Clinical Practice ImprovementNetwork for Early Psychosis
CPIN-EP
Stanley Catts
University of Queensland
Tenth NSW Early Psychosis Forum
30 October 2007
Westmead
Overview
Evaluating early psychosis intervention– Naturalistic evidence– Clinical trials– Observational studies
Description of CPIN-EP
CPIN-EP results– Performance of CPIN-EP methodology– Cohort description– Impact of service model on patient registration
rates and DUP– Impact of clinical practice and family outcomes
Interim conclusions
EPI: Naturalistic evidence
Early disease course is the strongestpredictor of long-term outcome forpsychotic disorders1
Remission of the first psychotic episodeand avoidance of relapse during the firsttwo years of treatment may reduce long-term disability up to 30% irrespective ofbaseline patient characteristics2
1Harrison et al., 2001, Br J Psychiatry, 178,506-5172Harrison et al., 1996, Psychol Med, 26,697-705
EPI: Naturalistic evidence
The chance of full recovery reduces witheach relapse of psychosis1
The likelihood of relapse increases overtime2
1Wiersma et al., 1998, Schizophr Bull, 24,75-852Robinson et al., 1999, Arch Gen Psychiatry, 56,241-247
EPI: Naturalistic evidence
More patients respond well to medicationin the first episode (about 80%)1 comparedto subsequent episodes (about 50%)2
Initial response occurs at lower doses ofmedication3 and psychosocial intervention
1Lieberman et al., 1993, Arch Gen Psychiatry, 50,369-3762Lieberman et al., 1996, J Clin Psychiatry, 57,suppl 9,5-93Merlo et al., 2002, J Clin Psychiatry, 63,885-891
EPI: Clinical trials
“There was insufficient or no evidencefrom RCTs to support the benefits ofspecialist EPI teams, and no evidencefrom clinical trials to support the benefits ofearly detection of patients in their firstepisode of psychosis”1
“There is now quantitative evidence … thatenriched interventions for patients withrecent-onset psychosis are significantlymore effective than standard care … overa period of about one year”2
1Marshall & Rathbone, Cochrane Database, 2006, issue 42Harvey et al., (2007) Can J Psychiatry 52:464-472
Clinical trials:The OPUS Study
Design1,2
– 547 patients aged 16/18 - 45 presenting (1/98-12/00) for the first time with ICD-diagnosed nonaffective non-organic psychosis living in 2 districtsof Copenhagen and one in Aahus wererandomised to:
• Integrated Treatment (IT: specialist team using ACT,1:10; family therapy for 18 months; SST fro 12months)
• Standard Treatment (ST; standard casemanagement, 1:25; no family therapy or SST
– Patients have been followed-up for five years
1. Jorgensen et al (2000) Soc Psychiatry Psychiatr Epidemiol 35:283-287
2. Thorup at al (2005) Schiz Res 79:95-105
Clinical trials:The OPUS Study
163 (60%)206 (75%)Patients interviewed
At 24 months:
87 (62%)137 (74%)Carers interviewed4
192 (71%)227 (83%)Patients interviewed3
At 12 months:
140185Eligible carers
65%67%Schizophrenia
5345.5Median DUP (wks)
60%58%Males
26.626.6Mean age
272275At baseline (n)
STITSUBJECTS
3. Petersen et al (2005) Brit J Psychiatr 187 (Suppl 48):s98-103
4. Jeppesen at al (2005) Brit J Psychiatr 187 (Suppl 48):s85-90
Clinical trials:The OPUS Study
Results at 12 monthsOutcomes favoured IT significantly IT ST
M=1.13SD=0.4
M=1.02SD=0.04
Carer burden (SBAS subscale score)4
81561Days in hospital in last 12 months
73%357%Any ‘poor outcome’
53%342%No work/study
17%310%Homeless/supervised accommodation
22%316%Patients with SUDS
17%310%Patients with GAF –s <30
35%322%Patients with SANS ss >3
20%310%Patients with SAPS s.s >3
3. Petersen et al (2005) Brit J Psychiatr 187 (Suppl 48):s98-103; 4. Jeppesen at al (2005) Brit J
Psychiatr 187 (Suppl 48):s85-90; 5. Norden et al (2003) Schiz Res 60 (Suppl): s297
Carer satisfaction (Mean IT vs ST difference score on modified CSQ) = 3.4
Clinical trials:The OPUS Study
Results at 24 months:The following outcomes significantly favoured IT vs ST:
• Mean SAPS score 1.07 vs 1.29
• Mean SANS score 1.42 vs 1.84
The following outcomes favoured IT vs ST at level of a trend:
• Days in hospital in past 2 years
Differences in treatment at 24 months:Number of contacts with primary staff: IT = 77 vs ST = 27
Relatives involved in treatment: IT = 61% vs ST = 20%
Similar medication doses were used and similar number of
patients in IT vs ST had stopped medication at 2 years (30%)
Thorup at al (2005) Schiz Res 79:95-105
Clinical trials:The OPUS Study
Results at 5 years:“The results show no difference between treatment
groups at five-year follow-up. … It seems that twoyears of intensive treatment in the early phases ofpsychosis is not enough to ensure a good long-term outcome.”
