Mental health challenges are often overshadowed by adverse physical health impacts, though psychopathologies have major adverse impacts globally. Depression is associated with increased risks for both suicide and all-cause mortality, with the WHO ranking depression as the third leading cause of burden of disease worldwide in 2004.Reference Chesney, Goodwin and Fazel1, Reference Mathers and Ties2 Consistently, studies have found the same core risk factors for depression among populations worldwide, regardless of country and region. In these studies, being of female sex,Reference Lee, Shin and Kim3–Reference Brown, Hambleton and Sobers-Grannum5 less educated,Reference Anstey, Von Sanden and Sargent-Cox4, Reference Brown, Hambleton and Sobers-Grannum5 a younger adult,Reference Ferrari, Charlson and Norman6 divorced/separated and/or widowed,Reference Lee, Shin and Kim3–Reference Brown, Hambleton and Sobers-Grannum5 unemployed,Reference Paul and Moser7 having low food security,Reference Lee, Shin and Kim3 and having perceived poor health or chronic diseases have been associated with increased risk of depression.Reference Carroll, Cassidy and Côté8, Reference Moussavi, Chatterji and Verdes9
Serious pandemics and epidemics are associated with increases in burden of depressive symptoms, anxiety disorders, and stress.Reference Preti, Di Mattei and Perego10, Reference Hossain, Tasnim and Sultana11 The COVID-19 pandemic is similar, with a related psychopathological pandemic calling for global attention.Reference Hossain, Tasnim and Sultana11 The determinants for COVID-19-related depression and other psychopathologies not only include exposure to illness and death due to COVID-19,Reference Holman, Thompson and Garfin12 but also pandemic-related government regulatory controls that has frequently led to social isolation,Reference Lei, Huang and Zhang13, Reference Luo, Guo and Yu14 job loss, and shortage of necessities.Reference Holman, Thompson and Garfin12 The COVID-19 pandemic-related lockdown regulations of extended social distancing, self-isolation, border closures, and travel restrictions has translated into workforce reductions and declining demand for products, leading to enhanced anxiety of impending economic crisis.Reference Nicola, Alsafi and Sohrabi15
Characterizing long-term adverse COVID-19 mental health impacts are even more critical given that the virus in some diseased patients has penetrated and caused deterioration of the central nervous system, producing neuroinflammation and other neurological manifestations.Reference Szcześniak, Gładka and Misiak16 It is unknown what impact this will have in the long-term, given lack of longitudinal studies of the current pandemic.
Long-term trends of depression and other psychopathologies following pandemics are poorly characterized given their infrequent occurrence and the lack of longitudinal studies.Reference Szcześniak, Gładka and Misiak16 Predominantly, studies assessing mental health risk factors during serious epidemics and pandemics such as COVID-19 are cross-sectional in design,Reference Preti, Di Mattei and Perego10, Reference Hossain, Tasnim and Sultana11 which are limited to assessing demographic, behavioral, and situational risk factors, and determinants at the same point in time. Cohort studies designed to study follow-up periods are limited to short time frames, typically a year-or-less.Reference Hossain, Tasnim and Sultana11
In contrast, there are extensive published research on single-episodic disaster mental health,Reference Pfefferbaum and North17 where survivor cohorts were followed over many years, sometimes decades, after a disaster episode.Reference Guo, He and Qu18–Reference Bosmans and Van der Velden20 These studies uncovered some unexpected findings. Two years following the 1988 earthquake in Armenia, greater earthquake-related material loss was associated with an increased prevalence of depression, but earthquake-related deaths in the family was not.Reference Armenian, Morikawa and Melkonian21 Twenty-four years after this earthquake, researchers found a declining prevalence of depression among survivors, but the prevalence of earthquake-related depression was still elevated and persistent.Reference Khachadourian, Armenian and Demirchyan19 They also found that depression was associated with increased risk of chronic diseases, especially heart diseases.
