Article Text

PDF
Does the risk of hospitalisation for ischaemic heart disease rise already before widowhood?
  1. Elina Einiö1,
  2. Heta Moustgaard1,
  3. Pekka Martikainen1,2,3,
  4. Taina Leinonen1,4
  1. 1Population Research Unit, Department of Social Research, University of Helsinki, Finland
  2. 2Centre for Health Equity Studies (CHESS), Stockholm University/Karolinska Institutet, Stockholm, Sweden
  3. 3Max Planck Institute for Demographic Research, Rostock, Germany
  4. 4Finnish Institute of Occupational Health, Helsinki, Finland
  1. Correspondence to Dr Elina Einiö, Population Research Unit, Department of Social Research, University of Helsinki, Helsinki, P.O.Box 18, FIN-00014 Finland; elina.einio{at}helsinki.fi

Abstract

Background The death of a spouse has been shown to increase mortality from various causes, including ischaemic heart disease. It is unclear, however, whether cardiac problems are already on the rise before widowhood.

Methods Using longitudinal register data of Finnish widows-to-be aged 65 and over at baseline (N=19 185), we assessed the risk of hospitalisation for ischaemic heart disease 18 months before and after widowhood. Hospital admissions were derived from national hospital discharge registers between 1996 and 2002. Analyses used population-averaged and fixed-effects logistic models, the latter of which controlled for unobserved time-invariant characteristics, such as genetic susceptibility, personality and behavioural and medical history.

Results For men, fixed-effects model revealed that hospitalisation for ischaemic heart disease increased twofold already 0–3 months prior to the death of a spouse (OR=2.09, 95% CI 1.22 to 3.60), relative to the period of 15–18 months before widowhood. It stayed at a heightened level up to 6 months following bereavement (OR=2.15, 95% CI 1.07 to 4.30). Among women, the fixed-effects analysis detected no statistically significant increase in hospitalisation for ischaemic heart disease before or after widowhood.

Conclusions These findings indicate that men are already vulnerable to cardiac problems before the death of a wife. Medical interventions and health counselling could be targeted to the husbands of terminally ill patients, in order to improve their cardiovascular health over the transition to widowhood.

Statistics from Altmetric.com

Introduction

Several studies show that the risk of death from all causes,1–8 unnatural causes,9–11 certain cancers and ischaemic heart diseases10 ,11 rise after the death of a spouse. Most studies indicate that the risk of death is highest during the first weeks6 ,9 or months1 ,2 ,4 ,7 ,8 ,12 ,13 following the spouse's death, while one study from the USA indicates the highest risk as late as 7–12 months following bereavement.5 Despite contextual and methodological differences between the studies, the effect of a broken heart after having lost a spouse has become a widely accepted paradigm in medical and social sciences.11 ,14 ,15 One recent study from the USA even showed that the likelihood of an acute myocardial infarction (MI) was elevated during the first month following the death of a significant person.16 It has been suggested that bereavement increases susceptibility to physical illnesses,14 which can be observed in an older individual's immune system shortly after the loss.17 ,18 Most studies, however, compare bereaved individuals with their non-bereaved counterparts,14 as if the death of a spouse was an external unexpected shock for the surviving spouse. Yet, the death of a fragile or ill spouse may be expected, and its anticipation may already trigger heart problems.

While having many similar characteristics to those of the typical grief reaction after the death of a spouse, anticipatory grief has a unique characteristic of groundless hope, as the exact timing of spousal death is still uncertain.19 Furthermore, the process of widowhood may involve intensified caregiving burdens before the loss,20 and reduce participation in other activities. Widowhood is thus not necessarily an unexpected external shock for the surviving spouse; rather, it is a complex process that could cause acute and chronic stress even prior to the actual loss. Although acute stress has previously been linked to changes in the nervous, endocrine and immune systems following a stressful event, chronic stress has been shown to be especially harmful for older people due to the loss of the immune function associated with ageing.21 Therefore, the overall process of widowhood could cause severe cardiac manifestations of deteriorating well-being.

