Social support is a complex multidimensional construct that is integral to an individual's physical and psychological well-being. A number of authors have proposed that social support buffers the effect of stressful life events through coping mechanisms,1-3 making it an important construct for clinical nurse specialists (CNSs) and other advanced practice nurses (APNs) interested in preventing illness and promoting health among their patients. The use of well-tested instruments that purport to measure social support, demonstrate reliability, and produce valid data is critical for testing interventions designed to improve patient and client outcomes, improve quality, and decrease healthcare costs.4
BACKGROUND
Social support is a relational transaction between individuals.5 Three primary properties of social support include social embeddedness, perceived social support, and enacted support.6 In other words, an act of support occurs in the context of a social interaction and must be "perceived" as having been offered. From the position of the person offering, social support has been broadly defined as an individual's dependability and ability to show love toward others.3
In a considerably large body of literature, researchers have examined the role of social support in physical and emotional well-being, coping, and stress management.7-10 However, knowledge of how social support differs across gender, age, and racial ethnic groups is lacking, as is the much needed testing of social support interventions. Measuring social support using a reliable instrument that produces valid research data across different groups of patients is of critical importance in moving this research forward.
Clinical nurse specialists continue to contribute to the improvements in patient outcomes and cost-effective care through performing advanced nursing roles such as conducting research.11 Research is an important process that will enable CNSs and other APNs to examine the relationship between social support and healthcare outcomes among different groups of patients and across settings. Such studies have the potential to establish new approaches to care provision that incorporate social support. In anticipation of this much needed research, the purpose of this article is to critique the psychometric properties of the Multidimensional Scale of Perceived Social Support (MSPSS).
DESCRIPTION OF INSTRUMENT
Zimet et al5 developed the MSPSS to subjectively assess the perception of social support adequacy. In other words, the instrument measures levels of perceived social support by the patient or client completing the instrument. It was designed to be self-administered and brief. The MSPSS specifically assesses perceptions of social support from 3 different sources: family, friends, and significant other. As a concept, social support has been previously measured using a number of different scales. Other instruments include the Duke-UNC Functional Social Support Questionnaire, Medical Outcomes Study Social Support Survey, Norbeck Social Support Questionnaire, Perceived Social Support Scale, The Social Provisions Scale, and Social Support Questionnaire. Beyond the scope of this article, details about the weaknesses in these instruments have been identified in other publications.5,12,13
MSPSS Instrument Scoring
The initial version of MSPSS consisted of 24 items focused on measuring social support relationships between an individual and their family, friends, and significant other. Based on a number of pilot studies, repeated exploratory factorial analyses (EFAs) were conducted, resulting in the exclusion of 12 items that did not psychometrically form into consistent factors related to the categories of family, friends, and significant others. Therefore, the current version of the MSPSS consists of 12 items. Three social support subscales contain 4 items each: (1) family (FA) (ie, "My family is willing to help me make decisions"), (2) friends (FR) (ie, "I can talk about my problems with friends"), and significant other (SO) (ie, "There is a special person who is around when I am in need"). Early in its development, in order to enhance variability and avoid a ceiling effect of the responses, Zimet et al5 changed the MSPSS's responses from a 5- to 7-point Likert scale ranging from "very strongly disagree" (1) to "very strongly agree" (7). Each subscale score can range from 4 to 28. Items are summed, and a total score is also calculated and ranges from 12 to 84. Higher subscale and total scores indicate high levels of perceived social support.
RELIABILITY AND VALIDITY
Zimet et al5 used different procedures to establish the psychometrics of the current version of the MSPSS by using the scores of 275 undergraduate university students from different years of enrollment. Initially, in order to establish the measure's construct validity, students were asked to complete the MSPSS and the Hopkins Symptom Checklist. In later studies, MSPSS psychometrics were confirmed by Zimet et al14 and extended by Dahlem et al.15 In addition, the instrument's psychometric properties were tested and validated among different populations such as psychiatric outpatients,16 older adults,17 and adolescents.18 Of note, the majority of testing has occurred in nonclinical populations. Furthermore, MSPSS psychometric testing has been conducted on 2 different language versions: Thai19 and Polish.20 The following sections highlight the MSPSS psychometric properties through a discussion of reliability and validity, including factor analysis.
Reliability
Data collected using the MSPSS were analyzed for internal consistency using Cronbach's [alpha] reliability coefficient. According to Waltz et al,4 Cronbach's [alpha] is the preferred index of internal consistency reliability. The reliability can range from between .00 and 1.00, with .8 or greater as a highly desirable value.21 The MSPSS internal reliability has been widely tested, showing strong internal consistency for the measure's total score (0.93-0.98) and for the subscales (0.91-0.81).5,14,15,17-19,22
The MSPSS's stability over time has been examined using the test-retest procedure. Because this testing requires 2 administrations, the MSPSS was repeated 2 to 3 months after the initial administration by Zimet et al,5 and after a 4-week period by Wongpakaran et al.19 Both repeated administration times were beyond the recommended time span of 2 weeks,4 indicating the instrument is stable for an even longer period. The MSPSS total scores and 3 subscales demonstrated adequate stability over time, with strong correlations between the 2 administrations ranging between .72 and .85.5,19 Based on the previously mentioned studies, the MSPSS has strong internal reliability.
Validity
Ideally, it is best to determine to what extent the content of the instrument items and subscales reflects the construct measure.23 As a reminder, "Validity is about the interpretation of the scores generated from an instrument."24(p136) A variety of approaches can be used to determine the validity of the data generated from an instrument.
The MSPSS has been examined across different populations and settings. This is important because the extent to which an instrument's data are valid can differ across different groups and settings.24 The MSPSS validity has been established in a few populations where researchers have taken different approaches to validity testing among these populations.