Whether the long-term trajectory of the extended COVID-19 pandemic psychopathologies is similar with the trajectory of single-episodic disasters or will follow a different trajectory is unknown. The study of factors for COVID-19 pandemic mental health and the capacity for successful adaptation and recovery can inform planning of more effective treatment programs and interventions for mitigating long-term mental health needs due to this and future pandemics.Reference Beaglehole, Mulder and Frampton22
To investigate the long-term trajectory of Covid -19 related depression, we undertook a survey in the 2 largest cities in the Central Asian country of Kazakhstan. This city-stratified randomized telephonic survey serves as a baseline for a planned longitudinal panel study. The survey follows a stratified random sampling design to estimate COVID-19 related risk factors for probable depression. Identifying and characterizing risk factors for depression during the pandemic may inform planning for improved interventions and policies for treatment and prophylaxis of the population after the pandemic, for future pandemics, and other enduring crises.Reference Beaglehole, Mulder and Frampton22
The survey aimed to answer the following research questions:
-
1) What are the associations of direct COVID-19 pandemic-related social factors such as COVID-19 deaths among family members, friends, and coworkers, and having disease among household members and others with depression?
-
2) What are the associations of indirect COVID-19 pandemic-related economic factors such as employment status before and during the pandemic and food security with depression?
-
3) What are the associations of perceived beliefs of COVID-19 threat, fear of self or other household family members developing COVID-19 disease, and stress in the family with depression?
Methods
Study Design and Participants
We designed a city-stratified randomized baseline survey by cell phone for a planned longitudinal panel study on the trajectory of probable depression associated with the pandemic in the 2 cosmopolitan cities of the high middle-income Central Asian country of Kazakhstan, formerly part of the Soviet Union. The survey participants were randomly sampled from Kazakhstan citizens 18-75 years-of-age. The survey was designed by faculty from Nazarbayev University School of Medicine and Graduate School of Business and was conducted by the professional BISAM-Central Asia company, which has more than 20 years of experience conducting surveys in Kazakhstan. Cell phone interviews were conducted from October 27-November 22, 2020. The survey was conducted during the “soft lockdown” regulatory period (starting from July 5, 2020, during which time the survey was conducted), which followed the “hard lockdown” (March 16-May 17, 2020) after the first confirmed COVID-19 case in the country on March 13, 2020 (Figure 1). The “hard lockdown” state-of-emergency regulations included closures of all non-essential businesses and schools (only online operations were allowed), regulated isolation in homes, required face masks, and required physical distancing during brief visits to purchase food and medicines. Travel from and to Kazakhstan cities and towns was banned and all flights were cancelled. “Soft lockdown” allowed for limited openings of businesses, a lifting of household isolation, domestic travel to and from cities and towns, and restricted international travel.
The 2-city randomized survey was conducted five-and-a-half months after the end of the government "hard lockdown" period and during the government “soft lockdown” COVID-19 period in Kazakhstan.
Two-thousand Kazakhstan citizens 18-75 years-of-age who are residing in the 2 largest cities of Kazakhstan were interviewed via telephone. Random sampling phone-calling was stratified by the 2 largest cities of Almaty (total estimated population of 2 055 724 as of January 1, 2020) and Astana (total estimated population of 1 136 156).23 From these populations, a total of 64.5% of the study population was estimated to be 18-75 years of age, the age range eligible for the study population.24 Astana, the smaller city, was over-sampled with 900 surveys to assure adequate sample size for sub-analyses, with the remaining 1100 surveys being conducted among Almaty city residents.
A total of 0.06% of the eligible population was randomly sampled in Astana and 0.08% in Almaty. No study respondent shared a household with any other study respondent; thus, the 2000 respondents also represented 2000 different households. We obtained verbal informed consent from persons eligible to participate.
Ethical Standards
Our study protocol for the 2-city randomized telephonic baseline survey was approved by the Institutional Research Ethics Committee of the Nazarbayev University, Astana, Kazakhstan (№316/29092020). All data received by the researchers were anonymized, with no personal identifiers such as name, birthdate, address, or phone number.
