Mora, Pablo A. MS; Robitaille, Chantal PhD; Leventhal, Howard PhD; Swigar, Mary MD, and; Leventhal, Elaine A. PhD, MD
CI = confidence interval;, NA = negative affect;, OR = odds ratio;, SAH = self-assessed health.
Two types of variables were related to the reporting of symptoms and to the use of medical care in this study of over 700 community-dwelling, well-educated, mostly retired, white elderly subjects (mean age = 73): 1) stable characteristics of the participants (trait NA, that is, reports that one is usually tense, anxious, sad, and/or blue (1), and SAH (2, 3)) and 2) the perceived properties of specific illness episodes. Our data allowed us to make two types of comparisons. One type of comparison explored the relationship of trait NA with the reporting of symptoms, illness episodes, and use of medical care to the relationship of SAH with the same health variables. These comparisons are of interest because trait NA is assumed to be related to biased, over-reporting of symptoms, whereas SAH is a powerful predictor of critical health outcomes such as mortality. The second type of comparison explored the relationship of stable, person characteristics (trait NA and SAH) with the use of medical care during illness episodes to the relationship of the features of the episode (eg, its duration, novelty or unexpectedness of its symptoms, and episode-specific worry) with this outcome.
Expectations regarding the relationship between these predictors and outcomes were generated from the commonsense model of illness cognition. This framework proposes that people are active problem solvers who assess the meaning of somatic sensations by forming hypotheses about their identity (ie, label), cause (ie, stress or virus), timeline, controllability, and consequences (4). These hypotheses guide the selection, performance, and evaluation of coping procedures for the management of illness episodes. Feedback from these procedures can reshape the hypotheses and alter subsequent coping actions (5, 6). The constructionist or “problem solving” emphasis of this framework suggests that the perceived features of illness episodes (ie, whether the symptoms exceed their expected duration, are novel or unexpected, and pose images and thoughts of severe consequences) will be more important as predictors of the use of medical care than will personality characteristics such as trait NA or SAH (7, 8). Thus, our overall expectation was that the use of medical care would be predicted mainly by properties of the episode.
We also expected that the relationships of trait NA and SAH to care seeking would differ for acute and chronic episodes. Episode properties should dominate care seeking for common, acute conditions because they have no special meaning for the self system. Thus, NA and SAH are of less consequence for care seeking for such conditions in comparison to the features of the specific episode, that is, whether it proves overly long and severe or has novel and unexpected features. Person characteristics, particularly SAH, should contribute to decisions to seek care for episodes of chronic conditions because these long-lasting conditions have implications for the overall health status of the self (9).
A second set of expectations was designed to clarify the processes underlying the relationship of trait NA to reports of everyday symptoms, illness episodes, episode-specific symptoms, and the use of health care. Multiple studies have shown that trait NA has a positive relationship to current symptom reports but has no significant relationship to objective measures of health (eg, 10, 11), indicating that correlations of trait NA and symptom reports cannot be used as evidence that trait NA is a cause of illness (10, 12). These relationships led to the formulation of the symptom perception hypothesis (10), which states that high levels of vigilance and awareness of internal stimuli and signs of threat (13) underlie the association of trait NA to the reporting of current symptoms in the absence of disease. Because similar hypotheses have been advanced to account for the relationship of hypochondriasis and somatization to current symptom reports (14, 15), it might be expected that individuals who are high on trait NA would also be more likely to report being sick and more likely to use medical care during illness episodes.
The commonsense framework suggests a different perspective on the relationship of NA to symptom reporting by elderly individuals. Because the elderly have a substantial history of interpreting symptoms associated with acute and chronic conditions, both minor and possibly serious, it is reasonable to assume that they will develop a degree of expertise in identifying when they are sick vs. when they are well. When sick, they should be able to determine whether their condition is acute or chronic, novel or expected, highly symptomatic or relatively mild, and conforming to an expected time course and trajectory. From the commonsense framework, the vigilance associated with trait NA is an additional source of motivation for the acquisition of this type of self-knowledge. Decades of vigilance can be expected, therefore, to create a form of commonsense, although anxious, expertise. Thus, the elderly high trait NA respondent will have a detailed perspective of his or her usual, somatic background, that is, ongoing chronic symptoms, and the ability to identify departures from this background and develop appropriate strategies such as seeking health care to deal with illness-related departures and their associated emotional upset (4, 6, 16).
If high NA is associated with such commonsense expertise but is unrelated to vulnerability to illness, we can expect NA to be positively associated with current (ie, the “everyday” background) symptoms, particularly symptoms of a chronic nature, and to be unrelated to reporting of illness episodes, particularly those of acute, expected, common conditions (eg, colds). We would also expect trait NA to be unrelated to seeking medical care for acute episodes, although it might encourage care seeking for chronic episodes, particularly for difficult-to-control conditions. Although this expertise will seem psychologically “rational,” it may depart from rationality when the elderly, trait-anxious person believes the threat to be serious, that is, when control efforts fail, hence our inclination to label this expertise as “anxious.” SAH was included in our analyses because we expected that comparing its relationship with symptom reporting and care seeking to the similar relationships of trait NA would further support our interpretation of trait NA as a form of commonsense expertise when assessed in elderly respondents. Although SAH and trait NA are related to one another, with high levels of trait NA associated with low levels of SAH (r values ranging from −0.30 to −0.45), SAH has a strong relationship to health outcomes (17). Unlike many other psychological factors, which have a small or no relationship with health outcomes for people over 65 years of age (18), low self-ratings of health (poor and fair SAH) are related to both self-reported symptoms and mortality (eg, 19); both relationships replicate with regularity, and both are substantial in magnitude. Indeed, individuals reporting poor to fair SAH are two or more times as likely to be deceased 2 to 20 years after their self-assessments. These effects appear in analyses that control for multiple factors including age and objective measures of physical health (2, 3). The data for SAH suggest that some subjective reports can exceed objective indicators in predicting objective health outcomes.
Cross-sectional and longitudinal data were analyzed to assess the relationship of trait NA and of SAH with prior-week current (ie, background) symptoms (those judged to be acute and those judged to be chronic), illness episodes, and symptoms reported for these episodes. The use of medical care during episodes of acute and chronic illness was predicted using both stable self-appraisals (eg, trait NA and SAH) and episode-specific measures, such as number of symptoms, severity and duration of the episode, and the novelty of the episode. Additionally, we examined the relationship of these stable characteristics and episode-specific variables to care seeking in a subset of participants who reported ongoing illness episodes at the time of the interview. Our expectations were as follows:
1. Compared with their low trait NA peers, elderly individuals who reported high levels of trait NA were expected to be more aware of and report a greater number of “background” symptoms during the prior week, particularly of symptoms they self-identified as chronic. Because high and low trait NA elderly are believed to be equally susceptible to physical illness, we expected them to report similar frequencies of illness episodes during the prior 3 months.
