Research Collection – The Course of Grief


Photo of a tree re-growing after being felled in a storm

I took this photo 3 months after my son's death, hoping to be able to do what that tree had done – re-establish a thriving life after a devastating blow. When that blow had landed, one of the first things I had wanted to know was how long it usually takes for extreme distress to start easing. Of course, grief, like any personal experience, is very individual: no one could tell me how long it was going to be until my despair lessened – or even if it would. But I knew, from decades of developing evidence-based information for suffering people, that research should be able to provide better answers than I was reading in information for grievers.

I wrote recently at The Atlantic about my frustration with the conflicting information I was seeing on the course of grief. This research collection is the basis for the statements I made there.

This post is organized into a pyramid with several layers. First, there's a short summary of what I found. That's followed by a more detailed overview of the research results. And later I explain how I went about compiling the research collection, and include brief details of the studies. 

So here's the answer to the question I had at my grief's ground zero – although it's based on research that was mostly done in the US and Europe, with some in Asia, Australia, and Canada. On average, people's acute grief after losing someone close to them begins easing within a few weeks. For a lot of people – around half – grief isn't overwhelming, at least from a month after the loss. Most people are noticeably less distressed by 3 months, and many more by 6 months or so.

The evidence supports this take from a medical expert in grief:

"In so far as we never stop feeling sad that loved ones are gone, or stop missing them, grief is permanent. However, the acute, all-consuming intensity usually moderates over time, as grief becomes deeper, less intrusive, and integrated into our lives."

Medically, that's called uncomplicated grief. "Over time" doesn't mean each day will be better than the day before – it's a path that can have twists and turns. And even after grief isn't dominating our lives any more, there could still be tough times – around anniversaries for example. For most bereaved people, though, a year or so after the loss, their wellbeing and life satisfaction may be similar, or close to similar, to their pre-loss level.

However, many people are still suffering intensely more than a year after their loss. That's what's often called complicated or prolonged grief. The overall rate for complicated grief at 6 months may be roughly 10% of bereaved people, including a small percentage of people who were doing better earlier, but whose distress worsened. The rate of complicated grief is lower for people in some circumstances, though, more common when people experience loss of a spouse, and very common after a particularly shocking bereavement, like losing someone to an accidental death, homicide, or suicide; or losing their children of any age. Social support and other circumstances are critical, too: for example, bereavement can lead to major changes in people's living situation and income that make everything tougher.

"Grief doesn't have a timetable," people often say. And it's true: it doesn't. Even when there appear to be landmarks from this research – a month after a loss, say, or a year – these aren't boundaries, but typical times grief researchers have checked in on people. There are some patterns in people's experiences, though. Life after major loss usually does get easier. Relief, for most of us, starts to seep in. And it can be sooner than we might think is possible.

Hilda Bastian
October 2022

Digging into the research results

Between reviews and individual additional studies, this collection is based on 103 studies, with over 38,000 bereaved people. As I mentioned, they are mostly from the US and Europe, and though quite a few are from Asia and some are from Australia and Canada, there are none from Africa, the Middle East, or Latin America. That doesn't mean none exist. One reason for the absence is that for one category of research, I relied on a review that's several years old, and studies have been happening in other parts of the world since then.

While my process for searching for studies on the course of grief was fairly extensive, it was still limited, and most of my hunting was targeted to particular types of studies. I wanted to find studies that were more reliable because they could better deal with the many biases that could skew results on the course of grief.

For example, consider studies based on people volunteering for interviews or questionnaires across time. Who's invited and who accepts, or who finds out about a study and signs up, can be different from those who aren't included in ways that skew the picture of grief we get. That can happen, too, if people who are doing better – or particularly worse – don't return for follow-ups down the line, and if the interviews or questionnaires themselves are having an impact on how people feel.

So I was particularly interested in finding studies that could track people in whole communities, or close to it. When these not only covered a whole community, but also compared the wellbeing of bereaved people with the non-bereaved, my confidence in the patterns of experiences grew – especially when the data didn't rely on occasional interviews or questionnaires. Grappling with a death doesn't mean we're insulated from other sources of misery, so comparison is important for perspective: it helps us get a perspective on how much distress is added by bereavement.

It's even better if the tracking started before the bereavement. That can offer additional insight not only into the impact of a sudden loss, but also of an impending loss.

Only a few studies met all, or most, of those criteria. They, too, showed a pattern of improvement from early on, which increases my confidence in the conclusions you can draw from the body of research in this post.

This collection is made up of 3 categories of studies:

  1. Studies on the rate of complicated grief (usually measuring at a single time point);
  2. Studies measuring people's wellbeing at more than 1 time point, using statistical modelling to tease out the different "trajectories" within the averages (people who are never very distressed, deeply distressed then improved, and so on);
  3. Longitudinal studies (following people up across some time) other than trajectory modelling studies, that reported on at least 2 time points after the loss, and met some additional scientific quality criteria.
The rate of complicated grief varies quite a lot from study to study. An older systematic review of studies assessing the rates found in studies that assessed mainly people bereaved after non-violent deaths. The researchers concluded it was about 10% at 6 months. Rates were much higher after sudden, violent deaths, though – in some fortunately uncommon circumstances, grief is complicated for most of the people who experience it. Complicated grief also corresponds to some of the trajectories in the studies in my second category. The rates of the trajectories analogous to complicated grief in those studies ranged from 6% to 26%.

The trajectory at the opposite end of the spectrum – people who didn't have high levels of distress throughout the studies (typically measured from a month post-loss) – was much more common: mostly between 40% and 60%. That trajectory, along with more severe distress that improves across time, accounted for most bereaved people. And averages of signs of distress improving across time was the major pattern in the studies in my third category of studies.

What about the studies that met all, or almost all, the criteria I mentioned earlier for more reliability and perspective? There were 4 of these studies, all from the third category. They were based on analyzing healthcare use for close to 10,000 bereaved people in total, compared with over 240,000 non-bereaved people – and 3 of them included data from before the loss to at least 2 years afterwards. All but 1 of those studies were of parents who lost children, but they corroborated the rough patterns emerging from the other types of studies in this collection. Here's an overview of them, to give you an idea what that kind of study looks like:

  • Taiwan's national health insurance covers the whole population. Medical service usage for major depression was tracked for parents of the children over 1 and under 12 who died between 2002 and 2010, matching with parents of 4 randomly selected living children for each lost child. (Chen 2021)
  • Medical records of bereaved widows relying on the US military health service were matched with those of wives whose service-member husbands didn't die. Data on mental health diagnoses and service usage was analyzed from 1 year before the loss to 2 years afterwards. (Cozza 2019)
  • Data from national registries in Finland for a random 20% of the population were used to analyze what anxiety and depression prescriptions parents of children up to 14 years of age filled. Data for parents from 4 years before losing a child to 4 years after were compared to all the non-bereaved parents. (Rostila 2018)
  • Data from several healthcare usage sources related to mental health in a Canadian province were linked for all mothers who lost a child under 1 year of age across 4 years, matched with similar mothers who hadn't lost a child. Data were analyzed from 4 years before to 4 years after the loss. (Wall-Wieler 2018)

Technical details about how I found and chose the studies in this collection follow below. Before that section starts, though, there are links to the 3 categories of studies just below if you want to jump straight into the details of the studies. If you do, check out these abbreviations that I use in those summaries:
  • CI: confidence interval. This is an indicator of the range of statistical uncertainty around a result. It will look like this: 59.6% (CI 51-67). If the distance between the first and second numbers is small (51 and 67 in this example), then the associated uncertainty is fairly small. In this example, 55-62 would have been a very small difference, and 31-87 would have been massive.
  • RR: relative risk. An RR of 1 would mean the risk was similar, an RR of 2 is double the risk, a negative RR is a lowering of risk.
Here are the links to jump to each category of studies:
  1. Measuring the extent of high levels of grief at 6 months or later;
  2. Modelling trajectories of grief; and
  3. Following people for at least 2 points in time after loss, and meeting some criteria for scientific quality.

