Validation of the Hospital Anxiety and Depression Scale in patients with decompensated cirrhosis

Abstract

INTRODUCTION Patients with decompensated cirrhosis (DC) experience substantial psychological distress. Accurate screening of psychological distress in this population is a critical first step toward reducing their morbidity. The Hospital Anxiety and Depression Scale (HADS) is a self-administered questionnaire designed to identify psychological distress.1 However, there is no research into the performance of HADS in DC. The overall goal of this study is to evaluate the psychometric properties of the HADS for patients with DC. METHODS Study design and population This is a secondary data analysis of a longitudinal cohort study of adult outpatients with DC recruited from Massachusetts General Hospital between August 2018 and September 2022, details of which have been reported elsewhere.2 All patient-reported outcomes were collected consistently across the study period (baseline, weeks 6, 12, 24, 26, and 48)—this study used data collected at baseline and week 6. All patients provided informed consent, and the Mass General Brigham Institutional Review Board approved this study. Patient-reported outcome measures HADS We assessed self-reported anxiety and depression using the 14-item HADS.1 The HADS has two 7-item subscales assessing anxiety (HADS-Anxiety) and depression (HADS-Depression) symptoms, with subscale scores ranging from 0 (no distress) to 21 (maximum distress). Other patient-reported outcome measures included depression (Patient Health Questionnaire-9, [PHQ-9]) and health-related quality of life (Short-Form Liver Disease Quality of Life [SF-LDQOL] questionnaire).3,4 Statistical Methods We calculated Cronbach’s alpha at baseline and week 6 for HADS-Anxiety and HADS-Depression separately to evaluate reliability. We evaluated floor and ceiling effects using the commonly accepted thresholds of 15% of patients achieving total scores between 0–1 (floor effect) or 20–21 (ceiling effect).5 A confirmatory factor analysis with 2 correlated latent variables (depression and anxiety) was performed to evaluate structural validity. We evaluated model fit using the comparative fit index (cutoff value >0.95), Tucker-Lewis index (cutoff value >0.95), root mean square error of approximation (cutoff value <0.06), and/or standardized root mean square residual (cutoff value <0.08).6 Convergent validity was evaluated by calculating the correlations of HADS subscale scores with PHQ-9 scores at both baseline and week 6. Predictive validity was assessed through a regression analysis examining whether the change in overall SF-LDQOL was predicted by the changes in HADS-Anxiety and HADS-Depression subscale scores from baseline to week 6, adjusting for confounders as previously described (age; Model for End-Stage Liver Disease-Sodium score; the presence of ascites, HE, or HCC; diagnosis of alcohol-associated cirrhosis; transplant listing status; and cirrhosis-specific comorbidity scoring system comorbidity [CirCom] score).2 Known-groups validity was assessed using ANOVA to evaluate if baseline HADS subscale scores differed based on Child-Pugh classes as they reflect the varying degrees of DC. Internal responsiveness of HADS-Depression was evaluated by calculating the mean differences and 95% CIs in total scores between baseline and week 6 for patients who experienced clinically meaningful changes on PHQ-9 and SF-LDQOL.2 For HADS-Anxiety, responsiveness was assessed only using SF-LDQOL as a criterion. The minimal clinically important difference is 1.5 points for HADS-Anxiety and HADS-Depression, 5 points for PHQ-9, and 8 points for SF-LDQOL.7–9 RESULTS Overview In the 218 patients at baseline, the median age was 60 years (IQR: 51–65). Half of the patients were actively listed for liver transplantation (n = 109, 50%). Median Child-Pugh score was 9 (IQR: 7–10), with the following distribution of classes: A (n = 19, 8.7%), B (n = 129, 59.2%), and C (n = 70, 32.1%). Mean scores for HADS-Depression and HADS-Anxiety at baseline were 6.7 (SD: 4.1) and 7.2 (SD: 4.3), respectively. In total, 145 patients were included in the week 6 analysis. Reliability, floor, and ceiling effects Both HADS-Depression and HADS-Anxiety demonstrated strong internal consistency (Cronbach alpha value >0.8) at baseline and week 6. Both HADS-Depression and HADS-Anxiety demonstrated minimum floor (8.1% and 9.4%, respectively) and ceiling effects (0% for both) at