2. If we assume that high trait NA elderly are more attentive to their bodies and less likely to ignore somatic changes, especially when those changes signal potential threat, we could expect that they would detect and report more symptoms during chronic episodes (ie, conditions that are cyclical and impinge on individual functioning) but would not do so for typical, acute illness episodes. These assumptions led to two hypotheses: 1) compared with low trait NA individuals, high trait NA individuals would report more symptoms during episodes or flares of chronic conditions in the prior 3 months and 2) both high and low trait NA individuals would report similar numbers of symptoms for episodes of acute illness in this same prior 3-month period.
3. We expected that the properties of the illness episode would affect care seeking and would do so in similar degrees for high and low trait NA individuals. Thus, we hypothesized that 1) properties of the episode, particularly its duration and its novelty or unexpectedness, would have a greater impact on care seeking than would trait NA or SAH and 2) SAH but not trait NA would relate to care seeking for chronic conditions because poor SAH implies vulnerability to decline in function from chronic conditions.
4. We also examined the relationship of emotional reactions during ongoing illness episodes (ie, episode-based worry) to care seeking in a subset of our respondents, allowing us to assess and contrast the relationship of trait NA and emotional states to care seeking. We expected that episode-specific worry would lead to care seeking and trait NA would not.
Data for these analyses were obtained during the second (1992) and third annual interviews (1993) of a 9-year longitudinal survey of community-dwelling older adults in central New Jersey (1992:N = 790; 1993:N = 719). The 1992 interview served as the baseline (wave 1), and the 1993 interview served as the 12-month follow-up (wave 2). 1 Subject loss averaged 9.75% per year due, in approximately equal parts, to death, moving away from the community, and withdrawal from the study. The mean age at baseline was 72.8 years (range, 49–93), and 83.6% of the participants were >65 years of age. The division by sex at baseline was 479 (56%) female and 311 male (44%), and 72% of the female subjects and 80% of the male subjects had >12 years of education.
Design and Procedure
The two waves of data allowed for the examination of cross-sectional and longitudinal relationships of NA and SAH to reports of acute and chronic symptoms in the prior week (current background symptoms), acute and chronic illness episodes experienced during the prior 3 months, the symptoms associated with each of these episodes, and the use of medical care for each episode. All participants answered questions about current acute and chronic symptoms during the prior week, and all subjects reported on the presence of acute and chronic illness episodes during the past 3 months. Approximately one-third of the respondents reported having had at least one illness episode in the past 3 months at each wave (1992: acute = 190, chronic = 264; 1993: acute = 197, chronic = 234).
Care seeking was predicted cross-sectionally at both waves (1992 and 1993) using both trait and episode-specific factors. 2 Longitudinal analyses assessing the effects of trait variables on change in both current acute and current chronic symptoms entered current 1992 acute or current 1992 chronic symptoms (ie, baseline symptom measure) as the first factor in their respective analyses. Longitudinal analyses for acute and chronic episodes (1993 outcome) used 1992 episode reports as the first control variable. Because different individuals reported illness episodes at each of the two annual waves, longitudinal analyses for episode-linked symptoms reported during longitudinal analyses for episode-linked symptoms reported during 1993 episodes used current (prior week) 1992 symptoms as a control because all participants had values for this control variable. In all of these analyses, age, illness burden, and sex were entered as controls after the baseline values for dependent measure and then were followed by trait NA and SAH from the 1992 interview.
A more detailed report of factors affecting care seeking was obtained from a subset of respondents who reported an acute and/or chronic illness episode that was ongoing at the time of the interview: 43 subjects reported an ongoing acute condition, and 121 subjects reported an ongoing chronic condition. These respondents provided information about the course of their episode and a measure of episode-linked emotional state, that is, how much they worried about the illness. Because these respondents are included in the overall analyses of responses during acute or chronic illnesses during the prior 3 months, we conducted a separate set of analyses from which this subset was excluded. Because these latter findings replicated findings using all participants, we report the data both for the subset and for the complete set, which includes these 164 participants.
Measures Assessed for all Participants: Trait NA and SAH
Trait NA was assessed with two, five-item adjective lists, one for depressed (eg, blue; sad) and one for anxious (eg, tense; worried) mood. Each question asked, “How ______ are you usually?”; responses were recorded on five-point Likert scales (range, 1 = not at all to 5 = very much). Cronbach’s alphas were very high for the five-item scales assessing depressed (α = 0.91 and 0.90, for 1992 and 1993, respectively) and anxious mood (α = 0.88 for both years) and higher yet for the combined scale of trait NA (1992: α = 0.93; 1993: α = 0.92). Analyses examining the stability of NA over time revealed test-retest figures of rtt = .74 over 1 year and rtt = .71 over 2 years. These figures match those for other widely used trait inventories such as the Neuroticism, Extraversion, Openness Personality Inventory (NEO PI) (20, 21) and the Positive Affect-Negative Affect Scale (PANAS) (22). Preliminary results from Latent Growth Curve analyses (23, 24) examining the intra-individual stability of NA indicated that the NA measure was very stable, showing that the average elderly participant realized an approximately 2% change in NA scores over a 1-year period and an approximately 7% change over a 2-year period.
Self-assessments of health were reported in response to a single item: “In general, how would you rate your health. . .excellent, very good, good, fair, or poor.” Responses to this item have proven to be a powerful and consistent predictor of health outcomes such as mortality (2, 3). The test-retest reliability for this single item was 0.66 over 1 year and 0.61 over 2 years. Latent Growth Curve analyses showed that an average individual realized a 3% change in SAH scores over 1 year and a 6% change over 2 years.
Health Status Measures
Current Symptom Check List.
Current symptoms were assessed by 45 questions that asked about problems experienced in specific parts of the body during the prior week. The stem for each question asked: “Have you had any problems or trouble with ______ in the past week?”; the blank was filled with specific areas of the body (eg, lung, arm, hand, or shoulder). Questions about specific symptoms (eg, chest discomfort or pain, stomachaches, or diarrhea) substituted, “symptoms” for “problems or trouble.” For each problem, respondents indicated whether it was acute or chronic. The seven response categories were not at all, yes acute (mild, moderate, or severe), or yes chronic (mild, moderate, or severe). Separate scores for acute and chronic symptoms were obtained by summing the number of yes responses in each category (actual range: acute, 0–12, chronic, 0–19). The questions and format were based on the standard review of systems used in internal medical practice (25).
