SIM AI

A STATISTICAL STUDY OF INCREASED
MORTALITY AMONG WIDOWERS

A STATISTICAL STUDY OF INCREASED
MORTALITY AMONG WIDOWERS

Have you ever wondered what it really means when someone tells you they have a “broken heart”? Most people understand that having a “broken heart” references a state of extreme grief or sorrow, typically caused by the death of a loved one or the ending of a romantic relationship. But is there any additional truth to this saying? This study examines the medical implications of having a ‘broken heart’ by establishing a correlational relationship between the loss of a romantic partner and the mortality rates of widowers due to heart-related diseases. By tracking the mortality rates of 4,486 widowers of 55 years of age and older, over a period of 9 years, the researchers made some interesting findings: that within the first 6 months of bereavement, the mortality rate for widowers was 40% above that for matched controls, and that the greatest increase in mortality during these first 6 months came from heart-related diseases. No significant conclusions could be made about the relationship between the cause of wife’s death and the cause of her widower’s, nor about the impact of social class on widower’s mortality rate. We also examine the study’s methodology and provide reasonable recommendations to some of its shortcomings.

One practical application of statistics and probability is in the area of medical research. In ‘Broken Heart: A Statistical Study of Increased Mortality Among Widowers’, published by the British Medical Journal, researchers wanted to find out if one could really ‘die of a broken heart’. To test this theory, they decided to obtain data on the mortality of recently bereaved widowers, and examine if their cause of death was indeed from a heart-related illness. Due to the difficulty in manipulating such a study, not to mention its ethical considerations, the researchers settled on conducting an Observational Study. This was done by identifying the population of interest as it naturally occurred, then obtaining data without manipulating the group. Researchers identified widowers aged 55 or older whose wives had died in 1957. Following this, they obtained data on their mortality over a period of 9 years, without any interference or manipulation. The researcher’s population of interest were recently bereaved widowers aged 55 years or older. As the study was conducted in England and Wales, the sample population were widowers of 55 years of age or older in England and Wales. The sample size of 4486 was set based on the total number of candidates identified from the death certificates of their wives. The widowers were chosen through Simple Random Sampling. All widowers from England and Wales had an equal chance of being chosen, since death certificates are issued to all citizens of the country. In proposing the theory that one can die of a broken heart, the researchers put 2 sets of hypotheses to the test. These were: Null Hypothesis 1 (H0): Sample mean, μ = μ0
Alternative Hypothesis 1 (H1): Sample mean, μ > μ0 (In both hypotheses, Sample mean, μ, refers to the number of widowers who died from a heart-related illness within the first 6 months.) Interestingly, the researchers attempted to test another related theory in the course of the study. In the cases where the wives and widowers died from the same disease group, they wanted to know if it was anything other than coincidence. The null hypothesis in this case is the expected number of similar deaths based on pure chance alone, which is in other words, being due to chance association. To test the theory that this phenomenon was not due to chance association, the researchers came up with another set of hypotheses. These were: Null Hypothesis 2 (H0): Sample proportion, p = p0
Alternative Hypothesis 2 (H1): Sample proportion, p > p0 (In both hypotheses, p refers to the proportion of wives and their widowers who died from the same cause of death.) The researchers then examined the relationship between social class and the increase in mortality rate of widowers in their first six months of bereavement. Null Hypothesis 3: Sample proportion, p = p0
Alternative Hypothesis 3: Sample proportion, p ≠ p0 (In both hypotheses, p refers to the proportion of deaths of the widowers as is expected by their social class.)

Conclusion 1 (by the study): Reject Null Hypothesis 1. A widower in bereavement does have a higher chance of dying from heart diseases. Firstly, the researchers examined if there was indeed an increase in mortality rates for recently bereaved widowers as compared to the control group of married men. Their findings are displayed in Table 1 and Figure 1, which shows that there is a huge increase in mortality rates in the 1st year compared to the 9 years which they ran the study. In fact, the mortality rate dropped to around the same level as the control group after the 1st year, after which it even fell below the control group mortality rate. This shows that the initial mortality rate of the widowers was abnormal, having taken a sharp increase immediately after their wives’ deaths. | | | | | Coronary Thrombosis and other arteriosclerotic and degenerative heart disease | 77 | 46 | Influenza, pneumonia, and bronchitis | 29 | 20 | Other heart and circulatory diseases | 24 | 15 | Vascular lesions affecting C.N.S. | 22 | 22 | Cancer of other sites | 22 | 19 | Cancer of lung and bronchus | 8 | 7 | Infectious diseases | 4 | 1 | Other causes | 27 | 23 | Total | 213 | 153 | Table 1 – Cause of death among the 213 Widowers who died within six months of their wives compared with the number expected from the mortality rate of married males of the same age in England and Wales during 1957
-10
-5
0
5
10
15
20
25
30
35
40
ST
1
2
ND
3
RD
4
TH
5
TH
6
TH
7
TH
8
TH
9
TH
PERCENTAGE DIFFERENCE
Year Since Bereavement
-10
-5
0
5
10
15
20
25
30
35
40
ST
1
2
ND
3
RD
4
TH
5
TH
6
TH
7
TH
8
TH
9
TH
PERCENTAGE DIFFERENCE
Year Since Bereavement

