The effectiveness of web-based grief intervention for adults who lost a loved one: A systematic review and meta-analysis

Search strategy

We conducted a comprehensive systematic literature search across seven electronic databases and one clinical registration center: The Cochrane Library, PubMed, PsycARTICLES, Embase, Web of Science, CINAHL, MEDLINE, and Clinical Trials, from the start date of the index to February 1, 2024.We included literature without imposing any language restrictions.The references of the identified studies were searched manually for additional potential articles. Search strings or terms were derived from the following keywords: psychotherapy, bereavement, Internet-based. In the advanced search of the database, a search strategy was constructed based on the above search terms using the “Medical Subject Headings (MESH)” and “Title/Abstract” with appropriate “AND” and “OR” Boolean operators. The complete search strategy and number of search results for each database are presented in Appendix 1.

Study selection

The inclusion criteria were as follows: (1) participants were aged 18 years or older and bereaved; (2) the experimental group received a grief intervention primarily through the implementation of network technology; (3) the control group received conventional treatment, placebo, blank control, waiting list, or other grief interventions facilitated by non-web-based technology; (4) the study outcome indicators included grief, PTSD, depression, or anxiety, and the type of assessment scale was not limited; and (5) the study design was a randomized controlled trial (RCT). Trials that did not meet the inclusion criteria were excluded. In addition, studies lacking basic information or indicators of interest, baseline incomparable studies between the experimental and control group, and studies with duplicate data were excluded.

EndNote X9 was used to exclude duplicate records and manage the included studies. Two researchers (FQ and DYY) selected studies independently based on a two-step screening process consisting of title and abstract screening followed by full-text review to identify studies according to the eligibility criteria. Other disputes and disagreements were resolved through discussions with a third researcher (THT) to reach a consensus. For example, DYY believes that the web-based grief intervention study published by Wagner et al. [12], 2007 meets the standards, while FQ believes that the study has data duplication issues and should be excluded. After discussing with THT, a consensus was reached that there was data overlap between this study and another included article published by Wagner et al. [13], 2006, so the duplicated portion was excluded. Data other than duplicate publications were confirmed not to contribute further. Ultimately, the decision was made to exclude the study.

Quality assessment and data extraction

Two researchers (FQ and DYY) used the Cochrane Collaboration Tool for assessing the risk of bias in inclusion trials through Review Manager 5.3, which included the following: (1) random sequence generation, (2) allocation concealment, (3) blinding of participants and personnel, (4) blinding of outcome evaluation, (5) incomplete outcome data, (6) selective reporting, and (7) other sources of bias. Other disagreements were resolved through discussion with the third researcher (THT).

In the process of study quality assessment, the risk of bias evaluation for some studies was controversial. For the study by Van der houwen et al. [14], DYY believed that its randomization method met the criteria for low risk of bias, whereas FQ argued that while the study mentioned automatic randomization, it did not explicitly describe how the random order was generated, leaving the risk of bias unclear. Ultimately, we decided to invite a third researcher (THT) for evaluation, and based on their opinion, the random sequence generation of the study was rated as “unclear”.

Data were extracted and managed by one researcher (DYY) and verified by a second researcher (FQ). Any differences or inconsistencies between the two researchers were reached through discussion and consensus with a third researcher (THT). The extracted information included the author, year/location, research participants, number of participants, intervention measures for the experimental and control groups, frequency, duration, follow-up, measurement outcome indicators, and measurement tools.

Statistical analysis

RevMan 5.3 (The Nordic Cochrane Centre for The Cochrane Collaboration, Copenhagen, Denmark) was used to merge and analyze the extracted data. First, the heterogeneity of the studies was examined by the p-value and I². If the p-value was less than 0.1, the heterogeneity was considered to be high. Heterogeneity was divided into none (I² = 0–24.9%), low (I² = 25–49.9%), moderate (I² = 50–74.9%), and high (I² = 75–100%) [15]. A fixed-effects model was used if I² was less than 50% and the p-value is greater than 0.1. Second, when the I² was greater than 50% and the p-value was less than 0.1, we performed a sensitivity analysis by excluding trials that may bias the results, or by using the random effects model. Lastly, a p-value lower than 0.05 was considered statistically significant. We used the standard mean difference (SMD) and a 95% confidence interval (CI) to examine the intervention effect, which was evaluated by relevant scales after the intervention and follow-up. Funnel plots were used to examine potential publication bias when more than 10 studies were included for an outcome.

Literature search and study characteristics

Figure 1 illustrates the study selection process. We identified 2219 studies from seven electronic databases and one clinical registration center. Of these, 16 studies reported on the impact of web-based grief interventions on bereaved individuals. Table 1 listed the characteristics of each included study. This meta-analysis included 16 RCTs published between 2006 and 2023. The studies were conducted in Germany (seven studies), the Netherlands (four studies), USA (three studies), Switzerland (one study), and South Korea (one study). One study tested two interventions (exposure and behavioral activation) and compared them to the same control group [16]. For both grief and depression, Wagner et al. [17] used two different scales for assessment. Kaiser et al. [18] used two scales to assess grief of the bereaved. Sample sizes ranged from 26 to 455 participants, with a total of 726 participants in the intervention group and 771 participants in the control group.

