A longitudinal study of bereavement phenomena in recently widowed elderly men more

Psychological Medicine, 1994, 24, 411-421. Copyright © 1994 Cambridge University Press A longitudinal study of bereavement phenomena in recently widowed elderly men G. J. A. BYRNE1 AND B. RAPHAEL From the Department of Psychiatry, University of Queensland, Australia SYNOPSIS A three-phase longitudinal design was used to study bereavement phenomena in a cohort (N = 78) of recently widowed elderly Australian men. One group of 57 widowers was compared with a second group of 21 widowers on a waiting list. A brief, interviewer-administered, structured questionnaire was used to rate the frequency of 22 self-reported bereavement phenomena. The prevalence and temporal evolution of these phenomena are described. Half of a subgroup of elderly widowers reporting high levels of bereavement phenomena at 6 weeks post-bereavement went on to exhibit persistent or chronic grief at 13 months post-bereavement. The prevalence of persistent or chronic grief throughout the first 13 months post-bereavement was 8-8%. Income, education and expectedness of the death were all negatively correlated with frequency of selfreported bereavement phenomena at 6 weeks post-bereavement. In a multiple regression analysis only expectedness of the death contributed significantly to prediction of the frequency of bereavement phenomena at 6 weeks post-bereavement. Widowers who were unable to anticipate their wife's death, even when their wife had suffered a long final illness, had a more severe bereavement reaction. INTRODUCTION Impact of conjugal bereavement on morbidity in the elderly It is generally accepted that the death of a spouse is the most significant loss event which commonly affects the lives of older adults (Gallagher et al. 1981). However, there is divided opinion about the nature and magnitude of the impact of conjugal bereavement on the elderly. We were able to identify four morbidity studies relevant to this issue in which a prospective design was employed. Two of these studies suggest that conjugal bereavement may have little impact on physical and emotional wellbeing in the elderly (Heyman & Gianturco, 1973; Murrell et al. 1988). Both studies used prebereavement status as the comparison condition. However, the findings of two other studies suggest that elderly persons may suffer significant decrements in their physical and emotional well-being following conjugal bereavement Address for correspondence: Dr Gerard Byrne, Department of Psychiatry, The University of Queensland, Room El5, Clinical Sciences Building, Royal Brisbane Hospital, Herston, Queensland, 4029, Australia. 1 (Goldberg et al. 1988; Van Zandt et al. 1989). One of these studies (Goldberg et al. 1988) used pre-event status as the comparison condition whereas the other used a parallel groups design. From a somewhat different perspective, there is evidence that the onset of clinical depression and psychiatric hospitalization in the elderly is preceded by adverse life events such as bereavement more often than would be expected by chance alone (Stein & Susser, 1969; Emmerson et al. 1989; Bruce et al. 1990). In this context, Bruce et al. (1990) using data from the Epidemiologic Catchment Area (ECA) Study, found that conjugal bereavement greatly increased the risk of depressive episodes and dysphoria in older adults. Almost two-thirds of the newly widowed reported a 2-week period or longer of dysphoria and almost one-third of all newly bereaved reported depressive symptoms severe enough to have met the DSM-III criteria for a Major Depressive Episode. Gender differences in physical morbidity and mortality While the empirical literature suggests that elderly women report more bereavement-related 411 412 G. J. A. Byrne and B. Raphael While numerous studies have looked at aspects of coping or adaptation in the face of conjugal bereavement in the elderly (for example, Gerber et al. 1975; Hill et al. 1988; Gallagher et al. 1989; Bass & Bowman, 1990), there have been Bereavement phenomenology in elderly men few published descriptions of the specific phenoThus, from these seemingly disparate strands menology of bereavement reactions in the of evidence emerges a pattern which suggests elderly. Neither has the pattern of grief over time that the impact of bereavement on the health of been adequately described in this group. elderly men may have been seriously underBereavement phenomena have often been estimated. Before this can be rigorously investi- construed as depressive equivalents or other gated it is necessary to be able to describe in non-specific psychological symptoms (Sheldon detail the phenomenology of bereavement in the et al. 1981; Vachon et al. 1982; Feinson, 1986) elderly. This is particularly necessary to allow a and most published studies have not separated distinction to be made between normal and out the affects of grief and depression. Unfortuabnormal, or 'pathological', bereavements as nately, these studies do not allow identification the latter may be accompanied by more negative of whether or not the bereavement-specific health outcomes than normal bereavements phenomena of grief are present or their intensity (Kim & Jacobs, 1991). or pattern of change over time. Traditionally, negative health effects have Two important exceptions are the reports of been attributed to the acute stressor effect of Jacobs et al. (1986) and Lund et al. (1986). Lund bereavement, assumed to be a simple and single et al. (1986) described a 2-year longitudinal event, and little attempt has been made to study of 192 elderly bereaved spouses, average measure its specific components or to identify age 67 years. The subjects were asked about 26 which, if any, may contribute. Such an under- bereavement related feelings (for example, standing of specific components is critical be- 'dazed', 'panic') and 16 