| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
ORIGINAL ARTICLES |
From the St. John’s University, Jamaica, New York (E.B.); University of Medicine and Dentistry of New Jersey Robert Wood Johnson Medical School (R.C.R., J.B.K.); and State University of New York, Stonybrook, New York (J.E.S.).
Address reprint requests to: Dr. Elizabeth Brondolo, Department of Psychology, St. John’s University, 8000 Utopia Pkwy., Jamaica, NY 11439. Email: brondole{at}stjohns.edu
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMETHODSRESULTSDISCUSSIONACKNOWLEDGMENTSREFERENCES |
|---|
METHODS: Participants included 54 mildly hypertensive men who were participating in a placebo-controlled, double-blind study of the quality-of-life effects of antihypertensive therapies. Survey data and BP measurements were obtained during a series of clinic visits.
RESULTS: Mixed-model analysis of variance was used to evaluate both between- and within-person relations of psychological to physiological state. Results revealed significant within-person associations between predicted and actual BP. Negative mood was closely related to predicted, but not actual, BP. Participants were also relatively accurate in rating active medications as more effective than placebo. Between-persons analyses did not show relations of symptoms or moods to actual BP.
CONCLUSIONS: The significant within-person relations of estimated to actual BP suggest that some individuals may be able to estimate their own BP, although the accuracy of these estimates is limited. The findings may explain patients’ belief that they can self-monitor BP. The results have implications for theories of the mental representation of illness and for efforts to improve compliance with antihypertensive therapy.
Key Words: Blood pressure • symptoms • mixed models • placebo • self-regulation • mood
Abbreviations: BP = blood pressure; SBP = systolic blood pressure; DBP = diastolic blood pressure; ANOVA = analysis of variance; SE = standard error; SD = standard deviation.
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONACKNOWLEDGMENTSREFERENCES |
|---|
Hypertensive patients may be at particular risk for treatment noncompliance because hypertension is commonly believed to be a lanthanic syndrome (ie, one that does not have consistent or readily perceptible symptoms). The presence of acute symptoms is a key trigger motivating individuals to seek help or to adhere to recommended treatments (6). Conversely, the lack of perceptible symptoms may inhibit a patient’s willingness to comply with medical advice, particularly with the life-long medication regimens that are frequently recommended for hypertension treatment. Furthermore, pharmacological therapies are often associated with adverse effects on mood, sleep, and sexual performance (7, 8).
Limited evidence exists that individuals appear to use information about BP symptoms to make decisions about seeking help or adhering to treatment. When patients believe their hypertension is associated with physical symptoms and that the medication they are taking is effective in alleviating these symptoms, they are more compliant with treatment. On the other hand, when patients believe they have symptoms but that their medication is ineffective in reducing symptoms (even if it may actually be effective in reducing BP), they are less compliant (9, 10).
Investigators have used both between- and within-person designs to investigate potential symptoms associated with fluctuations in BP and to investigate the ability to estimate changes in BP. Cross-sectional, between-persons studies assume that the level of BP at which symptoms will be perceptible is constant across individuals and have been used to identify symptoms associated with elevated BP or hypertensive status. These studies have generally failed to find significant associations of specific symptoms with high BP (11–13), reinforcing physicians’ perceptions that hypertension is not accompanied by salient or consistent symptoms.
In contrast, within-person designs have been used to test a different hypothesis. These studies have assessed whether an individual’s symptom reports and BP estimations vary as a function of changes in actual BP, regardless of the level of BP (9, 14–16). Positive findings from these studies seem to support many patients’ beliefs that they can detect fluctuations in their BP and that their judgments about these variations in BP are based on perceptions of symptoms.
Several methods have been used to conduct these within-person evaluations of the relationship of both estimated BP or symptoms to actual BP. Baumann and Leventhal (9) examined the correlations of estimated BP and symptoms to actual BP obtained at a work site over a period of several days. These authors report small (average, r = 0.14) relations of predicted to actual BP, significant correlations of symptoms and moods to predicted SBP, but weaker relations of symptoms to actual SBP.
Pennebaker et al. (14, 16) conducted two laboratory studies in which normotensive participants were presented with a variety of physically and mentally challenging tasks (14, 16). Participants had a limited ability to estimate changes in BP associated with different tasks (16). Significant associations of symptoms to SBP emerged for most participants, although the specific symptoms associated with BP varied among participants (14, 16).
As Stewart and Olbrisch (15) point out, the tasks used in the Pennebaker et al. studies were quite varied, making it difficult to determine whether participants’ estimates of their BP were based on internal changes in symptoms and moods or on situational cues. In a replication and reanalysis of these laboratory studies, Stewart and Olbrisch (15) report that significant symptom–BP relations were found primarily in tasks involving marked physical activation. In this case, participants’ estimates of their own BP could be influenced both by more easily detectable symptoms, such as faster heart rate and shortness of breath, and by the knowledge that exercise can increase BP. Because individuals may use several sources of information when making estimates of their BP (6, 9, 10, 17, 18), interpretation of the accuracy of BP estimates is aided by consideration of both the internal and situational cues available to the patient at the time of BP estimation.