Figure 2 presents a summary of the bias risk assessment using the Cochrane Collaboration Tool. Since blinding participants in intervention trials involving active participation is not feasible, and most trials require feedback from treatment staff or other researchers, all studies were judged to have a high risk of bias regarding participant and personnel blinding.

Anxiety

Seven studies [13, 18–23] examined the impact of a web-based grief intervention on anxiety. The pooled results showed no statistical significance (Z = 0.33, p = 0.74) and high heterogeneity (χ² = 39.32, I² = 85%), necessitating a sensitivity analysis (Fig. 3A). Since the study by Dominick et al. [19] used the State-Trait Anxiety Scale, where a higher score indicated a lower level of anxiety (opposite to the scales used in other studies), we excluded this study from the sensitivity analysis. After excluding Dominick et al.’s [19] study, the remaining trails were analyzed and using a fixed-effects model due to low heterogeneity (χ² = 2.55, I² = 0%). Anxiety in the web-based grief intervention group were significantly less severe than those in the control group (SMD = -0.37, 95% CI [-0.54, -0.20], p < 0.0001) (Fig. 3B). Three of the seven studies did not provide follow-up data; therefore, we analyzed the four studies [13, 21–23] that provided three-month follow-up data. A fixed effects model was used due to low heterogeneity test (χ² = 0.99, I² = 0%). Anxiety did not significantly improved at the three months follow-up compared to post-intervention (SMD = -0.02, 95% CI [-0.22, 0.19], p = 0.86) (Fig. 3C).

PTSD

Nine studies [16–18, 21–26] examined the impact of web-based grief interventions on PTSD. Data from the selected trials were analyzed using a fixed-effects model due to low heterogeneity (χ² = 16.99, I² = 47%). The pooled result indicated that web-based grief interventions had a significant effect on PTSD in bereaved individuals immediately after the intervention (SMD = -0.64, 95% CI [-0.78, -0.50], p < 0.00001) (Fig. 4A). For three studies [21–23] that provided follow-up data at three-months, a fixed-effects model was used (χ² = 3.53, I² = 43%). PTSD significantly improved at the three months follow-up compared to post-intervention (SMD = -0.23, 95% CI [-0.45, -0.01], p = 0.04) (Fig. 4B).

Depression

Fifteen studies [13, 14, 16–18, 20–29] focused on the impact of web-based grief interventions on depression. The pooled results showed statistically significant differences (Z = 5.22, p<0.00001) and high heterogeneity (χ² = 37.41, I² = 57%), necessitating a sensitivity analysis (Fig. 5A). After sensitivity analysis, the study by Lenfrink et al. [24] was excluded. The remaining studies were analyzed using a fixed-effects model (χ² = 29.08, I² = 48%). Web-based grief interventions had a significant effect on depression in bereaved individuals (SMD = -0.37, 95% CI [-0.47, -0.27], p < 0.00001) (Fig. 5B). Seven studies [13, 14, 17, 21–23, 27] that provided three-month follow-up data were analyzed. A fixed-effects models was used due to low heterogeneity (χ² = 6.46, I² = 0%). Depression showed significant improvement at the three-month follow-up compared to post-intervention (SMD = -0.15, 95% CI [-0.26, -0.04], p = 0.009) (Fig. 5C).

Grief

Thirteen studies [14, 16–18, 21–29] focused on the impact of web-based grief interventions on grief. The pooled results showed statistically significant differences (Z = 5.29, p<0.00001) and high heterogeneity (χ² = 45.27, I² = 67%), necessitating a sensitivity analysis (Fig. 6A). After sensitivity analysis, the studies by Lenfrink et al. [24], Litz et al. [23], Kaiser et al. [18], and Treml et al. [29] were excluded. A fixed effects model was used due to the low heterogeneity (χ² = 16.71, I² = 40%). Grief in the web-based grief intervention group was significantly less severe than in the control group (SMD = -0.30, 95% CI [-0.40, -0.19], p < 0.00001) (Fig. 6B). For six studies [14, 17, 21–23, 27] that provided three-month follow-up data, a fixed-effects model was used due to low heterogeneity (χ² = 5.92, I² = 0%). Grief was significantly less severe at three-month follow-up than post-intervention (SMD = -0.19, 95% CI [-0.31, -0.08], p = 0.001) (Fig. 6C).

Publication bias

We conducted a publication bias analysis of the literature on PTSD, grief, and depression as more than 10 articles measured these outcomes. Funnel plot analysis showed relatively symmetrical left and right plots, suggesting a small publication bias (Fig. 7). The results reported in the literature appeared relatively stable.

Competing interests

The authors declare no competing interests.

Ethical approval

Not applicable.

Consent to participate

Not applicable.

Consent to Publish

Not applicable.