Survey Instrument
The phone survey consisted of 70 questions, with skip patterns reducing the number of questions for individual participants, and was designed to be completed within 25 minutes. A 16-question version of the internationally validated Likert-scale 20-question Center for Epidemiologic Studies Depression Scale (CESD-20) was utilized for this study, which was designed to identify probable depressive disorders in the general population.Reference Radloff25 The validated 16-question variant (CESD-16), with individual items scored from 0-3 on a Likert scale, excluded the 4 positive affective questions that were included in the CESD-20; this exclusion improved scale validity for identifying probable depression for many Asian countries.Reference Demirchyan, Petrosyan and Thompson26, Reference Iwata, Umesue and Egashira27 Scores greater than 12 are considered probable depression.Reference Demirchyan, Petrosyan and Thompson26
Other questions related to demographics, economic status, and perceptions of the pandemic were adapted from questions used in our prior research and from other previously published research.Reference Lei, Huang and Zhang13 All English-language questions, including CESD-16 questions, were translated into Russian and then back-translated into English for further corrections, to assure meaning was aligned with the original questions. The final Russian-language survey was pre-tested for further corrections.
Missing data were sporadic, with no systematic missing data in any participant’s reported answers. Among variables used in analyses, there were no missing data in questions for age, sex, city of residence, residential status, number in household, having a chronic disease, household family member being 65 years or older, being the primary household wage earner, currently sufficient income to buy food, number of household family members diagnosed with COVID-19, household family members with diagnosed or suspected COVID-19, and employment status before, during the pandemic, and at the time of the survey. No data were missing for Likert-scale factors associated with attitudes or perceptions, such as stress in the family, fear of COVID-19 infection, and perceived threat of COVID-19.
There were 3 answers missing (0.01% of all 214 participants reporting having a chronic disease) to the question asking for type of chronic disease. These 3 were included in the “other” category. There were 3 refusals (0.15%) to answer the marital status question (these were added to the married/common law category based on their relationship to the primary household earner), 4 refusals (0.2%) for the educational level achieved question (added to the lowest educational category based on job identification), and 1 refusal (0.05%) for the ethnicity question, (added to the collapsed “other” category), which was included into the collapsed “other” category. For household monthly income, as measured in the local currency tenge (at the time of the survey, 420 tenge was approximately equal to 1 USD), 214 participants reported not knowing or refused to answer (10.74%). The household monthly income “other” category was composed to include these non-responders.
There were no missing data for the following factors, but answers “no” and “don’t know” were collapsed into a single no/don’t know category in each of these factors: friends or coworkers diagnosed with COVID-19 (“no” at 54.7%, “don’t know” at 5.3%), COVID-19 deaths among household family member or anyone close (“no” at 89.1%, “don’t know” at 5.3%), and COVID-19 deaths among acquaintances or coworkers (“no” at 75.6%, “don’t know” at 8.8%).
Analyses
Stata 16.1 was utilized for all bivariate and multivariate analyses. To adjust survey-sampling effects in bivariate and multivariate analyses for city-stratified random sampling, probability weights were calculated for each city. The probability weight was computed to be 1205.8 (1 326 379 eligible population/1 100 sample size) for Almaty and 814.5 (733 062/900) for Astana. These stratified probability weights, along with the randomized sampling design, were utilized in Stata’s survey adjustment for stratified random sampling, using simple and multivariate Poisson regression models with robust variances to produce appropriately survey-adjusted prevalence ratio point estimates with confidence intervals and P values.Reference Barros and Hirakata28
The probability cut-off α was set at 0.05 for rejecting the null hypotheses. Any independent socio-economic and factors that showed bivariate associations of P ≤ 0.10 with the depression outcome (CESD-16) were included in the intermediate multivariate model. Any independent covariates with P ≥ 0.10 were removed to form the final survey-adjusted multivariate Poisson model of prevalence ratios for depression. Associations for self-report Likert scale questions concerning attitudes and perceptions, including stress in the family, fear of COVID-19, and perceived threat of COVID-19, were also analyzed using survey-adjusted Poisson regression models with probable depression as the outcome.
