Construct validity and factor structure of sense of coherence (SoC-13) scale as a measure of resilience in Eritrean refugees living in Ethiopia

Mental health research is increasingly focusing on the importance of investigating protective factors such as coping style and resilience in people who are vulnerable to developing mental health problems in the context of a humanitarian crisis [1‐3], as well as those with chronic physical health conditions [4‐6]. Knowledge about such personal resources is vital for the purpose of making informed decisions while planning an intervention. Following the salutogenic model of human resilience, which is considered to be a paradigm shift from the adversities (pathogenic) model to the strengths model, understanding resilience is becoming the focus of research in humanitarian settings [7]. There is also a shift in attention to positive human functioning and the means of achieving an optimal level of wellbeing, although a number of research studies deal with the negative effect of trauma [8].

Resilience refers to the process of negotiating, managing and adapting to significant sources of stress or trauma [9]. Resilience is considered to be common and the normal response of people to conditions of adversity [7]. Unlike the pathogenic paradigm, which focuses on the aetiology of disease, the focus of the salutogenic paradigm is on sources of health, and hence it deals with the mechanisms which underlie management of stress to achieve health [6]. Among others, sense of coherence is a personality-focused collective attribute employed for predicting health, which is protective against the negative consequences of adverse events [10].The definition of sense of coherence provided by Antonvsky (1987, p.19) is: “a global orientation that expresses the extent to which one has pervasive, enduring though dynamic feelings of confidence that (1) the stimuli deriving from one’s internal and external environments in the course of living are structured and predictable and explicable; (2) the resources are available to meet the demands posed by the stimuli; and (3) these demands are challenges worthy of investment and engagement” [11]. There are three underlying constructs within sense of coherence. These include: comprehensibility, which refers to an enduring way of conceptualizing circumstances in an orderly, consistent, structured and clear way; manageability which refers to understanding the availability of adequate resources to cope with demands, whereas meaningfulness refers to the values that individuals place one vents irrespective of their effect, and which, therefore, deserve effort and commitment [12]. Although a number of studies employ the long version of Antonovsky’s SoC scale having twenty nine items (SoC-29), some studies have used the adapted version of the short form of the Sense of Coherence Scale with 13 items (SoC-13) to study resilience among adult forced migrants in humanitarian settings [1, 2, 13]. The validity and robustness of the SoC-13 scale is confirmed by its increased application in different places of the world, including: North America, Europe, Australia, South Africa, and the Middle East [2]. Although there is a debate in the literature regarding the extent to which sense of coherence and resilience are distinct concepts [14], the consensus is that sense of coherence is a comprehensive and over-arching concept, which includes resilience and hardiness [9].

Sense of coherence has been found to have a significant inverse relationship with mental health problems such as Post Traumatic Stress Disorder (PTSD) and depression [15], as well as adverse health conditions such as: rheumatoid arthritis [4], coronary heart disease [5] and congenital heart disease [6].

In a previous study amongst Eritrean internally displaced persons (IDPs), the sense of coherence scale (SoC-13) was found to have adequate psychometric properties to measure resilience [2].There are no consistent findings regarding the factor structure of SoC-13 when tested using Confirmatory Factor Analysis (CFA) across different cultures [16‐18]. In this regard, a previously validated instrument is not guaranteed to remain valid in another time, culture or context [19]. For example, in a study examining the dimensionality of SoC-13 using CFA in an Italian sample, a one-factor model best fit their data [16]. In contrast, a three factor model of SoC-13 had a better fit than a single factor model in Peruvian college students [20]. In a systematic review of 458 scientific publications and 13 doctoral theses, it was reported that the factorial structure of the scale remain unclear, with some evidence supporting Antonovsky’s single factor solution while other evidence supported two or three factor solutions [17].

Although extensive studies on the topics of resilience and sense of coherence have been carried out among Eritreans in humanitarian settings [2, 7, 10, 21‐23], empirical studies on the adaptation of resilience measures and their validity have not been given due attention. The SoC-13 has been employed to study resilience among displaced Eritreans in all nine languages of Eritrea [2]. Although the SoC-13 has been reported to be contextually appropriate for Eritrean culture and adequate to measure resilience in this population, the basis for this conclusion is qualitative evidence [2, 22]. There has been no quantitative investigation of the psychometric properties of the instrument in this community. Hence, the aim of the present study was to identify the factor structure and examine other psychometric properties of the SoC-13, including internal consistency, construct validity (divergent convergent and discriminant validity) and to address this evidence gap.

