Narrating Final Memories From Spousal Loss: The Role of Place and Quality of Death

Abstract

Objective:

Personal memories of the death of a spouse can guide bereavement adjustment. Place of death and quality of death are end-of-life factors that are likely to influence death experiences and formation of subsequent personal memories. The current study employs narrative content-analysis to examine how place and quality of death relate to affective sequences present in older adults’ final memories from the death of their spouse.

Method:

Based on power analyses, 53 older adults were recruited and completed a Final Memory Interview. They also reported place of spouse’s death (ie, in hospital, out of hospital) and quality of death across four subscales. Final memory narratives were reliably content-analyzed (interrater agreements >.70), revealing positive and negative affective sequences, including: redemption, contamination, positive stability, and negative stability.

Findings:

Experiencing the death of a spouse in hospital was related to narrating final memories with contamination. In terms of quality of death, reporting a less comforting social environment at time of death was related to the presence of redemption in final memories. Reporting that one’s spouse received appropriate medical care related to narrating memories that showed positive stability.

Conclusions:

Final memories are carried with the bereaved long after their loss. Positive final memories appear to stem from witnessing a comfortable, medically appropriate death outside of a hospital setting. End-of-life ‘that is’ between care and aligned with patients’ values for place and treatment may be critical for spouses’ formation of constructive final memories and bereavement adjustment.

Personal memories are recalled versions of experiences¹ that scaffold life stories and can influence current wellbeing.² Memories from the dying days of a deceased close other influence bereavement adjustment and functioning across late life.^3–5 In older adulthood, individuals are commonly cared for by spouses,⁶ who provide support for activities of daily living, decision-making, and existential and emotional needs, and thus have the capacity to develop salient narratives of the dying days. The current study identifies end-of-life care factors (i.e., place of death, quality of death) that are associated with positive and negative affective sequences in older adults’ memories of the final time spent with their spouse.

End-of-Life Care Factors: Place of Death and Quality of Death

The two end-of-life care factors we investigated recur in theoretical and empirical work on a good death⁷: the place of death (e.g., in hospital or out of the hospital) and the quality of the death.^7–9 Place of death is considered to “set the stage” for access to resources, comfort from pain, availability of professional care, and social visiting opportunities.^10,11 Patients and close others have historically reported preferences for a home death as opposed to a hospital death, even if dying outside hospital will shorten the lifespan.¹² Home death has also been considered one indicator of quality palliative care delivery,¹³ but it has been unclear how place of death affects the bereaved over time.¹⁴ Recent studies highlight the complexity of determining a preferred place of death: qualitative research demonstrates that both patients and their family caregivers are concerned that dying at home undermines access to quality or expert care,^14,15 and a longitudinal cohort study similarly showcases ambivalence about dying at home.¹⁶

Beyond place of death, quality of death refers to the deceased experiencing physical and psychological comfort, receiving social support, feeling a sense of life completion and death preparation, and maintaining dignity.^7–9,17 While quality indicators are expected to align with patients’ values, recently, attention has also been paid to understanding close others’ perspectives on death quality,^15,18 as better patient quality of death is considered to promote positive bereavement adjustment for close others.¹⁹

The Dying Days of a Deceased Spouse: Examining Final Memories

Using a life story approach to investigate the impact of end-of-life factors on narration of memories from the dying days acknowledges that loss reactions develop over time^20,21 and poises researchers to consider links between loss experiences, grief, and health behaviors for the bereaved.^3,22 Specifically, final memories, one’s memories of very last moments or experiences with the deceased, are considered likely to be central to bereavement adjustment processes.²³ Final memories may be imbued with affect, in other words, emotional tone overlayed when constructing a narrative from a lived experience.^24,25 Affect can be measured in terms of affective sequences in narratives²⁶: redemption, where affectively bad scenes are made better in light of ensuing good, contamination, where good or neutral scenes are spoiled in light of resulting bad, and positive or negative stability, where scenes are consistently affectively good or bad. Affect imbued in recalled memories from the loss of one’s spouse is likely to have lasting implications for distress,^4,27 therapeutic processing,²⁸ and bereavement adjustment.^4,29–32

Current Study

The current study examines how two end-of-life factors (i.e., place of death, quality of death) are associated with affective sequences that manifest in older adults’ final memories of from the death of their spouse. To provide inferential evidence of these links and build upon existing qualitative research,¹⁵ a quantitative, content-analytic approach was chosen. The aims were to 1) characterize and examine frequencies of affective sequences in the final memories, 2) examine associations between place of death (i.e., in hospital, out of hospital) and final memory affective sequences, and 3) examine associations between quality of death, as reported by the bereaved, and final memory affective sequences, controlling for place of death.

