Independent importance of psychosocial factors for prognosis after myocardial infarction


: Professor Lars Wilhelmsen, Section of Preventive Cardiology, Drakegatan 6, SE-412 50 Göteborg, Sweden (fax: + 46 31 7031890; e-mail:


Abstract. Welin C, Lappas G, Wilhelmsen L (Heart and Lung Institute, Göteborg University, Göteborg, Sweden). Independent importance of psychosocial factors for prognosis after myocardial infarction. J Intern Med 2000; 247: 629–639.

Aims. As a primary aim it was tested whether the 10-year prognosis after a myocardial infarction is related to psychological stress, lack of social support, anxiety, and/or depressive tendency. A secondary aim was to analyse the prognostic importance of a series of other psychosocial factors as well as interactions.

Methods. Non-selected patients aged below 65 years with a first infarction (230 men and 45 women) were followed for 10 years with 100% assessment of morbidity and cause-specific mortality. Baseline somatic and psychosocial variables were collected with the aid of standard, validated questionnaires.

Results. In multivariate analysis, factors increasing risk for coronary mortality included female sex (hazard ratio, ± 95% confidence interval) 2.47 (1.06, 5.71), signs of left ventricular failure 3.93 (1.87, 8.26), ventricular dysrhythmia 3 months after the infarction 5.45 (2.21, 13.42), high depression scores 3.16 (1.38, 7.25) and lack of social support 2.75 (1.29, 5.89). All-cause mortality was significantly related to left ventricular failure, ventricular dysrhythmias, and high depression scores with borderline significance for female sex and social support. Prognosis was affected during the entire follow-up period. It was not significantly associated with age, marital status, education, extra work, mental strain at work or in the marriage, anxiety, dissatisfaction with family life, problems with children, dissatisfaction with the financial situation, life events, anger-in, irritability, type A behaviour, or health locus of control. Incidence of nonfatal infarction was not associated with any of the baseline variables.

Conclusion. In addition to known somatic predictors of prognosis after a myocardial infarction, prognosis is strongly influenced by depression and lack of social support, but not to a series of other psychosocial factors. It is recommended to use self-reporting scales to detect prognostically important psychosocial problems.


Mortality after a myocardial infarction is highest during the first months of the infarction, after which it levels off. It is well established that a series of factors related to the size of the index infarction and reduction of myocardial contractility are related to deteriorated prognosis [1] as well as presence of angina pectoris [2], and ventricular dysrhythmias [3]. Long-term prognosis is also dependent on pre-infarct risk factors such as diabetes [45], hypertension [2], cholesterol levels [6], as well as smoking habits after the infarction [7].

In addition to these somatic factors several psychosocial factors have been implicated in prognosis as recently reviewed [8]. Prognosis has been significantly related to depression, anxiety and social support, but not to the type A personality pattern [8]. The prognostic significance of social factors has varied markedly in different studies [9].

Patients for the present study were enrolled between 1985 and 1987 in a case-control study analysing the importance of a series of somatic and psychosocial factors for the risk of a first myocardial infarction amongst patients <65 years old. In addition to well-known somatic risk factors, psychological stress at work, negative life events, irritability or low social activity were more common amongst patients than controls [10–14]. In the present long-term follow-up of these patients with a first infarction we studied factors that were important for prognosis after an event. Based on previous reports, the primary hypothesis was to test whether prognosis is related to psychological stress, lack of social support, anxiety and a tendency to be depressed. A secondary aim was to perform some explanatory analyses on the prognostic importance of marital status, education, extra work, mental strain at work, mental strain in the marriage, dissatisfaction with family life, problems with children, dissatisfaction with the financial situation, life events, anger-in, irritability, type A behaviour and health locus of control, as well as sleep problems. This second group of analyses was performed with consideration of the risk of significance occurring by chance when many variables are tested. The comprehensive baseline evaluation performed at a defined time after the infarction enabled determination of the prognostic importance, if any, of many psychosocial variables. Relationships between baseline variables were studied as well as the prognostic importance during the first years after the index event compared to the latter part of the 10-year follow-up period.