bereavement related cause each may potentially have different psycho- behaviours (for example, 'visit grave', 'talk to physiological and health behavioural conse- spouse'). These 42 phenomena were reduced by quences. For example, post-bereavement bio- factor analysis to five global scales (Caserta et logical events, particularly psychoneuroendo- al. 1985). Significant effects for time but not for crine and psychoneuroimmunological changes, gender were noted for four of these scales: may be associated with particular bereavement- 'emotional shock', 'helplessness/avoidance', specific phenomena. To date, assessment of the 'anger/guilt/confusion' and 'grief resolution importance of these changes has been limited by behaviours'. There were no significant gender the lack of specific information about bereave- by time interactions. Temporal change was ment phenomena. Most such studies have greatest for the 'emotional shock' and 'grief emotional distress than men, there is evidence from a variety of sources which indicates that it is elderly men who are more likely to suffer physical health decrements following bereavement. Jacobs & Ostfeld (1977) have shown that ' widowers are at higher risk than widows for ill health during acute bereavement', and have also noted that 'the risk does not lessen with increasing age as it does among widows'. Indeed, it is among recently widowed elderly men that excess 'all causes' mortality following bereavement has most consistently been demonstrated, although the size of this effect is probably only modest (Jacobs & Ostfeld, 1977; Niemi, 1979; Bowling & Benjamin, 1985; Stroebe & Stroebe, 1987). Although 'all causes' mortality is increased in elderly widowers, it is well established that, for disaffiliated males, advancing age brings with it a markedly increased propensity to death by suicide (Shulman, 1978; Blazer et al. 1986). employed measures of psychological distress which are not specific to bereavement. For instance, in their study of immune function in bereavement, Irwin & Weiner (1987) employed the Hamilton Rating Scale for Depression as their measure of bereavement-related distress. Furthermore, in relation to the adverse health consequences of bereavement, including bereavement-related mortality, it may be important to be able to integrate knowledge about specific bereavement phenomena with information about changed health-related behaviour and conventional risk factors for the major causes of mortality in this group. Methodological issues Bereavement phenomena in elderly men 413 resolution behaviours' scales. Low levels of 'anger/guilt/confusion' were found throughout the 2 years post-bereavement. Individual phenomena were not described separately. Using a cross-sectional design, Jacobs et al. (1986) interviewed 114 bereaved spouses one month after a loss. Sixty-eight subjects were aged 61 years or older and 44 subjects were male. Subjects were asked about 38 individual phenomena consisting of 20 items from the Center for Epidemiological Studies Depression Scale (CES-D), 6 items developed to assess 'numbness and disbelief, and 12 items developed to assess 'separation anxiety'. Elderly bereaved persons scored significantly lower than middle-aged bereaved on 'numbness and disbelief but were otherwise quite similar. Individual phenomena were not described separately. While it is usually accepted that reactions to conjugal bereavement in the elderly are similar in nature to bereavement reactions in adults generally, most studies of conjugal bereavement have either used samples of bereaved adults of all ages (for example, Zisook et al. 1987 a, b) or have specifically excluded older adults (for example, Parkes & Brown, 1972). Those studies which have focused on older persons have often considered persons of 50 years and over or 55 years and over to be elderly (Lund et al. 1985; Murrell et al. 1988), resulting in cohorts with low mean ages. Understandably, many studies of bereaved elderly adults have used samples biased significantly in favour of female subjects (Gerber et al. 1975; Vachon et al. 1982; Lund et al. 1985; Gass & Chang, 1989). Unanswered questions to the death of spouse associated with a better or worse outcome of the bereavement reaction? Can a subgroup of elderly widowers be identified with absent grief? Can a subgroup be identified with chronic or persistent grief (grief persisting at high intensity for longer than a year)? Is the nature or intensity of bereavement-specific phenomena early in the post-bereavement period predictive of the nature or intensity of the bereavement reaction at the end of the first year post-bereavement? Obtaining answers to these questions is a necessary first step in exploring the relationships between bereavement phenomena and psychophysiological and health-related behaviour changes in elderly widowers. In this report we outline a longitudinal study of the impact of conjugal bereavement on elderly men and describe bereavement-specific phenomenology in this group over the first 13 months post-bereavement. Future reports will focus on mental health outcomes, physical health-related behaviours and biological concomitants of bereavement in this group in comparison with an individually-matched group of married men. METHOD Study design A three-phase longitudinal design was used. Widowers were interviewed 6 weeks after the death of their wives (Tl) and then again at 6 months (T2) and 13 months (T3) post-bereavement. Data used in this report were collected using structured interviews and self-report questionnaires. Subjects Overall, the limited empirical evidence that is available suggests that conjugal bereavement in elderly males may be associated with excess physical morbidity and mortality. However, a large number of questions remain unanswered about the nature of the reaction to bereavement in elderly men: What bereavement-specific