Method

Participants

Data collection was approved by the university’s Institutional Review Board. Community-dwelling older adults in the Southeastern US were recruited before the COVID-19 pandemic (2019–2020) using community bulletins, retirement community email listservs, and snowball sampling. Inclusion required that older adults were past the initial grieving period (i.e., loss occurred between 3–12 years prior) and not clinically distressed.³³ The range of time since the loss was intentional and reflective of bereavement as an ongoing, lifelong process.²⁰ The upper cut-off of 12 years was selected to maintain a sample who had lost spouses in late life, as opposed to mid-life or earlier. Individuals whose spouses experienced sudden or medically unexpected deaths were not included. Participants were also excluded due to possibility of cognitive impairment (i.e., more than six errors on the Orientation-Memory-Concentration Test).³⁴ Screening for each of the inclusion criteria occurred over the phone, prior to study enrollment. Power analyses conducted using G*Power 3.1.7 confirmed adequacy of a sample of 53 participants for testing inferential relations (i.e., Aim 2 and 3) and detecting small-medium effects with the ANOVA and regression analyses specified below.

Procedure

Participants were screened for study eligibility over the phone by a trained research assistant. Data collection was conducted individually, in person, in a quiet, comfortable room. The research assistant guided participants through the informed consent process and provided verbal and written study instructions while collecting self-report responses and final memories. To control for order effects, counterbalanced orders of administration were used.

The Final Memory Interview was designed for this study, grounded in autobiographical memory interviewing practices refined in previous work by the study team.³⁵ Participants were first given up to two minutes to think of the memory. When they were ready, they had up to seven minutes to narrate their memory aloud. On average, participants spent 4:20 minutes narrating their memory (range = 1:10 – 6:17 minutes). Instructions included:

The memory I’m hoping you can share is the very last time you saw your spouse… you might have exchanged some words or just been together without talking… All deaths are different – this might be a memory with simple or very complicated emotions… please tell me everything that you can remember about what happened, including what you were thinking, feeling, saying, doing…

Measures

Scalar study measures assessed place and quality of spouse’s death. Final Memory Interview narratives were content-analyzed for affective sequences. Descriptive demographic variables were also collected.

Place of Death.

Responses to the item, “In what environment did your spouse pass away?” were organized into two categories: in hospital and out of hospital. Out of hospital included several locations, predominantly one’s own home (20.4% of participants), a care facility (e.g., assisted living, skilled nursing; 20.4% of participants), or a hospice facility (24.1% of participants).

Quality of Death.

Participants completed subscales of the Good Death Inventory⁹ using Likert Scale ratings (1 = absolutely disagree, 7 = absolutely agree), with higher scores indicating greater perceived quality of spouse’s death. Due to conceptual and statistical associations between the measure’s original subscales, some subscales having very few (i.e., two) items, or subscales having low inter-item reliability, original subscales were combined into four broader subscales: appropriate medical care (6 items; physical and psychological comfort, natural death that avoids aggressive treatment; Cronbach’s α = .68), dignity and positivity maintained (6 items; maintaining hope, pleasure, and respect from others; Cronbach’s α = .71), comforting social circumstances (6 items; ability to maintain good relationships with family, avoid social disturbance; Cronbach’s α = .66), and end-of-life resolution (9 items; feelings of life completion, preparation for death, and control over the future; Cronbach’s α = .76).

Content Analysis: Affective Sequences in Final Memories.

Final memory narratives were transcribed verbatim and analyzed using standard content-analysis practices, including established codebooks,³⁶ for affective sequences^29,37,38: redemption, contamination, positive stability and negative stability. Narratives that did not contain affective sequences did not receive any code.³⁰ Detailed descriptions of each coding scheme can be found in Supplemental File 1. Coders were an undergraduate and a post-baccalaureate research assistant led by the first author, who has expertise in narrative content-analysis. Each of the two research assistant coders was on a separate coding team (ie, redemption and positive stability, contamination and negative stability), and the first author was on both teams. This allowed for minimizing coding errors (ie, because each code would be verified by two coders) and accidental coding of both positive and negative affect in one memory (ie, because the first author reviewed codes from both teams and scrutinized occasions where both positive and negative codes were suggested). A reliability check consisted of independently coding 27 narratives and indicated good reliability (all kappa scores ≥.70). To prevent coder drift and resolve disagreements, coders met weekly while coding all study memories. Final codes thus reflect agreement between coders.