To the best of our knowledge, no study has analysed cardiovascular health shortly before widowhood, which could provide an opportunity to enhance our understanding of the complex process of losing a spouse. Most studies that examine the association between prewidowhood and health lack repeated measures of health, or focus on the analysis of depressive symptoms. In a study of Americans over 50 years old, Vable et al22 showed that widows-to-be had worse depressive symptoms, mobility and cognitive problems than those continuously married. Similarly, Williams et al23 suggested that widows-to-be differed from those continuously married by demonstrating their higher levels of anxiety and depressive symptoms, as well as poorer self-reported health. Also the few studies using repeated measures of health before and after widowhood indicate that depressive symptoms are already on the rise before the death of a spouse.24 ,25 As depression and emotional distress in general are suggested to predispose individuals to ischaemic heart disease,26 ,27 it is likely that heart problems also aggravate or occur prior to widowhood.

This study used unique register-based longitudinal data on Finnish widows to analyse the risk of hospitalisation for ischaemic heart disease 18 months before and 18 months after the death of a spouse, at 3-month intervals. We used (1) a standard population-averaged model that controls for observed characteristics, such as education, gender, age and calendar year, and (2) a fixed-effects logistic model that also controls for unobserved time-invariant characteristics of the individuals, such as genetic makeup, personality and behavioural and medical history. The fixed-effects approach reduces the selection bias related to unobserved differences between the individuals, since each individual serves as his or her own control.28

Methods

Data

The data were based on a 40% random sample of older persons aged 65 and older who resided in Finland on 31 December, 1997. The sample from Statistics Finland was linked with death records and hospital discharge records from the National Research and Development Centre for Welfare and Health. The sample was further linked to the medicine reimbursement data of the Social Insurance Institution (permission number: TK-53-1236-15). The linkage was carried out using 11-digit personal identification codes.

Of the 301 265 persons aged 65 and over captured in the random sample, 45.2% lived with a spouse on 31 December, 1997. To focus our analyses on those who experienced spousal bereavement, we excluded those whose spouse survived from 1 January, 1998 to 31 December, 2002. We also excluded those few persons who died on the same day as their spouse. Thus, the analytical sample (standard population-averaged model) consisted of 19 185 persons whose spouse died within the 5-year follow-up period. The subsample for the second analysis (individual fixed-effects models) included those 1611 widows who had at least one hospital admission for ischaemic heart disease 18 months before or after the death of their spouse. The restriction of the analysis to this subsample was due to the constraints of fixed-effects modelling, which require within-individual variation in the outcome. The characteristics of the samples are shown in table 1. Of the 19 185 widows-to-be persons, 8.4% were hospitalised for ischaemic heart disease 18 months before or after widowhood (table 1). A corresponding 8.4% of the total population aged 65 and over resident in private households at baseline were hospitalised for ischaemic heart disease during a 36-month follow-up period from 1 January, 1998 to 31 December, 2000.

Table 1

Sample characteristics of persons who experienced the death of a spouse between 1998 and 2002, Finland

Ischaemic heart disease

The data on the principal cause of hospitalisation were based on the Tenth Revision of the International Statistical Classification of Diseases and Related Health Problems (ICD-10).29 Hospitalisation for ischaemic heart disease included admissions due to the following diagnoses: angina pectoris (ICD-10 code I20), MIs (I21-I22), complications following MIs (I23), other acute ischaemic heart diseases (I24) and chronic ischaemic heart diseases (I25). The diagnoses and admission dates were identified from the national hospital discharge records between 1996 and 2002.

For sensitivity analyses, we used the medicine reimbursement data provided by the National Insurance Institution to assess whether a person had been diagnosed with chronic ischaemic heart disease before the follow-up period. The study persons were classified as having prior chronic ischaemic heart disease if they were ever granted the right to receive reimbursement for drug costs due to their chronic ischaemic heart disease by December, 1995 (Special Refund Category Code 206).30 The medicine reimbursement system covers all residents of Finland, regardless of age and wealth.31

Statistical analysis

Individuals were followed for 18 months before and 18 months after the death of their spouse. The proximity to widowhood was divided into six 3-month (91-day) intervals around the date of widowhood (altogether 12 intervals). The outcome was categorised as 1 if a study person had at least one hospital admission for ischaemic heart disease during a 3-month interval and 0 otherwise. If a study person died during an interval, censoring occurred at the beginning of the next interval.