phenomena are reported by recently bereaved elderly men? What is the nature of the temporal evolution of these phenomena? Does the suddenness of the death or the degree to which the death was expected predict outcome in terms of the intensity or pattern of bereavement phenomena? Is the adoption of a care-giving role prior Subjects were recently widowed white males aged 65 years and over. Consecutive death certificates of married females aged 63 years and over were identified from repeated searches at the office of the State Registrar-General over a 5-month period (June to October, 1989). All surviving husbands identified in this manner who resided in the greater Brisbane metropolitan area (N = 156) were initially approached by mail and followed-up by telephone. Potential subjects not listed in the telephone directory were paid personal visits. Multiple attempts were made to contact each potential subject. Only consenting, English-speaking, non-institutionalized subjects were enrolled. 414 G. J. A. Byrne and B. Raphael Table 1. Response rates for widowers Recently bereaved widowers Response category Potential subjects Ineligible Eligible Refusals Consenting subjects N 117 31 86 29 57 % 100 26-5 100(73-5) 33-7 (24-8) 66-3 (48-7) Wait-listed widowers N 39 11 28 7 21 % 100 28-2 100(71-8) 250(17-9) 75-0 (53-8) N 156 42 114 36 78 All widowers % 100 26-9 100(731) 31-6(23-1) 68-4 (50-0) Figures in parentheses represent percentages of potential subjects. death of their wives. These men represented Table 2. Potential subjects who were ineligible: 66-3% of those eligible (see Table 1). An reasons for their ineligibility additional 21 of the next 39 recently widowed elderly men to be identified in the same manner Recently W;ait-listed were enrolled 13 months after the death of their bereaved widowers widowers AH widowers wives. This group was recruited to control for the potential effects of the research interviews. N ineligibility N (%) (%) (%) N These men represented 75 % of those eligible. 1 (3-2) Deceased 5 (11-9) 4 (36-4) Reasons for ineligibility are summarized in — 10 (23-8) 10 (32-3) Age < 65 yr Table 2. About one third of the ineligible 5 (161) Not living in the 6 (14-3) 1 (91) catchment area widowers were either too young (< 65 years) or Unable to speak 2 (4-8) 2 (6-5) deceased. Only limited information was available English on potentially eligible widowers who declined to 10 (23-8) 1 (9-1) Severely ill 9 (290) Institutionalized 4 (9-5) 2 (18-2) 2 (6-5) participate. Non-participants were contacted by Miscellaneous 5 (11-9) 2 (6-5) 3 (27-3) telephone and asked their reasons for declining. Total 42 (100-0) 11 (1001) 31 (1001) Just over 10% of eligible non-participants gave emotional distress as their reason for not Table 3. Eligible widowers who refused to participating. The remainder gave a variety of reasons including serious illness, insufficient time participate: reasons given for refusal and issues relating to privacy (see Table 3). Recently bereaved widowers Reason given for refusal to participate Too busy Not interested On relative's advice Privacy issues Feels too upset Serious illness Mistrusts doctors Miscellaneous reasons* Totalf Wait-listed widowers N All widowers N (%) (%) N (%) (22-2) (8-3) (8-3) (111) (111) (5-6) (8-3) (250) (99-9) 6 (20-7) 1 (3-4) 2 (6-9) 3 (103) 4 (13-8) 2 (6-9) 3 (10-3) 8 (27-6) 29 (99-9) (28-6) 8 2 2 (28-6) 3 1 (14-3) 3 1 (14-3) 4 — 4 — 2 — 3 1 (14-3) 9 7 (100-1) 36 * Some of this residual group were probably suffering from cognitive impairment. t Figures do not total 100 due to rounding. Two groups of widowed men were studied: recently widowed men (Group 1); and waitlisted widowed men (Group 2). Fifty-seven (57) of the first 117 recently widowed elderly men to be identified were enrolled within 6 weeks of the Measures Data were collected on: socio-demographic details; aspects of the death; bereavementspecific phenomena; psychiatric diagnosis; a battery of psychological self-report measures; self-reported health-related behaviour; and selected biological measures obtained from fasting blood samples. This paper will focus on bereavement-specific phenomena and their relationship to aspects of the death. Data were collected on widowers' age, education, income, occupational prestige and selfrated general physical health. Occupational prestige was measured on a seven-point scale with scores near seven representing occupations with the lowest occupational prestige and scores near one representing occupations with the highest occupational prestige (Daniel, 1984). As the subjects were all elderly, the Mini-Mental Bereavement phenomena in elderly men 415 Table 4. Sociodemographic data, cognitive function and self-rated general physical health Variable Age (yr) at Tl range: 65-90 Cognitive function MMSE30 at T3 range: 22-30 Occupational prestige Daniel's Scale range: 2-3-6-7 Recent widowers Mean (s.D.) 74-52(4-65) 27-58(1-80) Wait-listed widowers Mean (s.D.) 73-80(4-71) 26-90(1-87) All widowers Mean (s.D.) 74-33(6-01) 27-38(1-83) 4-66(1-19) 4-71(0-96) 4-67(1-13) N* (%) Income $5000-9900 $10000-14900 $15000-19900 $20000 + Education - Age at leaving school < 15 yr >15yr Self-rated general physical health Excellent Good Fair Poor 33 (660) 10(20-0) 1 (20) 6(12-0) N* (%) 16(76-2) 3(14-3) 0 (00) 2 (9-5) N* (%) 49 (690) 13(18-3) 1 (1-4) 8(11-3) 39 (750) 13(25-0) 9 (42-9) 12(57-1) 48 (65-8) 25 (34-2) 10(17-5) 30 (52-6) 16(28-1) 1 (1-8) 3(14-3) 11 (52-4) 6 (28-6) 1 (4-8) 13(16-7) 41 (52-6) 22 (28-2) 2 (2-6) * Sample size differs for some variables due to missing data. There were no significant differences between recent widowers and wait-listed widowers on any of these variables. State Examination (MMSE) (Folstein et al. scale was used to record the frequency of 1975) was employed to estimate cognitive status. phenomena in the 2 weeks prior to the interview Likert-type scales were used to measure duration as occurring 'never', 'rarely', 'sometimes' or of wife's final illness and degree of expectedness 'often'. These anchor points were scored 0, 1, 2 of wife's death. For those widowers whose wives and 3, respectively, giving a maximum possible had not died suddenly, a categorical scale was unweighted scale score of 66. The BPQ was used to collect information on who provided the highly acceptable to distressed elderly persons, majority of care for the deceased wife during her exhibited good internal consistency (Cronbach's final illness. 