Mental health is reciprocally linked to sleep quality, specifically in patients with COVID-19 who are commonly susceptible to sleep disturbances (Deng et al., 2021; Marvaldi et al., 2021). Perceived stress, referring to the extent to which a person perceives their daily life situations as stressful, was also found to be positively associated with GHQ-12 in a large cross-national sample of COVID-19 patients (Bonsaksen et al., 2022). Patients infected with COVID-19 experienced the burden of job loss (Crayne, 2020), death anxiety (Korkut, 2022), and may react with heightened stress (Bonsaksen et al., 2022).

Coronavirus may severely impair the subsequent physical functioning in some patients, especially the elderly (Carfì et al., 2020; Halpin et al., 2020). The Activities of daily living (ADL) and the Instrumental Activities of Daily Living (IADL; Katz et al., 1970; Lawton, 2000) are basic skills necessary for independently taking care of oneself (Edemekong et al., 2017) and environmental adaptation (Roehrig et al., 2007). Although few in number, some studies have shown the link of mental health to ADL and IADL scores (de Castroe Costa et al., 2008).

Due to the negative effect of COVID-19 on several aspects of people’s lives (Santomauro et al., 2021), research should document the association of mental health status in COVID-19 patients with demographics and medical features. Previous studies have suggested the positive link of mental health status with medical features, such as change in sleep time before and after COVID-19 and use of sleeping pills (Becker et al., 2018), hospitalization duration (Liao et al., 2020), psychiatry disorders (Kaufman et al., 2020), hyperlipidemia (HLP; Chang et al., 2021), diabetes (Moradian et al., 2021), cardiovascular disease (CVD; de Paiva Teixeira et al., 2020), substance use history (Czeisler et al., 2020), as well as demographics characteristics, namely, educational level (Dalgard et al., 2007), unemployment (Achdut and Refaeli, 2020), and the number of family members (Hendriksen et al., 2021). COVID-19 pandemic was shown to result in greater mental distress in women (Giorgi et al., 2014; Bucciarelli et al., 2021). As an example, in a study on large data of 49,156 participants, Proto and Quintana-Domeque (2021) found that after the COVID-19 pandemic, women manifested higher elevation in GHQ-12 scores (higher psychological distress) than men. Aging is believed to be associated with decrease in mental distress (Hoeymans et al., 2004), while COVID-19 pandemic have led to greater mental distress in younger patients (Bruine de Bruin, 2021).

Iran is among the worst-hit countries by Coronavirus, with heavy death tolls (more than 19,000 deaths until August 2020; Shahriarirad et al., 2021). Challenging factors, namely, the shortage of hygiene and medical supplies and equipment (i.e., masks and disinfectants), economic constraints (Zandifar and Badrfam, 2020), and the incapacity of government to formulate and enforce effectual social distancing and lockdown measures have led to the mental distress in the Iranian general population (Moghanibashi-Mansourieh, 2020; Zandifar and Badrfam, 2020; Shahriarirad et al., 2021). The Iranian population has been no exception to the global trend of increased mental issues. In a group of 5,328 individuals from the general population of Iran, the prevalence rates of anxiety, depression, and comorbid depression-anxiety were determined to be 30.1, 33.4, and 22.1%, respectively (Nakhostin-Ansari et al., 2020). Moreover, in another recent study by Maroufizadeh et al. (2022), the prevalence of mild-to-severe anxiety and depression in Iranian medical students was found to be 38.1 and 27.6%, showing a significant impact on sleep patterns.

In Iran, two studies have assessed the psychometric properties of GHQ-12. In a study on emerging adults, Montazeri et al. (2003) findings confirmed the two-factor model, comprising “depression” and “social dysfunction.” Their study showed the negative association of GHQ-12 with global quality of life, supporting its satisfactory convergent validity. Similarly, the results of Najarkolaei et al. (2014) study supported a two-factor model including “distress” and “social dysfunction” in freshmen university students. Nevertheless, participants of these two studies had limited age range (18–26 years of age) and were recruited from university students, which prevent their results from being generalized to the general or clinical Iranian population.

