Skip to main content
Download PDF

Measuring a family sense of coherence: a rasch-based study extending dyadic data analyses

BMC Palliative Care volume 24, Article number: 8 (2025) Cite this article

Abstract

Background

Family sense of coherence (FSOC) seems to reduce distress in the family and promote the well-being of the family. Therefore, getting accurate measurements for families with long-term illnesses is of particular interest. This study explores dyadic data analysis from the dyadic- and single-informant perspectives, and the measurement properties of the FSOC-S12 according to the Rasch model.

Methods

Racked and stacked data from 151 dyads were analyzed according to the polytomous Rasch model.

Results

Notably, both the dyadic- and single-informant perspectives (i.e., racked and stacked data set-ups) showed measurement properties with minor deviations from the Rasch model according to fit statistics. However, most items had disordered thresholds and some problems with local dependency. Item hierarchies were similar in both set-ups and there was no differential item functioning (DIF) by role from the dyadic informant perspective. Four items showed DIF by informant role in the single-informant perspective.

Conclusions

Our approach to handling dyadic data has shown both strengths and limitations in the evaluation of FSOC-S12, and the understanding of FSOC as a construct from the family’s view of the family’s ability as a whole (dyadic-informant perspective) and patient’s and family member’s separate views of the family’s ability as a whole (single-informant perspective).

Peer Review reports

Background

Living with a family member with a long-term illness at the end of life is for most people, a stressful situation [1,2,3]. Therefore, health and well-being are important outcomes for family members caring for a person with a long-term illness [3,4,5]. Sense of coherence (SOC), the core concept in Antonovsky’s salutogenic model, was developed to explain why some people remain healthy in stressful life situations. SOC consists of three interrelated concepts: comprehensibility (i.e., ability to understand the situations), manageability (i.e., access to sufficient resources to manage situations), and meaningfulness (i.e., challenges are worthy of the corresponding energy investment). Antonovsky [6] argues that SOC can be described as a health-protective behaviour pattern and an effective stress buffer, thereby influencing individuals’ ability to manage to live with long-term illness meaningfully.

SOC was initially developed at the level of the individual but has since been extended to the family level as the family sense of coherence (FSOC) construct [6]. There is growing evidence that FSOC reduces psychological distress in the family [7,8,9] and promotes family well-being and functioning [9,10,11]. For instance, proper person/family-centered care for children with cancer and their families has strengthened FSOC and thereby increased the quality of family life [12]. Therefore, identifying families with weak FSOC can be a way for healthcare professionals to identify families in need of support [7].

The Sense of Coherence Scale (SOC-S) was developed to assess SOC at an individual level [13]. Later, Antonovsky and Sourani developed the Family Sense of Coherence Scale; a long 26-item version (FSOC-S26) [14] and a short 12-item version (FSOC-S12) [15]. The FSOC-S12, which is the most commonly used, has been psychometrically evaluated using classic test theory with satisfactory results [7, 10, 16, 17]. However, these studies have some significant limitations; (i) the data were treated as independent even if they were obtained from two persons within the same family [18], and (ii) the studies did not take into account that the data were ordinal, and (iii) that the responses depend on both the item and agent attributes [19,20,21].

As families consist of at least two parts, dyadic studies are common in family research. Nested observations within dyads (e.g., two family members) cannot be assumed to be mutually independent as they share a common context [18]. Violating the independence assumption can create a bias in the test of statistical significance and measures of associations [22]. There are several strategies to handle nonindependence: (i) collect data from one person, (ii) collect data from two persons and treat them as if they were independent (dyadic-informant perspective), or (iii) conduct separate analyses for the two dyad members (single-informant perspective) [23]. At a theoretical level, FSOC as a construct needs to be equally informed by at least two parties in the family [14, 15]. When using the FSOC-S12, this is further referred to as the dyadic-informant perspective. However, families that may need more support can be identified if they have a low level of FSOC or when patients and family members disagree when comparing their levels of FSOC [7]. This can be done by treating family members as unique individuals and then comparing the patients’ and family members’ perspectives, referred to as the single-informant perspective. For the latter one, the power in both groups is reduced and some results might be missed.

According to Kenny et al. [23, 24], analysis of dyadic data can be summarized as shown in Table 1. In line with this table, family as a whole corresponds to the underpinnings of FSOC [14, 15] and can be assessed using the FSOC-S12 either from a dyadic- or single-informant perspective. However, there is no strong agreement how to handle non-independence in the psychometric literature [25].

Table 1 Methods of data collection and analysis relating to different family composition conceptual models
Full size table

To yield reliable and valid measures and be able to make accurate decisions about FSOC, there is a need for well-designed scales with satisfactory measurement properties. This study explores the measurement properties of the 12-item FSOC-S12 according to the Rasch model, from the dyadic- and single-informant perspectives.

Methods

Study design

This psychometric study used data from a previous research project regarding families’ life situations when living with cancer [7, 26]. This study included only complete dyads, with one patient and one family member. The Regional Ethical Review Boards in Linköping, Sweden, approved the study (No. 2014/70-31).