'Final illness' was defined as 'the alpha = 0-83) and a mean item-total correlation of 0-41. In a separate sample (N = 25) of elderly illness which led to the wife's death'. widows and widowers, an excellent level of interA brief, interviewer-administered, structured questionnaire was used to rate the frequency of rater reliability (Pearson's R = 0-98, P = 00006) 22 self-reported bereavement phenomena. The and a satisfactory level of concurrent validity items in the Bereavement Phenomenology Ques- against a clinical interview (Spearman's rho = tionnaire (BPQ) (Raphael & Middleton, 1989) 0-45, P < 005) were found (Moylan et al. were derived (by factor analysis) from a much unpublished manuscript, 1993). A study of the larger group of items generated from three predictive validity of the BPQ in elderly comseparate lines of enquiry: by reference to the munity-residing widows and widowers is underinternational literature; from a survey of be- way and will be the subject of a separate report. reavement experts; and from detailed interviews of several hundred recently bereaved individuals Data analysis (Middleton et al. 1993). This atheoretical ap- Statistical analyses were undertaken with the proach produced a measure with high face Statistical Package for the Social Sciences validity. In the current study, a four-point Likert (SPSS/PC + 4-0) (Norusis, 1990), with the 416 G. J. A. Byrne and B. Raphael The Group 2 (wait-listed) widowers did not differ significantly from Group 1 widowers on age, income, education, occupational prestige or self-rated general physical health. Neither did the wait-listed group differ significantly from the RESULTS Group 1 widowers on aspects of the death. Three widowers (3-8%) withdrew and two There was no significant difference between the widowers (2-6 %) died during the first year of the mean BPQ total scores of the Group 1 and study. Baseline data on sociodemographic vari- Group 2 (wait-listed) widowers at T3 (15-1 ables, cognitive function and self-rated general (S.D. 10-6) v. 15-8 (S.D. 10-1); F(l,70) = 0-065, physical health are summarized in Table 4. Most P = 0-799), suggesting that the research widowers had left school before the age of 15 interviews had not affected the frequency of years and most were subsisting on low fixed reported bereavement phenomena. incomes. Mean MMSE score for cognitive performance was 27-38 (S.D. 1-83) out of a Prevalence of bereavement phenomena at Tl possible score of 30. Sixty-nine per cent of widowers described their general physical health Although a four-point Likert scale was used to measure self-reported frequency of bereavement as 'good' or 'excellent'. phenomena, the second point on the scale, Data on aspects of the death are summarized 'rarely', turned out to be too semantically close in Table 5. In two-thirds of cases, the duration to the first point, 'never', and was not often of the wife's final illness exceeded 1 month, and utilized by the widowers. We have therefore in one-quarter of cases it exceeded 1 year. In dichotomized our results and scored only items only 13-7 % of cases did the illness leading to the checked as 'sometimes' or 'often' as being wife's death last less than 24 h. Nevertheless, in present. 39-7% of cases the widower rated his wife's Some bereavement phenomena were reported death as 'unexpected'. In almost 60% of cases to have been experienced by more than 75 % of in which the wife did not die suddenly, the the widowers in the 2 weeks prior to the Tl widower provided the majority of care for her interview: a sense of nostalgia, intrusive during the illness which led to her death. thoughts concerning the lost person, feelings of sadness concerning the loss, distressing thoughts Table 5. Aspects of the death concerning the loss, preoccupation with images of the lost person, yearning or pining for the lost Recent Wait-listed person, and distress at reminders of the lost widowers All widowers widowers Variable N (%) N (%) N(%) person (see Table 6). Other bereavement phenomena were reported Duration of wife's final to have been experienced by between 50 % and illness > 1 year 15(28-8) 3(14-3) 18(24-7) 75 % of the widowers in the 2 weeks prior to > 1 month- < 1 year 21 (40-4) 9 (42-9) 30(411) the Tl interview: a need to talk about the lost > 1 week- < 1 month 3 (5-8) 0 (00) 3(4-1) person, crying, a sense of presence of the lost 6 (28 6) > 24 hours- < 1 week 12(16-4) 6(11-5) 3(14-3) < 24 hour 7(13-5) 10(13-7) person, looking for the lost person in familiar Degree of expectedness places and feelings of anxiety following the loss. of wife's death Some bereavement phenomena were reported Expected 23 (44-2) 8(38-1) 31 (42-5) Fairly expected 6(11-5) 2 (9-5) 8(110) to have been experienced by between 25 % and Fairly unexpected 1 (4-8) 4(7-7) 5(6-8) 50 % of the widowers in the 2 weeks prior to the 19 (36-5) Unexpected 10 (47-6) 29 (39-7) Tl interview: feelings of depression following Provision of care for the loss, acting as if the lost person was still the deceased during her final illness* alive, feelings of unreality concerning the death, Widower 5 (41-7) 27 (57-4) 22 (62-9) hallucinations of the lost person, and physical 13(37-1) Others 7 (58-3) 20 (42-6) symptoms following the loss. * This question was only asked of widowers whose wives suffered The remaining bereavement phenomena were a final illness lasting longer than one week. There were no significant reported to have been experienced by fewer than differences between recent widowers and wait-listed widowers on any 25 % of the widowers in the 2 weeks prior to of these variables. exception of confidence intervals for proportions which were calculated by the method of Fleiss (1981). Bereavement phenomena in elderly men 417 Table 6. Temporal evolution of bereavement phenomena: proportion of widowers reporting each phenomenon the previous fortnight* Tl (AT = 57) Over the past two weeks... 