We first used a standard population-averaged logistic model to predict the adjusted 3-month prevalence of hospitalisation for ischaemic heart disease before and after widowhood. These standard models use between-individual and within-individual variation in the exposure and outcome to estimate the effect of the proximity to widowhood. Level of education and gender were added to the model as time-invariant covariates, and calendar year and age as time-varying covariates. The results are presented as the adjusted prevalence of hospitalisation for ischaemic heart disease within each 3-month interval before and after widowhood by gender (figure 1). The unadjusted prevalences were not of notable interest to us, and they are in online supplementary appendix 1.

Supplementary appendix 1

Figure 1

Hospitalisation for ischaemic heart disease 18 months before and after the death of a spouse by gender, adjusted (a) 3-month prevalence (%), Finnish widows, N=19,185. Note: (a) The population-averaged model controlled for age and calendar year (time-varying), gender and education (time-invariant), proximity to widowhood and the interaction between proximity to widowhood and gender.

Second, we used individual-level fixed-effects logistic models to analyse hospitalisation for ischaemic heart disease before and after the death of a spouse. The fixed-effects approach controlled for all unobserved time-invariant characteristics of individuals, such as genetic makeup, personality and behavioural and medical history, and allowed for controlling of observed time-varying covariates, such as age and calendar year. Despite several advantages, fixed-effects models do not control for unobserved time-varying characteristics of the individuals (eg, changes in health behaviour).28 ,32 Fixed-effects models were fitted separately for men and women, and the results are presented as ORs with their 95% CIs. The first 3-month interval, from 15 to 18 months before widowhood, was chosen as the reference category (figures 2 and 3).

Figure 2

ORs for men of hospitalisation due to ischaemic heart disease 18 months before and after the death of a wife (ref.=15–18 months before), fixed-effects model (a), widowed men, N=650. Note: (a) The fixed-effects logistic model controlled for observed time-varying age and calendar year and all unobserved time-invariant confounders.

Figure 3

ORs for women of hospitalisation for ischaemic heart disease 18 months before and after the death of a husband (ref.=15–18 months before), fixed-effects model (a), widowed women, N=961. Note: (a) The fixed-effects logistic model controlled for observed time-varying age and calendar year and all unobserved time-invariant confounders.

The Hausman test was used to determine whether a simpler random-effects model could have been used instead of a fixed-effects model. The test showed that an individual-level fixed-effects model was a better choice to analyse hospitalisations for ischaemic heart disease (men: p<0.0001; women: p=0.0107).

For a sensitivity analysis, we fitted a population-averaged model with three-way interactions (proximity to widowhood, gender and prior chronic ischaemic heart disease) to examine whether the association between the proximity to widowhood and hospitalisation for ischaemic heart disease was different among those who had had prior chronic ischaemic heart disease before the follow-up period. The predicted prevalences are presented by gender in online supplementary appendix 2.

Supplementary appendix 2

Observed control variables

We used age, calendar year, gender and level of education as observed control variables. Age and calendar year were measured at each 3-month interval. Age was calculated from the date of birth and expressed in completed years. Gender and education, which were measured on 31 December, 1997, were added to the models as time-invariant categorical variables. The educational categories were based on the highest educational qualification or degree completed. The three categories were: tertiary education, intermediate education and basic education or less.

Results

Population-averaged model

Among men, the prevalence of hospitalisation for ischaemic heart disease showed a marked increase a few months before the death of a wife, independent of observed confounders such as age, calendar year and education (figure 1). The 3-month prevalence was between 1.0 and 1.1% from 3 to 18 months before widowhood, increased to 1.9% from 0 to 3 months prior to widowhood and stayed above 1.5% for the 12 months following the death of a wife.