Results

This study included 53 participants (M_age = 81.59; SD = 7.57; range = 70–96). On average, participants had lost their spouse 6.81 years prior (SD = 2.98; range = 3–12). Just over half (62.3%) were female, all identified as White, and three (5.56%) identified as Hispanic/Latinx. All identified as having had a heterosexual relationship with the now-deceased spouse. Preliminary analysis showed one-third (n = 18) of participants’ spouses died in hospital and two-thirds (n = 36) died outside of hospital. Bivariate correlation and one-way ANOVA analyses demonstrated that years since the death and order of administration were not associated with other variables (rs < .25, ps > .05; Fs < 1.90, ps > .05). See Table 1 for preliminary correlations.

Table 1.

Bivariate Correlations Between Place of Death, Quality of Death Variables, Affective Sequence Variables, and Years Since Loss.

	Variable	1	2	3	4	5	6	7	8	9	10
1	Place of death	-
2	QoD1: Appropriate medical care	−.28	-
3	QoD2: Dignity and positivity	−.03	.54**	-
4	QoD3: Comforting circumstances	−.24	.61**	.63**	-
5	QoD4: End-of-life resolution	−.40**	.33*	.19	.37**	-
6	Redemption	−.02	−.16	−.16	−.39**	−.27	-
7	Contamination	.27*	−.13	−.08	−.29*	.12	−.18	-
8	Positive stability	−.22	.43**	.10	.19	.24	−.20	−.18	-
9	Negative stability	.19	−.22	−.12	.12	−.03	−.17	−.16	−.18	-
10	Years since loss	.20	−.22	−.08	−.12	.007	.15	.03	−.25	.21	-

QoD refers to Quality of Death subscales.

^*= P < .05.

^**= P < .01.

Aim 1: Frequency of Affective Sequences in Final Memories

Of the 53 participants, 62.26% narrated their memory with one of the four affective sequences (See Table 2).¹ The remaining participants told memories with no affect, where accounts of final moments with the deceased were factual rather than emotional. Table 3 presents excerpts from older spouses’ narratives to illustrate each of the affective sequences.

Table 2.

Percent of Affective Sequence Types in Final Memories.

Affective sequence type	Percent
Redemption: Negative to positive shift	27.28
Contamination: Positive to negative shift	24.24
Positive stability	27.28
Negative stability	21.21

Note. Percentages are of total memories that included an affective sequence.

Table 3.

Exemplars of Final Memories Exhibiting Each Affective Sequence and no Affect.

Redemption Affectively Negative beginning shifts to positive ending	She was diagnosed late in her life, uh, her doctor -- And the doctor she found was, was Not a good one. She collapsed at work one day… so I took her to her doctor’s office… and the doctor said, “Put her in the hospital.” And her doctor then went on a two-week vacation. What he had failed to diagnose was that she had a chest tumor the size of a Football. The treatment recommended, even though it seemed Futile, was chemotherapy. She had one chemotherapy treatment which almost killed her, she wasn’t strong enough to withstand that. She was, she was so terribly weak. We saw a Number of tests being performed, which appeared to be Futile. And I asked my wife if she’d just like to go home and she said “Oh yes please”. We brought in hospice, and I’d swear by them. They’re absolutely wonderful people… [the Nurse] would come with her Bible and read, read to [my wife] and, and she was just wonderful. We had a hospital bed in our bedroom and our regular bed was right next to her, so we could, we could lie together and hold hands as we went to sleep. Just before she went to sleep I bent over to kiss her and said uh, “Save me a place with you in heaven”. And she will. That’s it
Contamination Affectively positive or neutral beginning shifts to negative ending	It was December, it was nearing christmas and her birthday, we were at a computer sending emails to her friends, to family and all of sudden she said “Something’s happened to me,” and she kind of slumped over in the chair. I Tried to get her out to get her to the bedroom so she wouldn’t fall on the floor, I finally got the chair out of the door with much difficulty, called 911… they kind of tested her and did things, I thought we should just get her in the ambulance and take her and they eventually did… I got to the hospital, they told me she had passed. I Fully thought that they would bring her back, I knew she had cancer, and I just thought - I knew it was bad, I knew she really had a 50/50 chance of surviving. But I thought she’d make it through that cause she’d always been very healthy and strong
Positive stability	The backdrop of this last minute of her life goes back to the two years when she was losing her memory. She would go to bed early so I would always come in and lay next to her for a while. I’d kiss her on the bridge of her nose… and I’d always say, “I love you, I love you, you’re my dear, dear.” So, as her breath was gradually getting softer and softer. I Knew this was the way you could tell that she was about to pass on because she had been taken off the feeding tube. I Was basically trying to be patient because I knew there was nothing I could do. Nothing that needed to be done. She did not want a feeding or breathing tube… so I didn’t have any really bad emotional experiences, it was part of the pattern and what was expected. So as [her breathing] got shorter and shorter, I would get up, go to the bed, kiss her on the bridge of the nose, say “I love you, I love you, you’re my dear, dear.” and within a few seconds, she passed away
Negative stability	When my husband died in hospice, I was with him. During the night he was extremely distressed and uncomfortable. The next morning, it was quite obvious that the end was near. The doctor did come into the room with a nurse, and my husband was no longer speaking, or really cognizant. I Was very hopeful that it was going to end soon. I Felt guilty about that. I Mean it was obviously difficult, for sure. I Was hoping that he’d still hear me saying that I love him. But I’m not sure that he could. I’m not sure. And while the three of us were at his bedside, he passed quietly. Well, maybe I shouldn’t say quietly because he was rather distressed. I Guess that’s it. I Was obviously very upset. The whole night was very upsetting. it
Description without affective tone	She slept a lot. And she was sleeping, I guess, when she died… hospice was there, it was at home, and then the hospice nurse was in the room also, and I lay down on the bed next to her. And… I took her in my arms. And then, I can remember she took 1, 2, 3 breaths. I can’t- I don’t know why 3. But she took 3 breaths, and then that was it. She stopped. And I-I looked at the nurse and said “she stopped breathing.” she came over and… took her pulse and, she said “I’m sorry, she’s gone.”

