Impact of early intervention services on duration of untreated psychosis: data from the National EDEN prospective cohort study

Page 1

Schizophrenia Research xxx (2014) xxx–xxx

SCHRES-05954; No of Pages 6

Contents lists available at ScienceDirect

Schizophrenia Research

j ourna l homepage: www.e lsev ie r .com/ locate /schres

Impact of early intervention services on duration of untreated psychosis:Data from the National EDEN prospective cohort study

Max Marshall a,⁎, Nusrat Husain a, Natalie Bork a, Imran B. Chaudhry a, Helen Lester b, Linda Everard b,Swaran P. Singh c, Nick Freemantle d, Vimal Sharma e, Peter B. Jones f, David Fowler g, Tim Amos h,Barbara Tomenson a, Max Birchwood b

a School of Medicine, University of Manchester, United Kingdomb University of Birmingham, Birmingham, United Kingdomc Health Sciences Research Institute, University of Warwick, Warwick, United Kingdomd Department of Primary Care and Population Health, University College, London, United Kingdome University of Cheshire, United Kingdomf Department of Psychiatry, University of Cambridge, Cambridge, United Kingdomg School of Medicine, University of East Anglia, United Kingdomh Academic Unit of Psychiatry, University of Bristol, Bristol, United Kingdom

⁎ Corresponding author at: The LANTERN Centre, Vicar8DY, United Kingdom. Tel.: +44 1772 773500; fax: 44 17

E-mail address: [email protected] (M. M

http://dx.doi.org/10.1016/j.schres.2014.07.0050920-9964/© 2014 Published by Elsevier B.V.

Please cite this article as: Marshall, M., et al.EDEN prospective cohort study, Schizophr. R

a b s t r a c t

a r t i c l e i n f o

Article history:

Received 5 December 2013Received in revised form 5 July 2014Accepted 10 July 2014Available online xxxx

Keywords:Early Intervention ServicePsychosisDuration of untreated psychosis

Objective: This study aimed to determine if the inception of Early Intervention Services (EISs) is followed by animprovement in the prompt treatment of people with first episode psychosis.Method: A prospective cohort study of referrals to new and established EISs was conducted at 1, 2, 3, and 4 yearsafter inception of new EIS.The study was conducted with 14 (seven new and seven established) secondary care EIS within geographicallydefined catchment areas in England between 2005 and 2009. Participants included 1027 consecutive referrals toEIS aged 14–35 with a first episode of psychosis.Duration of untreated psychosis (DUP) and number of participants treated adequately within 6 months of onsetwere the main outcome measures.

Results: A significant downward trend across yearly cohorts for DUP for new EIS (F1,549= 8.4, p= 0.004) but notfor established EIS (F1,429= 1.7, p= 0.19) was observed. There was a significant upward trend across cohorts inthe proportion of referrals treated within 6 months for new EIS (X2 = 8.0, df = 1, p = 0.005), but not forestablished EIS (X2 = 0.1, df = 1, p = 0.72).Conclusion: The introduction of new EIS was followed by a reduction in DUP and an increase in the proportion ofpatients treated within 6 months of onset. These trends were not present in the catchment areas of establishedservices where DUP was initially lower, suggesting that there was no general tendency for DUP to fall over time.Hence, the introduction of an EIS was followed by an improvement in the prompt and proper treatment of firstepisode psychosis.

© 2014 Published by Elsevier B.V.

1. Introduction

The duration of untreated psychosis (DUP) is the time from the firstpsychotic symptom to the initiation of adequate neuroleptic treatment(Norman and Malla, 2001). There are three compelling reasons forshortening DUP. The first is to avoid the social consequences of activepsychosis such as homelessness, unemployment, and social isolation

age Lane, Fulwood, Preston PR272 718268.arshall).

, Impact of early interventiones. (2014), http://dx.doi.org/1

(Addington et al., 2002, Bertelsen et al., 2009). The second is to improveprognosis because a shorter DUP is associated with a better recoveryacross a broad range of outcomes (Marshall et al., 2005, Perkins et al.,2005). The third is to reduce the risk that a person with psychosis willseriously harm himself or other people because this risk is greatest dur-ing the untreated period of their first episode, and increases with thelength of the episode (Nielssen and Large, 2008, Barrett et al., 2010,Nielssen et al., 2011).

It has been proposed that specialist teams for the treatment of earlypsychosis (known as Early Intervention Services or EISs) could reducethe duration of untreated psychosis by promoting prompt and

services on duration of untreated psychosis: Data from the National0.1016/j.schres.2014.07.005

Page 2

2 M. Marshall et al. / Schizophrenia Research xxx (2014) xxx–xxx

proper treatment (Bertolote and McGorry, 2005). However, thisproposition is difficult to test in a randomized controlled trial as itwould require a complex cluster randomized design in which EIS wasrandomly allocated to localities (Lloyd-Evans et al., 2011). Whilst theproposition has some support from small before and after designsconducted in single localities, a recent systematic review of the limitedevidence concluded that “establishing early intervention services prob-ably does not on its own reduce DUP (Lloyd-Evans et al., 2011)”.

