Sample

Consecutive patients admitted to four inner-city hospitals with suspected myocardial infarction were screened for this study, on average 3.6 (s.d.=3.3) days after admission. Individuals were recruited if they had no history of previous myocardial infarction, were aged less than 80 years and met World Health Organization criteria for myocardial infarction (Reference GutzwillerGutzwiller, 1994). Two of the following were required: history of typical chest pain, characteristic echocardiogram changes, a serial rise in cardiac enzyme creatine phosphokinase (CPK). People with severe mental illness (including severe depression with suicidal ideas) were excluded from the study. Written, informed consent was provided by participants after the procedures involved in the study had been fully explained.

Participants received treatment as usual, as determined by the attending hospital physician and primary care doctors. As this was an observational study, participants did not receive any additional therapy (e.g. antidepressants) based on their research assessments.

Baseline assessments

We recorded age, gender (male=1, female=2), ethnicity (White=1, Black and minority ethnic=0), marital status (single: yes=1, no=0), years of education, smoking history (past or current smoker=1, never smoked=0), past psychiatric history (yes=1, 1, no=0) and comorbid illnesses for which participants were receiving regular treatment (for each illness yes=1, no=0). Socioeconomic status, based on current or latest employment, was categorised into high or low according to the criteria of Goldthorpe and Hope (Reference Goldthorpe and HopeGoldthorpe & Hope, 1974), (lower status=1, higher status=2). Serum cholesterol was measured during admission.

Details of the myocardial infarction were recorded from the notes, with discussion with the attending cardiologist if clarification was required. We recorded site of infarction as determined by echocardiogram, categorised as anterior (including anterior, anterolateral, anteroseptal and apical myocardial infarctions, scored as 1) or other (scored as 2). Details of treatment, including reperfusion therapy, surgical intervention during index admission (angioplasty or coronary artery bypass graft) and discharge medication were also noted. We used three measures of severity of myocardial infarction. The Killip class, a 4-point scale, rated left ventricular function based on the presence of pulmonary rales, S3 gallop rhythm and peripheral hypoperfusion, with high scores representing worse function (Reference Killip and KimballKillip & Kimball, 1967). Creatine phosphokinase levels were measured on admission and on 2 subsequent days as an indication of the extent of myocardial damage. Finally, left ventricular ejection fraction measures (using echocardiogram) were recorded for 217 patients.

Anxiety and depression were assessed using the self-rated, 14-item Hospital Anxiety and Depression Scale (HADS), which scores the severity of the symptoms of anxiety (scored 0 to 21) and depression (scored 0 to 21). This measure was developed specifically for medically ill populations, and excludes bodily symptoms such as sleep disturbance, fatigue and pain that may be owing to physical illness (Reference Zigmond and SnaithZigmond & Snaith, 1983). The HADS has been widely used in people with medical illnesses, including myocardial infarction, and its factor structure has been confirmed (Reference HerrmannHerrmann, 1997; Reference Herrmann, Brand and KaminskyHerrmann et al, 1998; Reference Mayou, Gill and ThompsonMayou et al, 2000). At baseline, participants were instructed to complete the questionnaire to reflect their mood during the week before myocardial infarction, in order to avoid recording mood changes that occur as immediate reactions to the infarction and during time spent on the coronary care unit. In addition to using HADS subscores to measure severity of depression and anxiety, a HADS total score of 17 or above was taken to indicate probable major depressive disorder, as we have validated this score against a standardised research interview (Reference Dickens, McGowan and PercivalDickens et al, 2004b ). We did not aim to measure distress in the period immediately after infarction, and we have recorded the fact that HADS scores rated for the days following admission were significantly higher than those rated for the week before, which were used in this report (Reference Dickens, McGowan and PercivalDickens et al, 2004b ).