Methods for assembling this collection

Ideally, there would be up-to-date, high quality systematic reviews of studies to rely on. Systematic reviews are research projects using very thorough processes to try to reduce bias in finding, selecting, and evaluating studies and their results. However, this collection mostly relies on individual studies. When there were multiple papers from a single data source, I used only 1 report of data for a category of study (more on issues of overlap below).

The 3 categories were mentioned earlier in this post. Category 1 is assessing the rate of complicated grief – studies could be cross-sectional (assessing people at one time point), or longitudinal (assessing at more than one time point). Category 2 is a particular style of longitudinal study that models major trajectories of grief, exploring the different patterns within the averages. And Category 3 is longitudinal studies other than "trajectory" studies.

I searched for systematic reviews at Epistemonikos, a database specializing in this type of research, as well as Google Scholar. I found 4 relevant systematic reviews that I judged to be recent enough and of adequate scientific quality. Of those, 3 were for Category 1 (the rate of complicated grief), and 1 was for Category 2 (trajectories of grief). There were none for Category 3 (longitudinal studies, other than the type in Category 2).

I checked the included studies in the systematic reviews for studies that could also be included in Category 3, except for a systematic review on sudden and violent loss from Category 1. The review reported didn't tabulate the detail I needed to do this. However, I did check for overlap (whether studies after sudden and violent loss from my other categories were included in that review).

For Category 1, I relied only on the systematic reviews and did not search for additional studies. I searched for trajectory (Category 2) and longitudinal studies (Category 3) at Google Scholar. To find more trajectory and longitudinal studies, I followed citations (both of key papers and within them). For Category 2, which was a specific type of longitudinal study, I included all I found.

For Category 3, I also ran this search in PubMed (the major biomedical database) from 1995 until August 13, 2022:

grie*[tiab] OR bereave*[tiab] AND longitudinal[tiab]

I searched from 1995 to focus on studies where both community experiences and research methods were more likely to reflect current experience/practice, including ways of studying grief. (I chose 1995 because that was the year the Inventory of Complicated Grief was published, which became commonly used in grief research.)

The further criteria for including studies in Category 3 aimed to limit the group to papers reporting data on the course of grief and more likely to contribute to a more reliable picture of grief in their community:
  • At least 200 bereaved adults;
  • Outcomes relevant to chart the course of grief gathered for at least 2 time points post-loss;
  • Either based on a whole population (regional, national, or healthcare) or a representative sample (or likely to be reasonably close to that), and/or a non-bereaved control or comparison group;
  • Reasonably low attrition rate for a grief study (for example, a response rate for the first post-loss measurement that was close to 40% or higher for a defined population).
As some of these criteria were somewhat vague, I kept studies that were borderline, but included them separately as a "lower quality" group. Because studies with multiple time points in the early weeks after bereavement were so rare, I included all the studies I found that included data on acute grief, whether they met the other criteria or not – as long as they were after 1945, and included 2 time points up to 3 months post-loss. (I've separated those out into ones meeting my quality criteria, and those that didn't.)

Because most of this collection isn't based on systematic reviews with extensive searches and thorough screening of studies, I am sure to have missed studies – though I think it's unlikely that my searches and trawl of cited studies missed ones large enough to dramatically affect the overall picture.

There is some overlap between the studies in this collection. There's a mapping for this below this post. However, I didn't include mapping for the studies in the systematic review on complicated grief after sudden and violent loss (Category 1). There was no overlap in studies, although the review did include a study that might have overlapping participants with Kristensen 2020 (Category 2): there were 67 bereaved people in a study of the same terrorist attack in the review, and 129 people in Kristensen 2020. 

Category 1: High levels of grief at 6 months or later

This category includes 3 systematic reviews, including 13,723 participants (after deducting those included in 2 reviews). I relied on the first for "non-violent" deaths (Lundorff 2017), even though with just over 6 years since its search date, it is a little out-of-date. However, the quality is high, and the scope of the review is relatively broad for people at low or average risk of prolonged grief disorder.

The second is another high quality systematic review, and it isn't out of date (Djelantik 2020). The focus here was people at very high risk of prolonged grief disorder, after losses from sudden and violent deaths.

The third review focuses only on older people, and it is not as high quality as the other couple (for example, there was no published protocol for the review – a protocol sets out the "rules" established ahead of time for finding, evaluating, and analyzing results).

Details of the 3 systematic reviews on prolonged grief:

Marie Lundorff and colleagues, 2017. Protocol for this systematic review: Lundorff 2016.

This is a systematic review including 14 studies found as of May 2016, mostly from Australia, China, Europe, Japan, and the USA, including 8,035 participants. The authors were only looking for studies predominantly in bereavement after "non-violent" deaths – that's age or illness, not for example accidents (including accidental overdoses), suicide, or homicide.

There was a lot of variation in these studies. The authors concluded the rate of prolonged grief disorder at six months was about 10% (CI 7-14). 

Manik Djelantik and colleagues, 2020. Protocol for this systematic review: Boelen 2017.

The search for studies for this systematic review reached to December 2017, and resulted in 25 included studies with 4,774 participants altogether: 10 studies from Europe (including 3 from Kosovo), 7 from Australasia, 6 from the USA, and 1 from Africa. The focus was on sudden and violent deaths, caused by accident, homicide, suicide, disaster, war, and terrorism.

There was a lot of variation between the studies, so the authors caution about the high level of uncertainty around the combined result: 49% (CI 34-65). All but 2 of the studies were a year or longer after the death, some decades later. Rates of complicated grief were lower in the studies where it was longer since the loss. Rates were also lower in deaths after disasters.

Pia Thiemann and colleagues, 2021 [PDF].

This search for studies for this systematic review was between January 2009 and November 2019. The review includes 9 studies, mostly from Europe, plus 2 from China and 1 from the USA: 3 of them were also included in the systematic review above, accounting for most of the participants (1,797 out of the total of 2,711) (Lundorff 2017). The authors looked at the rate in people aged 65 or older, excluding studies on more traumatic causes of death: the authors’ best estimate was about 9%. That's not a summary of research they found: it's from a single study the authors considered the most reliable. That study involves a random sample of a country’s population (Germany), estimating the proportion of bereaved people in the country who met the criteria for elevated grief at the time of the study. It’s not about a specific time point after the loss.

Category 2: Modelling the trajectory of grief

This is a particular type of study of grief, which use different statistical methods to model patterns of grief experiences from longitudinal studies – they typically suggest there are 4 or 5 groups that all but a few percent of people fall into. I've included a systematic review with 6 studies (up to June 2015) (Su Ching Kuo and colleagues, 2017), plus 16 extra studies from my own search for grief trajectories on Google Scholar, together with those I found via citations and on my search for longitudinal studies (Category 3 below). Between them, these studies include 10,047 people, with nearly 4,200 from the USA. Just over 1,000 of the people were from Asia. The rest were from northern and western Europe.

The 6 studies in the Kuo systematic review included 1,102 bereaved people altogether. All the studies were done in the USA. The authors concluded that based on measurements of depressive symptoms, people's grief trajectories were:
  • Low depressive symptoms throughout: 54.2%
  • Severe depressive symptoms at first with quick recovery: 8.8%
  • Distress lasting 7 to 12 months, then returning to pre-bereavement levels: 7.7%
  • Prolonged depressive symptoms, which eventually improved: 19.4%
  • Chronic depressive symptoms, which did not improve during the study: 9.9%
There have been so many of this type of study since that review, that there are far more studies and participants that aren't in the Kuo systematic review than are in it. There's a lot of variation in them, including how likely they are to be representative of people in their communities, but I didn't filter them. Just under half of the studies have fewer than 200 participants.

The 16 studies that aren't in the Kuo review don't all chart the same outcomes, or the same 5 trajectories as listed above. However, with few exceptions, they include groups that roughly correspond to the first and last trajectories.

The first group in common is those people who had low or mild distress at each measurement. That was 54% in the Kuo review: in most of the other studies that was from 40% to 60%. (Almost all were in the range of a third to two-thirds.)

