racial discrimination and inequity, Black Americans shoulder an oversize burden for lost sleep and resultant health problems (e.g., Stamatakis et al., 2007;Williams, 2012). ...
Article
Objective: Discrimination disrupts sleep and contributes to race-based health inequities for Black Americans, but less is known about the psychological mechanisms underlying this relation. The current work tests whether emotion invalidation, termed Social Pain Minimization (SPM), mediates discrimination's negative effects on sleep quality. We focus on the experiences of Black Americans because of racism's disproportionate effect on the health outcomes of Black individuals in the U.S. Methods: Four studies with cross-sectional (Studies 1a-1b) and cross-lagged panel designs (Studies 2-3; Ntotal = 1,176) were used. Discrimination and SPM were assessed with established measures and self-reported sleep quality was assessed with three different operationalizations including the short form Patient-Reported Outcomes Measurement Information System (PROMIS) Sleep-Related Impairment (SRI) subscale and composites of the Insomnia Severity Index (ISI). Results: Cross-sectional (Studies 1a-1b) and longitudinal mediation analyses (Studies 2-3) indicated discrimination's negative effect on sleep quality partially operated through SPM (Study 1a, 95% CI = [-.24, -.016]); Study 1b, 95% CI = [-.22, -.03]; Study 2, 95% CI = [-.268, -.007]]; Study 3, 95% CI = [-0.043, -.002]). Conclusion: Beyond its direct negative effect on sleep, discrimination also operates through SPM to reduce sleep quality. Emotion invalidation stemming from discrimination damages sleep.
... 12,13 Furthermore, it is noteworthy that over two-thirds of individuals with insomnia do not seek medical assistance and instead resort to over-the-counter medications or self-treatment. 14 As a result, there has been growing interest in exploring complementary and alternative medicine (CAM) approaches to improve sleep quality and duration, including the use of herbal remedies. [15][16][17][18] Chamomile, derived from the German chamomile, Matricaria chamomilla L. (Asteraceae), has a long history of traditional use as a medicinal herb, particularly for its calming and sedative properties. ...
Article
Full-text available
Sleep problems were examined in archival data from 18,646 users of a commercial service that provided online health risk screening, educational resources, and self-guided computerized therapy lessons for insomnia and other mental health disorders. The sample was split between college students and working adults and represents a growing modern segment of adults who voluntarily seek out digital support for common behavioral issues. The goals were to explore the prevalence and possible correlates of insomnia among this unique sample. The cognitive behavioral-based therapy from this service has evidence of its clinical effectiveness and value to users in past research. Results revealed that 36% of all users were at risk for a clinical insomnia disorder. The severity of insomnia was significantly (all _p _<.001) associated with the severity of depression (_r_ =.65; 43% clinical); anxiety (_r_ =.54; 40% clinical); stress (_r_ =.54; 25% clinical); social phobia (_r_ =.34; 27% clinical); and general health status (_r_ = -.26; 15% clinical). Younger age was weakly associated with insomnia (_r_ = -.14; avg. 32 years; range 18-83), while both gender (_r_ = -.05; 76% female) and race (_r_ =.00; 81% White) were unrelated to insomnia. More severe insomnia was associated with lower work performance and greater work absenteeism (_r_ = -.30; _r_ =.17, respectively). The conclusions are that insomnia was commonly experienced among users of digital tools, often comorbid with other common mental health conditions, and linked to work performance problems. Thus, online self-help health and related digital services should screen for multiple disorders, including insomnia, rather than focusing on the specific disorder emphasized in the care program.