A unique aspect of the English National Health Service is thatEIS looks after all young adults with a first episode of psychosis, follow-ing a national policy introduced in 2000 (Department of Health, 2000).In 2003, the English Department of Health set two national targets forEIS: to reduce the median DUP to 3 months, and to minimize the num-ber of people with a DUP of greater than 6 months (Norman andMalla,2001, Department of Health, 2003).

Whilst the policy of establishing EIS was announced in 2000, imple-mentation was initially patchy due to cost and complexity. In 2005,when the National EDEN longitudinal study of people receiving carefrom EIS started, national coverage was still variable. This afforded anopportunity to observe the effects of introducing new EIS on the dura-tion of untreated psychosis. We hypothesised that over four years theduration of untreated psychosis and the proportion of participantswith a DUP of greater than 6 months would fall significantly in thecatchment areas of new EIS, whilst the same parameters would remainstable in the catchment areas of established EIS.

2. Methods

National EDEN (2005–16) is a longitudinal cohort study of referralsto early intervention services in five socioeconomically diverse sitesacross England: Birmingham, Cornwall, Cambridgeshire, Norwich, andLancashire/Cheshire. We chose sites purposively to reflect urban/ruraldifferences. One of our core research objectives for National EDENwas to determine the duration of untreated psychosis in referrals toearly intervention services in England (Birchwood et al., 2014). For thepresent study, we identified, from National EDEN, a prospectivecohort of first referrals to EIS at 1, 2, 3, and 4 years after the establish-ment of new EIS and compared this cohort with a contemporaneous co-hort of first referrals to EIS. Fourteen EISs participated at the five sites,each within a defined geographical catchment area, from where theyaccepted all new cases of first episode psychosis in people aged 14–35.We classified participating EIS into those that had been set up for lessthan one year when data collection started (new services), and thosethat had been set up for more than one year when collection started(established services) (see Table 1).

Table 1New and established Early Intervention Services in National EDEN.

Team Catchment area size(100 k pop)

Set update

StudyEntry date

N in study

NewLancashirea 1,306,000 2005 2006 189Birmingham BENa 383,300 2005 2005 98Birmingham Southa 392,800 2005 2006 79Kings Lynn, Norfolk 188,600 2007 2007 11Birmingham–Solihull 205,000 2006 2007 31Cameo North, Cambridge 311,274 2007 2008 23Cornwalla 536,000 2005 2006 122

EstablishedBirmingham Centrala 310,000 1995 2005 66Birmingham Easta 132,800 2002 2005 67East Anglia Norfolka 633,400 2003 2005 146Cameo Southa 466,911 2002 2005 98Wirral 468,000 2005 2007 27West Cheshire 260,000 2004 2007 18East Cheshire 315,000 2005 2007 11

a Teams with data for more than 3 years (included in sensitivity analysis).

Please cite this article as: Marshall, M., et al., Impact of early interventionEDEN prospective cohort study, Schizophr. Res. (2014), http://dx.doi.org/

We approached all referrals accepted by participating EISs fromAugust 2005 to April 2009. However, not all EISs were recruited forthe whole period (see Table 1). Our inclusion criteria were the follow-ing: aged 14 to 35 and accepted into a participating EIS following afirst episode of psychosis. We obtained informed consent from allparticipants.

Research assistants, not directly involved in clinical care, assessedparticipants at intake to the study and performed 6 and 12-monthfollow-ups. We evaluated DUP using an internationally accepted stan-dardized definition and methodology (Larsen et al., 1996) that definedthe onset of psychosis as one symptom from the positive scale ofthe Positive and Negative Syndrome Scale (PANSS) (Kay et al., 1987)at a level of 4 or above, or a cluster of symptoms including either delu-sions, conceptual disorganisation, or hallucinations, with a total score of7 ormore (excluding ‘absent’ scorings). Symptomshad to be present fora period of twoweeks ormore (unless remissionwas due to treatment).We defined the end of the period of untreated psychosis as the onset ofcriterion treatment with antipsychotic medication (Joint FormularyCommittee, 2005).

In some cases participants never received criterion treatment de-spite being under the care of an EIS. Somewho did not receive criteriontreatment recovered nonetheless, so that theywere no longer psychoticwhen assessed at intake to the study or at 6 or 12-month follow-ups. Forthese participants we added 0, 6, or 12 months to the duration ofuntreated psychosis depending on the follow-up point at which theywere judged to be no longer psychotic. The remainder of the groupwhonever received criterion treatment remained psychotic throughoutthe study, and their duration of untreated psychosis was increased by12 months.