Figure 1 – Percentage difference between mortality rate of widowers over 54 and that of married men of the same age, by years of bereavement. (N = 4,486) Rate as proportion of rate for married men of same age | | | |
1.71.651.61.551.51.451.41.351.31.251.2 C . T & O T H E R A . S . H . D O T H E R H E A R T A N D I N F L U E N Z A , A L L O T H E R C A U S E SC I R C U L A T O R Y B R O N C H I T I S A N D D I S E A S E S P N E U N O M I A | Figure 2 – Mortality rate of widowers during first six months of bereavement as a proportion of rate for married men of the same age, by cause of death Then, they charted the cause of death of the widowers, most of whom had died in the first six months of bereavement (Figure 2). They realised that there was a sharp increase in fatal heart-related diseases as compared to the control group. In fact, death due to heart-related diseases accounted for ⅔ of the increase in mortality rate in the first 6 months of bereavement. Additionally, based on the expected number, the 99% confidence interval for the expected number of deaths from heart problems was 25 - 67. The actual number of widower’s deaths was 77, which is significant at the 1% level. This shows that the mortality rate due to heart diseases is not the norm as it is well outside the expected number even at the 99% confidence interval. Conclusion 2 (by the study): Do not reject null hypothesis 2. There is insufficient evidence to prove that there is a relationship between the increase in mortality rates in the first 6 months and the homogamy theory. While the proportion of wives and widowers dying from the same disease group was higher than expected, the total (23.9% more than expected) was not high enough to support the homogamy theory. | | | | Arteriosclerotic and Degenerative
Heart Disease Arteriosclerotic and Degenerative
Heart Disease | Other heart and
Circulatory
Disease Other heart and
Circulatory
Disease | Vascular lesions of C.N.S. Vascular lesions of C.N.S. | Influenza, pneumonia and bronchitis Influenza, pneumonia and bronchitis | Cancer Cancer | All other causes All other causes | | Arteriosclerotic and Degenerative Heart Disease | 22 (18.79) | 5 | 5 | 5 | 7 | 8 | 52 | Other heart and Circulatory Disease | 10 | 5 (2.81) | 2 | 4 | 2 | 2 | 25 | Vascular lesions of C.N.S. | 19 | 6 | 5 (5.68) | 10 | 9 | 6 | 55 | Influenza, pneumonia and bronchitis | 5 | 0 | 3 | 3 (2.04) | 1 | 3 | 15 | Cancer | 13 | 5 | 6 | 4 | 7 (5.77) | 6 | 41 | All other causes | 8 | 3 | 1 | 3 | 4 | 6 (3.64) | 25 | Totals | 77 | 24 | 22 | 29 | 30 | 31 | 213 | Table 2 – Actual number of cases of widowers bereaved in the first six months in which the cause of death is the same as for their wives. Number expected is in ().

| | | | | Arteriosclerotic and Degenerative Heart Disease | 22 | 18.79 | Other heart and Circulatory Disease | 5 | 2.81 | Vascular lesions of C.N.S. | 5 | 5.68 | Influenza, pneumonia and bronchitis | 3 | 2.04 | Cancer | 7 | 5.77 | All other causes | 6 | 3.64 | Total | 48 | 38.73 | Table 3 (Extrapolated from Table 2) – Number of actual and expected cases of widowers bereaved in the first six months in which the deaths of husband and wife were in the same diagnostic group. The study charted the cause of death of the widowers with that of their wives to see whether they had the same cause of death. This is because a possible explanation for the increase of widower’s mortality rates may be due to homogamy, which is the idea that the ‘unfit marry the unfit’. This theory suggests that the wife’s death may be related to her widower’s death from the same disease as both live similar unhealthy lifestyles or had similar characteristics which made them predisposed to the disease. The researchers were thus interested to examine if homogamy, rather than bereavement itself, was indeed the main factor accompanying increased mortality rates. Researchers charted the actual number of widower’s deaths from the same disease as their wives against the expected number which would be due to chance association or coincidence. Table 2 shows that while 38.7 cases were expected to coincide, a total of 48 did. If homogamy were to have an actual relationship to the increase in widower’s mortality rate, the number of similar deaths would have to be much higher and closer to an outlier. Since the article did not mention the criteria required to support the homogamy theory, we assumed that this would be if the actual number of similar deaths is different from the expected number by more than 50%. As shown in Table 2, however, this is not the case. Thus, there is insufficient evidence to show a strong relationship between homogamy and the increase in widower’s mortality rate. Conclusion 3 (by the study): Do not reject Null Hypothesis 3. There is insufficient evidence to prove that there is a relationship between social class and the increase in widower’s mortality rate. While the proportion of wives and widowers dying from the same disease group was higher than expected, the total (23.9% more than expected) was not high enough to support the homogamy theory. Using data from the control group (refer to Table 4), researchers calculated the expected proportion of deaths by social class. They then compared this with the proportion of deaths by social class for the sample of widowers. They expected that there would be a marked increase in proportion of deaths from the higher social classes compared to the lower social classes. However, the sample findings did not show any relationship between the social classes and the mortality rates. It would seem that the correlation between bereavement and increased mortality is one that cuts across social classes. | | | | | | | I | 4 | 2.30 | 1.74 | II | 25 | 17.90 | 1.40 | III | 107 | 69.41 | 1.54 | IV | 47 | 34.15 | 1.37 | V | 19 | 15.97 | 1.19 | Unknown | 11 | 13.26 | 0.83 | Total | 213 | 152.99 | 1.39 | Table 4 – Mortality in the first six months of bereavement for widowers over the age of 55 by social class, expressed as a proportion of mortality rate for married men of the same age and social class