The second group in common is those people who had severe distress at each time point. That very roughly corresponds with complicated grief. In category 1 above, systematic reviews estimated the rate of complicated grief at roughly 10% for people bereaved mainly by non-violent loss, and 49% for people who suffered sudden, violent loss – though there was a lot more variation in studies than there was among  the non-violent loss studies. The individual trajectory studies here included people with a mixture of risk levels. The rate for the more chronically severe outcome measures ranged from 6% to 26%.

Some of these studies (4) included a group of people who got worse over time, ranging from 4% to 10%.

Details of the 16 extra studies:

Kate Mary Bennett and colleagues (2019studied 403 people in Switzerland widowed after marriages of 15 years or longer, first after a maximum of 5 years post-loss, with a follow-up 2 years later. These authors explored their data differently to the other trajectory studies. They reported relatively little movement between groups at the second assessment, and it was usually to move towards improved well-being. According to their preferred data model, by the second assessment people fit into 3 profiles:
  • Resilient: 53%
  • Coping: 39%
  • Vulnerable: 8%

George Bonanno and colleagues (2002 [PDF]) studied 185 older widowed people in the USA, and mapped these trajectories pre-loss, and then at 6 and18 months post-loss:
  • Low depression pre-loss and at 6 and 18 months: 51%
  • Depressed pre-loss, improved at 6 and 18 months: 11%
  • Low depression pre-loss, depression at 6 but not 18 months: 12%
  • Low depression pre-loss, depression at both 6 and 18 months: 17%
  • Depression pre-loss that continued at 6 and 18 months: 9%
(Note: This is the CLOC study [Changing Lives of Older Couples]. There's at least a further analysis by Bonanno of these data (in 2004). I haven't included other analyses based on longer-term CLOC data, where the drop-off in participation at the later time point was very high, such as Boerner 2005, Holland 2013, and Lotterman 2014 [PDF].)

George Bonanno and colleague (2019 [PDF]) studied 282 people in the USA widowed when they were under 65, at 3, 14, and 25 months post-loss. They mapped these trajectories:
  • Low grief at 3 months declining slightly over time ("resilience"): 58%
  • Moderate grief at 3 months that declined steadily but not completely by 25 months ("moderate-improving"): 18%
  • High level of grief at 3 months that improved dramatically, reaching the same low level as the "resilience" group by 25 months ("acute-recovery"): 7%
  • Moderate grief at 3 months without major change by 25 months ("moderate-stable"): 13%
  • High level of grief at 3 months that grew markedly worse over time ("prolonged-worsening"): 4%

Chuqian Chen and colleagues (2019studied 198 older people in a longitudinal study in China who were widowed between 2011 and 2013, with data collected in 2011, 2013, and 2015. They mapped these trajectories:
  • Low depression across time ("resilience"): 55%
  • High pre-loss depression and constant decrease ("depressed-improved"): 12%
  • Low pre-loss depression and constant increase ("chronic grief"): 24%
  • High depression across time ("chronic depression"): 10%

Manik Djelantik and colleagues (2017) studied symptoms of prolonged grief disorder in 166 bereaved people in the Netherlands, and mapped these trajectories (between 6 and 18 months post-loss):
  • Low symptoms throughout: 26%
  • Slightly decreasing moderate symptoms: 33%
  • Persistent moderate symptoms: 35%
  • Persistent high symptoms: 6%

Manik Djelantik and colleagues (2022 [PDF]) studied 259 people in the Netherlands who had been bereaved less than a year at the outset, with follow-ups every 6 weeks up to 10 times over 15 months (which was up to 27 months post-loss). They mapped these trajectories:
  • Low and slightly decreasing grief ("resilient"): 66%
  • High level of grief decreasing over time (usually between 6 and 18 months) ("acute recovery"): 8%
  • High level of grief slightly increasing over time ("chronic"): 25%

Pål Kristensen and colleagues (2020studied 129 parents and siblings bereaved by the 2011 mass shooting on Utøya Island, Norway, at 18, 28, and 40 months post-loss. They mapped these trajectories:
  • Moderate, decreasing grief: 23%
  • High level of grief, slowly decreasing: 64%
  • High level of chronic grief: 13%

Lonneke Lenferink and colleagues (2018) studied 172 people from the Netherlands who lost a spouse when flight MH17 was downed by a missile in 2014, and mapped these trajectories up to 42 months post-loss:
  • For grief: 82% mild and stable across time, and 18% prolonged grief
  • For depression: 86% mild and stable, 8% high and decreasing across time ("recovered"), 6% high and remaining high ("chronic")
  • For PTSD: 85% mild and stable, 4% "recovered", 10% "chronic"

Marie Lundorff and colleagues (2020 [PDF]) mapped these trajectories up to 11 months for the Aarhus Bereavement Study in Denmark, including 857 bereaved spouses:
  • Low symptoms throughout: 64.4%
  • Moderate and stable symptoms: 20.4%
  • Elevated symptoms decreasing over time: 8.4%
  • Continuing elevated symptoms (prolonged grief): 6.8%

Fiona MacCallum and colleagues (2015) analyzed rates of depression in 2,512 people who lost a child or spouse from a longitudinal study of older adults in the USA, and mapped these trajectories:
  • Little or no depression: 68.2%
  • Onset of depression following loss: 13.2%
  • High pre-loss depression that improved following loss: 11.2%
  • Pre-existing and continuing depression: 7.4%

Anthony Mancini and colleagues (2015asked 116 bereaved people (in the USA) to self-identify which of a set of 4 trajectories best represented their experience up to 2 years post-loss. The results:
  • Initially high distress, then quickly stabilizing ("resilient"): 28%
  • Initially high distress, then gradual improvement ("gradual recovery"): 35%
  • High level of distress, with some improvement in the second year ("prolonged grieving"): 27%
  • Continuing high level of distress ("continuous distress"): 10%

Mette Kjærgaard Nielsen and colleagues (2019studied 1,735 people bereaved after loss caused by illness in Denmark. They mapped 2 types of trajectories from pre-loss to 3 years post-loss: 5 groups of people who lost a spouse (from 1,138 participants), and 4 for others (from 597 participants).

The first trajectory, for spouses, was:
  • Low grief: 34%
  • Moderate, decreasing grief: 30%
  • High, decreasing grief: 20%
  • High grief: 7%
  • Late high grief: 10%
The second trajectory, for others, was:
  • Low grief: 45%
  • Moderate, decreasing grief: 31%
  • High, decreasing grief: 16%
  • High grief: 8%

Kirsten Smith and colleague (2020) studied 264 bereaved people at 3 time points in the first 12 to 18 months post-loss (in the UK), and mapped these trajectories up to 12 months:
  • Modest grief symptoms, reducing to low over time: 41%
  • High level of initial grief, reducing close to low by 12 months: 13%
  • High level of initial grief, with some gradual decrease: 38%
  • High level of initial grief, with some increase: 8%

Josefin Sveen and colleagues (2018) studied 132 Swedish tourists who had lost a family member during the 2004 tsunami in Asia, at 3 time points up to 6 years after the disaster. They mapped the following trajectories:
  • Moderately low level of grief at 6 months and steady decline over time ("resilient"): 41%
  • Initially high level of grief, reducing over time ("recovering"): 48%
  • High level of grief at each time point: 11%

Ágnes Szabó and colleagues (2020studied 686 older people participating in the Longitudinal Aging Study Amsterdam whose spouse died between 1992 and 2015, analyzing data on depressive symptoms and loneliness from 12 years before the bereavement to 12 years after. Interviews were every 3 years. They mapped these trajectories (based on depression):
  • Depressive symptoms low before bereavement, rose a little at the first post-loss assessment, then stabilized pre-bereavement level: 16%
  • Depressive symptoms low 12 years pre-loss, but rose gradually, peaked in the first post-loss assessment, then declined, but never reached the earlier low level: 54%
  • Depressive symptoms just below the level of clinical depression 12 years pre-loss, increased gradually to clinical depression, which peaked at first post-loss assessment, then declined somewhat but remained at clinically depressed level: 31%