1 Have you found that thoughts of her keep coming into your mind? 2 Have thoughts of her made you feel distressed or upset? 3 Have you found that you are preoccupied with images or mental pictures of her? 4 Have you felt as though you have seen her, heard her, or felt as though she has touched you? 5 Have you felt as though she is still present? 6 Have you dreamt about her as though she were still alive? 7 Have you found yourself yearning or pining for her? 8 Have you found yourself looking for her in familiar places? 9 Have you felt distressed when faced with reminders of her? 10 Have you had feelings of sadness about your loss? 11 Have you had feelings of unreality following your loss? 12 Have you had feelings of anger about your loss? 13 Have you had feelings of guilt about your loss? 14 Have you had feelings of nostalgia when thinking about her? 15 Have you had feelings of anxiety following your loss? 16 Have you had feelings of depression about your loss? 17 Have you been crying about her? 18 Have you found yourself acting as though she were still alive? 19 Have you felt a need to talk about her? 20 Have you found yourself searching for her? 21 Have you found that you are experiencing physical symptoms in your body such as pain or other discomfort? 22 Have you found that you are less able to organize your daily life? 930 80-7 860 351 59-6 21-1 78-9 57-9 77-2 930 43-9 15-8 24-6 930 54-4 47-4 50-9 45-6 70-2 140 421 15-8 83-6 67-3 691 27-3 63-6 21-8 691 54-5 56-4 800 27-3 12-7 18-2 90-9 16-4 41-8 25-5 38-2 41-8 7-3 10-9 16-4 76-5 490 431 9-8 29-4 15-7 41-2 25-5 23-5 62-7 5-9 9-8 11-8 74-5 19-6 25-5 9-8 15-7 15-7 20 00 00 66-7 28-6 47-6 9-5 381 33-3 52-4 28-6 23-8 61-9 14-3 190 9-5 85-7 00 23-8 14-3 4-8 95 00 9-5 00 T2 (A1 = 5 5 ) T3 (AT = 5 1 ) T3-WL (AT = 21) • For this table, responses were dichotomized with reports of'sometimes' or 'often' being scored as present and reports of'never' or 'rarely' being scored as absent. T3-WL represents the responses of wait-listed widowers at T3. frequently at T2 than at Tl or T3, but which exhibited little temporal variation between Tl and T2. Individual bereavement-specific phenomena exhibited widely varying temporal patterns (see Table 6). Temporal evolution of bereavement phenomena On inspection of Table 6 there appears to be When mean BPQ total score is plotted against a positive association between the frequency of time there is a highly significant diminution in self-reported bereavement phenomena at Tl and the overall frequency of reported bereavement the tendency for phenomena to persist throughphenomena from Tl to T3 (Tl :33-4 (s.D. 11-9); out the first 13 months post-bereavement. To T2:24-8 (s.D. 11-7); T3:15-1 (S.D. 10-6); F(2,48) assess the significance of this observation, ranked = 98-78, P < 0-000) indicating that the BPQ is percentage change scores ((Tl score—T3 score)/ sensitive to change. Post hoc comparisons using Tl score) were compared between phenomena Tukey's test confirmed significant reductions in with high and low prevalence rates at Tl by the total BPQ score between Tl and T2, and between Mann-Whitney U test. No significant difference in the tendency of phenomena to persist was T2 and T3. Of the 22 individual bereavement phenomena found (U = 40; critical value of U = 30). Although the strong general trend of BPQ recorded by the BPQ, 19 were reported by the widowers to have occurred less frequently with scores was for them to diminish with time, a time. The exceptions were '(dreaming) about small subgroup of widowers (N = 7) reported a her as though she were still alive', '(feeling) as higher frequency of bereavement-specific phenothough she is still present' and '(finding) that mena at T2 (6 months) than at Tl (6 weeks) you are less able to organize your daily life', all (F(l,6) = 34-78, P = 0-001). These seven of which were reported to have occurred more widowers had mean BPQ scores of 27-9 (s.D. 4-6) the Tl interview: disorganization following the loss, guilt about the loss, anger about the loss, dreaming of the lost person as though she were still alive, and searching for the lost person. 418 G. J. A. Byrne and B. Raphael Absent grief Table 7. Correlations between BPQ scores and selected independent variables BPQ total score (Tl) N = 51 Age (yr)t Cognitive function (MMSE)t Occupational prestige (1-7 scale)! General physical health (1-4 scale)J Education (1-5 scale)! Income (1-5 scale)! Expectedness of wife's death (1-4 scale)! Duration of wife's final illness (1-4 scale)! -0-0229 -0-0012 01697 -01068 BPQ total score (T2) N=55 00668 -01150 01836 -01141 BPQ total score (T3) N=5l -00799 00342 -00044 -0-1444 -00881 01418 -00512 00724 -0-3582* -03167* -0-2393" -0-2482** -0-2654** -01953 -01711 -01375 t Pearson product-moment correlation; J Spearman rank correlation; *P<0-01; **P<005. at Tl, 33-6 (S.D. 5-2) at T2, and 12-1 (S.D. 8-3) at T3. On inspection of the data, the bereavement phenomena which seemed to contribute most to this peak at T2 were 'hallucinations of the lost person', 'yearning or pining for the lost person', 'a sense of presence of the lost person' and 'anger about the loss'. This late-peaking grief had a prevalence of 7/57 (12-3 %) in this cohort (95% confidence interval: 5-3%-23-9%). Persistent grief at 13 months post-bereavement The concept of 'absent grief has