Spiritual beliefs may affect outcome of bereavement: prospective study

Data analysis

We analysed the data using the statistics software package SPSS (versions 6 and 9). We used Student's t test for normally distributed data and the χ² statistic for categorical data. On an a priori basis, we divided the sample into three groups on the basis of their beliefs: no spiritual belief, low strength of belief, and high strength of belief. The latter two groups were divided on the mean score of the spiritual scale (the mean and median scores were 27). We compared the three groups by using analysis of variance and the Pearson χ² statistic. We analysed our principal outcomes at 1, 9, and 14 months by using multivariate, repeated measures analyses. We present Greenhouse-Geisser adjusted F ratios for the effect of time at 1, 9, and 14 months and the interaction of time and strength of belief to test whether the three strength of belief groups differed in their course of grief.

We conducted further exploratory analyses of the effects of the hypothesised confounders in the relation between spiritual belief and outcome of bereavement by using repeated measures, multivariate analyses of covariance in which age, depression, locus of control, and degree of social isolation were included as covariates. Sex of the participant and emotional closeness to the patient were included as between subjects factors. We entered the covariates and factors into the analysis, beginning with demographic variables and following with psychological and then social variables. The final model included all variables that showed important (P<0.2) within subject or between subject effects. We retained strength of belief in the model because this addressed our main hypothesis.

We estimated that 170 participants would be sufficient to detect a moderate correlation of 0.25 between the scores on the core bereavement items scale and strength of spiritual beliefs with 90% power and at the 5% level of significance.

Response rates

We approached 216 people, 153 (71%) of whom agreed to be interviewed (fig 1). We found no significant difference in sex of people agreeing or refusing to take part. We did not undertake interviews with 18 participants because their relative or close friend died before the interview could be arranged. Therefore, a total of 135 people (90 women and 45 men) took part in the baseline interviews. In all, 129 patients (96%) died within the time available for follow up and thus the relatives and friends of these 129 comprised the study cohort. Ninety five (74%) of these completed all follow up assessments. No significant differences were found between non-responders, responders with incomplete data, and those who completed all assessments (table 1), although there was a tendency for people with no spiritual beliefs to fail to complete all follow ups.

Fig 1.

Flow chart for studying the effect of spiritual beliefs on outcome of bereavement

• Download figure
• Open in new tab
• Download powerpoint

Baseline characteristics

Fifty one participants (40%) were the spouse or partner of the dying patient and 52 (40%) were their adult children. The remaining 26 (20%) were made up of 9 (7%) friends and 17 (13%) other relatives. Fifty two (40%) named the dying patient as the person they had felt closest to over the previous 12 months, with a further 13 (10%) naming the dying patient as the second closest person to them. Of the 114 patients for whom data were not available, the mean age at death was 67 years (SD 12.3, range 28-94); only eight (7%) of the people whose age was known were under 50. In all, 72/129 (56%) patients died within one month of the baseline interview with their participating friend or relative, a further 37 (29%) died within three months, and 20 (16%) died more than three months later. At first follow up one month after the death, participants were asked about the circumstances of the death. Of the 114 who were asked, 65 (57%) were present at the death, whereas 88 (77%) had wanted to be there. Thirty-nine (80%) of those who were not present at the death had seen the patient in the preceding 24 hours. These findings indicate that we had been successful in recruiting people who were closely involved with the dying patient.

A total of 21/129 people (16%) reported no religious or spiritual belief; 53 (41% of all participants) reported spiritual beliefs of low intensity and 55 (43%) reported strongly held beliefs (table 2). Marital status and ethnic origin were significantly related to spiritual belief: people with partners had stronger beliefs, and almost everyone in the low belief group was white.

Spiritual belief and outcome of bereavement

The main result of this study arose from the analysis of the 95 participants (74%) who completed all follow up assessments (table 3). Participants with strongly held spiritual beliefs recovered from their bereavement in a linear fashion (fig 2), whereas those with low strength of belief showed little change by nine months but recovered rapidly thereafter. Participants with no spiritual beliefs showed a temporary gain at nine months but their symptoms of grief had intensified again by the final assessment.

Fig 2.

