Context Depression is a leading cause of disability worldwide, but treatment
rates in primary care are low.
Objective To determine the cost-effectiveness from a societal perspective of 2
quality improvement (QI) interventions to improve treatment of depression
in primary care and their effects on patient employment.
Design Group-level randomized controlled trial conducted June 1996 to July
1999.
Setting Forty-six primary care clinics in 6 community-based managed care organizations.
Participants One hundred eighty-one primary care clinicians and 1356 patients with
positive screening results for current depression.
Interventions Matched practices were randomly assigned to provide usual care (n =
443 patients) or to 1 of 2 QI interventions offering training to practice
leaders and nurses, enhanced educational and assessment resources, and either
nurses for medication follow-up (QI-meds; n = 424 patients) or trained local
psychotherapists (QI-therapy; n = 489). Practices could flexibly implement
the interventions, which did not assign type of treatment.
Main Outcome Measures Total health care costs, costs per quality-adjusted life-year (QALY),
days with depression burden, and employment over 24 months, compared between
usual care and the 2 interventions.
Results Relative to usual care, average health care costs increased $419 (11%)
in QI-meds (P = .35) and $485 (13%) in QI-therapy
(P = .28); estimated costs per QALY gained were between
$15 331 and $36 467 for QI-meds and $9478 and $21 478 for QI-therapy;
and patients had 25 (P = .19) and 47 (P = .01) fewer days with depression burden and were employed 17.9 (P = .07) and 20.9 (P = .03) more
days during the study period.
Conclusions Societal cost-effectiveness of practice-initiated QI efforts for depression
is comparable with that of accepted medical interventions. The intervention
effects on employment may be of particular interest to employers and other
stakeholders.
Depression is common among primary care patients and practice guidelines
are available, yet quality of care and outcomes remain poor.1-10
Improving quality of care for depression in primary care could potentially
increase well-being for patients, their families, and society at large.11,12
Studies suggest that practice-based interventions to improve quality
of treatment for depression in primary care can improve short-term clinical
outcomes relative to usual care, at modest cost.13-15
One such study of high utilizers found that a practice-based quality improvement
(QI) program for depression was cost-effective relative to usual care, with
longer-term improvements that included physical functioning.16
In addition, less intensive interventions, such as nurse telephone contact,
also improve clinical outcomes, but those relying on clinician training alone
may have limited benefits.17,18
Most prior studies have focused on interventions implemented under research
protocols in academically affiliated, organized care settings, so the extent
to which results can be generalized to more naturalistic practice conditions
and diverse types of practices is unclear.
Findings from Partners in Care (PIC) suggest that diverse managed primary
care practices can implement such intervention programs under naturalistic
practice conditions that include choice of treatment by patients and clinicians,
and that such interventions improved quality of care, quality of life, clinical
outcomes, and retention in employment over 1 year of follow-up.19
However, this work did not evaluate the cost-effectiveness of the practice-initiated
interventions.
The current study was designed to fill those gaps. We examine societal
cost-effectiveness over a 2-year period of implementation of the PIC interventions,
relative to usual care, in diverse managed care practices. In addition, we
examine impacts on patients' employment because of the strong interest in
this outcome among purchasers.
Experimental Design and Sample
PIC is a group-level, randomized controlled trial of practice-initiated
QI programs for depression.19 PIC was fielded
in 6 nonacademic managed care organizations. Forty-six of 48 primary care
practices (clinics) and 181 of 183 clinicians participated. Within organizations,
practices were matched into blocks of 3 clusters based on factors expected
to affect outcomes (specialty mix, patient socioeconomic and demographic factors,
and having on-site mental health specialists). Within blocks, practices were
randomized to usual care or 1 of 2 QI programs (QI-meds or QI-therapy).
Study staff screened 27 332 consecutive patients in participating
practices over a 5- to 7-month period between June 1996 and March 1997 (Figure 1). Patients were eligible if they
intended to use the practice over the next 12 months and screened positive
for depression, based on stem items from the World Health Organization's 12-month Composite International Diagnostic Interview (CIDI).20 Patients were considered positive
for depression if they reported at least 1 week of depression in the last
30 days, plus 2 or more weeks of depressed mood or loss of interest in pleasurable
activities in the last year or persistent depression over the year. This indicator
has a 55% positive predictive value for 12-month major depressive or dysthymic
disorder by the full CIDI.9
Patients were ineligible if younger than 18 years, not fluent in English or
Spanish, or lacking in insurance coverage for intervention therapists.