Participants

Participants for the overall research project were recruited from two palliative centres and two oncology clinics in the south of Sweden, between May 2015 and October 2016. The care units were selected through convenience sampling in one large city, a mid-sized city, and two small towns. Patients recruited consecutively were Swedish-speaking, older than 18 years with a diagnosis of cancer in the palliative stage, and each patient invited one family member to participate. Family members were defined as individuals to whom the patient felt linked to via a sense of belonging and engagement in their lives (e.g., spouse, sibling, children or friend) [27]. In total, 179 patients and 165 family members were recruited and took part in the research project [7, 26]. For this study, only complete dyads including one patient and one family member from the same family were selected (n = 151).

Procedure and data collection

Nurses at the palliative centres and oncology clinics distributed oral and written information regarding the study, and each patient and family member were asked to complete a study questionnaire. The questionnaire could be completed by paper-and-pencil or online. Paper based questionnaires were returned in a pre-paid envelope to the research group. The study questionnaire included demographic characteristics and the FSOC-S12 [15]. The FSOC-S12 is constructed as a unidimensional scale, including 12 items representing all three core components of SOC (i.e., comprehensibility, manageability, meaningfulness). An example of an FSOC-S12 item is: ‘To what extent does it seem to you that your family rules are clear to you?’. Responses were collected using a seven-point numerical rating scale [1,2,3,4,5,6,7] with item-specific anchor descriptors. The summed total score has a possible range between 12 and 84; higher scores imply a higher level of FSOC (15). In the Swedish version all items are scored in the same direction [26].

Data analysis

Participants’ demographic characteristics and study variables were presented using descriptive statistics.

To assess the measurement properties of FSOC-S12 from a dyadic- and single-informant perspective, data were analysed according to the polytomous (partial credit) Rasch model [28] using Winsteps® 4.3.1 [29]. Details of the analyses are presented in Table 2 [30,31,32,33,34].

Table 2 Summary of measurement properties of the FSOC-S12, its interpretation from the initial analyses
Full size table

The Rasch model makes two main assertions: (i) the lower the location of an item, the more likely it will be affirmed, and (ii) the higher the location of the attribute an agent has, the more likely they will affirm an item. In turn, the Rasch model enables separate measures of the agent attribute (here the family’s ability to feel FSOC) and the item attribute (here the FSOC task difficulty) on a conjoint interval scale corresponding to the measurement continuum of the FSOC.

Data were racked and stacked to assess the two different perspectives for this study, i.e. a dyadic- and single-informant perspective [35], as illustrated in Figure 1. Racking (Fig. 1a) refers to placing items for the patient and the family member together horizontally, thus providing a dyadic-informant perspective on the FSOC-S12 where each item is treated as two respondent-role specific items. This allows for separate estimates of FSOC task (item) difficulties for patients and family members, respectively. However, the Rasch model assumes that items are locally independent, but when the ‘same’ item is used twice, a potential risk of local dependence is apparent (similar to discussions by Andrich & Krainer [36], Andrich et al [37], Olsbjerg & Christensen [38]. Thus, local dependence needs to be examined. Stacking (Fig. 1b) refers to vertically placing items from patients and family members together, thus representing the single-informant perspective. When stacking data, item difficulties are assumed to be equal across respondent groups (which is formally tested by differential item function, DIF), but differences in dyad-member ability to feel FSOC are allowed. Stacking data also allows differences in perceived FSOC within the dyad (i.e., between patients and family members) to be assessed.

Fig. 1
figure 1

Conceptual visualization for (A) racked data (dyadic-informant perspective), and (B) stacked data, (single informant perspective)

Full size image

To further assess relations, similarities, and differences in FSOC-S12 between a dyadic- and a single-informant perspective, we:

  1. I.

    Compared the hierarchical ordering of items and correlated (Pearson correlation) item measures from the different perspectives as well as plotted agreements (Bland–Altman plot).

  2. II.

    Correlated (Pearson correlation) person measures from the patient and family members, plotted agreements (Bland–Altman plot), and compared (t-tests) individual person measures from patient and family members based on the single-informant perspective.

Results

Sample characteristics

In total, 151 dyads with one patient and one family member were included. The number of women and men was equally distributed among persons with cancer (51% vs. 49%), while there were more women than men among family members (64% vs. 36%). The mean age was 68.4 years for patients and 62.5 years for family members. The majority (69.5 %) of family members had a partner relationship to the patient (Table 3).

Table 3 Characteristics of the dyads (n = 151)
Full size table

Measurement properties of the FSOC-S12: A dyadic-informant perspective

Table 2 provides a summary of the analyses of both disordered and collapsed threshold of FSOC-S12 with a dyadic-informant perspective. The initial analyses revealed problems with reversed thresholds for all except four items. Ordered thresholds were obtained by collapsing response categories 1 and 2 and categories 3 and 4 into two categories for most items (items 2-7 and 9-12). Items 1 and 8 required further collapses; response categories 1, 2, 3 and 4 into one category for item 1 and 1, 2, 3 and 4 as well as response categories 5 and 6 for item 8 into two categories.

Table 4 gives the item measures and fit statistics from the revised analysis with resolved threshold ordering. The items are ordered from lower to higher measures, thus representing a hierarchy from the lowest to the highest item locations. It is clear that the ‘same’ items responded to by either the patient or the family member were located close to each other. For example, item 10 has the lowest location for both patients (-0.75; 2SE, 0.22) and family members (-0.95; 2SE, 0.24) and Item 1 has highest location for both patients (1.01; 2SE, 0.8) and family members (1.11; 2SE, 0.18).

