Improving depression outcomes in older adults with comorbid
medical illness
Linda H.Harpole,M.D.,M.P.H.a,*,John W.Williams Jr.,M.D.,M.H.S.a ,b ,
Maren K.Oln,Ph.D.b ,c ,Karen M.Stechuchak,M.S.b ,Eugene Oddone,M.D.,M.H.S.a ,b ,Christopher M.Callahan,M.D.d ,Wayne J.Katon,M.D.e ,f ,Elizabeth H.Lin,M.D.,M.P.H.f ,
Lydia M.Grypma,M.D.g ,J q rgen Un q tzer,M.D.,M.P.H.e
a
Department of Medicine,Duke University Medical Center,Durham,NC 27709,USA
b
Center for Health Services Rearch in Primary Care,Durham Veterans Affairs Medical Center,Durham,NC 27705,USA
c
Department of Biostatistics and Bioinformatics,Duke University Medical Center,Durham,NC 27705,USA d
Indiana University Center for Aging Rearch,Regenstrief Institute Inc.,Indianapolis,IN 46202,USA
e
Department of Psychiatry,University of Washington,Seattle,WA 98195,USA f
Center for Health Studies,Group Health Cooperative,Seattle,WA 98101,USA g
Kair Permanente of Southern California,San Diego,CA 92120,USA
Received 19April 2004;accepted 8September 2004
Abstract
street是什么意思Background:Depression is common in older adults and often coexists with multiple chronic dias,which may complicate its diagnosis and treatment.
Objective:To determine whether or not the prence of multiple comorbid medical illness affects p
atient respon to a multidisciplinary depression treatment program.
Design,Setting and Participants:Preplanned analys of Improving Mood-Promoting Access to Collaborative Treatment (IMPACT),a randomized controlled trial of 1801depresd older adults (z 60years),which was performed at 18primary care clinics from eight health care organizations in five states across the United States from July 1999to August 2001.
Intervention:Intervention patients had access for up to 12months to a depression care manager,supervid by a psychiatrist and a primary care expert,who offered education,care management and support of antidepressant management by the patient’s primary care physician,or provided brief psychotherapy (Problem-Solving Treatment in Primary Care).
Measurements:Depression,quality of life (QOL;scale of 0–10)and mental health component score (MCS)of the Short-Form 12assd at baline,3,6and 12months.
Results:Patients suffered from an average of 3.8chronic medical conditions.Although patients with more chronic medical conditions had higher depression verity at baline,the number of chronic dias did not affect the likelihood of respon to the IMPACT intervention when compared to care as usual.Intervention patients experienced significantly lower depression during all follow-up time poi
nts as compared with patients in usual care independent of other comorbid illness (P b .001).Intervention patients were also more likely to experience substantial respon (at least a 50%reduction in depressive symptoms)regardless of the number of comorbidities,to experience improved MCS-12scores at 3and 12months,and to experience improved QOL.
Conclusions:The prence of multiple comorbid medical illness did not affect patient respon to a multidisciplinary depression treatment program.The IMPACT collaborative care model was equally effective for depresd older adults with or without comorbid medical illness.Published by Elvier Inc.
Keywords:Depression;Comorbidity;Primary care
0163-8343/$–e front matter.Published by Elvier Inc.doi:10.hosppsych.2004.09.004
*Corresponding author.Tel.:+19194837434;fax:+19193150984.E-mail address:Linda.h. (L.H.Harpole).Psychiatry and Primary Care
Recent epidemiologic studies have found that most patients with mental illness are en exclusively in primary care medicine.The patients often prent with medically unexplained somatic symptom
s and utilize at least twice as many health care visits as controls.There has been an exponential growth in studies in this interface between primary care and psychiatry in the last 10years.This special ction,edited by Wayne J.Katon,MD.,will publish informative rearch articles that address primary care-psychiatric issues.
General Hospital Psychiatry 27(2005)4–
12
1.Introduction
Major depression and dysthymia affect between5%and 10%of older adults who are en in the outpatient tting [1–3].As compared with younger patients,older adults often suffer from multiple coexisting chronic dias, which can complicate the diagnosis and treatment of depression in the elderly.The adver effects of depression and chronic medical conditions on functioning are additive [4]and are associated with incread health care costs[5].