Study population

A myocardial infarction register and an outpatient clinic for follow-up and treatment of all surviving infarct patients <65 years of age and registered as residents in Göteborg, Sweden (population 450 000), have been in operation at the Section for Preventive Cardiology since 1968 [15]. During the period of the present study, there were two hospitals that accepted emergencies: Sahlgrenska Hospital and Östra Hospital. Lists of patients admitted with suspected myocardial infarction were scrutinized daily by specially trained nurses. In addition, all wards in which patients with acute infarction were primarily treated were visited by one of the nurses at least twice a week. A physician from the Myocardial Infarction Clinic [16] established the definite diagnosis using predetermined criteria.

Acute myocardial infarction according to the definitions of the Swedish Cardiac Society was considered to be present when two or more of the following criteria were fulfilled: (a) central chest pain of at least 15 min duration; (b) transient elevation of transaminases in serum to levels exceeding the upper normal limit for the laboratory; and (c) an electrocardiogram series with addition of pathological Q wave and/or disappearance of localized ST elevation in combination with development of T inversion in at least two of 11 routinely registered standard leads (I, II, III, aVR, aVL, aVF, V1, V2, V4, V5, V7).

The study included consecutively registered patients with a first myocardial infarction during the period October 1 1985 to October 31 1987 (excluding a period of 8 weeks interruption because of leave from work for the principal investigator). Fifty-five woman and 288 men took part in the investigation at hospital. During 3 months after the infarction one woman and seven men died. Due to refusal to participate in the psychosocial investigation or some missing baseline data from the hospital period, as well as one and 3 months after the infarction data were not available for nine women and 51 men. Thus, 230 men and 45 women were available for the present prospective study starting 3 months after the infarction.


All clinical data were obtained according to a predefined protocol [16]. After receiving information about the purpose of the study, the patients filled in a self-administered questionnaire 3–6 days after the onset of the event. Most of the questions concerned conditions during the last week, month or year. The patients were strongly urged to answer the questions with reference to conditions before the onset of the event. At follow-up assessments after 1 and 3 months, the conditions were assessed again. The same person (C.W.) provided all information and administered all questionnaires. The follow-up started 3 months after the infarction.

Left ventricular failure is known to have important prognostic implication. The criteria included clinical signs of heart failure (New York Heart Association Class III–IV), and pulmonary congestion at chest X-ray during hospitalization. Standardized forms were used for the assessment. At the time of this baseline examination in 1985–87, left ventricular ejection fraction was not measured routinely.

Ventricular dysrhythmias were, for the present analysis, assessed 3 months after the acute episode with routine long strips (including about 30 beats) of a 12-lead electrocardiogram. Single or multiple beats of ventricular origin were accepted as ventricular dysrhythmias.

Participants were classified as smokers before infarction if they smoked at least one cigarette per day or the equivalent in cigar, or pipe tobacco, and at 3 months after the infarction using the same criteria.

Systolic and diastolic pressure phase 5 were measured with a mercury manometer to the nearest 2 mmHg. Measurements were taken in the right arm with the patient in the supine position after 5 min rest. Hypertension was considered to be present if the patient had been treated for hypertension or if blood pressure levels exceeded the cut-off 3 months after onset of the infarction. Blood pressure at that time is known to be on average 14/10 mmHg lower than the pre-myocardial infarction level [17] and the cut-off point for hypertension was set at 146/85.

Serum cholesterol was determined after a 12-h fast according to routine laboratory methods 3 months after the onset of the infarction, when the values had stabilized to the pre-infarct level [17].

Diabetes was defined as a clinical diagnosis of diabetes during hospitalization.

Psychosocial variables

As far as available, previously published and validated questionnaires were used. If available, the variables were scored and cut-off points used according to suggestions in the original publication. If such cut-points were not available, predetermined cut-points from the case-control study (to avoid bias) were also used in this prospective study. These cut-points, as well as ranges, are stated below.

Marital status, education, overtime work (hours per month), extra work and shift work were assessed according to questionnaires that are presented in detail elsewhere [11]. Mental strain at work was assessed according to the following questions: ‘How often does it worry you that you have to try very hard in order to take care of your present work load?’, and ‘How frequently does it worry you that you are often hurried in your work?’ Responses were scored from 2 to 10 and ≥7 was regarded as mental strain. Mental strain in the marriage was defined as a score ≥7 (range 2–10). Dissatisfaction with family life, problems with children, work interfering with family life, and dissatisfaction with financial situation were each scored from 1 to 5 with 5 as the worst situation and scores >4 defined as mental strain. Life events (score ≥7) were assessed using a scale comprising 10 negative events with a maximum total score of 20. Emotional strain (score ≥15) in connection with life events was scored 0–50.

