We conducted a cross-sectional study from 14th February to May 31, 2020. The study was conducted at 15 hospitals and three health centers across Vietnam, including 10 hospitals and one health center in the northern region (Military Hospital 103, E Hospital, General Hospital of Agricultural, Thai Nguyen National Hospital, Bac Ninh Obstetrics and Pediatrics Hospital, Hai Phong University of Medicine and Pharmacy Hospital, Kien An Hospital, Quang Ninh General Hospital, Bai Chay Hospital, Quang Ninh Obstetrics and Pediatrics Hospital, Kien Thuy District Health Center), one hospital and one health center in the central region (Da Nang Oncology Hospital, Trieu Phong District Health Center), four hospitals and one health center in the southern region (Thu Duc District Hospital, Hospital District 2, Tan Phu District Hospital, Can Tho University of Medicine and Pharmacy Hospital, Thu Duc District Health Center).

The study was reviewed and approved by the Institutional Ethical Review Committee of Hanoi University of Public Health, Vietnam (IRB No. 029/2020/YTCC-HD3, and IRB No. 133/2020/YTCC-HD3).

Consecutively convenience sampling was used to recruit the participants who visited the outpatient departments (OPD) at each hospital and health center. Participants recruited were those aged 18–85 years, be able to communicate using Vietnamese language, without any emergency medical conditions (e.g., stroke, respiratory failure, serious injuries). The total sample of 8,291 participants was analyzed in this study. The number of studied participants in each hospital and health center by geographic location was described in detail in a previous study [23, 24].

The underlying health problems were assessed using the Charlson Comorbidity Index items [28]. Participants were asked whether they had one of the following medical conditions, including myocardial infarction, congestive heart failure, peripheral disease, cerebrovascular disease, chronic pulmonary disease, diabetes without complications, taking anticoagulants (for cardiovascular disease, thrombosis), hemiplegia, diabetes with complications, moderate or severe renal disease, tumor without metastasis, moderate or severe liver disease, solid metastatic tumor, HIV/AIDS. We categorized underlying health conditions (UHC) into two groups: with UHC vs. without UHC. Patients were classified as having UHC if they had at least one of these above diseases.

The short form of the health literacy questionnaire (HLS-SF12) was used to measure health literacy (HL). This questionnaire has been validated and used in Asia [29], including in Vietnam [30] with satisfactory reproductivity and validity. The HLS-SF12 consists of 12 items assessing participant’s perceived difficulty on 4-point Likert responses from 1 (very difficult) to 4 (very easy). We transformed the HL score into a standardized index ranging from 0 to 50, with higher scores indicating better health literacy, using the formula: Index = ( mean − 1 ) × ( 50/3 ) (1)

In which, Index is the standardized HL index score, mean is the average score of 12 questionnaire items for each participant, one is the lowest value of the mean, three is the range of the mean, and 50 is the highest value of the index. In the current study, the Cronbach’s α score of the HLS-SF12 was 0.94.

Patients provided their information about current lifestyle behaviors in comparison to before the epidemic. Smoking, drinking, and physical activity were assessed on five responses: never, stopped, less, unchanged, and more. Eating behavior also was reported on three responses scale from less healthy, unchanged, and healthier. Based on the WHO’s guideline [9], people were recommended to keep their positive lifestyles (e.g., physical exercise, balanced diet) unchanged or better to stay healthy during the COVID-19 crisis. Thus, we classified the “unchanged or healthier” eating and “unchanged or more” physical activity as healthy behavior changes and “unchanged or more” smoking or drinking as unhealthy behavior changes. Patient’s responses were classified into two groups: “never/stopped/less” vs. “unchanged or more” for smoking, drinking, and physical activity, “less healthy” vs. “unchanged or healthier” for the eating behavior.

Preventive behaviors were evaluated through three items based on the COVID-19 prevention guidelines of the world health organization [9], including 1) washing hands with soap or alcohol-based liquid, 2) wearing masks when getting outside, 3) adherence to at least 2 m physical distance rule. Participants reported how often they followed those preventive behaviors during the pandemic on a 5-point Likert scale ranging from 0 (Never) to 4 (Always). The total score ranges from 0 to 12, with a higher score indicates higher compliance with preventive behaviors. We categorized PB into two categories based on the median value of 9 [31], including low PB score (< 9) and high PB score (≥ 9). The Cronbach’s Alpha score of this questionnaire in this study was 0.81.

Depression was assessed using the Patient health questionnaire (PHQ-9), which was widely used in Vietnam [24]. The PHQ-9 has nine items regarding the frequency of given symptoms that participants had encountered over the past two weeks, and the score of each item is given on a 4-point scale from 0 (not at all) to 3 (almost every day). The total PHQ score ranges from 0 to 27, and scores ≥10 were classified as having depression symptoms [32].

