Symptoms of Anxiety
Participants completed the Generalized Anxiety Disorder 7-item (GAD-7) questionnaire for screening and measuring severity of generalized anxiety disorder [39]. The GAD-7 asks about such symptoms as feeling nervous, anxious, or on edge over the past two weeks. Scores from all 7 items are added to obtain a total score [40]. Cut-points of 5, 10, and 15 represent mild, moderate, and severe levels of anxiety symptoms. Respondents indicate how difficult the symptoms make it for them to engage in daily activities from 0 (Not difficult at all) to 3 (Extremely difficult). The internal consistency of the GAD-7 is excellent (α = 0.92). Test-retest reliability is also good (Pearson r = 0.83) as is procedural validity (intraclass correlation = 0.83). The GAD-7 has excellent sensitivity and specificity to identify generalized anxiety disorder [41].
Health-Related Quality of Life
Participants completed the SF-36, which asks about the patient’s health status in the past week [14]. Eight subscales include physical functioning, role-physical, bodily pain, general health, vitality, social functioning, role-emotional and mental health. Two summary measures, the Physical Component Summary and the Mental Component Summary, are calculated from 4 scales each. Use of the summary measures has been shown to increase the reliability of scores and improve the validity of scores in discriminating between physical and psychosocial outcomes [14]. Higher scores represent better HRQL, with a mean score of 50 (SD = 50) for the general population. Internal consistency estimates for the component summary scores are α > 0.89, item discriminant validity estimates are greater than 92.5% and 2-week test-retest reliability was excellent. Scores on the SF-36 have been divided into categories of HRQL functioning [42,43]. Participants in the impaired to very impaired category have scores ≤ mean – 1 SD while participants with average to above average functioning have scores > mean – 1 SD.
The SF-36 has been used extensively in observational and randomized studies for a range of illness conditions. In SCD, some aspects of HRQL as measured by the SF-36 improved for adult patients who responded to hydroxyurea [44]. Participants in the Pain in Sickle Cell Epidemiology Study scored lower than national norms on all SF-36 subscales except psychosocial functioning [45]. HRQL decreased significantly as daily pain intensity increased [45]. Further, women reported worse bodily pain compared with men [46].
Data Analyses
All biostatistical analyses were conducted using Stata 13 [47]. Continuous variables were examined for normality with measures of skewness and peakedness. All variables satisfied the assumptions of normality with the exception of barriers to health care and ED utilization. The variable barriers to health care was transformed using a square root transformation, resulting in a more normally distributed variable. ED utilization was dichotomized as 0–2 versus 3 or more ED visits in the previous 12 months, based on the distribution of utilization in the sample. The cutpoint of ≥ 3 annual ED visits is consistent with other literature on SCD clinical severity [48].
Descriptive statistics were computed to include means, standard deviations and frequencies. Sociodemographic variables (age, sex, insurance status [public or private] and income) were examined as potential covariates using Pearson correlations and t tests. Associations among emotional distress (anxiety and depression symptoms), clinical complications and ED utilization, barriers to health care, and the outcomes of the Physical and Mental Component Summary scores from the SF-36 were examined using Pearson correlations. We conducted stepwise regression with forward selection to determine models predictive of physical and mental HRQL. We tested the addition of each chosen variable (anxiety symptoms, depression symptoms, clinical complications, ED utilization, barriers to health care, age, sex, insurance status, and income), adding the variables (if any) that were most correlated with the outcome, and repeated the process until the model was not improved. A significance level of 0.05 was used for all statistical tests.