Rasch analysis of the SAMHSA recovery inventory for Chinese (SAMHSA-RIC)

Abstract

Aims:
To generate a short version of a newly developed inventory that adopted the conceptual framework of the Substance Abuse and Mental Health Services Administration (SAMHSA) consensus statement on recovery.

Methods:
Through Rasch analysis, this paper presents how this recovery inventory (SAMHSA-RIC), with its original 111 items, can be reduced to a much shorter version with only 41 items.

Results:
Although internal consistency is slightly lowered because of item reduction, the short version maintains satisfactory and significant correlations with quality of life measures. Overall, the canonical correlation between the scale and WHOQOL-BREF was virtually the same, with only a 0.2% decrease.

Conclusions:
SAMHSA-RIC (short version) has strong potential to become a general tool for evaluating rehabilitative services for persons with persistent and severe mental illness. A validation study of the short version with clinical samples is warranted.

Full text

International Journal of
Social Psychiatry
2014, Vol. 60(3) 254 –262
© The Author(s) 2013
Reprints and permissions:
sagepub.co.uk/journalsPermissions.nav
DOI: 10.1177/0020764013485327
isp.sagepub.com
E
CAMDEN
SCHIZOPH
Introduction
The Substance Abuse and Mental Health Services
Administration (SAMHSA) statement on recovery pre-
sented a comprehensive framework for the consumer-based
recovery concept, and signalled the gradual extension of
recovery from a pure medical concern of symptom control
and restoration of functions to a personal and psychosocial
process. The statement identified 10 components as essen-
tial to recovery. These components not only covered aspects
such as clinical improvement and functional normalization,
but also consumers’ subjective experiences of optimism,
empowerment, interpersonal support, peer support and
stigma reduction (SAMHSA, 2005, 2006). This consumer-
based recovery model is considered closer to the lived
experience of recovery among many who suffered from
persistent and severe mental illness (PSMI). However, the
general recovery framework is so all encompassing that a
great deal of empirical work still remains to be done regard-
ing its many facets.
Empirical research into recovery can be broadly divided
into two different lines: one focusing on the demarcation
of recovery stages and the development of inventories to
identify these stages; the other, a process model without
assumption of stages. For the stages model, Andresen,
Oades and Caputi (2003) presented an overall review of
existing major studies on recovery stages and developed
the five-stage model for recovery based on its conclusions:
(1) moratorium; (2) awareness; (3) preparation; (4) rebuild-
ing; and (5) growth. This model encompassed the several
three- or four-stage models that had been previously cre-
ated. However, the model must refer to the variable fea-
tures of each stage before they become clinically
meaningful. A common problem with stage division by
cluster analysis is the overlapping features of adjacent
stages, which make certain clusters easily absorbed by oth-
ers. For this probable reason, Andresen, Caputi and
Oades’s (2006) later empirical study on stage modelling,
which includes measures of mental health, psychological
well-being, hope, resilience and recovery, named three
major stages instead of five.
The other line of research work attempts to test out
major recovery components without making assumptions
Rasch analysis of the SAMHSA Recovery
Inventory for Chinese (SAMHSA-RIC)
Marcus Yu Lung Chiu,1 Frank Ho Ting Wong2 and
Winnie Wing Nan Ho3
Abstract
Aims: To generate a short version of a newly developed inventory that adopted the conceptual framework of the
Substance Abuse and Mental Health Services Administration (SAMHSA) consensus statement on recovery.
Methods: Through Rasch analysis, this paper presents how this recovery inventory (SAMHSA-RIC), with its original 111
items, can be reduced to a much shorter version with only 41 items.
Results: Although internal consistency is slightly lowered because of item reduction, the short version maintains
satisfactory and significant correlations with quality of life measures. Overall, the canonical correlation between the scale
and WHOQOL-BREF was virtually the same, with only a 0.2% decrease.
Conclusions: SAMHSA-RIC (short version) has strong potential to become a general tool for evaluating rehabilitative
services for persons with persistent and severe mental illness. A validation study of the short version with clinical
samples is warranted.
Keywords
Recovery, Chinese, SAMHSA, HRQOL, psychometrics, Rasch analysis
1Department of Social Work, National University of Singapore, Singapore
2 Department of Geography, The University of Hong Kong, Hong Kong,
PR China
3Beat Drugs Fund Association, Hong Kong, PR China
Corresponding author:
Marcus Yu Lung Chiu, Department of Social Work, Faculty of Arts &
Social Sciences, Block AS3, Level 4, 3 Arts Link, 117570 Singapore.
Email: mchiu@nus.edu.sg
485327ISP60310.1177/0020764013485327International Journal of Social PsychiatryChiu et al.
2013
Article