Fur-Hsing Wen and colleagues (2020studied 849 caregivers of people who died of cancer across the first 2 years of bereavement (in Taiwan), and mapped trajectories for prolonged grief disorder and major depression:

Prolonged grief:
  • Low levels throughout: 37%
  • High level initially, decreasing to low levels at 1 to 3 months: 43%
  • Very high level initially, decreasing to low levels by 7 to 8 months: 10%
  • Very high level initially, decreasing steadily but not completely by 2 years: 4%
  • High level initially, decreasing gradually to the end of 12 months, but then increasing steadily from about 18 months: 7%
Major depression:
  • Low levels throughout: 32%
  • High level initially, decreasing to low levels at 3 to 6 months: 44%
  • Very high level initially, decreasing to low levels by 12 months: 3%
  • Very high level initially, decreasing steadily but not completely by 2 years: 17%
  • Very high level with minimal improvement across 2 years: 4%

Category 3: Following people for at least 2 points in time after bereavement

There are 3 groups in this category of longitudinal studies (that aren't "grief trajectory" studies). The first is a group of higher quality studies with outcomes measured for at least 2 time points after loss – all from 1995, and it's the scientifically strongest group of studies in this collection. There are 13 studies, including 15,585 bereaved people: 5 of these studies have control or comparison groups of non-bereaved people – totalling 242,905 non-bereaved people. There are 5 from Europe (France, Finland, Netherlands, Sweden); 4 from Asia (Japan, Korea, Taiwan); and 4 from North America (Canada, USA).

The second is a group of 5 studies of lower quality ("borderline") studies, also from 1995. (I kept these in, because some of the quality criteria I was using were vague.) There are 1,369 bereaved people in these studies, with 2 from the Netherlands, and 1 each from Australia, Taiwan, and USA.

The third is a group of studies that track acute grief. These studies include at least 2 time points in the first 3 months after the loss. There are 7 studies in this group, 3 of which are also included above, as they track beyond 3 months as well (2 higher quality, and 1 lower "borderline" quality). The other 4 studies didn't meet the criteria for the category – they each have below 200 participants, and 1 is before 1995 (1989).  However, as studies of acute early grief are so uncommon, I kept all I found.

There are 1,154 bereaved people in these 7 studies tracking acute grief: 4 are from Asia (3 from Taiwan, 1 from India), and there's 1 each from Australia, Germany, and USA. 

Details of the 13 higher quality longitudinal studies:

Hsin-Hung Chen and colleagues, 2021.

Place: Taiwan (national).

Recruitment methods: National health insurance data (covers whole population). Medical service usage for major depression of parents of all children aged over 1 and up to 12 who died between 2002 and 2010. Each child was age- and gender-matched with 4 randomly selected living children.

Number eligible: 5,881 children who died, and 23,524 matched living children.

Number participating: At least 1 parent was identified for 91% of the children who died: 2,987 mothers and 4,258 fathers (6,945 bereaved parents). In the comparison group, 96% were identified: 16,512 mothers and 17,753 fathers (34,265 non-bereaved parents).

Study methods: Medical records linkage, measuring major depression reports after the death.

Nearly 5% of bereaved mothers and over 2% of bereaved fathers had treatment for major depression recorded in the 3 years after their child's death. The overall risk for mothers (hazard ratio) was 4.71 (CI 3-7) and the risk for fathers was 1.93 (CI 1-3).

The risk for mothers dropped steeply in the first 4 months, then more slowly after that, reaching a level a little higher than non-bereaved mothers close to 14 months post-loss. It remained at that somewhat higher level to the end of the study (36 months).

The risk for fathers dropped steeply across the first 8 months, then fluctuated to the end of the 22nd month, where the level approached that of non-bereaved fathers before fluctuating again. At 36 months, the risk was similar to non-bereaved fathers.

Stephen Cozza and colleagues, 2019. (A further paper on this study: Joscelyn Fischer and colleagues, 2021.)

Place: USA (national).

Recruitment methods: Medical records in the military health service, matched with demographic data of service members who died between 2001 and 2006.

Number eligible: Of the 8,363 service members who died, there were 3,627 military-dependent spouses. After excluding the small number of widowers and national guard and reserve spouses who may not have used the military healthcare services exclusively, there were 3,355 military-dependent widows. Only those in that group who had used military healthcare services at least once at multiple time points of the study period were then included, to decrease the chances that the person used civilian healthcare services. There were 1,521 eligible widows. The authors attempted to match them by age, service member rank, and deployment status during the period with a control group of non-bereaved military wives. (Matches could not be found for 146 widows.)

Number participating: 1,375 widows (90%) matched with 1,375 non-bereaved wives.

Study methods: Medical records analyzed for relevant mental health diagnoses and associated use of health care.

Time points and participation:

  • 3 years of healthcare data, from 1 year pre-loss to 2 years post-loss: all participants.

The rates of mental health problems remained fairly constant for the non-bereaved control group across all 3 years, and it was similar to that for the wives the year before their husbands died. In the year after the loss, though, the widows' rate of depression rose from 12% to 29%: that dropped a little after 1 year, down to 24%. At that time, there was no diagnostic coding for prolonged grief disorder or complicated grief. However, the rate of "adjustment disorder" appears to have stood in: that rose from 4% to 21%, and then reduced to 11% after a year.

Benjamin Domingue and colleagues, 2017.

Place: USA (national).

Recruitment methods: The participants whose spouses died during the nationally representative longitudinal study of people aged over 50 and their spouses, the US Health and Retirement Study, in which depression was regularly measured.

Number eligible: Not reported.

Number participating: 1,647.

Study methods: Modeling of depression scores across time. (The focus of the paper is on genetics.)

Time points and participation:

  • Continuous data from 5 years before the loss to 5 years afterwards.
The model showed a sharp increase in depression shortly after the spouse's death, which reduced within 2 years to a level that remained to 5 years: that was a bit higher than it had been in the years before the spouse's death. (More specific data on this not included in this study.)

Celia Harris and colleagues, 2021.

Place: Aarhus, Denmark.

Recruitment methods: County-wide study, where all people aged 65-80 who lost their spouse in 2006 were contacted 2 months after the loss. They were also asked to invite their adult children to participate.

Number eligible: Not reported for this paper.

Number participating: 1,187.

Study methods: Questionnaires.

Time points and participation:

  • 2 months post-loss: 1,187.
  • 18 months post-loss: 892.

This is another paper from the Aarhus Bereavement Study. The level of grief was higher for people that lost their spouse than it was for people who lost a parent: their grief decreased at a similar rate, though. The mean score on grief symptoms was 27 for spouses and 22 for those bereaved of their parent at 2 months, and 22 for spouses and 18 for the parentally bereaved at 18 months.

Nancy Kentish-Barnes and colleagues, 2015.

Place: France.

Recruitment methods: Designated surrogates (eg spouse or adult offspring) of everyone who died after at least 48 hours in one of 41 intensive care units and who had visited at least once and could speak French, were approached. 

Number eligible: 647 eligible: 104 "missed" (couldn't be contacted?) and 68 declined.

Number participating: 475 enrolled (73%).

Study methods: Questionnaires.

Time points and participation:

  • Day 21 post-loss (about end-of-life management): 430.
  • 3 months post-loss: 386.
  • 6 months post-loss: 282.
  • 12 months post-loss: 215.

The first assessment did not relate to grieving or wellbeing. At 3 months, 36% were depressed and 52% had elevated anxiety scores, but later levels weren't reported. The rate of significant PTSD symptoms reduced from 45% at 3 months to 36% at 12 months. The rate of complicated grief at 6 months was 52% and at 12 months it was 53% – but these weren't the same people: 19% of those with complicated grief at 6 months no longer did at 12 months, while 21% who did not have complicated grief at 6 months did at 12 months.

Ulrika Kreicbergs and colleagues, 2004

Place: Sweden (national).

Recruitment methods: All parents who had lost a child to cancer from 1992-1997 were contacted. A control group of non-bereaved parents was recruited by randomly selecting living children from the population register.