Among women, the adjusted 3-month prevalence of hospitalisation for ischaemic heart disease was between 0.6 and 0.8% from 9 to 18 months before the death of a spouse and rose gradually to 1.1% from 0 to 3 months before widowhood. This increase observed among women shortly before widowhood was less pronounced than that among men. Additionally, the prevalence for women stayed above 1.0% for at least the 18-month period after widowhood.

The sensitivity analysis indicated that the modest increase in female hospitalisation observed before widowhood was largely related to the vulnerability of women who already had prior chronic ischaemic heart disease before the follow-up period, whereas male hospitalisation showed a marked increase shortly before widowhood, regardless of prior chronic ischaemic heart disease (see online supplementary appendix 2).

Fixed-effects models

For men, the risk of hospitalisation due to ischaemic heart disease showed a clear increase shortly before widowhood, independent of unobserved time-invariant characteristics, as well as the time-varying measures of age and calendar year (figure 2). The odds of being hospitalised 0–3 months prior to widowhood was twofold higher compared with that observed earlier, 15–18 months before the death of a wife (OR=2.09, 95% CI 1.22 to 3.60). The odds of hospitalisation for ischaemic heart disease remained at a heightened level for 6 months following the death of a wife (OR=2.15, 95% CI 1.07 to 4.30), after which it was no longer significantly higher, compared with the reference period of 15–18 months prior to widowhood (6–9 months after widowhood: OR=2.14, 95% CI 0.99 to 4.63, 10–12 months after widowhood: OR=2.30, 95% CI 0.98 to 5.39).

Among women, fixed-effects model revealed that hospitalisation for ischaemic heart disease was not significantly different with proximity to widowhood (figure 3). However, the odds of hospitalisation seemed somewhat elevated 0–3 months before widowhood (OR=1.37, 95% CI 0.87 to 2.15) and 0–3 months after widowhood (OR=1.45, 95% CI 0.87 to 2.41), compared with that observed 15–18 months prior to widowhood.

Discussion

The novel contribution of our study was to shed new light on the complex process of losing a spouse with regard to changes in cardiovascular health. In many cases, widowhood is not an unexpected exogenous shock for the surviving spouse but can include an anticipatory period before the loss, the actual death and the time after the loss. However, most studies on widowhood do not include the period before the loss in the assessment of health changes.

This study was the first to include the period before a spouse's death in the assessment of the relationship between ischaemic heart disease and the proximity to widowhood. Using unique longitudinal data of Finnish widows and widowers, our analysis showed that, for men, the risk of hospitalisation due to ischaemic heart disease increased markedly 0–3 months prior to the death of a wife, and remained at a heightened level for 6 months following the loss. This pattern was similar in the population-averaged and fixed-effects model for men. On the contrary, among women, the prevalence of hospitalisation for ischaemic heart disease appeared to increase a few months prior to widowhood in the population-averaged model, but this result could not be verified with the fixed-effects approach which accounted for unobserved time-invariant characteristics, such as genetic makeup, personality and medical and behavioural history. Hence, this study demonstrated a clear relationship between the proximity to widowhood and hospitalisation for ischaemic heart disease among men only (figure 2 vs figure 3).

Our findings for men are consistent with results from a previous study in the USA showing an increased risk of heart problems after the death of a significant person.16 In their study of surviving MI patients who were able to complete a structured interview, Mostofsky et al16 indicated that the risk of an acute MI was elevated within 24 hours of the death of a significant person, declining steadily on each subsequent day, but remained elevated for at least 1 month following the loss. The result of our study showed that men's risk of hospitalisation for ischaemic heart disease, which also included acute MIs, remained elevated up to 6 months following bereavement, compared with that observed 15–18 months prior to widowhood. However, in addition to MIs, our study also included chronic ischaemic heart disease, the mortality of which has been shown to be elevated for at least 6 months following the death of a spouse, compared with individuals with a spouse who is still alive.10