Anxiety was assessed using the Generalized Anxiety Disorder (GAD-7) [33], which was validated and used in Vietnam [34]. The GAD-7 comprises seven items about how often they have been encountered seven symptoms over the last two weeks on the 4-point Likert scale ranging from 0 (not at all) to 3 (nearly every day). The total GAD score ranges from 0 to 21, and scores ≥8 were classified as having anxiety disorders [35]. The Cronbach’s Alpha index for the PHQ-9 and the GAD-17 in this study were 0.90 and 0.94, respectively.

Research assistants (physicians, nurses, and medical students) received a 4 h training session for data collection which conducted by researchers in each medical facility. The infection control training (e.g., mask-wearing, handwashing, distancing) were also provided using guidelines of the Centers for disease Control and Prevention (CDC) [36], and World Health Organization [37].

Research assistants (e.g., physicians, nurses, medical students) invited and asked for voluntary participation of OPD visitors. All eligible participants have signed the informed consent form before conducting the survey. We collected data through a self-administered questionnaire. In the initial period, only printed self-administered questionnaires were used for data collection. However, at the peak of the pandemic, patients can choose to complete the survey by filling in the printed questionnaire or scanning the printed QR codes and completed the survey using their smartphone while waiting for the doctor’s visit. During the questionnaire administration, research assistants provided guidance and support if it’s necessary. Each survey took about 20–30 min to complete. Finally, the data was cleaned and analyzed by researchers confidentially.

The sample of 4,348 participants was used for analysis of preventive behaviors and anxiety as data of these variables were collected in the peak stage of the pandemic. A sample of 8,291 participants was used for analysis of all other variables which investigated in two stages of the pandemic. First, the Chi-square test was performed to examine the distribution of depression and anxiety among different studied variables. Secondly, the associations of underlying health conditions, S-COVID-19-S, health-related behavior changes and PB with depression and anxiety were tested separately in different models using the simple and adjusted logistic regression models. To reduce the impact of residual confounders, age, gender, and factors associated with outcomes at p < 0.20 in the bivariate models were chosen to adjust in multivariate models. To eliminate multicollinearity, the Spearman correlation was utilized to examine correlations between potential confounders. If the correlation between two confounders at rho ≥0.30, one representative confounder was adjusted in final models. The results were presented in Supplementary Tables S1–S3 in the Supplementary File. Finally, the interaction analyses were performed (using the interaction terms) to examine the effect modification of S-COVID-19-S, health-related behavior changes, and preventive behaviors on the associations of UHC with anxiety and depression. Model 1 (unadjusted model) was performed with three terms, including X1, X2, and X1*X2 (where X1 is the main effect of UHC, and X2 is the main effect of modifying factors. Model 2 (adjusted model) was conducted with three terms in model 1 and adjusted for potential confounders. The interaction models were presented with the odds ratio and the 95% confidence interval of three terms (X1, X2, and the interaction of X1*X2). The p-value < 0.05 was considered significance. We analyzed data using the IBM SPSS Version 20.0 (IBM Corp, Armonk, NY, United States).

The mean age, HL of the population were 43.6 ± 16.9, 28.1 ± 9.4, respectively. The prevalence of underlying health conditions (UHC), depression, and anxiety disorders were 22.3% (1867/8,291), 12.5% (1,033/8,291), and 16.7% (726/4,348), respectively. Of 8,291 participants, 41% were men, 79.9% ever married, 45.5% had college/university or higher degree, 92.5% had a job, 45.9 reported felt very or fairly easy to pay for treatments, 28.7% participated in surveying during the lockdown period, 11.7% were overweight or obese, and 37.7% had S-COVID-19-S. In the pandemic, the proportions of patients who responded “unchanged/healthier” eating behavior, “unchanged/more” smoking, drinking, and physical activity were 94.9, 9.0, 14.9, and 66.1%, respectively (Table 1).

Table 2 presents the association of underlying health conditions with depression and anxiety symptoms. After adjusted for confounders, multiple logistic regression models showed that participants with UHC had higher odds of experiencing depression (OR, 2.11; 95% CI, 1.81–2.46; p < 0.001) and anxiety symptoms (OR, 2.86; 95% CI, 2.37–3.46; p < 0.001), as compared to those without comorbidity (Table 2). The odds of having depression and anxiety symptoms were significantly lower in patients who had a high PB score (OR, 0.36; 95% CI, 0.30–0.44; p < 0.001) and (OR, 0.28; 95% CI, 0.23–0.34; p < 0.001), who had physical activity at “unchanged or more” level (OR, 0.53; 95% CI, 0.45–0.61; p < 0.001) and (OR, 0.48; 95% CI, 0.39–0.59; p < 0.001), who ate “unchanged or healthier” (OR, 0.22; 95% CI, 0.17–0.28; p < 0.001) and (OR, 0.20; 95% CI, 0.15–0.26; p < 0.001), respectively, as compared to counterparts. The odds of depression and anxiety symptoms were significantly higher in patients with S-COVID-19-S (OR, 2.84; 95% CI, 2.45–3.29); p < 0.001) and (OR, 2.92; 95% CI, 2.42–3.52; p < 0.001), respectively. Besides, the odds of suffering from anxiety symptoms were significantly larger in patients who drunk at “unchanged or more” level (OR, 1.99; 95% CI, 1.45–2.72; p < 0.001).