Table 4 FSOC-S12 item statistics with a dyadic-informant perspective following collapsing of thresholds
Full size table

Two items (items 3 and 11) demonstrated misfit among the patients; both showed OUTFIT ZSTD values outside the expected range, and item 11 also exhibited a larger INFIT ZSTD than expected (Table 4). No DIF by role was detected in the dyad, i.e., patient vs. family member (Table 4).

The eigenvalue of unexplained variance in the first contrast was 3.60 and, by examining the three item clusters derived from the loadings in the principal component analysis (PCA) of the fit residuals, it was evident that items were mainly grouped according to the two respondent groups, i.e., patients in cluster 3 and family members in cluster 1 and 2 (Table 4).

Problems with local dependency were identified when residual correlations were above the relative Q3* cut-off (0.24) as found within the dyads (i.e., residual correlations between patients and family members) for item 8 (0.33) and item 9 (0.41). Furthermore, local dependencies were identified for patients between items 7 and 11 (0.26), and for family members between items 2 and 3 (0.29), 3 and 4 (0.26), 5 and 6 (0.25), 3 and 10 (0.27), 9 and 10 (0.37), and 7 and 11 (0.42).

Measurement properties of the FSOC-S12: A single-informant perspective

Table 2 summarises the analyses of both disordered and collapsed threshold of FSOC-S12 from a single-informant perspective. Regarding the analysis from a single-informant perspective, disordered thresholds were present for all items except item 7. Thus, response categories were collapsed using the same strategy as that for the dyadic-informant perspective.

Table 5 provides the item measures and fit statistics from the revised analysis with resolved threshold ordering. As in Table 4, items in Table 5 are ordered from lower to higher measures and the item hierarchy is similar for the analyses of a single-informant perspective compared with a dyadic-informant perspective. This is further elaborated below.

Table 5 Item fit statistics for FSOC-S12 analysis of a single-informant perspective when disordered thresholds are resolved
Full size table

Only item 12 demonstrated misfit (OUTFIT ZSTD value 2.20). The eigenvalue of the unexplained variance in the first contrast was 1.75, and the correlation between clusters 1 and 3 was 1.00. As shown in Table 5, the clusters did not fully correspond to the three concepts proposed by Antonowsky [6], and the three concepts were not strictly hierarchically ordered, although, items stemming from meaningfulness tended to be easier, followed by more challenging items from manageability and the most challenging items from comprehensibility (Table 5).

Four items showed significant DIF by role of the informant. As shown in Table 2, item 8 showed a slight to moderate DIF by role with a slightly higher estimate of task difficulty from family members than from patients, while the DIF size was smaller for items 4, 6 and 12. For item 12 the family members had the highest task location, while for items 4 and 6 patients had the highest task location.

Comparisons between the dyadic- and single-informant perspectives

As described above and presented in Tables 4 and 5, the item hierarchy is similar when comparing analyses based on single- and dyadic-informant perspectives. This is further illustrated in Figures 2A-C, where measures of task location based on the different perspectives and respondents are plotted. The Pearson correlation coefficients ranged from 0.96 to 0.97. Item 4 deviated from the hierarchical structures when comparing the item measure from the family members from the dyadic-informant perspective with the item measure from single-informant perspective, which is shown as an outlying dot in Figure 2C. Thus, estimation of task location for item 4 based on family members from the dyadic-informant perspective set up (y-axis) was higher compared with when estimated from the single-informant perspective (x-axis).This result is also corroborated in the Bland-Altman plot (Figure 2D), showing all items within +/-1.96SD except item 4 when comparing family members from the dyadic-informant perspective with the single-informant perspective.

Fig. 2
figure 2

Pearson correlation plots for task difficulty estimates based on (A) family members vs. patients in analyses with a dyadic-informant perspective, (B) patients from a dyadic-informant perspective vs. a single-informant perspective, and (C) family members from a dyadic-informant perspective vs. a single-informant perspective. Error bars indicate 2SE. Plotted agreements (D) Dyadic informant patient vs family member, single informant vs dyadic informant patient and single informant vs dyadic family member

Full size image

Person locations correlated moderately (0.58) between patients’ and family members’ ratings from a single-informant perspective (Figure 3A) and almost all comparisons where within +/-1.96SD in the Bland-Altman plot (Figure 3B). In seven cases (4%), the patient showed a significantly lower measure than the family members and part of this was reflected in person fit statistics (e.g., three dyads with INFIT ZSTD > 2SD). In contrast, 11 (7%) family members showed a significantly lower measure than the corresponding patients and five of those dyads had INFIT ZSTD > 2SD.

