In-Hospital Symptoms of Psychological Stress as Predictors of Long-Term Outcome After Acute Myocardial Infarction in Men Nancy Frasure-Smith, PhD
The impact of high levels of psychological stress symptoms in the hospital after an acute myocardial infarction (AMI) was examined over 5 years among 461 men who took part in a trial of psychological stress monitoring and intervention. Psychological stress was assessed using the 20.item GenerAI Health Questionnaire (GHQ) 1 to 2 days before hospital discharge. Once discharged, patients in the treatment group responded to the GHQ by telephone on a monthly basis and, when they reported high levels of stress symptoms (GHQ ZS), received visits from nurses to help them deal with their life problems. Control patients received routine medical care after discharge. Post-hoc subgroup analyses based on life-table methods showed that, for patients receiving routine care after discharge, high stress (GHQ 25) was associated with a close to threefold increase in risk of cardiac mortality over 5 years (p = 0.0003) alid an approximately 1.5 fold increase in risk of reinfarction over the same period (p = 0.09). In contrast, highly stressed patients who took part in the l-year program of stress monitoring and intervention did not experience any significant long-term increase in risk. Although program impact was significant in terms of reduction of both cardiac mortality (p = 0.006) and AMI recurrences (p = 0.004) among highly stressed patients, there was little evidence of impact among patients with low levels of stress in the hospital. These results add to the growing body of research implicating psychosocial factors in postAMI outcomes, and suggest that the patients who can benefit most from interventions to alter these factors may be identified before hospital discharge. (Am J Cardiol 1991;67:121-127)
From the Department of Psychiatry and School of Nursing, McGill University, Montreal, Quebec, Canada. This research was supported by the National Health Research and Development Program of Canada through Projects 605-1303-44,6605-2388-44 and 6605-2022-48. Manuscript received August 9, 1990, revised manuscript received and accepted September 12, 1990. Address for reprints: Nancy Frasure-Smith, Department of Psychiatry, McGill University, 1033 Pine Avenue West, Montreal, Quebec, Canada H3A 1A 1.
vv
e recently completeda long-term study of 461 men with acute myocardial infarction (AMI) who had taken part in a l-year randomized, controlled trial of a program of monthly monitoring of psychological stress levels coupled with home nursing visits for highly stressedpatients.1,2The program significantly reduced stressscores,had a marginal impact on cardiac mortality over the first postinfarct year, and significantly reduced long-term AM1 recurrences. The trial’s rationale was that peaks in life stressoccurring after hospital discharge have important consequencesfor AM1 patients, and that intervention at the time of these peaks tiuld make a difference in outcomes.It was not expected that high levels of stressin the hospital would have any prognostic significance. However, recent evidence of the long-term impact of life stressbefore an AMI suggested2 post-hoc hypotheses:first, that among the control patients who received routine care, high levels of in-hospital stresswould have long-term negative prognostic significance; and second, that treatment group participation would buffer the long-term impact of in-hospital stress, so that, with treatment, high levels of stresswould not be accompanied by an increase in long-term risk. METHODS Treatment
program: A major issue in program development was determining how to monitor life stress. Although inferring life stress from the occurrence of major life changes, such as losing a job or getting a divorce, can be a useful approach in studying groups,4 evidence of individual variation in the meaning of life events5led to the selection of a more patient-specific approach for clinical purposes.Stress was conceptualized in terms of the psychological symptomsthat people having difficulty with life eventsfrequently report (such as sleepdifficulties, problems concentrating and feelings of inadequacy). A standardized questionnaire, the General Health Questionnaire (GHQ),6 was administered by telephone on a monthly basis in order to assess20 such symptoms. The GHQ was originally developedto identify those medical patients most likely to benefit from psychiatric treatment. However, research crossvalidating the GHQ with psychiatric ratings has shown that, at times of adverse life events, people without psychiatric illness often report as many psychological symptoms as do psychiatric patients.7 This has led to the interpretation of GHQ scoresas indexes of psycho-
THE AMERICAN JOURNAL OF CARDIOLOGY JANUARY 15, 1991
121