Note: Identifying details (e.g., names, locations) and additional transcription details (e.g., pauses in speech, crying, shaking voices) have been removed.

Aim 2: Association of Place of Death and Affective Sequences

Four one-way ANOVAs were conducted,² each with place of death (in hospital, out of hospital) as the independent variable and affective sequence (redemption, contamination, positive stability, and negative stability) as dependent variables. Place of death was associated with narrating final memories with contamination: spousal deaths that occurred in hospital were more likely to be narrated with contamination than deaths that occurred outside of hospital, F (1, 51) = 4.06, P < .05, np² = .07. Place of death was unrelated to narrating a final memory with other affective sequences: redemption, F (1, 51) = .02, P > .05 = .90, np² = .00, positive stability, F (1, 51) = 2.55, P > .10, np² = .05, negative stability, F (1, 51) = 1.93, P > .10, np² = .04.

Aim 3: Association of Quality of Death and Affective Sequences

Hierarchical linear regressions were conducted for each affective sequence. In each model, place of death was entered in Step 1, and the four variables representing quality of death were entered in Step 2, with affective sequences as criterion variables. Significant final regression models appear in Table 4. The final regression model for redemption, F (5, 47) = 2.98, P < .05, showed change in R² from Step 1 (.00) to Step 2 (.27; P < .05), indicating that the quality of death variables entered in Step 2 had a significant effect beyond place of death. Participants who reported a less positive social environment for their spouse’s death more often narrated their final memory with redemption (t = −3.05, P < .01). For positive stability, the final regression model, F (5, 47) = 2.62, P < .05, also showed change in R² from Step 1 (R² = .06) to Step 2 (R² = .24; P = .06). Participants who reported more appropriate medical care for their spouse more often narrated their final memory with positive stability (t = 2.60, P < .01). The final regression models for contamination, F (5, 47) = 2.43, P > .05, and negative stability, F (5, 47) = 1.90, P > .10, were not significant.

Table 4.

Summary of Significant Regression Models Predicting Affective Sequences.

Variable	B	SE	β	t
Criterion: Redemption
Place of death	−.07	.05	−.21	−1.38
QoD1: Appropriate medical care	.00	.02	.00	.02
QoD2: Dignity and positivity	.05	.03	.30	1.53
QoD3: Comforting circumstances	−.10	.03	−.61	−3.05**
QoD4: End-of-life resolution	−.03	.02	−.19	−1.22
Criterion: Positive stability
Place of death	−.02	.04	−.08	−.51
QoD1: Appropriate medical care	.04	.02	.49	2.60*
QoD2: Dignity and positivity	−.02	.03	−.13	.53
QoD3: Comforting circumstances	−.01	.03	−.05	−.26
QoD4: End-of-life resolution	.02	.02	.13	.83

Note: Regression statistics are presented for the final model, at Step 2, for significant models. Place of death is entered into models as a dichotomous variable (0 = out of hospital, 1 = in hospital). QoD refers to Quality of Death subscales.

