Introduction

Most cancer survivors adjust well to life after cancer but some experience emotional disorders. Anxiety and depression are the most frequent problems, with a prevalence superior to that found in healthy control groups (^{Yi & Syrjala, 2017}). It is common for cancer survivors, having completed oncological treatment, to suffer from physical complaints, tiredness and a reduction in attention span and ability to concentrate. These or other physical repercussions frequently lead to patients reducing their involvement in relevant roles and/or pleasant activities (^{Cataldo & Brodsky, 2013}; ^{Fernández, Padierna et al., 2011}). Moreover, when faced with the oncological process and fear of cancer recurrence, survivors frequently experience unpleasant thoughts, feelings and memories (^{McGinty, Small, Laronga, & Jacobsen, 2016}). A common reaction to these experiences, encouraged and reinforced by the cultural context, is to try to avoid them. These attempts not only do not achieve the desired relief but generate more discomfort, contribute to perpetuating it, and limit involvement in important areas of life (^{Bardeen, 2015}). Distancing oneself from day-to-day activities reduces the chances of maintaining contact with the rewarding situations and valuable conditions of life. This situation may be at the root of their emotional problems (^{González-Fernández et al., 2017}).

The behavioral theories of depression posit that decrease in access to environmental rewards and/or the reinforcement of depressive behaviors and the punishment of healthy ones are causal factors predicting the beginning and the maintenance of clinical depression (^{Lewinsohn, 1974}). The relationship between response-contingent positive reinforcement (RCPR) and emotional distress has been established (^{Cuijpers, van Straten, & Warmerdam, 2007}; ^{Manos, Kanfer, & Bush, 2010}); in particular, a low level of RCPR is one of the critical predictors of clinical depression. ^{Lewinsohn, Sullivan and Grosscup (1980}) explain the decrease in RCPR as a consequence of the combination of the following conditions: a decrease in the number of reinforcement events; a decrease in the availability of these reinforcers in the environment; the absence of appropriate instrumental behaviors to experience gratifying contingencies; and an increase in exposure to aversive environmental experiences. For many cancer survivors, these four conditions are assimilated in their own experience of the oncological process. In line with this model, behavioral treatments for depression have been developed to facilitate increased access to reward while decreasing the intensity and frequency of punishing events ((^{Lewinsohn, 1974}). Currently, Behavioral Activation (BA) is a behavioral intervention for depression that shows promise in effectively treating depression by increasing goal and value-based activity levels, thereby eliciting increased response-contingent reinforcement (^{Lejuez, Hopko, Acierno, Daughters, & Pagoto, 2011}; ^{Martell, Adis, & Jacobson, 2001}). The use of BA in treating emotional problems in cancer patients has also been supported by controlled studies (^{Fernández, Villoria, Fernández, González, & Pérez, 2017}; ^{Hopko et al., 2013}; ²⁰⁰⁸; ²⁰⁰³).

In order to assess the relationship between RCPR and emotional state, it is vital to carry out an objective and valid evaluation of the former. During the therapy, observation and behavior reports (e.g. daily diaries, activity schedules, home observations) represent invaluable strategies with regard to programming activities aimed at recovering reinforcement (^{Hopko & Mullane, 2008}). However, it is not easy to measure RCPR directly as this involves observation of relevant behaviors in the person’s day-to-day environment over extended periods of time (^{Manos et al., 2010}). Those researchers who have developed measures of RCPR have focused on measuring environmental rewards and exposure to pleasurable events. With the same objective in mind, ^{Armento and Hopko (2007}) developed the Environmental Reward Observation Scale (EROS). This scale is intended to be a proxy measure of RCPR by assessing the subjective experience of rein forcement. Items were designed to evaluate RCPR according to the formulation of ^{Lewinsohn (1974}). In its original version, with a sample of American students, the scale showed a unidimensional structure. Compared to commonly used depression and anxiety measures, as ^{Beck Depression Inventory (BDI-II; Beck, Steer, & Brown, 1996}), The Center for Epidemiologic Studies Depression Scale (CESD; ^{Radloff, 1977}), or The State-Trait Anxiety Inventory (STAI; ^{Spielberger et al., 1983}), correlations with the EROS ranged between medium and large. In other validation studies which included the general population and also populations with emotional disorders, carried out with participants who were Spanish (^{Barraca & Pérez-Álvarez, 2010)}, Colombian (^{Valderrama-Díaz, Bianchi-Salguero, & Villalba-Garzón, 2016}), or French-speaking Belgians (^{Wagener & Blairy, 2015}), in all cases there were found to be appropriate levels of internal consistency and also consistent validity evidence of internal structure. These findings indicate that the EROS could be a valid tool of environmental reinforcement, is congruent with the behavioral theories of depression and can facilitate the implementation and measuring of BA interventions. Indeed, the EROS has been used both to measure the notion of environmental reinforcement (^{Becoña et al., 2017}; ^{MacPherson et al., 2010}), and to evaluate the effect of behavioral treatment of emotional disorders in subjects with health problems (^{Bombardier et al., 2017}). In oncological populations, the EROS has been used when carrying out behavioral interventions and as a measure of the efficacy of Behavioral Activation in the treatment of emotional disorders (^{González-Fernández, Fernández-Rodríguez, Paz-Caballero & Pérez-Álvarez, 2018}; ^{Lejuez et al., 2011}, 2001). Despite this, and despite its evident usefulness, the EROS has not, however, been validated with this population.