It is well established that depression adverly affects outcomes from chronic medical conditions.Patients with coronary artery dia and comorbid depression have functional disability[
6],poorer outcomes following coronary artery bypass surgery[7],a wor prognosis following an episode of unstable angina[8]and incread mortality[9]as compared to tho without depression.The biologic rationale for incread vulnerability of depresd patients with coronary artery dia is thought to be a manifestation of hypothalalmic–pituitary–adrenocortical axis hyperactivity, decread heart rate variability and changes in platelet receptor function[10].Depression is also associated with wor glycemic control in diabetics[11]and is a risk factor for the development of stroke[12].
Although prior rearch had suggested that treatment of depression in late life is not very effective[13],recent studies have demonstrated that depression symptomatology can be successfully treated in older adults[14,15].How effective treatment is in older patients with medical comor-bid illness is not clear.Many trials of antidepressants in psychiatric ttings have explicitly excluded patients with significant medical and psychiatry comorbidity[16],or accepted tho with only stable chronic dia[14].The few studies,which have included patients with comorbid medical illness,have suggested that treatment for depression may be less effective than for tho without comorbid medical illness[17–21].
A recent study[22,23]of a dia management model for the treatment of depression in1801older a
dults in the primary care tting was successful at improving outcomes. Becau enrollment criteria did not exclude patients with multiple medical illness,we have a unique opportunity to determine whether or not comorbid medical illness modu-lates the effectiveness of depression treatment in this intervention trial.To address the effect of comorbid medical illness on patient outcomes,we analyzed data from the Improving Mood-Promoting Access to Collaborative Treat-ment(IMPACT)trial of collaborative care management of depression to asss if increasing comorbid medical illness negatively affects respon to a successful depression intervention.
2.Methods
The IMPACT study was conducted at18participating primary care clinics belonging to ven study sites across the United States.At one site,two different healthcare organizations participated in the study,resulting in a total of eight different health care organizations.Participating organizations include two staff-model health maintenance organizations(HMOs),two regions of a large group-model HMO,the Department of Veterans Affairs,two university-affiliated primary care systems and one private practice physician group.Each institution’s review board approved the study procedures and all participants gave written informed connt.A total of1801participants were enrolled in the study.A detailed description of the sites,recruitment procedures,overall sample and intervention outc
飞行器动力工程omes are described in early publications[22,23].
Study inclusion criteria include age of60years or older, current major depression or dysthymia diagnod by a Structured Clinical Interview bad upon the Diagnostic and Statistical Manual of Mental Disorders,Fourth Edition (SCID)[24,25],and plans to u one participating primary care clinic as the main source of general medical rvices for the coming year.Exclusion criteria include history of bipolar disorder or psychosis,ongoing treatment by a psychiatrist,current drinking problems[26],vere cogni-tive impairment[27]or acute risk of suicide.Baline,3-,6-and12-month data were utilized for the preplanned analys described in this paper.
2.1.Intervention
Intervention participants received depression care man-agement by a depression clinical specialist(DCS)who was either a nur or psychologist working in the primary care tting,who collaborated with the patient,his primary care physician,a liaison primary care expert and a psychiatrist. The DCS conducted a psychosocial history,provided education and behavioral activation and helped patients identify treatment preferences.Treatment options included antidepressant medications prescribed by the patients’primary care provider,or six to eight ssions
of psycho-therapy designed for primary care,namely,Problem-Solving Treatment[28–32],which was delivered by the DCS.A stepped-care algorithm was utilized to guide treatment[23].The DCS met weekly with the supervising psychiatrist and the liaison primary care physician to monitor progress and adjust treatment plans as needed. Patients had either weekly or biweekly contact with the DCS during the acute treatment pha and less frequently (monthly)once symptoms remitted.Contacts were either in person or by telephone.
2.2.Data collection
Baline interviews were conducted by trained lay interviewers using structured computerized interviews. Follow-up interviews were conducted by a trained telephone survey rearch group using computer-assisted telephone interviews.All interviewers were blind to study assignment. Survey respon rates were90%at3months,87%at6 months,and83%at12months.