Strength of spiritual belief and outcome of bereavement

• Download figure
• Open in new tab
• Download powerpoint

To explore the potential effect of confounders on the main outcomes, our first multivariate analysis concerned the 95 people who completed all assessments. The first line of table 4 describes the effects of time and the interaction between time and strength of belief, when strength of belief is the only variable entered into the model. We found a borderline significant effect (P=0.058) for the interaction between strength of belief and time (fig 2).

We then examined three groups of potential confounders of the relation between spiritual belief and outcome of bereavement. When we entered the demographic variables of sex and age into the model, the effect of strength of belief remained significant (table 4). Although sex of the participant was important (men had higher grief scores than women initially but not by nine months), it did not remove the effect of strength of belief on the course of grieving. Age was not a significant covariate in initial grief or recovery with time. We then examined the effects of psychological status and locus of control. Mean scores on the hospital anxiety and depression scale before the death were significantly associated with initial levels of grief. Higher depression scores at baseline also predicted higher scores on the core bereavement items scale at all follow ups. However, scores on the hospital anxiety and depression scale or mean locus of control scale failed to have a major impact on the effect of strength of belief (P=0.063) or the interaction between sex and time. Next, we examined the effects of social isolation and emotional closeness of the participant to the deceased. Adding these variables reduced the effect of strength of belief (P=0.263). Closeness to the dying patient predicted higher grief scores at all follow ups. The final model included strength of belief, sex, and closeness to the deceased as between subject factors and hospital anxiety and depression scale score at baseline as a covariate. Strength of spiritual belief retained its borderline significance. At 14 months the difference between the group with no beliefs and the combined low and high belief groups was 7.30 (95% confidence interval 0.86 to 13.73) points on the core bereavement items scale. Adjusting for confounders in the final model (table 4) reduced this difference to 4.64 (1.04 to 10.32) points. These figures correspond to an effect size of 0.65 standard deviations before adjustment for confounding and 0.41 standard deviations after adjustment.

We undertook the same analysis, using all 114 people who had completed the baseline and first follow up assessments, imputing missing values thereafter by using the last observation carried forward. The interaction between time and strength of belief was confirmed (P=0.034) (table 5). Results for the whole model were similar using this larger sample, with strength of belief remaining significant (P=0.021) when all significant covariates and factors of interest were included.

Limitations of the study

One limitation of our study is sample size, which restricts the number of potential confounders we can explore. Set against that is the difficulty of recruiting a large sample of people before the death of their loved one and following them for a considerable time afterwards. Another limitation is a lack of qualitative detail to help in our interpretation of the results. It would also have been useful to know more about participants' experiences of palliative care and counselling before the death. However, to avoid overburdening the participants, we restricted ourselves to data that were germane to our hypothesis.

Great sensitivity was required in approaching participants, and it was inevitable that a proportion would refuse to participate. This may simply have been because they moved address. But we cannot know whether some dropped out because of unresolved bereavement issues or dissatisfaction with the circumstances of the death.

Strengths of the study

The strengths of our study are that it was prospective over a long follow up period and that we recruited participants in a secular palliative care unit that has no religious affiliation. A further strength is that we recruited participants before they were bereaved. Most participants, however, appreciated that their loved one was very sick and that their death was imminent. Furthermore, not all participants completed all follow ups.

Spiritual belief

Spiritual beliefs may provide an existential framework in which grief is resolved more readily. Most spiritual beliefs, whether or not associated with religious practice, contain tenets about the course of human life and existence beyond it. Strong beliefs may be a proxy for better adjustment and less psychological distress. But our analysis suggests that strength of belief affected the course of bereavement, independently of psychological status. Relating to a person with terminal illness may sharpen our focus on spiritual or existential matters. Participants in this cohort, however, held beliefs that were similar in character and strength to those of other populations in which this standardised measure of spiritual and religious beliefs has been used. 1011

Concluding remarks

If our results were replicated, this would show that the absence of spiritual belief is a risk factor for delayed or complicated grief. It has been thought that sensitive discussions between palliative care staff and relatives before a death have a positive impact on subsequent bereavement.12 Most palliative care units try to involve family members and friends who are important and close to the person dying. Attention to spiritual matters may be a component in this work that is often overlooked or avoided by secular services. We are not suggesting that an intervention concerning spiritual matters is appropriate for people with no professed beliefs. Rather, our finding might help in identifying people who are having difficulty in readjusting to life after their loss.