Study design and data source
This study was conducted using data from the 2017 US National Health and Wellness Survey (NHWS; N = 75,004). The NHWS is a self-administered, internet-based survey of a sample of adults (aged ≥ 18 years) that provides “real world” patient-level information over 165 therapeutic conditions. Potential respondents for the survey are recruited through a general-purpose web-based consumer panel. The panel recruits its members via opt-in e-mails, co-registration with panel partners, e-newsletter campaigns, banner placements, and affiliate networks. All the respondents who explicitly agreed to be a panel member registered through a unique e-mail address and completed an in-depth demographic registration profile. A quota sampling procedure (using data from the Current Population Survey of the US Census) was used to ensure that the final NHWS sample was representative of the US’ adult population in 2017 with respect to age, gender, and race/ethnicity. Informed consent was obtained from all the respondents and all parties ensured protection of patients’ personal data. The study protocol and questionnaire were reviewed by the Pearl Institutional Review Board and granted exemption status.
Study sample
Respondents aged 18-64 years ‘with depression diagnosis’ (n = 8853) or ‘without depression diagnosis’ (n = 30,478) were included in the analysis. Respondents with depression diagnosis: those who self-reported physician diagnosis of depression and reported experiencing depression in the past 12 months) [2]. These respondents were further stratified by severity of depression as determined by Patient Health Questionnaire-9 (PHQ-9) scores: none/minimal (score = 0–4; n = 1876), mild (score = 5–9; n = 2801), moderate (score = 10–14; n = 1938), moderately severe (score = 15–19; n = 1376), or severe (score = 20–27; n = 862). Respondents without depression diagnosis: those who had no self-reported physician diagnosis of depression, reported not experiencing depression in the past 12 months, and had PHQ-9 scores ≤ 4 [17] (Fig. 1). Respondents diagnosed with bipolar disorder and those who reported not experiencing depression in the past 12 months but had a diagnosis were excluded from the study.
Measures
Demographics and health characteristics
Demographic variables including age, gender, employment status, race/ethnicity, marital status, education, household income, insurance status, body mass index (BMI), smoking status, alcohol use, exercise behavior, and Charlson Comorbidity Index (CCI) were collected. The CCI represents a weighted sum of multiple comorbid conditions predictive of mortality with greater scores indicating greater comorbid burden on the patient [18]. Disease-specific diagnoses including depression, anxiety, and sleep difficulties were also analyzed.
Depression symptoms, anxiety and sleep problems
Depression symptoms assessed that prompted respondents to see their doctor included self-reported depressed mood and other emotional problems, changes in eating and sleep patterns, mental changes (e.g., forgetfulness, difficulty thinking, difficulty concentrating), and social and physical problems. Sleep problems including self-reported difficulty falling asleep, difficulty staying awake, daytime sleepiness, leg cramps/leg problems, night sweats/hot flashes, and poor quality of sleep were evaluated. Anxiety was assessed according to the self-reported diagnoses of anxiety disorders and self-reported experiences of anxiety. Additionally, anxiety was measured by the Generalized Anxiety Disorder-7 (GAD-7) scale (Supplementary Table 1) [19].
Health-related quality of life (HRQoL) and health utilities
Short Form Survey Instrument version 2 (SF-36v2)
HRQoL was assessed using the SF-36v2 [20], which is a multipurpose, generic health status instrument comprised of 36 questions. The instrument is designed to report eight health domains (Physical Functioning, Role-Physical, Bodily Pain, General Health, Vitality, Social Functioning, Role-Emotional, and Mental Health) and two summary scores (Physical Component Summary [PCS] and Mental Component Summary [MCS]). Each domain and PCS and MCS scores are normed to a mean of 50 and a standard deviation of 10 for the US’ population. Higher scores are indicative of better health status [20]. SF-36v2 related parameters were studied based on past 4 weeks health status. Additionally, health state utility index was calculated using the Short-Form 6 Dimensions (SF-6D) form. The SF-6D is a preference-based single index measure for health using general population values and provides scores on a theoretical 0–1 scale with higher scores indicating better health status [21].