Participants completed the Illness Perceptions Questionnaire (Reference Weinman, Petrie and Moss–MorrisWeinman et al, 1996), which assesses patients’ perception of their illness in terms of:

1. (a) its bodily symptoms (‘identity’ scale – scored 0 to 14, with high scores indicating more reported symptoms);

2. (b) its likely duration (‘time-line’ – scored 1 to 5, with low score reflecting longer anticipated duration);

3. (c) the perceived consequences (scored 1 to 6, with low score indicating more serious perceived consequences);

4. (d) the likelihood of the illness being cured or controlled (scored 11 to 5, with to 5, with low score indicating more perceived control).

This scale has been used previously for people who have experienced myocardial infarction (Petrie et al, Reference Petrie, Weinman and Sharpe1996, Reference Petrie, Cameron and Ellis2002).

We used the self-reported Medical Outcomes Study SF–36 (Reference Ware and SherbourneWare & Sherbourne, 1992) to assess both the physical and mental aspects of HRQoL at baseline, 6 months and 12 months following myocardial infarction. The SF–36 is a generic measure of HRQoL that provides scores on eight sub-scales, reflecting functioning over the month preceding the assessment. The SF–36 has been well validated in populations with coronary heart disease (Reference Failde and RamosFailde & Ramos, 2000; Reference Morrin, Black and ReidMorrin et al, 2000; Reference Rumsfeld and HoRumsfeld & Ho, 2005). We deliberately chose a generic measure of HRQoL, as opposed to a cardiac-specific measure, as our main outcome; people after myocardial infarction frequently experience multiple health problems, and we wanted a measure that was sensitive to the impact of noncardiac illnesses on HRQoL. A cardiac-specific measure of HRQoL would be less sensitive to such non-cardiac illnesses.

We used the SF–36 Physical Component Score (PCS) as the primary outcome measure, because it has greater sensitivity to change over the original eight scales of the SF–36, involves fewer statistical comparisons and eliminates floor and ceiling effects (Ware et al, Reference Ware, Kosinski and Keller1994, Reference Ware, Kosinski and Baylis1995). The PCS was derived from four sub-scales (physical function, role limitation, physical pain and health perceptions), using the method described by Ware and colleagues (Reference Ware, Kosinski and KellerWare et al, 1994). The PCS is standardised to the general population (mean score, 50: standard deviation, 10), with high scores representing higher levels of functioning. A change on the PCS of between 3 and 5 (equivalent to 0.3 to 0.5 of the standard deviation) is considered to be clinically significant (Reference Beck, Joseph and BelisleBeck et al, 2001). We did not use the mental component score as an outcome measure, as its items overlap with those of the HADS; however, we did use the SF–36 energy and vitality sub-scale as a measure of fatigue, for the purposes of the mediator analyses.

Follow-up assessment

The HADS, Illness Perceptions and SF–36 questionnaires were mailed to our study participants at 6 and 12 months following myocardial infarction. At 12 months, participants were asked whether they experienced chest pain and, if appropriate, whether they had returned to work. They were also asked to complete the Seattle Angina Questionnaire (Reference Spertus, Winder and DewhurstSpertus et al, 1995). This 19-item self-rated questionnaire assesses cardiac health status as perceived by the patient on 5 scales: physical limitations owing to chest pain, anginal stability, anginal frequency, treatment satisfaction and perceived burden of cardiac illness.

Participants were also asked to give details of hospital attendances during followup period (dates, hospitals attended and nature of health problem). Hospital attendances were verified by inspection of the appropriate hospital records, during which we confirmed whether participants had experienced any significant cardiac events during the 12 months following the myocardial infarction. These were defined as admissions to hospital or visits to accident and emergency departments with an acute cardiac problem, confirmed by investigations, e.g. chest pain associated with acute changes on echocardiogram. Both angioplasty and coronary artery bypass graft following an acute deterioration in cardiological state during the follow-up period were included.

Statistics

We used hierarchical multiple regression techniques to identify, at baseline and 6 months, variables that contributed to SF–36 PCS at 12 months after myocardial infarction, which was the dependent variable. The following independent variables were entered in blocks in the order presented below.

Block 1 (socio-demographic factors): age, gender, marital status (widowed, divorced, separated or other), education (more or less than 12 years), socio-economic status and baseline SF–36 PCS.