logical distress that are more or less transient, depending on the life circumstancesand personality of the respondent.8For example, patients are asked, “Over the past week, have you recently been able to concentrate on whatever you’re doing? Would you say: better than usual, same as usual, less than usual, or much lessthan usual?” When the symptom is worse or much worse than usual, the patient is considered to have 1 stress symptom. Preliminary research using the GHQ among post-AM1 patients showedthat those reporting 15 such symptoms were more likely than others to be readmitted to hospital in the next few months.9 Therefore, for the present study, patients with 25 GHQ stresssymptoms were considered to have high stresslevels and to need interventions to reduce stress. Interventions for highly stressed patients involved home visits from nurses who received no special training, but who had experiencein coronary care and were sensitive to patients’ psychological needs. Becausethe monitored stress symptoms could have been triggered by a variety of life problems needing a variety of solutions, the treatments were matched, in terms of type, amount and timing, to the needsof each highly stressed patient. The nurses’ mandate was to visit each highly stressedpatient and to attempt to determine the aspects of his life that contributed to his stress.Later, in consultation with the project team (a psychiatrist, a psychologist and other project nurses), nurses did whatever was relevant to help each patient with his problems. Nurses had their own caseloads, and the nursing interventions involved an individually tailored combination of teaching, support and consultation or referral strategies.Visits continued until the patient’s stressscorewas brought down to normal and until the nurse believed that his problems were resolved. Although the average highly stressedpatient required 5 to 6 hours of nursing over 5 to 6 months, only about half of the patients had stress scoreshigh enough to require visits from nurses during the year of stressmonitoring. Patient population: Sample selection procedures have been outlined in previous reports,lj2 and methodologic issues and study limitations have also been discussed at length.lJO In brief, from November 1977 to May 1981, 539 men recovering from an AM1 at 3 McGill University hospitals were randomly assignedto treatment or control (routine medical care) conditions and asked to take part in the study. Diagnosis of AM1 was based on the cardiologist’s diagnosis at the time of discharge from coronary intensive care, and was verified by the presence of 2 of the following criteria: typical infarct-related pattern of chest pain lasting 230 minutes, usual evolutionary electrocardiographic changes, and elevation of the creatine phosphokinaselevel above normal for each hospital. Informed consent was given by 83% of treatment patients and 89% of control patients, resulting in a final sample size of 461 men (232 treatment and 229 control) ranging in age from 28 to 86 years (average 58). Only 25% had attended a university and 82% were married. Twelve percent were diabetic and 58% reported that they smoked. Twenty-four percent had non-Q-wave AMIs. In almost 55%, the 122
THE AMERICAN JOURNAL OF CARDIOLOGY VOLUME 67
AM1 was complicated (congestiveheart failure, pulmonary edema,pericarditis, arrhythmias after 48 hours in the hospital, ventricular tachycardia, recurrent chest pain, pulmonary emboli). Seventy-five percent had no previous cardiac hospitalizations. Group comparisonsof baselinevariables showedthat the control group had a larger proportion of low socioeconomic status patients, more patients with elevated creatine phosphokinaselevels and more patients who neededdiamorphine for cardiac pain after 48 hours in the hospital. Although, as discussedelsewhere,1,2J0 the most likely explanation for these imbalances involves the timing of randomization and a consequentdifference between treatment and control groups in study participation rates, these baseline differences require statistical control. Data collection: All patients were interviewed 1 to 2 days before discharge using a semistructured questionnaire that included the GHQ as well as questions to assessbehavioral risk factors for myocardial infarction and demographic, social and psychological variables. Other medical and risk factor data, seen as prognostitally valuable at the time of study design,” were obtained from patients’ hospital charts. In addition, becauseof recent evidencesuggestingthe importance of Q waves in determining prognosis,r2 electrocardiograms were coded blindly with the modified Minnesota codes used in the Diltiazem Reinfarction study.” Only 6% of patients had indeterminate electrocardiograms.Patient records in study hospitals, Quebec Medicare data and death certificates were used to obtain data on AM1 recurrences and deaths during the period of the program for 98% of the sample.Proceduresusedto code the data were detailed previously.’ Thesesamerecordswere used to follow patients up to 25 years after the index AMI, at which point lossesto follow-up amounted to only 12% in each group. Data analysis: Record-basedoutcome data for all subjects (including the 27 treated patients who withdrew from the study during the program year) were assessedby life-table methods. Times to the first occurrence of each major outcome (cardiac mortality and AM1 recurrence regardlessof survival) were the dependent variables. Censoring occurred at the time of loss to follow-up. Patients were subdivided into high- and low-stress groups on the basisof GHQ scoresin the hospital. As in the treatment program, those with GHQ scores 15 were consideredhighly stressedpatients and those with GHQ scores