Acute and chronic illness episodes experienced during the 3 months before the interview were assessed by asking “In the past 3 months, did you have any episodes of illness, such as a cold, pneumonia, or upset stomach?” and “In the past 3 months, did you have the onset of a new chronic condition or the recurrence, worsening or flare-up of an existing chronic condition or health problem?” The computer interview allowed for recording of up to four different episodes in each category. Separate scores were generated for acute and chronic episodes.
Measures of Illness Episodes: Prior and Ongoing
Subjects who reported having had an acute or chronic episode in the prior 3 months answered a series of questions covering the properties of those episodes. The questions asked about the duration of the episode: “About how long did it last (or has it lasted)?” (responses transposed to hours); its symptoms: “What symptoms went along with it?” (scored in number of symptoms); and severity: “How severe was it?” (five-point scale from not at all to very).
Current (Ongoing) Episodes.
An additional series of questions was posed to the subset of individuals who reported an episode that was ongoing at the time of interview. Questions were asked about the speed of onset, severity (separately at onset and at worst point), responses to control the episode and their effectiveness (eg, use of over-the-counter medications), and perceived causes of the episode (eg, illness, age, and stress); however, we examined only the data from the four questions asking about emotional reactions. Worry at episode onset was assessed by two questions, “When you first noticed any symptoms, how nervous were you about it and its consequences?” and “. . .how anxious were you about it and its consequences?” Worry when the episode was at its worst was assessed by two additional items, “When it was at its worst, how nervous were you about it and its consequences?” and “. . .how anxious were you about it and its consequences?” Five-point scales (not at all, a little bit, moderately, quite a bit, and very) were used to assess worry, severity, and cause. Alpha for the four-item worry scale was high (0.88).
Novelty or Unexpectedness of the Illness Episode.
Each person’s self-defined illness episode was reviewed by a board-certified geriatric physician and classified for its likely perceived novelty or unexpectedness. Two factors were involved in the decision: 1) the ambiguity vs. the distinctiveness of the symptoms as signs of a specific illness and 2) how likely it would be for an elderly person to decide that she or he was ill and needed medical care given the label and symptoms reported. An illness episode was classified as not novel (coded as 0) if the symptoms were not distinctive indicators of serious disease and/or people would be likely to discount the symptoms as indicating a benign condition (eg, cold or aging). All other illness episodes were categorized as novel (coded as 1).
A measure of lifetime illness burden was created from a detailed review of each individual’s illness history. Each reported condition was rated for 1) level of functional impairment (scale of 1–100) and 2) threat to life (scale of 1–100). The ratings by six internists were averaged to form a severity index that served as the estimate of the burden imposed by that illness. Cronbach’s α for the six ratings across all diseases was 0.97. Although the alpha coefficient was high, the decision was made to drop the highest and lowest ratings, the average of the remaining four serving as the final severity score. The individual’s illness burden was the sum of these scores. At baseline, participants reported an average of 6.60 (range, 0–20) illness conditions for their lifetime, and their average illness burden score at baseline was 170.21 (range, 0–692).
Five items assessed limitations in daily function (α = 0.85): 1) “Does your health limit the kinds or amounts of vigorous activities you can do such as running, lifting heavy objects, or participating in strenuous sports or activities?”; 2) “Does your health limit the kinds or amounts of moderate activities you can do such as moving a table, carrying groceries, bending, or lifting?”; 3) “Do you have any trouble walking one block, uphill, or a few flights of stairs?”; 4) “How much do problems with your health stand in the way of doing the things you would like to do?”; and 5) “How much do problems with your health stand in the way of doing the things you need to do?” The response scale ranged from 1 = not at all to 5 = very much. The total score for each subject was computed by averaging responses to the five items.
Self-Appraisals (NA and SAH) and Prior-Week Symptoms
The zero-order correlations between trait NA and reports of current (prior week) acute and chronic symptoms are shown in Table 1 for both waves of data (1992 and 1993: all p values <.01). The magnitude of the associations replicated prior research (10) by showing that higher levels of trait NA were related to increased symptom reporting. These correlations were essentially the same for both women and men and changed no more than ±0.05 units when age and illness burden were entered as controls. The results were very much the same for the correlations of SAH to current acute and chronic symptom reports (Table 1). Additionally, we examined the relationships of NA and SAH to the perceived severity of current symptoms. These results showed that NA was related to the perceived severity of prior-week chronic symptoms (r values = .14, p < .01 for both waves) but was unrelated to the perceived severity of acute symptoms in both waves (r values <.08, NS). SAH, on the other hand, was related to the perceived severity of both acute and chronic symptoms (1992: acute r = −.15, chronic r = −.25; 1993: acute r = −.19, chronic r = −.28; all p values <.01).
Regression models assessed the joint relationship of trait NA, the factor imputed to be a source of reporting error, and SAH, the factor known to be a powerful predictor of health outcomes, to the number of both current acute and current chronic symptoms in both waves. The models accounted for significant variance at both waves for both current acute and current chronic symptoms (acute: 1992 R2 = 0.09, F (5,783) = 15.06, p < .01; 1993 R2 = 0.10, F (5,713) = 16.37, p < .01; chronic: 1992 R2 = 0.28, F (5,783) = 59.56, p < .01; 1993 R2 = 0.28, F (5,713) = 56.27, p < .01). The analyses showed that trait NA and SAH accounted for more variance in chronic symptoms than for variance in acute symptoms at both waves (Table 2). Although trait NA and SAH were moderately correlated (1992:r = −.34; 1993:r = −.37), there was little overlap in the variance they accounted for in symptom reporting; each accounted for a similar amount of variance whether entered before or after the other in the hierarchical models.
Both trait NA and SAH reported in wave 1 significantly predicted reports of current (prior week) symptoms 1 year later (Table 2, prospective analyses). The amount of variance accounted for by trait NA and SAH was statistically significant but small, less than 2% for each of the two factors after controlling for reports of symptoms at wave 1 (1992), age, sex, and illness burden. The total models were significant both for acute (R2 = 0.08, F (6,712) = 9.61, p < .01) and chronic symptoms (R2 = 0.35, F (6,712) = 63.98, p < .01). In addition to the significant increases in current acute and chronic symptoms predicted by both trait NA and SAH over a 1-year interval, sex predicted increases in acute symptoms (women reporting more changes in symptoms) 1 year later, and older age predicted an increase in the number of current chronic symptoms reported 1 year later.