Chiu et al. 255
about stages and their associated features. Measures in the
major recovery components can be directly compared and a
sensible empirical model can be built and tested. The
SAMHSA model of recovery, although comprehensive,
requires tremendous effort to operationalize, because many
of its building blocks were considered subjective and vague
(Bellack, 2006). Chiu, Ho, Lo and Yiu (2010) operational-
ized the SAMHSA recovery concept, and based on 11 exist-
ing psychometric scales or sub-scales, they developed a
111-item Chinese Recovery Inventory (SAMHSA-RIC) to
measure the recovery of 200 community-residing people
with schizophrenia spectrum disorders. These items were
found to correlate significantly with health-related quality
of life (HRQOL) measures, and subsequent analysis of its
best-fit factor structure by using structured equation model-
ling technique showed that the final structural model
explained as much as 81% variance in the quality of life
measures. HRQOL was chosen in the study because of its
popularity in schizophrenia research and its proximity to
the term ‘recovery’ from the consumer movement (Ho,
Chiu, Lo & Yiu, 2010). The high percentage of variance
explained by the structural model helped to reveal the rela-
tion between the recovery components and HRQOL,
although originally we only had a unified statement for the
concept of recovery instead of its linkage with HRQOL.
The developed scale did not refer to the Psychosis Recovery
Scale (Chen, Tam, Wong, Law & Chiu, 2005) because it
has a different notion of recovery (i.e. clinical recovery),
nor to existing recovery scales such as the Recovery
Assessment Scale (RAS) (Corrigan, Giffort, Rashid, Leary
& Okeke, 1999; Corrigan, Salzer, Ralph, Sangster & Keck,
2004; Song & Hsu, 2011), which adopted only part rather
than the whole of the SAMHSA recovery framework. For
example, RAS also tapped on reliance on others and will-
ingness to seek help.
Back in the local Asian context of a generally under-
resourced and understaffed mental health system (Chiu,
2012; Tse, Siu & Kan, 2011) and the overwhelming con-
cern over symptom control in conventional psychiatry, it is
not surprising for mental health workers to be unfamiliar
with the recovery concept and how these recovery princi-
ples could be implemented (Mak, Lam & Yau, 2010). Even
though there has been the initial suggestion of the use of
user participation and the grooming of leaders from within
service users as a means to promote recovery practice (Tse,
Cheung, Kan, Ng & Yau, 2012), most discussions are con-
ceptual-based rather than evidence-based. It is fair to say
that the recovery concept issue is still quite new in Hong
Kong and elsewhere in Asia.
Although some researchers once doubted if the individ-
ualistic American concept of recovery could be used for
societies in other cultures (Tse, 2004; Yee, 2003), the latest
work in this line of research has lent empirical support to
the SAMHSA recovery concept (Chiu et al., 2010; Ho
et al., 2010). Recovery components like hope, desire for
personal agency, frustration with stigmas, and so on, are
probably universal experiences during the recovery jour-
ney, irrespective of cultural differences.
These empirical studies, though promising, have yet to
overcome the various difficulties associated with
SAMHSA’s lengthy list of items. The many items of the
inventory incur a considerable administration cost and
probably would not allow those with limited memory and
concentration to take part in the study. This possible exclu-
sion may deprive the researchers of the opportunity to test
their models on lower-functioning persons with PSMI.
Moreover, the 111-item inventory is so long that it is not
considered optimal for frequent clinical usage. Therefore,
there is both a practical and a research need to develop a
shorter version.
Rasch model
In this research, Rasch analysis was conducted to reduce
the number of items on SAMHSA-RIC. As the response to
the items in the scale was given in ordinal outcomes, this
study employed the polytomous Rasch model, which is a
generalization of the dichotomous Rasch model (RUMM,
2005). The dichotomous Rasch model is considered to be
the simplest, with a one-item parameter equivalent to an
item-response theory model (Meads & Bentall, 2008).
The polytomous Rasch model assumes that the probabil-
ity of a given response of an item is a logistic function of
the relative distance between the item location and the
respondent location on a linear scale. Mathematically, it
can be written as:
PX xe
e
ni
xk
x
ki
x
mx
ni
i
ni
k
x
ki
=
()
=
∑
∑
−
()
−=
=
−
()
−∑=
βδ
βδ τ
τ
1
0
1
′
′
′
where βn and δi are the location of the nth person and ith
item respectively. τkik=1,2,…,mi are the thresholds that par-
titioned the latent continuum of item i into mi+1 ordered cat-
egories. Xni is the random variable of item score. For a
polytomous Rasch model involving three category coeffi-
cients 0, 1, and 2, we have:
PX
ee
ni
iniii
ni
=
()
=
++
−+−−−
+−
()
01
1112
2
τβδττ βδ
PX e
ee
ni
ini
iniii
ni
=
()
=
++
−+−
−+−−−
+−
()
1
1
1
112
2
τβδ
τβδττ βδ
PX e
ee
ni
ii ni
iniii
ni
=
()
=
++
−−+−
()
−+−−−
+−
(
2
1
12
112
2
2
ττ βδ
τβδττ βδ
))

