Changes in prevalence of mental health problems during the first year of the COVID 19 pandemic a systematic review and dose-response meta-analysis with associated control measures

Results

The flow of study selection is shown in Figure 1 (to be included). Overall, we included 40 studies with data from 7 conditions and 117 timepoints and condition combinations (references of included studies in appendix). The studies were conducted in 16 different countries. The pre-pandemic data were collected as early as 2014, while the most recent data were from 01-2021.
The median sample size across the timepoints was 1993 participants. The median of the mean participant age was 46 years (ranged from 5 to 72 years), and over half of the participants across all studies were women. The cumulative COVID-19 cases and deaths, the stringency, economic support and containment and health indices varied widely across time points (table 1b). Most studies (23/40) were repeated cross-sectional surveys. The risk of an unrepresentative sample and non-response bias was high or unclear in most studies. In contrast, most studies showed a low risk of information bias (table 1a).

Table 1a: Characteristics of study participants, countries, and the COVID-19 pandemic

	No. of timepoints	Median	Low	High
No of participants	117	1993	38	90798
Mean age (years)	93	46	5	72
Percentage of females (%)	117	53	50	80
GDP per capita in US$	117	46406	11371	70920
Gini index	117	36	25	52
Days since first case	117	102	1	340
Stringengy	117	73	29	94
Cumulative cases	117	158	0	2701
Cumulative deaths	117	9	0	75
Note:
Cumulative cases and deaths are per 100.000 people

Table 1b: Characteristics of studies

Subgroups	Number of studies	Number of timepoints
Population
Adolescents	1	2
Adolescents,Adults	1	4
Adults	30	85
Children	2	6
Children,Adolescents	2	8
Elderly	4	12
Condition
ADHD	1	2
Alcohol/Substance Abuse	3	10
Anxiety	12	26
Depression	14	30
Mental Wellbeing	2	4
Psychological Distress	13	31
Sleep Disturbance	7	14
Design
1	23	63
2	17	54
Risk of unrepresentative sample
0	13	40
1	9	24
2	12	35
NA	7	18
Risk of information bias
0	29	87
1	1	4
2	3	8
NA	7	18
Risk of non-response bias
0	9	22
1	10	36
2	14	41
NA	7	18
Country
Canada	1	2
Chile	1	4
China	6	16
Czechia	1	2
Ecuador	1	2
Germany	2	6
Hong Kong	1	4
Iran	1	2
Italy	1	2
Japan	2	5
Netherlands	2	8
Spain	1	2
Sweden	1	2
Switzerland	1	2
United Kingdom	8	26
United States	10	32

Meta-analysis of pre- versus during-pandemic prevalence of mental health problems

Of the 40 included studies, 25 provided measurements before and during the pandemic and contributed 38 ORs to the pre-during meta-analysis. The summary ORs for each condition are shown in Figure 2.

Figure 2: Meta-analysis of odds ratios (OR) for people above a threshold on a symptoms scale during the pandemic compared to before the pandemic. OR>1 means that the odds of people above the threshold is larger during the pandemic.

There was substantial heterogeneity in in all conditions, as shown in the width of the prediction intervals. We explore whether heterogeneity could be potentially explained by differences in the scale used to measure the symptoms, age, sex, GDP per capital,GINI inequality index,and risk of bias int he studies (Appendix in https://rpubs.com/geointheworld/APPENDIX_MHCOVID_dichotomous). There was limited evidence that the use of short versions of scales instead of longer (e.g. using PHQ-2 instead of PHQ-9) was associated with larger ORs.

Dose-response meta-analysis

Dose-response meta-analyses were performed only for anxiety, depression, and psychological distress which presented enough data across different timepoints. Figures 3-5 show the trajectory of the OR as a function of the days since the days of the first case in the study country, the stringency index, the cumulative number of cases and the cumulative number of deaths. The odds to have a symptoms score above a threshold increases on average up to 90 days after the first reported cases for the three conditions; thereafter they decrease or remain at a stable level but with large uncertainty. The odds also increase with increasing stringency, number of cases and deaths reported.

Sensitivity analyses after excluding a study with much longer follow-up than the other studies and after changing the location of the knots in the splines did not materially change the dose-response shapes (see Appendix in https://rpubs.com/geointheworld/APPENDIX_MHCOVID_dichotomous).

Figure 3 Dose-response meta-analysis plots of the odds ratios (OR) for depression as function of the days since the days of the first case in the study country, the stringency index, the cumulative number of cases and the cumulative number of deaths. Confidence intervals are shows as dashed lines. OR>1 means that the odds of people above the threshold increases over time.

Figure 4 Dose-response meta-analysis plots of the odds ratios (OR) for anxiety as a function of the days since the days of the first case in the study country, the stringency index, the cumulative number of cases and the cumulative number of deaths. Confidence intervals are shows as dashed lines. OR>1 means that the odds of people above the threshold increases over time.

Figure 5 Dose-response meta-analysis plots of the odds ratios (OR) for psychological distress as a function of the days since the days of the first case in the study country, the stringency index, the cumulative number of cases and the cumulative number of deaths. Confidence intervals are shows as dashed lines. OR>1 means that the odds of people above the threshold increases over time.