Article
Study Objectives Insomnia has profound negative effects on behavioral health, physical health, and functional domains. Leveraging claims data from one of the nation’s largest payor-provider systems, the current study examined the real-world prevalence of insomnia, comorbidity of insomnia with behavioral health and other sleep disorders, and the impact of insomnia on total health care costs. Methods Prevalence and costs associated with insomnia were assessed by examining claims data on approximately 3 million insured members during the year 2022. Using propensity score matching, total health care expenditures were calculated and compared for members with insomnia relative to matched cohorts without insomnia. Generalized linear modeling tested for differences between the cohorts. Results Nine percent of members were identified as having insomnia; 64% of those also had depression, anxiety, and/or substance use disorder. Median total health care costs among individuals with insomnia were 4–6 times greater than among those without insomnia. A disproportionate amount (21.1%) of total claims spend came from members with insomnia. Conclusions Findings demonstrate a high degree of clinical need and behavioral health comorbidity associated with insomnia within a large insured cohort. Beyond the clinical significance, the current results demonstrate substantial financial need and opportunity for adequately treating insomnia. This is especially the case for the high proportion of members with insomnia and cooccurring depression, anxiety, and/or substance use disorders. Overall, the findings point to the important role payors and providers may have in promoting greater attention to sleep and insomnia.
Article
Full-text available
The frequencies of five common sleep complaints--trouble falling asleep, waking up, awaking too early, needing to nap and not feeling rested--were assessed in over 9,000 participants aged 65 years and older in the National Institute on Aging's multicentered study entitled "Established Populations for Epidemiologic Studies of the Elderly" (EPESE). Less than 20% of the participants in each community rarely or never had any complaints, whereas over half reported at least one of these complaints as occurring most of the time. Between 23% and 34% had symptoms of insomnia, and between 7% and 15% percent rarely or never felt rested after waking up in the morning. In multivariate analyses, sleep complaints were associated with an increasing number of respiratory symptoms, physical disabilities, nonprescription medications, depressive symptoms and poorer self-perceived health. Sleep disturbances, particularly among older persons, oftentimes may be secondary to coexisting diseases. Determining the prevalence of specific sleep disorders, independent of health status, will require the development of more sophisticated and objective measures of sleep disturbances.
Conference Paper
Study Objectives: To assess the direct economic costs of insomnia in the United States in 1995. Methods: The costs of prescription medications were based on 1995 data compiled by IMS America, Ltd. (Plymouth Meeting, PA). Non-prescription medication expenditures were provided by Information Resources, Inc. (Chicago, IL). The costs of physician visits related to insomnia were estimated from unpublished data of the 1994 National Ambulatory Medical Care Survey conducted by the National Center for Health Statistics and from the America Medical Association Center for Health Policy Research. Several other sources were used for other cost estimates. Results: Total cost for substances used to treat insomnia was 1.97billion,lessthanhalfofwhichwasforprescriptionmedication.Healthcareservicesforinsomniatotaled1.97 billion, less than half of which was for prescription medication. Health care services for insomnia totaled 11.96 billion, 91% of which is attributable to nursing home care. The total direct costs in the United States for insomnia in 1995 were estimated to be $13.9 billion. Conclusions: Increased efforts are needed in several domains to offset the cost of insomnia including clinical research on the consequences of untreated and treated insomnia, development and implementation of curricula to provide knowledge about sleep and sleep disorders for medical students, physicians, and other health professionals, education to increase public awareness of insomnia and sleep disorders, and more support for basic research on neural mechanisms involved in healthy and disordered sleep.
Article
Sleep difficulties affect about one-third of American adults, yet these symptoms are not often addressed by patients and their physicians. Unresolved insomnia that impairs daytime function may be associated with significant psychiatric morbidity, predominantly major depression. Clinicians who are aware that these disorders frequently co-exist will be able to facilitate prompt diagnosis and initiate appropriate pharmacotherapeutic intervention. Antidepressant therapies may interfere with sleep quality; therefore, the management of insomnia in depressed patients should be addressed separately. Traditionally, benzodiazepines have been prescribed to treat sleep disturbances, but certain drugs in this class have limited benefits due to residual sedative effects that impair cognitive function, memory and general daytime performance. Zaleplon, a new, quick-acting, non-benzodiazepine sleep medication, may be clinically advantageous in promoting sleep without residual impairment.