256 International Journal of Social Psychiatry 60(3)
One of the characteristics of Rasch analysis is locating
the responses of each respondent on an item-person map
according to the above probabilistic relation between the
items’ location βn (difficulty’) and persons’ locationδi
(‘ability’). This enables further analysis of items’ dis-
criminative power on the respondents by using the item-
person map (Figure 1). Moreover, in contrast to the
traditional modelling approach, which normally requires
tuning the model parameters to fit the data, the Rasch
analysis requires the data to fit the Rasch model. The
Rasch analysis provides different kinds of statistical indi-
ces for assessing the fitness of the data to the Rasch
model, such as individual items’ χ2 fit statistics, item
residual score, and item–trait interaction χ2 fit statistics.
Once the fitness of the data has been confirmed, their uni-
dimensionality will also be confirmed. Hence, the item
scores can justifiably be added together as a single com-
prehensive total score (Bond, 2007).
Item reduction by means of Rasch analysis is not uncom-
mon in other research fields. In psychology, Meads and
Bentall (2008) shortened the 48-item Hypomanic
Personality Scale to 20 items, while the Cronbach’s α was
maintained at a high level (α = 0.8). Moreover, Dreer et al.
(2009) also successfully shortened the 25-item Social
Problem Solving Inventory (Revised Scale) to 10 items and
uni-dimensionality was also confirmed on the 10-item
scale. Similar analysis using the Rasch model can also be
found in the field of optometry (Pesudovs, Garamendi,
Keeves & Elliott, 2003; Ryan, Court & Margrain, 2008),
rehabilitation (Siegert, Tennant & Turner-Stokes, 2010;
Vidotto, Carone, Jones, Salini & Bertolotti, 2007) and edu-
cation (Waugh, 2010). There is a sizeable volume of litera-
ture that provides a comprehensive review on the Rasch
model and its application (Andrich, 2011; da Rocha,
Chachamovich, de Almeida Fleck & Tennant, 2013;
Hagquist, Bruce & Gustavsson, 2009), and the model’s
increasing popularity demonstrates that Rasch analysis
might be an alternative method of item reduction for
SAMHSA-RIC on top of the conventional use of factor
analysis.
Methods
Sample
A total of 204 eligible participants were recruited from
two psychiatric outpatient clinics in Hong Kong. The
inclusion criteria were: (1) aged 18–60; and (2) primary
diagnosis of ‘schizophrenia’, ‘schizo-affective’ or ‘schiz-
ophreniform’ disorder. The exclusion criteria were: (1)
inability to communicate in mother tongue (Cantonese);
(2) global score less than 4 in the Capacity to Report
Subjective Quality of Life (CapQOL) inventory screening
assessment (a score of 4 or above indicates the ability to
complete QOL measures and give valid and reliable
answers (Wong et al., 2005)); and (3) discharged from the
psychiatric ward during the 30 days preceding the interview.
Written consent was obtained from the participants after
Figure 1. Person-item map of Adult State Hope Scale items.

Chiu et al. 257
they were given a detailed description of the study. Ethics
approval of the original study had been obtained from the
Hospital Authority Cluster Research Ethics Committees
before collecting the data. This Rasch analysis is essen-
tially an extended study, this time examining only the
measurement items.
Instruments
Eleven (sub-)scales were included in the Rasch analysis.
All of the scales were filled out by trained research staff
through face-to-face interview. These scales included
Adult State Hope Scale (ASHS) (Snyder et al., 1996),
Recovery Attitude Questionnaire (RAQ-7)(Ralph, Kidder
& Philips, 2000), Health Care Climate Questionnaire
(HCCQ) – a scale assessing the degree of autonomy sup-
port that clients perceive their psychiatrists to provide
(Williams, Rodin, Ryan, Grolnick & Deci, 1998), Self-
responsibility sub-scale of the Exercise of Self Care
Agency scale (ESCA; Kearney & Fleischer, 1979; Riesch
& Hauck, 1988), the personal competence sub-scale of
the Resilience Scale (RS) (Wagnild & Young, 1993;
Bengtsson-Tops, 2004; Rosenfield, 1992), the self-esteem,
self-efficacy sub-scale of the Making Decision
Empowerment Scale (MDES; Rogers, Chamberlin,
Ellison & Crean, 1997), the alienation, perceived discrim-
ination and the social withdrawal sub-scale of the
Internalized Stigma of Mental Illness (ISMI) scale
(Ritsher, Otilingam & Grajales, 2003).
Holistic recovery was assessed in three aspects: (1) psy-
chosocial symptoms (mind and emotion); (2) social sup-
port (community); and (3) spirituality. The frequency of
psychosocial symptoms was measured by a 15-item psy-
chosocial sub-scale of the Schizophrenia Quality of Life
Scale (SQLS; Wilkinson et al., 2000). Social support was
measured by the Multidimensional Scale of Perceived
Social Support-Chinese version (MSPSS-C) (Chou,
2000). Spirituality was measured by the World Health
Organization Spirituality Religion and Personal Belief
Scale Hong Kong version (WHOQOL-SRPB-HK) )
(WHOQOL SRPB Group, 2006). Among the eight facets
of WHOQOL-SRPB-HK, only three were selected for this
study: connectedness to a spiritual being or force, spiritual
strength and faith sub-scales. This was because only these
three provide a ‘pure’ measure of a respondent’s spiritual-
ity; that is, they are not complicated by a respondent’s
mental health condition (Moreira-Almeida & Koenig,
2006). Respondents’ quality of life was measured by the
Hong Kong Chinese World Health Organization Quality
of Life Measure abbreviated version (WHOQOL-
BREF(HK)). The WHOQOL perception among people
with schizophrenia has been found to be negatively cor-
related with psychiatric ratings (Chan, Ungvari, Shek &
Leung, 2003; Chan & Yu, 2004).
All scales were translated directly from English to
Cantonese by an experienced linguist, except the
MSPSS-C and WHOQOL-SRPB-HK, which were already
in Chinese. The English version of ISMI, MDES, MS,
SQLS and RAQ-7 were validated in psychiatric samples,
whereas RS, ESCA, ASHS and HCCQ were validated in
healthy samples. The reasons for the choice of scales and
the reliability and validity issues have already been
reported elsewhere (please refer to Chiu et al, 2010 for
details).
General steps of Rasch analysis
The Rasch analysis was conducted using the RUMM 2020
software (http://www.rummlab.com.au) on each of the 11
sub-scales of SAMHSA-RIC individually. The 204 partici-
pants in this study was considered an adequate number
because Linacre (1994) showed that a sample size of 27–61
gives 99% confidence of item-calibration stability within a
logit, which is accurate enough.
In general, four item-reduction steps were followed
throughout the analysis. First, the individual item’s χ2 fit
statistics were checked to ensure that they fitted the Rasch
model. Any item with a p-value less than .05 was consid-
ered to be a misfit item and was discarded from the scale.
Moreover, researchers also checked the item residual score,
which represents the error on the fit of the data to the model
from the perspective of the item (RUMM, 2005). As a rule
of thumb, residuals greater than ±2.5 were considered to be
a misfit to the Rasch model and would be discarded if the
item had significant item χ2 fit statistics.
Second, differential item functioning (DIF) by gender
was evaluated to ascertain if any item was subject to gender
bias. Similar to individual-item χ2 fit statistics, a DIF with a
p-value less than .05 was considered to be significant and
the item would be discarded.
Third, the item-person maps of the scales were checked
in order to ensure that the subjects’ responses were evenly
spread out, so that the scales could distinguish between
subjects at different levels of recovery.
Finally, item–trait interaction χ2 fit statistics were
checked for the remaining chosen items after deletion.
Insignificant χ2 statistics were obtained and ensured the
overall fit of the Rasch model, which also implied the uni-
dimensionality of the scale. Once the uni-dimensionality
had been confirmed, the item scores could justifiably be
added together (Bond, 2007).
Result validation
A series of pre-post comparisons were done before and
after item reduction in order to ascertain the quality of the
shortened SAMHSA-RIC. First, the change of Cronbach’s
α, Person Separation Index and item–trait interaction χ2 fit
statistics were assessed to ensure scale reliability. Second,