Diagnosis was established using the OPCRIT (Operational Criteria)computerized diagnostic system. OPCRIT is a 90-item checklist of symp-toms rated by a researcher from the participant's clinical case notes.OPCRIT generates diagnoses according to 12 operational diagnostic sys-tems (including ICD 10, DSM-III, and DSM-IIIR). OPCRIT has been usedin a wide range of psychiatric research and has proven to be both reli-able and valid (McGuffin et al., 1991).

2.1. Statistical analysis

We present descriptive statistics for the whole sample and for newand established EISs separately. We compared new and establishedEISs using chi-squared tests for categorical variables, and t-tests andMann–Whitney tests for ages, duration between key dates, and log-transformed duration of untreated psychosis. We present medians andgeometric means with 95% confidence intervals for duration of untreat-ed psychosis as it is positively skewed and lognormally distributed.Nested one-way analysis of variancewas used to estimate the differencebetween new and established services aswell as the difference betweenindividual services within these two groups. We calculated proportionsof referrals with DUP less than 6 months by service, by new versusestablished service, and for the whole group. For comparisons betweenservices and between new and established services we used Fisher'sexact test.

We investigated whether DUP was stable over time in the catch-ment areas of established services whilst falling over time in the catch-ment areas of new services as they became established, by studyingduration of untreated psychosis between new and established servicesacross four yearly cohorts, using as baseline for each service the datewhen the first subject from that service entered the study. Some ser-vices were being offered through National EDEN from the beginningand had patients in all 4 yearly sequential cohorts; later services hadfewer. Multiple regression was used with log (DUP) as the dependentvariable and yearly cohorts, new/established team status, and the inter-action of these as the independent variables.

Since the interaction termwas significant, we used one-way analysisof variance to estimate a linear trend in geometric mean duration of

services on duration of untreated psychosis: Data from the National10.1016/j.schres.2014.07.005

Page 3

Table 2Comparison of participants in established and new EISs.

Whole group(n = 986)

Established EIS(n = 433)

New EIS(n = 553)

Sig.a

Risk or protective factor n/N % n/N % n/N % p

Female 308 31.2 136 31.4 172 31.1 0.95Ethnic origin:

White 714 72.4 288 66.5 426 77.0 X2 = 16.7Asian 153 15.5 85 19.6 68 12.3 df = 3Black 71 7.2 40 9.2 31 5.6 P = 0.001Mixed/other 48 4.9 20 4.6 28 5.1

Born in the UK 884/972 90.9 374/425 88.0 510/547 93.2 0.007Fluent in English 942 95.5 408 94.2 534 96.6 0.088Marital status:

Single 835 84.7 362 83.6 473 85.5 X2 = 0.9Married/cohabiting 123 12.5 59 13.6 64 11.6 df = 2Separated or divorced 28 2.8 12 2.8 16 2.9 p = 0.63

Living situation:Alone 126 12.8 54 12.5 72 13.0 X2 = 8.3With parents/guardians 623 63.4 257 59.6 366 66.3 df = 3With partner 104 10.6 49 11.4 55 10.0 p = 0.040Other 130 13.2 71 16.5 59 10.7

Educated to A level or higher 341 34.6 161 37.2 180 32.5 0.14In paid work 185 18.8 87 20.1 98 17.7 0.37Ever used drugs 626/951 65.8 263/412 63.8 363/539 67.3 0.27Schizophrenia in 1st degree relative 91/880 10.3 40/379 10.6 51/501 10.2 0.91

n = number of patients in the group with the risk or protective factor. N = total number of patients in the group. N = 433 for established teams, and N = 553 for new teams unlessotherwise stated.

a Comparison used Fisher's exact test for dichotomous variables, and chi-squared test for other variables.

3M. Marshall et al. / Schizophrenia Research xxx (2014) xxx–xxx

untreated psychosis across the yearly cohorts in new and establishedservices separately. We used logistic regression analysis with DUPless than 6 months as the dependent variable and yearly cohorts,new/established EIS, and the interaction of these as the independentvariables. Since the interaction term was significant, the proportion ofsubjects with DUP less than 6 months in each cohort was comparedusing the chi-squared test with the test for linear trends across cohortsfor new and established teams separately.

We based our power calculation on the ability to detect a clinicallysignificant difference of 10% in the number of people treated in lessthan 6 months in new and established teams. A two-group continuitycorrected chi-squared test with a 0.050 two-sided significance levelhas 80% power to detect the difference between 70% for establishedteams and 60% for new teams (odds ratio of 1.556) when the samplesize in each group is 376.

3. Results

3.1. Participants

Over the recruitment period 2097 new patients were incepted intoparticipating EIS. Of these 1027 (49%) were recruited into NationalEDEN.Womenweremore common amongst thosewhodid not consent(34% versus 31%) but otherwise there were no significant differences

Table 3Comparison of new and established EISs for age and prodrome.