Article
The objective of this study was to develop a prospectively applicable method for classifying comorbid conditions which might alter the risk of mortality for use in longitudinal studies. A weighted index that takes into account the number and the seriousness of comorbid disease was developed in a cohort of 559 medical patients. The 1-yr mortality rates for the different scores were: "0", 12% (181); "1-2", 26% (225); "3-4", 52% (71); and "greater than or equal to 5", 85% (82). The index was tested for its ability to predict risk of death from comorbid disease in the second cohort of 685 patients during a 10-yr follow-up. The percent of patients who died of comorbid disease for the different scores were: "0", 8% (588); "1", 25% (54); "2", 48% (25); "greater than or equal to 3", 59% (18). With each increased level of the comorbidity index, there were stepwise increases in the cumulative mortality attributable to comorbid disease (log rank chi 2 = 165; p less than 0.0001). In this longer follow-up, age was also a predictor of mortality (p less than 0.001). The new index performed similarly to a previous system devised by Kaplan and Feinstein. The method of classifying comorbidity provides a simple, readily applicable and valid method of estimating risk of death from comorbid disease for use in longitudinal studies. Further work in larger populations is still required to refine the approach because the number of patients with any given condition in this study was relatively small.
Article
It is generally accepted that diagnosis-related groups (DRGs) for alcohol, drug, and mental disorders are inappropriate for inpatient prospective payment. To address this issue, the Veterans Administration (VA) supported a project to construct alternative classes that are more clinically meaningful, more homogeneous in their resource use, and that account for more variation in resource use among psychiatric and substance use cases than existing DRGs. This paper reports on this project. Using a data set containing universally available discharge data plus behavioral, social, and functional information obtained by a survey of 116,191 discharges from VA psychiatric beds, and with AUTOGRP as the classifying algorithm, a classification system was formed. Twelve psychiatric diagnostic groupings (PDGs) were identified, analogous to major diagnostic groups in the DRG system. Within each PDG, from 4 to 9 terminal groups of Psychiatric Patient Classes (PPCs) were formed and validated. The 12 substance abuse PPCs explain greater than 31% of the variation in length of stay; for the mental disorder PPCs the variance explanation is greater than 11%, a substantial improvement over DRGs that, for the same data set, explain less than 2 and 3%, respectively. With the addition of only 5 variables beyond those presently included in discharge data sets, greater precision for payment purposes can be achieved. Implications for adoption of this classification system are discussed.
Article
As part of the National Institute of Mental Health Epidemiologic Catchment Area study, 7954 respondents were questioned at baseline and 1 year later about sleep complaints and psychiatric symptoms using the Diagnostic Interview Schedule. Of this community sample, 10.2% and 3.2% noted insomnia and hypersomnia, respectively, at the first interview. Forty percent of those with insomnia and 46.5% of those with hypersomnia had a psychiatric disorder compared with 16.4% of those with no sleep complaints. The risk of developing new major depression was much higher in those who had insomnia at both interviews compared with those without insomnia (odds ratio, 39.8; 95% confidence interval, 19.8 to 80.0). The risk of developing new major depression was much less for those who had insomnia that had resolved by the second visit (odds ratio, 1.6; 95% confidence interval, 0.5 to 5.3). Further research is needed to determine if early recognition and treatment of sleep disturbances can prevent future psychiatric disorders.
Article
The objective of this study was to develop a prospectively applicable method for classifying comorbid conditions which might alter the risk of mortality for use in longitudinal studies. A weighted index that takes into account the number and the seriousness of comorbid disease was developed in a cohort of 559 medical patients. The 1-yr mortality rates for the different scores were: "0", 12% (181); "1-2", 26% (225); "3-4", 52% (71); and "greater than or equal to 5", 85% (82). The index was tested for its ability to predict risk of death from comorbid disease in the second cohort of 685 patients during a 10-yr follow-up. The percent of patients who died of comorbid disease for the different scores were: "0", 8% (588); "1", 25% (54); "2", 48% (25); "greater than or equal to 3", 59% (18). With each increased level of the comorbidity index, there were stepwise increases in the cumulative mortality attributable to comorbid disease (log rank chi 2 = 165; p less than 0.0001). In this longer follow-up, age was also a predictor of mortality (p less than 0.001). The new index performed similarly to a previous system devised by Kaplan and Feinstein. The method of classifying comorbidity provides a simple, readily applicable and valid method of estimating risk of death from comorbid disease for use in longitudinal studies. Further work in larger populations is still required to refine the approach because the number of patients with any given condition in this study was relatively small.