258 International Journal of Social Psychiatry 60(3)
canonical correlation analysis was conducted between
SAMHSA-RIC and WHOQOL-BREF before and after
item reduction so that the stability of their relationship
could be tested.
Results
An example of item reduction
In order to illustrate the item-reduction process but to save
the lengthy repetition on all items, ASHS was chosen to
show reduction steps. First of all, the χ2 fit statistics of the
six items were checked. As items 2, 5 and 6 had significant
p-values of .0083, .0210 and .0006, respectively, they
should be removed from the scale. However, after deleting
the items, Cronbach’s α dropped from 0.78 to 0.56 and was
not considered acceptable. As item 5 had the highest
p-value among the three, it was selected to be retained in
the scale and Cronbach’s α thereby increased to 0.69. For
this set of items (i.e. items 1, 3, 4 and 5), the item residual
scores were within the acceptable range, no significant DIF
statistics were found, item–trait interaction χ2 fit statistics
showed the overall fit of the Rasch model and the items in
the person-item map basically covered nearly all responses
(Figure 1). Hence, items 1, 3, 4 and 5 formed the new group
after item reduction.
The items of ASHS before and after item reduction are
listed in Table 1. The original version of ASHS had six
items that could be divided into two groups of questions:
items 1, 3 and 5 were related to pathways (i.e. belief in
one’s capacity to generate routes); and items 2, 4, and 6
were related to agency (i.e. belief in one’s capacity to initi-
ate and sustain actions) (Snyder et al., 1996). The result of
item reduction showed that items 2 and 6, related to agency,
were deleted from the scale. Although retaining item 4 did
not meet our expectation because it was also related to
agency, the item-person map (Figure 1) suggested that the
discriminant power of item 4 was weak. However, it was
retained in the scale because deleting it would have low-
ered Cronbach’s α from 0.69 to 0.61, an unacceptable level
for clinical use.
Number of items reduced
After item reduction by Rasch analysis, the number of
items of each sub-scale of SAMHSA-RIC was reduced.
Although some only had minor reductions from six to five
items, many long scales were shortened to half of their
original length. Overall, the Chinese recovery scale was
shortened from 111 to 41 items, around just one-third of its
original length (Table 2).
Scale reliability
From Table 3 we see that the item–trait interaction χ2 fit
statistics of five sub-scales have a p-value less than .05; this
means that the data of the scales did not fit the Rasch model.
However, after item reduction, all p-values were above .05,
and hence the reduced scales fit the Rasch model. All of the
sub-scales had different degrees of decrease in Cronbach’s
α. However, the final α values were still within the accept-
able range (α > 0.6) and suitable for clinical use (Moss
et al., 1998). In addition, some of the sub-scales, such as
WHOQOL-SPRB, maintained a high level of internal con-
sistency (α = 0.89). The result of the Person Separation
Index was similar to that of Cronbach’s α, with all values
also being above 0.6. Actually, such a result is expected
because the Person Separation Index is considered as an
analogue of Cronbach’s α (Curt, 2007; Piquero, Macintosh
& Hickman, 2002).
Canonical correlation analysis between
SAMHSA-RIC and WHOQOL-BREF
Table 4 shows the result of canonical correlation analysis
between SAMHSA-RIC and WHOQOL-BREF before and
after item reduction. The overall canonical correlation,
which is the correlation between the canonical variates of
SAMHSA-RIC and WHOQOL-BREF, was virtually the
same with only a 0.2% decrease. Regarding the cross-
canonical loadings, which is the correlation between a sub-
scale score and the opposite scale’s canonical variate (e.g.
the correlation between the sub-score of hope in SAMHSA-
RIC and the canonical variate of WHOQOL-BREF), seven
out of the 16 cross-canonical loadings were strengthened.
For the nine weakened loadings, only three belonging to
SAMHSA-RIC had a decrease larger than 5% after item
reduction. They represented holistic well-being/social sup-
port, empowerment and holistic well-being/spirituality,
which had a decrease of 14.8%, 11.3% and 7.8%, respec-
tively. We considered the decreases to be acceptable
because the magnitude of the three loadings after item
reduction was still above the standard acceptable threshold
of 0.3.
Discussion
The study successfully shortened the 111-item SAMHSA-
RIC to 41 items while maintaining the scope of the original
scale, with the trade-off of a slightly lowered internal con-
sistency because of the significant reduction in the number
of questions. This demonstrated, similar to Meads and
Bentall’s (2008) study, that although many original scales
have good internal consistency, item reduction would
inevitably reduce the level of internal consistency.
Nevertheless, the result of canonical correlation analysis
provides a more comprehensive picture on the effect of the
item reduction. The magnitude of the canonical correlation
after item reduction was virtually the same as before.
Generally, around half of the cross-canonical loadings
obtained were stronger than the original scale. In other

