Reminiscence and grief resolution through online journalling: mixed-methods analysis of the Living Memory Home bereavement cohort

TL;DR

Online journalling on the Living Memory Home website helped reduce symptoms of prolonged grief disorder by encouraging reflection on memories of the deceased.

Contribution

This study shows that guided online journalling, including reflection on negative memories, can reduce prolonged grief disorder symptoms.

Findings

PG-13-R scores decreased significantly after one week and one month of online journalling.

Reflection on negative traits of the deceased was negatively associated with PG-13-R scores.

Expression of regret and guilt was linked to further reductions in PGD symptoms over time.

Abstract

Prolonged grief disorder (PGD) is a chronic, impairing state of intense grief that is responsive to specialised intervention. Online journalling has the potential to reduce PGD symptoms.

To assess whether reminiscence via online journalling facilitates bereavement adjustment – specifically, to identify key themes in online grief journals and examine associations between types of reminiscence and changes in PGD symptom severity.

A cohort of 96 bereaved adults completed 7 days of online journalling on the Living Memory Home (LMH) bereavement website. Participants were recruited from clinics, bereavement support groups and a National Institutes of Health-funded, web-based platform. The Prolonged Grief Disorder-Revised scale (PG-13-R) was administered at baseline and at 1-week and 1-month follow-up.

Descriptive analysis revealed a reduction in PG-13-R scores from baseline to 1-week follow-up (mean difference −3.7, 95% CI: −4.9, −2.5, P < 0.001) and 1-month follow-up (mean difference −5.0, 95% CI: −6.4, −3.7. P < 0.001). Mixed-methods analysis revealed significant negative associations between reflection on negative traits of the deceased and PG-13-R score at all three time points, and between reflection on past experiences with the deceased and PG-13-R score at baseline. Expression of regret and guilt was significantly associated with reduction in PG-13-R scores from 1-week to 1-month follow-up.

Engagement in the LMH website demonstrated significant declines in PGD symptom severity after 1 week of online journalling and at 1-month follow-up. Guided reflection on memories of the deceased, and even on negative or emotionally challenging memories, shows potential for reducing symptoms of PGD.

Figures

Full text

Reminiscence, or the recollection of memories, is characteristic of mourning. In an adaptive mourning process, the bereaved person maintains connection with the deceased through reminiscence and, over time, integrates the reality of the loss with existing memories.^ 1 ^ Reminiscence is thus a topic of potential relevance to the design of interventions aimed at reducing symptoms of prolonged grief disorder (PGD), a chronically intense, impairing state of distressing grief.^ 2,3 ^

Background studies

Research shows that avoiding reminders and memories of the deceased is associated with poor bereavement outcomes.^ 4–6 ^ In addition, grief therapies that target cognitive reminder avoidance and encourage reminiscence of memories of the deceased person have been found to be more effective than supportive counselling, mindfulness-based cognitive therapy and antidepressant monotherapy in improving grief outcomes.^ 7–10 ^ Nonetheless, not all types of reminiscence may facilitate bereavement adjustment.

Reminiscence can be categorised as either functionally reflective – motivated by identity construction and problem-solving, or ruminative – motivated by bitterness revival and boredom reduction.^ 11,12 ^ Grief rumination is further characterised by negative focus on challenging emotions in grief (e.g. bitterness, regret, guilt), thoughts on the injustice or consequences of the loss and counterfactual thinking (i.e. dwelling on alternative scenarios such as preventing the death).^ 13 ^ Although both reflection and rumination engage with past-oriented thoughts, rumination is associated with poor adaptation to grief – manifesting as preoccupation, difficulty establishing trust relationships and excessive distress.^ 5,14,15 ^ It has thus been hypothesised that rumination is an avoidant coping style that provides a defensive escape from confronting the painful reality of the loss. Rumination is thought to hinder the direct reflection and processing of salient memories, an essential element of ‘grief work’ – mental processing of the loss.^ 16–18 ^ In contrast, reflection that functions as meaning reconstruction – the process of making sense of the loss and reconfiguring new relationships with the deceased, others and oneself – is associated with better bereavement outcomes.^ 14,19–21 ^

While the relationship between reminiscence functions and bereavement adjustment has been well described,^ 11,12,14,19–21 ^ studies have not examined, on a more granular level, whether the content or emotional valence of reminiscence is related to bereavement outcomes. It remains unclear, for example, whether reflection on negative, distressing memories is linked to worse outcomes than reflection on mostly positive ones. Here, we posit that what matters for adaptive mourning is not necessarily the emotional valence of memories but rather the act of engaging with memories – good or bad.

