Intimate partner violence during COVID-19: systematic review and meta-analysis according to methodological choices

Abstract

Background: Intimate Partner Violence (IPV) is the most common form of interpersonal violence and a major public health problem. The COVID-19 pandemic might have contributed to an increase in IPV experiences. To evaluate changes in IPV prevalence during the pandemic, it is important to consider studies' methodological characteristics such as the assessment tools used, samples addressed, or administration modes (e.g., face-to-face, telephone or online interviews), since they may influence disclosure and were likely affected by pandemic-imposed mobility restrictions.

Methods: Systematic review and meta-analysis of empirical studies addressing IPV against women, men, or both, during the COVID-19 period. We searched six electronic databases until December 2021, including articles in English, German, Spanish, French or Portuguese languages. We extracted and synthesised characteristics of studies related to sampling (clinical, community, convenience), type assessment tool (standardised questionnaire, specifically created questions), method of administration (online, telephone, face-to-face), and estimates of different forms of IPV (physical, sexual, psychological). IPV estimates were pooled stratified by study characteristics using random-effects models.

Results: Of 3581 publications, we included 103 studies. Fifty-five studies used a standardized instrument (or some adaptations) to assess IPV, with the World Health Organisation Questionnaire and the Revised Conflicts Tactics Scales being the most frequent. For 34 studies, the authors created specific questions to assess IPV. Sixty-one studies were conducted online, 16 contacted participants face-to-face and 11 by telephone. The pooled prevalence estimate for any type of violence against women (VAW) was 21% (95% Confidence Interval, 95%CI = 18%-23%). The pooled estimate observed for studies assessing VAW using the telephone was 19% (95%CI = 10%-28%). For online studies it was 16% (95%CI = 13%-19%), and for face-to-face studies, it was 38% (95%CI = 28%-49%). According to the type of sample, a pooled estimate of 17% (95%CI = 9%-25%) was observed for studies on VAW using a clinical sample. This value was 21% (95%CI = 18%-24%) and 22% (95%CI = 16%-28%) for studies assessing VAW using a convenience sample and a general population or community sample, respectively. According to the type of instrument, studies on VAW using a standardized tool revealed a pooled estimate of 21% (95%CI = 18%-25%), and an estimate of 17% (95%CI = 13%-21%) was found for studies using specifically created questions.

Conclusions: During the pandemic, IPV prevalence studies showed great methodological variation. Most studies were conducted online, reflecting adaptation to pandemic measures implemented worldwide. Prevalence estimates were higher in face-to-face studies and in studies using a standardized tool. However, estimates of the different forms of IPV during the pandemic do not suggest a marked change in prevalence compared to pre-pandemic global prevalence estimates, suggesting that one in five women experienced IPV during this period.

Keywords: COVID-19; Intimate partner violence; Methods; Prevalence; Systematic review.

Fig. 1

PRISMA flow diagram

Fig. 2

a Forest plot of (“any” type of) intimate partner violence against women (VAW) prevalence pooled by method of administration in telephone, online, face-to-face, or other, between-study variance, τ² = 0.00609. I²: Telephone = 97.32%, p = 0.00; Online = 99.04%, p = 0.00; Face to face = 97.33%, p = 0.00; Overall = 99.04%, p = 0.00. b Forest plot of (“any” type of) intimate partner violence against men (VAM) prevalence pooled by method of administration (all were conducted “online”), between-study variance, τ² = 0.00748. I²: Online = 96.67%, p = 0.00. c Forest plot of (“any” type of) intimate partner violence against women and men prevalence pooled by method of administration in telephone, online, face-to-face, or other (from studies where sex-disaggregated prevalence estimates were not available, VAW&M), between-study variance, τ² = 0.03787. I²: Online = 99.89%, p = 0.00; Overall = 99.84%, p = 0.00

Fig. 3

a Forest plot of (“any” type of) intimate partner violence against women (VAW) prevalence pooled by type of sample in clinical, convenience or general population/community sample, between-study variance, τ² = 0.00609. I²: Clinical = 94.22%, p = 0.00; Convenience = 99.02%, p = 0.00; General pop/community = 99.26%, p = 0.00; Overall = 99.04%, p = 0.00. b Forest plot of (“any” type of) intimate partner violence against men (VAM) prevalence pooled by type of sample (all used a “convenience” sample), between-study variance, τ² = 0.00748. I²: Convenience = 96.67%, p = 0.00. c Forest plot of (“any” type of) intimate partner violence against women and men prevalence pooled by type of sample in clinical, convenience or general population/community sample (from studies where sex-disaggregated prevalence estimates were not available, VAW&M), between-study variance, τ² = 0.03787. I²: Convenience = 99.86%, p = 0.00; Overall = 99.84%, p = 0.00

Fig. 4

a Forest plot of (“any” type of) intimate partner violence against women prevalence pooled by instrument used for assessment in standardized tool, specifically created questions, or other, between-study variance, τ2 = 0.00609. I²: Standardized tool = 98.63%, p = 0.00; Created Quest = 99.30%, p = 0.00; Overall = 99.04%, p = 0.00. b Forest plot of (“any” type of) intimate partner violence against men prevalence pooled by instrument used for assessment in standardized tool and specifically created questions or other, between-study variance, τ² = 0.00748. I²: Standardized tool = 91.41%, p = 0.00; Overall = 96.67%, p = 0.00. c Forest plot of (“any” type of) intimate partner violence against women and men prevalence pooled by instrument used for assessment in standardized tool and specifically created questions or other (from studies where sex-disaggregated prevalence estimates were not available), between-study variance, τ² = 0.03788. I²: Standardized tool = 99.90%, p = 0.00; Created Quest = 99.21%, p = 0.00; Overall = 99.84%, p = 0.00