Established EIS (n = 433) N

Mean sd n M

Age accepted by EIS 22.2 4.63 433 22Age at onset of psychosis 21.3 4.80 433 21

Duration of intervals in daysProdromec 593 903 414 60Duration of untreated psychosis (DUP) 300 636 433 32

a Comparison using t-test and Mann–Whitney test.b Unequal variance version of the t-test used.c No date for the onset of non-specific symptoms so the prodrome cannot be calculated for

Please cite this article as: Marshall, M., et al., Impact of early interventionEDEN prospective cohort study, Schizophr. Res. (2014), http://dx.doi.org/1

between those who did not consent and those who did. For the currentstudy, we included 986 participants from the 1027 enrolled in NationalEDEN.

We excluded 14 participants because they were not first episodeand had been receiving criterion treatment for psychosis for someyears before acceptance into EIS. We excluded an additional 10 whodid not appear to have experienced psychosis according to our detailedbaseline assessment, and another three who were over the age limit of35 years at the onset of psychosis. In addition, we excluded six partici-pants transferred from EIS who were not participating in NationalEDEN and eight for whom no data on duration of untreated psychosiswere available.

3.2. Socio-demographic and psychiatric history data

The final sample included 308 (31.2%) females and 678 males (seeTable 2), with ages ranging from 14 to 35 years (mean = 22.5 years,standard deviation = 4.8 years). The mean age at the onset of prodro-mal (non-specific) symptoms was 19.8 years (standard deviation =5.2), and at the onset of psychosis was 21.4 years (standarddeviation = 4.9). Majority had an OPCRIT diagnosis in the schizophre-nia spectrum (n = 874, 88.6%); 44 (4.5%) had bipolar disorder, and in68 (6.9%) a diagnosis could not be assigned from the available records.Participants in the new EIS were significantly more likely to be white

ew EIS (n = 553) Comparisona

ean sd n t df p M–W p

.4 5.02 553 0.9b 958 0.38 0.67

.5 5.06 553 0.6 984 0.58 0.98

6 903 537 0.2 949 0.83 0.975 642 553 0.6 984 0.56 0.25

35 patients (19 in established teams and 16 in new teams).

services on duration of untreated psychosis: Data from the National0.1016/j.schres.2014.07.005

Page 4

Table 4Duration of untreated psychosis in days for each EIS.

n Min Max Median Mean 95% CI formean

Geometricmean

95% CI forgeometric mean

Number and percentageof patients with DUPunder 6 months

Established EISBirmingham Central 66 0 2905 45 237 127 to 346 40 22 to 72 47 (71.2%)Birmingham East 67 0 2022 141 296 195 to 398 85 51 to 142 37 (55.2%)East Anglia Norfolk 146 0 5652 102 385 252 to 518 90 66 to 124 90 (61.6%)CAMEO South 98 0 4748 47 257 130 to 384 43 28 to 66 72 (73.5%)Wirral 27 0 3598 113 322 47 to 596 69 28 to 165 17 (63.0%)West Cheshire 18 0 298 73 93 52 to 133 52 25 to 109 16 (88.9%)East Cheshire 11 5 783 133 261 67 to 455 73 15 to 347 6 (54.5%)All established teams 435 0 5652 77 300 240 to 361 64 53 to 78 285 (65.8%)

New EISLancashire 189 0 5435 146 438 333 to 544 133 103 to 173 100 (52.9%)Birmingham BEN 98 0 4821 34 208 94 to 321 24 15 to 40 76 (77.6%)Birmingham South 79 0 1900 141 307 217 to 397 100 66 to 154 43 (54.4%)Norfolk Kings Lynn 11 0 1471 12 300 −18 to 617 15 1 to 157 7 (63.6%)Solihull 31 3 2807 87 357 144 to 569 92 45 to 187 19 (61.3%)CAMEO North 23 0 857 110 200 100 to 299 68 29 to 159 13 (56.5%)Cornwall 122 0 6185 67 272 141 to 402 57 40 to 82 82 (67.2%)All new teams 556 0 6185 89 325 271 to 378 72 60 to 86 340 (61.5%)All teams 986 0 6185 82 314 274 to 354 68 60 to 78 625 (63.4%)

4 M. Marshall et al. / Schizophrenia Research xxx (2014) xxx–xxx

and born in the UK than those in the established EIS (see Table 2), butnot significantly different in terms of sex, marital status, education(see Table 2), age, or length of prodrome or DUP (see Table 3).

3.3. DUP for new and established EISs

Descriptive statistics for DUP for each EIS and for new andestablished EISs are shown in Table 4. The median DUP was less than3 months for both new and established EISs (77 days for establishedEIS and 89 days for new EIS). Overall, the new and established EISshad similar durations of untreated psychosis (F1,12 = 0.27, p = 0.60for logDUP), although there were significant differences between EISs,and within new and established EIS groups (F12,972 = 5.6, p b 0.001for logDUP). The percentage of participants with DUP less than6 months in the various EIS groups ranged from 52.9% at Lancashire to88.9% at West Cheshire. This difference between EISs was highlysignificant (X2 = 35.0, df = 13, p = 0.001). However, overall, thenew and established EISs did not have significantly different propor-tions of participants with DUP less than 6 months [63.4% overall, 340(61.5%) in the new EIS compared with 285 (65.8%) in the establishedEIS, Fisher's exact p = 0.16].