Anger-in (score ≥17; range 10–20) was assessed according to the Karolinska Scales of Personality [18], sense of guilt (score ≥17; range 9–18), according to the Buss Durkee Hostility Inventory [1920], irritability (score ≥17; range 11–22), according to the same instrument, and anxiety (score ≥40; range 20–80), according to the Trait Anxiety Inventory [21]. Depression (score ≥40; range 20–80), was assessed by the Zung Self-Rating Depression Scale [22], which assesses the level of depression using emotional, biological, and physiological symptoms using 20 items. This part of the investigation was carried out 1 month after the infarction. The cut-point used by us conforms to the one originally suggested by Zung to indicate clinically significant depression. Meaningfulness (score ≥80 ; range 20–120) was measured by Beck’s Hopelessness Scale [2324].

The Jenkins’ Activity Survey form N [2526] measured type A behaviour ( score ≥5; range –20 to +20), The Wallston 11-item scale [27] measured health locus of control (high score of ≥46 indicates strongest external locus of control; range 11–66). Sleep problems ( score ≥12 (indicating more sleep problems than those with lower scores); range 0–20), were evaluated by four six-point questions [28]. Alcohol consumption was measured using questions concerning the frequency and the amount of beer, wine, and/or liquor drinking ( score ≥16 (higher degree of reported alcohol consumption); range 0–30).

Social relationships were measured using three different scales: (a) a condensed version of the Interview Schedule for Social Interaction (ISSI) [2930]; (b) The Broadhead questionnaire [31]; and (c) another questionnaire which includes eight items about social activities during the previous year [3233].


A nonfatal recurrent myocardial infarction was diagnosed with the same criteria as those used for entry into the study as described above. For fatal events the death certificate diagnosis was used. The criteria for fatal coronary heart disease were either myocardial infarction on the death certificate or sudden death not due to noncoronary causes. The other groups analysed were stroke, cancers and other causes of death.

Statistical methods

All dichotomizations were performed prior to analysis of outcome according to previous experience [10–14]. Analysis was carried out using SAS. Survival curves were constructed for the different baseline variables to assess effects over time on outcome during the 10-year follow-up (assumption of proportional hazards over time) and whether censoring was uniform across groups. Hazard ratios (HR) with 95% confidence intervals (95% CI) were calculated using Cox regression analyses. The χ2 statistic was used to test associations between baseline variables. Baseline variables that were significantly (P < 0.05) related to outcome in univariate analysis were included in the multivariate Cox regression, and conditional hazard ratios with 95% CIs were calculated for these predictors. The log–rank test was used to compare probability curves in Figs 1 and 2.

Figure 1.

Cumulative probability of avoiding coronary mortality amongst 98 infarct (MI) patients with depression scores ≥40 after their first myocardial infarction according to Zung’s Depression Scale, compared with 169 patients with lower scores. Numbers below figure relate to patients available for assessment at beginning of each year.

Figure 2.

Cumulative probability of avoiding coronary mortality amongst 79 infarct (MI) patients with social support scores ≤17 compared with 196 patients with higher scores. Numbers below figure relate to patients available for assessment at beginning of each year.


There were in total 230 men and 45 women with a first myocardial infarction for whom data were available for prospective analysis at 3 months after the infarction. Sixty-seven (24%) of these patients died; 41 (61% of all deaths) from coronary heart disease, two from stroke, 10 from cancer and 14 from other causes during the mean follow-up of 10 years. Fifty-five patients had a recurrent nonfatal infarction. Out of these 15 died (10 from coronary disease) during follow-up.

Table 1 shows bivariate associations between baseline variables and the endpoints fatal coronary events and all-cause mortality, during 10 years follow-up. Amongst somatic variables, female sex, diabetes mellitus, left ventricular failure during hospitalization, dyspnoea at 1 month after infarction, and ventricular dysrhythmia recorded 3 months after infarction were significant predictors for a fatal coronary event as well as all-cause mortality. A recurrent nonfatal infarction was also significantly related to outcome (P = 0.018) according to time-dependent Cox analysis. Age was not significant (range 30–65 years).

