The novel coronavirus disease (COVID-19) outbreak that emerged in Wuhan, China, in 2019 soon spread to the rest of the world, causing more than 4.2 million infections and about 85 thousand deaths within the first year in 2020. As a result, the World Health Organization (WHO) declared COVID-19 a global pandemic in March 2020. After that, many countries started adopting data collection protocols from municipalities and local counties to help them make informed decisions to curb the spread of the causative agent severe acute respiratory syndrome coronavirus 2 (SARS‑CoV‑2). However, limited information availability led many countries to implement large-scale control measures such as lockdowns.

The state and federal governments in the United States implemented several mandates to manage the spread of SARS‑CoV‑2. The federal government, however, opted for a hands-off approach allowing individual states to decide how to limit the spread of the disease. There were a variety of mandates, ranging from school and work closures to stay-at-home orders. Moreover, the increase in COVID-19 cases and deaths led to increased pandemic-related worry among people. Policies determining pandemic-related decisions can adversely affect mental health leading to depression and anxiety. As a result, a jump in mental health conditions was reported in the second quarter of 2020 compared to the last quarter of 2019.

Study: Understanding mental health trends during COVID-19 pandemic in the United States using network analysis. Image Credit: rudall30 / Shutterstock

Several studies indicated that lockdowns and many COVID-19-related policies could increase mental health burdens, especially for vulnerable groups. However, some policies were shown to positively impact both psychological and physical health. Moreover, vaccines were also reported to reduce mental health issues following their initial roll-out in late 2020. However, a large proportion of the population was hesitant to receive vaccination and continued to experience similar mental distress levels. Therefore, dynamic interpretation of data is vital across time concisely and clearly.

One previous study by Bulai and Amico implemented a network analysis to determine COVID-19 interactions among various regions of Italy and the impact of the Italian government policies to control the spread of the disease. They used six indicators to form a correlation network known as “Covidome,” which showed the north-south clustering of the regions in Italy. Moreover, they also observed a significant difference in Covidome fluctuations between the first and second pandemic waves based on political choices between different regions.

A new study posted in the pre-print server medRxiv* aimed to apply clustering and network analysis to determine the connectivity between the states and used COVID-19-related mental health indicators to understand how COVID-19 impacts mental health across the United States.

About the study

The study was based on survey results from Carnegie Mellon University’s Delphi Group. Survey questions varied from the economic impact of COVID-19 and physical health to behavioral prompts and mental health. The participants’ responses were aggregated, collected, and made available publicly. Three indicators that indicated the impact of COVID-19 on mental health were the percentage of participants who experienced feelings of anxiety within the past seven days, the percentage of participants who felt worried about their finances for the next month, and the percentage of participants who felt depressed within the past seven days. Information on daily confirmed COVID-19 cases and deaths was also obtained.

The results collected from the survey were categorized into two-time frames, the first ranging from 8th September 2020 to 2nd March 2021 and the second from 2nd March 2021 to 10th January 2022. The plotting of daily COVID-19 cases, hospitalizations, and deaths indicated three waves. Based on them, the data was split into three different periods, 1st April to 1st July 2021, 2nd July to 11th November 2021, and 2nd November 2021 to 10th January 2022. In addition to the individual states, four geographical regions (south, west, northeast, and midwest) of the United States and political party reference were used to determine any mental health trends among the states clustered as a result of similar politically or geographically established communities. Additionally, the survey results on mental health indicators were plotted based on political preference and geographical regions.

The policies implemented to control the COVID-19 outbreak were categorized as worried about finances or related to depression and anxiety. Clustering and correlation networks were used to determine the connectivity of states. Finally, dynamic connectome analysis was used to understand the link between government policy, mental health indicators, and their relationship across political parties and geographical regions. Both correlation values and eigenvector centrality values were analyzed for mental health indicators. The minimum and maximum correlation values were determined for each period and verified using the eigenvector centrality values.