As the first study looking at the period before widowhood, our results indicate that men's risk of hospitalisation for ischaemic heart disease began rising even prior to widowhood. This prewidowhood increase cannot be compared with the findings of Mostofsky et al16 or those of mortality studies10 ,11 because these lack repeated measures of health outcomes before bereavement. However, our findings for men are somewhat consistent with the results of an earlier Dutch study on healthcare expenditures 42 months before and after widowhood suggesting that the average expenditure level rises a few months before widowhood among individuals over 80 years old, and remains at a heightened level after widowhood.20 Rolden et al20 suggested that the expenditure levels predominantly rose because of an increased need for formal care after the death of an informal caregiver, rather than an increased need for medical treatment related to the health impact of bereavement. This argument was justified by the fact that the increase in expenditure levels after widowhood was especially high in the long-term care sector. However, they did suggest that the increase in expenditure levels observed 7 months before widowhood, among the oldest widows-to-be, could be related to the consequences of emotional stress and a higher burden of informal care. Additionally, in spite of a longitudinal study design, the Dutch study had no information on the diagnoses that were associated with the increase in expenditure levels over the transition to widowhood. The present study adds to the existing knowledge by showing that older men's need for healthcare shortly prior to widowhood was related to ischaemic heart disease.

In our study, there are several possible explanations for the elevated risk of hospitalisation due to ischaemic heart disease observed in men 0–3 months before the death of a wife. First, psychosocial mechanisms that are suggested to operate after the death of a spouse may well be in place before the actual loss. The surviving husband may be aware of the approaching death of his wife and experience an increase in psychological stress, life dissatisfaction and depressive symptoms,24 ,25 which can increase susceptibility to physical illnesses.15 Additionally, the fear of losing a spouse may cause insomnia, which has been shown to increase the risk of an acute MI.33 Furthermore, a distressed husband may experience high blood pressure or develop harmful coping mechanisms, such as an increase in smoking and alcohol intake, which can affect his cardiovascular health.34–36

Second, it is possible that a spouse's approaching death intensifies informal caregiving burdens,20 which can cause health problems for the caregiver. Caregiving has previously been suggested to increase the risk of ischaemic heart disease.37 However, a recent study from England indicated that caregiving was only associated with ischaemic heart disease among caregivers who were already in poor health at baseline.38 In our sample, the surviving husbands were, on average, over 75 years old at the study entry; thus, the caregiving burden could be of importance in explaining the results for the widowers.

Finally, our results may also be consistent with a scenario in which the heart problems of a husband contribute to the death of a wife. The hospitalisation of a husband for a non-fatal heart disease may leave a fragile wife without a caregiver, may cause depression and anxiety, and, in the most severe case, precipitate the process leading to her death. Accordingly, Christakis and Allison39 showed that a spouse's hospitalisation for ischaemic heart disease was associated with an increased risk of death among women, but not men, in the USA. Furthermore, in a recent Danish study, Fosbol et al40 showed that hospitalisation of a spouse for an acute MI increased the use of antidepressants and benzodiazepine medication in the non-hospitalised spouse after the event, even among those whose spouse survived. A spouse's hospitalisation for other non-fatal causes had no such effect. Fosbol et al40 argued that a heart attack has psychological consequences for a spouse, because he or she has to live through the serious event alongside the patient. These psychological consequences may be devastating for an already ill or fragile person, possibly accelerating the natural progression of a disease, leading to an earlier death.

There are several possible explanations for the non-significant finding for women in the fixed-effects analysis. First, it seems reasonable to assume that the different findings on gender relate to men's higher tendency to have more complicated clinical manifestations of atherosclerosis,41 which leads to immediate hospitalisation, whereas stress attendant on widowhood could cause women less severe physiological changes. This argument is in line with a recent Norwegian study, in which Albrektsen et al42 showed that men have a higher risk of incident MI's than do women, even independent of established risk factors, such as age, high blood pressure, cholesterol level, smoking, body mass index, physical inactivity and diabetes. It is thus possible that the same risk factor of stress could cause men more severe cardiac problems due to biological mechanisms that are not yet fully understood.