The impact of high levels of psychological stress symptoms in the hospital after an acute myocardial infarction (AMI) was examined over 5 years among 461 men who took part in a trial of psychological stress monitoring and intervention. Psychological stress was assessed using the 20.item GenerAI Health Questionnaire (GHQ) 1 to 2 days before hospital discharge. Once discharged, patients in the treatment group responded to the GHQ by telephone on a monthly basis and, when they reported high levels of stress symptoms (GHQ ZS), received visits from nurses to help them deal with their life problems. Control patients received routine medical care after discharge. Post-hoc subgroup analyses based on life-table methods showed that, for patients receiving routine care after discharge, high stress (GHQ 25) was associated with a close to threefold increase in risk of cardiac mortality over 5 years (p = 0.0003) alid an approximately 1.5 fold increase in risk of reinfarction over the same period (p = 0.09). In contrast, highly stressed patients who took part in the l-year program of stress monitoring and intervention did not experience any significant long-term increase in risk. Although program impact was significant in terms of reduction of both cardiac mortality (p = 0.006) and AMI recurrences (p = 0.004) among highly stressed patients, there was little evidence of impact among patients with low levels of stress in the hospital. These results add to the growing body of research implicating psychosocial factors in postAMI outcomes, and suggest that the patients who can benefit most from interventions to alter these factors may be identified before hospital discharge. (Am J Cardiol 1991;67:121-127)
From the Department of Psychiatry and School of Nursing, McGill University, Montreal, Quebec, Canada. This research was supported by the National Health Research and Development Program of Canada through Projects 605-1303-44,6605-2388-44 and 6605-2022-48. Manuscript received August 9, 1990, revised manuscript received and accepted September 12, 1990. Address for reprints: Nancy Frasure-Smith, Department of Psychiatry, McGill University, 1033 Pine Avenue West, Montreal, Quebec, Canada H3A 1A 1.
vv
e recently completeda long-term study of 461 men with acute myocardial infarction (AMI) who had taken part in a l-year randomized, controlled trial of a program of monthly monitoring of psychological stress levels coupled with home nursing visits for highly stressedpatients.1,2The program significantly reduced stressscores,had a marginal impact on cardiac mortality over the first postinfarct year, and significantly reduced long-term AM1 recurrences. The trial’s rationale was that peaks in life stressoccurring after hospital discharge have important consequencesfor AM1 patients, and that intervention at the time of these peaks tiuld make a difference in outcomes.It was not expected that high levels of stressin the hospital would have any prognostic significance. However, recent evidence of the long-term impact of life stressbefore an AMI suggested2 post-hoc hypotheses:first, that among the control patients who received routine care, high levels of in-hospital stresswould have long-term negative prognostic significance; and second, that treatment group participation would buffer the long-term impact of in-hospital stress, so that, with treatment, high levels of stresswould not be accompanied by an increase in long-term risk. METHODS Treatment
program: A major issue in program development was determining how to monitor life stress. Although inferring life stress from the occurrence of major life changes, such as losing a job or getting a divorce, can be a useful approach in studying groups,4 evidence of individual variation in the meaning of life events5led to the selection of a more patient-specific approach for clinical purposes.Stress was conceptualized in terms of the psychological symptomsthat people having difficulty with life eventsfrequently report (such as sleepdifficulties, problems concentrating and feelings of inadequacy). A standardized questionnaire, the General Health Questionnaire (GHQ),6 was administered by telephone on a monthly basis in order to assess20 such symptoms. The GHQ was originally developedto identify those medical patients most likely to benefit from psychiatric treatment. However, research crossvalidating the GHQ with psychiatric ratings has shown that, at times of adverse life events, people without psychiatric illness often report as many psychological symptoms as do psychiatric patients.7 This has led to the interpretation of GHQ scoresas indexes of psycho-
THE AMERICAN JOURNAL OF CARDIOLOGY JANUARY 15, 1991
121