Chiu et al. 259
words, for the short version some sub-scales have a slightly
increased loading with the opposite scale’s canonical vari-
ate, while some sub-scales have a slightly decreased load-
ing. This balanced picture has provided us with better
confidence that the short version’s efficacy will not be sig-
nificantly reduced. Removing some of the less effective
items actually not only caused less of a burden for the
respondents, but also added strength to some sub-scales
Table 1. Items of ASHS before and after item reduction.
Item Item description Decision
1 I can think of many ways to get out of a jam retained
2 I energetically pursue my goals deleted
3 There are lots of ways around any problem retained
4 I have been pretty successful in my life retained
5 I can think of many ways to get the things in life that are most important to me retained
6 I meet the goals that I set for myself deleted
Table 2. Number of items in the SAMHSA-RIC before and after item reduction.
(Sub-)scales Number of items
Before item reduction After item reduction
Hope: ASHS 6 4
Non-linear recovery: RAQ-7 4 2
Person centred: HCCQ 6 3
Self-responsibility: ESCA-ISR 10 4
Strength based: RS-PC 6 5
Self-direction: MS 5 4
Empowerment: MDES-SESE 18 4
Respect-stigma: ISMI 17 6
Holistic well-being / social support: MSPSS 12 2
Holistic well-being / spirituality: WHOQOL-SPRB 12 4
Holistic well-being /psychosocial symptoms: SQLS 15 3
Total 111 41
Table 3. Scale reliability indices of the SAMHSA-RIC before and after item reduction.
Scale Cronbach’s αPerson Separation Index p-value of the item–trait
interaction χ2 fit statistics
Before item
reduction
After item
reduction
Before item
reduction
After item
reduction
Before item
reduction
After item
reduction
Hope: ASHS 0.78 0.69 0.83 0.75 < .001 .09
Non-linear recovery: RAQ-7 0.65 0.62 0.71 0.57 .2 .17
Person centred: HCCQ 0.85 0.74 0.86 0.75 .23 .87
Self-responsibility: ESCA-ISR 0.84 0.74 0.84 0.75 < .001 .28
Strength based: RS-PC 0.77 0.70 0.8 0.73 .61 .75
Self-direction: MS 0.69 0.64 0.68 0.63 .52 .83
Empowerment: MDES-SESE 0.92 0.71 0.91 0.70 .42 .12
Respect / stigma: ISMI 0.91 0.82 0.92 0.80 .01 .94
Social support: MSPSS 0.92 0.70 0.94 0.83 < .001 .09
Spirituality: WHOQOL-SPRB 0.95 0.89 0.96 0.90 .13 .71
Psychosocial symptoms: SQLS 0.93 0.75 0.93 0.76 < .001 .11
All items 0.96 0.90 N/A N/A N/A N/A