Number eligible: 561 bereaved parents, 659 non-bereaved parents.

Number participating: 449 bereaved parents (80%), 457 non-bereaved parents (69%).

Study methods: Posted questionnaires.

Time points and participation:

  • 4 to 6 years post-loss
  • 7 to 9 years post-loss

This is a study of psychological wellbeing, not grief specifically. Relative risks (RR) of anxiety and depression, for example, were assessed.

The psychological wellbeing of bereaved parents was lower than non-bereaved parents at 4 to 6 years post-loss, but wellbeing and quality of life was similar at 7 to 9 years. For example, the RR for high scores for depression-related symptoms was 1.9 at 4-6 years (CI 1-3), and 1.0 at 7-9 years. (CI 0.6–1.6). Every measure of wellbeing didn't show such a high relative risk, but all the measures had improved at the 7-9 year mark. Mothers' wellbeing was lower than fathers'.

To a question about having come to terms with the loss, 32% answered "no, not at all" or "yes, a little" at 4-6 years; this had reduced to 20% at 7-9 years.

The risk of psychological morbidity was higher if the child was aged over 9 when they died, than if they were younger (relative risk (RR) of anxiety: 1.5 [CI 1-2]; of depression: 1.6 [CI 1-2]).

Takeshi Nakagawa and colleagues, 2021.

Place: Japan (national).

Recruitment methods: Nationally representative longitudinal survey of adults aged 60 and over.

Number eligible: 492 were married at first assessment, then were widowed and had not remarried.

Number participating: 481 (12 excluded because of missing data).

Study methods: Home interviews, with year and month of spouse death reported, but the gap of 3-4 years between interviews means this study has rough estimates.

Time points:

  • 7 interviews between 1987 and 2006, 3-4 years apart. 

Time to or from widowhood was calculated for each interview's time point, and a model was developed for categorizing whether the interview represented "transition to widowhood" (within a year either side of the loss), or post-widowhood. Life satisfaction dropped in the transition to widowhood, and stabilized at the lower level around a year after the loss. The authors hypothesize the reason for this may be cultural approaches to bereavement that are not adaptive in the longer term, and/or related to the older age of the group. However, there was only a single outcome here – life satisfaction (measured by level of (dis)agreement with 3 questions). Unless other studies and other outcomes find a similar pattern, there's too much uncertainty to draw a conclusion. The size of the study may also affect results. For example, the pattern wasn't the same for widowed people who had one of their children living with them: life satisfaction improved for them at a year post-loss.

Maja O'Connor and colleagues, 2015.

Place: Aarhus county, Denmark.

Recruitment methods: 4-year follow-up of county-wide study, where all people aged 65-80 who lost their spouse in 2006 were contacted 2 months after the loss. (The same study as Harris above.)

Number eligible: 839.

Number participating: 330 at the start (41%). 

Study methods: Questionnaires. One on prolonged grief began at the study's second time point, 6 months post-loss.

Time points and participation:

  • 6 months post-loss: 237.
  • 13 months post-loss: 198.
  • 18 months post-loss: 192.
  • 48 months post-loss: 213 (90% of those participating at 6 months).

The authors did not apply a cut-off score for prolonged grief in this study. Mean scores on the 2 measures collected across time points were:

  • Prolonged grief average: 29.54 at 6 months, 29.48 at 13 months, 28.64 at 18 months, 26.62 at 48 months; and
  • Post-traumatic stress average: 32.79 at 6 months, 31.53 at 13 months, 30.3 at 18 months, 24.98 at 48 months.

Mikael Rostila and colleagues, 2018.

Place: Finland (national).

Recruitment methods: Data linkage from national registries for a random 20% of the population with at least 1 child aged 0-14 at the end of 2000.

Number eligible: All with available data from a randomly selected group – 902. The 205,206 non-bereaved parents in the sample were used for comparison.

Number participating: All.

Study methods: Analysis of purchase dates for several psychotropic drugs.

Time points:

  • Data on filled prescriptions from 4 years before the child's death to 4 years after (plus filled prescriptions for non-bereaved parents).

Between 20-25% of bereaved mothers and 10-15% of bereaved fathers used antidepressants or anxiety medications around a year after the death, compared to around 10% of non-bereaved mothers and 5% of non-bereaved fathers. The percentage peaked around a year, presumably indicating medical intervention for complicated grief towards a year post-loss. Usage then dropped, until 4 years later it was a few percentage points higher than non-bereaved parents (from the graphs, it appears to be about 2% or so higher). The increase began for mothers before the death where the child was dying of disease, and after the death for other causes.

Siew Tzuh Tang and colleagues, 2021.

Place: Taiwan.

Recruitment methods: Families of patients in palliative care in 2 intensive care units with about 1/15 of all the national ICU capacity, with all eligible families approached. 

Number eligible: 389.

Number participating: 326 (84%).

Study methods: Questionnaires.

Time points and participation:

  • 1 month post-loss: 278 (85% of 326).
  • 3 months post-loss: 256.
  • 6 months post-loss: 211. (Study report prepared when 38 participants were not yet 6 months post-loss.)
The proportion of people with severe distress reduced at each time point:
  • Severe depression symptoms: 42% at 1 month, 21% at 3 months, and 11% at 6 months.
  • Severe anxiety symptoms: 22% at 1 month, 9% at 3 months, and 3% at 6 months.

Elizabeth Wall-Wieler and colleagues, 2018.

 Manitoba province, Canada.

Recruitment methods: All mothers in the province who had lost 1 child at under 1 year of age in the 4 years after April 1, 1999. Matched with mothers who had never lost a child, of similar age, with children born at similar times, and similar living circumstances. Registry data for births and deaths, linked to medical, hospital, pharmaceutical, and Census data.

Number eligible: 534 bereaved mothers, 1,602 non-bereaved mothers.

Number participating: All.

Study methods: Data relevant to mental health outcomes analyzed.

Time points and participation:

  • Each 6 month-period, beginning 4 years before the loss and ending at 4 years after the loss.

Measured in various ways, the mental health of bereaved mothers 4 years before the births and from about a year after the death was similar to the mothers who hadn't lost a baby. 

The highest risk was depression. The relative risk (RR) for bereaved mothers was 4.94 in the 6 months after the death (CI 4-6) (that is, nearly 5 times as high as for non-bereaved mothers with a child of the same age). Between 6 months and a year, the RR was 1.57 (CI 1-2), and from after 1 year it was similar.

Heesoo Yoon and colleagues, 2022.

Place: Korea (national).

Recruitment methods: Korean Longitudinal Study of Ageing, representative sample of people aged 45 or more surveyed every 2 years since 2006. Study analyzes data from 685 people who lost a spouse between 2006 and 2018.

Number eligible: 685 (551 women, 134 men).

Number participating: All.

Study methods: Questionnaires. Data model analyzing measuring depression and participation in social activities.

The rate of depression increased in the year after the loss of a spouse, and then declined to pre-loss levels within about 3 years. However, the increase in rate of depression was greater for men, and for this relatively small group of men, the rate did not return to pre-widowhood levels in 4 years or more. Women's social participation remained fairly constant, however, there was a major ongoing reduction in men's social activities.

JoAnne Youngblut and colleagues, 2013 [PDF].

 South Florida, USA.

Recruitment methods: Families of babies and children (up to age 18) who had died in intensive care (other than twin or other multiple pregnancy, child in foster care, suspected child abuse, parent dying with the child eg in a motor vehicle accident or childbirth).

Number eligible: 752 families affected, but 32 not eligible and 372 families could not be found, leaving 348 eligible families that could be contacted.

Number participating: 188 of 348 families (54%): 249 individuals, with 176 mothers and 73 fathers (including 55 couples).

Study methods: Interview and questionnaires, in English or Spanish.

Time points and participation:

  • 1 month post-loss: not reported.
  • 3 months post-loss: not reported.
  • 6 months post-loss: not reported.
  • 13 months post-loss: 92% of families.