Fig. 3
figure 3

Correlations of person ability estimates within the dyads from a single informant perspective (A) and the comparisons in a Bland-Altman plot (B)

Full size image

Discussion

To the best of our knowledge, this is the first evaluation of the FSOC-S12 using the Rasch model and from a dyadic analytical approach. The choice of the Rasch model in favour to any other model, more commonly used in studies with dyadic data [25], in line with a measurement science perspective, rather than data modelling [19, 39]. We have thus advanced previous work, on the FSOC-S12 by applying the Rasch model, which enable linear measurement based on ordinal observations, separation of person and item attributes, and detailed insights on the measurement properties. Overall, the FSOC-S12 demonstrated minor deviations from the Rasch model among dyads of patients and family members in palliative cancer care. Furthermore, our findings provide novel insights into the FSOC construct and how to measure it from different perspectives using the FSOC-S12. Both data set-ups are useful, and the choice depends on the clinical or research question, the FSOC-S12 can be used to measure the family’s overall perception of FSOC (dyadic-informant perspective) or to measure the patient’s and family member’s individual perceptions of FSOC (single-informant perspective).

Despite the conceptual difference between the dyadic- and single-informant perspectives, the item hierarchy ordering appears to be similar. This provides further support for the construct validity of the FSOC-S12 and could be the start of a coherent construct theory of what less to more sense of coherence in the family means. Practically, this implies that the item hierarchy in FSOC-S12 can inform where a family is located on the continuum and provides clinicians and families with a ‘compass’, pointing the way forward for actions needed [40].

Additional key questions are whether a single-informant perspective is enough to measure FSOC, and what is measured when only one party gives his or her voice about the family as a whole. A single-informant perspective does not consider that the whole family’s experiences are more than the sum of each family member’s experiences [27, 41] and that the family´s reality can be understood as multidimensional, which means that family members’ different descriptions of the same situation may be equally valid [42]. However, the dyadic informant perspective indicates some dimensionality issues with the FSOC-S12. Multidimensionality always exists to some extent [43, 44], although the critical question is whether it is significant enough [45]. At this stage we can only speculate about this; the dimensionality issues may be a consequence of a lack of agreement within families and could be explained in terms of the FSOC referring to different constructs for patients and family members. Furthermore, in the present study, we only included two informants in the family dyads, and one may ask if more informants are needed to provide a reliable measure for the family as a whole [27, 41]. With only two informants in some families, all members have not had the chance to raise their voice.

From a dyadic-informant perspective, the same statement is repeated twice: once when the patient responds and once when the family member responds. It is, therefore, likely that patients’ and family members’ responses are dependent upon each other, thus causing local dependency within the item-pair [36,37,38]. In this study, local dependency was, however, only apparent for two items within the dyads: item 8 [Do you have the feeling that you are being treated unfairly by your family?] and item 9 [When you think about your family, you very often feel how great it is to be alive]. While there are more sophisticated methods for investigating and accommodate local dependencies [36,37,38, 46, 47] , this suggest further qualitative investigations on why local dependency only was found for items 8 and 9 within the dyads and not for the other items. It should also be noted that local dependency is based on fit residuals [48, 49], which is affected by the sample size [50, 51]. Thus, at the present stage with a somewhat small sample size, we recommend a mixed-method approach combining statistical results from a larger sample size with the qualitatively meaning of the items from patients and family members, respectively, to better understand local dependency issues when applying a dyadic-informant perspective on FSOC.

Given the different roles of patients and family members, items may be interpreted differently between these groups in the same way as it might differ between groups of different sex and age and potential DIF may suggest a lack of measurement invariance. In this study, DIF was found present in four items when comparing patients and family members. At the same time, the consequences on person measures were almost negligible. This warrants further exploration and at present we can only speculate about why DIF was only present in some items. DIF may be due to informal roles in the dyads [3, 52], e.g., if one partner has the power to stipulate rules in the family and therefore scores higher on item 4 [Your family rules are clear to you] and on item 8 [Feeling that you are being treated unfairly]. The DIF may also be due to if one of the dyads had a good day and the other one had a bad day, e.g., if one partner has a bad day, therefore, they score lower on item 6 [Your family life seems to you: full of interest]. In further studies, we encourage deeper investigations on the relation between the patient and family member and disease severity, which also warrants larger samples, to get a clearer picture of DIF.

Methodological considerations

There are some methodological consideration with the present study to bear in mind. First, 151 dyads can be considered to be a small sample [53]. One consequence is that it may have affected threshold ordering negatively due to the few respondents using each response category [54]. Disordered thresholds may have other explanations, for example that only the extreme response categories are labelled [55] and difficulties differentiating between seven levels [56]. Therefore, further studies with larger samples are needed to better understand potential problems with model fit, DIF and local dependency [50, 51] and before any firm conclusions can be drawn. The small sample is a consequence of the fact that the present study is based on data from a previous study that was not designed to address dyadic data analyses. Another limitation is that the type of relationship between the patient and family member has not been considered in the present study. The reason was that a vast majority were partners. The design of the present study did not allow any drop out analysis. Therefore, we cannot exclude any type of attrition bias. There is signs of attrition bias due to socioeconomic status since a large share of the participants had a university degree and a high income level. Even if this is a threat for the external validity, it is of minor importance for the psychometric properties. This is particularly true for the Rasch model which is sample independent in contrast to models under classical test theory. Therefore, the result should be carefully generalised. However, the insights from this study can be of value to better understand and improve the FSOC-S12 as well as for designing future studies using dyadic data.