260 International Journal of Social Psychiatry 60(3)
like hope, person-centred recovery, self-responsibility and
personal strength in terms of the relations with the
WHOQOL-BREF’s canonical variate. Selecting items (by
deletion) and putting them under the SAMHSA recovery
framework may render some of the question items less
effective than others in measuring all the facets of recov-
ery that the original full scale intended to measure.
However, the remaining items that survived the deletion
now have a new mission, which is to measure the recovery
concept rather than individual traits.
Limitations
Obviously, one of the limitations of this research was the
aspect of uni-dimensionality (Falissard, 1999). Because the
11 sub-scales of SAMHSA-RIC originally consisted of 11
independent scales that may be used for other purposes, the
number of options in each scale was different (ranging from
four to eight). The different number of items in each sub-
scale made fitting the whole scale into the Rasch model as
a block unfeasible and the issue of uni-dimensionality
therefore could not be tested. The use of a single total score
to represent the extent of recovery, although technically
feasible, would not be justifiable. It will be meaningful
only when the response range of all items has been aligned
to the same scoring options.
Another limitation is related to the nature of sample sub-
jects. The participants were consecutive pickups instead of
random samples of those suffering from schizophrenia.
Their age covered a wide range from 20 to 60, with a mean
of 42. The sample characteristics will be very different
from a younger cohort who in general had a shorter period
of delay in first consultation and treatment, and who had a
better treatment outcome on symptoms and disruption.
What this study represents may be those with more chro-
nicity and might not be applicable to younger groups.
Unfortunately, the spread of age range and the resulting
small number of subjects per group do not allow us to ana-
lyse the age effects comfortably.
Conclusion
In connection to this aspect of the survey, one future
research direction is to align the number of options for all
sub-scales and re-analyse them with the Rasch model. Once
uni-dimensionality has been confirmed by Rasch analysis,
the score of the items can be justifiably added together and
a comprehensive total score representing the overall
recovery stage can then be obtained. On the other hand, it is
also necessary to test out in subsequent validation study
whether a straightforward single total score is possible or
not. The total score will not only be useful for clinicians,
but also serve as a direct reference index for persons with
PSMI to understand their recovery progress. At the current
stage, however, it is suggested to calculate separate scores
for each of the 11 sub-scales as the overall uni-dimension-
ality has not been confirmed.
To make the survey a valid and handy tool to measure
consumer-oriented recovery (Bellack, 2006), a validation
study is warranted to establish its divergent, convergent and
predictive validity. If proven valid, a user-friendly tool like
SAMHSA-RIC would allow many health care and social
care programmes to be evaluated beyond the traditional
outcome indicators of symptom control and hospital read-
missions, and possibly serve as an important criterion of
psychiatric rehabilitation (Schrank & Slade, 2007).
Table 4. Cross-canonical correlation analysis between SAMHSA-RIC and WHOQOL-BREF before and after item reduction.
Variables Before item
reduction
After item
reduction
% difference
Canonical correlation Scale Domain 0.87 0.87 -0.2
Cross-canonical loadings SAMHSA-RIC Hope: ASHS 0.58 0.60 3.7
Non-linear recovery: RAQ-7 0.41 0.39 −4.7
Person centred: HCCQ 0.43 0.45 6.1
Self-responsibility: ESCA-ISR 0.58 0.61 4.1
Strength based: RS-PC 0.55 0.55 0.4
Self-direction: MS 0.56 0.55 −1.3
Empowerment: MDES-SESE 0.70 0.62 −11.3
Respect / stigma: ISMI −0.63 −0.61 −4.0
Social support: MSPSS 0.46 0.39 −14.8
Spirituality: WHOQOL-SPRB 0.35 0.32 −7.8
Psychosocial symptoms: SQLS −0.66 −0.63 −5.0
WHOQOL-BREF Overall 0.66 0.67 1.8
Physical health 0.69 0.71 2.6
Psychological 0.84 0.84 −0.8
Social 0.59 0.58 −1.9
Environmental 0.61 0.61 0.5