The current study examined symptom–BP relationships within a sample of mildly hypertensive middle-aged men, a group likely to benefit from information about the accuracy of their BP perceptions. Previous studies have included normotensive subjects (14–16), although Baumann and Leventhal (9) also included a limited sample of medicated and unmedicated hypertensive subjects.
Participants were enrolled in a controlled evaluation of the quality-of-life effects associated with drug and nondrug antihypertensive regimens (8); consequently, estimates of BP were made when individuals were both on and off medication. Multiple measures of symptoms, moods, perceived treatment effectiveness, estimated BP, and actual BP were obtained from participants at a series of clinic visits made over a period of several months.
Advances in statistical procedures, specifically mixed-model ANOVA, have improved the ability to evaluate both between- and within-person relations of symptoms and BP (19, 20). When the previous within-person studies were conducted, there were limits to the types of repeated-measures analysis that could be used. Investigators reported correlations computed separately for each participant, enhancing the probability of Type I errors (9, 14–16). These analyses did not permit evaluations of the strength of the relations across participants.
In mixed-model analyses, both between- and within-person effects can be tested in a single analysis, a significant advantage over prior evaluations of symptom–BP relations. Between-persons analyses address questions about the cross-sectional relations of self-reported variables (eg, symptoms or estimated BP) to physiological state (eg, actual BP) across individuals. Multiple measures of both the psychological and physiological variables improve the reliability and sensitivity of the analyses. Longitudinal within-person analyses evaluate the relations of psychological to physiological variables within each participant over time. This makes it possible to determine, for example, whether symptom reports or estimates of BP vary in a consistent manner as the participant’s actual BP fluctuates. Effects are calculated for each participant, and the estimates are pooled across individuals, controlling for multiple observations from each participant.
The first set of analyses in the current study were designed to test hypotheses about the relations of estimated to actual BP and addressed the following question: Can hypertensive patients tell when their BP is relatively elevated? The second set of analyses tested hypotheses concerning the between- and within-person relations of symptoms to predicted and actual BP and addressed two questions: (1) Across participants, do those with higher levels of BP also have higher levels of symptoms in comparison to those with lower levels of BP? (2) Within participants, are variations in symptoms or moods associated with changes in estimated or actual BP? The last set of analyses tested hypotheses about the relations of actual antihypertensive treatment effectiveness to perceived treatment effectiveness and addressed this question: Can patients determine whether their treatment is effective in reducing BP?
These analyses were conducted with and without a series of situational control variables that may influence participants’ estimates of their BP independent of changes in internal state. These variables included the participant’s actual treatment status (ie, whether he was taking medication), perceived treatment status (ie, whether he believed he was taking an active medication or a placebo), and use of home BP monitoring, which might increase accuracy through experience.
|
METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONACKNOWLEDGMENTSREFERENCES |
|---|
Participants took each medication for a 12-week period, followed by a 4-week wash-out phase, at which time use of the next medication (or placebo) was initiated. Before beginning treatment with the first medication, participants who had previously been on antihypertensive medication underwent a 4-week wash-out period.
Procedure
Participants visited the clinic at the beginning and end of each treatment phase (ie, baseline, medication or placebo, and wash-out period). Participants made an average of 7.5 visits to the clinic (range = 2–13). The frequency of visits was unrelated to the participant’s actual BP. For 42 men, data collection began once they had initiated the first phase of active treatment since this study was not part of the original protocol.
At each clinic visit, participants were provided with information about their BP on the previous visit. They were then asked to estimate their current BP (in mm Hg). A registered nurse then obtained two readings of the participant’s current BP using a random zero sphygmomanometer but did not give this information to the participant until he had completed the health questionnaire.
Measures
Actual BP was the mean of the two readings taken by the nurse at the time of the clinic visit. Predicted BP was the participant’s prediction of his own BP, and previous BP was the first of the two readings obtained on the previous clinic visit.
The health questionnaire contained items concerning estimated changes in BP, current symptoms and moods, perceived medication side effects, and other questions related to perceptions of treatment. Specifically, participants were asked to indicate the degree to which they were currently feeling each of 14 physical symptoms and five moods on a five-point Likert-type scale. Five symptoms (ie, headache, fast heart rate, feeling warm, feeling flushed, and shortness of breath) had been identified as associated with perceived or actual BP in other studies (13, 14) and were combined to form a physical symptoms scale with moderate internal consistency (Cronbach’s 
= 0.67). Five mood items (ie, upset, happy, nervous, relaxed, and sad) were combined, with the positive items reverse coded, to form a negative mood scale with moderate internal consistency (Cronbach’s = 0.65). The physical symptoms scale and the negative mood scale were positively correlated (time 1: r = 0.44, p < .01).