Results
Based on the January 1, 2020 population estimates, the sample ethnic distribution among Kazakh citizens was similar with the distribution found in the national population.23 Kazakh ethnicity contributed to 68.5% of the sample as compared to 70.5% nationally, Russian ethnicity was 18.9% versus 20.1%, and other ethnicities contributed to 12.6% versus 9.4% (Table 1). Other ethnicities in the sample included Uyghur, Tatar, Korean, German, Uzbek, and Ukrainian (not shown on table). Mean age in the random sample of 18-75 years-of-age was 34 years, whereas on the January 1, 2020 in the same age range, the general population was estimated to have a mean age of 43 years.24 The sample had a percentage of 59.9% women, versus a general population percentage of women of 51.5%. The response rate for contacts consenting to participate in the survey was 37.4%, where 5348 cell phone contacts were made to achieve the study target of 2000 participants providing consent to participate in the survey.
Socio-demographic characteristics of participants (n = 2000)
Bivariate analyses models were run for each factor, utilizing simple Poisson regression with city-stratified randomized survey adjustment for potential risk factors with having probable depression. Table 2 showed bivariate analyses of individual demographic factors. Female sex, lower education, being divorced/separated or widowed, having a major chronic disease (cardiovascular disease or diabetes mellitus), and being divorced/separated or widowed, or not having a job before, during the pandemic, or at the time of the interview all showed statistically significant positive associations with probable depression. The prevalence ratio for the age groups 20-30 and 31-40 years-of-age, as compared to 18-19 year-of-age group, showed bivariate statistically significant protective associations.
Separate bivariate simple Poisson regression models, with survey adjustment for city-stratified randomized sampling, showing associations of individual socio-demographic factors with probable depression
Table 3 showed bivariate associations for household demographic factors for depression among participants. The factor sufficient income to buy food at time of interview was strongly inversely associated with probable depression. Table 4 provides bivariate analysis of associations forCOVID-19 episodic, psychological and belief factors with probable depression. Diagnosed or suspected COVID-19 in the household, friends or coworkers diagnosed with COVID-19, and COVID-19 death of acquaintances or coworkers were all statistically significantly associated with increased prevalence of probable depression. In contrast, the factor COVID-19 death of household member or anyone close was not statistically significant. The self-reported belief that COVID-19 is a serious threat, and fear of self or family members developing COVID-19, was not associated with probable depression, but perceived stress level in the family was associated, showing a dose response. Very high family stress was found to have a PR = 13.42 (95% CI 7.11-25.32; P < 0·001) for probable depression as compared to very low family stress. However, this family stress factor showed high multicollinearity with the 3 episodic COVID-19 factors that were statistically significant in the bivariate analysis; the COVID-19 episodic factors were introduced into the multivariate Poisson regression model in lieu of the family stress factor.
Separate bivariate simple Poisson regression models, with survey adjustment for city-stratified randomized sampling, showing associations of household socio-demographic factors with probable depression
Bivariate simple Poisson regression models, with survey adjustment for city-stratified randomized sampling, showing associations of COVID-19 related episodic, psychological, and belief factors with probable depression
The bivariate associations of individual and household demographic factors and COVID-19 episodic factors with at least 1 value with a P value ≤0.10 were introduced into the initial survey-adjusted multivariate Poisson regression model (Table 5). Factors that showed P values >0.10 in the initial model, including decreased household income during pandemic (P = 0.802) and primary city of residence (P = 0.318), were excluded from the final multivariate model. The factor household monthly income was also excluded from the final model because the only value for this factor was a P value < 0.10 for “do not know or refused,” which is not interpretable.
Multivariate Poisson regression initial and final models, with survey adjustment for city-stratified randomized sampling, showing associations of individual and household demographics and COVID-19-related episodic associations with depression
a Factors with no P value ≤ 0.10 in the initial model were removed for the final model.