Aims

We aimed to examine the relationship between reminiscence and bereavement adjustment using free-text data from the Living Memory Home (LMH) prospective cohort study. The LMH is a digital platform hosted on the Cornell Center for Research on End-of-Life Care website for bereaved individuals as a resource for memorialising deceased family members and friends.^ 22 ^ With the increasingly common usage of digital media (digital photographs, videos, blogs) to maintain connections with deceased loved ones,^ 23,24 ^ the LMH was designed as a controlled, accessible website for bereaved individuals to maintain bonds and express emotions. This study aimed to assess associations between online journal writing and bereavement adjustment (i.e. PGD symptom severity over time). A mixed-methods approach was used to identify key themes in LMH journal entries, and to explore associations between types of reminiscence – thematically coded by content and emotional valence – and changes in PGD symptom severity before and after LMH engagement.

Method

Recruitment

Adults living in New York State and bereaved by the death of a close person (e.g. family member, friend) were recruited through flyers at bereavement support groups and relevant clinics and websites at Weill Cornell Medicine (e.g. Wright Center on Aging, Department of Geriatrics and Palliative Medicine, cancer clinic bulletin boards, Acute Care for the Elderly Unit, the Medical Intensive Care Unit, Cornell Center for Research on End-of-Life Care website), by direct mailing to people referred from these clinics and via advertisements on ResearchMatch.org – an online platform sponsored by the National Institutes of Health (NIH) to connect people with NIH-funded research studies. Potential participants completed the validated Prolonged Grief Disorder-Revised (PG-13-R) grief severity scale,^ 2 ^ Columbia Suicide Severity Rating Scale^ 25 ^ and an established sociodemographic and loss characteristics questionnaire.^ 26 ^ Exclusion criteria included not having internet access or knowledge of how to operate a computer or smartphone; not being fluent in English; belonging to a specific vulnerable population (prisoners, pregnant persons, children); and endorsement of suicide attempt or suicidal ideation with plan and/or intent within the past 12 months (for risk management reasons). Verbal consent was witnessed via telephone and formally recorded by a research staff member.

Study procedure

Eligible participants were given a secure personal account and a brief tutorial on using the LMH website. The LMH journal writing feature offered a selection of 44 open-ended writing prompts that encouraged reflection, expression of grief-related thoughts and feelings and imagined dialogues with the deceased, expanded from the therapeutic prompts by Neimeyer’s ‘Correspondence with the Deceased’ in Techniques in Grief Therapy. ^ 27 ^ Writing prompts elicited specific memories (e.g. ‘One of my most treasured memories of you is …’) and emotions in grief (e.g. ‘Now that you’re gone I feel …’). The full list of writing prompts is provided in Supplementary Appendix 2 available at https://doi.org/10.1192/bjo.2025.10882. Participants completed 7 consecutive days of journal writing in the LMH for at least 20 min per day. Participants chose writing prompts at will, and there was no limit on the length or number of entries that could be entered. Data were collected from 29 April 2021 to 30 April 2023.

Ethical standards

The authors assert that all procedures contributing to this work comply with the ethical standards of the relevant national and institutional committees on human experimentation, and with the Helsinki Declaration of 1975 as revised in 2013. All procedures involving human subjects/patients were approved by the Weill Cornell Medicine Institutional Review Board (Approval no. 1810019629).