Because studies have shown that older women report more symptoms and have higher frequencies of arthritic conditions than older men (26), separate analyses were conducted for men and women, respectively. These analyses showed that both trait NA and SAH predicted small yet significant increases in chronic symptoms for women (NA: R2 = 0.03, F (1,437) = 19.20, p < .01; SAH: R2 = 0.03, F (1,436) = 17.45, p < .01), but not for men (R2 values = 0.00, F values <2.64, NS). The pattern was different for acute symptoms. Trait NA was not a significant predictor of acute symptoms one year later for women (R2 = 0.00, F < 1.00, NS) but was for men (R2 = 0.07, F (1,272) = 20.57, p < .01). SAH showed the opposite pattern; it was a significant predictor of subsequent changes in acute symptoms for women (R2 = 0.02, F (1,436) = 6.55, p < .05) but was not for men (R2 = 0.00, F < 1.00, NS). Once again, it is important to note that the effect sizes were very small over a 1-year time frame. In summary, three of the four prospective relationships were significant for women, and those for changes in chronic symptoms were stronger than those for changes in acute symptoms, whereas for men, only one of the four possible relationships was significant.
Self-Appraisals (NA and SAH) and Illness Episodes
If trait NA does nothing more than bias self reports, it should have the same relationship to reports of illness episodes as reports of current symptoms. The cross-sectional results did not confirm the hypothesis that trait NA is a general, biasing factor because it was unrelated to reports of acute illness episodes in both waves of data (Table 3): trait NA was not associated with an over-reporting of acute illness episodes. Because SAH is a judgment of an attribute of the self system and a valid predictor of major health outcomes, we expected it to predict chronic episodes but not acute episodes. The relationship of SAH to acute illness episodes was similar to that of NA: it failed to predict acute episodes in wave 1, but it had a very small, although statistically significant relationship, to acute episodes at wave 2 (1993: R2 = 0.01, F (1,713) = 7.32, p < .01).
The cross-sectional analyses showed that both trait NA and SAH were significantly associated with reports of chronic illness episodes at each wave, although they accounted for modest amounts of variance (Table 3, bottom half). Separate analyses by sex showed that trait NA was related to reports of chronic illness episodes for women at both waves (1992: R2 = 0.04, F (1,474) = 16.99; 1993: R2 = 0.02, F (1,438) = 9.74; both p values <.01), but not for men (R2 values = 0.00, F values <1.46, NS). SAH, on the other hand, was related to reports of chronic illness episodes for both women and men at both waves: for women in 1992, R2 = 0.03, F (1,473) = 17.10; for women in 1993, R2 = 0.02, F (1,438) = 7.22 (both p values for women <.01); for men in 1992, R2 = 0.04, F (1,306) = 12.09; for men in 1993, R2 = 0.03, F (1,272) = 8.84 (both p values for men <.01). In summary, for reports of acute illness episodes, only one of eight possible associations with trait NA and SAH was statistically significant: SAH to acute episodes for men. By contrast, for chronic episodes, 6 of the 8 possible associations with trait NA and/or SAH were significant: all four were significant for women (NA and SAH to chronic episodes in 1992 and 1993), and two were significant for men (SAH to chronic episodes in 1992 and 1993).
Longitudinal analyses were conducted to provide partial insight into the possible directionality of the relationships of trait NA and SAH to illness episodes. The control variables in these analyses (ie, baseline episodes, age, sex, and illness burden) accounted for small amounts of variance in the subsequent number of acute and chronic illness episodes (Table 3). Trait NA (1992), however, failed to predict changes in the number of either acute or chronic illness episodes reported 1 year later (1993, wave 2). The picture was identical for men and women. SAH also failed to predict onset of acute illnesses 1 year later, but it did predict onset and/or flares of chronic illness episodes 1 year later. Separate analyses by sex showed similar effects for men and women, although neither was statistically significant due to smaller sample size. There was no evidence, therefore, of a bias in reporting of illness episodes for trait NA or for SAH. Individuals reported acute illness episodes with equal frequency regardless of their level of trait NA or SAH, and this held in both cross-sectional and longitudinal analyses. Both trait NA and SAH were associated with reports of chronic illness episodes in cross-sectional data, but only SAH predicted chronic illness episodes prospectively when the regression equation controlled for baseline episodes, age, sex, and illness burden. It is important to note that the variance accounted for was small.
Use of Health Care
Wave 1 (1992) and wave 2 (1993) measures of trait NA and SAH were used to predict care seeking for those respondents reporting acute illness episodes at each respective wave (1992 or 1993). As hypothesized, neither trait NA nor SAH predicted use of health care, but the attributes of the episodes did (Table 4). The odds ratios were very large for novelty or unexpectedness (1992: OR, 16.10; 1993: OR, 13.46) and duration (1992: OR, 1.07 for each of the 29 steps in duration; 1993: OR, 1.08 for each step), and the odds ratios for severity were substantial (1992: OR, 1.47 for each of the five steps in severity; 1993: OR, 2.15 for each of the five steps in severity).
A possible corollary of the symptom perception hypothesis is that high levels of trait NA would increase care seeking for individuals with poor self-perceptions of health. Thus, the combination of poor SAH and high trait NA would lead to increased attention to the body, worry about symptoms, and increased seeking of medical care. We explored this idea by examining the relationship of the interaction of trait NA with SAH to care seeking. This interaction did not predict use of medical care in either wave of data in our sample of elderly participants (1992: OR, 0.85, CI, 0.24–3.04; 1993: OR, 0.61, CI, 0.19–1.89).
Two sets of models were tested for care seeking during chronic illness episodes. In one set, trait NA and SAH were measured the same year the episodes were reported (1992 or 1993, respectively), and in the other set, they were measured the previous year (ie, prospective models using 1991 trait NA and SAH to predict care seeking for episodes in 1992 and 1992 trait NA and SAH to predict care seeking for episodes in 1993). It was important to compare these models because trait NA and SAH could be affected by the illness episodes when measured at the same interview (1992 or 1993), that is, the presence of a chronic condition may lead to higher levels of NA and a decrease in SAH when these variables are measured concurrently.