Of those completing the screener, 3918 were eligible, 2417 were available
to confirm insurance eligibility, and 241 were ineligible. Of those reading
the informed consent, 1356 (70%) enrolled: 443 in usual care, 424 in QI-meds,
and 489 in QI-therapy practices. The study was approved by the institutional
review boards of RAND and the practices.
Intervention design and implementation are described elsewhere, and
all QI materials are available from RAND (http://www.rand.org/organization/health/pic.products/order.html).19,21 Prior to implementation,
practices committed to implementing the programs and the study provided a
payment of up to half of the estimated practice participation costs ($35 000-$70 000).
Usual care clinics received depression practice guidelines by mail. The interventions
provided practices with training and resources to initiate and monitor QI
programs according to local practice goals and resources. Patients and clinicians
retained choice of treatment, and their use of intervention resources was
optional. In effect, the study served as an external disease management firm,
designing the materials, hosting initial training, and offering limited support
during implementation.
For both QI-meds and QI-therapy interventions, local practice teams
were trained in a 2-day workshop to provide clinician education through lectures,
academic detailing, or audit and feedback, and to supervise intervention staff
and conduct team oversight. Practice nurses were trained as depression specialists,
following a written protocol, to assist in initial patient assessment, education,
and motivation for treatment. Practice teams were given patient education
pamphlets and videotapes, patient tracking forms, and clinician manuals and
pocket reminder cards and were encouraged to distribute them. The materials
described guideline-concordant care for depression and presented psychotherapy
and antidepressant medication as equally effective.
In the QI-meds program, nurse specialists were trained to support medication
adherence through monthly telephone contacts or visits for 6 or 12 months,
randomized at the patient level. In the QI-therapy program, practice therapists
were trained to provide individual and group cognitive behavioral therapy,
following a protocol developed at the San Francisco General Hospital Depression
Clinic.22,23 This therapy was
available at the primary care co-payment rate (usually $5-$10) for 6 months
after enrollment. All patients could have other therapy at the usual co-payment
rate. Clinical supervision was provided by local experts assisted by study
experts in cognitive behavioral therapy. In all conditions, patients could
have medication, therapy, or both. However, the extra QI-meds or QI-therapy
resources made it easier to obtain appropriate medication or therapy, respectively.
Patients were asked to complete the screener, the telephone CIDI for depression, a detailed economic survey by telephone at baseline
and 24 months of follow-up, and mail surveys at baseline and 6, 12, 18, and
24 months. Completion rates were 95% and 85% for the baseline and 24-month
telephone surveys and 90%, 86%, 84%, 83%, and 85%, respectively, for the mail
surveys. Subjects were eligible for follow-up unless they disenrolled from
the study. Claims and encounter data from the practices were only consistently
available for the first 6 months of patient follow-up, and most practices
did not have pharmacy data.
Intervention Costs. These included screening, intervention materials, initial nurse specialist
assessments, and 20 minutes of supervision of nurses and therapists per enrolled
patient. We assigned costs to intervention activities based on data from the
practices about the average cost of clinic staff. Research-specific costs
were excluded. For main analyses, we assumed that follow-up visits to intervention
staff were included in patient reports of outpatient visits. In sensitivity
analyses, we used data from intervention logs to include such visits as intervention
costs (which double counts them if they were also reported by patients directly);
our results did not change substantively.
Health Care Costs. We assigned costs to patient-reported counts of emergency department
visits, medical and mental health visits, and psychotropic medications used,
for each follow-up. Patient report was selected due to the limitations in
the available claims and encounter data. In addition, the number of outpatient
visits was higher for patient surveys than claims data over the first 6 months,
probably due to out-of-practice use or incomplete claims data. We excluded
inpatient costs because we did not expect or observe intervention effects
on them and had limited precision to analyze them.24
Average costs in 1998 dollars were assigned to each component of patient-reported
health care use using a national database of about 1.8 million privately insured
individuals (provided by Ingenix, a benefits consulting firm in New Haven,
Conn). The Ingenix data included information on provider reimbursements (ie,
patient and plan payments, plus coordination of benefits), which we used as
a proxy for health care costs. Specifically, we calculated the mean cost per
outpatient medical visit ($46), mental health visits ($96), and emergency
department visits ($450), respectively, for adults in the Ingenix data; these
costs include facility charges, professional fees, and ancillary services
associated with the visits, as applicable. We then multiplied the visit counts
reported by PIC patients by these mean costs.
