第一年的研究目的乃在發展「復元階段量表」，該研究成果已發表於國外期 刊：Song, L .& Hsu, S. (2011).The Development of the Stages of Recovery Scale for Persons with Persistent Mental Illness. Research on Social Work Practice, 21(5), 572-581. First published on March 14, DOI: 10.1177/1049731511402218
在此直接以刊登的版本呈現該研究的設計、具體的方法和研究結果。
The Development of the Stages of Recovery Scale (SRS) for Persons with Persistent Mental Illness
Li-yu Song & Su-Ting Hsu Abstract
This study aimed to develop a scale which could be used as a valid way to show the evidence of recovery-oriented services. A 51-item scale was developed to assess both the component processes and outcomes of recovery. A sample of 471 participants administered the questionnaire. The factor analysis yielded a 45-item scale with six subscales, including three components of process and three outcomes. The construct validity was confirmed. Each subscale has very good internal consistency (α
=0.80-0.95), and the 3 to 5 weeks test-retest reliability was 0.72. The scale could significantly differentiate the rehabilitation sample and the better functioning sample.
The results indicated four stages of recovery. The external construct validity was also ensured. The results supported the psychometric property of the Stages of Recovery Scale (SRS). It could be utilized for both assessment and evaluation to document the evidence of a recovery-oriented program, collectively or individually.
Keywords: Recovery, Stage of Recovery Scale, Persistent Mental illness
Introduction
Recovery has become the ultimate goal of treatment for persons with persistent mental illness in the West (Anthony, Cohen, Farkas, & Gagne, 2002; Liberman &
Kopelowicz, 2002). As recovery becomes the primary treatment goal, one issue that emerges is how we assess the progress. Such a measure is desirable as evidence-based practice is now emphasized. For example, in the document prepared by The Evaluation Center@HSRI, four measures were included to assess the recovery-promoting environments, including personal, administrative, and treatment levels (Campbell-Orde, Garrett, & Leff, 2005). In addition, some researchers have focused on the differentiation of stages of recovery and the measurement of those stages (Andersen, Oades, & Caputi, 2003; Davidson & Strauss, 1992; Spaniol, Wewiorski, Gagne, & Anthony, 2002).
The development of such a measurement needs the guidance of a clear definition of recovery. From the consumer’s perspective, personal recovery is a unique and individual journey of self-discovery; it does not mean “to be cured," but it is about recovering to a new sense of self and of purpose within and beyond the limits of disability (Deegan, 1988). Anthony, Cohen, Farkas & Gagne (2002) defined recovery as “a unique personal process of changing one’s attitudes, values, feelings, goals, skills, and/or roles. It is a way of living a satisfying, hopeful, and contributing life, with or without limitations caused by the illness” (p.31). These descriptions imply that recovery is a holistic concept, which covers both processes and outcomes. However, the existing measures of recovery have not fully assessed both, and none has the cutoff scores to differentiate the stages of recovery. Therefore, the investigators aimed to develop a scale to fully measure both the components of process and outcomes of recovery, and to further use the scale score to indicate a stage of recovery. The goal was to create an instrument which can be used to show the evidence of recovery-oriented services.
Conceptual definition of recovery
Based on the literature, recovery is a holistic concept, including component process and outcomes. The important components in the recovery process include the emergence of hope, acceptance of the disability and becoming able to cope with the symptoms, taking responsibility of own wellness, effective management of own disability, developing a self-identity with potential, and developing a new meaning and sense of purpose for life (Andersen, et al., 2003; Anthony et al., 2002; Jenkins &
Carpenter-Song, 2006; Kelly & Gamble, 2005; Markowitz, 2001; Onken, Craig, Ridgway, Ralph, & Cook, 2007; Spaniol, et al., 2002; Turner-Crowson & Wallcraft, 2002). Song & Shih (2009) summed up the components as emerging sense of self, management of disability, and hope, willingness, and action.
The recovery outcomes can be evaluated from both subjective feelings and attitudes and objective performance of functioning. Subjective feelings and attitudes include an enhanced sense of self (Davidson & Strauss, 1992; Deegan, 1988; Fisher, 1991), feeling balanced and having a sense of wholeness (Fisher & Ahern, 1999), improved quality of life (Liberman & Kopelowicz, 2002), personal growth, self-acceptance, and autonomy (Andersen et al., 2003). Among these, sense of self is a key element. According to Davidson and Strauss (1992), it connotes self-efficacy, internal control, and self-esteem. Objective performance of functioning is demonstrated by involvement with the outside world and includes: enhanced social functioning, regaining social role, and participating in the community (Deegan, 1988;
Fisher, 1991; Fisher & Ahern, 1999; Liberman, Kopelowicz, Ventura, and Gutkind, 2002; Song & Shih, 2009).