Second, it is possible that the lower rate of female hospitalisation prior to widowhood relates to women's being less likely than men to anticipate a spouse's death. This idea is consistent with well-known differences in the cause-of-death structures between the sexes in the general population, with accidental, violent and MI deaths being more prevalent among men. However, these distributional differentials are relatively small for married older individuals,43 and prior studies also show that the impact of a spouse's cause of death on the health of the surviving spouse does not markedly vary between men and women.6 ,11 We thus believe that these differentials are too small to account for gender difference in hospitalisation before spousal death. Unfortunately, our data did not contain information on the spouse's cause of death, which could have provided indirect information on whether the death was anticipated before its actual occurrence. Thus, examining the contribution of a spouse's cause of death offers promise as a direction for future studies.

Third, the main limitation of the fixed-effect models is the imprecision of estimates as compared with the standard models due to reduced statistical power.28 ,44 Thus, a larger sample size of widow-to-be women would have been required to draw firmer conclusions on prewidowhood having no harmful effect on female cardiovascular health.

The main strength of our study was the identification of the date of widowhood and of the hospitalisations for ischaemic heart disease before the death of a spouse. Diagnoses based on hospital registers are less vulnerable to self-report bias than surveys and the attrition rate is minimal. Of course, minor symptoms of ischaemic heart disease, which did not lead to hospital use, were not captured. The fixed-effects models used in the study reduced the selection bias related to unobserved differences between the individuals, since each individual served as his or her own control. Controlling for unobserved characteristics is important because they are likely to be correlated with the timing of widowhood and health, and hence bias the standard estimates. For example, time-invariant behavioural characteristics, personality and medical history could contribute to an individual's health and the timing of a spouse's death. However, unobserved time-varying characteristics of an individual, such as behavioural changes that are independent of widowhood, could still confound the results. Unfortunately, our register-based data did not contain any information on an individual's or a couple's health behaviour. It is plausible, however, that most short-term behavioural changes occurring during the transition to widowhood are related to the process of widowhood. Therefore, more research is needed to assess changes in stress, for example, biomarkers such as blood pressure, and various health behaviours among the surviving spouses prior to the death of a spouse. In addition, examining changes in other aspects of physical health is required to broaden our understanding of the overall process of widowhood.

Conclusions

In summary, we found that men's risk of hospitalisation for ischaemic heart disease increased 0–3 months before the death of a wife, and remained at a heightened level for 6 months following widowhood. This result was independent of the influences of individuals' unobserved time-invariant characteristics, such as genetic susceptibility, personality and behavioural and medical history, and of observed time-varying characteristics, such as age and calendar year. These results suggest that men are already physically vulnerable to bereavement-related cardiac problems prior to the actual loss of a wife. Medical interventions and health counselling could be targeted to the husbands of terminally ill patients, in order to improve their cardiovascular health over the transition to widowhood.

What is already known on this subject?

  • Mortality from ischaemic heart disease is known to be elevated after widowhood.

  • It is unclear whether cardiac problems are already on the rise before widowhood.

What this study adds

  • This is the first study to assess the risk of hospitalisation for ischaemic heart disease shortly before widowhood.

  • This study uses fixed-effects models to control for the confounding influences of unobserved time-invariant characteristics, such as genetic susceptibility, personality and behavioural and medical history.

  • The findings indicate that men are already vulnerable to cardiac problems a few months before the actual loss of a wife.

References

View Abstract

Footnotes

  • Contributors EE planned the study, conducted the analyses and wrote the paper. HM contributed to planning and revision of the study. PM and TL contributed to revising the paper.

  • Funding The study was funded by the Academy of Finland and Signe and Ane Gyllenberg Foundation.

  • Competing interests None declared.

  • Ethics approval The permission number for the register-based data is TK-53-1236-15.

  • Provenance and peer review Not commissioned; externally peer reviewed.

  • Data sharing statement The longitudinal register-based data file can be accessed at Statistics Finland by agreeing to confidential conditions and restrictions of the register authorities involved (TK-53-1236-15).

Request permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.