Measures

Participants self-reported sociodemographic and loss-related characteristics. PGD symptom severity was measured by PG-13-R score at baseline (prior to LMH login), at 1 week (on completion of the 7-day LMH journal writing protocol) and at 1 month (30 days after the first recorded journal entry). The PG-13-R scale ranges from 10 to 50, with scores of 30 or above indicating a clinically significant threshold of grief severity that may serve as a diagnostic criterion for PGD.^ 2 ^ We used change in this score to capture bereavement adjustment.

Thematic analysis

We used thematic analysis to capture key themes and codes in free-text data from LMH journal entries, for use as variables in statistical analysis. Details regarding data extraction, coding team, codebook development, coding protocol, final coding and code frequency are provided in Supplementary Appendix 1.

Derived variables of reminiscence

A team approach was used to derive variables based on codes from the thematic analysis, used as proxies for key constructs in relation to the process of reminiscence. From the codebook we identified codes that captured reflective reminiscence and grief rumination.

Statistical analysis

We quantified reminiscence variables by their code frequency, the total number of text entries expressing that code and participant. Spearman partial correlation coefficients were used to estimate associations between code frequency and two sets of outcomes: (a) PG-13-R scores at baseline, 1 week and 1 month, (b) and change in PG-13-R scores from baseline to 1 week, baseline to 1 month and 1 week to 1 month. Correlation coefficients controlled for the total number of journal entries per participant over the course of LMH use. We identified potential confounders by estimating correlations between participant characteristics and outcomes (PG-13-R scores and code frequencies). Four participants had missing or incomplete PG-13-R assessments; for missing values we imputed their mean values.

Results

Participant characteristics

Of 137 participants, 96 completed the study, 41 withdrew or were lost to follow-up and none were excluded. Our analysis included 2390 journal entries from 96 participants. Characteristics of study completers are summarised in Table 1. The mean age was 44.5 years (s.d. = 16.5), ranging from 20 to 89 years. Regarding demographic factors, 77 (80.2%) were female, 73 (76.8%) were heterosexual, 71 (74.0%) were White, 53 (55.2%) had attained a graduate or professional degree and 52 (54.2%) lived in New York City. Regarding loss-related characteristics, 42 (43.8%) had lost a parent, 15 (15.6%) a grandparent, 14 (14.6%) a spouse or partner, 7 (7.3%) a child, 4 (4.2%) a sibling and 11 (11.5%) a friend or other non-family other. Mean time since loss was 58.8 months (s.d. = 85.7), with a median of 24 months, ranging from 2 months to 38 years. In 55 (57.3%) cases the death was rated by the participant as unexpected, including causes of natural sudden death (e.g. heart attack), COVID-19, homicide, suicide, accidental overdose, fatal accident (e.g. car accident), natural disaster (e.g. earthquake) and others.

Table 1. Sample demographic and loss-related characteristicsDemographic or loss-related characteristic^ a ^

n (%) (N = 96)Age (years), mean (s.d.) [range]44.5 (16.5) [20–89]Gender^ b ^  Male18 (19.0) Female77 (80.2)Sexual orientation^ c ^  Heterosexual73 (76.8) Lesbian6 (6.3) Gay4 (4.2) Bisexual3 (3.2) Queer7 (7.4) Other2 (2.1)Race American Indian or Alaska Native1 (1.0) Asian7 (7.3) African American or Black12 (12.5) White71 (74.0) Other5 (5.2)Ethnicity Hispanic, Latino or of Spanish origin7 (7.3) Not Hispanic, Latino or of Spanish origin89 (92.7)Highest education level High school8 (8.3) College35 (36.5) Graduate/professional53 (55.2)Location New York City (NYC)52 (54.2) New York State (outside of NYC’s five boroughs)44 (45.8) Time since loss (months),^ d ^ mean (s.d.) [range]58.8 (85.7) [2–456]Relationship of deceased to participant Spouse/partner14 (14.6) Child7 (7.3) Parent42 (43.8) Grandparent15 (15.6) Sibling4 (4.2) Friend3 (3.1) Other11 (11.5)Cause of death Natural anticipated death (e.g. terminal illness)41 (42.7) Natural sudden death (e.g. heart attack)22 (22.9) COVID-195 (5.2) Homicide1 (1.0) Suicide4 (4.2) Accidental overdose2 (2.1) Fatal accident (e.g. car accident)7 (7.3) Natural disaster (e.g. earthquake)1 (1.0) Other13 (13.5)a.Sample characteristics were self-reported.b.Missing data from one participant.c.Missing data from one participant.d.Missing data from 14 participants.

