Reliability, Validity and Responsiveness of the Syncope Functional Status Questionnaire

This study is part of the Fainting Assessment Study (FAST); FAST was designed to asses the diagnostic yield and accuracy of attending physicians in patients presenting with TLOC (Van Dijk et al., submitted for publication).6,7 The Academic Medical Center (AMC) is a first-line referral center for patients from the surrounding area and a speciality center for patients with TLOC from the Dutch population. Adult patients presenting to the AMC, between January 2000 and July 2002, with at least one episode of TLOC in the 12 months before presentation were eligible for the study. TLOC was defined as self-limiting episodes of loss of consciousness, not lasting longer than 1 hour. Patients incapable of reading Dutch and patients with a physical/mental impairment that prevented their completing the questionnaires were excluded. All patients gave informed consent. This study was approved by the Medical Ethical Committee of our institution.

All HR-QoL questionnaires were entered into an SPSS database by two different people. Differences were checked and corrected according to data on the original questionnaire. All analyses were performed using SPSS version 11.5.

From January 2000 till July 2002, 503 patients were enrolled in FAST. Thirty-five patients were excluded from this HR-QoL study because of language problems (n = 14) or being physically or mentally unable to complete the questionnaires (n = 21). Of the remaining 468 patients, 385 (82%) returned their questionnaires. At 1-year follow-up, 268 (72%) of the original respondents filled in the questionnaires again. Twelve patients were excluded because of death during the follow-up period (n = 10), severe dementia (n = 1), and detention (n = 1). Fifty-two patients completed the SFSQ two times for the test–retest analysis with a mean interim period of 8 days (SD 6.9). The sociodemographic and clinical characteristics of the respondents and nonrespondents at baseline and 1-year follow-up are displayed in Table 1. Nonrespondents were significantly younger, more often female, and had experienced fewer episodes of TLOC than respondents.

Table 2 lists the means, SDs, medians, and ranges for the summary scores of the SFSQ. For all scales, the full range of possible scores was seen. Score distributions were asymmetrical, with large peaks at 0 (no impairment) for all scales. The number of missing values ranged from 2.1% (relationship with family/ friends) to 5.2% (use of public transportation). Internal consistency of the impairment score was 0.88, and it was 0.92 of the fear–worry score.

Correlations between the scale scores of the SFSQ and the SF-36 are low to moderate, ranging from 0.30 to 0.62 (Table 3). The impairment score shows moderate correlation with the PF, RP, and SF scores (0.52, 0.54, and 0.58). The correlation between the fear–worry score and MH is moderate (0.48) but higher than the correlation with any other scale. The SDS shows moderate correlation with GH (0.50) and shows similar correlations with other scales scores of the SF-36.

The SDS did not discriminate between patients differing in age and gender. Patients with more than one episode in the year before presentation showed a significantly poorer HR-QoL on all SFSQ scales than patients with one episode before presentation. The impairment score was significantly poorer in patients with comorbid conditions than in patients without comorbid conditions (Table 4). These differences were comparable to the minimally important difference for these scales7 and, therefore, can also be considered clinically relevant. The scores on the physical component summary of the SF-36 were different for subgroups of patients differing in gender, number of episodes, and Charlson score. The mental component summary (MCS) scores were not different between any of the subgroups.

The changes in the scale scores of the SFSQ between baseline and 1-year follow-up, per category of perceived change in health status, are displayed in Figure 1. Changes are linearly related to the change in health status and significantly different from each other (ANOVA; p = <0.01). The improvement in impairment score (mean difference 11.8; p < 0.01) and in SDS (mean difference 8.2; p = 0.02) was larger in patients without recurrences during follow-up than in patients with recurrences. The changes were not different between patients with or without a diagnosis or with different diagnoses.

In this study we examined the reliability, validity, and responsiveness of a disease-specific functional status questionnaire for patients with syncope. The internal consistency of the impairment score and fear–worry score items was high with alphas of 0.88 and 0.92, respectively. Test–retest reliability was moderate to good for the 11 yes/no impairment questions, moderate for the Likert-scale fear–worry questions, and high for the summary scale (SDS). Validity was moderate when comparing the summary scales of the SDS with comparable scale scores of the SF-36. The SFSQ showed adequate discriminative properties between some of the clinically defined subgroups of patients but not on sociodemographically defined patient subgroups. This finding was in concordance with the results of the SF-36. Responsiveness to perceived change in health status and the presence of recurrence during follow-up was good.