3.4. DUP and time since establishment of EIS

We hypothesised that there would be a significant downwardtrend in mean DUP in the new EIS across the annual cohorts, but notin the established EIS. This hypothesis was confirmed. There was asignificant interaction effect for logDUP between new/established EISand linear trends across the yearly cohorts (t = 3.04, p = 0.002),suggesting a significantly different linear trend across the cohorts fornew and established EISs. Analysing new and established EISs separate-ly showed that there was a significant downward trend in geometric

Table 5Duration of untreated psychosis in 4 yearly cohorts since establishment of EIS for new and est

Established teams

Geometric meanDUP in days

95% confidenceinterval

Within 1 year 56.9 39.1 to 82.91–2 years 57.8 38.4 to 87.02–3 years 67.9 46.9 to 98.23–4 years 83.3 53.8 to 128.6

Please cite this article as: Marshall, M., et al., Impact of early interventionEDEN prospective cohort study, Schizophr. Res. (2014), http://dx.doi.org/

mean DUP for new EIS (b = −0.13, 95% confidence interval −0.22 to−0.04, p= 0.004), but no significant linear trend across the four yearlycohorts for DUP in established EIS (b = 0.05, 95% confidence interval−0.03 to 0.13, p = 0.19).

Transforming these estimates back from the log scale to the originalscale gives a ratio change per year of 0.74 with 95% confidence interval,0.60 to 0.91 for new EIS, and 1.13, 95% confidence interval 0.94 to 1.36for established EIS. The cohort geometric means and 95% confidence in-tervals are shown in Table 5.

3.5. Proportion of participants with DUP less than 6 months and time sinceestablishment of EIS

We also hypothesised that there would be a significant upward lin-ear trend across the cohorts with respect to the proportion of patientswith DUP less than 6 months for the newEIS, but not for the establishedEIS. There was a significant interaction between yearly cohort and new/established EIS (Wald = 5.49, p = 0.019), again suggesting a sig-nificantly different linear trend across the cohorts for new andestablished EISs. Separate analyses showed no significant linear trendfor established EIS (X2=0.1, df=1, p= 0.72) and a significant upwardtrend for new EIS (X2 = 8.0, df = 1, p = 0.005). The number and per-centage of patients with DUP less than 6 months for each yearly cohortare shown in Table 6 for new and established teams separately.

3.6. Sensitivity analyses

To test the robustness of our findings we carried out three sensi-tivity analyses. First, we controlled for the fact that six EISs did notcontribute at all to the 4th annual cohort by excluding data from partic-ipants admitted in the 4th yearly cohort (n=108) and all the data fromthe six EISs (Wirral, West and East Cheshire, CAMEO North, Kings Lynn,

ablished EIS teams.

New teams

Geometric meanDUP in days

95% confidenceinterval

Within 1 year 86.7 65.9 to 114.11–2 years 74.8 56.7 to 98.62–3 years 57.9 35.3 to 94.63–4 years 24.5 10.0 to 58.7

services on duration of untreated psychosis: Data from the National10.1016/j.schres.2014.07.005

Page 5

Table 6Number and percentage of patients with duration of untreated psychosis less than6 months in 4 yearly cohorts since establishment of EIS for new and established EIS teams.

Established teams New teams

N % N %

Within 1 year (n = 122) 82 67.2 Within 1 year (n = 214) 119 55.61–2 years (n = 122) 80 65.6 1–2 years (n = 229) 143 62.42–3 years (n = 112) 73 65.2 2–3 years (n = 79) 54 68.43–4 years (n = 77) 50 64.9 3–4 years (n = 31) 24 77.4

5M. Marshall et al. / Schizophrenia Research xxx (2014) xxx–xxx

and Solihull) with no data for this cohort (n = 121), leaving data forthe first three year cohorts from eight of the fourteen EISs (n = 757).The linear trend for the geometric mean was then non-significant,(b = 0.05, 95%CI −0.08 to 0.19, p = 0.43) for the established EIS,and b = −0.09, 95% CI −0.21 to 0.02, p = 0.11 for the new EIS, i.e.no significant downward trend for either group. However, thereremained a significant upward linear trend across the remaining threeyearly cohorts with respect to the proportion of participants with DUPless than 6 months for the new EIS (X2 = 4.2, df = 1, p = 0.040), butnot for the established EIS (X2 = 0.1, df = 1, p = 0.77).