The results of the cross-sectional analyses (trait NA and SAH at the same year as the episode) essentially repeated the effects for acute episodes because the attributes of the episodes showed strong relationships to the use of medical care. Novelty was a strong predictor at both waves (1992: OR, 13.47; 1993: OR, 6.99) as was duration (1992 and 1993: OR, 1.06 for each of 28 steps of duration). Severity was also a substantial predictor (1992: OR, 1.66; 1993: OR, 2.33). Trait NA was not related to the use of medical care at either wave, although poor SAH was related at wave 1 (1992) but not at wave 2 (1993). Models using trait NA assessed a year before these chronic episodes showed that NA was unrelated to care seeking in 1992 (OR, 0.40, CI, 0.14–1.15) but was significantly related in 1993 (OR, 0.26, CI, 0.09–0.79). It is important to note, however, that the relationships were opposite to those predicted by the symptom perception hypothesis, that is, each unit increase in trait NA was related to a decrease in the odds of using medical care. Cross-sectional analyses showed the expected negative relationship of SAH to the use of medical care, which was significant only in wave 1 (1992: OR, 0.51, CI, 0.34–0.77; 1993: OR, 0.89, CI, 0.58–1.36): an increase in SAH was associated with a decrease in the use of medical care. The same pattern appeared using the prior-year ratings of SAH (SAH in 1991 and use of care in 1992: OR, 0.68, CI, 0.47–0.98; SAH in 1992 and use of care in 1993: OR, 0.84, CI, 0.54–1.30). We also examined whether the trait NA by SAH interaction predicted care seeking for chronic episodes. The interaction term was not significant for either wave of data whether the variables were measured at the same or the prior year (1992: OR, 2.22, CI, 0.86–5.69; 1993: OR, 0.54, CI, 0.17–1.71).
Symptom Amplification and Episode-Based Anxiety
Although trait NA was unrelated to the reporting of acute illness episodes and to the use of medical care for these illness episodes and was only marginally related (negatively) to the use of medical care for chronic illness, it is still possible that trait NA is related to the attributes of the episodes, including those that were strong predictors of the use of care. The symptom perception hypothesis would predict that trait NA would be positively associated with the number of symptoms reported during acute and chronic episodes. The data provided little support for this hypothesis because trait NA had no relationship to the average number of symptoms reported during episodes of acute or chronic illness at either wave of data (Table 5). Thus, there was no indication that trait NA results in increased amplification during episodes of acute or chronic illness when the average number of symptoms reported for illness episodes is used as the outcome measure. The result was the same for SAH. In addition, NA was unrelated to reported duration or to reported severity for acute conditions at both waves (r values ranged between .00 and .10, all p values >.22). Results for chronic conditions showed that NA was unrelated to duration but was modestly related to reported severity (r = .13 for both waves, p < .05).
A more elaborate version of the symptom perception hypothesis would predict that increases in emotional distress by high trait NA during illness episodes would lead to increased use of medical care. A more detailed assessment of episode-related variables, including a report of episode-specific worry (ie, anxiety) conducted for the subset of participants who reported an ongoing chronic illness at the time of the 1992 (wave 1) interviews allowed us to partially test this hypothesis. The number of participants reporting an ongoing chronic condition (N = 121) was sufficient to test a seven-variable model predicting the use of medical care. All episode attributes in the model were statistically significant or nearly significant: duration (OR, 11.22, CI, 3.69–34.15), novelty (OR, 2.64, CI, 0.85–8.24, p = .09), and episode-specific anxiety (OR, 27.68, CI, 1.49–511.75). In addition, each of the individual difference factors was significant: an increase in age was associated with a decrease in the odds for the use of health care (OR, 0.92, CI, 0.86–0.99), the odds in favor of using care were larger for men than for women, (OR, 0.31, CI, 0.11–0.91), increases in NA were associated with a decrease in the use of medical care (OR, 0.16, CI, 0.03–1.08, p < .06), and increases in SAH were related to a decrease in the odds for the use of health care (OR, 0.46, CI, 0.24–0.91). Thus, whereas episode-specific worry was positively related to care seeking, generic measures of trait NA were related negatively such that high trait NA was associated with less care seeking. The surprising and opposite relationship of trait NA and episode-specific worry to care seeking is not inconsistent with the data because trait NA and episode-specific worry were only moderately correlated (r = .34, p < .01). Finally, we tested whether the interaction of NA with episode-specific worry was predictive of care seeking for ongoing episodes. These results did not support the more elaborate version of the symptom perception hypothesis as this interaction term was unrelated to the use of medical care for ongoing chronic conditions (OR, 0.07, CI, 0.00–35.19).
Because participants with an ongoing, chronic condition were included in the previously reported analyses of the wave 1 care seeking data, the earlier wave 1 analyses were repeated excluding these respondents. The new analyses were nearly identical to the earlier ones, although the significance levels declined somewhat due to reductions in sample size.
The positive, cross-sectional associations of trait NA with prior-week symptoms reported by our elderly participants replicated prior findings for middle-aged and younger persons (eg, Ref. 10), as did the negative association of SAH with these same symptom reports (19). By allowing our respondents to classify their symptoms as acute or chronic, we also determined that the relationships of both trait NA and SAH are more robust to symptoms regarded as chronic than to those regarded as acute. In addition, the small, but significant, prospective associations to increases in chronic symptoms 1 year later were due primarily to increases in chronic symptoms for our elderly, female participants, an effect that is likely to reflect a bidirectional influence of chronic somatic distress from arthritic conditions. A 1-year time frame was sufficient for a small number of the >700 elderly individuals sampled (mean age of 72.8 years with an average of slightly more than 10 lifetime illnesses) to experience changes in somatic sensations. We have no explanation for the unexpected, positive relationship of trait NA to acute symptoms in the prospective data for male respondents. Given the absence of cross-sectional associations for these variables, the prospective association may well be a random effect.
Although the data replicate the often-reported positive relationship of trait NA and symptoms, they provide little support for expectations derived from the symptom perception hypothesis (ie, that trait NA biases all health reports). Neither trait NA nor SAH was related to reports of acute illness either cross-sectionally or longitudinally. Colds, stomachaches, and other such minutiae are acute, self-limited events that may have little or no implication with respect to one’s overall emotional or health status (19). The absence of significant relationships of trait NA and SAH to reports of episodes of acute illnesses further suggests that the occasional unanticipated prospective relationships of trait NA and SAH to symptoms during acute episodes may reflect chance relationships. By contrast, self-appraisals of emotionality (trait NA) and health (SAH) were correlated with stable chronic conditions that have implications for the elderly person’s self system (9): higher levels of trait NA were associated with reports of more chronic episodes, and excellent SAH ratings were associated with reports of fewer chronic episodes. The relationship of chronic episodes to SAH was somewhat stronger than the relationship to trait NA because trait NA was not related to chronic episodes for men. The prospective analysis yielded a small but significant relationship of SAH to chronic episodes 1 year later. Neither measure, however, was related to the number of symptoms reported during these episodes.
The similar pattern of relationships between these two self-ratings and reports of symptoms and illness episodes suggests that both factors are correlated with reports of somatic status, SAH more so than trait NA. That SAH is the more robust predictor is not surprising because it is a global appraisal of health status (17, 19). By contrast, when participants report the extent to which their “usual” mood is anxious and/or depressed, they are not directly assessing their health status. As Smith et al. (27) noted in their study of disease impact and function in patients with rheumatoid arthritis, “…[R]esearchers who rely on self-reports to study stress, coping, and health ought to be concerned with NA…NA was neither the dominant factor in our respondents’ reports, nor was it able to account for the majority of relations that were observed among the self-report variables we examined.”
