Stages of Recovery Instrument
The final consumer-oriented measure of recovery described in this section is the Stages of Recovery Instrument (STORI: Andresen, Caputi and Oades, 2006).
The final consumer-oriented measure of recovery described in this section is the Stages of Recovery Instrument (STORI: Andresen, Caputi and Oades, 2006).
The STORI is a 50-item self-report measure that yields five subscales corresponding to the five stages of recovery, Moratorium, Awareness, Preparation, Rebuilding and Growth. The development of the STORI is grounded in the stage model of recovery (Andresen, Caputi and Oades, 2003). As outlined in Andresen, Caputi and Oades (2006), item generation was significantly influenced by five key studies that identified stages or phases of recovery (Davidson and Strauss, 1992; Baxter and Diehl, 1998; Pettie and Triolo, 1999; Young and Ensing, 1999; Spaniol et al, 2002).
In particular, these studies provided evidence for the four component processes associated with recovery – finding and maintaining hope, re-establishing a positive identity, finding meaning in life and taking responsibility for one’s life (Andresen, Caputi andOades, 2006, p. 974). Conceptually, the STORI has a complex structure reflecting the four component processes that are assumed to vary across five stages of recovery.
Recently, the STORI has received growing attention in the psychological and psychiatric literature. For instance, Deane and Andresen (2006) have used the STORI in a review of a volunteer programme to increase and support social contact and friendships with people with mental illness. Puschner et al. (2010) have included a 30 item version of the STORI in an international study of the outcomes of clinical decision making with individuals with serious mental illness. Given its increased use in research and practice, what can be said about the psychometric properties of the STORI? We provided initial support for the internal consistency of the subscales representing the stages of recovery with alpha coefficients ranging from 0.88 to 0.94 (Andresen, Caputi and Oades, 2006). Weeks, Slade and Hayward (2010) also reported strong alpha coefficients for these subscales with Cronbach’s alpha coefficients between 0.81 and 0.87. They also found strong evidence for the temporal stability of the subscales with re-test reliability coefficients of at least 0.90 for each stage subscale.
There is also evidence for the concurrent validity of the STORI. In our original paper (Andresen, Caputi and Oades, 2006) we correlated the STORI subscales with a set of measures that included the Mental Health Inventory – five item version (Stewart, Hays and Ware, 1988) – the RAS, the Psychological Well-Being Scales (Ryff and Keyes, 1995), the Adult State Hope Scale (Snyder et al., 1996) and the Connor–Davidson Resilience Scale (Connor and Davidson, 2003). A distinct pattern of relationships emerged from this analysis. We found negative correlations between the subscale for Moratorium (stage 1) and the other measures (ranging from 0.44 to 0.68); on the other hand significant positive associations were observed between the subscale for Growth (stage 5) and the other measures (ranging from 0.53 to 0.79) (Andresen, Caputi and Oades, 2006). There were no significant associations observed between stage 3 subscale and the other measures. There were only some negative associations found for the stage 2 subscale, and some significant positive correlations between the stage 4 subscale and the other measures. The pattern of correlations, increasing from negative in stage 1 through to highly positive at stage 5, lend some support to the ordinal nature of the stages of the model.
Moreover, we concluded that these patterns of findings suggest that stages 2, 3 and 4 subscales may be measuring variables not detectable by the other measures used in their study (Andresen, Caputi and Oades, 2006).
Weeks, Slade and Hayward (2010) observed a similar pattern of relationships between the subscales of the STORI and the RAS, finding a strong negative association between the RAS and the Stage 1 subscale and moderate to strong positive associations between the RAS and the subscales of Stages 3, 4 and 5, with the correlations in increasing size from Stage 3 (r¼.458) to Stage 5 (r¼.735). Weeks,Slade and Hayward (2010) argued that these findings suggest the sequential or ordinal nature of stages. We also found evidence for ordinality when examining the inter-correlation of stage subscales (Andresen, Caputi and Oades, 2006). Stage 1 scores were positively correlated with stage 2 and 3 scores (r¼.33 and .22, respectively), but negatively correlated with stage 5 scores (r¼ .05). There was no correlation between stage 1 and stage 4 scores. Stage 2 scores were positively correlated with stage 3 and 4 scores (r¼.88 and .66, respectively), and stage 3 scores were positively correlated with stage 4 and 5 scores (r¼.82 and .29, respectively).
Stage 4 and 5 scores were also correlated (r¼.52). We argued that this pattern of correlations provided support for the validity of the subscales of the STORI – proximal stages were positively correlated, while distal stages had weak association, with the most distal stages being negatively correlated.
In our original investigation, we conducted a cluster analysis of the items of the STORI to determine whether they clustered into the five theoretically determined stages of recovery (Andresen, Caputi and Oades, 2006). Interestingly, the most interpretable solution contained only three clusters. The first cluster comprised stage 1 items; the second cluster consisted of stage 2 and 3 items as well as four stage 4 items; the third cluster consisted of the stage 5 items and six stage 4 items. Similarly, Weeks, Slade and Hayward (2010) also found a three-cluster solution for the items of STORI. These findings point to further research focusing on whether the STORI items discriminate between the stages (Andresen, Caputi and Oades, 2006). However, at a more fundamental level it does raise the question (as noted by Weeks, Slade and Hayward, 2010) of whether recovery is best described by a three or five stage model. Conceptually, we have argued for a five-stage model of recovery (Andresen, Caputi and Oades, 2003). However, empirically, the evidence points to a three-stage model. We addressed this anomaly by considering a general issue with psychological change models (Andresen, Caputi and Oades, 2006) and drew on Smedslund (1997), who argued the stage models of change, such as the Transtheoretical Model, would describe the behavioural outcome for smoking cessation as either smoking or non-smoking. However, there are processes that described the intermediate stages between smoking and non-smoking. Similarly, a three-stage model of recovery may represent a distinction between stages hopelessness, loss of identity and meaning in life; transition; and finally growth.