logical distress that are more or less transient, depending on the life circumstancesand personality of the respondent.8For example, patients are asked, “Over the past week, have you recently been able to concentrate on whatever you’re doing? Would you say: better than usual, same as usual, less than usual, or much lessthan usual?” When the symptom is worse or much worse than usual, the patient is considered to have 1 stress symptom. Preliminary research using the GHQ among post-AM1 patients showedthat those reporting 15 such symptoms were more likely than others to be readmitted to hospital in the next few months.9 Therefore, for the present study, patients with 25 GHQ stresssymptoms were considered to have high stresslevels and to need interventions to reduce stress. Interventions for highly stressed patients involved home visits from nurses who received no special training, but who had experiencein coronary care and were sensitive to patients’ psychological needs. Becausethe monitored stress symptoms could have been triggered by a variety of life problems needing a variety of solutions, the treatments were matched, in terms of type, amount and timing, to the needsof each highly stressed patient. The nurses’ mandate was to visit each highly stressedpatient and to attempt to determine the aspects of his life that contributed to his stress.Later, in consultation with the project team (a psychiatrist, a psychologist and other project nurses), nurses did whatever was relevant to help each patient with his problems. Nurses had their own caseloads, and the nursing interventions involved an individually tailored combination of teaching, support and consultation or referral strategies.Visits continued until the patient’s stressscorewas brought down to normal and until the nurse believed that his problems were resolved. Although the average highly stressedpatient required 5 to 6 hours of nursing over 5 to 6 months, only about half of the patients had stress scoreshigh enough to require visits from nurses during the year of stressmonitoring. Patient population: Sample selection procedures have been outlined in previous reports,lj2 and methodologic issues and study limitations have also been discussed at length.lJO In brief, from November 1977 to May 1981, 539 men recovering from an AM1 at 3 McGill University hospitals were randomly assignedto treatment or control (routine medical care) conditions and asked to take part in the study. Diagnosis of AM1 was based on the cardiologist’s diagnosis at the time of discharge from coronary intensive care, and was verified by the presence of 2 of the following criteria: typical infarct-related pattern of chest pain lasting 230 minutes, usual evolutionary electrocardiographic changes, and elevation of the creatine phosphokinaselevel above normal for each hospital. Informed consent was given by 83% of treatment patients and 89% of control patients, resulting in a final sample size of 461 men (232 treatment and 229 control) ranging in age from 28 to 86 years (average 58). Only 25% had attended a university and 82% were married. Twelve percent were diabetic and 58% reported that they smoked. Twenty-four percent had non-Q-wave AMIs. In almost 55%, the 122
THE AMERICAN JOURNAL OF CARDIOLOGY VOLUME 67
AM1 was complicated (congestiveheart failure, pulmonary edema,pericarditis, arrhythmias after 48 hours in the hospital, ventricular tachycardia, recurrent chest pain, pulmonary emboli). Seventy-five percent had no previous cardiac hospitalizations. Group comparisonsof baselinevariables showedthat the control group had a larger proportion of low socioeconomic status patients, more patients with elevated creatine phosphokinaselevels and more patients who neededdiamorphine for cardiac pain after 48 hours in the hospital. Although, as discussedelsewhere,1,2J0 the most likely explanation for these imbalances involves the timing of randomization and a consequentdifference between treatment and control groups in study participation rates, these baseline differences require statistical control. Data collection: All patients were interviewed 1 to 2 days before discharge using a semistructured questionnaire that included the GHQ as well as questions to assessbehavioral risk factors for myocardial infarction and demographic, social and psychological variables. Other medical and risk factor data, seen as prognostitally valuable at the time of study design,” were obtained from patients’ hospital charts. In addition, becauseof recent evidencesuggestingthe importance of Q waves in determining prognosis,r2 electrocardiograms were coded blindly with the modified Minnesota codes used in the Diltiazem Reinfarction study.” Only 6% of patients had indeterminate electrocardiograms.Patient records in study hospitals, Quebec Medicare data and death certificates were used to obtain data on AM1 recurrences and deaths during the period of the program for 98% of the sample.Proceduresusedto code the data were detailed previously.’ Thesesamerecordswere used to follow patients up to 25 years after the index AMI, at which point lossesto follow-up amounted to only 12% in each group. Data analysis: Record-basedoutcome data for all subjects (including the 27 treated patients who withdrew from the study during the program year) were assessedby life-table methods. Times to the first occurrence of each major outcome (cardiac mortality and AM1 recurrence regardlessof survival) were the dependent variables. Censoring occurred at the time of loss to follow-up. Patients were subdivided into high- and low-stress groups on the basisof GHQ scoresin the hospital. As in the treatment program, those with GHQ scores 15 were consideredhighly stressedpatients and those with GHQ scores