Second, we controlled for differences in DUP between EISs using lin-ear regression for the geometric mean, and logistic regression for theproportion of participants under 6 months. For established EIS, multipleregression with log (DUP) as the dependent variable found that neitherEIS nor year cohort was significant, whereas for new EIS, year cohortremained significant even after adjusting for differences between EISs(b=−0.12, t = 2.9, p= 0.003). For established EIS, logistic regressionon the proportion of patients with DUP under 6 months found thatneither EIS nor year cohort was significant, whereas for new EIS, yearcohort was significant when no adjustment was made for differencesbetween EISs (Wald= 7.9, df = 3, p= 0.048), but not when EIS differ-ences were accounted for (Wald = 2.3, df = 1, p = 0.51).

Third, we controlled for differences in ethnicity (white vs non-white) or place of birth (born in UK or elsewhere) between new andestablished EISs for all the analyses which tested the relationshipbetween time since establishment of EIS and both length of DUP andproportion of participants with DUP under 6 months. The results werevirtually unchanged.

4. Discussion

We found that the introduction of new early intervention serviceswas followed by a significant decrease in the duration of untreated psy-chosis and a significant increase in the proportion of patients treatedwithin six months of the onset of psychosis, in the catchment area ofthose services over the subsequent four years. These same parametersremained stable in the catchment areas of established early interventionservices. On both outcomes, new EIS started from a lower baseline thanestablished EIS, but gradually caught up. Since the EIS in our study hadno specific early detection components, the most plausible explanationis that after the introduction of an EIS into a catchment area, people pre-senting with first episode psychosis in that area receive more promptand proper treatment.

4.1. Strengths and limitations of the study

Themain limitation of our study is that it was not randomized, sowecannot rule out an unknown confounding factor associated with boththe late introduction of EIS to an area, and a falling rate of DUP in thesame area. Other possible confounding factors were a significant differ-ence in DUP length between teams and the fact that not all teams con-tributed data to all four annual cohorts. However, sensitivity analysessuggest that our findings are largely robust despite these limitations. Afurther limitation is that nearly 50% of potential participants declined

Please cite this article as: Marshall, M., et al., Impact of early interventionEDEN prospective cohort study, Schizophr. Res. (2014), http://dx.doi.org/1

to participate. However, our analysis of the limited available data onnon-participants suggests that they were similar to the participants.

Our study is one of the largest cohort studies of people with firstepisode psychoses conducted yet. It involves representative sam-ples of first episode patients from defined catchment areas includingthose not admitted to hospitals. It uses a standardized method ofassessing DUP, and it observes the real-world effects of fundamentallytransforming clinical practice on a large scale.

Our results were not consistent with the findings of a systematicreview that concluded that EISs were unlikely to reduce DUP unlessthey included an intensive public awareness campaign, which our EISdid not (Lloyd-Evans, 2011). However, this conclusion was based ondata from two before and after studies that examined DUP followingthe introduction of a single EIS without a public awareness campaign.In the first study from Ontario, Canada, DUP actually fell from a medianof 24.8 weeks to 11.6 weeks over a three-year period following theintroduction of an EIS, but the fall was non-significant, suggestingthat the study (n = 125) was underpowered (Scholten et al., 2003).The second study, from Copenhagen, was large enough to exclude aclinically significant effect on DUP (n = 578), but had a median DUP(52 weeks) over four times longer than National EDEN (Nordentoftet al., 2008). The authors attributed this very long DUP to differencesin measurement methods, but it could also reflect unknown but impor-tant differences in the care pathway for first episode psychosis betweenDenmark and other countries.

Our studywas not designed or powered to determine preciselywhatcaused the reduction in duration of untreated psychosis. However, onthe basis of other researchfindings it is possible to rule out two explana-tions. First, it is unlikely that the reduction was caused by earlier detec-tion of people with psychosis in the areas served by new EIS. We knowthis because the TIPs project has established that earlier detection canonly be achieved by a sustained campaign of public education, andnone of our participating EISs were engaging in such campaigns (Joaet al., 2008). Nor is it likely that the reductionwas caused by EIS encour-agingmore rapid referrals from primary to secondary care, as a large UKrandomized controlled trial (REDIRECT) has shown that even intensiveengagement of EIS with primary care does not result in more rapid re-ferrals, largely because referrals are already quite prompt (Lester et al.,2009) We therefore suspect that the explanation lies in the fact thatmost people in England with a first episode of psychosis have substan-tial contact with secondary mental health care services before they re-ceive treatment. For example, a recent study in Birmingham foundthat delay within secondary mental health services accounted for 35%of mean DUP (Nordentoft et al., 2008). We think it likely that the actof setting up EIS within the organizations that provide secondary men-tal health care has been a catalyst for improvement leading to a moreprompt response to first episode psychosis across the whole secondarysystemof care (Nordentoft et al., 2008). This conclusion contradicts pre-vailing opinion which asserts that EISs do not affect the duration of un-treated psychosis unless they actively seek out “undetected” psychosisin the community through public education campaigns or in-reachinto schools and colleges (Brunet et al., 2007; Lloyd-Evans et al., 2011).

