Psychometric properties of the 45-item supportive care needs survey—partners and caregivers - Dutch (SCNS-P&C45-D) in partners of patients with breast cancer

Procedure

Partners of patients with BC (hereafter named participants) were invited to participate in this cross-sectional psychometric study,. The method of inviting participants varied based on the preferences of the hospital. In five hospitals, participants were invited either directly via a research nurse or indirectly via the patient who was asked by a nurse to hand over the questionnaire to their partner. All participants were asked to return the questionnaires to the researcher (EP) in a Reply Paid envelope (Fig. 1).

Participants

To be eligible, participants had to be 18 years or older and have an intimate relationship with a patient diagnosed with BC who was receiving curative treatment or had received primary treatment ≤5 years ago. All participants were native Dutch speakers who could adequately read, speak and understand Dutch. We specified no criteria regarding the patient’s and partner’s gender since breast cancer is very rare in males.

Data collection

Other measures for discriminant and convergent validity

Partners’ demographics and questions related to the patients’ illness such as time since diagnosis and type of treatments were reported. The Hospital Anxiety and Depression scale (HADS) was used to examine psychological distress and to determine convergent validity [21]. Each of the 14-items was rated on a scale from zero (not at all) to three (very much). Higher scores imply more psychological distress. The total score was used to describe psychological distress with a cut-off point of 11 [21]. The HADS revealed Cronbach’s alpha’s from 0.71 to 0.90 for the total scale and both subscales [21]. The EDIZ (Self-Perceived burden from Informal Care) is a Dutch instrument to measure the self-perceived burden from informal care with 9-items on a 5-point Likert scale. This instrument was validated for partners of patients with dementia [22] and also used for partners of cancer patients in palliative care [23]. Higher scores imply higher burden due to informal care. The EDIZ revealed Cronbach’s alpha of 0.79 [23]. The EDIZ was used to determine convergent validity too.

Ethical considerations

Permission to conduct the study was obtained from the regional ethics review board (Radboud university medical center, no. 2013–350). The committee concluded that no detailed ethical approval was necessary. Returning the questionnaire was considered as an implicit informed consent. All procedures performed in studies involving human participants were in accordance with the ethical standards of the institutional and/or national research committee and with the 1964 Helsinki declaration and its later amendments or comparable ethical standards.

Data analysis

Psychometric analyses of the SCNS-P&C45-D were performed using SPSS 22 (SPSS Inc., Chicago, IL, USA). When a participant left more than 33.3% of the SCNS-P&C45 questions blank, the participant was excluded from further analysis due to lacking data. This was done to maintain the reliability of remaining data. Remaining missing data were omitted per item, leading to a fluctuating number of participants answering a certain item. Descriptive statistics were calculated to identify the mean distress level, caregiving burden and mean number of unmet needs. Unmet needs were defined as moderate (response 4) or high (response 5) needs. For data analysis of the EDIZ, the 5-point scale was recoded to a dichotomous score. Score 1–2 (no) were recoded in score 0, score 3–5 (yes) were recoded in score 1. Total scores 0–3 indicate little self-perceived burden, scores 4–6 moderate self-perceived burden and scores 7–9 many self-perceived burden [22]. For the descriptive statistics of the SCNS-P&C45-D we focused on the mean number of moderate/high unmet needs (response category 4 and 5), consistent with previous research [12]. Following the example of Girgis et al. (2011) the responses 1 and 2 were re-coded as 1 ‘No need’ and the other response categories were re-scored (2–4) to ensure a linear response format for the factory analysis [18]. As in the study of Girgis et al. (2011), flooring effects were identified with descriptive statistics, if more than 90% of the participants answered ‘no need’, that item was deleted from further analysis [18]. To identify underlying need domains of the SCNS-P&C45-D, exploratory factor analysis was used. To discover the suitability of the data, the correlation matrix was inspected for extreme high (r > .90) or no significant correlations [24]. The Kaiser–Meyer–Olkin (KMO) measure of sampling adequacy (>.6) and Bartlett test of Sphericity (p < 0.05) were calculated to test whether the sample size was appropriate [24]. Principal component analysis (PCA) was performed using oblique rotation, because of expected inter-correlations [24]. Factors were identified using Kaiser Criterion (based on Cattell’s scree plot and eigenvalues of 1.0 or greater) [18]. Items were included in a factor if: 1) the minimum loading was > 0.30; 2) the items scored highest on this certain factor; and 3) the factor contained at least three items [25]. Internal reliability of each subscale was calculated using Cronbach α, correlations between 0.70–0.95 were preferred [26]. Missing values were managed using listwise deletion in the factor analysis.

Construct validity was measured with convergent and discriminant validity [27]. Validity refers to whether an instrument actually measures what it is supposed to measure [28]. Validity was measured by calculating Pearson correlation coefficients for the SCNS-P&C45-D with distress (HADS) and burden from care (EDIZ). Many studies suggested that more needs were reported by participants with higher levels of distress or more burden from informal care [12, 13]. In concordance with these results, it was hypothesized that the number of total needs (moderate or high needs) was correlated with the total score on the HADS and with the total score on the EDIZ (Pearson’s r > .30: moderate; Pearson’s r > .60: large) [29]. It was expected that HADS and EDIZ scales that are conceptually related correlate moderately to highly with one another (r ≥ 0.40). Conversely, scales with a less conceptual relation are expected to show weak correlations (r < 0.40) [27]. We tested known-group validity, as a form of discriminant validity, in different groups of which higher needs were expected based on the literature. Although not consistent, studies have shown that those having more needs are often younger [12, 18]. Our hypothesis was that participants younger than sixty would report more often at least one unmet moderate or high need across domains. The cut-off point of 60 years of age was chosen based on the literature that describes significant relationships between age and unmet needs of informal caregivers [18]. Although unmet needs are not uncommon in (long-term) survivors and their informal caregivers [1, 15, 17, 18], a cohort study [30] and a longitudinal study [12] showed that unmet needs decrease in the course of time since diagnosis. Our hypothesis was that partners of patients who were diagnosed with BC less than one year ago, more often had at least one unmet moderate or high need than those diagnosed more than one year ago.