Table 1.  Relationships between baseline variables and events during 10 years’ follow-up
Coronary mortalityAll cause mortality
(n = 41)
95% CI
(n = 67)
95% CI
  • *

    Numbers that do not add up to 275 are due to missing information.




(1.22, 4.67)



(1.07, 3.30)

 ≥55 years
 ≤54 years
(0.53, 1.87)
(0.73, 2.05)
Left ventricular failure
(1.94, 6.62)
(1.73, 4.59)
Dyspnea after infarction
(1.16, 4.05)
(1.35, 3.63)
Ventricular dysrhythmia at 3 months
(2.19, 11.32)
(1.34, 6.49)
Diabetes mellitus
(1.25, 5.93)
(1.03, 3.98)
(0.52, 2.88)
(0.73, 2.28)
Smoking at 3 months after MI
(0.18, 1.43)
(0.71, 2.31)
Serum cholesterol
 ≥7.00 mmol L–1
 ≤6.99 mmol L–1
(0.47, 1.80)
(0.50, 1.44)
Depression score
(1.85, 6.79)
(1.49, 4.02)
Social support
(1.13, 3.88)
(0.92, 2.52)
Social activities
(1.07, 3.98)
(1.12, 3.20)
Anxiety before MI
(0.59, 2.00)
(0.56, 1.50)
Sleep problems
(1.03, 3.74)
(0.78, 2.35)

Depression after the infarction (score ≥40; range 0–80), lack of social support (score ≤17; range 0–65) and sleep problems (score ≥12; range 0–20) were significant predictors of coronary mortality in bivariate analysis, and these variables, with the exception of social support and sleep problems, were also significant predictors for all-cause mortality. Anxiety (score ≥40; range 20–80), before the infarction was associated with significantly worse prognosis during the first 5–6 years of follow-up, but the curves merged at the latter part of follow-up and there was no significant difference for the entire follow-up period. Kaplan–Meyer curves for survival free of coronary deaths in relation to the presence or absence of depression after the infarction are shown in Fig. 1 and the presence or absence of social support, in Fig. 2. It is evident that depression and social support had an impact on prognosis during the entire follow-up period according to the log–rank test (P = 0.001 for depression and P = 0.0001 for social support).

Nonfatal myocardial infarction during follow-up was not significantly related to any of the baseline variables.

Fatal coronary events were not related to age (P = 0.98), marital status (P = 0.73), education (P = 0.37), extra work (P = 0.26), mental strain at work (P = 0.99), mental strain in the marriage (P = 0.91), dissatisfaction with family life (P = 0.80), problems with children (P = 0.11), dissatisfaction with the financial situation (P = 0.83), life events (P = 0.84), anger-in (P = 0.28), irritability (P = 0.60), type A behaviour (P = 0.99), or health locus of control (P = 0.76). Neither were these variables predictive of all-cause mortality.

Some of the baseline variables were correlated. Women tended to have diabetes more often, compared to men; 16.3% vs. 7.7% (P = 0.074) and were significantly more often depressed after their infarction compared to men; 55.6% vs. 34.4% (P = 0.013). Diabetic patients, more often than nondiabetics, tended to have hypertension; 54.5% vs. 37.7% (P = 0.122), more often complained of dyspnoea; 52.2% vs. 21.2% (P = 0.003), and tended to be depressed more often; 54.2% vs. 35.7% (P = 0.081). Patients with ventricular dysrhythmia were more often depressed compared to those without; 64.3% vs. 35.9% (P = 0.033), and reported social support to a lesser degree; 50.0% vs. 76.5% (P = 0.026). Patients with few social contacts reported less social support; 48.3% vs. 18.9% (P = 0.001), smoked more often 3 months after the infarction; 27.6% vs. 15.6% (P = 0.037), and more often complained of dyspnoea; 35.6% vs. 21.6% (P = 0.028). High scores for depression were associated with dyspnoea after infarction; 35.0% vs. 19.2% (P = 0.004), high score for anxiety before the infarction; 62.1% vs. 30.0% (P = 0.0001) as well as sleep problems; 37.8% vs. 13.6% (P = 0.001).