The researchers identified 5 considerations to take when interpreting the results of the findings. Firstly, they made clear that the average age of widowers in any age group is about one year older than the average age of married men within the same age group. This may explain why the rate of mortality among widowers is slightly higher. However, they did not provide any evidence to show a relation between a one-year age gap and an increase in mortality rate. Secondly, the researchers also recognise that there could have been miscalculations in the process of data collection. For example, the number of widowers may be under-enumerated in the census provided by the N.H.S Central Register, while the number of widowers may be over-enumerated in the deaths recorded. This may result in inaccurate findings. Thirdly, they acknowledge that some of the widowers in the group may get remarried, but were not removed from the initial sample data of widowers. As such, this may lead to an issue of bad sampling in later years as the widowers who remarry can be considered “fitter” than the general population of widowers who do not remarry. This will result in a more varied sample data and skewed results in later years, and thus reduce the validity of the findings. Fourthly, some of the tagged cases made at the start of the observational study may become lost in transit over the years. For example, some of the widower’s demise may not be reported to the Registrar General Office, due to the difficulty in ensuring the validity of the samples at the time. This would directly interfere with the sample size and reduce the effectiveness of their study, but may be one possible reason in explaining the spuriously lower mortality rate in the years following the initial year of bereavement. Lastly, to draw the possibility between dying from a ‘broken heart’ through heart diseases, the researchers looked into the relationship between emotional stress (from the loss of a loved one, or the metaphorical ‘broken heart’) and psychological stress, which ultimately leads to the increase in mortality rate by heart diseases. Researchers cited a study conducted in 1957 where emotional stress was found to be present in 86% of 43 coronary thrombosis cases as compared to 9% in the control group. Emotional stress has also been found to affect pulse rate, stroke volume, cardiac output, and to produce arrhythmias and electrocardiographic changes. Therefore, the study concludes that there is a possibility that emotional stress caused by the effects of bereavement leads to psycho-endocrine functions, which increases the mortality rate by heart diseases. This would explain the link between bereavement (the metaphorical ‘broken heart’) and the increased mortality rate of widowers due to heart-related illnesses (the literal ‘broken heart’).

The group would like to make 3 recommendations that we hope will address the shortcomings as observed in the study. Firstly, in order to increase the statistical validity of the study, researchers could conduct the study multiple times to ensure that the trends of findings are consistent. This would allow them to establish a definite correlative relationship between bereavement and increased mortality rate from heart-related illnesses. Furthermore, the study could be conducted with different samples from various geographical locations albeit within the same time-period, and with comparable medical treatment and technology available. This would increase the external validity of the study by removing any locational bias and ensuring that the study’s findings are more applicable to the population. Next, to reduce the problems researchers faced with tracking of the widowers data over the long 9 year period, researchers could conduct regular checks on the sample data list to ensure that it is updated. They could also make use of the computerised data collection to make the data collection and recording process more accurate and reliable. Lastly, to increase the accuracy and validity of the study’s findings on the correlation between homogamy and couples dying of a similar cause, the researchers could have provided the statistical tvalue, p-value, and the power for the test. Such statistical calculations will help to directly indicate whether the two groups of data come from the same group (p-value) and whether the difference between the two data groups is significant (t-value). Such statistical information will help to add validation to the concluded findings. In addition, researchers could have also disclosed the possibility of a type 1 or type 2 error. By doing so, readers of the study will be able to see the probability of the study making a wrong conclusion, and provide validation to the concluded findings (assuming error is small).

Parkes, C. M., Benjamin, B., & Fitzgerald, R. G. (1969). A Statistical Study of Increased Mortality Among Widowers. British Medical Journal, 1969(1), 740-743. Retrieved from http://www.ncbi.nlm.nih.gov/pmc/articles/PMC1982801/pdf/brmedj02024-0030.pdf