For psychotropic medications, we matched patient-reported data of medication
names, daily dosages, and months of use to average costs for that combination
from the Ingenix data, pooling data on generic and brand names for the same
medication according to their relative proportion in the Ingenix data and
summing all medications used (for reference, 20 mg of fluoxetine cost $2.20
per pill, on average).
Indirect costs of treatment include patient time costs for obtaining
health care.25 We assumed an average time of
30 and 45 minutes for outpatient medical and mental health visits, respectively,
and added average travel and waiting times reported by patients at baseline.
In addition, we assumed 3 hours for emergency department visits and 1.5 hours
to fill prescriptions in a month of use. We priced patients' time using reported
hourly wage at baseline and sex-specific mean wage for those not working at
baseline (this may slightly overstate the value of time for nonworking patients).
Quality-Adjusted Life-Years. To measure quality-adjusted life-years (QALYs), we calculated a health
utility index from the Short-Form, 12-Item Health Survey (SF-12) items collapsed
into 6 health states that had been identified through cluster analysis of
SF-12 physical and mental component scores.26,27
Utility weights from this index were derived from a convenience sample of
primary care patients with symptoms of depression using a standard gamble
approach.26 QALY weights were calculated for
each 6-month follow-up time period, and we analyzed patterns over time. We
call this measure "QALY-SF."
In addition, following an approach developed by Lave et al,28 we developed a measure of depression-burden days
and assigned utility scores from the literature to estimate QALYs.28,29 We call this measure "QALY-DB." Specifically,
for each survey from baseline through 24 months, we developed a count of positive
scores on the following 3 measures: probable major depressive disorder, based
on a repeat of the baseline screener19; significant
depressive symptoms, based on a modified Center for Epidemiologic Studies
Depression Studies (CES-D) scale19,30;
and poor mental health-related quality of life (HRQOL), based on being more
than 1 SD below the population mean on the mental health subscale of the SF-12.3 We averaged the count for the beginning and end of
each 6-month follow-up period and multiplied by 182 to estimate depression-burden
days during the period. We summed across periods to get the 24-month total.
We then used findings from the literature that a year of depression is associated
with losses of 0.2 to 0.4 QALYs to convert the intervention effect on depression-burden
days into the QALY-DB estimates.9,31-34
Employment. We created a measure of days worked in each 6-month follow-up by taking
the average of employment status at the start and end of each period and multiplying
by 116 (the number of workdays in 6 months). We summed across periods to calculate
the 24-month total. We also examined days missed from work due to illness,
which patients reported for the 4 weeks preceding each follow-up survey.
All multivariate models controlled for baseline measures of patient
age, sex, marital status, education, rank in the distribution of household
wealth, employment status, ethnicity, medical comorbidity, depressive disorder
status, the SF-12 aggregate HRQOL measures, presence of comorbid anxiety disorder,
and practice randomization block.
To estimate the effects of practice-initiated QI on patients, we conducted
patient-level intent-to-treat analyses, controlling for baseline patient differences
that could remain after group-level randomization. We examined intervention
effects on health care costs using 2-part models, due to the skewed distribution
of costs. The first is the probability of positive costs, using logistic regression.
The second is the log of costs given any, using ordinary least squares.35 We used the smearing estimate for retransformation,
applying separate factors for each intervention group to ensure consistent
estimates.36,37 We did not adjust
cost models for clustering by clinic because we know of no existing software
to do so for 2-part models. We expected the interventions to increase health
care costs, relative to usual care; not accounting for clustering is thus
conservative from a policy perspective, since not adjusting for clustering
is likely to overstate the statistical significance of cost differences.
For the QALY-SF measure, we specified 3-level (repeated measurements
nested within patients, and patients nested within clinics) mixed effects
linear time-trend regression models, controlling for the baseline utility
value in addition to the covariates listed above (except HRQOL). We calculated
the area under the curve to derive values over 24 months. For days of depression
burden and employment, respectively, we specified 2-level (patients nested
within clinics) mixed effects linear regression models, to account for patient
clustering at the practice level. For these outcomes, we examined the 24-month
value directly.
Significance of comparisons across intervention groups is based on the
regression coefficients. We illustrate average intervention effects relative
to usual care, adjusted for patient characteristics using a direct method,
ie, standardized predictions generated from each regression model. Specifically,
we used the regression parameters and each individual's actual values for
all covariates other than intervention status to calculate the predicted outcome
assuming the patient had been assigned to usual care or to either intervention,
respectively. We then calculated the mean prediction under each scenario.