It is worth reviewing the ways in which trait NA can be related to elderly participants’ reports of symptoms and illness episodes. First, the physiological processes underlying mood experiences could involve immune suppression and the occurrence of illness; Watson and Pennebaker (10) labeled this the “psychosomatic hypothesis.” Data support this hypothesis for the common cold (28) and chronic conditions such as coronary events (29). The current analyses did not yield similar relationships because trait NA was not related to acute conditions cross-sectionally nor longitudinally and showed unimpressive relationships to chronic conditions longitudinally, although the sample size and limited time frame are less than optimal for the occurrence of a substantial number of new, major stress-related chronic events in an elderly sample. We conclude, therefore, that trait NA fails to assess the affective processes underlying the stress effects for acute and chronic illness reported, respectively, by Cohen et al. (28) and Kamarck and Jennings (29), rather than suggest that these data provide negative evidence with respect to the psychosomatic hypothesis.
Second, the similarity of the positive association of trait NA and symptom reports to that of somatization to symptom reports suggests that both somatization and trait NA amplify somatic cues due to their shared tendency to attend to somatic sensations (30–32). The current data offer little comfort, however, to two variants of the symptom perception and/or amplification hypotheses, namely that trait NA is associated with biases in self reports, and/or that the amplification process involved with trait NA increases symptom reporting because individuals confuse emotional symptoms with those of illness (28, 33). A simple bias hypothesis would predict increased reporting of both illness episodes and illness-linked symptoms, however neither occurred. The version of the hypothesis stating that individuals high in trait NA attribute emotional symptoms to illness would suggest increased reporting of symptoms linked to chronic episodes because such episodes may be emotionally distressing. There was, however, no relationship between trait NA and the number of symptoms reported for chronic episodes. Because self reports of trait NA do not contain questions regarding individuals’ beliefs on symptom attribution, they may not be a sufficient measure to examine this process (cf. Ref. 31). To assess the attributional features of this hypothesis, we examined the joint relationship of trait NA and SAH to episode-linked symptoms, assuming that poor SAH ratings and high trait NA in combination would contain both the cognitive and affective factors needed for amplification. Neither term alone nor their combination was a significant predictor of episode-linked symptoms. Whether a direct measure of somatization would fare better, particularly in prospective analyses, is unclear because measures of this construct may be no more stable than the measures of trait NA and SAH used here (34). Moreover, because studies of community samples report less than 1% of respondents meeting diagnostic criteria for somatization (35), there would be fewer than a dozen “true” somatizers in a community sample of this size.
Our detailed data on care seeking clearly show that specific features of the illness episode do affect care seeking, although these processes are not directly indexed by trait NA. The most powerful predictors of seeking medical care were long duration and unfamiliarity, with illness severity a close third. The dominant role of these attributes emphasizes the psychological rationality of these decisions and is consistent with prior data (36, 37). These features were potent determinants both for episodes that were ongoing and for recent episodes that had already terminated. A separate analysis that excluded the ongoing cases from the larger set of acute and chronic conditions showed precisely the same effects as that for the combined cases. Examination of the ongoing episodes also indicated that episode-specific worry associated with chronic episodes also predicted care seeking, although trait NA did not. Thus, cognitive and affective properties of the illness episode predict coping, and general features of the self system do not (38, 39).
Conceptualizing self-reports such as trait NA and SAH within the framework of the commonsense model of illness (4) clarifies their role as measures of self-knowledge, that is, trait NA indicates whether the respondent experiences him or herself as an anxious person, and SAH indicates whether the individual perceives him or herself as healthy or unhealthy. Not surprisingly, these forms of self-knowledge are associated with knowledge of one’s somatic status. Because the extent of detailed knowledge of somatic conditions will be greater with advances in chronological age and age-related increases in illness experience, we can expect factors accounting for specific forms of somatic attention (eg, trait NA and SAH) to relate to this growth of self-knowledge. Given their lengthy history of psychophysiological activation and extended experience with symptoms (33), elderly individuals reporting high levels of trait NA are likely to have, to be aware of, and to report vague, psychophysiological symptoms and to attribute these symptoms to emotional activation rather than to illness (8, 31). Thus, high scores on trait NA would indicate the presence of psychophysiological symptoms, the ability to detect and report these symptoms, and the ability to distinguish these symptoms from symptoms of illness. By contrast, low levels of trait NA would indicate the absence of and, therefore, failure to detect and report the presence of psychophysiological symptoms and the failure to detect and report symptoms of illness. Thus, trait NA would be associated with reporting of psychophysiological symptoms but not with reporting of symptoms specific to illness or treatment unless the illness or treatment activated the endocrine system underlying psychophysiological activation.
The above conclusions are consistent with existent data. For example, in an early tamoxifen clinical trial, this hormonal agent stimulated endocrine responses leading to greater increases in reporting of tamoxifen symptoms by high trait NA women in the tamoxifen arm of the trial (33). It was clear that high trait NA women distinguished the somatic changes of this intervention from the set of somatic symptoms typically associated with emotional activation because the high trait NA women in the trial reported higher levels of psychophysiological symptoms in both the tamoxifen and placebo arms of the trial. High trait NA women made the distinction more sharply than did low trait NA women because the latter had both fewer psychophysiological and fewer tamoxifen symptoms; additional data suggest that their physiological systems were likely less responsive. A similar pattern likely existed in the study by Cohen et al. (28) where exposure to cold viruses activated a sequence of immune and endocrine reactions leading to a slight increase in symptom reporting for high trait NA individuals. Although high trait NA may lead to reporting of more vague, psychophysiological symptoms on an ongoing basis and to the reporting of more specific symptoms when drug or viral interventions activate the high trait NA individual’s immune and highly responsive hormonal systems, that does not mean that they are always more likely than low NA persons to have and to report illness- or treatment-related symptoms (40). If a treatment impacts with equal strength on high and low trait NA persons during conditions where concern about treatment impact is high for both groups (eg, cancer chemotherapy, which attacks multiple systems in addition to endocrine reactions) no group differences should be expected in the accuracy of symptom reporting (41). The expertise reflected in trait NA among elderly persons may be due to familiarity with one’s emotional reactivity and the ability to distinguish the effects of this reactivity from the effects of physical illness (42).