Both scale scores and the SDS show a large floor effect. This indicates that the SFSQ is not very sensitive to detecting differences between patients with a low impact of the episodes on their lives, i.e., good HR-QoL scores. Thus, when studying a relatively healthy population, the SFSQ might not be adequately sensitive. This could in part be the consequence of the dichotomous choices (yes/no) for the impairment questions. To make the instrument more sensitive to small changes in relatively healthy subjects, one could replace the yes/no response scale with a five-point Likert scale. This would likely lead to the instrument being more sensitive to changes in clinical conditions in all patients. Furthermore, in the existing scoring system, all areas of impairment are considered equally important; these questions may benefit from a weighting system. In future use, we also recommend not combining the impairment score and fear–worry score to obtain the SDS, as the separate scale scores may provide more valid information.

The number of missing items is low, indicating a high acceptability of the questionnaire. Although we expected the question on sexual functioning would be considered inappropriate by some, this question did show equal acceptability.

The impairment score is initially based on 11 questions. In cases where patients have “not applicable” responses (e.g., when the patient does not drive, does not work, has no partner, is isolated socially, and is not sexually active), the utility of the instrument may be diminished. For future iterations of the questionnaire, the term “partner” could be added to the question “does syncope interfere with your relation with spouse/boyfriend/girlfriend” to make the question applicable to more patients.

The internal consistency of the measure is comparable to that found in the earlier study by Linzer et al.9 The test–retest method shows that the within-subject variability is quite small when comparing the results to the mean changes seen in patients with different changes in health status (Fig. 1). This should enable researchers to use the SFSQ in clinical trials with relatively small sample sizes,22 keeping in mind that the test–retest reliability is related to the variance of the study population in which it was tested. Our population was quite large and heterogeneous; therefore, these results might differ in a more homogeneous population.23 Furthermore, the respondents in our population differed from the nonrespondents, which also might affect the generalizability of our results.

The correlation was moderate between the scores of the SFSQ and the preselected scales of the SF-36. Although these correlations are among the highest correlations, they are not markedly different from the correlations with the other scale scores of the SF-36. The relatively high correlation of the impairment score with the RE scale can be explained by the fact that the questions that comprise the RE scale are work-related. The fear–worry score shows moderate correlation with the MH scale of the SF-36. A possible explanation for this moderate correlation can be that the items of the SF-36 focus on MH in general, whereas the SFSQ focuses on fear and worry on account of episodes of TLOC. Individuals are likely to have another source of emotional distress in addition to their syncope episodes.

The SFSQ has been shown to differentiate between patients with one versus those with more than one episode and between patients with or without comorbid disorders; it does not differentiate between age groups and gender. This could indicate that the questionnaire is not very discriminative or that the difference between the selected subgroups was not large enough for any differences to be detected. The second explanation seems plausible because the SF-36, particularly the MCS, does not discriminate between the selected groups either. Another view could be that the lack of difference is actually a strength of the instrument because influences from age and gender may arise from other than syncope-related causes.

The results of the changes in scores related to changes in health status (Fig. 1) and the differentiation between patients with and without recurrences during follow-up indicate that the responsiveness of the SFSQ is appropriate, especially because recurrences have been shown to be one of the main influences on the HR-QoL of patients with TLOC.7,20,21,24 Although the SFSQ was not responsive to changes in HR-QoL in patients with different diagnoses, earlier studies have shown that the difference in the HR-QoL of patients with different diagnoses is modest.7 It therefore seems that other factors are more important contributors to QoL. One possible explanation could be that patients interpret the questions to be related to the moment of an episode instead of their daily life. To prevent this issue, a clearer time frame could be added to the questionnaire, for instance, focusing patients on the last 4 weeks.

The SFSQ can provide important information on the effect of interventions on HR-QoL of patients, especially in moderately or severely affected patients. With minimal amendments to the instrument, the SFSQ could also be reliable in less affected groups of patients. We recommend that this instrument be used primarily in group-comparisons in trials comparing diagnostic tests or treatment effects. As well, it could be a viable option in cohort-studies, where patients are compared over different time points. Furthermore, we feel that individual patient SFSQ scores could be informative for the physician, giving the physician an impression of the patient’s fear and worry and general experience as a result of the episodes.