4.2. Implications for clinicians and policy makers

Since the implementation of EIS in England, new evidence hasemerged in favour of the cost-effectiveness of the approach (McCroneet al., 2009), the more favourable illness course of those who receive it(Mihalopoulos, 2009), and the fact that EIS improves the engagementand treatment of young people with first episode psychosis comparedto traditional community mental health teams (National CollaboratingCentre for Mental Health, 2010). Our study adds to this evidence bysuggesting that the introduction of EIS in Englandmay have led to a sig-nificant fall in DUP. This is important for both health economies thinkingof introducing EIS and also those wonderingwhether, at a time of inter-national economic recession, to decommission them. It also begs the

services on duration of untreated psychosis: Data from the National0.1016/j.schres.2014.07.005

Page 6

6 M. Marshall et al. / Schizophrenia Research xxx (2014) xxx–xxx

question what further reductions in DUP might be achieved if EnglishEIS were commissioned to actively seek out “undetected” psychosis inthe community through public education campaigns or in-reach intoschools and colleges (Melle et al., 2004).

Ethical approval

Ethical approvalwas given by Suffolk Local Research Ethics Commit-tee, UK. REC reference number: 05/Q0102/44. All participants gave in-formed consent before taking part.

Data sharing

No additional data available.

Copyright licence

The corresponding author has the right to grant on behalf of allauthors and does grant on behalf of all authors, an exclusive licence(or non-exclusive for government employees) on a worldwide basis tothe BMJ Publishing Group Ltd and its Licensees to permit this article(if accepted) to be published in BMJ editions and any other BMJPGLproducts and sub-licences to exploit all subsidiary rights, as set out inour licence (http://resources.bmj.com/bmj/authors/checklists-forms/licence-for-publication).

Role of funding sourceThis research was funded by the National Institute of Health Research as part of the

National EDEN study (National Institute for Health Research Programme Grants for Ap-plied Research PO261680 and RP-PG-0109-10074). MB is part-funded by the National In-stitute for Health Research through the Collaborations for Leadership in Applied HealthResearch and Care for Birmingham and Black Country (CLAHRC-BBC) programme. Theviews expressed are those of the author(s) and not necessarily those of the NHS, theNIHR, or the Department of Health. The funding source had no role in study design, datacollection, analysis, interpretation, or writing of the report.

ContributorsAll authors contributed to the protocol for this work. All authors other than BT and NF

participated in the recruitment of participants. MM and BT analysed the data with assis-tance from NF, LE and NB. All authors contributed to the interpretation of the findings.MMwrote the first draft of themanuscript, which was subsequently edited by all authorswho have also approved the final version. All authors had full access to the data (includingstatistical results and tables) and take responsibility for the integrity of the data and accu-racy of the analysis. MM will act as guarantor.

Conflict of interestAll authors have completed the Unified Competing Interest form at www.icmje.org/

coi_disclosure.pdf (available on request from the corresponding author) and declare sup-port from the National Institute of Health Research for the submittedwork. IC is amemberof the national expert panel that gives advice to pharmaceutical companies including BMS,Astra Zeneca, Jansen, and Lilly from whom he has received meeting expenses in additionto payment for talks and lectures from BMS, Astra Zeneca, and Jansen. PJ is a member ofa scientific advisory board for Roche. There are no other financial relationships with anyorganizations that might have an interest in the submitted work in the past three years.No spouses, partners, or children have financial relationships that may be relevant to thesubmitted work. MM, MB, IC, NH and PJ have worked in early intervention services.

AcknowledgementsWe thank all the EIS staff who assisted in the study.Wewould like to thank Dr James

Kirkbride for the help with catchment area populations.

Please cite this article as: Marshall, M., et al., Impact of early interventionEDEN prospective cohort study, Schizophr. Res. (2014), http://dx.doi.org/

References

Addington, J., Van Mastriqt, S., Hutchinson, J., Addington, D., 2002. Pathways to care: helpseeking behaviour in first episode psychosis. Acta Psychiatr. Scand. 106, 358–364.

Barrett, E.A., Sundet, K., Faerden, A., 2010. Suicidality before and in the early phases of firstepisode psychosis. Schizophr. Res. 119, 11–17.

Bertelsen, M., Jeppesen, P., Petersen, L., 2009. Course of illness in a sample of 265 patientswith first-episode psychosis-five-year follow-up of the Danish OPUS trial. Schizophr.Res. 107, 173–178.

Bertolote, J., McGorry, P., 2005. Early intervention and recovery for young people withearly psychosis: consensus statement. Br. J. Psychiatry Suppl. 48, 116–119.