Multivariate Cox analysis was performed for the above-mentioned endpoints. In Table 2, variables significantly related to coronary mortality and all-cause mortality, respectively, are listed. Fatal coronary disease was more common amongst women, patients with left ventricular failure, patients with ventricular dysrhythmia, patients with high depression scores after the infarction as well as those with a lack of social support. Diabetes became nonsignificant due to correlations between several baseline variables, as mentioned above, as well as the relatively few diabetic patients. A nonfatal reinfarction during follow-up (analysed by time-dependent Cox analysis) was no longer significantly associated with a fatal outcome (P = 0.070). All-cause mortality was significantly related to left ventricular failure, ventricular dysrhythmia, and high depression scores and was close to significantly related to female sex and social support.

Table 2.  Multivariate, Cox regression analysis of risk factors for coronary mortality and all-cause mortality, respectively. Endpoint numbers available for the multivariate analyses, as well as hazard ratios (HR) and 95% confidence intervals (95% CI) are given
Coronary mortality (n = 33) All-cause mortality (n = 54)
VariableHR(95% CI)PHR(95% CI)P
Female/male2.47(1.06, 5.71)0.0361.85(0.99, 3.47)0.056
Left ventricular failure3.93(1.87, 8.26)0.0032.90(1.70, 4.95)0.001
Ventricular dysrhythmia5.45(2.21, 13.42)0.00022.49(1.09, 5.67)0.030
Depression after infarction3.16(1.38, 7.25)0.00071.75(1.02, 2.99)0.041
Social support2.75(1.29, 5.89)0.0091.67(0.97, 2.89)0.066

Table 3 provides an example based on the same results as Table 2 for coronary mortality and all-cause mortality in relation to signs of left ventricular failure, depression score and lack of social support. The table is based upon men only because of the small number of women, and the variable ventricular dysrhythmia was not selected because only 14 patients were reported with dysrhythmia according to the present criteria. Coronary mortality increased from 3% when none of the factors was present to 88% when all were present, and all-cause mortality from 13% to 100%.

Table 3.  Coronary mortality and all-cause mortality, respectively, in relation to left ventricular failure, depression after the infarction, and social support
Lack of
Yes Yes8225450

There were too few women to enable multivariate analyses amongst them. At bivariate analyses of baseline factors associated with fatal coronary or all-cause mortality during follow-up, the same relative risks were found for women as for men for all factors listed in Table 1, with one exception. Lack of social support did not seem to have the same negative prognostic effect amongst women, but again small numbers preclude strong conclusions regarding this finding.


This study shows that female sex, left ventricular failure, ventricular dysrhythmia, depression, as well as lack of social support independently increased the risk for fatal coronary events amongst men and women who had suffered a first infarction. All-cause mortality was significantly related to the same factors, but female sex and lack of social support did not reach statistical significance. Diabetes mellitus, which was significant in bivariate analysis, was no longer significant in multivariate analysis due to correlation with other baseline factors and small numbers. A series of other psychosocial and behavioural characteristics had no predictive power.

Amongst psychosocial variables it was found that patients who reported anxiety before the infarction had higher depression scores after it.

In a previous case-control study including the present infarct patients, it was found in multivariate analysis that mental strain at work, negative life events, irritability as well as few social contacts (lack of social support), but not depression, were significantly more common than expected in infarct patients [10–14]. There were (positive) associations amongst infarct patients between depression score and mental strain at work, life events, irritability, sleep problems, as well as lack of social support. Whether lack of social support is due to depression or vice versa could not be evaluated. In the present prospective analysis of prognosis after the infarction it is noteworthy that lack of social support and high depression scores had independent effects on prognosis.

A limitation of this study is the small number of women that had an infarction during the period, and the lack of information regarding ejection fraction. However, a standardized clinical assessment of functional class was performed, and the impact of psychosocial factors was seen over the whole range of myocardial function. Unfortunately routine death certificates did not allow a scientifically valid decision about whether deaths were sudden or not. Strengths of the study are: the standardized evaluation of all variables with use of well-established questionnaires; assessment of the importance of many psychosocial variables; and the complete follow-up.

Similarly to the present study, others have found that social isolation is associated with a worse prognosis after a diagnosis of coronary disease [8 34–36]. However, Frasure-Smith et al. [37] recently completed a sizeable randomized trial of improved support to infarct patients, but was not able to show any benefit from their intervention.