Study findings

The results indicated no clear community distinction for the three mental health indicators except for a slight South region clustering. An allegiance matrix constructed using the three mental health indicators showed three main clusters, out of which the south geographic region was most interesting. However, the south region did not include Arkansas, West Virginia, and Virginia, while it included Nevada, North Carolina, and California, which are non-southern states.

The lowest minimum and highest maximum correlation values were observed from the northeastern region in the first period for the feeling anxious variable. In the second period, the northeast showed the lowest correlation values, while the south showed the highest. The west had the lowest correlation values, while the midwest had the highest.

In the first period, the minimum and maximum correlation values for feeling depressed were observed in the midwest region. In the second period, the maximum value was observed in the south and the minimum in the northeast. During the final period, northeastern states had the highest correlation, and western states had the lowest.

The worried about finances variable had a higher maximum correlation in the first period and a decreasing correlation in subsequent periods. In the first period, the midwest region had the lowest correlation values while the south had the highest values. During the second period, the south had the highest values, and the northeast had the lowest values. Finally, third-period values were observed to be lowest in the northeast and highest in the west.

On the feeling anxious variable, the midwest exhibited minimal eigenvector centrality and northeast maximum values. In the second period, maximum values were observed in the south and minimum values in the northeast. The south had the lowest values during the third period, while the northeast had the highest.

There were only two regions with minimum and maximum eigenvector centrality values for the feeling depressed variable. The midwest was observed to have maximum values for the first period and minimum and maximum values for the second period. Maximum values were observed in the northeast during the third and second periods, while minimum values were observed during the third period.

For the worried about finances variable, the midwest was observed to have the maximum values and the south as the minimum values in the first period. For the second period, the south was observed to have both minimum and maximum values. Finally, the west had maximum values for the third period, and the midwest had minimum values.

Furthermore, the correlation values increased for the Democratic states but not for Republican states. For the feeling anxious variable, both the maximum and minimum correlation values were observed in the Democratic states for the first period. For the second period, the maximum values were observed in the Republican states and minimum values in the Democratic states. For the third period, the minimum values were observed in the Republican states and the maximum in the Democratic states.

The Republican states were observed to have minimum values for all three periods for the feeling depressed variable. The maximum correlation values were observed in the Democratic states for the first and third periods while in the Republican states for the second period. Moreover, for the worried about finances variable, all maximum and minimum values were observed in the Republican states for all the periods.

For the feeling anxious variable, minimum eigenvector centrality values were observed in the Republican states, and the maximum values were observed in the Democratic states during the first wave. Both maximum and minimum values were observed in the Republican states for the second wave while in the Democratic states for the third wave.

For the feeling depressed variable, minimum eigenvector centrality values were observed in the Republican states in the first and second periods and the Democratic states in the third period. On the other hand, the maximum eigenvector centrality values were observed in the Democratic states in the first and second periods and Republican states in the second period. For the worried about finances variable, all maximum and minimum eigenvector centrality values were observed in the Republican states.

Therefore, the current study demonstrated a similar trend for worried about finances and feeling anxious among the Republican and southern states between 3rd March 2021, and 10th January 2022. However, no identifiable communities that resembled political parties or geographical regions were reported for the feeling depressed indicator. Furthermore, the depressed and anxious feelings variables overlapped with the increased COVID-19 cases, hospitalizations, and deaths and the spread of the Delta variant.

Limitations

The study has certain limitations. First, anxiety can be caused by other sources, such as media exposure and negative COVID-19 experience. Second, most lockdown measures had ended during this study’s data collection. Third, certain states were loosening restrictions which could influence the participants’ response to the survey. Fourth, the closeness of specific policies makes it difficult to determine their impact on mental health. Fifth, the study might not be sensitive to short-term changes in correlation. Finally, the interpretation of the feelings of depression variable is difficult.

*Important notice

medRxiv publishes preliminary scientific reports that are not peer-reviewed and, therefore, should not be regarded as conclusive, guide clinical practice/health-related behavior, or treated as established information.