Bertelsen et al (2006) Schiz Res 86:S43
EPI: Observational studies
Demonstrating treatment effects of EPI in RCTs is difficultbecause– A proximal element of EPI operates through a chain of processes
aiming to improve a distal health outcome
– The proximal application of EPI is temporally distant from long-termoutcomes of interest
– In the absence of agreement about the essential ingredients foreffective EPI, the composition of programs is usually dissimilar,preventing ready meta-analytic evaluation
Observational studies can clarify essential components ofintervention and allow examination of the effect of adiscrete intervention on a relevant proximal process,providing information for designing effective EPI programsand sound RCTs
EPI: Observational studies
Pros– Real-world treatment can be observed
without experimental intervention– All patients may be included– Relatively inexpensive– Can be carried out in a quality assurance
framework
Cons– Confounding and biases are inherent,
necessitating complicated analyticapproaches
– Any finding made has to be confirmedexperimentally in an RCT
An Australian multi-siteevaluation of EP programs
Chief Investigators
Stanley V. Catts*
Brian I. O’Toole**
Vaughan J. Carr***
Associate Investigators
Terry Lewin***
Amanda Neil***
Meredith Harris****
The NHMRC Clinical Practice Improvement
Network for Early Psychosis
*U of Queensland; **U of Sydney; *** U of Newcastle; ****ORYGEN Youth Health
The NHMRC Clinical PracticeImprovement Network
for Early Psychosis (CPIN-EP)
CPIN-EP is an epidemiological study ofthe effect of exposure to guideline-adherent early psychosis intervention
The study used a prospectiveobservational cohort design
Only routinely collected data was used
CPIN-EP was approved as a qualityassurance project, not requiring patientconsent
CPIN-EP has amulti-level hierarchical design
EPI teams
Observations over 6 months
Patient 2 Patient n
Clinician 1 Clinician 2 Clinician n
Level 3
Level 1
Level 2
Patient 1
Level 4
C-PINEP Data Sources
Case Manager
Feedback
Questionnaire
Service Self
Review
Instruments
Service
Census
Service Level
Patient Level
Consumer and
Carer Feedback
Questionnaires
Chart
AuditHoNOS, LSP,
Consumer
Measures
Service
Contact
Forms
Recording
Forms
Data collection protocol
Patient level data acquisition
Structured file audit by medical studentsof photocopied de-identified file noteson first six months (or till service exit) toextract:– Demographics
– CPIN-EP indicator codes
– Verbatim file recorded diagnosis (at anytime within 6 months)
– CPG-EP adherence
– Documentation quality indicators
Bulk scanning of Service Contact Formintervention list (reverse side of SCF)
Early Psychosis Indicator
On the basis of available information, are you confident that at some
point ever (including now) a diagnosis of a psychotic disorder* could be
made for this patient?
Has the patient received a
diagnosis of a psychotic disorder
more than 12 months ago?
Do you think the patient is/has
been highly likely to be
prodromal in the last 12 months?