Stage of recovery and measurement
Regarding the stages of recovery, Davidson & Strauss (1992) categorized the stage based on the sense of self, including discovering a more active self, taking stock of self, putting the self into action, and appealing to self. Spaniol et al. (2002) differentiated the stages from how well the consumer handled the disability. They derived four stages: overwhelmed by the disability, struggling with the disability, living with the disability, and living beyond the disability. Andersen et al. (2003) focused on psychological recovery and categorized the progress of recovery into five stages based on four key component processes: finding and maintaining hope, the reestablishment of a positive identity, finding meaning in life, and taking responsibility for one’s life. The five stages they derived were: moratorium, awareness, preparation, rebuilding, and growth. Although there are different ways of categorizing the stages of recovery, there are similarities among them according to Andersen et al. (2003). Moreover, it is noteworthy that the stage a consumer reaches might be tentative since recovery is a journey of spiral progress (Deegan, 1988; Song
& Shih, 2009).
There are existing instruments which measure recovery. Among them, the stage of recovery instrument (STORI) (Andersen et al., 2003) is the only one which attempted to differentiate the stages of recovery. The scale is comprised of 50 items and has established an important basis for assessment to capture the rich information on the recovery process. STORI also has good concurrent validity in that the scale scores had a significant and high correlation with other measures, such as Recovery Assessment Scale (RAS) and Psychological Well-Being Scale. However, in the
There are other existing measures of individual recovery. Campbell-Orde, Garrett,
& Leff (2005) collected nine measures in the document Measuring the Promise: A Compendium of Recovery Measures, Volume II. Among these measures, seven of them had been empirically tested. The Consumer Recovery Outcomes System (CROS 3.0) was a 38-item scale developed by the Colorado Health Networks Partnership. It contains such domains as hope, coping with symptoms, daily functioning, quality of life, and treatment satisfactions. The initial test was based on a sample of 576 consumers and staff. The scale domains had good internal consistency (range=0.79-0.90), and the test-retest reliability at eight days interval ranged from 0.69 to 0.76. The scale’s external construct validity was demonstrated by a high correlation between the scale scores of the Behavior and Symptom Identification Scale (BASIS-32) and Brief Psychiatric Rating Scale (BPRS), and the Wisconsin Quality of Life Index (WQLI).
The Illness Management and Recovery (IMR) scale was initially tested on 50 adults with severe mental illness and 20 clinicians. It contains 15 items with good test-retest reliability (0.82) but not very satisfactory internal consistency (0.70 < 0.80).
The client version scale score significantly correlated with the Colorado Symptom Inventory (r=0.38), the Recovery Assessment Scale ((r=0.54), and the Multnomath Community Ability Scale (r=0.48). Thus, the concurrent validity of IMR was partially confirmed based on the rule that the correlations should be larger than 0.50 (Monette, Sullivan, & DeJong, 2008).
The RAS (Corrigan, Salzer, Ralph, Sangster, & Keck, 2004) consisted of 41 items and was tested with 35 consumers with psychiatric disability. It is a five-factor scale, including personal confidence and hope, willingness to ask for help, not dominated by symptoms, goal and success orientation, and ability to rely on others.
The RAS has the same drawbacks as the STORI.
The Mental Health Recovery Measure (MHRM) (Bullock, 2005) is a 30-item scale and was tested on a sample of 279 cases. The scale is comprised of eight subscales: overcoming stuckness, self-empowerment, learning and self-redefinition, basic functioning, overall well-being, new potentials, spirituality, and advocacy/enrichment. The scale covers some recovery outcomes and has good concurrent validity.
The Ohio Mental Health Consumer Outcomes System (Ohio Outcomes System) focused on measuring recovery outcome instead of process. The adult consumer form A contains 67 items with domains of quality of life, health and symptom distress, overall empowerment, etc. The scale was tested on a large size of sample (nearly 1,500) with good internal consistency on the subscales (0.77-0.93). The scale had discriminate validity with other constructs, such as the Beck Depression Inventory
and the Minnesota Multiphasic Personality Inventory.
The Peer Outcomes Protocol (POP) measures outcomes of recovery and other related concepts such as employment satisfaction, community satisfaction, program quality of life, program satisfaction, etc. It is a 241-item scale and was initially tested on 100 consumers. It had good internal consistency on the subscales, but less satisfactory test-retest reliability (0.47-0.85, interval: two-weeks or less).