Clinical depression as a predictor of worse prognosis after an infarction has been described already in papers by Garrity & Klein [38] and Stern et al. [39] and later by Carney et al.[40], but follow-up of their patients was restricted to about 1 year. The later studies by Ladwig et al. [41] and by Frasure-Smith et al. [42] covered a follow-up period of 6 months and 18 months, respectively. In the Cardiac Dysrhythmia Pilot Study, 502 patients with ≥10 ventricular premature complexes per hour or ≥5 episodes of ventricular tachycardia as well as previous myocardial infarction were studied and it was found that depression was an independent predictor of mortality [43]. Similarly Barefoot et al. [44] found that high scores of the Zung Self-Rating Depression Scale were associated with an increased risk of subsequent cardiac death as well as total mortality amongst 1250 coronary patients (including 52 recent and 607 older infarctions) referred to angiography during and up to 19.4 years of follow-up. However, the time of the baseline investigation differed between patients, compared to the present one in which the baseline was always 3 months after the infarction. Denollet & Brustaert [45] studied 87 patients with first as well as recurrent infarctions. A tendency to suppress negative emotions, which is a combination of ‘trait anxiety’, also used in the present study, and social inhibition, was the strongest negative prognostic factor. The results were questioned by Carney [46]. Trait anxiety did not predict prognosis in our study, but social inhibition was not studied by us, which is why a direct comparison is not possible.

The mechanisms underlying the increased mortality rate amongst postinfarct patients with limited social contacts and depressive symptoms are not well elucidated. It is attractive to believe that persons with a lack of social contacts and support are less interested in controlling their cardiovascular risk factors. However, we did not see that these patients had a worse risk factor profile. Another possibility is that they respond later to premonitory symptoms. This was not studied, but the strength of the association indicates that more basic influences have to be sought.

It may seem unlikely that high scores for depression assessed up to 3 months after the infarction had predictive significance during the 10 years after the event. It is known that a tendency to react with depression is more common in certain persons and may be regarded as a personality trait. Thus, this trait may be related to events during a long period. There are several disturbed humoral mechanisms in depression which may be potential pathways leading to cardiac and primarily arrhythmic death. Increased sympathetic and decreased vagal tone, as well as major disturbances of the serotonergic neuronal function have been demonstrated in depressed patients [47–51]. Reduced heart rate variability has been detected in depressed patients with heart disease [52], and low heart rate variability has predicted mortality in postinfarct patients [53].

It has been documented that acute psychological threats can trigger sudden coronary death [54]. Experimentally it was shown, for example by Parker et al. [55], that central beta-adrenergic mechanisms could modulate ischaemic fibrillation in pigs. In the present study ventricular dysrhythmia (but not with any sophisticated methods) were detected in 14 patients at baseline, and depression was more common amongst them. Seven of these 14 patients died during follow-up. Due to the lack of a more detailed assessment of dysrhythmias at baseline and exact mode of death, we cannot throw more light on the mechanisms leading to death in the present patients.

Krishnan and McDonald [56] hypothesized that “late onset depression is secondary to subtle cerebrovascular changes due to atherosclerosis… We do not hypothesize that the cerebrovascular changes lead directly to depression, but that they are vulnerability factors which, in the context of psychological factors, lead to the development of depression.” They also present evidence for increased atherosclerosis amongst depressed patients. Following their idea it may be further hypothesized that patients who become depressed after a myocardial infarction have more advanced (but often difficult to detect) cerebral as well as coronary atherosclerosis than nondepressed patients. This in turn would be associated with an increased mortality risk.

The cardiologist is usually not trained in detecting premonitory symptoms of depression and even major depression may sometimes go undetected. Specific treatment is consequently not prescribed. In order to make comprehensive prognostic assessment and to quantify depressive symptoms, it is recommended to use a depression scale such as that used in the present study (The Zung Self-Rating Depression Scale) [22]. Anti-depressant treatment may improve the adverse prognosis, but randomized trials amongst patients who are depressive after an infarction are needed to assess this possibility.


The valuable check of the language by Mr Steve Fullerton and secretarial work by Miss Ingela Thorlin are acknowledged.

This study was supported by grants from the Swedish Heart and Lung Foundation, the Göteborg University, the Knut and Alice Wallenberg Foundation, and the Swedish Labour Market Insurance Company.

Received 29 September 1999; revision received 22 December 1999; accepted 1 March 2000.