L.H.Harpole et al./General Hospital Psychiatry27(2005)4–125
Baline interviews captured sociodemographic charac-teristics,verity of depressive symptoms via the mean score of the 20depression items from the Symptom Checklist-90(SCL-20)[33],SCID [24,25]diagnosis of major depression or dysthymia,overall quality of life (QOL)on a 0–10scale and th
e mental component score (MCS)of the Short-Form 12(SF12)[34].The component scores range from 0to 100with lower scores indicating poorer functional status.Cognitive impairment was assd using a six-item cognitive screener derived from the minimental status examination [27].The prence of panic attacks in the past 4weeks was measured [35]as was neuroticism with a subscale of the NEO [36].Additional questions were asked about the u of antidepressant medication,counling or psychotherapy in the 3months prior to enrollment.
A history of diagnosis or treatment for common chronic medical problems over the prior 3years was determined by the baline survey.Conditions were collapd into 11general categories that were lected to reprent the most common or significant chronic medical conditions in older adults.The included asthma,emphyma or chronic bronchitis (chronic lung dia);high blood pressure or hypertension (hypertension);high blood sugar or diabetes (diabetes);arthritis or rheumatism (arthritis);loss of hearing or vision (nsory deficit);cancer —excluding skin cancer (cancer);neurological condition such as epilepsy,izures,Parkinson’s dia or stroke (neurological dia);heart dia such as angina,heart failure or valve problems
(cardiac dia);chronic back problems,headache or other chronic pain problems (chronic pain);stomach ulcer,chronic inflamed bowel,enteritis or colitis (gastrointestinal dia);chronic probl
ems with urination,chronic bladder infections or prostate problems,incontinence or inability to hold your urine (urologic).Skin cancer was excluded from the total count due to presumptive lack of chronicity.A summary score was created for each participant reprenting the total number of chronic dias.
Additionally,we calculated a chronic dia score (CDS)[37,38],which is bad upon the number of prescribed medications.This score increas with the number of different chronic dias inferred from the subject’s medication profile.In this study,we utilized two different weighting systems,one that predicted health care costs and one that predicted mortality [39].The two weighting systems of the CDS esntially produced the same result as the simple dia count;therefore,for ea of interpretation and adaptability to clinical scenarios,we cho to prent the results utilizing simple dia counts.2.3.Outcomes examined
In this study,we examined five different outcomes.The continuous outcomes included the SCL-20,the MCS,and the QOL.All were assd at baline,3,6and 12months.The dichotomous outcomes included complete remission of depressive symptoms (SCL score b 0.5)[40]and treatment respon (z 50%decrea in SCL-20from baline).Remission was assd at all four time points while
Table 1
Patient characteristics Sample characteristics
All (N =1801)Usual care (n =895)Intervention (n =906)Group test P value Female
1168(64.9)587(65.6)581(64.1).52Mean (S.D.)age,years 71.2(7.5)71.4(7.6)71.0(7.4).33Marital status
.23
Married or living with partner 834(46.3)432(48.3)401(44.3)Divorced/parated/never married 521(28.9)248(27.7)273(30.1)Widowed
446(24.8)215(24.0)232(25.6)
Ethnic background .16
Caucasian/White
1388(77.1)679(75.9)709(78.2)African American/Black 222(12.3)108(12.1)114(12.6)Other 191(10.6)107(12.0)83(9.2)
Education
.34
Less than high school graduate 347(19.2)170(18.9)177(19.5)High school graduate or GED 408(22.7)209(23.4)199(22.0)Some college
637(35.4)327(36.6)309(34.2)College graduate/graduate degree 409
(22.7)
whitehor189(21.1)221
(24.3)
Depression status (SCID diagnosis).35
Major depression 306(17.0)146(16.3)160(17.7)Dysthymia
542(30.1)283(31.6)259(28.6)Major depression and dysthymia
953(52.9)466(52.1)487(53.7)Two or more prior episodes of depression
1274(70.7)632(70.6)642(70.9).90Mean (S.D.)SCL-20depression score (range,0–4) 1.7(0.6) 1.7(0.6) 1.7(0.6).75Positive result on anxiety screener 518(28.7)260(29.0)258(28.5).79Mean (S.D.)NEO neuroticism scalecorner怎么读
22.5(5.2)22.5(5.3)22.5(5.2).88Mean (S.D.)physical component score (PCS-12)40.3(7.4)40.1(7.4)40.4(7.4).35Any depression treatment in lifetime
1189(66.0)
577
(64.5)
611(67.5)
.19
Values are N (%)unless otherwi indicated.
The numbers prented in this table are bad on the multiple imputed data ts and are subject to rounding discrepancies.