Finally, the Reciprocal Support Scale measures only one element of the recovery outcome. It is a 14-item scale and was tested on 80 adults with mental illness. It had high internal consistency (0.95) and significant correlation (r=0.28) with the self-esteem scale derived from the Ohio Outcomes System.
The existing scales have established some important items of recovery. However, there were some limitations among the scales. First, some scales focused on the component of process (e.g. STORI), some on outcomes (e.g. Ohio Outcome System), and some covered only part of both (e.g. RAS, MHRM); however, it seems that none fully measured both the process and outcomes of recovery. Second, except for the CROS 3.0 and Ohio Outcomes System, the sample size used did not reach the standard of 300 as required for scale development (DeVellis, 1991). Third, the external construct validity had been tested in most of the scale. However, neither discriminant validity among different stages or levels of recovery has been examined, nor cutoff scores developed for stages. Since recovery involves both process and outcomes, the investigators maintain that both should be included in a measure to fully reveal a consumer’s status. Since recovery-oriented programs have gained popularity in this field, the instrument with acceptable psychometric property is necessary to demonstrate the effectiveness of various programs. The investigators aimed to expand the examination on different types of validity and to develop the cut-off scores for stages of recovery.
Method
This study followed the eight steps of scale development proposed by DeVellis (1991), including clarifying the definition of the concept, generating an item pool, determining the format of measure, having initial item pool examined by experts, inclusion of validation items, administering items to a development sample, evaluating the items, and optimizing scale length. The goal was to develop a self-administered scale. This study has approved by the Institutional Review Board at the Chang-Gung Hospital in Taiwan.
willingness to cope, taking responsibility, management of disability, sense of self, new meaning/purpose, self-determination, awareness and potentiality, and competence (confidence, pursuing life goals). The latter included four elements: better social functionning and social roles, overall wellbeing, life satisfaction, and helping others.
During the process, the existing three scales (STORI, RAS, and MHRM) served as important references for the investigators to generate the items. Initially, 70 items were derived and submitted for review by five experts who have some knowledge in recovery. Two of the experts have studied recovery at the Center for Psychiatric Rehabilitation in Boston; one has read extensively on the subject and has promoted recovery-oriented service; and two have been practicing recovery-oriented service.
They were asked to rate the adequacy (yes or no) and importance (1 to 5) of each item and to suggest the changes in wording and the new items. The items were retained if at least three experts rated it as adequate and with an average score of importance at least four. As a result, sixteen items were deleted during the process. Four items were found redundant and were combined into two, and four additional items were suggested by the experts.
A 56-item scale was thus formed and pilot tested on 35 consumers in two rehabilitation centers, one in Northern and one in Southern Taiwan, to gather their opinions and input on the items and to ensure that wordings were comprehensible for them. The frequency distribution of each item was examined to ensure enough variability captured. The discriminatory power (DP) score of each item was calculated (Monette, Sullivan, & DeJong, 2008). Four items with a DP score less than 0.50 were deleted. The process yielded a 51-item scale for the final test which included the 31 component process items and 20 outcome items (see Table 1). Responses were measured with a four-point Likert-type scale, with response categories as: never (0), seldom (1), sometimes (2), and often (3).
Participants
Participants for this study were persons with severe mental illness (consumers) living in the community. Criteria for sample selection were: 1) Consumers must have been at least 18 years old and hospitalized at least once since the onset of illness; 2) Consumers must have a diagnosis other than neurosis, substance abuse, personality disorder, or dementia due to any cause. To increase the variance in the status of recovery and for the examination of discriminant validity, two types of samples were recruited. One sample consisted of those who were participating in the activities in a psychiatric rehabilitation center (Rehab sample). The rehabilitation center in Taiwan serves the purpose of enhancing participants’ social functioning. The members in the rehabilitation center are comprised of consumers in various states in terms of symptom control and functioning. The second sample consisted of those who once
were members of a rehabilitation center but are now living in the community with a life goal, engaging in continuous social participation, and have not been hospitalized within a year (Better Function sample).
A list of all the rehabilitation centers was established and the centers were contacted for their participation in the study. There were 34 centers in Taiwan, and 24 of them agreed to collaborate with the investigators. They asked their members’
willingness to fill in the questionnaires and contacted those who fit the criteria for the Better Function sample. They also helped arrange the time for data collection in the center. The questionnaire was filled out by participants, with a research assistant present to clarify questions. As a result, 370 consumers completed data for the Rehab sample, and 101 for the Better Function sample. Among them, 55 agreed to take the test again within three to five weeks. Each subject was given a voucher (worth US$6.30) to a convenience store as a payment.