We analyzed patients completing at least 1 follow-up (92% of the enrolled
sample; N = 1248 total [422 in usual care, 393 in QI-meds, and 433 in QI-therapy]).
The data are weighted for the probability of study enrollment and follow-up
response to the characteristics of the eligible sample. We used multiple imputations
for missing items at each wave.38,39
For outcomes, we averaged predictions from 5 randomly imputed data sets and
adjusted SEs for uncertainty due to imputation.39,40
Because many tests are in the same direction as hypothesized, a formal
Bonferroni correction for multiple statistical comparisons is too conservative,
so we report actual P values and interpret results
with multiple comparisons in mind.41
At baseline, the percentage of patients with a college education was
lower for usual care (15.0%) than QI-meds (22.9%, P
= .004) or QI-therapy (21.5%, P = .01). Usual care
patients were more likely to have current symptoms with lifetime disorder
(26.1%) compared with QI-meds (18.5%, P = .02) and
QI-therapy (19.4%, P = .03) vs current disorder or
symptoms but no lifetime disorder. QI-therapy patients were slightly older
(by 3 years on average) (P = .02). Intervention and
control patients did not differ with respect to the other baseline covariates
listed above at P = .05.
Table 1 reports average
per patient costs and outcomes over 24 months (including patient time costs,
but not inpatient care and nonpsychotropic medications). Average total costs
for usual care patients were estimated to be $3835, increasing by $419 (11%)
among QI-meds participants and by $485 (13%) among QI-therapy participants.
Neither intervention effect on total costs is statistically significant. Patient
time costs represented 22% of the average total cost under usual care. Increases
in time costs represented 3% of the incremental increase due to QI-meds and
25% of the incremental increase due to QI-therapy. The intervention effects
on patient time costs were also statistically nonsignificant (details available
from authors on request).
For the QALY-SF measure, the incremental increase due to QI-meds was
0.0115 QALYs over 24 months (P = .15), while the
increase due to QI-therapy was 0.0226 (P = .006).
Combining these point estimates with our point estimates of the incremental
intervention costs yields an estimated cost per QALY of $36 467 for QI-meds
and $21 478 for QI-therapy.
For the QALY-DB measure, we assumed that depression reduces the value
of a life-year by 0.2 to 0.4 QALYs. 9,28,31-34
Compared with usual care, QI-meds reduced the number of depression-burden
days by 25 (P = .19), or 0.0137 to 0.0274 QALYs.
QI-therapy yielded 47 fewer depression-burden days over 24 months (P = .01), or 0.0258 to 0.0515 QALYs. These point estimates yield a
cost-per-QALY range of $15 331 to $30 663 for QI-meds and $9478
to $18 953 for QI-therapy.
As shown in Table 1, participants
from QI-meds clinics had 17.9 more employed days relative to usual care over
24 months (P = .07), on average, and QI-therapy participants
had 20.9 more employed days (P = .03). Intervention
and usual care patients who were working did not differ substantively or statistically
with respect to sick days at any follow-up period. For instance, intervention
patients were significantly more likely to be working at the 12-month follow-up
survey (65.7% in QI vs 60.8% in usual care; 95% confidence interval [CI] for
difference: 0.01-0.09; P = .03). Among employed patients,
however, the number of reported sick days in the previous 4 weeks was virtually
identical (1.2 days in QI vs 1.1 in usual care; 95% CI for difference: −
0.5 to 0.6; P = .81). Results for other follow-up
periods were similar.
We found that practice-initiated, locally implemented programs that
encourage guideline-concordant care for depression can substantially reduce
the individual suffering and economic consequences of depression. The point
estimates for incremental costs per QALY relative to usual care were within
the range of many accepted medical interventions25,42
and substantially below the estimated value of a year of life.25,43
Our data suggested that QI-therapy may have a better overall value in terms
of cost per QALY than QI-meds, suggesting that there may be particular value
to improving access to structured psychotherapy such as cognitive behavioral
therapy for depressed primary care patients.
We found significant intervention effects on patients' labor supply,
on the order of 1 additional month of employment over 2 years. In addition
to its importance to patients, this result could suggest broader economic
benefits of the intervention to families and purchasers—benefits that
may not be fully captured in standard measures of QALYs, which are based on
patients' HRQOL. If so, the true societal cost-effectiveness of the interventions
may be more favorable than what we report.25
This study has important limitations. For instance, we studied 6 practice
networks, although they were chosen to be diverse. We relied on patient self-report
for most measures, which would bias intent-to-treat analyses if the interventions
affected patient reports. We had a relatively low enrollment rate, which we
partially account for by weighting back to the eligible population. Despite
a large sample size relative to most clinical trials, our cost estimates lacked
precision.