There is a risk that items are regarded as misfitting due to too large sample sizes (type I error), and conversely, there is a risk of not identifying misfitting items correctly with too small sample sizes (type II errors). Those risks in relation to sample size are also affected by weather conditional or unconditional infit or outfit statistics are used [57]. Winsteps, which was used in the present study, provides unconditional models which may be associated with inflated type I error rates at sample sizes of 250-500 or more [57]. While our sample is not associated with any obvious risks unreliable fit statistics, those results should be interpreted with some caution.

Conclusion

Our approach to handling dyadic data has shown both strengths and limitations in the evaluation of FSOC-S12. This study provides important insights into the dyadic- and single-informant perspectives when using the FSOC-S12 in family research. Notably, both perspectives showed minor deviations from the Rasch model. Depending on the clinical or research question to respond, at present, the FSOC-S12 may be used to provide meaningful measures of family’s view of the family ability as a whole (dyadic-informant perspective) or measures of patient’s and family member’s own view of the family ability as a whole (single-informant perspective). However, we encourage further studies to consider three closely related conceptual and methodological aspects, (i) conceptual differences between the dyadic- and single-informant perspectives, (ii) if a single-informant perspective is enough to measure family as a whole, and (iii) if more than two informants are needed to measure the family as a whole.

Data availability

The datasets generated and analysed during the current study are not available for public use, due to confidentiality, but are available from the corresponding author on reasonable request.

Abbreviations

FSOC:

family sense of coherence

FSOC-S12:

Family sense of coherence scale 12 items

FSOC-S26:

Family sense of coherence scale 26 items

DIF:

differential item functioning

INFIT:

Inlier-pattern sensitive statistics

LD:

Local dependency

MNSQ:

Mean square fit statistics

OUTFIT:

Outlier-pattern sensitive statistics

PCA:

principal component analysis

SE:

Standard error

SOC:

Sense of coherence

SOC-S:

Sense of Coherence Scale

ZSTD:

z-standardized fit statistics

References

  1. Fjose M, Eilertsen G, Kirkevold M, Grov EK. A valuable but demanding time family life during advanced cancer in an elderly family member. Adv Nurs Sci. 2016;39(4):358–73. https://doiorg.publicaciones.saludcastillayleon.es/10.1097/ANS.0000000000000145.

    Article  Google Scholar 

  2. Möllerberg ML, Sandgren A, Lithman T, Noreen D, Olsson H, Sjövall K. The effects of a cancer diagnosis on the health of a patient’s partner: a population-based registry study of cancer in Sweden. Eur J Cancer Care. 2016;25(5):744–52. https://doiorg.publicaciones.saludcastillayleon.es/10.1016/S0959-8049(15)30025-3.

    Article  Google Scholar 

  3. Möllerberg ML, Sandgren A, Swahnberg K, Benzein E. Familial interaction patterns during the palliative phase of a family member living with cancer. J Hosp Palliat Nurs. 2017;19(1):67–74. https://doiorg.publicaciones.saludcastillayleon.es/10.1111/ecc.12487.

    Article  Google Scholar 

  4. Regonal Cancer Center. National palliative care guidelines in Sweden. Available from: https://kunskapsbanken.cancercentrum.se/diagnoser/palliativ-vard/vardprogram/.

  5. Thomas PA, Liu H, Umberson D. Family relationships and well-being. Innov. Aging. 2017;1(3):igx025. https://doiorg.publicaciones.saludcastillayleon.es/10.1093/geroni/igx025.

    Article  Google Scholar 

  6. Antonovsky A. Unraveling the mystery of health: How people manage stress and stay well. San Francisco, CA, US: Jossey-Bass; 1987.

    Google Scholar 

  7. Möllerberg ML, Årestedt K, Swahnberg K, Benzein E, Sandgren A. Family sense of coherence and its associations with hope, anxiety and symptoms of depression in persons with cancer in palliative phase and their family members: A cross-sectional study. Palliat Med. 2019;33(10):1310–8. https://doiorg.publicaciones.saludcastillayleon.es/10.1177/0269216319866653.

    Article  PubMed  Google Scholar 

  8. Ngai FW, Ngu SF. Family sense of coherence and quality of life. Qual Life Res. 2013;22(8):2031–9.

    Article  PubMed  Google Scholar 

  9. Ngai FW, Ngu SF. Family sense of coherence and family adaptation among childbearing couples. J Nurs Scholarsh. 2014;46(2):82–90.

    Article  PubMed  Google Scholar 

  10. Larsen Moen Ø, Hall-Lord ML. Reliability and validity of the Norwegian family sense of coherence scale. Open J Nurs. 2016;06(12):1075–86. https://doiorg.publicaciones.saludcastillayleon.es/10.4236/ojn.2016.612102.

    Article  Google Scholar 

  11. Lustig DC, Akey T. Adaptation in Families with adult children with mental retardation: impact of family strengths and appraisal. Educ Train Ment Retard Dev Disabil. 1999;34(3):260–70.

    Google Scholar 

  12. Heydari F, Nikfarid L, Oujian P, Nasiri M, Motefakker S. Predictors of perceived family sense of coherence in parents of children with cancer. Indian J Palliat Care. 2022;28(1):28–36. https://doiorg.publicaciones.saludcastillayleon.es/10.25259/IJPC_107_21.