Given this fact, the aim of this study is to obtain data regarding the reliability and validity evidence of the Spanish version of the EROS in a representative and relatively numerous sample of breast cancer survivors. At the same time, given the aforementioned peculiarities of populations of cancer survivors, the study also aims to analyze the sensitivity, specificity and ability of this scale to discriminate between survivors with and without depressive disorders.

Method

Results

Data quality analysis

Comparing all items means and standard deviations are like almost all the items except items 5 and 7. The item response was high with a small number of missing answers (1.4-1.8%). Floor effect was small in almost all the items (range 2.3-14.4% in eight out of ten cases. Floor effects were found in the cited items 5 and 7) and all items had a ceiling effect larger than 15% (range 18.6-56.3%), and all the response choices were used in all items. (table 1)

Table 1.  Data quality analysis. 

Psychometric properties of the EROS

First of all, validity evidence based on internal structure was checked via CFA, showing the following indexes: χ ²/df = 3.90; CFI = .95; RMSEA = .116 (.096 - .137; CL = 90%); UniCo = .896; ECV = .819; MIREAL = .270. Additionally, factor loadings of each item, and measurement errors (represented by a double-headed arrow) are shown in Figure 1. This unidimensional solution shows a reliability coefficient of α = .91.

Figure 1.  Standardized solution of unidimensional model of EROS. 

IRT Analysis

Parameters a and b were estimated for the EROS instrument (Table 2). As can be seen, all items have a moderate to very high discrimination power (a-parameter). More specifically, half of the items exhibit a very high discrimination (values larger than 1.7). In relation to b-parameters, when using Samejima’s GRM, the number of b-parameters derives from 1 minus the number of alternatives for that item. As there are four alternatives for the items in the EROS, there are three b-parameter values for each item. The largest differences (b ₃ − b ₁ ) in the EROS were found in items 10 and 2, and the smallest differences correspond to items 5, 9 and 6.

Table 2.  Item Response Theory (IRT) parameter estimates for Environmental Reward Observation Scale (EROS). 

Regarding the measurement precision of the EROS, the IF exhibits maximum information between -1.5 and 0.5 trait levels, with the accuracy diminishing particularly in individuals with a latent trait level above +1 (Figure 2).

Figure 2.  Environmental Reward Observation Scale Information Function. 

Validity evidence in relation to other variables

Table 3 shows the Pearson correlations between the EROS and the rest of the instruments described previously, considering their subscales in those cases in which the instruments are multidimensional. Medium to large correlations were found between the EROS and almost all the dimensions assessed in this sample.

Table 3.  Pearson correlation coefficients between the Environmental Reward Observation Scale (EROS)and scales of anxiety, depression, general distress and psychological flexibility. 

Discussion

The emotional problems of cancer survivors have been related to a progressive loss of rewards resulting from adverse experiences in the oncological process. With this in mind, the aims of this study were, firstly, to determine whether the Environmental Reward Observation Scale is a valid and reliable instrument for evaluating the notion of environmental reinforcement in breast cancer survivors and, secondly, to examine the capacity of this scale to discriminate between survivors with and without depressive disorders.