Chiu et al. 261
Funding
This research received no specific grant from any funding agency
in the public, commercial or not-for-profit sectors.
References
Andresen, R., Caputi, P., & Oades, L. (2006). Stages of recovery
instrument: Development of a measure of recovery from seri-
ous mental illness. Australian and New Zealand Journal of
Psychiatry, 40, 972–980.
Andresen, R., Oades, L., & Caputi, P. (2003). The experience of
recovery from schizophrenia: Towards an empirically vali-
dated stage model. Australian and New Zealand Journal of
Psychiatry, 37, 586–594.
Andrich, D. (2011). Rating scales and Rasch measurement.
Expert Review of Pharmacoeconomics & Outcomes
Research, 11, 571–585.
Bellack, A. S. (2006). Scientific and consumer models of recovery
in schizophrenia: Concordance, contrasts, and implications.
Schizophrenia Bulletin, 32, 432–442.
Bengtsson-Tops, A. (2004). Mastery in patients with schizo-
phrenia living in the community: Relationship to sociode-
mographic and clinical characteristics, needs for care and
support, and social network. Journal of Psychiatric and Men-
tal Health Nursing, 11, 298–304.
Bond, T. G. (2007). Applying the Rasch model: Fundamental
measurement in the human sciences, 2nd edn. Mahwah, NJ:
Lawrence Erlbaum.
Chan, G., Ungvari, G. S., Shek, D. T. L., & Leung, J. J. P. (2003).
Hospital and community-based care for patients with chronic
schizophrenia in Hong Kong: Quality of life and its corre-
lates. Social Psychiatry and Psychiatric Epidemiology, 38,
196–203.
Chan, S., & Yu, I. W. (2004). Quality of life of clients with schizo-
phrenia. Journal of Advanced Nursing, 45, 72–83.
Chen, E. Y., Tam, D. K., Wong, J. W., Law, C. W., & Chiu, C. P.
(2005). Self-administered instrument to measure the patient’s
experience of recovery after first-episode psychosis: Develop-
ment and validation of the Psychosis Recovery Inventory. Aus-
tralian and New Zealand Journal of Psychiatry, 39, 493–499.
Chiu, M. Y. L. (2012). Recovery and advocacy: Contextualiz-
ing justice in relation to recovery from mental illness in East
Asia. In A. Rudnick (ed.), Recovery of people with mental ill-
ness: Philosophical and related perspectives (pp. 279–303).
Oxford: Oxford University Press.
Chiu, M. Y. L., Ho, W. W. N., Lo, W. T. L., & Yiu, M. G. C.
(2010). Operationalization of the SAMHSA model of recov-
ery: A quality of life perspective. Quality of Life Research,
19, 1–13.
Chou, K. L. (2000). Assessing Chinese adolescents’ social sup-
port: The Multidimensional Scale of Perceived Social Sup-
port. Personality and Individual Differences, 28, 299–307.
Corrigan, P. W., Giffort, D., Rashid, F., Leary, M., & Okeke, I.
(1999). Recovery as a psychological construct. Community
Mental Health Journal, 35, 231–239.
Corrigan, P. W., Salzer, M., Ralph, R. O., Sangster, Y., & Keck,
L. (2004). Examining the factor structure of the recovery
assessment scale. Schizophrenia Bulletin, 30, 1035–1041.
Curt, H. (2007). The psychometric properties of the self-
reported SDQ: An analysis of Swedish data based on the
Rasch model. Personality and Individual Differences, 43,
1289–1301.
da Rocha, N. S., Chachamovich, E., de Almeida Fleck, M.P.,
& Tennant, A. (2013). An introduction to Rasch analysis
for psychiatric practice and research. Journal of Psychiatric
Research, 47, 141–148.
Dreer, L. E., Berry, J., Rivera, P., Snow, M., Elliott, T. R., Miller,
D., & Little, T. D. (2009). Efficient assessment of social prob-
lem-solving abilities in medical and rehabilitation settings:
A Rasch analysis of the Social Problem-Solving Inventory-
Revised. Journal of Clinical Psychology, 65, 653–669.
Falissard, B. (1999). The unidimensionality of a psychiatric scale:
A statistical point of view. International Journal of Methods
in Psychiatric Research, 8, 162–167.
Hagquist, C., Bruce, M., & Gustavsson, J. P. (2009). Using the
Rasch model in nursing research: An introduction and illus-
trative example. International Journal of Nursing Studies, 46,
380–393.
Hansson, L., & Bjorkman, T. (2005). Empowerment in people
with a mental illness: Reliability and validity of the Swedish
version of an empowerment scale. Scandinavian Journal of
Caring Sciences, 19, 32–38.
Ho, W. W. N., Chiu, M. Y. L., Lo, W. T. L., & Yiu, M. G. C.
(2010). Recovery components as determinants of the health-
related quality of life among patients with schizophrenia:
Structural equation modelling analysis. Australian & New
Zealand Journal of Psychiatry, 44, 71–84.
Kearney, B. Y., & Fleischer, B. J. (1979). Development of an
instrument to measure exercise of self-care agency. Research
in Nursing & Health, 2, 25–34.
Linacre, J. M. (1994). Sample size and item calibration stability.
Rasch Measurement Transactions, 7, 328.
Mak, W. W. S., Lam, B. Y. H., & Yau, S. S. W. (2010). Recov-
ery knowledge and recovery-oriented services in Hong
Kong. Psychiatric Services, 61, 1164. DOI: 10.1176/appi.
ps.61.11.1164.
Meads, D. M., & Bentall, R. P. (2008). Rasch analysis and item
reduction of the hypomanic personality scale. Personality and
Individual Differences, 44, 1772–1783.
Moreira-Almeida, A., & Koenig, H. G. (2006). Retaining the
meaning of the words religiousness and spirituality: A com-
mentary on the WHOQOL SRPB group’s ‘a cross-cultural
study of spirituality, religion, and personal beliefs as compo-
nents of quality of life’ (62: 6, 2005, 1486–1497). Social Sci-
ence & Medicine, 63, 843–845.
Moss, S., Prosser, H., Costello, H., Simpson, N., Patel, P., Rowe,
S., … Hatton, C. (1998). Reliability and validity of the PAS-
ADD Checklist for detecting psychiatric disorders in adults
with intellectual disability. Journal of Intellectual Disability
Research, 42, 173–183.
Pearlin, L., & Schooler, C. (1978). The structure of coping. Jour-
nal of Health and Social Behavior, 19, 2–21.
Pesudovs, K., Garamendi, E., Keeves, J. P., & Elliott, D. B.
(2003). The activities of Daily Vision Scale for cataract
surgery outcomes: Re-evaluating validity with Rasch anal-
ysis. Investigative Ophthalmology & Visual Science, 44,
2892–2899.
Piquero, A. R., Macintosh, R., & Hickman, M. (2002). The valid-
ity of a self-reported delinquency scale. Sociological Methods
& Research, 30, 492–529.