Unmodified measure of sense of coherence with thirteen items (SoC-13) has demonstrated acceptable convergent as well as divergent validity. However, when evaluated the overall fit of the unmodified model (SoC-13), with uncorrelated error terms, it demonstrated poor fit to the present data (CFI < 0.95; TLI < 0.95 and RMSEA > 0.08) compared to the fit indices of the same scale with specified model in samples from South Africa [54].

After permission of error terms to correlate, the present data fit best with single latent factor model of Sense of Coherence scale with twelve items (SoC-12) significantly loaded onto a single latent factor (χ² = 57.54, df = 24; χ²/df = 1.69; CFI = 0.98; RMSEA = .035) compared to other computing models of SoC-13 tested based on relevant literature. Hence our data supported the single factor structure of the short form of SoC proposed by the original scale developer [34]. However, when the thirteen items were evaluated for the strength of their loadings, item-1, item-2, item-3 and item-12 demonstrated poor loadings (β < 0.40), and item-2 demonstrated non-significant loading (β = 0.06, p>0.05).

Hence, the internal consistency as measured by Cronbach’s alpha remain stable, even after deleting item-2, and changed from 0.736 for SoC-13 to a slight improvement of 0.748 for SoC-12. Therefore, item- 2 which reads: “Has it happened in the past that you were surprised by the behavior of people whom you thought you knew well?” (see Additional file 1: Table S3) is not valid to Eritrean culture in its current form of presentation, because it has demonstrated very low item loading that doesn’t reach significance level. Similarly, this item was also found to have relatively lower item loadings (β<0.40) as well as considered as problematic items in previous studies done in different cultural contexts and populations [6, 16, 20]. For example, item-2 demonstrated insignificant loadings to a three latent factor model of SoC-11, which fit data for a Dutch speaking Beligian sample [6]. For this particular item, the intact rich social connectedness among Eritreans, being from a collectivist society, may explain why respondents may not have understood the item as part of their concern.

The current findings with respect to lower item loadings of item-1 can be justified by evidence from previous qualitative investigation of how displaced Eritreans reacted to each items of sense of coherence [2]. Their findings indicate that Eritrean participants reacted to item-1: “do you have feeling that you don’t really care about what goes on around you?” by responding to the question itself with feelings of surprise saying: “how can you ask such a question? I wouldn’t be human if I didn’t care about what goes on around me…, this is not in our culture … of course, I care and try to do something about it, if I can.”(Alemdom,2007; p.97) [2]. It means Eritreans value a strong social life such that life for them is entirely interdependent in contrast to people from western cultural background, who may have grown up in a culture where indivudualistic culture is so deeply rooted. For item-3, which reads: “Has it happened that people whom you counted on disappointed you?” the relatively lower item loading onto a single latent factor has a similar justification as in the case for item-1. Hence the deep rooted trust in others, as a matter of a shared cultural norm for Eritreans, may best justify the lower convergent validity of this item too. However, the weaker item loadings of the first three items seen in Eritrean refugee sample in the present study are not unique to the type of population as well as study context. Cross comparison of item loadings across studies for the observed weaker loadings of item-1, item-2 and item-3 indicated that these items also demonstrated relatively weaker convergent validity (weaker loadings of β<0.40) in previous studies conducted in western settings, such as CFA studies conducted in Italy [16], Belgium [6] as well as a non-western setting, such as Peru [20].

Attaining a better fit for a model through model modification and omission of an item or a few items is also a practice of amending measures to be valid in studies conducted in different cultural settings [6, 13, 19, 53, 54]. For example, only 11 items from the Sense of Coherence scale were suggested as justifiable measures of capturing what SoC-13 can perform in Dutch speaking Belgians, after disregarding the two items (i.e. item-2 and item-3) having insignificant loadings [6] (see Table 3). Similar to the Tigrigna version of SoC-12 model with correlated error terms in the present study, error terms were also allowed to correlate in a one factor model of SoC-13 suggested as best fit to data of an Italian sample [16].

However, when comparison of the current data with previous studies done across cultures was made, differences were observed on the strength of loadings for item-12, which asks: “How often do you have the feeling that there’s little meaning in the things you do in your daily life?”. The item loading for this item onto both the one factor as well as the three factors models of SoC-13 is poor (β < 0.40) in Eritrean sample in the current study, while this item sufficiently loaded onto one factor as well as three factors in the Peruvian sample [20] (see Table 3).