  13. Antonovsky A. The structure and properties of the sense of coherence scale. Soc Sci Med. 1993;36(6):725–33. https://doiorg.publicaciones.saludcastillayleon.es/10.1016/0277-9536(93)90033-Z.

    Article  CAS  PubMed  Google Scholar 

  14. Antonovsky A, Sourani T. Family Sense of Coherence and Family Adaptation. J Marriage Fam. 1988;50(1):79–92. https://doiorg.publicaciones.saludcastillayleon.es/10.2307/352429.

    Article  Google Scholar 

  15. Sagy S. Effects of Personal, Family, and Community Characteristics on Emotional Reactions in a Stress Situation: The Golan Heights Negotiations. Youth Soc. 1998;29(3):311–29. https://doiorg.publicaciones.saludcastillayleon.es/10.1177/0044118x98029003003.

    Article  Google Scholar 

  16. Çeçen AR. The Turkish Version of the Family Sense of Coherence Scale- Short Form (FSOC-S): Initial Development and Validation. Educational Sciences: Theory and Practice. 2007;7(3):1211–8.

    Google Scholar 

  17. Ngai FW, Ngu SF. Translation and Validation of a Chinese Version of the Family Sense of Coherence Scale in Chinese Childbearing Families. Nurs Res. 2011;60(5):295–301.

    Article  PubMed  Google Scholar 

  18. Stemmler M, von Eye A, Wiedermann W. Dependent data in social sciences research. Springer Proceedings in Mathematics and Statistics, vol. 145. Springer: Berlin; 345–367. 2015. Available from: http://link.springer.com/https://doiorg.publicaciones.saludcastillayleon.es/10.1007/978-3-319-20585-4

  19. Andrich D. Controversy and the rasch model: a characteristic of incompatible paradigms? Med Care. 2004;42(1):I7–16.

    Article  PubMed  Google Scholar 

  20. Hobart JC, Cano SJ, Zajicek JP, Thompson AJ. Rating scales as outcome measures for clinical trials in neurology: problems, solutions, and recommendations. Lancet Neurol. 2007;6(12):1094–105. https://doiorg.publicaciones.saludcastillayleon.es/10.1016/S1474-4422(07)70290-9.

    Article  PubMed  Google Scholar 

  21. Pendrill L. Man as a Measurement Instrument. NCSLI Meas. 2014;9(4):24–35.

    Article  Google Scholar 

  22. Kenny DA, Judd CM. Consequences of violating the independence assumption in analysis of variance. Psychol Bull. 1986;99:422–31. https://doiorg.publicaciones.saludcastillayleon.es/10.1037/0033-2909.99.3.422.

    Article  Google Scholar 

  23. Kenny DA. Commentary: Dyadic Analyses of Family Data. J Pediatr Psychol. 2011;36(5):630–3. https://doiorg.publicaciones.saludcastillayleon.es/10.1093/jpepsy/jsq124.

    Article  PubMed  Google Scholar 

  24. Kenny DA, Kashy DA, Cook WL. Dyadic data analysis. New York, NY, US: Guilford Press; 2006.

    Google Scholar 

  25. Claxton SE, DeLuca HK, Van Dulmen MHM. Testing psychometric properties in dyadic data using confirmatory factor analysis: current practices and recommendations. TPM - Test Psychom Methodol Appl Psychol. 2015;22(2):181–98. https://doiorg.publicaciones.saludcastillayleon.es/10.4473/TPM22.2.2.

    Article  Google Scholar 

  26. Möllerberg ML, Årestedt K, Sandgren A, Benzein E, Swahnberg K. Adaptation and psychometric evaluation of the short version of family sense of coherence scale in a sample of persons with cancer in the palliative stage and their family members. Palliat Support Care. 2020;18(1):24–32. https://doiorg.publicaciones.saludcastillayleon.es/10.1017/S1478951519000592.

    Article  PubMed  Google Scholar 

  27. Wright LM, Leahey M. Nurses and families: a guide to family assessment and intervention. Philadelphia: F.A. Davis; 2013.

    Google Scholar 

  28. Wright BD, Masters GN. Rating scale analysis. Chicago: Mesa Press; 1982.

    Google Scholar 

  29. Linacre JM. Winsteps® Rasch measurement computer program. Beaverton, Oregon: Winsteps.com; 2020. https://www.winsteps.com/index.htm.

  30. Hobart JC, Cano SJ. Improving the evaluation of therapeutic interventions in multiple sclerosis: the role of new psychometric methods. Health Technol Assess. 2009;13(12). https://doiorg.publicaciones.saludcastillayleon.es/10.3310/hta13120.

  31. Tennant A, Conaghan PG. The Rasch measurement model in rheumatology: What is it and why use it? When should it be applied, and what should one look for in a Rasch paper? Arthritis & Rheumatism-Arthritis Care & Research. 2007;15;57(8):1358–1362. https://doiorg.publicaciones.saludcastillayleon.es/10.1002/art.23108.

  32. Andrich D. A rating formulation for ordered response categories. Psychometrika. 1978;43(4):561–73. https://doiorg.publicaciones.saludcastillayleon.es/10.1007/BF02293814.