Block 2 (comorbid, non-cardiac illnesses): current rheumatic, respiratory, gastrointestinal or other disorder (including neurological, renal, cerebrovascular disease or endocrine disorder other than diabetes).

Block 3 (cardiac factors): ongoing risk factors for further coronary heart disease (diabetes, serum cholesterol, hypertension, previous cardiac disease, previous or current smoking v. never), measures of severity of index myocardial infarction (Killip class, CPK), site of infarction, use of thrombolysis or surgical intervention during index admission, cardiac medication on discharge and cardiac events during subsequent 12 months.

Block 4 (anxiety and depression): HADS anxiety and depression scores at baseline.

This analysis was then repeated using block 4 depression and anxiety scores recorded 6 months after infarction, in place of the same scores measured at baseline.

Collinearity diagnostics were performed and showed that, although the independent variables were associated, the degree of association was not so great as to threaten the reliability of the regression analyses (variance inflation factors all < 2.5, tolerances all > 0.4) (Reference FieldField, 2000).

In order to check that we had adjusted adequately for cardiac function, the analyses were repeated including the subgroup of 217 participants who had undergone echocardiogram, with left ventricular ejection fraction (categorised as < 30%, 30–50% and > 50%) entered into the regression equation in place of Killip class. In addition, we assessed whether chest pain, fatigue (energy and vitality sub-scale of the SF–36), illness perceptions or self-rated cardiac health status at 12-month follow-up could act as possible mediators of any association between depression/anxiety and SF–36 PCS at 12 months. These potential mediators were added after block 3, to see whether the subsequent contribution of depression/anxiety to the model was reduced, according to the method of Baron & Kenny (Reference Baron and Kenny1986).

Statistics were performed using the Statistical Package for the Social Sciences for Windows, version 11.5.0 (2002).

Follow-up assessment

At 6 months, follow-up data were collected for 268 participants. Mean HADS depression score at 6 months was 5.6 (s.d.=4.0) and mean anxiety score was 7.7 (s.d.=4.6). Mean SF–36 PCS score was 39.0 (s.d.=11.3).

At 12-month follow-up, 16 patients had died and follow-up data were collected from 260 (83% of original sample; 87% of survivors). Participants who did not complete follow-up were significantly younger than those who did, i.e. 53.3 (s.d.=10.4) v. 57.9 (s.d.=11.3) years, respectively, more likely to have low socioeconomic status (73.7% v. 54.3%, respectively), and less likely to have received aspirin (90% v. 96%, respectively). There were no statistically significant differences in other demographic variables, medical history, severity of myocardial infarction, depression and anxiety scores or SF–36 PCS between those who completed assessments and those who dropped out of the study.

Mean HADS depression score at 12 months was 5.4 (s.d.=4.2) and mean anxiety score was 7.5 (s.d.=4.7). The mean SF–36 PCS score at 12-month follow-up was 40.4 (s.d.=11.9); participants who were depressed at this time had lower PCS scores than those who were not, i.e. mean scores=32.7 (s.d.=9.9) v. 43.8 (s.d.=11.1), respectively, t=7.8, P<0.0005.

At the time of their myocardial infarction, 137 patients were employed (part time or full time). Their mean PCS at baseline (adjusted for age and gender) was 48.7 (s.e.m.=0.86) compared with 31.8 (s.e.m.=1.9) for those on sick leave (n=29, F=43.0, P<0.0005). Of those employed at baseline, 3 died during the subsequent 12 months, and we collected complete follow-up data on 100 of the remainder (75%). For the 66 who had returned to work at 12-month follow-up, the mean PCS at follow-up (adjusted for the variables that predicted PCS at 12 months – listed in Table 1) was 46.5 (s.e.m.=1.3) and for the 30 who had not returned to work, it was 39.5 (s.e.m.=1.7, F=9.1, P=0.003).