With respect to factor structure, the present CFA finding is not in line with previous findings, which supported a three factor model of SoC-13 in old samples of Netherlands [18], patients with morbid obesity [55], and Dutch speaking Belgians [6]. The present study supported a previous study in an Italian sample, whose data best fit a one factor model of SoC-13 [16]. Despite the conflicting findings regarding the dimensionality of SoC-13, the data in the present study for Eritrean refugee sample best fit with one factor structure of sense of coherence scale with 12 items (CFI =0 .982, RMSEA = 0.035 [90%CI = 0.018,0.050], which supported the factor structure proposed by the original scale developer, Antonovsky [34].

The fact that the present study demonstrated very high co-variances among the three correlated latent factors (r ≥ 0.80, p < 0.001) (Fig. 4) may serve as additional evidence to propose that the factors seem to measure similar or same construct in the context of Eritrean refugees living in Ethiopia.

The mean value in the current study suggested that individuals’ sense of coherence is compromised by being a refugee (mean = 39.91) compared to those internal displaced persons as well as non-displaced people whose mean were reported to be 48.94 and 54.84 respectively in the previous study [2] (see Additional file 4: Table S4).

The internal consistency for SoC-13 (Cronbach’s alpha = 0.74) as well as the internal consistency for SoC-12 (Cronbach’s alpha = 0.75) is high, and these are within the range of alpha values reported (i.e. 0.74 to 0.91) [1]. The current alpha coefficient is also in line with a study involving systematic review of 127 studies, which reported the reliability range from 0.70 to 0.92 [17]. Reliability coefficients which range from 0.70 to 0.90 are demonstrating high reliability [56].

Given that there is a paucity of locally adapted measures to measure resilience factors for Eritreans in humanitarian settings, the present measurement study, following rigorous procedures of adaptation filled gaps noted in the previous studies. Making use of a comparative analysis with other previous CFA studies done across cultures with findings of the present study can also be taken as the strength of the study. A comparison of data across studies gives the reader a clearer picture of contrast to the relative relevance of items with weaker loadings across cultures and type of population. However, caution should be taken while making a comparison of findings obtained from the 5-point response format of the Tigrigna version of SoC-13 in the present study with other findings which used a 7- point response format of the same scale in other cultural contexts. Another limitation of the present study is that we did not conduct a sub-sample analysis which would have permitted us to see if the factor structure is confounded by some basic demographic variables, such as age and gender. This study would also have profited if data from comparable groups of sample had been collected, because it would have strengthened the external validity of the findings.

The short form of SoC with twelve items seems an appropriate measure of sense of coherence for Eritrean refugees living in Ethiopia, which should be understood as a uni-dimensional construct. Therefore, the Tigrigna version of sense of coherence, with 12-items is a valid measure with its acceptable internal consistency. Additional inference derived from covariance of the three theoretical latent factor structures (r ≥ 0.80), which demonstrated above the maximum cut-off point for discriminant validity, implies the likelihood of the present data to support a single factor structure of the Tigrigna version of sense of coherence as a more reasonable factor structure compared to a three factor structure to the Eritrean cultural context. Omission of item-2 substantially improved fit indices as well as item loadings for other items. Hence the reduced Tigrigna version SoC-12 is a good measure for assessing resilience and can be taken as a proxy measure of mental wellbeing for Eritreans living in Ethiopian emergency settings. It can be employed by psychiatrists, counselors, social workers, and researchers in clinical as well as non-clinical settings for assessing resilience for the ultimate purpose of generating data helpful to make an informed decision in primary mental health care for the community-based psychosocial intervention as well as counseling. It may also provide a supplementary source of information for clinical decision making.

Future studies should undertake a qualitative study on the phenomenology of sense of coherence in Eritrean communities so that Antonovsk’s problematic items in the Tigrigna version will be further improved and adapted using valid concepts from Eritrean culture. Hence adaptation should consider their frame of understanding to the holistic and overall Eritrean way of life style, including their inner psyche, spiritual, social, familial and community life, their collective culture, trust, hope, tradition and their belief in rituals, etc. In addition, future clinical practice and interventions regarding refugee mental health in humanitarian settings of Africa, like in refugee camps of Ethiopia for Eritreans, should be geared towards alternative use of this measure to assess mental wellbeing rather than being confined solely on the assessment and diagnosis of pathology. Unfortunately this trend of measuring pathology only is primarily practiced as the predominant means of assessment in mental healthcare practices by humanitarian institutions and practitioners working to assist the mental health of refugees or displaced people in such settings. In order for the current findings to be replicable, future research should be carried out using longitudinal study designs. Furthermore CFA study on the validity of sense of coherence based on data from multiple samples of Eritreans is needed in the future to fairly generalize the factorial structure and construct validity of this tool in Eritrean culture.