In the interaction analysis models, we found that as compared to participants without UHC and without S-COVID-19-S, those without UHC and with S-COVID-19-S had 1.92 times higher likelihood of having depressive symptoms (OR, 1.92; 95% CI, 1.59–2.31; p < 0.001), and 1.50 times higher likelihood of anxiety symptoms (OR, 1.50; 95% CI, 1.15–1.96; p = 0.003); while those with both UHC and S-COVID-19-S had 2.62 times higher depressive symptoms likelihood (OR, 2.62; 95% CI, 1.81–3.79; p < 0.001), and 3.12 times higher anxiety symptoms likelihood (OR, 3.12; 95% CI, 1.92–5.07; p < 0.001; Table 3).

In comparison to patients without UHC and drinking at “never/stopped/less” level, those without UHC and drinking at “unchanged or more” level had 3.26 times higher likelihood of having anxiety symptoms (OR, 3.26; 95% CI, 2.28–4.65; p < 0.001); those with UHC and drinking at “never/stopped/less” level had 3.31 times higher likelihood of developing anxiety disorders (OR, 3.31; 95% CI, 2.71–4.05; p < 0.001); while participants with UHC and drinking at “unchanged or more” level had 66% lower likelihood of experiencing anxiety symptoms (OR, 0.34; 95% CI, 0.16–0.69; p = 0.003; Table 4).

In comparison to patients without UHC and had physical activity at “never/stopped/less” level, those without UHC and did physical activity at “unchanged or more” level had a 23% lower likelihood of depressive symptoms (OR, 0.77; 95% CI, 0.64–0.94; p = 0.01) and a 25% lower anxiety symptoms likelihood (OR, 0.75; 95% CI, 0.58–0.97; p = 0.027); those with UHC and had physical activity at “never/stopped/less” level had 3.03 times higher likelihood of depressive symptoms (OR, 3.03; 95% CI, 2.45–3.74; p < 0.001) and 3.46 times higher anxiety symptoms likelihood (OR, 3.46; 95% CI, 2.76–4.35; p < 0.001); while patients with UHC and had physical activity at “unchanged or more” level had 58% lower depressive symptoms likelihood (OR, 0.42; 95% CI, 0.30–0.56; p < 0.001) and 58% lower anxiety symptoms likelihood (OR, 0.42; 95%CI, 0.28–0.64; p < 0.001; Table 4).

In comparison to people without UHC and ate “less healthy”; those without UHC and ate “unchanged or healthier” had a 64% lower likelihood of developing depressive symptoms (OR, 0.36; 95% CI, 0.25–0.52; p < 0.001), and a 64% lower likelihood of having anxiety symptoms (OR, 0.36; 95% CI, 0.23–0.56; p < 0.001); those with UHC and ate “less healthy” had 4.27 times higher likelihood of depressive symptoms (OR, 4.27; 95% CI, 2.57–7.08; p < 0.001) and 6.27 times higher likelihood of anxiety disorders (OR, 6.27; 95% CI, 3.46–11.35; p < 0.001); while patients with UHC and ate “unchanged or healthier” had 58% lower likelihood of depressive symptoms (OR, 0.42; 95% CI, 0.25–0.72; p = 0.001), and a 63% lower anxiety symptoms likelihood (OR, 0.37; 95% CI, 0.20–0.68; p = 0.002; Table 4).

In comparison to participants without UHC and with a low PB score; those without UHC and with a high PB score had a 43% lower likelihood of depressive symptoms (OR, 0.57; 95% CI, 0.44–0.73; p < 0.001) and a 45% lower anxiety symptoms likelihood (OR, 0.55; 95% CI, 0.43–0.71; p < 0.001); those with UHC and with a low PB score had 5.32 times higher likelihood of depressive symptoms (OR, 5.32; 95% CI, 4.11–6.89; p < 0.001) and 6.24 times lower anxiety symptoms likelihood (OR, 6.24; 95% CI, 4.80–8.11; p < 0.001); while participants with UHC and with a high PB score had 69% lower depressive symptoms likelihood (OR, 0.31; 95% CI, 0.21–0.45; p < 0.001) and 82% lower anxiety symptoms likelihood (OR, 0.18; 95% CI, 0.12–0.27; p < 0.001; Table 5).

In the current study, we found that people with underlying health conditions are more vulnerable to depression and anxiety symptoms. Having S-COVID-19-S made these patients more depressed and anxious. However, positive changes in lifestyles (e.g., doing more physical exercising, healthier eating) and adherence to protective measures showed benefits for mental health in those with underlying diseases. Unexpectedly, the constant or increase drinking behavior was found as a coping behavior to protect patients against psychological illnesses during the pandemic. Our findings could be helpful for governments and health organizations to develop appropriate strategies to alleviate the harmful impact of the pandemic on mental health, especially in patients with underlying health conditions.