Months 1 to 3 are discussed in Category 3b (including reduction in rate of depression). Depression and PTSD were high at 1 month, though half the mothers with PTSD no longer had it at 13 months (from 71% to 35%). At 1 month, 31% of mothers had no depression, increasing at each time point to 65% by 13 months. For 22%, depression was moderate/severe at 13 months (and 13% were mildly depressed). For fathers, of whom 44% were not depressed at 1 month, this also increased at each time point, up to 76% at 13 months.

Details of the 5 lower quality ("borderline") longitudinal studies:

Paul Boelen, 2015.

Place: Netherlands.

Recruitment methods: A group via healthcare workers and a group from internet advertising.

Number eligible: Not known. Healthcare workers handed out 1,128 questionnaires: 492 returned (44%). 490 questionnaires were sent to people responding to the internet call, with 260 returned (53%). Some were not eligible as they were already more than 12 months post-loss.

Number participating: 230. 

Study methods: Questionnaires.

Time points and participation:

  • Within first year post-loss: 230. (Average time post-loss: 6.6 months)
  • 6 months later: 159.
  • 15 months later: 136.

At first measurement, scores on the grief scale were high enough for prolonged grief disorder for 67%, decreasing to 47%, then 42%. That score was >25. The mean scores for grief and depression decreased at each time point:

  • Grief: 32 at first measurement, then 27 at 6 months later, and 24 at the third.
  • Depression: 39 at first measurement, 35 at the second, and 33 at the third.

Kairi Kõlves and colleagues, 2020.

Place: Queensland (state), Australia.

Recruitment methods: All on the state's police Suicide Register who had indicated on routine police form that they were willing to be contacted for research: about 60% of all deaths by suicide. Plus a comparison group of people bereaved by sudden death, contacted via coronial inquiry register for recent closed cases.

Number eligible: 276 people bereaved by suicide. 233 people bereaved by a sudden death.

Number participating: 142 bereaved by suicide, 63 bereaved by a sudden death (205 in total).

Study methods: Interviews.

Time points and participation:

  • Approximately 6 months post-loss: 142 suicide bereaved, 63 sudden death bereaved.
  • 12 months post-loss: 128 suicide bereaved, 59 sudden death bereaved.
  • 24 months post-loss: 110 suicide bereaved, 48 sudden death bereaved.

The authors concluded that, "Irrespective of the cause of the bereavement, family members tended to show significant reductions over time in somatic reactions, searching for an explanation, feelings of stigmatization and rejection, and symptoms of depression, anxiety and stress."

The relevant data is shown in graphs without specific endpoints shown, including mean scores for the depression scale. The score used for depression for suicide bereaved is 5 to 6 for mild depression. Mean scores were:

  • Suicide bereaved: just over 5 at 6 months (CI roughly under 4.5 to just over 6), 4.5 at 12 months, and a little under 4 at 24 months.
  • Sudden death bereaved: at 6 months, 4.5 for sudden death bereaved (CI over 3 to just over 6), 4 at 12 months, and a little under 4 at 24 months.

Su-Ching Kuo and colleagues, 2017.

Place: Taiwan.

Recruitment methods: Referral by oncologists of caregivers of people terminally ill with cancer.

Number eligible: Not known. 

Number participating: 263 at post-loss, of 344 with data collected before the death.

Study methods: Questionnaires.

Time points and participation:

  • 1 month post-loss: 263
  • 3 months post-loss: 246
  • 6 months post-loss: 240
  • 13 months post-loss: 224
  • 18 months post-loss: 203
  • 24 months post-loss: 190
The authors reported on the percentage of people with severe depressive symptoms at each time point. This measure dropped at each time point, from 73% at 1 month, to 38% at 6 months, 23% at 13 months, with smaller decreases after that: 18% at 18 months, and 15% at 24 months.

The drop between 1 month and 3 months was from 73% to 50%. The authors discuss the very high rate early, as potentially because a high proportion would not have been prepared (as prognostic information is not routinely shared), Taiwanese cultural grieving practices (with 3 ceremonies in the first 3 months), and the group registered very high caregiver burdens – the authors suggest that might indicate they had over-stretched their personal resources before the death.

Paul Maciejewski and colleagues, 2007.

Place: Connecticut, USA.

Recruitment methods: 2 groups: smaller group recruited in various ways, majority group from a fairly eligible register of widows. 

Number eligible: Total not known. 575 identified and invited.

Number participating: 317 (55%) who were more likely to be male and older than the others invited. Data presented in this study was for 233 (41%) who had a loss from natural causes, never met the criteria for complicated grief, and had at least 1 complete set of data. They were older and more likely to be white.

Study methods: Interviews. Questions partly developed for this study.

Time points and participation:

  • Between 1 and 6 months post-loss: responses to 5 items ranged from 143 to 174 people
  • 6 to 12 months post-loss: responses to 5 items ranged from 209 to 213 people
  • 1 to 2 years post-loss: responses to all 5 items from 205 people

This is the Yale Bereavement Study: there are multiple papers published on it. The aim of this one was to see whether grieving followed a 5 stages model (not the Kübler-Ross one): disbelief, yearning, anger, depression, and acceptance, on a scale of 1 to 5. Mean scores on all 4 measures of distress reduced across time. Acceptance was high at the start and kept increasing, starting with a mean of 4.1 and ending with 4.7. Anger was the lowest. The proportion of people who experienced each of these was not reported. The authors included calculations which they said support the 5-stages running in sequence, but they don't provide enough of the basic data: what they do provide, also supports the opposite conclusion.

Leonik Wijngaards-de Meij and colleagues, 2005 [PDF].

Place: Netherlands.

Recruitment methods: Parent couples who lost a child (up to 18 years) and published an obituary in local Utrecht and national newspapers were contacted.

Number eligible: 463 couples.

Number participating: 219 couples – 438 parents (47%).

Study methods: Interviews and questionnaires.

Time points and participation:

  • 6 months post-loss: 438.
  • 13 months post-loss: 378.
  • 20 months post-loss: 361.

Measures of grief and depression reduced somewhat at each time point. Mean scores (out of 100) were:
  • Grief: 45 at 6 months, 43 at 13 months, and 42 at 20 months.
  • Depression: 24 at 6 months, 22 at 13 months, and 21 at 20 months.

Details of the 7 studies with data on acute grief:

The 2 high quality studies

Siew Tzuh Tang and colleagues, 2021(Also included in Category 3a.) 

Place: Taiwan.

Recruitment methods: Families of patients in palliative care in 2 intensive care units with about 1/15 of all the national ICU capacity, with all eligible families approached. 

Number eligible: 389.

Number participating: 326 (84%).

Study methods: Questionnaires.

Time points and participation:

  • 1 month post-loss: 278 (85% of 326).
  • 3 months post-loss: 256.
  • 6 months post-loss: 211. (Study report prepared when 38 participants were not yet 6 months post-loss.)
The reported proportion of people with severe distress reduced at each time point:
  • Severe depression symptoms: 42% at 1 month, 21% at 3 months, and 11% at 6 months.
  • Severe anxiety symptoms: 22% at 1 month, 9% at 3 months, and 3% at 6 months.

JoAnne Youngblut and colleagues, 2013 [PDF]. (Also included in Category 3a.)

 South Florida, USA.

Recruitment methods: Families of babies and children (up to age 18) who had died in intensive care (other than twin or other multiple pregnancy, child in foster care, suspected child abuse, parent dying with the child eg in a motor vehicle accident or childbirth).

Number eligible: 752 families affected, but 32 not eligible and 372 families could not be found, leaving 348 eligible families that could be contacted.

Number participating: 188 of 348 families (54%): 249 individuals, with 176 mothers and 73 fathers (including 55 couples).

Study methods: Interview and questionnaires, in English or Spanish.

At 1 month after the loss, 31% of mothers were not depressed, which increased to 47% at 3 months. For fathers, 44% were not depressed at 1 month, rising to 59% at 3 months. At 1 month after the loss, 48% of mothers and 40% of fathers had moderate/severe depression; at 3 months this had reduced to 35% of mothers and 24% of fathers. 