Our findings suggest that practice-initiated interventions to improve
quality of care for depression can substantially increase patients' and societal
welfare, even when implemented locally and under flexible, naturalistic practice
conditions that support patient and clinician treatment choices. These interventions
increase costs for clinicians and insurers, suggesting that their widespread
adoption may require increases in consumer demand or public policy initiatives
that provide incentives for implementing them. But the gains observed in such
an applied and real-world context suggest that improved medical care has much
to offer depressed patients and their families and communities if we can create
the conditions necessary to put such programs in place.
1.Katon W, Schulberg HC. Epidemiology of depression in primary care.
Gen Hosp Psychiatry.1992;14:237-247.Google Scholar 2.Spitzer R, Williams JBW, Kroenke K.
et al. Utility of a new procedure for diagnosing mental disorders in primary
care: the Prime-MD 1000 study.
JAMA.1994;272:1749-1756.Google Scholar 3.Wells K, Stewart A, Hays R.
et al. The functioning and well-being of depressed patients: results from
the Medical Outcomes Study.
JAMA.1989;262:914-919.Google Scholar 4.Simon G, VonKorff M. Recognition, management, and outcomes of depression in primary care.
Arch Fam Med.1995;4:99-105.Google Scholar 5.Simon GE, Unützer J. Health care utilization and costs among patients treated for bipolar
disorder in an insured population.
Psychiatr Serv.1999;50:1303-1308.Google Scholar 6.Henk HJ, Katzelnick DJ, Kobak KA.
et al. Medical costs attributed to depression among patients with a history
of medical expenses in a health maintenance organization.
Arch Gen Psychiatry.1996;53:899-904.Google Scholar 7.Unützer J, Patrick D, Diehr P.
et al. Quality adjusted life years in older adults with depressive symptoms
and chronic medical disorders.
Int Psychogeriatr.2000;12:15-33.Google Scholar 8.Wells KB, Sturm R, Sherbourne CD.
et al. Caring for Depression. Cambridge, Mass: Harvard University Press; 1996.
9.Wells K, Sherbourne C. Functioning and utility for current health of patients with depression
or chronic medical conditions in managed, primary care practices.
Arch Gen Psychiatry.1999;56:897-904.Google Scholar 10.Young AS, Klap R, Sherbourne CD.
et al. The quality of care for depressive and anxiety disorders in the United
States.
Arch Gen Psychiatry.2001;58:55-61.Google Scholar 11.Murray CJ, Lopez AD. The Global Burden of Disease: A Comprehensive Assessment
of Mortality and Disability from Disease, Injuries, and Risk Factors in 1990
and Projected to 2020. Boston, Mass: The Harvard School of Public Health on behalf of the
World Health Organization and The World Bank; 1996.
12.Sturm R, Wells KB. How can care for depression become more cost-effective?
JAMA.1995;273:51-58.Google Scholar 13.Katon W, Von Korff M, Lin E.
et al. Collaborative management to achieve treatment guidelines: impact on
depression in primary care.
JAMA.1995;273:1026-1031.Google Scholar 14.Katon W, Robinson P, Von Korff M.
et al. A multifaceted intervention to improve treatment of depression in primary
care.
Arch Gen Psychiatry.1996;53:924-932.Google Scholar 15.Von Korff M, Katon W, Bush T.
et al. Treatment costs, cost offset and cost-effectiveness of collaborative
management of depression.
Psychosom Med.1998;60:143-149.Google Scholar 16.Simon GE, Manning WG, Katzelnick DJ.
et al. Cost-effectiveness of systematic depression treatment for high utilizers
of general medical care.
Arch Gen Psychiatry.2001;58:181-187.Google Scholar 17.Thompson C, Kinmonth AL, Stevens L.
et al. Effects of a clinical practice guideline and practice-based education
on detection and outcome of depression in primary care: Hampshire-Depression
Project randomized controlled trial.
Lancet.2000;355:185-191.Google Scholar 18.Hunkeler EM, Meresman J, Hargreaves WA.
et al. Efficacy of nurse telehealth care and peer support in augmenting treatment
of depression in primary care.
Arch Fam Med.2000;9:700-708.Google Scholar 19.Wells KB, Sherbourne C, Schoenbaum M.
et al. Impact of disseminating quality improvement programs for depression
in managed primary care: a randomized controlled trial.