The present study was first-of-its-kind that aimed to examine the psychometric properties of the GHQ-12 in Iranian patients with COVID-19. In specific, we aimed to examine: (1) the factor structure by conducting Confirmatory Factor Analysis (CFA) based on 13 empirically-derived factorial models, (2) the internal consistency, (3) the criterion validity through the relationship of GHQ with perceived stress, sleep quality, ADL/IADL, and demographic and medical variables, and (4) the comparison of the average GHQ-12 scores among age and gender groups (if any). We hypothesized that higher GHQ-12 score—that reflects lower mental health—has a positive relationship with: (1) poor sleep quality, (2) higher perceived stress, and (3) lower level of ADL/IADL functions.

Participants comprised a total of 328 patients with COVID-19 (55.8% men), aged 21 to 92 (Mean_age (SD) = 50.49 (14.96); 73.6% 60 years old or younger). As for educational level, 19.8% of participants were illiterate, 22.9% had primary education, 16.5% had secondary education, 23.5% had diploma level, and 17.3% had higher education. Their job status included 15.3% employee, 6.7% skill-worker, 20.7% self-employed, 39.6% unemployed, and 13.1% retired. Most patients (88.6%) were living with their spouse and/or their children, while 10.5% were living alone. A majority of 86.6% had no history of smoking, while 10.1 and 2.4% reported smoking in the past and at the present time. In addition, 93.9 and 96.6% reported no history of alcohol and drug use, respectively. Among participants, 4.6% reported using sleeping pills—mostly Alprazolam, Chlordiazepoxide, and Asentra. Patients under and over 60 reported underlying diseases, including Diabetes (13 and 44.6%), High Blood Pressure (HTN; 15.2 and 28.9%), HLP (11.7 and 31.3%), psychiatry disorders (4.8 and 3.6%), immune deficiency disease (IDD; 1.3 and 4.8%), and Cardiovascular disease (9.1 and 28.9%), each. Finally, the mean of hospitalization duration was 6.35 days (SD = 5.87) for patients under 60 and 8.8 days (SD = 5.9) for patients over 60.

The current cross-sectional study was carried out on patients hospitalized due to Coronavirus infection at the Baharloo and Ziaeian Hospitals from March to October 2020 in Tehran, Iran. After being discharged from the hospital, those who accepted to take part in the current study were asked to sign the consent form. The demographic information of those who consented to take part and their contact number was collected in a registration form. Then, three psychologists collected the data (demographic and medical variables, GHQ-12, PSS-10, PSQI, ADL and IADL), using telephone survey. Patients were informed about their optional participation in the current research and that they can leave the research any time they wish. This study received ethic approval from the Review Board of Tehran University of Medical Sciences (Ethical Code: IR.TUMS.VCR.REC.1399.156).

Data screening was performed via IBM SPSS Statistics (Version 28). CFA tests of the GHQ-12 were conducted using Mplus version 8.8. Evaluating the assumption of normality revealed a mostly positive but non-substantial skewness in all items; thus, transformation was not required (Gravetter et al., 2020). We applied CFA using the Weighted Least Square Mean and Variance Adjusted (WLSMV) estimator. Statistical strategies were as follows: First, we used the following statistical tests and indices (MacCallum et al., 1996; Hu and Bentler, 1999; Hooper and Coughlan, 2008) to assess the models’ “goodness-of-fit” (acceptable values in parenthesis): the Chi-square (χ²; desired p > 0.05), the Comparative Fit Index (CFI > 0.95), the Tucker–Lewis Index (TLI > 0.95), the Standardized Root Mean Square Residual (SRMR <0.08), the Normal Chi-square (χ²/df < 5), the Root Mean Square Error of Approximation (RMSEA <0.10), and its 90% Confidence Interval (Bentler and Bonett, 1980; MacCallum et al., 1996; Loehlin, 2004; Miles and Shevlin, 2007). The exact fit is defensible when the Chi-square is not significant, regardless of the SRMR value. Approximate fit is tenable when Chi-square is significant, SRMR ≤0.08, and standard residuals are all small (|r_res| < 0.1), and finally poor fit is concluded if Chi-square is significant, and SRMR >0.08 (Satorra and Bentler, 2010).