Conviction in the correctness of these conclusions must be tempered by the properties of this study. First, the psychological reasonableness of the effects likely depends on the age of the respondents and their functional status. Individuals the ages of our participants have years of experience with illness allowing them to differentiate between signs of stress and signs of illness (7, 8) and between signs of illness and signs of aging (43–45). Different effects might be found in samples of respondents who are younger or in older respondents who are seriously ill. Second, our community-based sampling frame was well educated, relatively well-off economically, and white and was not representative of the county, state, or nation at large. This study also likely ignored individuals suffering from severe states of depression and/or anxiety and likely included few if any individuals with high scores on measures of hypochondriasis (35). It remains to be seen whether the same process will appear in other samples. Third, our method of recording illness episodes may overlook medical visits for benign conditions. For example, high trait NA individuals might not report visits to the doctor for an acute or chronic illness if they went for care in response to emotionally upsetting symptoms of uncertain meaning and were given a clear explanation that the symptoms were benign. On the other hand, it is our belief that the detailed reports elicited by our questions offer a more accurate and valid picture of care seeking than do reports asking for estimates of visits in the prior year. Fourth, although several of our variables assess fairly objective factors (eg, sex, age, and medical history) our data consist exclusively of self reports. Objective measures might reveal different relationships. It is important to remember, however, that tests of the implications of the symptom perception hypothesis require the use of self-report.
If our conclusions prove valid, we propose at least four critical directions for research that would build on these analyses. First, it would be worthwhile to identify whether differences in endocrine response to stress are the basis for trait NA self-descriptions. Second, it would be valuable to extend the analysis and to identify the environmental conditions and individual propensities that facilitate the formation of symptom memories (46, 47), the templates for discriminating illness from emotion, and the contribution of individual differences to their formation. Third, it would be interesting to examine the factors that contribute both to the accuracy and the inaccuracy of self-appraisals of health in predicting health outcomes such as mortality and morbidity (17). Fourth, there is much to learn as to how illness, dysfunction, and beliefs about how best to sustain health (ie, self-regulatory processes) affect the formation of affective states and traits (48).
Finally, we should not overlook practice implications. There is substantial risk in attributing symptoms to personality factors, such as affective traits rather than to illness. The risk of doing so may be particularly great for somewhat atypical presentations of life-threatening conditions, such as myocardial infarction in the absence of chest pain, when the presenting individuals fit the complainer or illness amplifier stereotype, that is, when they are female (49) and/or older (50). An additional aspect of the attribution problem is the frequent conflating of symptoms of physical illness with those of depression. The very notion, however, that physical symptoms are conflated with or “mask” depression assumes that patients are suffering from two, distinctly different diseases: physical illness and depression. Looking at symptoms from the commonsense perspective of the patient could suggest a more parsimonious model, ie, that physical illness and the behavioral features of depression are intertwined in a common package, so that successful treatment and/or control of the physical illness and/or of its associated dysfunction will be the quickest and best form of “psychotherapy.” Our data are consistent with the conclusion of the Costa et al. (51) analysis of a national sample of elderly individuals “The present data provide striking support for the contention that aging individuals are not hypochondriacs.”
The present research was supported by Grant AG03501 from the National Institute on Aging. Analyses and write-up were aided by Fulbright and AGCI-CHILE scholarships awarded to P.M. Portions of this study were based on the Master’s thesis of Pablo Mora. We also would like to thank Melissa Crouch, Susan Brownlee, Linda Patrick-Miller, and Frances Sisack for their assistance on various aspects of the research reported in this article.
1 The 1992 interview was selected as the baseline for two reasons: 1) prior-week symptoms were obtained from only half of the respondents in 1991 and 2) 1991 measures of trait NA and SAH were used for prospective analyses of care seeking in 1992. Cited Here...
2 For the analyses predicting medical care seeking we used only one illness episode per individual in each category (ie, acute or chronic). For participants who reported more than one illness episode, we decided to randomly select one illness condition for analytic purposes. The numbers of people who reported more than one illness episode in each wave were as follows: wave 1, acute = 31 and chronic = 63; wave 2, acute = 30 and chronic = 59. Cited Here...
1. Watson D, Clark LA. Negative affectivity: the disposition to experience aversive emotional states. Psychol Bull 1984; 96: 465–90.
2. Idler E, Benyamini Y. Self-rated health and mortality: a review of twenty-seven community studies. J Health Soc Behav 1997; 38: 21–37.
3. Benyamini Y, Idler EL. Community studies reporting association between self-rated health and mortality: additional studies, 1995–1998. Res Aging 1999; 21: 392–401.
4. Leventhal H, Meyer D, Nerenz D. The common sense representation of illness danger. In: Rachman S, editor. Contributions to medical psychology.Vol 2. New York: Pergamon Press; 1980. p. 7–30.
5. Leventhal H, Diefenbach M. The active side of illness cognition. In: Skelton JA, Croyle RT, editors. Mental representation in health and illness. New York: Springer-Verlag; 1991. p. 247–72.
6. Leventhal H, Diefenbach M, Leventhal EA. Illness cognition: using common sense to understand treatment adherence and affect cognition interactions. Special issue: cognitive perspectives in health psychology. Cog Ther Res 1992; 16: 143–63.
7. Cameron L, Leventhal EA, Leventhal H. Symptom representations and affect as determinants of care seeking in a community-dwelling, adult sample population. Health Psychol 1993; 12: 171–9.
8. Cameron L, Leventhal EA, Leventhal H. Seeking medical care in response to symptoms and life stress. Psychosom Med 1995; 57: 37–47.
9. Leventhal H, Idler EL, Leventhal EA. The impact of chronic illness on the self system. In: Contrada RJ, Ashmore RD, editors. Self, social identity, and physical health: interdisciplinary explorations. Rutgers series on self and social identity.Vol 2. New York: Oxford University Press; 1999. p. 185–208.
10. Watson D, Pennebaker JW. Health complaints, stress, and distress: exploring the central role of negative affectivity. Psychol Rev 1989; 96: 234–54.
11. Brenner B. Depressed affect as a cause of associated somatic problems. Psychol Med 1979; 9: 737–46.
12. Costa PT, McCrae RR. Hypochondriasis, neuroticism, and aging: when are somatic complaints unfounded? Am Psychol 1985; 40: 19–28.
13. Gray JA. The neuropsychology of anxiety: an enquiry into the functions of the septo-hippocampal system. New York: Oxford University Press; 1982.
14. Barsky AJ, Klerman GL. Overview: hypochondriasis, bodily complaints, and somatic styles. Am J Psychiatry 1983; 140: 273–83.
15. Kirmayer L. Somatization and the social construction of illness experience. In: McHugh S, Vallis TM, editors. Illness behavior: a multidisciplinary model. New York: Plenum Press; 1986. p. 111–33.