fascinated both clinicians and theoreticians for a long time. Our data were examined to establish whether any widowers had reported the total absence of bereavement phenomena at any of the three interviews during the first 13 months postbereavement. No widower had reported a total absence of bereavement-specific phenomena at 6 weeks post-bereavement and only one widower had reported no phenomena at 6 months and 13 months post-bereavement. Thus, the prevalence of 'absent grief at Tl in this cohort of elderly widowers (with 'absent grief being defined as the total absence of self-reported bereavementspecific phenomena on the BPQ) may be less than 1/57 (1-8%). Clearly a much larger cohort would be needed to investigate the true prevalence of such an apparently uncommon state in elderly widowers. Correlates and predictors of grief We were interested in whether a subgroup of elderly widowers with persistently elevated total scores on the BPQ throughout the first year following bereavement could be identified. Ten subjects achieved scores on the BPQ at Tl which were greater than one standard deviation above the mean. Five (50%) of these subjects were still achieving scores on the BPQ at T2 and T3 which were greater than one standard deviation above the mean. In this subgroup of five widowers, total BPQ scores decreased across time in a similar fashion to the whole cohort. The mean BPQ score for this subgroup at T3 (38-6) was similar to the mean BPQ score for the whole cohort at Tl (33-4), suggesting that these men were experiencing chronic grief. Thus, the prevalence of chronic or persistent grief among the elderly widowers in our cohort throughout the first 13 months post-bereavement was 5/57 (8-8%) (95% confidence interval: 34% -20-1%). There was no significant correlation between age or cognitive function and mean total BPQ score at Tl to T3 (see Table 7). Neither were self-rated physical health, occupational prestige or the duration of the wife's final illness significantly associated with total BPQ score during the first 13 months following bereavement. However, both education and income were negatively correlated with BPQ total score at Tl and T2. Widowers with more years of education or a higher current income reported lower mean BPQ scores, suggesting some modifying or protective effect of these sociodemographic variables. Expectedness of wife's death was negatively correlated with total BPQ score, but only at Tl. Widowers who rated their wives' deaths as 'unexpected' or 'fairly unexpected' reported higher mean total BPQ scores at Tl. Duration of wife's final illness was not significantly correlated with total BPQ score at T1-T3, although there was a linear association between duration of wife's final illness and expectedness of wife's death (Mantel-Haenszel x2 = 14-4915, df = 1, P = 0-00014). Widowers were more likely to report that their wife's death was expected if the duration of her final illness had been long. Those widowers (N = 10) who reported both that their wife's death had been 'unexpected' or 'fairly unexpected' and that the duration of Bereavement phenomena in elderly men 419 their wife's final illness had been greater than one month were identified. Mean total BPQ scores for these widowers were significantly higher at Tl and T2, but not at T3, than for the remainder of the widowers (Tl: 41-9 v. 326, F(l,48) = 5-97, P = 0-018; T2: 328 v. 24-1, /^l,48) = 4-72, P = 0-035; T3: 17-0 v. 14-7, F(l,48) = 0-40, P = 0-531). These results suggest that widowers who were unable to anticipate their wife's death, even when their wife had suffered a long final illness, have a more severe bereavement reaction. A standard multiple regression analysis was performed with Tl BPQ score as the dependent variable and income, education, and expectedness of wife's death as independent variables (IVs). As a result of missing data on some variables, the available sample size was reduced to 50 Group 1 widowers. Two of the IVs were positively skewed (education and income) and two multivariate outliers were identified in initial data screening. Analyses were run with and without the outliers and with and without logarithmic transformation of the skewed IVs. There being no significant differences between the results of these analyses, only the original analysis is presented here. Multiple R was significantly different from zero (F(3,46): 3-367, P = 0-033). Only one of the IVs contributed significantly to prediction of Tl BPQ score, expectedness of wife's death (semi-partial correlation, sr2 = 009). The three IVs in combination contributed another 009 in shared variability. However, altogether only 18% of the variability in Tl BPQ scores was accounted for by knowledge of these three IVs. Similar multiple regression analyses were performed with T2 and T3 BPQ score as the dependent variable. The three IVs did not predict T2 or T3 BPQ score. Widowers whose wives had suffered a final illness lasting longer than 1 week were asked who had provided the majority of her care. Sixteen of the 51 Group 1 widowers for whom complete T1-T3 data was available had experienced a relatively sudden bereavement and were thus excluded from this analysis. The categorical responses from the remaining 35 widowers were recorded to a dummy variable (1 = widower, 0 = other) and used in a repeated measures ANOVA. Adoption of the principal care-giving role by the widower was found not to be significantly associated with total BPQ score at Tl through T3 in Group 1 widowers (F(2,32): 2-03, P = 0-148). DISCUSSION In this study we deliberately set out to study elderly widowers, a group rarely subject to systematic research. Thus, care should be taken in generalizing the results to widows or to younger widowers. Our findings are also limited to those experiencing spousal bereavement, as our study did not address other forms of bereavement. Our sample, although