Second, for internal consistency—as recommended for ordinal Likert-type scales, the equivalents of Cronbach’s alpha coefficient (Ordinal Theta and Omega reliability coefficients) using R version 4.1.2 (Team, R. C, 2013; Revelle, 2017) were conducted, which instead of the Pearson correlation matrix, apply the poly-choric correlation matrix (Zumbo et al., 2007; Gadermann et al., 2012). A reliability coefficient of 0.70 or higher was considered an acceptable level (Cicchetti, 1994).

Third, the criterion validity was evaluated by the Spearman coefficient of rank correlation of GHQ-12 with PSQI, PSS-10, and ADL/IADL, since the data showed evidence of non-normality. Correlation coefficients were interpreted based on the effect size classification of Cohen (1988): 0.10 = small, 0.30 = medium, 0.50 = large, and 0.70 = very large.

Forth, Multivariate Analysis of Variance (MANOVA) and effect size (Hedge’s g) were used to compare the mean and standard deviation of the GHQ-12 scores across gender. According to a rule of thumb suggested by Cohen (1988), effect sizes were classified into small (<0.20), medium (0.21–0.50), large (0.51–0.80), and very large (>0.80).

To test the GHQ-12 factor structure, CFA was conducted and the goodness of fit for 13 models was examined (Table 2). Model 1 (M₁) examined a general factor, in which, the total of the 12 items were loaded on a single common factor of general mental health (Goldberg et al., 1997; Romppel et al., 2013; Gnambs and Staufenbiel, 2018; Hystad and Johnsen, 2020; Alaminos-Torres et al., 2021) to test the unidimensional model of assumed latent factor and included just random measurement error and indicator-specific variance (Gustafsson and Åberg-Bengtsson, 2010). If the general factor model fitted the data well, it meant that the assumption of the multidimensionality of the measurement tool was violated. Models two to seven (M₂ to M₇) consisted of a the first-order two-factor oblique models that suggested two subscales measuring two distinct dimensions (Politi et al., 1994; Schrnitz et al., 1999; Montazeri et al., 2003; Kalliath et al., 2004; Najarkolaei et al., 2014; Zhong et al., 2021; Hamad, 2022). Models 8–13 (M₈ to M₁₃) examined first-order three-factor oblique models, resembling the Exploratory Factor Analysis (EFA) according to the literature (Graetz, 1991; Farrell, 1998; Martin, 1999; Daradkeh et al., 2001; Gao et al., 2004; del Pilar Sánchez-López and Dresch, 2008; Liang et al., 2016; Lee and Kim, 2020). Model 8 (M₈) included general dysphoria, lack of enjoyment, and social dysfunction (Daradkeh et al., 2001). Model 9 (M₉) consisted of anxiety, depression, and social dysfunction (Farrell, 1998). Model 10 (M₁₀) was loaded by all three first-order factors, which included social dysfunction, anxiety and depression, and loss of confidence (Graetz, 1991; Gao et al., 2004; Lee and Kim, 2020). Model 11 (M₁₁) included cope, stress, and low self-esteem (Martin, 1999). Model 12 (M₁₂) comprised low level of social function, anxiety/depression, and poor self-confidence (Liang et al., 2016). Finally, Model 13 (M₁₃) included successful coping, self-esteem, and stress (del Pilar Sánchez-López and Dresch, 2008).

The Ordinal Theta and the Omega reliability coefficients for the subscales of GHQ-12 are presented in Table 3. The means of inter-item correlation were 0.10, 0.53, 0.48, and 0.58 for the total score, successful coping, self-esteem, and stress, respectively. Almost all items within the three subscales had a moderate positive relationship with each other (based on the corrected item-total correlation for subscale’s items), with values ranging from 0.34 to 0.83, 0.23 to 0.78., and 0.53 to 0.66 for successful coping, self-esteem, and stress, respectively.

Table 4 demonstrates that the inter-correlation between GHQ-12 total score and subscales ranged from 0.40 to 0.83 (p < 0.01). Criterion validity was estimated by the correlation of GHQ-12 total score and its subscales with PSQI and PSS (Table 4). PSQI had significant positive correlations with total GHQ-12 (r = 0.28, p < 0.01), successful coping (r = 0.24, p < 0.01), self-esteem (r = 0.21, p < 0.01), and stress (r = 0.20, p < 0.01). Also, PSS had significant positive correlations with the total score of GHQ-12 and its subscales (r = 0.31 to 0.58, p < 0.01). In the over 60 years of age group, negative correlations of the total GHQ-12 score were found with the ADL (r = −0.34, p < 0.01; r = −0.37, p < 0.01) and IADL scores (r = −0.42, p < 0.01, r = −0.46, p < 0.01) before and after the infection of COVID-19, respectively (Table 4).