Initially, a hypothesis-testing approach was adopted by Zimet et al5 to examine MSPSS divergent validity, a form of construct validity, among 275 undergraduates. The psychometric testing team hypothesized that perceived social support test scores would be negatively correlated with reported anxiety and depression symptoms, 2 subscales on the Depression and Anxiety subscales of the Hopkins Symptom Checklist. In other words, when anxiety and depression scores were high, the social support scores would be low. The results of these hypotheses were partially supported. The MSPSS total score demonstrated moderate divergent validity based on a significant negative correlation with depression (r = - 0.25, P < .01). Despite the fact that all 3 MSPSS subscales were significantly inversely correlated to the depression subscale (family: r = -0.24, P < .01; friends r = -0.24, P < .01) and significant other (r = -0.13, P < .05), only 1 subscale, the family subscale, was significantly inversely correlated with anxiety (r = -0.18, P < .01).
Two years later, Zimet et al14 used a population-specific, contrasted group approach and confirmed 2 of the MSPSS psychometric subscale properties held up in 2 different groups: (1) male and female adolescents (n = 74), and (2) and male pediatric residents (n = 55). The FA subscale validity was confirmed using the adolescent sample when the participants were asked only to compare social support from their mothers to the other 2 groups (friends and significant others) (F1,68 = 34.47, P < .001). Next, hypothesizing that married residents compared with unmarried residents would score higher on the SO subscale, this subscale was deemed valid (F1,46 = 16.50, P < .001) for use in that population.
In 2011, Wongpakaran et al19 established both divergent and convergent validity using data from 462 Thai participants. A negative correlation of the total score with both the well-known State Trait Anxiety Inventory (r = - 0.20, P = .004) and the Thai Depression Inventory (r = -0.19, P = .007) was found. Expecting that those with high social support would also have high self-esteem, the MSPSS total score was positively correlated with the Rosenberg Self-esteem Scale (r = 0.33, P < .0001), establishing convergent validity.
More recently, Khalil22 examined validity for use among Arab American women. Divergence was established through a significant negative correlation (r = -0.27, P < .01) with the Psychological Stress Measure 9 using data collected from 267 Arab American women.
Factor Analysis. According to Waltz et al,4 factor analysis is one of the procedures that is used to assess construct validity. As background, literature on the construct of social support has presented 4 hypothesized models (constructs) of perceived social support sources. First, Zimet et al5 proposed 3 distinct forms of perceived social support (family, friends, and significant others). Second, Stanley et al17 and Chou25 proposed 2 factors, which included (1) friends and (2) combination of family and significant others. Finally, Clara et al26 proposed a single global social support factor. Because the MSPSS is presented here, the MSPSS factor analysis is the focus.
As mentioned previously, the initial version of MSPSS consisted of 24 items focused on measuring relationships with family, friends, and significant others. Repeated EFAs were conducted, resulting in 3 factors (the FA, FR, and SO subscales) with eigenvalues exceeding 0.5.5,7,15,18 An eigenvalue is a statistic used during the EFA procedure to determine how many total factors best explain the construct (eg, perceived social support).27 For the MSPSS's EFA procedure, there were 3 factors that resulted in having eigenvalues greater than 0.5. The 24 items were then reduced because only 12 of the 24 items correlated with those 3 factors.
Zimet et al14 also used confirmatory factor analysis (CFA) with varimax rotation to confirm the MSPSS 3 subscales' structure. The CFA demonstrated participants' responses to the items clearly differentiated between 3 sources of social support (family, friends, and significant others). Edwards28 also examined factorial validity using EFA with varimax rotation in a group of Mexican American youth. The same 3 subscales were extracted; consistent with the studies of Zimet et al.5,14 Similarly, Vaingankar et al29 confirmed the scale's structure of 3 subscales through CFA among outpatients with schizophrenia in Singapore.
Social desirability poses a threat to both reliability and validity. This phenomenon occurs when a respondent or research participant responds to items in order to appear socially acceptable. Social desirability is a greater threat when administering instruments that are based on social constructs such as in the MSPPS. Therefore, estimates for measuring social desirability are recommended as part of psychometric testing14 using such instruments as the Marlowe-Crowne (MC) Social Desirability Scales.4 Dahlem et al15 demonstrated that there was a nonsignificant correlation between the MSPSS subscales and total scores with the short version of the MC Social Desirability Scale (total MSPSS-MC, r = -0.01, P > .05; FA MSPSS-MC r = 0.02, P > .05; FR MSPSS-MC, r = -0.01, P > .05; SO MSPSS-MC, r = 0.03, P > .05).
SUMMARY, LIMITATIONS, AND RECOMMENDATIONS
In summary, the MSPSS is a brief 12-item instrument measuring perceived social support from family members, friends, and a significant other. The MSPSS was designed for easy and quick self-administration. Therefore, it may be an appropriate instrument to administer in clinic and hospital settings.
The instrument has been translated into a number of different languages and tested in populations in and outside the United States. Clinical nurse specialists and other APNs should consider using the instrument in diverse populations, keeping in mind to match those populations where the appropriate psychometric testing has been completed, as the meaning and sources of social support may differ across gender, cultures, and age groups.
The MSPSS has been tested using both cardiac and psychiatric patients and therefore demonstrates reliability and validity in clinical populations. Therefore, it is recommended that CNS researchers use the MSPSS in future research. However, conducting psychometric testing across different clinical populations would further enhance the generalizability of the MSPSS. Clinical populations for such psychometric testing might include patients and clients with pain, nausea, fatigue, early-stage dementia, frequent falls, and declining functional status.30-32 In summary, the MSPSS has been shown to be psychometrically sound across a number of populations, having demonstrated adequate reliability and validity, including robust factorial validity.
References