Participants

Three hundred two participants answered the survey (700 invited, response rate = 43%), 26 participants were excluded because of predetermined illness exclusion criteria (e.g. recurrence; metastases; > 5 year after diagnosis). Four surveys had more than 33.3% items of the SCNS-P&C45-D missing and were therefore excluded from further analysis, resulting in 272 participants eligible for analysis.

Factor analysis

All items on the SCNS-P&C45-D were significantly correlated, correlations ranged between r = .13 and r = .85. The KMO statistic (.94) was above the recommended value of .6, Bartlett’s test was significant (χ²(861) = 11,588.67, p < 0.001). Flooring effects were identified for item 18 (Accessing information about possible fertility problems of the patient with cancer) (96.7% no need), 19 (Looking after the patient with cancer) (93.4% no need) and 43 (Exploring your spiritual beliefs) (92.6% no need). These items were deleted from further factor analyses. The original factor structure and loading pattern of the SCNS-P&C45 was not replicated. The PCA with oblimin rotation showed five factors meeting Kaiser’s eigenvalue criterion (above 1), explaining 72% of the total variance (Additional file 1: Table S1).

One factor included one item (item 4: information concerning alternative therapies) which was therefore transferred to the factor with the second highest loading (factor 2). The four remaining factors explained 69.4% of the variance.

Factor 1 explained 55% of variance and includes 14 items regarding “Emotional and relational needs”.

The 16 items concerning “Health Care and Illness related needs” included in factor 2 accounted for 7.5% of the total variance. Due to cross loading and a better conceptual fitting, item 15 ‘look after own health’ and item 1 ‘caregiver information needs’ were replaced into factor 1.

Factor 3 included 4 items regarding “Practical needs” explaining 3.6% of variance.

Factor 4 relates to “Work and Social needs” which accounted for 3.3% of variance and included 7 items.

Reliability

Convergent validity

Discriminant validity

All domains and the total SCNS-P&C45-D showed a significantly greater proportion of younger participants (< 60 years) (mean age 52, median 53 and range 25–60 years) with at least one unmet moderate or high need in comparison to older participants (≥60 years) (mean age 69, median 68 and range 60–93 years) (SCNS-P&C45-D 25.5% and 20.3%, respectively p = 0.004).. Partners of patients diagnosed less than 1 year ago, reported slightly more often, but significantly one unmet moderate or high need in the Health Care and Illness related needs subscale in comparison with partners of patients diagnosed more than 1 year ago, significantly; 19.5% and 18%, respectively, p = 0.029, and the total SCNS-P&C45-D; respectively 23.2% and 22.4%, p = 0.018). In the domains of Emotion and Relational needs, Work and Social needs and Practical needs, no significant differences were found between participants whose partners were diagnosed less than 1 year ago and those of whose partners were diagnosed more than 1 year ago (Additional file 2: Table S2).

Clinical implications

The hypotheses concerning the convergent validity of the SCNS-P&C45-D were confirmed and partially confirmed concerning discriminant validity. The total SCNS-P&C45-D score was moderate/high positively correlated with the level of distress and burden from informal care. These relations have been found in other studies as well (for example in brain-, esophageal-, lung-, prostate-, testicular-, ovarian-, bowel-, and head and neck-cancer) [9, 36–38]. Experiencing any need or even just one can be quite distressing. All domains of the questionnaire had a moderately positive correlation with the distress and burden from informal care. All domains and the total SCNS-P&C45-D showed a significantly greater proportion of younger participants (< 60 years old) with at least one unmet moderate or high need in comparison to older participants. Other studies support this finding [1, 8, 12, 15]. Participants whose partner was diagnosed less than 1 year ago, often showed one or more unmet moderate or high need in the subscale of Health Care and Illness related needs, and overall at the SCNS-P&C45-D. Health care professionals should be aware of present but latent unmet needs of the patients and the informal caregiver, especially for younger patients and informal caregivers that might have more needs than older informal caregivers. During consultation not only the patient, but the informal caregiver should regularly be asked if the information was understandable, helpful and whether personal needs have been met in order to unravel persisting unmet needs.

Study limitations

This study has a number of strengths: the sample size, the distribution of age and the time since diagnosis. There is a good diversity between the type of hospitals and the regional spread of these hospitals in the Netherlands. In addition, we had a sufficient response rate. However, our study focused only on the intimate partner of the cancer patient and not on informal care givers in general. In addition, our sample was limited to (predominantly) male partners of breast cancer patients. Further study should validate the survey in partners of patients with other cancer types and stages. Future research on female patients with female informal caregivers should be performed to explore if the current factor structure is also valid for female partners of women suffering from BC. To expand on the psychometric properties of the SCNS-P&C45-D, further studies should investigate other indicators such as reproducibility of responsiveness measured by test-retest reliability and minimal clinical important difference [9, 27]. Since we used paper and pencil questionnaires we had some missing data. There is the also the possibility of selection bias as we do not know why the participating partners filled in the questionnaires and others did not. In addition, with the questionnaires sent by mail we had no control of whether the questionnaires were actually received by the partner of the patient. Lastly, in our study recall bias may have occurred since a considerable proportion of the population has been in follow-up longer than two years. Further research should investigate unmet needs of informal caregivers of patients that are in the early follow-up phase.