Variables and Instruments
A self-administered questionnaire was designed to collect the data. In addition to the recovery scale instrument, the questionnaire also included demographic variables and illness variables to depict the sample characteristics. Since social functioning, empowerment, and life satisfaction were individual elements of recovery outcomes based on the Unity model of Recovery proposed by Song & Shih (2009), they were used as the criteria variables for testing the external construct validity. If the scale truly measure recovery, the scores should highly correlated with these three measures.
Demographic variables included sex, actual age, education (less than high school, high school, and at least some college), marriage (yes/no), and living arrangement (living with family members, relatives, or friends; living alone; living in halfway house).
Illness variables included diagnosis (schizophrenia, affective disorder, other), age of onset, number of hospitalizations since onset, number and length of hospitalizations within the past two years, and taking psychiatric medication (yes/no).
The information was based on participants’ self report.
Empowerment was measured by the scale developed by Song (2006). It is based on the original scale of Rogers, Chamberlin, Ellison, & Crean (1997) and further expanded and tested by using the samples in Taiwan. There are eight dimensions among the 34 items, including self-efficacy and internal control, external control, interpersonal communication skills, interpersonal assertiveness, social assertiveness, social political resources and influence, social political power, and social political
discriminate the power scores among social work educators, social workers, and clients. In addition, the external construct validity was ensured by the significant association with participation in group activities, role opportunity and support, and life satisfaction. The internal consistency of the entire scale was 0.95, and the three weeks test-retest reliability was 0.80 (Song, 2006). The Cronbach’s alpha for the data in this study was 0.92.
Social functioning was measured using a 36-item scale designed for testing on persons with mental illness (Song, 2001). The scale had seven subscales:
social/withdrawal, interpersonal communication, independence-ability, independence-performance, recreation, pro-social, and occupation/employment. Each subscale had an acceptable reliability (α=0.52–0.86), given the number of items, and each had good test-retest reliability (γ=0.75–0.94). The interrater reliability between consumers and their caregivers, internal construct validity, and discriminant validity (consumers vs. their siblings) have been confirmed. The scale has been widely utilized as an evaluation tool in both research and practice in Taiwan. The Cronbach’s alpha for the data in the current study was 0.88.
Life satisfaction was measured by a scale of seven items developed by Song (2006). The scale captured the level of satisfaction on various life aspects of living status, work, finance, interpersonal relations, children’s status, self-competence, and external environment. Items were phrased in ways such as: “I am satisfied with my living status.” The participants rated each item among four categories: strongly disagree (1), disagree (2), agree (3), and strongly agree (4). The construct validity of this scale was ensured by its significant correlation with level of empowerment (γ
=0.66), and the Cronbach’s alpha was 0.80. The Cronbach’s alpha for the current study was 0.86.
Data Analysis
The investigator used SPSS software to conduct the analyses. In addition to the descriptive analysis, the following analyses were performed:
Factor analysis. Common factor analysis with varimax rotation was conducted to examine the factor structure of the scale. The criteria for the decision of the number of factors was an eigenvalue greater than 1, and for the retention of items was factor loading greater than 0.4.
Reliability test. Cronbach’s alpha was calculated to test the internal consistency of items for the entire scale and factors.
Cluster Formation. Two steps of cluster analyses were performed. First, initial cluster analysis, using Ward’s method (hierarchical cluster analysis), was conducted on the entire sample to derive the initial cluster solutions. The classification was based
on the Euclidean distances between consumers on the score of recovery dimensions.
Cluster centers (the means for each factor) were examined for each cluster solution alternative. Also, discriminant function analyses and one-way ANOVA were performed to help decide the best solution according to the following considerations:
1) whether significant differences existed among clusters on each factor of the recovery scale, 2) the number of consumers in each cluster, and 3) the rate of accurate classification (hit ratio). The cluster solution that maximized differences among clusters was decided to be the best solution. Second, the cluster solution derived from the previous stage was internally validated using K-Means cluster analysis. In this procedure, a priori cluster solutions and cluster centers were specified with constrained solutions (not allowing updating the cluster centers). The coefficient of agreement, Kappa, was computed for the previous classification and K-Means
1) whether significant differences existed among clusters on each factor of the recovery scale, 2) the number of consumers in each cluster, and 3) the rate of accurate classification (hit ratio). The cluster solution that maximized differences among clusters was decided to be the best solution. Second, the cluster solution derived from the previous stage was internally validated using K-Means cluster analysis. In this procedure, a priori cluster solutions and cluster centers were specified with constrained solutions (not allowing updating the cluster centers). The coefficient of agreement, Kappa, was computed for the previous classification and K-Means