We collected sociodemographic data for 36 of the 41 participants who withdrew or were lost to follow-up. This group had a mean age of 53.7 years (s.d. = 18.1). Of those 36 participants who withdrew or were lost to follow-up, 27 (75.0%) were female, 32 (88.9%) were heterosexual, 17 (47.2%) had attained a graduate or professional degree and 19 (52.8%) lived in New York City. Regarding race and ethnicity, 21 (58.3%) were White, 9 (25.0%) were African American or Black and 7 (19.4%) were Hispanic/Latino. Regarding loss-related characteristics, 13 (36.1%) had lost a parent, 2 (5.6%) a grandparent, 5 (13.9%) a spouse or partner, 4 (11.1%) a child, 3 (8.3%) a sibling and 9 (25.0%) a friend or other non-family other. In 18 (50.0%) cases, the death was rated by the participant as unexpected.

Thematic analysis

We identified 22 codes (subthemes) and three main themes: (a) reflection, (b) continued bonds and (c) psychosocial adjustment. Reflection captured reminiscence on memories of the deceased, as well as present- and future-oriented musings. Continued bonds described ways of maintaining a connection with the deceased, such as imagining dialogues, emulating the traits of the deceased or memorialising the deceased through significant objects, places or actions. Psychosocial adjustment captured emotional reactions to the loss, coping mechanisms and changes in one’s life and relationships following the loss. Organisation, definitions and textual examples of all codes are provided in Supplementary Table 1.

Derived variables of reminiscence

We identified five codes as conceptually meaningful variables for examination of reminiscence in the LMH journals: experiences with the deceased, traits of the deceased, death circumstances, bitterness, and regret and guilt. Code frequencies are provided in Supplementary Table 2.

Experiences, traits and death circumstances were selected as measures of reflective reminiscence: past-oriented reflections on specific memories of the deceased and the loss event. Experiences captured memories of experiences with the deceased, further categorised by emotional valence. Positive, negative and neutral/unclear experiences appeared in 338, 129 and 51 entries, respectively, out of the total 2390 entries. Traits captured reflection on specific characteristics of the deceased, further categorised by positive or negative emotional valence. Positive traits appeared more frequently than negative, in 541 and 134 entries, respectively. We did not identify traits with neutral/unclear valence. Death circumstances, coded in 241 entries, captured reflection on the death event and the circumstances preceding the death (e.g. a hospital course leading up to the death) or following the death (e.g. making funeral arrangements).

Bitterness and regret and guilt were selected to explore the negative emotional reactions associated with grief rumination.^ 13 ^ Bitterness captured negative emotions towards the deceased, other people or one’s circumstances, often using exaggerated language to express anger, disappointment or resentment. Regret and guilt captured repentance, remorse or the wish that different actions had been taken in the past. Often, regret and guilt coincided with expressions of pity, empathy and ways in which the bereaved individual would have changed their behaviour in hindsight (e.g. wishing they had spent more time with the deceased before the death). Bitterness appeared in 159 entries, regret and guilt in 188.

Changes in PGD symptom severity

Table 2 shows the main outcomes of PGD symptom severity: mean PG-13-R scores at baseline, 1 week and 1 month and mean changes in PG-13-R scores from baseline to 1 week and 1 month. At baseline, PG-13-R scores were below the PGD diagnostic threshold of 30 for 84 (87.5%) of the 96 participants, with an overall mean score of 24.9 (s.d. = 8.2) at baseline, 21.2 (7.0) at 1-week and 19.8 (6.8) at 1-month follow-up. Results revealed significant reductions in PG-13-R scores from baseline values at 1-week (mean difference [s.d.], −3.7 [6.0], P < 0.001) and at 1-month follow-up (mean difference [s.d.], −5.0 [6.9], P < 0.001).