one of the largest groups of elderly widowers described in this manner, is relatively small, limiting the power of some of our conclusions. In particular, our estimate of the prevalence of chronic or persistent grief is accompanied by a rather wide confidence interval. A much larger cohort would be required to measure reliably the prevalence of uncommon bereavement outcomes. Potential for bias was introduced by the participation rate of 66%, although this compares favourably with much previous bereavement research (Stroebe & Stroebe, 1989). We believe that obtaining higher voluntary participation rates in community-residing non-clinical elderly populations would be very difficult. In Australia, by no means all elderly persons have a single, identifiable general practitioner. Thus, a study aiming to look at undifferentiated community residing elderly bereaved persons could not have been conducted from general practitioner records. For practical and ethical reasons we only selected surviving spouses of women aged 63 years and over. In our State, data collected upon the registration of death does not include the current age or date of birth of surviving spouses. Thus, the alternative to the present approach would have been to contact the surviving spouses of all deceased married women, regardless of age. We considered this ethically unacceptable. Thus, it is likely that we missed a small number of elderly widowers who had been married to much younger women. However, we know of no evidence that the nature of bereavement in elderly men is a function of the conjugal age difference. Stroebe & Stroebe (1989) found that less distressed widowers were more likely to participate in bereavement research, while more distressed 420 G. J. A. Byrne and B. Raphael widows were more likely to participate. If this finding applied to the present sample, then it would be the severe end of the phenomenological spectrum which would be most likely to be under-represented. However, only 4/29 (13-8 %) of our potentially-eligible refusers reported that they felt 'too upset' to participate. There was little evidence from the responses given to us by widowers who refused to participate that they were biased towards being either severely or minimally distressed. A wide range of reasons was given for non-participation. Bereavement phenomena in this group of elderly widowers exhibited a pattern similar to that conventionally described for younger persons and for women. Some phenomena were almost universally reported, while others were less commonly reported. In particular, feelings of guilt or anger and physical symptoms were less commonly reported. There was clear evidence of ongoing bereavement phenomena at 13 months post-bereavement, with nostalgia and sadness, as well as intrusive thoughts about the deceased person, being reported by more than half of the widowers at this time. A subgroup of elderly widowers was identified which exhibited chronic or persistent grief, in that their mean BPQ total score at 13 months post-bereavement was similar to that of the total group at 6 weeks post-bereavement. These individuals consisted of half of a group of widowers who, at 6 weeks post-bereavement, had BPQ total scores more than one standard deviation above the mean for the group. This suggests a strategy for identifying individuals at high (50 %) risk of persistent grief by means of a brief questionnaire. If absent grief does occur in elderly widowers, we were not able to demonstrate it in our relatively small cohort. While absent grief may be a genuinely rare condition, widowers with absent grief may have declined to participate in the study. It is also possible that the social desirability of reporting bereavement phenomena is so strong that entirely negative responses are unlikely even among widowers with absent grief. In a correlational analysis, education and income were found to be associations of low BPQ score, suggesting a protective role for these variables. Widowers who reported their wives' deaths as 'unexpected' or 'fairly unexpected' reported a higher frequency of bereavement- specific phenomena at T l . Multiple regression analysis demonstrated that expectedness of the wife's death was a more important predictor of Tl BPQ score than income or education. This finding is in keeping with the observation of Clayton et al. (1973) who found that anticipating the death of a spouse had a positive effect on immediate post-bereavement depressive symptoms. Widowers who did not expect their wife's death despite a long final illness had more severe bereavement reactions. Care-giving by widowers prior to bereavement was not associated with increased total BPQ score. Future reports will examine the relationship between bereavement-specific phenomena and psychiatric diagnoses, standard measures of psychopathology, health-related behaviour, and biological measures. The authors gratefully acknowledge the assistance of Mrs Kathy Jong and Mrs Neroli Whiteman. This research was supported by grants from the National Health and Medical Research Council and the Australian Rotary Health Research Fund. REFERENCES Bass, D. M. & Bowman, K. (1990). The transition from caregiving to bereavement: the relationship of care-related strain and adjustment to death. Gerontologist 30, 35-42. Blazer, D. G., Bachar, J. R. & Manton, K. G. (1986). Suicide in late life. Review and commentary. Journal of the American Geriatrics Society 34, 519-525. Bowling, A. & Benjamin, B. (1985). Mortality after bereavement: a follow-up study of a sample of elderly widowed people. Biology and Society 2, 197-203. Bruce, M. L., Kim, K., Leaf, P. J. & Jacobs, S. (1990). Depressive episodes and dysphoria resulting from conjugal bereavement in a prospective community sample. American Journal of Psychiatry 147,608-611. Caserta, M.S., Lund, D. A. & Dimond, M. F. (1985). Assessing interviewer effects in a longitudinal study of bereaved elderly adults. Journal of Gerontology 40, 637-640. Clayton, P. J., Halikas, J. A., Maurice, W. L. & Robins, E. (1973). Anticipatory grief and widowhood. British Journal of Psychiatry 122,47-51. Daniel, A. (1984). The measurement of social class. Community Health Studies 8, 218-222. Emmerson, J. P., Burvill, P. W., Finlay-Jones, R. & Hall, W. (1989). Life events, life difficulties and confiding relationships in the depressed elderly. British Journal of Psychiatry 155, 787-792. Feinson, M. C. (1986). Aging widows and widowers: are there mental health differences? International Journal of Aging and Human Development 23, 241-255. Fleiss, J. L. (1981). Statistical Methods for Rates and Proportions. John Wiley & Sons: New York. Folstein, M. F., Folstein, S. E. & McHugh, P. R. (1975). 