With a clinically and socio-demographically representative sample of breast cancer survivors containing an adequate representation of participants with and without emotional disorders, the results obtained confirm the psychometric guarantees of the EROS in this population. Regarding evidence of construct validity, the confirmatory factor analysis indicates the unidimensionallity of the scale. This result coincides with those reported by ^{Armento and Hopko (2007}) and by other validation studies (^{Barraca & Pérez-Álvarez, 2010}; ^{Valderrama-Díaz et al., 2016}; ^{Wagener & Blairy, 2015}). It should be noted that, in our study, most of the fit indexes were observed to be below the standard criterion, and only the MIREAL and CFI indexes can be considered appropriate. The literature in general recommends a flexible approach to evaluating overall model fit, using a global analysis rather than a strict reliance on arbitrary cut-off values that may not be relevant to the specific research context (^{West, Taylor, & Wu, 2012}). Consequently, in this case, considering fit indexes as a whole, the adequate factor loading shown by all items, the adequate reliability, which demonstrates a high relationship between all the items, and the fit indexes found in other papers, it seems plausible to define EROS as an essentially unidimensional instrument. It is, therefore, a test with sufficient psychometric guarantees, in the field of Psycho-oncology, for use in research into the relationships between the loss of response-contingent positive reinforcement and depression (^{Hopko et al., 2003}; ^{Lewinsohn, 1974}; ^{Martell, Addis, & Jacobson, 2001}).

With regard to the reliability of the test, it is to be underlined that this was greater than that found in the aforementioned previous validation studies. Not only was there found to be a high relationship between the individual items, but also, when analyzed within the Item Response Theory, all have a moderate to very high power of discrimination regarding the perception of available environmental reinforcement of breast cancer survivors. Those items which would allow a particularly precise differentiation between survivors with a greater or lesser subjective perception of environmental reinforcement are: “In general I am very satisfied with the way I spend my time _{(item 3)}”, “My life is boring _{(item 9)}”, “It is easy for me to find enjoyment in my life _{(item 4)}”, “A lot of activities in my life are pleasurable _{(item 1)}”. As might be expected, these items are the ones which have the highest factor loadings and item-test correlations (>.80). It should be pointed out that, in line with ^{Lewinsohn’s formulation of RCPR (Lewinsohn, 1974}), these items would be evaluating both the quantity and potential availability of environmental reward (items 9 and 1) and the competence of the person to obtain reinforcement from her environment (items 3 and 4). These results offer further evidence of the consistency of the concept of RCPR. Furthermore, in light of the relationship existing between emotional state and RCPR (^{Cuijpers, van Straten, & Warmerdam, 2007}; ^{Manos et al., 2010}), they would support the effectivity ofbehavioral interventions aimed at increasing reinforcement in order to reduce levels of emotional distress. With regard to future studies, the question arises as to whether evaluating only those items which have shown the greatest discriminatory capacity may be sufficient in order to make a precise estimation of the survivor’s subjective perception of environmental reinforcement. If this were so, it would be possible to elaborate a briefer instrument and reduce evaluation time. The test would in turn be made simpler and more efficacious.

In the analysis of those items with the smallest b ₃ -b ₁ difference values, it is also worth noting that the conditions which appear to be most susceptible to change are: “Other people seem to have more fulfilling lives _{(item 5)}”, “My life is boring _{(item 9)}”, “Activities that used to be pleasurable are no longer gratifying _{(item 6)}”. In contrast, those in which change appears less probable are: “Lately I have found that many experiences make me unhappy _{(item 2)}”, “The activities I engage in usually have positive consequences _{(item 10)}”, “I am satisfied with my accomplishments _{(item 8)}”. These results have clear clinical implications. It can be concluded that, when dealing with emotional disorders in breast cancer survivors, it would be more effective and efficacious to focus the first sessions of the intervention on that condition which can be more easily modified (according to items 5, 9 and 6), that is, on promoting an increase in rewarding activities which ensure the availability of environmental rewards. However, ultimately, the efficacy of the intervention depends on being able to modify also those conditions which are less probable to change (according to our results, items 2, 10, 8) as this would imply that subjects had acquired competences to ensure the contingent reinforcement of their actions. These findings are totally coherent with the principles and procedures of behavioral therapies of depression, and in particular of Behavioral Activation (^{Lejuez et al., 2001}; ^{Martell et al., 2001}). From the very first sessions, this therapy establishes a program of rewarding activities which are consistent with the person’s values with a view to facilitating an increase in RCPR. As in any contextual therapy, the ultimate aim is to provide the person with competences which enable him/her to identify and change those response patterns which lead to depression and/or anxiety. For that reason, special attention is paid to identifying and modifying strategies of behavioral avoidance. Avoidance patterns distance the person from those day-to-day situations on which his/her rewards depend and reduce the possibilities of undertaking other healthier strategies of interaction with the environment (^{Bardeen, 2015}; ^{Trindade, Ferreira & Pinto-Gouveia, 2017}).