L.H.Harpole et al./General Hospital Psychiatry 27(2005)4–12
6
respon is only examined at the three follow-up time points becau it reflects a change at each point from baline. 2.4.Analys
We conducted t tests or chi-square tests to compare demographic and clinical characteristics of intervention and usual-care patients at baline(Table1)and to compare frequency of chronic medical illness between the groups (Table2).Analysis of variance was ud to test the association between the numbers of medical comorbidities and depression status by diagnosis.For all continuous outcome variables,we fit linear mixed-effects models(PROC MIXED in SAS)including fixed and random effects.In all of the mixed-effects models,we treated time as a categorical variable,and we ud an unstructured covariance to account for the within-subject correlation over time.Unadjusted models simply included the fixed effects of time,treatment group,number of chronic dias as well as all two-way and the three-way interaction term.We ud predicted values from the model to examine the relationship between treatment group,number of chronic dias and time.We plotted expected SCL-20,MCS-12and QOL trajectories for patients with either two or five chronic dias,a
s the reprented the first and third quartiles of this variable. Adjusted models included a number of other covariates of interest,including site,age,race,gender,education,marital status,type of depression,anxiety,NEO neuroticism score, physical component score of the SF12(PCS-12)and any depression treatment in lifetime.For each of the covariates we explored their interactions with time and included tho higher-level terms if they were statistically significant.
For the two dichotomous outcomes,we utilized a population-average generalized estimating equation(GEE) model,with an unstructured covariance matrix.Similar to the mixed-effects models,we ran both unadjusted and adjusted models.Adjusted models included the baline effects of age, race,gender,education,marital status,type of depression, anxiety,NEO neuroticism score,PCS-12and any depression treatment in lifetime.All GEE models were fit using PROC GENMOD in SAS version8.2(SAS Institute,Cary,NC).
In this study,missing data occurred at both the item and subject level.We ud an extended hot deck multiple imputation technique that modifies the predictive mean matching method to impute item-level missing data[41–43].Rates of item-level missing data were less than2%for all variables discusd in this paper.Although there were no significant differences in the completion rate of follow-up interviews between the intervention and usual-care groups, we found somewhat different p
redictors of follow-up respon in intervention and usual-care patients.We ud an approximate Bayesian bootstrap multiple imputation method[44]to impute subject-level missing data at baline and each follow-up.Imputations were conducted parately in the intervention and usual-care groups.Using Rubin’s[43] rules,the results across five imputed data ts were combined by averaging,and standard errors were adjusted to reflect both within-imputation variability and between-imputation variability.
3.Results
The enrolled sample of1801patients was ethnically and clinically diver(Table1).The majority of patients had suffered from two or more prior episodes of depression in the past and had received depression treatment in the past. Evaluation of intervention and control groups demonstrated no statistically significant difference between the two groups at baline.On average,patients suffered from3.8chronic medical conditions,in addition to depression(Table2). The average number comorbidities for tho with major depression only,dysthymia only and major depression and dysthymia was3.4,3.6and3.9,respectively(P b.0001). More than50%of patients suffered from high blood pressure, arthritis,loss of hearing or vision,or chronic pain.Although the overall mean number of comorbidities was equivalent between intervention and control patients,intervention patients were
韩国婚礼歌曲more likely to suffer from lung dia in the past3years.
Table2
Self-report of chronic medical conditions treated or diagnod within past
3years
responsible的用法Usual care a Intervention a P value b
Mean number of
comorbidities c(S.D.)
3.8(1.9) 3.7(1.9).65
Asthma,emphyma
or chronic bronchitis
188(21.0)232(25.6).02
High blood pressure
or hypertension
516(57.6)527(58.2).82
High blood sugar
or diabetes
214(23.9)204(22.5).50
Arthritis or rheumatism495(55.4)506(55.8).83
Loss of hearing or vision511(57.1)484(53.4).12
Cancer diagnod
or treated in the last3years
(excluding skin cancer)
47(5.3)58(6.4).31
A neurological condition
such as epilepsy,izures,
Parkinson’s dia or stroke
67(7.5)84(9.3).18
Heart dia such as angina,filled
heart failure or valve problems
261(29.1)236(26.0).15
Chronic back problems,
headache or other chronic
pain problems
517(57.7)506(55.9).43
Stomach ulcer,
chronic inflamed bowel,
enteritis or colitis
197(22.0)180(19.9).28
Chronic problems with urination,
provement
chronic bladder infections
(prostate problems),
incontinence or inability
to hold your urine
349(39.0)348(38.5).80
a Unless otherwi indicated,data reported are the percentage of usual
care(intervention)subjects having the specified comorbidity.
b Comparing differences across intervention conditions for multiple
imputed data ts.
c Number of comorbidities excludes cancer if it was skin cancer.