EuroQol 5-Dimension Health Questionnaire
The EuroQol 5-Dimension Health Questionnaire (EQ-5D-5L) [21] consists of a descriptive system (EQ-5D) and a visual analogue scale (EQ VAS). The descriptive system is composed of five dimensions: mobility, self-care, usual activities, pain/discomfort, and anxiety/depression. The EQ VAS (score: 0 to 100) indicates the respondent’s self-rated health, with the endpoints being 'Best imaginable health state' (score = 100) and 'Worst imaginable health state' (score = 0). Lower overall scores on the EQ-5D-5L health utilities are indicative of higher disability. The most recent version with 5-point rating scales for each dimension was used in this study [22]. The EQ-5D-5L utility scores were calculated by mapping the five-level descriptive system (EQ-5D-5L) onto the three-level value set (EQ-5D-3L) using the mapping (“crosswalk”) approach developed by van Hout et al. [23]. Health states were mapped using country-specific value set.
Work productivity and activity impairment (WPAI)
Work productivity loss was measured using the WPAI questionnaire [24], a six-item validated instrument which consists of four metrics: absenteeism (the percentage of work time missed because of one's health in the past seven days), presenteeism-related impairment (the percentage of impairment experienced while at work in the past seven days because of one's health), overall work productivity loss (an overall impairment estimate that it is a combination of absenteeism and presenteeism), and activity impairment (the percentage of impairment in daily activities because of one's health in the past seven days). Only respondents who reported being employed full-time or part-time provided data for absenteeism, presenteeism, and overall work impairment; all respondents reported data for activity impairment.
Health-resource utilization (HRU)
Healthcare utilization was defined by the number of healthcare provider (HCP) visits (e.g., general practitioner, internist, cardiologist, gynecologist, etc.), the number of emergency room (ER) visits, and the number of times hospitalized in the past six months. All outcome measures and scales [19,20,21, 24, 25] used in this study are detailed in Supplementary Table 1.
Statistical analyses
Chi-square and analysis of variance (ANOVA) tests were used to determine the significant differences for categorical variables and continuous variables, respectively. These results served to characterize differences between respondents with and without a depression diagnosis as well as between no/minimal, mild, moderate, moderately severe, and severe diagnosed depression and informed the selection of covariates for multivariable models.
Generalized linear models (GLMs) were used to control for demographic, health characteristic and comorbidity variables to compare HRQoL, WPAI, and HRU between respondents with and without a depression diagnosis and across symptom severity among respondents with a depression diagnosis. Only variables that were statistically significant in the bivariate analysis and had clinical importance were included in the regression models. GLMs with a negative binomial distribution were used for skewed data (e.g., WPAI and HRU).
The covariates included in the multivariable models were: Age (continuous), gender (male vs. female), ethnicity (black, hispanic, other vs. white [reference]), marital status (single, decline to answer vs. married/living with partner [reference]), education (less than college, decline to answer vs. college education [reference]), income (< 50 k, 50-75 k, decline to answer vs. 75 k + [reference]), employment (employed vs. not), insured (yes vs. no), BMI (underweight, overweight, obese [combined obese and morbidly obese], decline to answer vs. normal weight [reference]), smoking (former, current vs. never smoked), CCI, and individual comorbidities. The individual comorbidities included: Diagnosed with anxiety, nasal allergies/hay fever, allergies, pain, hypertension, high cholesterol, migraine, generalized anxiety disorder, heartburn, social anxiety disorder, gastroesophageal reflux disease, asthma, arthritis, panic disorder, acne, post-traumatic stress disorder, urinary tract infection, irritable bowel syndrome, dry eye, sleep apnea, eczema, thyroid problem, diabetes, yeast infection, and bladder control condition.
In bivariate analyses, comparisons were made for: (a) ‘with depression diagnosis’ versus ‘without depression diagnosis’ groups, and (b) across all severity groups using an overall omnibus test. In multivariable analyses, comparisons were made for: (a) ‘with depression diagnosis’ group versus ‘without depression diagnosis’ group (reference), and (b) mild, moderate, moderately severe, or severe groups versus ‘no/minimal’ symptoms severity group (reference). P-value less than 0.05 was considered statistically significant.