The survey also contained a 29-item side effects scale. Participants were asked to rate on a five-point Likert-type scale the degree to which they had experienced any of the listed medication side effects during the current phase of treatment. The items were drawn from lists of side effects known to be associated with medications used in the study. Two subscales were formed from 20 of these items. A physical side effects scale having good internal consistency (Cronbach’s = 0.79), included the following 15 items: "I have difficulty falling asleep"; "I am constipated"; "I have trouble sleeping"; "I feel tired"; "I have headaches"; "I have a very slow heart beat"; "I feel dizziness, faintness"; "I wheeze, can’t catch my breath"; "I have diarrhea"; "I feel nauseated, like vomiting"; "I have stomach pains or cramps"; "I have dry mouth"; "I have a rash"; "I have blurred vision"; and "I sleep deeply." Positive items were reversed coded. A positive benefits scale had good internal consistency ( = 0.83) and included the following five items: "feeling calm"; "I feel happy, less depressed"; "I feel confident about my health"; "I feel alert"; "I feel healthy." The physical side effects and positive benefits scales were negatively related (time 1: r = -0.38, p < .01). The remaining items concerned cognitive side effects and did not cohere sufficiently to form a discrete scale.
Additional items addressed participants’ perceptions of the effectiveness of the current treatment (rated on a five-point scale). During visits that occurred when participants were receiving medication (either placebo or an active drug), they were also asked to indicate whether they believed the pills they were receiving were an active drug or placebo and whether they had monitored their BP at home on at least one occasion since the last visit.
Analytical Plan
To evaluate the degree to which reports of symptoms and moods fluctuated over time, intraclass correlations were computed for the physical symptoms and negative mood scales. The intraclass correlation is the preferred statistic for summarizing stability over time and is an estimate of the degree to which participants’ responses remained constant across observation periods (21). The degree of stability across the full study period can be considered an estimate of the degree to which the scale taps a stable traitlike characteristic. In contrast, within-person variation equals the difference between the squared intraclass correlation (ie, the percentage of variance associated with between-persons variation) and 1.00 and may reflect both random reporting error and situational variation, including treatment effects. To control for the possibility that within-person variation in psychological state reflected changes in actual treatment condition (ie, baseline, placebo, propranolol, and clonidine), treatment was included as a covariate in these analyses.
The remainder of the analyses concerned the relations between psychological (ie, estimated BP, symptoms, mood, estimates of medication effectiveness, and estimates of medication status) and objective (ie, actual BP and actual medication status) measures. Advances in statistical procedures, specifically multilevel mixed-model analyses of variance for unbalanced designs, have made it possible to examine relations among BP and symptoms obtained in field studies such as this (20). PROC Mixed, a recently developed statistical procedure from the SAS Institute, was used in these analyses (19). PROC Mixed is a generalization of the standard linear model designed to estimate and test the significance of between- and within-person effects, controlling for factors associated with repeated observations from the same subject. These effects can be evaluated even when the between- and within-person variances are not orthogonal, as occurs when there are different numbers of observations for each subject or for each level of the independent variables.
As suggested by Schwartz and Stone (20), it is possible to separate within-person effects from between-persons effects by including both the observation-level score for each time-varying independent measure and the participants’ mean score for the measure. This type of control is valuable, because accuracy in symptom detection may be influenced by the actual level of symptoms. Individuals who experience greater changes in BP (and potentially higher levels of BP overall) may be more accurate in detecting BP and BP-related symptoms (22). As an example, in analyses estimating the relationship of physical symptoms to predicted BP, the predictors include the participant’s observation-level physical symptom scale score and his average symptom score. The coefficient associated with the observation-level variable is an estimate of the degree to which, within individuals, fluctuations in symptom scores correspond to changes in predicted BP. In contrast, the estimate associated with the average symptom score reflects the degree to which the actual average level of symptoms covaries with average level of predicted BP across individuals.
PROC Mixed analyses are appropriate even when there are different numbers of observations for each participant, because the estimates reflect the actual number of visits made. This is an important feature, because the men differed in the number of times they completed the health questionnaire. The 54 men completed the questionnaire an average of 7.5 times (range, 2–13). Observations were collected when the men were both medicated and not medicated, with 52% of the observations obtained when the men were not receiving an active drug. In some analyses, the actual number of observations may be slightly less due to occasional missing data.