L.H.Harpole et al./General Hospital Psychiatry27(2005)4–127
3.1.Clinical outcomes
Table 3prents unadjusted and adjusted model results for all continuous outcome variables.Adjusted models operationalize the covariates as they are reported in Table 1.Site is operationalized as a categorical variable with eight levels.At baline,patients with higher numbers of chronic
dias had wor predicted depression verity (measured by SCL-20depression scores)than tho
with fewer chronic dias (P b .01).The number of chronic dias,however,did not alter the change in depression verity over time.Regardless of the number of chronic dias,intervention patients had significantly lower depression verity during all follow-up asssments (P b .001)as compared with patients in usual care.This is graphically demonstrated in Fig.1(unadjusted results)where within treatment groups,the two comorbid illness groups have parallel trajectories.The results were similar to the adjusted model.Columns 1and 2in Table 3demonstrate that in the adjusted analysis,l differences were esntially the same for tho with low vs.high numbers of comorbidities.A subgroup analysis evaluating the potential impact of gender upon outcome demonstrated that males and females followed similar patterns over time (results not shown).The effects of individual comorbid illness were examined in parate models (results not shown).Although the prence of arthritis,urologic problems,lung dia,chronic pain and diabetes were associated with wor depression scores at baline,the prence or abnce of the chronic illness did not alter the expected change in SCL-20score over time.
This relationship between number of comorbid medical illness and the MCS-12score was also evaluated (Table 3).
Table 3
Clinical outcomes
Unadjusted model estimates Adjusted analysis for intervention vs.usual care Usual care
Intervention
Between-group difference Number of comorbidities Number of comorbidities Number of comorbidities Low
High Low High Low a High b T P value c SCL-20depression score (range,0–4)Baline 1.57 1.75
1.62 1.730.02À0.01À0.74.463-Month follow-up 1.34 1.54 1.11 1.23À0.25À0.29À0.39.696-Month follow-up 1.10 1.280.870.98À0.25À0.29À0.23.8212-Month follow-up 1.29
1.46
0.90
1.06
À0.40
À0.38
0.76
.45
MCS-12(SF12mental health component score)Baline 42.4941.9342.74
42.130.420.30À0.24.813-Month follow-up 44.4544.3846.0445.82 1.63 1.43À0.11.916-Month follow-up 45.1345.0745.7645.100.560.09À0.51.6112-Month follow-up 45.0344.5646.5545.90 1.57 1.29À0.21.83
Overall QOL in past month (range,0–10)Baline 5.70 5.08 5.50 5.24À0.160.16 2.28.023-Month follow-up 6.10 5.49 6.56 5.970.490.48À1.65.106-Month follow-up 6.17 5.57 6.51 6.000.350.42À1.31.1912-Month follow-up 6.15
5.92
6.80 6.400.670.47À2.58.01
Linear mixed-effects regression was ud to fit all models prented in this table.Number of comorbidities is a simple count of chronic medical conditions in the past 3years excluding cancer if the type of cancer was skin.In the adjusted analysis,additional covariates included site,age,race,gender,education,marital status,type of depression,anxiety,NEO neuroticism score,PCS-12and any depression treatment in lifetime.
a
Estimate of treatment group difference for low comorbidities;note that low number of comorbidities is defined in this table as the first quartile,which is two chronic medical conditions.
b
Estimate of treatment group difference for high comorbidities;note that high number of comorbidities is defined in this table as the third quartile,which is five chronic medical conditions.
c
Reprents the test at each time point of whether the treatment group differences are the same for different numbers of comorbidities.(This is not a test of treatment group differences for the specific le
vels of two and five chronic medical
conditions.)
Fig.1.Estimated mean SCL-20scores over time for subjects receiving usual care with high and low numbers of comorbidities and for subjects receiving intervention with high and low numbers of comorbidities.High number of comorbidities is defined as the third quartile,which is five chronic medical conditions.Low number of comorbidities is defined as the first quartile,which is two chronic medical conditions.