baseline. Structural, convergent, known-groups, and predictive validity The confirmatory factor analysis model estimated for HADS had a good model fit (comparative fit index: 0.95, Tucker-Lewis index: 0.94, root mean square error of approximation: 0.05, and standardized root mean square residual: 0.06). The correlation between HADS-Depression and HADS-Anxiety subscales was 0.71 (p < 0.001). At both baseline and week 6, strong correlations were observed between the PHQ-9 scores and HADS-Depression and HADS-Anxiety (r ≥ 0.70). There was a statistically and clinically significant difference in HADS-Depression scores among patients with varying levels of DC severity at baseline (Child-Pugh A: 3.8 vs. B: 6.8 and C: 7.5; p < 0.001). However, there was no statistically or clinically significant difference in HADS-Anxiety scores (Child-Pugh A: 6.6 vs. B: 7.0 and C: 7.7; p = 0.45). Among patients who completed the week 6 assessment, results of regression analysis revealed that the changes in HADS-Depression (β = −1.71 [95% CI: −2.27 to −1.15], p < 0.001) and HADS-Anxiety (β = −1.08 [95% CI: −1.60 to −0.55], p < 0.001) scores were significantly and negatively associated with the change in SF-LDQOL (Table 1). TABLE 1 - Regression analyses of change in SF-LDQOL predicted by the changes in HADS-Depression and HADS-Anxiety from baseline to week 6 (N=145)a Predictors β (95% CI) t Value p Age at baseline 0.00 (−0.16, 0.17) 0.01 0.994 MELD-Na score at baseline −0.19 (−0.49, 0.11) −1.28 0.204 Ascites 1.53 (−4.37, 7.43) 0.51 0.609 HE −3.81 (−7.76, 0.13) −1.91 0.058 Alcohol-associated cirrhosis 1.40 (−2.01, 4.81) 0.81 0.419 Listed for transplant at enrollment −1.94 (−5.50, 1.62) −1.08 0.284 HCC at enrollment 5.13 (−0.52, 10.77) 1.80 0.075 CirCom score −2.21 (−3.83, −0.59) −2.70 0.008 Change in HADS-Depression −1.71 (−2.27, −1.15) −6.06 <0.001 Age at baseline 0.01 (−0.17, 0.19) 0.11 0.915 MELD-Na score at baseline −0.18 (−0.49, 0.13) −1.14 0.256 Ascites 1.64 (−4.52, 7.80) 0.53 0.600 HE −4.04 (−8.23, 0.15) −1.91 0.059 Alcohol-associated cirrhosis 1.27 (−2.40, 4.94) 0.68 0.495 Listed for transplant at enrollment −2.71 (−6.50, 1.08) −1.42 0.159 HCC at enrollment 3.93 (−2.00, 9.87) 1.31 0.192 CirCom score −2.56 (−4.27, −0.85) −2.96 0.004 Change in HADS-Anxiety −1.08 (−1.60, −0.55) −4.04 <0.0001 aUnless specified. For the analysis of HADS-Depression, 134 patients were included due to the missingness in the HADS-Depression score (n = 3), MELD-Na score (n = 1), and SF-LDQOL score (n = 7). For the analysis of HADS-Anxiety, 135 patients were included due to the missingness in the HADS-Anxiety score (n = 2), MELD-Na score (n = 1), and SF-LDQOL score (n = 7).Abbreviations: CirCom score, cirrhosis-specific comorbidity scoring system comorbidity score; HADS, Hospital Anxiety and Depression Scale; MELD-Na, Model for End-Stage Liver Disease-Sodium; SF-LDQOL, Short-Form Liver Disease Quality of Life. Responsiveness The HADS-Depression score demonstrated significant changes corresponding to clinically meaningful changes in depression as measured by PHQ-9 from baseline to week 6, with a mean difference of −4.0 (95% CI: −5.4 to −2.6) among patients who showed a decrease of 5 points or more in the PHQ-9 and a mean difference of 2.4 (95% CI: 1.2 to 3.5) among those with an increase of 5 points or more in the PHQ-9. Using SF-LDQOL as a criterion, we found that both HADS-Depression (2.2 [95% CI: 1.0 to 3.4]) and HADS-Anxiety (1.8 [95% CI: 0.5 to 3.2]) scores increased significantly in patients with clinically meaningful decreases in SF-LDQOL. In patients with clinically meaningful increases in SF-LDQOL, we did not find significant decreases in their HADS-Depression and HADS-Anxiety scores (Supplemental Table, https://links.lww.com/HC9/B92). DISCUSSION Both the HADS-Depression and HADS-Anxiety subscales showed strong psychometric performance, confirming the use of HADS as a screening tool for psychological distress among patients with DC. Both HADS-Anxiety and HADS-Depression domains demonstrated comparable floor/ceiling effects, reliability, structural validity, and convergent validity, which aligns with empirical evidence.10 Limitations include not evaluating the test-retest reliability of the HADS subscales and not having a second measure of anxiety to evaluate the convergent validity of HADS-Anxiety. CONCLUSIONS The HADS-Depression and HADS-Anxiety subscales within HADS are reliable, valid, and responsive tools for assessing psychological distress among patients with DC.

Topics & Concepts