A similar interpretation can be made of the strong and negative correlation of this measure with measures of depression (HADS-D, BSI-D; BDI-IA-SCA), of behavioral avoidance (AAQ-II, BADS-A/R, BADS-WSI, BADS-SI) and of anxiety (HADS-A, BSI-A). There is an equally strong, but positive, correlation with measures of activation (BADS-T). These results coincide with those reported by other authors, who, with different clinical populations and general population, have identified a relationship between depression and RCPS (^{Folke & Kanter, 2016}; ^{Hill, Buitron, & Pettit, 2017}), and between patterns of behavioral avoidance and emotional state (^{Brem, Shorey, Anderson, & Stuart, 2017}; González-^{Fernández et al., 2017}). Both conditions, as pointed out previously, are principal objectives of behavioral therapies for depression. Consequently, we suggest that the EROS could also be a suitable instrument for evaluating the results of behavioral treatment of emotional disorders in breast cancer survivors. This belief is also supported by the results, which confirm that participants with and without emotional problems differ with regard to their levels of perceived environmental reward. Furthermore, as predicted by the model on which the scale is based, the particularly large effect size of the differences between the scores of survivors with and without depression are especially noteworthy.

In our study, the Information Function of the EROS indicates that it would be for those breast cancer survivors with a score around the mean of the population (approximately between 17 and 27) for whom the scale would best estimate perception of available environmental reinforcement. It is precisely within this range that the cut-off point for depression (25) is extremely sensitive and most specific. The use of the Hospital Anxiety and Depression Scale as a gold standard measure is endorsed by its properties and by its widespread application in evaluating emotional distress free from the bias of physical symptomatology. ^{Walker et al. (2007}) found somatic symptoms of depression to be confounding factors in patients with medical illnesses. The same could occur in cancer survivors. It is, nevertheless, true that some authors have questioned the suitability of this instrument for differentiating between depression and anxiety (^{Burns, Hofer, Curry, Sexton, & Doyle, 2014}). In light of the lack of other studies into this matter, one possible subject of future study could be the use of other measures and/or procedures of clinical diagnosis as gold standard comparison criteria for confirming the adjustment of cut-off points in the EROS.

The degree of similarity shown between the survivors’ experiences during the oncological process would suggest that the results of this study could be extended to other cancer patient populations. However, the fact that this population is mainly female raises the question of whether the same observations could be applied to a population of male survivors. Although it appears unlikely that gender would affect the unidimensional structure of the scale, as has been shown in previous validation studies using samples with an adequate representation of both men and women, it could affect the scores. It is known that women suffer from and/or seek help for more emotional distress than men (^{Kuehner, 2017}). In oncological survivors, gender has also been related to different levels of depression (^{Bevilacqua et al., 2018}), emotional distress and fear of a relapse (^{Koch-Gallenkamp et al., 2016}). Furthermore, some studies suggest that women appear to be more sensitive to reinforcement and rewards (^{Ryba & Hopko, 2012}; ^{Tull, Gratz, Latzman, Kimbrel, & Lejuez, 2010}). To sum up, due to the great clinical value that use of this scale appears to have with oncological patients, and also in order to gain more profound insights into the relationships between RCPR and emotional distress, it would be of interest to carry out further studies using adequately-balanced samples in terms of gender and at different stages of different types of cancer. In order to overcome some of the limitations of this study, it would be recommendable to use a longitudinal evaluation design to examine the guarantees offered by the instrument.

Conclusion

The EROS is shown to be a valid and reliable measure for the evaluation of environmental rewards in breast cancer survivors. Its use is recommended for estimating the degree of loss of contact with reinforcers of day-to-day life which patients undergo as a result of the oncological treatment and disease and which increase the probability of emotional disorders.