YES NO
YESNO NO YES
2
(Definite EP)
0
(Not EP)
1
(Possible EP)
* Any Axis I disorder with psychotic symptoms
0
(Not EP)
CPIN-EP results:Performance of EP indicator
(flag)
Highly likely EP Definitely EP
(prodromal) (diagnosable)
Flag 1 Flag 2
Flagging rate (n=451) 26% (n=116) 74% (n=335)
Final diagnosis within 6
months: Psychotic disorder 75% 100%
Final diagnosis within 6
months: Non-psychotic disorder 25% 0%
Final psychotic diagnosis(by detailed audit)
Schizophreniaspectrum
Other psychotic
disorder
Affectivepsychoses
Non-psychotic
disorder
FLAG 2 FLAG 1
Baseline, 3-month, 6-month measures:
HoNOS
Social and Occupational Functioning (SOFAS)or GAF/GAS
DUP Indicator (only at baseline)
Service Dropout Indicator (only at serviceseparation)
Suicide Risk
Substance Use
Relational [e.g. family] Functioning (GARF)
CPIN Clinical Indicators
Results of Reliability Analysis
= 0.82 Social and Occupational Functioning Assessment Scale
= 0.90 Global Assessment of relational functioning
= 0.91 Suicide Risk
= 0.99 Duration of Untreated Psychosis
= 0.96 EP Indicator
EP TeamINDICATOR
= 0.84
= 0.84
= 0.30
= 0.96
= 0.79
Non EP
Missing data rates
Measure Baseline Six months
(completers)
(n=451) (n=320)
HoNOS 27% 43%
SOFAS 31% 58%
GARF 33% 58%
Cannabis 36% 64%
Stimulants 43% 67%
Diagnostic information
From structured file audit:
Proportion of cases where no definite diagnosis*recorded by 6 months or service exit
35% (157/451)
Proportion of cases given an indefinite diagnosisat first assessment who did not have adocumented diagnostic review by 6 months orservice exit
57% (89/157)
*Psychosis NOS included as a definitive diagnosis
Diagnostic information
From unstructured and detailed (readingevery word in the file) file audit bymedical/nursing-trained auditors:
Proportion of cases where a definite DSMIV diagnosis* could be made with relativecertainty
~ 100% (449/451)
*Psychosis NOS included as a definitivediagnosis
Total sample baselinecharacteristics (n=451)
Age and Gender
Mean Age (at first contact with service) 21.61
Age Range 13-56
Gender (Male) 64%
Indigenous
Non-indigenous 97.8%
Indigenous 2.2%
Total sample baselinecharacteristics (n=451)
Living Arrangements
With family 66.1%
With partner 9.3%
With friends 7.5%
Alone 7.3%
Homeless 9.8%
Employment
Employed 37.9%
Unemployed 62.1%
Total sample baselinecharacteristics (n=451)
Referral Sources
General Practitioners 16.4%
Police 6.0%
Family/Friends/Self 39.7%
Other 37.9%
First Contact Setting
Outpatient 55.6%
Inpatient/Emergency 44.3%
Total sample baselinecharacteristics (n=451)
Duration of Untreated Psychosis
Delusions 8.18 months (10.96 SD)
Hallucinations 9.77 months (12.52 SD)
Other Symptoms 14.46 months (12.84 SD)
EP detection rate by AMHS(sexes separated)
0
50
100
150
200
250
300
1 2 3 4 5 6 7 8
Males
Female
Fla
g 2
pe
r 1
00
,00
0 p
eo
ple
EP detection rate by AMHS(sexes combined)
0
50
100
150
200
1 2 3 4 5 6 7 8
Fla
g 2
am
on
gs
t 1
8-2
5 y
ea
r o
lds
pe
r 1
00
,00
0 p
eo
ple
AMHS 1 and 2 have relatively long-established EPI functions (shading)
Flag 1 + 2 DSM IV grouping byAMHS (up to 3 Dx per patient)
3%3%3%4%12%12%8%16%
Personality
disorders - B
21%47%31%49%30%44%15%34%
Substance use
disorders
26%11%15%12%30%26%8%10%
Anxiety and non-
psychotic
depressive disorders
18%21%8%7%32%9%0%14%Affective psychoses
34%21%20%14%19%25%0%22%
Other psychotic
disorders
42%52%64%75%36%51%92%62%
Schizophrenia
spectrum
87654321DSM IV grouping
Detection rate versusservice model
AMHSs with highest recruitment ratesof EP patients:– are long-established
– have specialised function
– are fully integrated across services
Relationship betweendetection rate and DUP
If an AMHS successfully recruitsincreased numbers of EP patients yearafter year, does this in itself achieveearlier detection, as indexed by shorterDUP?