The direct and indirect pandemic-related factors with statistically significant associations with probable depression included household diagnosed or suspected COVID-19 with a prevalence ratio of PR = 1.48 (95% CI 1.09-2.02; P = 0.013), friends and coworkers being diagnosed with COVID-19 with PR = 1.43 (1.04-1.95; P = 0.026), losing your job during the pandemic and not regaining it as compared to having a job with PR = 1.80 (1.04-3.11; P = 0.037), and having insufficient income to buy food with PR = 1.54 (1.15-2.06; P = 0.004). No factors related to knowing someone who died due to COVID-19 were found to be associated with probable depression.
Discussion
Determinants and risk factors for long term psychopathologies such as depression resulting from severe single-episodic disasters such as earthquakes are well studied, with long-term cohort and longitudinal studies following survivors.Reference Guo, He and Qu18, Reference Khachadourian, Armenian and Demirchyan19, Reference Armenian, Morikawa and Melkonian21 Little is known about psychopathologies after extended pandemics with regulatory closures and isolation.
Our 2-city stratified randomized telephonic survey, conducted in the high middle-income country of Kazakhstan, a former part of the Soviet Union, is a baseline study for a planned follow-up longitudinal panel study that focuses on pandemic-related social and economic risk factors for depression and its trajectory. The prevalence of depressive symptoms during the “soft lockdown” regulatory period of the first year of pandemic was 9.4% in our sample, which was lower than the reported global prevalence of depression (24%) during the COVID-19 pandemic.Reference Castaldelli-Maia, Marziali and Lu29 Recent meta-analyses showed a significant impact of the COVID-19 outbreak on population mental health and increased rates of depression in the general population as compared to baseline assessments before the pandemic; however, there was a wide variation in the rates across regions and countries, ranging from 7.45%-48.30%,Reference Bueno-Notivol, Gracia-García and Olaya30 with Asian countries having the lowest rates.Reference Castaldelli-Maia, Marziali and Lu29 This can be explained by cultural differences, as mental illness is more stigmatized and under-reported in Asian countries,Reference Zhang, Sun and Jatchavala31 and differences between settings, such as between high and low- and middle-income countires.Reference Semrau, Evans-Lacko and Koschorke32
We found many risk factors for depression unrelated to the pandemic that are also found worldwide. Being a woman and being divorced/separated (as compared to married/cohabitating) independently increased the prevalence by close to double, similar with other studies.Reference Lee, Shin and Kim3–Reference Brown, Hambleton and Sobers-Grannum5 We also found that younger individuals 18-19 years of age were at greater risk for having probable depression than those 31-40 years of age, similar with findings from other studies.Reference Ferrari, Charlson and Norman6 Those not completing secondary/vocational school were more likely to have probable depression than their better-educated counterparts.Reference Anstey, Von Sanden and Sargent-Cox4, Reference Brown, Hambleton and Sobers-Grannum5 Those with chronic cardiovascular disease (with a prevalence of more than 3-fold) or diabetes mellitus (an almost 2.5-fold increase) were more likely to be depressed, similar with other findings on chronic diseases and depression,Reference Moussavi, Chatterji and Verdes9 though the magnitude of these associations may have been enhanced with media reporting on COVID-19 complications linked with chronic conditions.
More importantly, we found 2 COVID-19 factors directly associated with probable depression. The diagnosis or suspected COVID-19 disease among family or self in the household and having had friends or co-workers diagnosed with COVID-19 were both independently associated with an increased prevalence of probable depression by half. This association may be related to a need to protect self, family, and friends. In contrast, we found no association with probable depression for factors related to participants knowing persons who died from COVID-19, including deaths in the family or someone close, and deaths among coworkers or acquaintances. This finding is consistent for survivors of earthquakes, where earthquake-related deaths of family members was not associated with depression.Reference Armenian, Morikawa and Melkonian21 This can be explained by previous findings, where the normal grieving process for the death of family members or friends is time-limited, generally having no association with depression.Reference Shear, Simon and Wall33 Having had COVID-19 disease among contacts, on the other hand, may exemplify the current and future potential harm of disease that may impact social and economic survival.