Table 1 Regression model: predictors of Medical Outcomes Study SF–36 physical component score 12 months¹ after myocardial infarction

Independent	Cumulative adjusted R 2	Significance of change (P)	Variables included in block	Unstandardised coefficients (s.e.)	Standardised beta	t	P
Block 1	0.21	<0.0005	Age in years	-0.23 (0.07)	-0.21	-3.47	0.001
			Baseline physical component score	0.17 (0.06)	0.18	2.86	0.005
Block 2	0.25	0.001	Gastrointestinal problems	-4.37 (1.88)	-0.12	-2.32	0.021
			Rheumatic disease	-3.91 (1.46)	-0.15	-2.68	0.008
			Other medical problems	-4.06 (1.70)	-0.13	-2.39	0.018
Block 3	0.30	0.025	Smoking	-4.44 (1.60)	-0.15	-2.77	0.006
			Site of myocardial infarction	-3.01 (1.42)	-0.12	-2.12	0.04
			Cardiac events during follow-up	-4.16 (1.87)	-0.12	-2.22	0.028
Block 4	0.39	<0.0005	Depression at 6 months	-0.67 (0.23)	-0.21	-2.91	0.004
			Anxiety at 6 months	-0.42 (0.20)	-0.15	-2.09	0.037

Predictors of SF–36 PCS at 12-month follow-up

The variables that were independently associated with low SF–36 PCS at 12-month follow-up are shown in Table 1. Addition of the cardiac variables (block 3) increased the proportion of the variance accounted for (adjusted R ²) by 5%, and further addition of depression and anxiety scores at 6 months (block 4) increased the adjusted R ² by a further 9% (F-change=10.5, P<0.0005). Anxiety and depression measured at baseline and reflecting patient mood in the weeks before first myocardial infarction did not contribute significantly to the regression model after controlling for demographic and medical details (F-change=0.89).

Repeating the above analyses using left ventricular ejection fraction in place of Killip class (n=217) did not change our findings. Depression and anxiety at 6 months continued to contribute significantly to PCS at 12 months (adjusted R ² attributable to block 4=11%, F-change=18.6, P<0.0005).

Assessment of potential mediator variables

Reported chest pain, SF–36 energy and vitality scores, Illness Perceptions Questionnaire and Seattle Angina Questionnaire scores at 12 months were potential mediators, as they all showed significant univariate associations with both depression and anxiety at baseline and 6 months and with PCS at 12 months after myocardial infarction.

When chest pain was entered as a potential mediator, the R ² for depression and anxiety at 6 months decreased from 9% to 5.0%, but the R ² attributable to depression and anxiety remained significant at P<0.0005. When SF–36 energy and vitality score was added, the R ² for depression and anxiety decreased to 0% (P=0.37). Addition of the Illness Perceptions Questionnaire sub-scores, either singly or together, reduced the R ² that was independently attributable to the depression and anxiety measures, although in all models the contributions of depression and anxiety remained significant (P≤0.01). Similarly, addition of the scores from the Seattle Angina Questionnaire reduced the magnitude of the R ² attributable to anxiety and depression, although it remained significant (all P≤0.002).

The effects of changing depression on SF–36 PCS

To clarify the effect of depression severity changing between baseline and 12-month follow-up, we compared changes in the SF–36 PCS of 38 participants who were not depressed at baseline but who became depressed during follow-up, with those of 143 participants who remained undepressed throughout the study period, using HADS total of ≥17 to indicate depression. The HADS depression scores of these two groups at baseline were 3.2 (2.2) and 2.7 (2.5), respectively, (t=–1.3, P=0.2), confirming the non-depressed status of both groups at baseline. There was no significant difference in SF–36 PCS at baseline between these two groups: scores were 42.1 (12.8) and 45.5 (11.5), respectively (t=1.6, P=0.12). The HADS depression score at 12 months for individuals who became depressed was 10.0 (3.2) whereas for those who remained nondepressed it was 3.0 (2.2), (t=–15.6, P<0.0005). The SF–36 PCS for these groups were 30.9 (7.4) and 44.1 (10.7) (t=8.8, P<0.0005).