As depicted in Table 5 for demographic and medical variables, the total score of GHQ-12 was significantly correlated with HLP (r = 0.16, p < 0.01), kidney failure (r = 0.12, p < 0.05), psychiatry disorders (r = 0.18, p < 0.01), the hospitalization duration (r = 0.15, p < 0.01), the change in sleep time (r = 0.31, p < 0.01), use of sleeping pills (r = 0.30, p < 0.01), educational level (r = −0.26, p < 0.01), and the number of family members (r = −0.12, p < 0.05). The correlation of the total score of GHQ-12 with Diabetes, HTN, MI, CVA, PD, obesity, IDD, CVD, and drug, alcohol, and cigarette history were non-significant (p > 0.01). Information for the correlation of GHQ-12 subscales is presented in Table 5.

Table 3 presents the mean and Standard Deviations (SD) of the GHQ-12 total score and subscales across gender and age groups. The female patients scored significantly higher than the males on total GHQ-12 scores [t (310) = −4.77, p < 0.001, Cohen’s d = −0.65, mean difference bootstrap 95% CI = −4.46 to-1.92]. The patients above 60 also scored slightly higher than patents under 60 on their total GHQ-12 scores [t (312) = −1.45, p = 0.08, Cohen’s d = −0.19, mean difference bootstrap 95% CI = −2.94 to 0.55]. A Multivariate Analysis of Variance (MANOVA) showed significant group differences by gender [F (3–308) = 10.80, p < 0.001, η² = 0.095] and age groups [F (3–310) = 7.27, p < 0.001, η² = 0.066] on mean scores of the subscales (see Table 3 for Mean scores). Subsequent tests of between-subjects’ effects showed that females scored significantly higher on successful coping [F (1–310) = 5.68, p < 0.05, η² = 0.018], self-esteem [F (1–310) = 28.11, p < 0.001, η² = 0.083], and stress [F (1–310) = 27.38, p < 0.081, η² = 0.081] than males. The patients over 60 scored significantly higher on successful coping [F (1–312) = 9.41, p < 0.01, ns, η² = 0.029], than patients under 60. However, non-significant differences in mean score were found for self-esteem [F (1–312) = 0.31, p = 0.58, ns, η² = 0.001] = and stress [F (1–312) = 1.60, p = 0.21, ns, η² = 0.005] across age.

The current study results provide insight into the general mental health status in patients with COVID-19. However, this study is not without limitations. First, GHQ-12 is a screening tool and was not designed for diagnosis objectives and distinguishing among mental disorders (Goldberg, 1986; Schrnitz et al., 1999). Researchers in future work can use semi-structured interviews to provide more in-depth information regarding high scored items of GHQ-12 and compare yielded scores of GHQ-12 with the additional probe questions (i.e., regarding symptom severity and duration) to evaluate the accuracy of GHQ-12. Second, the cross-sectional design of the current study has prevented causal inferences. It restricts our knowledge on the direction of the correlation of GHQ-12 with perceived stress and sleep quality. It also prevents us from measuring the stability of mental health scores over time. Longitudinal studies are recommended to explore the predictive role of the abovementioned variables on each other, and evaluate the stability of GHQ-12 scores over time. Finally, the present study did not perform the measurement invariance analyses across gender or age, because the sample sizes would start to become small when we divided the sample into age and gender subgroups. In the age case, sample size would be down to 87 for the younger segment. Such small sample sizes lacked sufficient power to detect any invariance.

Given the positive link of poor mental health with perceived stress, sleep disturbances, and impaired independent daily activities among Iranian COVID-19 patients, an important clinical implication for physicians, psychologists, and psychiatrists is to design psychological interventional courses for COVID-19 hospitalized patients that specifically target these problems. Access to such services via social media may be beneficial not only for their mental health, but also for their ability to improve their physical and mental functioning and independency (Shojaei and Masoumi, 2020).