Two economic factors related to the pandemic were also associated with probable depression. The first was job loss during the pandemic, with subsequent continued unemployment, and the second was not having sufficient funds to buy food at the time of the interview during the soft lockdown. Becoming unemployed during the pandemic and continuing to be unemployed as compared to being continuously employed increased prevalence of probable depression by 80%. Not having sufficient income for food independently increased risk by more than half. Both these risk factors are similar with previous studies.Reference Lee, Shin and Kim3, Reference Paul and Moser7
Beliefs and psychological factors measured on Likert scales, including perceived seriousness of threat of COVID-19, reported that level of fear of family developing COVID-19 disease and perceived stress levels in the family were assessed independently. Strikingly, perceived threat of COVID-19 and fear of developing COVID-19 disease in the family were not associated with probable depression. However, increasingly higher stress levels in the family showed a strong dose response in increasing the prevalence of probable depression. Participants reporting very high levels of stress in the family were 13 times more likely to be depressed than the very low stress reference group. Family stress has long been associated with risk of depression,Reference Phelan, Schwartz and Bromet34 though the much stronger associations found in our study may have been amplified by the pandemic. We find that the level of reported fear of developing disease and belief in the severity of disease do not contribute to probable depression, which stands in contrast to that of COVID-19 infection events and related economic duress, which are associated with depression.
Limitations of the survey included our inability to confirm causality given the cross-sectional design. Also, though the ethnic group distribution of the survey was similar with that of the distribution of ethnic groups found in the general population, somewhat larger proportions of men than women and of older adults than younger refused to participate in the survey; however, inter-group regression analyses of probable depression should not likely be biased due to disproportionate response rates among sex and age groups. Also, depression level could be underreported in our study due to the telephone mode of data collection, which is susceptible to underreporting when eliciting sensitive information, including mental health. However, we chose self-reported telephone survey for our baseline longitudinal study because it provides greater control over avidity and attrition bias compared to web-based survey (Milton et al., 2017). Strengths of our study are that it is a randomized sampling design that improves the likelihood of generalizability and that it provides a snapshot of the trajectory depression may follow over time following the serious COVID-19 extended pandemic.
Our findings provide new information on high-risk groups that may guide clinical practice for psychiatrists, therapists, and social workers in supplementing the development of improved treatment plans. The current findings may also be used to inform public health professionals on designing more effective prophylactic mental health policies and interventions for future pandemics and to address long-term population needs following the current pandemic.
The baseline survey also provides a foundation for the longitudinal panel study to follow the same cohort for determining psychopathological trajectories and mental health resilience long-term. Follow-up also provides an assessment of possible COVID-19 neuropathological chronic changes by comparing mental health status and symptoms of those who developed COVID-19 to those who showed no disease symptoms, while adjusting for confounding.
Characterizing social and economic factors associated with depression during the COVID-19 pandemic may inform the design and implementation of more effective programmatic interventions and policies for the treatment, prophylaxis, and recovery of the population during and after extended pandemics, and aid in the preparation of meeting new long-term demands for mitigating long-term mental health needs.Reference Beaglehole, Mulder and Frampton22
Data availability statement
The database will be made available to other researchers by the authors after approval of a methodologically sound proposal with public health and/or clinical added value.
Acknowledgements
We would like to acknowledge the support of Nazarbayev University, and the advice provided by colleagues in the health care sector, especially during the complications of the pandemic.
Author contribution
AR and BLC designed the survey and reviewed and verified the data during and at the end of the data collection period and conducted analyses. All authors, AR, BLC, TY, and RA, contributed to the interpretation of results, and writing, revising, and editing the manuscript. All authors had full access to all the data in the study and had final responsibility for the decision to submit for publication.
Funding statement
Nazarbayev University, the funding source of the study (Small Grant 090118FD5302), played no role in the conception and development of the study, data collection, data analysis, data interpretation, or writing of the manuscript. All authors had full access to all the data in the study and had final responsibility for publication submission decisions.
Competing interests
We declare no competing interests.
Ethical standard
The authors assert that all procedures contributing to this work comply with the ethical standards of the Declaration of Helsinki and that the study received ethical approval from the Institutional Research Ethics Committee of the Nazarbayev University, Astana, Kazakhstan.