262 International Journal of Social Psychiatry 60(3)
Ralph, R. O., Kidder, K., & Philips, D. (2000). Can we measure
recovery? A compendium of recovery measures. Cambridge,
MA: Human Services Research Institute.
Riesch, S. K., & Hauck, M. R. (1988). The exercise of self-care
agency: An analysis of construct and discriminant validity.
Research in Nursing & Health, 11, 245–255.
Ritsher, B. J., Otilingam, P. G., & Grajales, M. (2003). Internal-
ized stigma of mental illness: Psychometric properties of a
new measure. Psychiatry Research, 121, 31–49.
Rogers, E. S., Chamberlin, J., Ellison, M. L., & Crean, T. (1997).
A consumer-constructed scale to measure empowerment
among users of mental health services. Psychiatric Services,
48, 1042–1047.
Rosenfield, S. (1992). Factors contributing to the subjective qual-
ity of life of the chronic mentally ill. Journal of Health and
Social Behavior, 33, 299–315.
RUMM. (2005). RUMM L: Interpreting RUMM2020: Part II
polytomous data. Duncraig: RUMM Laboratory.
Ryan, B., Court, H., & Margrain, T. H. (2008). Measuring low
vision service outcomes: Rasch analysis of the Seven-Item
National Eye Institute Visual Function Questionnaire. Optom-
etry & Vision Science, 85, 112–121.
SAMHSA (Substance Abuse and Mental Health Services
Administration). (2005). Consensus on recovery achieved.
In Mental health transformation trends: A periodic brief-
ing. Rockville, MD: US Department of Health and Human
Services.
SAMHSA (Substance Abuse and Mental Health Services Admin-
istration). (2006). National census statement on mental health
recovery. Rockville, MD: National Mental Health Informa-
tion Center, US Government.
Schrank, B., & Slade, M. (2007). Recovery in psychiatry. Psychi-
atric Bulletin, 31, 321–325.
Siegert, R. J., Tennant, A., & Turner-Stokes, L. (2010). Rasch
analysis of the Beck Depression Inventory-II in a neuro-
logical rehabilitation sample. Disability & Rehabilitation,
32, 8–17.
Snyder, C. R., Sympson, S. C., Ybasco, F. C., Borders, T. F.,
Babyak, M. A., & Higgins, R. L. (1996). Development and
validation of the State Hope Scale. Journal of Personality &
Social Psychology, 70, 321–335.
Song, L. Y., & Hsu, S. T. (2011). The development of the stages
of recovery scale for persons with persistent mental illness.
Research on Social Work Practice, 21, 572–581.
Tse, S. (2004). Use of the recovery approach to support Chinese
immigrants recovering from mental illness: A New Zealand
perspective. American Journal of Psychiatric Rehabilitation,
7, 53–68.
Tse, S., Cheung, E., Kan, A., Ng, R., & Yau, S. (2012). Recovery
in Hong Kong: Service user participation in mental health ser-
vices. International Review of Psychiatry, 24, 40–47.
Tse, S., Siu, B. W. M., & Kan, A. (2011). Can recovery-oriented
mental health services be created in Hong Kong? Struggles
and strategies. Administration and Policy in Mental Health,
40, 155–158.
Vidotto, G., Carone, M., Jones, P. W., Salini, S., & Bertolotti,
G. (2007). Maugeri Respiratory Failure questionnaire
reduced form: A method for improving the questionnaire
using the Rasch model. Disability & Rehabilitation, 29,
991–998.
Wagnild, G. M., & Young, H. M. (1993). Development and psy-
chometric evaluation of the Resilience Scale. Journal of Nurs-
ing Measurement, 1, 165–178.
Waugh, R. (2010). Specialized Rasch measures applied at the
forefront of education. New York: Nova Science Publishers.
WHOQOL SRPB Group. (2006). A cross-cultural study of spiri-
tuality, religion, and personal beliefs as components of quality
of life. Social Science & Medicine, 62, 1486–1497.
Wilkinson, G., Hesdon, B., Wild, D., Cookson, R., Farina, C.,
Sharma, V., … Jenkinson, C. (2000). Self-report quality of
life measure for people with schizophrenia: The SQLS. Brit-
ish Journal of Psychiatry, 177, 42–46.
Williams, G. C., Rodin, G. C., Ryan, R. M., Grolnick, W. S., &
Deci, E. L. (1998). Autonomous regulation and long-term
medication adherence in adult outpatients. Health Psychol-
ogy, 17, 269–276.
Wong, J. G. W. S., Cheung, E. P. T., Chen, E. Y. H., Chan, R. C.
K., Law, C. W., Lo, M. S. M., … Lam, C. L. K. (2005). An
instrument to assess mental patients’ capacity to appraise and
report subjective quality of life. Quality of Life Research, 14,
687–694.
Yee, B. (2003). Asian mental health recovery: Follow-up to the
Asian report, vol. 3. Wellington: Mental Health Commission.