    Article  Google Scholar 

  33. Linacre JM. What do infit and outfit, mean-square and standardized mean – ScienceOpen. 2002. Available from: https://www.scienceopen.com/document?vid=2bac12d4-945c-4baf-bd16-725c824b40da

  34. Boone WJ, Staver JR. Principal component analysis of residuals (PCAR). In: Boone WJ, Staver JR, editors. Advances in rasch analyses in the human sciences. Cham: Springer International Publishing; 2020. Available from: https://doiorg.publicaciones.saludcastillayleon.es/10.1007/978-3-030-43420-5_2.

  35. Wright BD. Rack and Stack: Time 1 vs. Time 2 or Pre-Test vs. Post-Test. Rasch Meas Trans. 2003;17(1):905–6. https://www.rasch.org/rmt/rmt171a.htm.

  36. Andrich D, Kreiner S. Quantifying response dependence between two dichotomous items using the rasch model. Appl Psychol Meas. 2010;34(3):181–92. https://doiorg.publicaciones.saludcastillayleon.es/10.1177/0146621609360202.

    Article  Google Scholar 

  37. Andrich D, Humphry SM, Marais I. Quantifying local, response dependence between two polytomous items using the rasch model. Appl Psychol Meas. 2012;36(4):309–24. https://doiorg.publicaciones.saludcastillayleon.es/10.1177/0146621612441858.

    Article  Google Scholar 

  38. Olsbjerg M, Christensen KB. Modeling local dependence in longitudinal IRT models. Behav Res Methods. 2015;47(4):1413–24. https://doiorg.publicaciones.saludcastillayleon.es/10.3758/s13428-014-0553-0.

    Article  PubMed  Google Scholar 

  39. Cano SJ, Hobart JC. The problem with health measurement. Patient Prefer Adherence. 2011;5:279–90. https://doiorg.publicaciones.saludcastillayleon.es/10.2147/PPA.S14399.

    Article  PubMed  PubMed Central  Google Scholar 

  40. Barbic SP, Cano SJ, Mathias S. The problem of patient-centred outcome measurement in psychiatry: why metrology hasn’t mattered and why it should. J Phys Conf Ser. 2018;1044:012069. https://doiorg.publicaciones.saludcastillayleon.es/10.1088/1742-6596/1044/1/012069

  41. Bell JM, Wright LM. The illness beliefs model: advancing practice knowledge about illness beliefs, family healing, and family interventions. J Fam Nurs. 2015;21(2):179–85. https://doiorg.publicaciones.saludcastillayleon.es/10.1177/1074840715586889.

    Article  PubMed  Google Scholar 

  42. Bateson G, Brand S, Goleman D, May R, Rieber R, Wilden A. The pattern that connects : eco-cybernetic texts[Swedish: Mönstret som förbinder:eko-cybernetiska texter]. Stockholm: Mareld; 1998.

    Google Scholar 

  43. Cano SJ, Barrett LE, Zajicek JP, Hobart JC. Dimensionality is a relative concept. Mult Scler J. 2011;17(7):893–4. https://doiorg.publicaciones.saludcastillayleon.es/10.1177/1352458511406910.

    Article  Google Scholar 

  44. Linacre JM. Dimensionality: when is a test multidimensional?: Winsteps Help. 2023 Available from: https://www.winsteps.com/winman/dimensionality.htm.

  45. Melin J, Cano S, Pendrill L. The role of entropy in construct specification equations (CSE) to improve the validity of memory tests. Entropy. 2021;23(2):212. https://doiorg.publicaciones.saludcastillayleon.es/10.3390/e23020212.

    Article  PubMed  PubMed Central  Google Scholar 

  46. Andrich D. Recognizing a Violation of the assumption of local independence in the analysis of test data. Educ Psychol Meas. 1983;43(3):829–33. https://doiorg.publicaciones.saludcastillayleon.es/10.1177/001316448304300318.

    Article  Google Scholar 

  47. Marais I. Local Dependence. In: Christensen KB, Kreiner S, Mesbah M. Rasch models in health. Wiley; 2012. Available from: https://onlinelibrary.wiley.com/doi/https://doiorg.publicaciones.saludcastillayleon.es/10.1002/9781118574454.ch7

  48. Christensen KB, Makransky G, Horton M. Critical values for Yen’s Q3: identification of local dependence in the rasch model using residual correlations. Appl Psychol Meas. 2017;41(3):178–94. https://doiorg.publicaciones.saludcastillayleon.es/10.1177/0146621616677520.

    Article  PubMed  Google Scholar 

  49. Marais I. Response dependence and the measurement of change. J Appl Meas. 2009;10(1):17–29.

    PubMed  Google Scholar 

  50. Chen WH, Lenderking W, Jin Y, Wyrwich KW, Gelhorn H, Revicki DA. Is rasch model analysis applicable in small sample size pilot studies for assessing item characteristics? An example using PROMIS pain behavior item bank data. Qual Life Res. 2014;23(2):485–93.

    Article  PubMed  Google Scholar 

  51. Hagell P, Westergren A. Sample size and statistical conclusions from tests of fit to the rasch model according to the rasch unidimensional measurement model (RUMM) program in health outcome measurement. J Appl Meas. 2016;17(4):416–31.