Table 2. Grief severity (Prolonged Grief Disorder-Revised scale, PG-13-R) at baseline, 1 week, 1 month and change from baselineTime pointPG-13-R score, mean (s.d.)Change in PG-13-R scorefrom baseline, meandifference (s.d.)Baseline24.9 (8.2)Not applicable1 week21.2 (7.0)−3.7 (6.0)***1 month19.8 (6.8)−5.0 (6.9)******P < 0.001 (via paired-sample t-test).

Confounders such as education level, location (New York City versus outside the city) and relationship of deceased (spouse or child versus other kinship) were identified based on their significant bivariate associations with PG-13-R scores and code frequency (Table 3).

Table 3. Spearman correlations between participant characteristics and outcomes, grief severity (Prolonged Grief Disorder-Revised scale, PG-13-R) and code frequencyOutcomeEducationLocation (New York City)Relationship of deceased (spouse or child) r

P

r

P

r

P PG-13-R at baseline−0.2640.0100.1180.2600.2090.041PG-13-R at 1 week−0.1820.080−0.0010.9940.2320.023PG-13-R at 1 month−0.0730.488−0.0310.7680.3100.002**Change in PG-13-R,baseline to 1 week0.1440.167−0.2230.0320.0230.826Change in PG-13-R,baseline to 1 month0.2410.020*−0.1590.1270.0810.432Change in PG-13-R,1 week to 1 month0.1870.072−0.0080.9390.0860.407Experiences with the deceased (positive)−0.2190.0360.2120.043−0.0290.784Experiences with the deceased (negative)0.1450.168−0.0840.4280.0410.692Experiences with the deceased (neutral/unclear)−0.0590.5730.0280.7910.0500.633Experiences with the deceased (all valences)−0.1360.1950.1690.108−0.0180.865Traits of the deceased (positive)0.0110.913−0.0080.937−0.1600.121Traits of the deceased (negative)0.2260.030*−0.0150.889−0.2310.024Traits of the deceased (all valences)0.0810.4440.0250.810−0.1970.056Death circumstances−0.0500.6360.0090.9360.0050.965Regret and guilt0.0650.537−0.0830.4310.0420.689Bitterness0.1740.0980.1230.2430.0350.738**P < 0.05; **P < 0.01.

Correlational results: associations between reminiscence variables and PGD symptom severity scores

We examined ten variables of reminiscence, derived from five thematic codes, with the emotional valences of codes treated as separate variables: experiences with the deceased (positive, negative, neutral/unclear and all valences); traits of the deceased (positive, negative and all valences); death circumstances; regret and guilt; and bitterness. We adjusted estimates for correlations between codes and PG-13-R scores to control for the total number of journal entries per participant and potential confounders identified above. Table 4 shows significant negative correlations between traits (negative) and PG-13-R scores at each time point: baseline (r = −0.302, P = 0.004), 1 week (r = −0.276, P = 0.008) and 1 month (r = −0.218, P = 0.037). There was a significant negative correlation between experiences (all valences) and PG-13-R score at baseline (r = −0.211, P = 0.044), but not at 1 week or 1 month. Negative correlations between experiences (neutral/unclear) and change in PG-13-R scores were found from 1 week to 1 month (r = −0.292, P = 0.005), and between regret and guilt and change in PG-13-R score from 1 week to 1 month (r = −0.219, P = 0.036).

Table 4. Spearman partial correlations between grief severity (Prolonged Grief Disorder-Revised scale, PG-13-R, scores) and code frequencyCodePG-13-Rat baselinePG-13-Rat 1 weekPG-13-Rat 1 monthChange inPG-13-R,baseline to1 weekChange inPG-13-R,baseline to1 monthChange inPG-13-R,1 week to1 month r

P

r

P

r

P

r

P

r

P

r

P Experiences with the deceased (neutral/unclear)−0.0640.542−0.0260.806−0.1860.0760.0140.895−0.1140.281−0.2920.005Experiences with the deceased (all valences)−0.2110.044*−0.1220.245−0.1530.1460.1060.3150.0500.635−0.0930.377Traits of the deceased (negative)−0.3020.004−0.2760.008**−0.2180.0370.0880.4040.1940.0640.1490.155Regret and guilt−0.0170.8700.0040.970−0.0980.355−0.0200.851−0.1390.188−0.2190.036All correlations control for total number of journal entries; correlations also control for identified confounders: participant education level, living in New York City and relationship of deceased being a spouse or child.*P < 0.05; **P < 0.01.