For those who were employed before the loss, half of the mothers and 74% of the fathers had returned to work by the end of the first month. It's not reported how often that was because it was necessary, or a sign of improved wellbeing.

The 5 lower quality studies

Su-Ching Kuo and colleagues, 2017(Also included in Category 3a.)

Place: Taiwan.

Recruitment methods: Referral by oncologists of caregivers of people terminally ill with cancer.

Number eligible: Not known. 

Number participating: 263 at post-loss, of 344 with data collected before the death.

Study methods: Questionnaires.

Time points and participation:

  • 1 month post-loss: 263
  • 3 months post-loss: 246
  • 6 months post-loss: 240
  • 13 months post-loss: 224
  • 18 months post-loss: 203
  • 24 months post-loss: 19
There were 263 people at 1 month post-loss, and 246 at 3 months. The authors reported on the percentage of people with severe depressive symptoms. The drop between 1 month and 3 months was from 73% to 50%. The authors discuss the very high rate early, as potentially because a high proportion would not have been prepared (as prognostic information is not routinely shared), Taiwanese cultural grieving practices (with 3 ceremonies in the first 3 months), and the group registered very high caregiver burdens – the authors suggest that might indicate they had over-stretched their personal resources before the death.

Sing-Fang Ling and colleagues, 2013.

Place: Taoyuan, Taiwan.
Recruitment methods: Primary caregivers of people with terminal cancer at 1 hospital invited to participate.
Number eligible: 388 (by time of data collection).
Number participating: 186.
Study methods: Questionnaires.
Time points and participation:
  • Roughly 2 weeks before the death: presumably 186.
  • 1 month post-loss: 176.
  • 3 months post-loss: 152.
  • 6 months post-loss: 143.
  • 13 months post-loss: 119.
The mean score for depressive symptoms was above the rate for clinical depression at 1 month post-loss, but below it thereafter. The rate of clinical depression was 60% shortly before the death and 59% at 1 month post-loss, 47% at 3 months, and 39% at 13 months. The younger the person who died, the higher the risk of depression.

Warwick Middleton and colleagues1996 [PDF].

Place: Brisbane, Australia.

Recruitment methods: Identified via hospital personnel or official state register. Continued until there were roughly equal participants in 3 groups: bereaved spouses under 70, adults who lost a parent, parents who lost a child aged between 1 and 18.

Number eligible: Not reported.

Number participating: 53 bereaved spouses – 44 of whom completed all interviews (47%), 52 parent-bereaved adults – 40 of whom completed all interviews (44%), and 53 bereaved parents – 36 of whom completed all of them (32%).

Study methods: Interviews. A grief measurement ("Core Bereavement Items" [PDF]) that rated 76 items, which were compiled into a single CBI score (with items given varying weight).

Time points – analyses limited to those who completed all 4:

  • 1 month after loss
  • 10 weeks post-loss
  • 7 months post-loss
  • 13 months post-loss 

People rated 76 items which were compiled into a single score. The mean score reduced at each time point, for each of the 3 groups. Between 17% and 25% had low scores across the all time points, and about 9% were described as having chronic grief that did not reduce across the time. The mean score reduced at each time point, however, for each of the 3 groups.

The first 2 time points were at 1 month, and then 10 weeks, with the mean scores reducing at both time points:

  • Bereaved spouses: from 26 to 23;
  • Adults who lost a parent: from 22 to 18;
  • Parents who lost a child: from 30 to 27.

Carl Scheidt and colleagues, 2012.

Place: Freiburg, Germany.

Recruitment methods: All adult mothers who were admitted to the study hospital after a miscarriage later than 12 weeks, pregnancy termination for medical reasons, or perinatal death (death before, during, or after birth) between January 2006 and May 2008 were eligible.

Number eligible: 78.

Number participating: 33 (42%).

Study methods: Interviews.

Time points and participation:

  • Within 2 weeks after the loss: 33 mothers.
  • 4 weeks later: 33 mothers.
  • 4 months later: 31 mothers.
  • 9 months later: 31 mothers.

Most of the women reported that the pregnancy had been planned (84%). All the women had either had a stillbirth (7/33), medically-indicated termination (11/33), or a miscarriage after 12 weeks (13/33): 12 of the women had had a previous miscarriage or perinatal loss, and 18 already had a healthy child. All the grief, depression, and anxiety measurements improved on average at each time point – enough to be a noticeable difference. For example, between the time point at around 2 weeks after the loss and the time point 4 weeks later:
  • The average score on a German grief scale was 3.6 at the first time point and 3.3 at the second time point (and 3.0 by the last measurement 9 months after the first); and
  • The average score on the HADS depression scale, where below 7 is not regarded as depressed, was 7.7 at the first time point and 6.5 at the second – by the last measurement 9 months after the first, the average score was 4.4.

Gurmeet Singh, 1989.

Place: Nangal, India.

Recruitment methods: Parents of all the children who died in a single boating accident, contacted within 2-4 days of the accident.

Number eligible: Parents of 22 children.

Number participating: 44 parents.

Study methods: Interviews.

Time points and participation:

  • First week: 44 parents.
  • 1 month post-loss: 44 parents.
  • 2 months post-loss: 44 parents.
  • 3 months post-loss: 44 parents.
  • 6 months post-loss: 44 parents.
  • 1 year post -loss: 44 parents.

Several reactions were universal or near-universal in the first week (all 80-100%, in order from most common): sleep disturbance, sadness, exhaustion, sighing, loss of appetite, weeping, psychomotor slowing, loss of interest in work and social activities, preoccupation with the lost child.

By the end of the first month, these were no longer reported for from around 5-10% of the parents, except for 2 of them. Preoccupation with the child had risen a little (from 86 to 89%), and loss of interest in work and social activities was no longer reported for a larger proportion – from 90% in the first week, to 70% at the end of the month.

By the end of the second month, the reduction was more marked, and again by the end of the third month. By then, the proportion of parents reported to have a loss of interest in work and social activities was 16%. Sleep disturbance dropped from 100% in the first week, to 91% at 1 month, 82% at 2 months, 61% at 3 months, and 27% at 6 months.

Preoccupation with the deceased reduced, but stayed high: in the 80s to 2 months, 77% at 3 months, and 57% at 6 months (59% at a year). Sadness also stayed high: in the 90s or 80s up to 3 months, then in the 70s at 6 months and a year.

Loss of purpose changed dramatically as time went by: from 66% in the first week, to 48% at a month, 36% at 2 months, 14% at 3 months, 6% at 6 months, and 0 at 1 year.

Appendix: Mapping for overlap of studies

This mapping doesn't include all the studies from the Category 1 systematic review on complicated grief after sudden and violent loss (Djelantik 2020). There was no overlap in studies.

* = included in more than 1 systematic review within the category
bold = the same study in more than 1 category
(+ 3b) = a Category 3a study that's also included in Category 3b
(Aarhus) = Aarhus Bereavement Study
(CLOC) = Changing Lives of Older Couples Study (only 1 study included - noted for future reference)
(YBS) = Yale Bereavement Study (note Goldsmith 2008 includes YBS plus an additional group)

Category 1Category 2Category 3a+ Category 3b
Allen 2013Aneshensel 2004Boelen 2015Ling 2013
Middleton 1996  
Barbosa 2014Bennett 2019Chen 2021Scheidt 2012
Bartik 2013Bonanno 2002 (CLOC)    Cozza 2019Singh 1989
Byrne 1994
Bonanno 2005     Domingue 2017
Cardoso 2017    Bonanno 2019    Harris 2021 (Aarhus)
Dyregrov 2015, 2003Djelantik 2017Kentish-Barnes 2015
Goldsmith 2008 (incl YBS)   Djelantik 2022Kõlves 2020
Harms 2015Galatzer-Levy 2012Kreicbergs 2004
He 2014Kristensen 2020Kuo 2017 (+ 3b)
Hu 2015Lenferink 2018Maciejewski 2007 (YBS)
Kersting 2011*Levy 1994Nakagawa 2021 
Kim 2015Lundorff 2020 (Aarhus)O'Connor 2015 (Aarhus)
Kristensen 2012MacCallum 2015Rostila 2018
Li 2015Mancini 2015Tang 2021 (+ 3b)
Li 2016Nielson 2019Wall-Wieler 2018
Lundorff 2019Ott 2007Wijngaards-de Meij 2005
McDevitt-Murphy 2012Smith 2020Youngblut 2013 (+ 3b)
Middleton 1996Sveen 2018

Mitchell 2004Szabó 2020

Miyajima 2014Wen 2020

Mizuno 2012Zhang 2008

Morina 2011, 2010

Morina 2012

Neria 2007

Newson 2011*

O'Connor 2010* (Aarhus)

O'Connor 2019

Pan 2019

Prigerson 1999

Prigerson 2009 (YBS)

Schaal 2009

Shear 2006

Spooren 2001

Stammel 2013

Tang 2016

Tsutsui 2014

van Denderen 2016

Varga 2015

Williams 2015

Xu 2014

Yun 2018

Copyright, Hilda Bastian, October 2022. The last search for studies was in August 2022, and updates are planned.