'MiniMental State': a practical method for grading the cognitive state of patients for the clinician. Journal of Psychiatric Research 12, 189-198. Gallagher, D., Lovett, S., Hanley-Dunn, P. & Thompson, L. W. (1989). Use of select coping strategies during later-life spousal Bereavement phenomena in elderly men bereavement. In Older Bereaved Spouses (ed. D. A. Lund), pp. 111-121. Hemisphere Publishing Corporation: New York. Gallagher, D. E., Thompson, L. W. & Peterson, J. A. (1981). Psychosocial factors affecting adaptation to bereavement in the elderly. International Journal of Aging and Human Development 14, 79-95. Gass, K. A. & Chang, A. S. (1989). Appraisals of bereavement, coping, resources, and psychosocial health dysfunction in widows and widowers. Nursing Research 38, 31-36. Gerber, 1., Rusalem, R., Hannon, N., Battin, D. & Arkin, A. (1975). Anticipatory grief and aged widows and widowers. Journal of Gerontology 30, 225-229. Goldberg, E. L., Comstock, G. W. & Harlow, S. D. (1988). Emotional problems and widowhood. Journal of Gerontology 43, 206-208. Heyman, D. K. & Gianturco, D. T. (1973). Long-term adaptation by the elderly to bereavement. Journal of Gerontology 28, 359-362. Hill, D. C , Thompson, L. W. & Gallagher, D. (1988). The role of anticipatory bereavement in older women's adjustment to widowhood. Gerontologist 28, 792-796. Irwin, M. & Weiner, H. (1987). Depressive symptoms and immune function during bereavement. In Biopsychosocial Aspects of Bereavement (ed. S. Zisook), pp. 159-174. American Psychiatric Press, Inc.: Washington, DC. Jacobs, S. & Ostfeld, A. (1977). An epidemiological review of the mortality of bereavement. Psychosomatic Medicine 39, 344-357. Jacobs, S., Kasl, S., Ostfeld, A., Berkman, L. & Charpentier, P. (1986). The measurement of grief: age and sex variation. British Journal of Medical Psychology 59, 305-310. Kim, K. & Jacobs, S. (1991). Pathologic grief and its relationship to other psychiatric disorders. Journal of Affective Disorders 21, 257-263. Lund, D. A., Dimond, M. F., Caserta, M. S., Johnson, R. J., Poulton, J. L. & Connelly, J. R. (1985). Identifying elderly with coping difficulties after two years of bereavement. Omega 16, 213-224. Lund, D. A., Caserta, M. S. & Dimond, M. F. (1986). Gender differences through two years of bereavement among the elderly. Gerontologist 26, 314-320. Middleton, W., Moylan. A.. Burnett, P. & Martinek, N. (1993). An international perspective on bereavement related concepts. Australia and New Zealand Journal of Psychiatry 27, 457-463. Moylan, A.. Byrne, G. J. A. & Middleton, W. (1993). Psychometric 421 properties of a questionnaire to measure bereavement phenomena. (Unpublished manuscript.) Department of Psychiatry, University of Queensland, Brisbane, Australia. Murrell, S. A., Himmelfarb, S. & Phifer, J. F. (1988). Effects of bereavement/loss and pre-event status on subsequent physical health in older adults. International Journal of Aging and Human Development 27, 89-107. Niemi, T. (1979). The mortality of male old-age pensioners following spouse's death. Scandinavian Journal of Social Medicine 7,115-117. Norusis, M.J. (1990). SPSS/PC+4.0 Base Manual. SPSS Inc.: Chicago. Parkes, C. M. & Brown, R. J. (1972). Health after bereavement. A controlled study of young Boston widows and widowers. Psycho* somatic Medicine 34, 449—461. Raphael, B. & Middleton, W. (1989). Phenomenology of adult bereavement. (Unpublished manuscript.) Department of Psychiatry, University of Queensland, Brisbane, Australia. Sheldon, A. R., Cochrane, J., Vachon, M . L . S., Lyall, W. A. L., Rogers, J. & Freeman, S. J. J. (1981). A psychosocial analysis of risk of psychological, impairment following bereavement. Journal of Nervous and Mental Disease 169, 253-255. Shulman, K. (1978). Suicide and parasuicide in old age: a review. Age and Ageing 7, 201-209. Stein, Z. & Susser, M. (1969). Widowhood and mental illness. British Journal of Preventive and Social Medicine 23, 106-110. Stroebe, M. S. & Stroebe, W. (1989). Who participates in bereavement research? A review and empirical study. Omega 20, 1-29. Stroebe, W. & Stroebe, M. S. (1987). Bereavement and Health. The Psychological and Physical Consequences of Partner Loss. Cambridge University Press: Cambridge. Vachon, M. L. S., Rogers, J., Lyall, W. A., Lancee, W. L., Sheldon, A. R. & Freeman, S. J. J. (1982). Predictors and correlates of adaptation to conjugal bereavement. American Journal of Psychiatry 139, 998-1002. Van Zandt, S., Mou, R. & Abbott, D. (1989). Mental and physical health of rural bereaved and nonbereaved elders: a longitudinal study. In Older Bereaved Spouses (ed. D. A. Lund), pp. 25-35. Hemisphere Publishing Corporation: New York. Zisook, S., Schucter, S. R. & Lyons, L. E. (1987a). Predictors of psychological reactions during the early stages of widowhood. Psychiatric Clinics of North America 10, 355-368. Zisook, S., Schucter, S. R. & Lyons, L. E. (19876). Adjustment to widowhood. In Biopsychosocial Aspects of Bereavement (ed. S. Zisook), pp. 51-74. American Psychiatric Press: Washington, DC.