Birchwood, M., Lester, H., McCarthy, L., Jones, P., Fowler, D., Amos, T., Freemantle, N.,Sharma, V., Lavis, A., Singh, S., Marshall, M., 2014. The UK National evaluation of thedevelopment and impact of early intervention services (The National Eden Studies):study rationale, design and baseline characteristics. Early Interv. Psychiatry 8 (1),59–67.

Brunet, K.B., Lester, M., Thornhill, H.K., 2007. Delays in mental health services and dura-tion of untreated psychosis. Psychiatr. Bull. 31, 408–410.

The NHS Plan: A Summary. Department of Health, London.Improvement, Expansion and Reform: The Next 3 Years— Priorities and Planning Frame-

work 2003–2006. Department of Health, London.Joa, I., Johannessen, J.O., Auestad, B., Friis, S., McGlashan, T., Melle, I., Opjordsmoen, S.,

Simonsen, E., Vaglum, P., Larsen, T.K., 2008. The key to reducing duration of untreatedfirst psychosis: information campaigns. Schizophr. Bull. 34 (3), 466–472.

Joint Formulary Committee, 2005. British National Formulary, 49 ed. BMJ Group and Phar-maceutical Press, London.

Kay, S.R., Fiszbein, A., Opler, L.A., 1987. The positive and negative syndrome scale (PANSS)for schizophrenia. Schizophr. Bull. 13, 261–276.

Larsen, T.K., McGlashan, T.H., Moe, L.C., 1996. First-episode schizophrenia: I. Early courseparameters. Schizophr. Bull. 22, 241–256.

Lester, H., Birchwood, M., Freemantle, N., Michall, M., Tait, L., 2009. REDIRECT: clusterrandomised controlled trial of GP training in first-episode psychosis. Br. J. Gen.Pract. 59 (563), e183–e190.

Lloyd-Evans, B., Crosby, M., Stockton, S., Pilling, S., Hobbs, L., Hinton, M., Johnson, S., 2011.Initiatives to shorten duration of untreated psychosis: systematic review. Br. J. Psychi-atry 198, 256–263.

Marshall, M., Lewis, S., Lockwood, A., Drake, R., Jones, P., Croudace, T., 2005. Associationbetween duration of untreated psychosis and outcome in cohorts of first-episode pa-tients. Arch. Gen. Psychiatry 62, 975–983.

McCrone, P., Knapp, M., Dhanasiri, S., 2009. Economic impact of services for first-episodepsychosis: a decision model approach. Early Interv. Psychiatry 3, 266–273.

McGuffin, P., Farmer, A., Harvey, I., 1991. A polydiagnostic application of operationalcriteria in studies of psychotic illness: development and reliability of the OPCRIT sys-tem. Arch. Gen. Psychiatry 48, 764–770.

Melle, I., Larsen, T.K., Haahr, U., Friis, S., Johannessen, J.O., Opjordsmoen, S., Simonsen, E.,Rund, B.R., Vaglum, P., McGlashan, T., 2004. Reducing the duration of untreated first-episode psychosis: effects on clinical presentation. Arch. Gen. Psychiatry 61, 143–150.

Mihalopoulos, C., 2009. Why invest in early psychosis intervention services? Early Interv.Psychiatry 3, 237–238.

National Collaborating Centre for Mental Health, 2010. Schizophrenia: core interventionsin the treatment and management of schizophrenia in adults in primary and second-ary care. The British Psychological Society and the Royal College of Psychiatrists(Report No CG 82).

Nielssen, O., Large, M., 2008. Rates of homicide during the first episode of psychosis andafter treatment: a systematic review andmeta-analysis. Schizophr. Bull. 36, 702–712.

Nielssen, O., Yee, N., Millard, M., Large, M., 2011. Comparison of first-episode and previ-ously treated persons with psychosis found NGMI for a violent offense. Psychiatr.Serv. 62, 759–764.

Nordentoft, M., Thorup, A., Petersen, L., Ohlenschlaeger, J., Christensen, T.O., Krarup, G.,Jorgensen, P., Jeppesen, P., 2008. Does a detection team shorten duration of untreatedpsychosis? Early Interv. Psychiatry 2 (1), 22–26.

Norman, R.M., Malla, A.K., 2001. Duration of untreated psychosis: a critical examination ofthe concept and its importance. Psychol. Med. 31, 381–400.

Perkins, D.O., Gu, H., Boteva, K., Lieberman, J.A., 2005. Relationship between duration ofuntreated psychosis and outcome in first-episode schizophrenia: a critical reviewand meta-analysis. Am. J. Psychiatry 162, 1785–1804.

Scholten, D., Malla, A., Norman, R., McLean, T.S., McIntosh, E.M., McDonald, C.L., Eliasziw,M., Speechley, K.N., 2003. Removing barriers to treatment of first episode psychoticdisorders. Can. J. Psychiatr. 48 (8), 561–565.

services on duration of untreated psychosis: Data from the National10.1016/j.schres.2014.07.005