Article
To determine if implementation of a PPS for Medicare hospital outpatient department (HOPD) services will have distributional consequences across hospital types and regions, this analysis assesses variation in service mix and the provision of high-technology services in the HOPD. HCFA's 1990 claims file for a 5 percent random sample of Medicare beneficiaries using the HOPD was merged, by hospital provider number, with various HCFA hospital characteristic files. Hospital characteristics examined are urban/rural location, teaching status, disproportionate-share status, and bed size. Two analyses of HOPD services are presented: mix of services provided and the provision of high-technology services. The mix of services is measured by the percentage of services in each of 14 type-of-service categories (e.g., medical visits, advanced imaging services, diagnostic testing services). Technology provision is measured by the percentage of hospitals providing selected high-technology services. The findings suggest that the role hospital types play in providing HOPD services warrants consideration in establishing a PPS. HOPDs in major teaching hospitals and hospitals serving a disproportionate share of the poor play an important role in providing routine visits. HOPDs in both major and minor teaching hospitals are important providers of high-technology services. Other findings have implications for the structure of an HOPD PPS as well. First, over half of the services provided in the HOPD are laboratory tests and HOPDs may have limited control over these services since they are often for patients referred from local physician offices. Second, service mix and technology provision vary markedly among regions, suggesting the need for a transition to prospective payment. Third, the organization of service supply in a region may affect service provision in the HOPD suggesting that an HOPD PPS needs to be coordinated with payment policies in competing sites of care (e.g., ambulatory surgical centers).
Article
Subjects in observational studies of exposure effects have not been randomized to exposure groups and may therefore differ systematically with regard to variables related to exposure and/or outcome. To obtain unbiased estimates and tests of exposure effects one needs to adjust for these variables. A common method is adjustment via a parametric model incorporating all known prognostic variables. Rosenbaum and Rubin propose adjustment by the conditional exposure probability given a set of covariates which they call the propensity score. They show that, at any value of the propensity score, covariates are on average balanced between exposure groups. Thus matching on the propensity score leads to unbiased estimators and tests of exposure effect. However, the validity of the method depends on knowing the exposure probability. This quantity is usually not known in observational studies and needs to be estimated.
Article
Three diagnostic classifications for sleep disorders have been developed recently: the International Classification of Sleep Disorders (ICSD), the Diagnostic and Statistical Manual, 4th edition (DSM-IV), and the International Classification of Diseases, 10th edition (ICD-10). No data have yet been published regarding the frequency of specific diagnoses within these systems or how the diagnostic systems relate to each other. To address these issues, we examined clinical sleep disorder diagnoses (without polysomnography) in 257 patients (216 insomnia patients and 41 medical/psychiatric patients) evaluated at five sleep centers. A sleep specialist interviewed each patient and assigned clinical diagnoses using ICSD, DSM-IV and ICD-10 classifications. "Sleep disorder associated with mood disorder" was the most frequent ICSD primary diagnosis (32.3% of cases), followed by "Psychophysiological insomnia" (12.5% of cases). The most frequent DSM-IV primary diagnoses were "Insomnia related to another mental disorder" (44% of cases) and "Primary insomnia" (20.2% of cases), and the most frequent ICD-10 diagnoses were "Insomnia due to emotional causes" (61.9% of cases) and "Insomnia of organic origin" (8.9% of cases). When primary and secondary diagnoses were considered, insomnia related to psychiatric disorders was diagnosed in over 75% of patients. The more narrowly defined ICSD diagnoses nested logically within the broader DSM-IV and ICD-10 categories. We found substantial site-related differences in diagnostic patterns. These results confirm the importance of psychiatric and behavioral factors in clinicians' assessments of insomnia patients across all three diagnostic systems. ICSD and DSM-IV sleep disorder diagnoses have similar patterns of use by experienced clinicians.