16. Brownlee S, Leventhal H, Leventhal EA. Regulation, self-regulation, and construction of the self in the maintenance of physical health. In: Boekaerts M, Pintrich P, Zeidner M, editors. Handbook of self-regulation. San Diego (CA): Academic Press; 2000. p. 369–415.
17. Benyamini Y, Leventhal EA, Leventhal H. Gender differences in processing information for making self-assessments of health. Psychosom Med 2000; 62: 354–64.
18. Scheier MF, Bridges MW. Person variables and health: personality predispositions and acute psychological states as shared determinants for disease. Psychosom Med 1995; 57: 255–68.
19. Benyamini Y, Leventhal EA, Leventhal H. Self-assessments of health: what do people know that predicts their mortality? Res Aging 1999; 21: 477–500.
20. Costa PT, McCrae RR. The NEO Personality Inventory manual. Odessa (FL): Personality Assessment Resources; 1985.
21. Costa PT, McCrae RR. Personality in adulthood: a six-year longitudinal study of self-reports and spouse ratings on the NEO Personality Inventory. J Pers Soc Psychol 1988; 54: 853–63.
22. Watson D, Clark LA, Tellegen A. Development and validation of brief measures of positive and negative affect: the PANAS scales. J Pers Soc Psychol 1988; 54: 1063–70.
23. Bryk AS, Raudenbush SW. Applications in the study of individual change. In: Bryk AS, Raudenbush SW, editors. Hierarchical linear models: applications and data analysis methods. Newbury Park (CA): Sage Publications; 1992. p. 130–54.
24. Rogosa D, Brandt D, Zimowski M. A growth curve approach to the measurement of change. Psychol Bull 1982; 92: 726–48.
25. Bates B, Hoekelman RA. A guide to physical examination. Philadelphia: Lippincott; 1983.
26. Oman D, Reed D, Ferrara A. Do elderly women have more physical disability than men do? Am J Epidemiol 1999; 150: 834–42.
27. Smith CA, Wallston KA, Dwyer KA. On babies and bathwater: disease impact and negative affectivity in the self-reports of persons with rheumatoid arthritis. Health Psychol 1995; 14: 64–73.
28. Cohen S, Doyle WJ, Skoner DP, Fireman P, Gwaltney JM, Newson JT. State and trait negative affect as predictors of objective and subjective symptoms of respiratory viral infections. J Pers Soc Psychol 1995; 68: 159–69.
29. Kamarck TW, Jennings JR. Biobehavioral factors in sudden cardiac death. Psychol Bull 1991; 109: 42–75.
30. Barsky AJ, Goodson JD, Lane RS, Cleary PD. The amplification of somatic symptoms. Psychosom Med 1988; 50: 510–9.
31. Robbins JM, Kirmayer LJ. Attributions of common somatic symptoms. Psychol Med 1991; 21: 1029–45.
32. Katon W, Russo J. Somatic symptoms and depression. J Fam Pract 1989; 29: 65–9.
33. Cameron LD, Leventhal H, Love RR. Trait anxiety, symptom perceptions, and illness-related responses among women with breast cancer in remission during a tamoxifen clinical trial. Health Psychol 1998; 17: 459–69.
34. Gureje O, Simon GE. The natural history of somatization in primary care. Psychol Med 1999; 29: 669–76.
35. Robins LN, Helzer JE, Weissman MM, Orvaschel H, Gruenberg E, Burke JD, Regier DA. Lifetime prevalence of specific psychiatric disorders in three sites. Arch Gen Psychiatry 1984; 41: 949–58.
36. Stoller EP, Forster LE. The impact of symptom interpretation on physician utilization. J Aging Health 1994; 6: 507–34.
37. Stoller EP, Forster LE, Portugal S. Self-care responses to symptoms by older people: a health diary study of illness behavior. Med Care 1993; 31: 24–42.
38. Coyne JC, Racioppo MW. Never the twain shall meet? Closing the gap between coping research and clinical intervention research. Am Psychol 2000; 55: 655–64.
39. Haug MR, Wykle ML, Namazi KH. Self-care among older adults. Soc Sci Med 1989; 29: 171–83.
40. Diefenbach MA, Leventhal EA, Leventhal H, Patrick-Miller L. Negative affect relates to cross-sectional but not longitudinal symptom reporting: data from elderly adults. Health Psychol 1996; 15: 282–8.
41. Rabin C, Ward S, Leventhal H, Schmitz M. Explaining retrospective reports of symptoms in patients undergoing chemotherapy: anxiety, initial symptom experience, and posttreatment symptoms. Health Psychol 2001; 20: 91–8.
42. Kellner R. Functional somatic symptoms and hypochondriasis: a survey of empirical studies. Arch Gen Psychiatry 1985; 42: 821–33.
43. Prohaska TR, Keller ML, Leventhal EA, Leventhal H. Impact of symptoms and aging attribution on emotions and coping. Health Psychol 1987; 6: 495–514.
44. Leventhal H, Patrick-Miller L, Leventhal EA, Burns EA. Does stress-emotion cause illness in elderly people? In: Schaie KW, Lawton MP, editors. Annual review of gerontology and geriatrics. Focus: emotion and adult development.Vol 17. New York: Springer Publishing; 1998. p. 138–84.
45. Stoller EA. Dynamics and processes of self-care in old age. In: Ory MG, DeFriese GH, editors. Self-care in later life. New York: Springer Publishing Company; 1998. p. 24–61.
46. Leventhal H. Symptom reporting: a focus on process. In: McHugh S, Vallis TM, editors. Illness behavior: a multidisciplinary model. New York: Plenum Press; 1986. p. 219–37.
47. Van den Bergh O, Stegen K, Van de Woestijne KP. Learning to have psychosomatic complaints: conditioning of respiratory behavior and somatic complaints in psychosomatic patients. Psychosom Med 1997; 59: 13–23.
48. Katon W, Sullivan MD. Depression and chronic medical illness. J Clin Psychiatry 1990; 51: 3–11.
49. Martin R, Gordon EEI, Lounsbury P. Gender disparities in the attribution of cardiac-related symptoms: contribution of common sense models of illness. Health Psychol 1998; 17: 346–57.
50. Canto JG, Shjlipak MG, Rogers WH, Malmgren JA, Frederick PD, Lambrew CT, Ornato JP, Barron HV, Kiefe CI. Prevalence, clinical characteristics, and mortality among patients with myocardial infarction presenting without chest pain. JAMA 2000; 283: 3223–9.
51. Costa PT, Zonderman AB, McCrae RR, Cornoni-Huntley J, Locke BZ, Barbano HE. Longitudinal analyses of psychological well-being in a national sample: stability of mean levels. J Gerontol 1987; 42: 50–5.