  1. As someone who has experienced many bouts of grief following the loss of close family members, I appreciate this compilation of research. After one particular death of a loved one in 2008 I researched grief on the internet to give me some solace. I don't know if it helped. Part of the problem from my perspective was that I experienced what I view as a cumulative effect of sudden loss, a sort of piled on sense of loss as I lived longer, such that each loss brought back memories of prior loss and possibly accounted for some prolonged grief. Even when my mother died two years ago at a ripe old age of 96, I found all my past losses coming back to me with a fury, and contributing to my grief after her death.

    1. Thanks for that, and I'm so sorry for the weight of your losses. When I was felled by this grievous loss, it hit me that I had never appreciated that aging didn't just mean physical aches and pains, but this: that you outlive people around you, increasingly over time. My very best wishes to you. (And my apologies for the delayed response - I messed up the setup of the blog, and only just found comments were waiting for me.)

  2. I attended two presentations and a 5 day workshop with Elizabeth Kubler Ross in Austin TX in 1995. She told me in private (when I spotted her smoking under a stairwell during a break during one of her presentation) that she never considered her approach to dying and grief as a 5 Step program but rather saw what was popularized as The Five Steps more like Stations Of The Cross that could be skipped, repeated, visited in no rigid order to the extent useful for the griever at different times. One of the reasons I finally retired after 50-odd years as a mental health and social work professional was the requirement that I teach my clinical supervisees how to use the DSM for billing purposes and to qualify people for services, etc. From my clients, family, friends, and literature, etc., I learned that how we respond to death is never a one-size-fits-all affair. MS in Psychology from UT Austin 1967, MSW in Clinical Social Work from the U of M at Ann Arbor 1978.

    1. How special to have met such an amazing person! Even though I don't think this particular part of her work was as useful as many believe, I'm a great admirer of so much of what she achieved for reform of care of the dying, and many societies' understanding of it. Thanks for your comment and apologies for the delay in responding. (I messed up the blog setup, and only just found comments were waiting for me.)

  3. So sorry that you had to experience the loss of a child. Thank you for your scientific approach to studying grief. In the first 60 years of my life I have grieved the loss of friends, grandparents, parents, and other family, including pets. That grief seemed natural and short-lived with only occasional sad memories that sometimes return. However, 4 years ago, losing a child to suicide was the most painful experience ever. No physical pain could even compare. My first sense of feeling a moment of joy came 8 months later while vacationing with my grandchildren. It was 18 months before I could find some distractions from the pain. Now at 4 years, I am a different person, with joy in one hand and sorrow in the other. I have learned to live with the grief, but don't ever see it as "healed". I feel it could be compared to someone learning to live with a new disability after a car wreck. I have learned to honor my daughter and help others because of the grief. I know that grief never leaves completely because I lost my mother at age 4 and that grief still visits me occaionally. The stages of grief usually weren't helpful, but sometimes added more confusion to what I was feeling. Yet the EKR stages brought some comfort knowing that people cared about the pain of grief to want to help others. I'm not sure if there can be a timeline for grief, but I do appreciate that you care enough to study it and that you've chosen to honor your son.

    1. Thank you so much. I'm so very sorry for the loss of your daughter, and your loss of your mother young: that's just so much pain to have to go through and carry. Losing my parents and grandparents was similar for me, too - nothing about that grief prepared me for what losing a child is like, not really. I'm so very glad to hear you're out the other side - and have the joy of grandchildren. (And apologies for the delay in answering - I messed up the blog's setup, and only just found comments were waiting for me.)

  4. First of all, I am so sorry for your loss. I also want to say thank you for sharing your deep knowledge about grief and for writing about your experience and your research in this way. For many years I have read about grief and I am grateful to have found such thoughtful and thorough writing on this topic.

    1. Thank you, Anne. (Apologies for taking so long to answer - I messed up when I set up this blog, and only just found comments were waiting for me.)

  5. I have followed you on Twitter for some time. Thank you for this systematic and structured presentation of research on grief. I am so sorry for your loss of your son, and grateful that it has inspired you to apply your skills and expertise to the study of grief research. I also lost my son suddenly and completely unexpectedly. He took his own life four and a half years ago, he was 23.
    It was hugely traumatic and an ‘out of order’ death. He had no prior known mental health issues.
    I recognise the patterns of grief experience you describe. If I was to fill in questionnaires on my mood, coping, resilience etc, my responses would vary depending on the day I completed them. I have worked as a clinician and now as a researcher in palliative care for most of my adult life so am helped by the fact that I can talk about out of order death, and work with people who generally have high levels of emotional maturity who can also talk about death and that helps. I wonder how grief is experienced in communities when ‘out of order’ deaths are more common, when the loss of your own child, sibling, partner is compounded by many out of order deaths in other families in your community, because of violence, poverty, disease.
    I also wonder about the outcomes measured in bereavement research. Whether the outcomes discriminate enough between out of order deaths and deaths when expected. Are depression and anxiety the only attributes worth measuring? How else might the lives of people change? While I experience regular but intermittent episodes of profound sadness, I also experience a huge range of other emotions, both negative and positive. And my behaviours and expectations have definitely changed. Questionnaires can reduce the experience. Many things give me solace, and help me to cope. I’d love to see more research on this. The novel ‘Happiness’ by Aminatta Forna, has helped me reflect. The character Attila, a Ghanaian psychiatrist, is asked to assess a young woman charged with arson. The young woman has experienced trauma and her defence are certain that the trauma must have damaged her. Attila says ‘suffering, yes, … there will be suffering, but suffering and damage are not the same… most people who have endured trauma do so without lasting negative effects, but we overlook the ones who cope because we never see them”. The trauma that my son ended his own life, the huge missing of him that me and my family will feel forever has changed me. Its changed the way I live. I am not damaged but changed. Alongside many many other people, the trauma made me a different mix of strengths and flaws, joys and wounds. The world has always been so. I wish my son was still here with all my heart. Saudade, this Portuguese word describes how I live now.

    1. Oh... I so wish both our sons were still with us. Thank you for your kind words and for sharing your thoughts: I'm sure your words will be helpful to others. I'm so very sorry for your grievous loss. I agree there are many attributes worth measuring. More will unfold with other analyses - for this particular question, I sought depression and anxiety out, because they were the outcomes that were most common across these studies. You raise important points about the weaknesses of questionnaires. Those things that give us solace are definitely high on the list of topics for future work.Yes, the compounding of losses and life stresses makes everything so much harder. I'm so glad you and your family and doing well. I wish you the very best, and send a big hug. (Also apologies for the delay - I thought I'd set the blog up with notifications of comments, but I hadn't, and only just found comments were waiting for me.)

  6. Thanks so much for your dedicated research, and for sharing that with us here. I work as a therapist primarily with grieving parents, and am always looking for good evidence to support psychoeducation and recommendations with clients. The way you write of your experience and the summary of research you provide fits well with what I see clinically, too, but provides a language and stronger structure to rely on - which people need so badly. Thank you. I'm sorry for the loss of your son.


Post a Comment