    PubMed  Google Scholar 

  52. Muliira JK, Kizza IB, Nakitende G. Roles of family caregivers and perceived burden when caring for hospitalized adult cancer patients: perspective from a low-income country. Cancer Nurs. 2019;42(3):208–17. https://doiorg.publicaciones.saludcastillayleon.es/10.1097/NCC.0000000000000591.

    Article  PubMed  Google Scholar 

  53. Houts CR, Edwards MC, Wirth RJ, Deal LS. A review of empirical research related to the use of small quantitative samples in clinical outcome scale development. Qual Life Res. 2016;25(11):2685–91.

    Article  PubMed  Google Scholar 

  54. Linacre JM. Optimizing Rating Scale Category Effectiveness. J Appl Meas. 2002;3(1):85–106.

    PubMed  Google Scholar 

  55. Weijters B, Cabooter E, Schillewaert N. The effect of rating scale format on response styles: the number of response categories and response category labels. Int J Res Mark. 2010;27(3):236–47. https://doiorg.publicaciones.saludcastillayleon.es/10.1016/j.ijresmar.2010.02.004.

    Article  Google Scholar 

  56. Simms LJ, Zelazny K, Williams TF, Bernstein L. Does the number of response options matter? Psychometric perspectives using personality questionnaire data. Psychol Assess. 2019;31(4):557–66. https://doiorg.publicaciones.saludcastillayleon.es/10.1037/pas0000648.

    Article  PubMed  Google Scholar 

  57. Müller M. Item fit statistics for Rasch analysis: can we trust them? J Stat Distrib Appl. 2020;7(1). https://doiorg.publicaciones.saludcastillayleon.es/10.1186/s40488-020-00108-7.

  58. The World Medical Association. World medical association declaration of Helsinki – ethical principles for Medical Research Involving Human Subjects. JAMA. 2013;310(20):2191–4. https://doiorg.publicaciones.saludcastillayleon.es/10.1001/jama.2013.281053.

    Article  CAS  Google Scholar 

Download references

Acknowledgements

The authors want to acknowledge and thank all the participants in this study.

Funding

Open access funding provided by Malmö University. The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: Linnaeus University, Sweden; RISE (Research institutes of Sweden), Sweden; Kristianstad University, Sweden; Cancer Foundation in Kronoberg County, Sweden; and the Kamprad Family Foundation for Entrepreneurship, Research & Charity, Sweden, funded this study.

Author information

Authors and Affiliations

  1. Department of Care Science, Malmö University, Faculty of Health and Society, Malmö, Sweden

    Marie-Louise Möllerberg

  2. RISE Research Institutes of Sweden, Division Safety and Transport, Measurement Science and Technology, Göteborg, Sweden

    Marie-Louise Möllerberg & Jeanette Melin

  3. Faculty of Health and Life Sciences, Linnaeus University, Kalmar, Sweden

    Kristofer Årestedt

  4. Department of Research, Region Kalmar County, Kalmar, Sweden

    Kristofer Årestedt

  5. The PRO-CARE Group, Faculty of Health Sciences, Kristianstad University, Kristianstad, Sweden

    Peter Hagell

  6. Department of Leadership, Demand and Control, Swedish Defence University, Karlstad, Sweden

    Jeanette Melin

Authors
  1. Marie-Louise Möllerberg
    View author publications

    You can also search for this author inPubMed Google Scholar

  2. Kristofer Årestedt
    View author publications

    You can also search for this author inPubMed Google Scholar

  3. Peter Hagell
    View author publications

    You can also search for this author inPubMed Google Scholar

  4. Jeanette Melin
    View author publications

    You can also search for this author inPubMed Google Scholar

Contributions

M-LM: Conception, design, data acquisition and interpretation, drafting the article, all revisions and final approval. KÅ: Conception, design, interpretation, drafting the article, all revisions and final approval. PH: Conception, design, interpretation, drafting the article, all revisions and final approval. JM: Conception, design, data acquisition and interpretation, drafting the article, all revisions and final approval.

Corresponding author

Correspondence to Marie-Louise Möllerberg.

Ethics declarations

Ethics approval and consent to participate

The research committees at all involved palliative centres and oncology clinics and the Regional Ethical Review Board in Linköping, Sweden, approved the study (No. 2014/70-31). The study following the ethical principles of the Declaration of Helsinki and all participants received oral and written study information and provided written informed consent to participate [58].

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.

Additional information

Publisher’s Note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Rights and permissions

Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.

Reprints and permissions

About this article

Check for updates. Verify currency and authenticity via CrossMark

Cite this article

Möllerberg, ML., Årestedt, K., Hagell, P. et al. Measuring a family sense of coherence: a rasch-based study extending dyadic data analyses. BMC Palliat Care 24, 8 (2025). https://doiorg.publicaciones.saludcastillayleon.es/10.1186/s12904-024-01639-5

Download citation

  • Received:

  • Accepted:

  • Published:

  • DOI: https://doiorg.publicaciones.saludcastillayleon.es/10.1186/s12904-024-01639-5

Keywords

BMC Palliative Care

ISSN: 1472-684X

Contact us