Discussion

Remarkably, results demonstrated significant declines in PGD symptom severity long after the death of a significant other and after only 1 week of journalling in the LMH website, with effects sustained at 1-month follow-up. Given that the mean time since loss was 4 years and 9 months, these findings suggest that online journalling may have a therapeutic effect in reducing grief intensity, even some years after the death. Although the LMH was not developed intentionally to be psychotherapeutic, it nevertheless elicited narrative reflections on loss-related memories and emotions, and these reminiscences appear to have had some therapeutic benefits for those engaged in the journal. Consistent with other grief studies,^ 15,28,29 ^ we demonstrated that online daily writing provides a forum in which bereaved individuals can confront memories of the loss. Prompting bereaved individuals to confront grief-related thoughts and feelings was not shown to worsen PGD symptoms; rather, online journalling appeared to facilitate bereavement adjustment, suggesting promising new directions for the development of widely accessible grief therapies.

Associations between reminiscence and PGD symptom severity

Results indicated that greater PGD symptom severity at baseline was associated with less reflection on past experiences and negative traits of the deceased in the LMH. Reflection on negative traits of the deceased was also associated with lower grief severity following LMH engagement. Findings suggest that LMH users who did not reflect on memories of the deceased – specifically past experiences or negative traits – had more severe PGD symptoms, aligning with the hypothesis that avoiding specific memories results in worse bereavement outcomes.^ 4-6 ^ Thus, results complement prior grief research showing that reflection on memories, even negative or emotionally distressing ones, facilitates bereavement adjustment when the goal is to make meaning of the loss.^ 19,20,29,30 ^ Overall, our findings lend support to the potential efficacy of guided online reflection to process grief and reduce symptoms of PGD (i.e. intense yearning, disbelief, numbness, pain and difficulty in moving on with life).

Analyses revealed an unexpected significant association between expressions of regret and guilt and reduction in PGD symptom severity 1 month after the first day of journal writing, contrary to our initial hypothesis that ruminating on one’s own distressing emotions in grief would confer worse bereavement outcomes. A possible explanation is that the LMH was a productive space for bereaved individuals to confront challenging emotions through expressive writing – the act of writing about deep, emotional experiences – which is shown to have psychotherapeutic outcomes.^ 31 ^ This aligns with contemporary grief models, which predict that periodically confronting grief-related emotions is integral to coping with grief,^ 17 ^ whereas an inability, or lack of opportunity, to share grief-related emotions and experiences characterises an isolating, persistent liminal state of grief that impedes adjustment.^ 32 ^ The LMH offered an alternative to sharing uncomfortable emotions with therapists, family or friends and perhaps provided an outlet for those without strong social support. Regarding expressions of guilt specifically, it is theorised that guilt can be psychologically adaptive if it promotes empathy for others and reparative changes in behaviour.^ 33 ^ However, specific forms of guilt may inhibit bereavement adjustment,^ 34 ^ particularly if it perpetuates rumination on counterfactual scenarios (i.e. preventing the death).^ 4 ^ It is possible that expressing guilt through online journalling promoted clarity through reflection rather than further rumination or circular thinking; however, our analysis did not index different forms of guilt, so further research is needed to unpack these results.

We also found a significant association between reflection on past experiences with neutral valence and reduction in PGD symptom severity from 1 week to 1 month. The ability to reflect on loss-related memories with less emotional bias (overly positive or overly negative) may be characteristic of grief adaptation and could be a sign of emotion regulation. PGD is characterised by symptoms of emotional pain and numbness, and research indicates that emotion regulation difficulties contribute to PGD symptoms.^ 2,35–37 ^ Similar studies have shown that reminiscence therapy – a psychosocial intervention delivered primarily to older adults and those at the end of life – reduces depression and anxiety symptoms and improves cognitive emotion regulation.^ 38–40 ^ Our findings might similarly represent a link between reminiscence and improved emotion regulation in grief.