JAMA.2000;283:212-220.Google Scholar 20.World Health Organization. Composite International Diagnostic Interview (CIDI)
Core Version 2.1 Interviewer's Manual. Geneva, Switzerland: World Health Organization; 1997.
21.Rubenstein LV, Jackson-Triche M, Unützer J.
et al. Evidence-based interventions for depression in managed primary care
practices.
Health Aff (Millwood).1999;18:89-105.Google Scholar 22.Muñoz RF, Miranda J. Group Therapy for Cognitive Behavioral Treatment
of Depression, San Francisco General Hospital Depression Clinic, 1986. Santa Monica, Calif: RAND; 2000. Document MR-1198/4.
23.Muñoz RF, Aguilar-Gaxiola S, Guzmán J. Manual de Terapia de Grupo para el Tratamiento Cognitivo-conductal
de Depresión, Hospital General de San Francisco, Clínica de
Depresión, 1986. Santa Monica, Calif: RAND; 2000. Document MR-1198/5.
24.Sturm R, Unützer J, Katon W. Effectiveness research and implications for study design: sample size
and statistical power.
Gen Hosp Psychiatry.1999;21:274-283.Google Scholar 25.Gold M, Siegel J, Russell L.
et al. Cost-effectiveness in Health and Medicine. New York, NY: Oxford University Press; 1996.
26.Lenert LA, Sherbourne CD, Sugar C.
et al. Estimation of utilities for the effects of depression from the SF-12.
Med Care.2000;38:763-770.Google Scholar 27.Sugar CA, Sturm R, Lee TT.
et al. Empirically defined health states for depression from the SF-12.
Health Serv Res.1998;33:911-928.Google Scholar 28.Lave JR, Frank RG, Schulberg HC.
et al. Cost-effectiveness of treatments for major depression in primary care
practice.
Arch Gen Psychiatry.1998;55:645-651.Google Scholar 29.Schulberg HC, Block MR, Madonia MJ.
et al. Treating major depression in primary care practice: 8-month clinical
outcomes.
Arch Gen Psychiatry.1996;53:913-919.Google Scholar 30.Radloff LS. The CES-D Scale: a self-report depression scale for research in the
general population.
Appl Psychol Measurement.1977;1:385-401.Google Scholar 31.Revicki D, Wood M. Patient-assigned health state utilities for depression-related outcomes:
differences by depression severity and antidepressant medications.
J Affect Disord.1998;48:25-36.Google Scholar 32.Fryback D, Dasbach E, Klein R.
et al. The Beaver Dam Health Outcomes Study: initial catalog of health state
quality factors.
Med Decis Making.1993;13:89-102.Google Scholar 33.Pyne J, Patterson T, Kaplan R.
et al. Preliminary longitudinal assessment of quality of life in patient with
major depression.
Psychopharmacol Bull.1997;33:23-29.Google Scholar 34.Unützer J, Patrick DL, Simon G.
et al. Depressive symptoms and the cost of health services in HMO patients
aged 65 and older: a 4-year prospective study.
JAMA.1997;277:1618-1623.Google Scholar 35.Duan N, Manning WG, Morris CN.
et al. A comparison of alternative models for the demand of medical care.
J Bus Econ Stat.1983;1:115-126.Google Scholar 36.Duan N. Smearing estimate: a nonparametric retransformation method.
J Am Stat Assoc.1983;78:605-610.Google Scholar 37.Manning WG. The logged dependent variable, heteroscedasticity, and the retransformation
problem.
J Health Econ.1998;17:283-295.Google Scholar 38.Little RJA. Pattern-mixture models for multivariate incomplete data.
J Am Stat Assoc.1993;88:125-134.Google Scholar 39.Little RJA, Rubin DB. Statistical Analysis With Missing Data. New York, NY: John Wiley & Sons; 1987.
40.Schafer JL. Analysis of Incomplete Multivariate Data. London, England: Chapman & Hall; 1997.
41.Miller Jr RG. Simultaneous Statistical Inference. 2nd ed. New York, NY: Springer-Verlag; 1981.
42.Tengs TO, Adams ME, Pliskin JS.
et al. Five-hundred life-saving interventions and their cost effectiveness.
Risk Analysis.1995;15:369-390.Google Scholar 43.Keeler EB. The value of remaining life is close to estimated values of life.
J Health Econ.2001;20:141-143.Google Scholar