L.H.Harpole et al./General Hospital Psychiatry 27(2005)4–12
8
Similar to the SCL-20results,the number of comorbid illness had little impact on MCS-12scores over time. Patients within the intervention group experienced improved MCS-12scores as compared with tho in the usual-care group,regardless of the number of comorbidities,at3and 12months.
For QOL,patients with higher numbers of chronic dia had wor QOL(lower scores)at baline as compared with tho with fewer comorbid illness(Table3).Again,the number of comorbid illness did not alter the intervention vs.usual-care effect.Over time,intervention patients experienced greater improvement in QOL as compared with usual-care patients,regardless of the number of chronic dias.
Table4prents analysis results from the dichotomous clinical outcomes.Intervention patients experienced greater rates of depression remission(defined as SCL-20score b0.5) and respon(defined as at least50%decrea in SCL-20 score from baline)regardless of the number of comorbid-ities.The differences held over the12-month follow-up period(Table4).Absolute rates of respon and remission in patients were lower in patients with more vs.less comorbid illness.The trends were similar in both the usual care and intervention group.
To address the question of whether or not some of the items that compri the SCL-20are in fact measuring symptoms that should be attributed to medical illness instead of depression verity,for example,poor appetite, sleep disturbance and fatigue,we created two modified SCL scores,one which excluded7of the20items that addresd potential physical complaints,and another which excluded3,and built unadjusted and adjusted models using the modified SCL scores(results not shown).Although the differences between patients with comorbidities were not as striking when using the modified SCL score,the results were still statistically significant and similar to tho en with the original SCL score.
3.2.Intensity of intervention
In order to determine whether it took more intervention resources to improve treatment outcomes for patients with more vere medical illness,we compared the measured number of phone visits,number of clinic visits,and number of phone and clinic visits combined to the DCS.Interven-tion patients averaged15.3total visits,with a mean of9.2 clinic and6.1phone visits.The association between the number of comorbid illness and the b do Q of the intervention was examined graphically and analytically through Spearman correlations(results not shown).The Spearman correlations between number of chronic dias and total number of visits,clinic visits,and phone vi
sits were0.03,À0.03and0.06,respectively.
4.Discussion
This study demonstrates that a multidisciplinary dia management program for depression in older adults was equally effective for patients with and without multiple chronic comorbid medical dias.As expected,at baline,individuals with more chronic medical illness had higher rates of depression and lower QOL in both the intervention and control groups.Regardless of the number of chronic dias,however,depresd individuals who were randomized to the IMPACT intervention program experienced similar rates of respon to depression
Table4
Clinical outcomes
Expected probabilities—unadjusted model Adjusted analysis for intervention vs.usual care
Usual care Intervention Between-group difference
Number of comorbidities Number of comorbidities Number of comorbidities
Low High Low High Low a High b T P value c Complete remission of depression symptoms(SCL-20score b0.5)
3-Month follow-up0.070.030.180.14 3.03 5.58 1.22.22
整流器6-Month follow-up0.200.140.310.29 1.94 2.650.91.36
12-Month follow-up0.110.060.280.23 3.37 5.26 1.06.29 Respon(at least50%decrea in SCL-20depression score from baline)
3-Month follow-up0.170.130.320.31 2.38 3.02 1.10.27
6-Month follow-up0.330.290.520.47 2.15 2.20À0.76.45
12-Month follow-up0.220.170.490.41 3.43 3.49À0.74.46 Generalized estimating equation was ud to fit all models prented in this table.Number of comorbidities is a simple count of chronic medical conditions in the past3years excluding cancer if the type of cancer was skin.In the adjusted analysis,additional covariates included site,age,race,gender,education,marital status,type of depression,anxiety,NEO neuroticism score,PCS-12,and any depression treatment in lifetime.
6350
a Odds ratio of treatment group difference for low comorbidities;note that low number of comorbidities is defined in this table as the first quartile,which is two chronic medical conditions.
b Odds ratio of treatment group difference for high comorbidities;note that high number of comorbidities is defined in this table as the third quartile,which is five chroni
c medical conditions.
c Reprents the test at each time point of whether the treatment group differences are the same for different numbers of comorbidities.(This is not a test of treatment group differences for the specific levels of two an
d fiv
e chronic medical conditions.)
L.H.Harpole et al./General Hospital Psychiatry27(2005)4–129