However, such a conceptualization may not explain the complex nature of recovery from serious mental illness. Perhaps further effort should be directed to testing the ability of the STORI to discriminate between the five stages of the model.
In particular, these studies provided evidence for the four component processes associated with recovery – finding and maintaining hope, re-establishing a positive identity, finding meaning in life and taking responsibility for one’s life (Andresen, Caputi andOades, 2006, p. 974). Conceptually, the STORI has a complex structure reflecting the four component processes that are assumed to vary across five stages of recovery.
Recently, the STORI has received growing attention in the psychological and psychiatric literature. For instance, Deane and Andresen (2006) have used the STORI in a review of a volunteer programme to increase and support social contact and friendships with people with mental illness. Puschner et al. (2010) have included a 30 item version of the STORI in an international study of the outcomes of clinical decision making with individuals with serious mental illness. Given its increased use in research and practice, what can be said about the psychometric properties of the STORI? We provided initial support for the internal consistency of the subscales representing the stages of recovery with alpha coefficients ranging from 0.88 to 0.94 (Andresen, Caputi and Oades, 2006). Weeks, Slade and Hayward (2010) also reported strong alpha coefficients for these subscales with Cronbach’s alpha coefficients between 0.81 and 0.87. They also found strong evidence for the temporal stability of the subscales with re-test reliability coefficients of at least 0.90 for each stage subscale.
There is also evidence for the concurrent validity of the STORI. In our original paper (Andresen, Caputi and Oades, 2006) we correlated the STORI subscales with a set of measures that included the Mental Health Inventory – five item version (Stewart, Hays and Ware, 1988) – the RAS, the Psychological Well-Being Scales (Ryff and Keyes, 1995), the Adult State Hope Scale (Snyder et al., 1996) and the Connor–Davidson Resilience Scale (Connor and Davidson, 2003). A distinct pattern of relationships emerged from this analysis. We found negative correlations between the subscale for Moratorium (stage 1) and the other measures (ranging from 0.44 to 0.68); on the other hand significant positive associations were observed between the subscale for Growth (stage 5) and the other measures (ranging from 0.53 to 0.79) (Andresen, Caputi and Oades, 2006). There were no significant associations observed between stage 3 subscale and the other measures. There were only some negative associations found for the stage 2 subscale, and some significant positive correlations between the stage 4 subscale and the other measures. The pattern of correlations, increasing from negative in stage 1 through to highly positive at stage 5, lend some support to the ordinal nature of the stages of the model.
Moreover, we concluded that these patterns of findings suggest that stages 2, 3 and 4 subscales may be measuring variables not detectable by the other measures used in their study (Andresen, Caputi and Oades, 2006).
Weeks, Slade and Hayward (2010) observed a similar pattern of relationships between the subscales of the STORI and the RAS, finding a strong negative association between the RAS and the Stage 1 subscale and moderate to strong positive associations between the RAS and the subscales of Stages 3, 4 and 5, with the correlations in increasing size from Stage 3 (r¼.458) to Stage 5 (r¼.735). Weeks,Slade and Hayward (2010) argued that these findings suggest the sequential or ordinal nature of stages. We also found evidence for ordinality when examining the inter-correlation of stage subscales (Andresen, Caputi and Oades, 2006). Stage 1 scores were positively correlated with stage 2 and 3 scores (r¼.33 and .22, respectively), but negatively correlated with stage 5 scores (r¼ .05). There was no correlation between stage 1 and stage 4 scores. Stage 2 scores were positively correlated with stage 3 and 4 scores (r¼.88 and .66, respectively), and stage 3 scores were positively correlated with stage 4 and 5 scores (r¼.82 and .29, respectively).
Stage 4 and 5 scores were also correlated (r¼.52). We argued that this pattern of correlations provided support for the validity of the subscales of the STORI – proximal stages were positively correlated, while distal stages had weak association, with the most distal stages being negatively correlated.
In our original investigation, we conducted a cluster analysis of the items of the STORI to determine whether they clustered into the five theoretically determined stages of recovery (Andresen, Caputi and Oades, 2006). Interestingly, the most interpretable solution contained only three clusters. The first cluster comprised stage 1 items; the second cluster consisted of stage 2 and 3 items as well as four stage 4 items; the third cluster consisted of the stage 5 items and six stage 4 items. Similarly, Weeks, Slade and Hayward (2010) also found a three-cluster solution for the items of STORI. These findings point to further research focusing on whether the STORI items discriminate between the stages (Andresen, Caputi and Oades, 2006). However, at a more fundamental level it does raise the question (as noted by Weeks, Slade and Hayward, 2010) of whether recovery is best described by a three or five stage model. Conceptually, we have argued for a five-stage model of recovery (Andresen, Caputi and Oades, 2003). However, empirically, the evidence points to a three-stage model. We addressed this anomaly by considering a general issue with psychological change models (Andresen, Caputi and Oades, 2006) and drew on Smedslund (1997), who argued the stage models of change, such as the Transtheoretical Model, would describe the behavioural outcome for smoking cessation as either smoking or non-smoking. However, there are processes that described the intermediate stages between smoking and non-smoking. Similarly, a three-stage model of recovery may represent a distinction between stages hopelessness, loss of identity and meaning in life; transition; and finally growth.
However, such a conceptualization may not explain the complex nature of recovery from serious mental illness. Perhaps further effort should be directed to testing the ability of the STORI to discriminate between the five stages of the model.
The absence of empirical evidence for a five-stage model may be a measurement issue rather than a conceptual one.