^*= P < .05.

^**= P < .01.

Discussion

Our findings indicate that about two-thirds of our sample of conjugally-bereaved older adults narrated final memories of their spouse with affect. Our results challenge assumptions that memories from the dying days are inherently negative: final memories that were affectively negative constituted the smallest category. Affect imbued appears to depend in part on end-of-life care factors: death in hospital is associated with recalling a unique type of “bad death” represented by a contaminated affect. Quality of death factors predicted narrating final memories with positive affective sequences, including redemption and stable positive affect.

Hospital and Non-Hospital Deaths: Implications for Final Memories

When their spouse died in hospital, participants more commonly told contaminated narratives, where neutral or positive scenes were spoiled by distressing endings. Narrating a memory with contamination may reflect a unique type of recalled “bad death” where the bereaved had not fully understood that their spouse was close to death. Considering that none of the participants lost spouses due to sudden illness onset or injury, “unexpected” deaths may have been instances in which spouses did not have the resources (e.g., health care literacy, clear communication from clinicians) to anticipate their spouse’s death occurring.³⁹ Hospital services may have been involved in the last days or hours of life in an attempt to improve circumstances, temporarily boosting feelings of control or hope for health improvements (e.g., quote: I knew it was bad, I knew she really had a 50/50 chance of surviving. But I thought she’d make it through… Table 3). Our findings thus suggest that supporting positive end-of-life experiences relies on preparing close others so that the death is less unexpected whenever possible by, for example, openly communicating about prognosis, explaining signs of imminent death, describing the limited benefits of hospital transfer near end-of-life, and making a plan to shift care to home permanently following hospital discharge.^40–42 If patients’ and spouses’ health care priorities include death in hospital, contaminated memories may be avoided if medical teams communicate about the imminence of death prior to or upon hospital admission.

Quality of Death: Positive Affect in Final Memories

Redemption in personal narratives is indicative of coping with adversity. Of all aspects of quality of death, creating comforting social circumstances is the one for which spouses often consider themselves most responsible. When circumstances were not comfortable, spouses may grapple with feeling personally responsible for deficiencies (e.g., not encouraging family to visit, not finding compassionate clinicians; quote: the doctor said, “Put her in the hospital.” and… then went on a 2-week vacation; Table 3). These circumstances may invite redemptive narration, where positive endings redeem disappointing scenes. Redemption in final memories may reflect recognition that intimate final moments with the deceased can salvage socially troubling circumstances that play out across the dying days.

Among multiple end-of-life factors, participants’ reports of appropriate medical care at end of life were associated with narrating a final memory that was steadily positive. These positive memories showcase one’s spouse seemingly at peace, content, or to be acting “like themselves”. When a patient is imminently dying, optimal medical care often involves prioritization of palliation over curative treatments.⁴³ The current study demonstrates the significance of appropriate medical care, including care focused on palliation, at end-of-life (e.g., quote: I knew this was the way you could tell that she was about to pass on because she had been taken off the feeding tube… she did not want a feeding or breathing tube… so I didn’t have any really bad emotional experiences; Table 3).

Limitations

This study was a first investigation of the relation between end-of-life factors and affect in final memories, and as such was conducted with a modest sample to establish that affective sequences can be detected. Future work with larger samples could use different analyses to examine differences across participant subgroups (e.g., moderation effects). The study design was based on recent care policy focused hospital deaths, so was limited in its comparison of other places of death. Future work should examine differences in recall of final memories across other place of death contexts. The study sample comprised White, upper-middle class older adults. Findings from the current study cannot be extended to, for example, racially or ethnically diverse older adults, nor those with major barriers to health care resources.

Conclusion

Recognizing that the individuals make sense of life experiences through narrative, this study examined the association of end-of-life factors and narration of final memories. Although values for end-of-life sometimes vary between patients and their close others,¹⁷ both groups consistently report valuing physical and psychological comfort through appropriate medical care. Our results show that this quality of death factor promotes affectively positive final memories of deceased spouses, further warranting practices which ensure appropriate care for dying patients. In circumstances where key end-of-life factors are not met (e.g., unexpected death in hospital, uncomfortable end-of-life social environment, misaligned medical treatment), resources to help bereaved spouses process their loss may be crucial for constructive bereavement adjustment.