Predictors of DUP at thepatient level
Adjusted
hazard
ratioa 95% CI p value
Age at onset 1.021 0.998-1.044 0.074
Employed 1.282 0.972-1.692 0.079
Living with family 1.308 0.874-1.958 0.191
Acute mode of onset 2.012 1.319-3.068 0.001
Diagnostic group 0.007b
Affective psychoses 1.291 0.973-1.713 0.076
Other psychotic disorders 2.054 1.269-3.326 0.003
SES of residential suburb 1.001 0.999-1.003 0.194
Service model 0.030c
Extensive implementation 1.765 1.028-3.029 0.039
Some implementation 2.013 1.200-3.377 0.008aHazard ratio > 1 indicates higher risk of contact with services at any point in time and,
therefore, shorter average DUPbReference category: Schizophrenia spectrum disorderscReference category: No/Little EP model implementation
DUP and DUP-relatedcovariates, by service
1 2 3 4 5 6 7 8
DUP (months)* 10.6 7.3 11.0 7.0 7.9 16.2 9.0 6.2
mean (SD) 12.9 3.9 13.7 6.7 12.4 15.6 10.7 8.6
median 6 8 6 5 3 12 3 2
Sex (Males %) 61.2 66.7 60.3 50.0 61.6 80.6 69.4 75.0
Living with family** 88.3 81.8 91.4 92.3 69.4 74.3 88.6 86.1
Education level*** 90.8 50.5 65.3 93.9 79.1 84.4 73.5 65.6
SES (% below
median SEIFA)*** 71.5 91.7 82.4 42.5 13.7 55.6 2.8 34.3
*p< 0.05; **p<0.01; ***p<0.001
Analyses exclude service # 2, due to small cell sizes
Change (improvement) inSOFAS scores by AMHS
0
2
4
6
8
10
12
1 2 3 4 5 6 7 8
Change (improvement) inGARF scores by AMHS
0
0.5
1
1.5
2
2.5
3
3.5
4
1 2 3 4 5 6 7 8
Multi-Level Modeling
Advanced form of regression
Analyses data within naturally occurringhierarchies (e.g. child within school)
Treating all the data as though it is onelevel under-estimate standard errors –increasing chance of type I error
Hierarchy of CPIN
Level 4 Treating Team n = 19
1 :
Level 3 Clinician n =181
1 :
Level 2 Patient n = 464
1 :
Level 1 Observation n = 1086
Expressing the Hierarchy
Regression Equation
Residuals
Estimating Variance
Parameter Estimate
Standard Error
Where is the Variance ?
Level 4 Which EP Team?
0 – 2.3%
Level 3 Which EP Clinician?
0 – 1.5%
Level 2 Which EP Patient?
29 – 45 %
Level 1 What happened to the patient?
54 – 70 %
What does this mean ?
Up to 70% of the variance in six monthsSOFAS score is explained by whathappened to the patient over time(including treatment received)
If the type of service or nature of theclinician is important, the impact of thesevariables is mediated by how theyinfluence the patient’s clinical treatment
Specific conclusions (1)
Despite strong support from naturalisticstudies, so far clinical trials have notconfirmed the effectiveness of EPI inimproving long-term outcomes
CPIN-EP, an observational study, wasnot designed to evaluate theeffectiveness of EPI, rather whetherguideline-adherent EPI during the firstsix months of treatment influencesshort-term outcome
Specific conclusions (2)
CPIN-EP evaluation tools were applicableand reliable, but CPIN-EP procedureswere not followed consistently by clinicians(e.g., rating the HoNOS)
Clinicians did not derive a definitivediagnosis in more than one-third of cases
These two factors prevented CPIN-EP ingiving timely feedback as a qualityimprovement project should
Specific conclusions (3)
There were substantial differences inthe EP detection rates across services,and these differences appear to berelated to EPI service model
Differences in diagnostic profile andDUP across services appeared to besecondary to differences in detectionrate, not the other way around
Specific conclusions (4)
In the CPIN-EP study DUP and EPoutcomes were confounded, whichagain calls into question the broaderobservational literature concerning therelationship between DUP and outcome
There were substantial differences inpatient and family outcomes at 6months, and these were primarilydetermined by clinical practice andsecondarily by baseline patientcharacteristics
General conclusions
Published RCTs of EPI are not truetests of the efficacy of EPI becauseenhanced initial treatment is beingcompared to treatment as usual, notearly versus late intervention
Observational studies can informclinical practice and guide design offuture RCTs
Acknowledgements
Project teamKathy EadieRussell EvansAaron FrostBelinda Schaefer
This project is dedicated to mentalhealth consumers, carers and serviceproviders. They are the enduringinspiration for this work.