Our results highlight how reminiscence and expressive writing may facilitate bereavement adjustment. Specifically, reflection on specific memories of the deceased and expression of regret and guilt were associated with low PGD symptom severity and improvement over time. Our study emphasises an important distinction: reflection on negative memories and emotions is not necessarily maladaptive and should not be conflated with grief rumination. Rather, confronting negative memories and working through distressing emotions probably paves the way for coping with grief. Our study advocates for narrative disclosure as a key component of bereavement support. The LMH may serve as a model for the development of more accessible online bereavement resources that centre reminiscence and expressive writing.

Strengths and limitations

This study is novel in its application of mixed methods. Our collection of qualitative and quantitative data had the benefits of ease, accessibility and anonymity in using an online journal bereavement resource. We used a validated measure of grief symptom severity to explore any changes over the course of engagement in the LMH, and specifically the type of content entered. This was facilitated using a robust team approach to thematic analysis, which allowed us to identify and measure distinct patterns in LMH journals and any correlations with grief severity. The validity of our thematic analysis was supported by acceptable interrater reliability for most codes (Supplementary Table 3). However, thematic analysis of digital free text is limited, lacking the opportunity to ask follow-up or clarifying questions as in a semi-structured interview.

Another limitation of this study is the absence of a control group. The LMH journal was not designed specifically as a therapeutic intervention, and thus we refrain from drawing conclusions about its potential therapeutic benefits. Comparing an online expressive writing intervention with structured reflective prompts to a control condition lacking targeted guidance would enable a more rigorous assessment of reminiscence as a mechanism for bereavement adjustment. Furthermore, our study offers limited observation of PGD among LMH users, because most participants did not meet the diagnostic criteria for PGD based on either grief severity or the duration of time since the loss. Our sample captures bereavement experiences in the community, and grief generally varies in both severity and duration. Notably, the symptoms of grief (e.g. yearning for the deceased) were present in all subjects, as indicated by their responses to the PG-13-R, even if they did not meet the diagnostic criteria for PGD. The sample also encompasses a wide range of time since loss, from 2 months to 38 years, but is skewed towards those more recently bereaved, with a mean of approximately 5 years and a median of 2 years. This broad range introduces heterogeneity in grief trajectories, because participants may be at varying points in their mourning process. As a result, the study is limited in its ability to tailor recommendations to individuals experiencing prolonged grief. Future trials should evaluate a more targeted version of the LMH against a control intervention among individuals meeting diagnostic criteria for PGD.

Study participants were self-selected, predominantly White, female and highly educated, limiting study generalisability. In addition, 41 of the 137 participants withdrew from the study or were lost to follow-up. Regarding baseline characteristics, this group was older on average and had a higher percentage of Black and Hispanic/Latino participants than the group who completed the study. Further analysis may be warranted to evaluate the potential differential impact and acceptability of online bereavement interventions across age and race. It is possible that participants withdrew from the study due to time constraints and the level of effort required. This study required active participation in LMH writing for 20 min per day for 7 consecutive days. While more subjects might have completed the study if it had required less activity, this amount of engagement was associated with significant reductions in grief symptom severity, arguing in favour of this effort level. Future research should consider evaluation of the effects associated with lower ‘doses’ (i.e. time commitments).

Notwithstanding these limitations, this mixed-methods study of journal entries from the LMH bereavement website demonstrated that online journal writing exercises show promise in facilitating bereavement adjustment in a community-based, non-clinical sample. This study underscores the potential therapeutic benefits of reminiscence and expressive writing activities for bereaved adults. Future research may adapt the LMH into an exposure-based therapeutic intervention with more guided prompts for reflecting on memories of the deceased and processing emotional reactions to the loss.

Supporting information

Lee et al. supplementary materialLee et al. supplementary material