In comparison to repeated-measures analyses conducted using ANOVA or multiple ANOVA modeling, mixed models offer a more efficient and potentially more powerful strategy for significance testing (19, 20). In repeated-measures ANOVA tests, if the actual error structure differs from the assumption of compound symmetry, the significance levels should be adjusted (ie, with Huynh-Feldt or Greenhouse-Geiser corrections to the degrees of freedom). In a PROC Mixed analysis, the investigator can compare several different error structures and then choose the error structure that is most appropriate for the data. Strategies for choosing among variance models have been articulated by Wolfinger (23), Wolfinger and Chang (24), and Bagiella et al. (unpublished). The F statistic for the equation changes depending on the error structure chosen, which should obviate the need for further corrections based on deviations from assumed error structures.
In this study, preliminary analyses were performed using three error structures: compound symmetry, first-order autoregression, and a combination of the two. Compound symmetry is the basic assumption of repeated-measures ANOVA and assumes correlated residuals due to unexplained between-persons variation. Serial autocorrelation, AR1, is another type of error structure hypothesized to be relevant to this type of data, because measurements obtained closer in time may be more highly correlated than those obtained at more distal points in time. The combination of the two types of error structure is the most general error structure of the three cases and allows for both unexplained between-persons variation and serial autocorrelation. Comparisons among the models were made using Akaike’s information criterion and Schwarz’s bayesian criterion. Because the estimated serial autocorrelation parameter was not statistically significant and the compound symmetry error structure produced the best fit in each case, the compound symmetry error structure was used in all analyses, and the statistical results reported below used this error structure. All estimates reported are unstandardized to facilitate interpretation of the magnitude of changes in BP or symptoms.
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONACKNOWLEDGMENTSREFERENCES |
|---|
Relations Between Estimated and Actual BP
The first set of analyses tested the hypothesis that within-person variations in actual BP would be associated with variations in the participant’s estimates of BP. The participant’s BP estimates served as the outcome variable. The predictor variables included both observation-level scores for actual BP and the participant’s average BP scores across observations. Analyses were performed separately for SBP and DBP. Results revealed a significant within-person association of actual to predicted SBP (B = 0.18, SE of B = 0.06, df = 303, t = 3.14, p < .002) and a significant within-person association of actual to predicted DBP (B = 0.17, SE of B = 0.07, df = 303, t = 2.5, p < .02). There were significant between-persons effects, but these indicate only that, as expected, individuals with higher average actual BP levels predicted that their BP would be higher than those with lower actual BP levels.
To facilitate an understanding of the size of the estimates for the relationship of actual to estimated BP, consider that a B value of 1.0 would indicate that, on average, for every 1-mm Hg increase in actual BP, the average estimated BP would increase by 1 mm Hg as well. The size of the obtained estimates (between 0.17 and 0.18) suggests that the observed relationship was small.
Across all observations, the mean difference between actual and estimated SBP was -0.73 mm Hg (SD = 13.96, range = -55–44 mm Hg). For 34% of the observations, the difference between actual and predicted SBP was 
5 mm Hg.
To compare the size of the effects obtained in this study with that obtained by other investigators (9, 14–16), we computed an average correlation of actual to estimated SBP and obtained a correlation of 0.17. This is quite similar to the average correlation of 0.14 found in both laboratory studies of the relationship of symptoms to SBP (14, 15) and the Baumann and Leventhal analysis of the relationship of estimated to actual BP (9).
These analyses of estimated BP were repeated with a series of control variables, each of which may have influenced participants’ estimates of BP independent of actual BP. These variables included the participant’s previous BP; treatment status (ie, whether the participant was receiving medication vs. abstaining from medication); if taking medication, perceived medication status (ie, the patient’s belief that he was receiving placebo vs. an active drug); and use of home BP monitoring equipment. Previous BP was closely related to both predicted and actual BP, and participants might rely solely on knowledge of previous BP when estimating current BP. Participants might use knowledge of actual treatment status or perceived medication status to make inferences about their current BP. For example, a participant might assume (correctly) that his BP was higher if he was not receiving medication or if he believed he was receiving a placebo. Home monitoring of BP might provide participants with more experience in estimating their own BP, although the association between home monitoring and predicted or actual SBP or DBP was not significant (p > .15). Controlling for these four variables, the within-person effect of actual SBP on predicted SBP remained significant (B = 0.11, SE of B = 0.05, df = 298, t = 2.33, p < .03). The association of within-person variation in actual DBP to variation in predicted DBP was no longer significant (p < .13).
To determine whether there were significant individual differences in the ability to estimate BP, the analysis of the association of actual to estimated BP was repeated, with observation-level scores of actual BP treated as a random effect. In this random effects analysis, the slopes of the relationship of actual to estimated BP are implicitly calculated separately for each individual. An examination of the estimates associated with the variance parameter for the random effect revealed significant variability (Z = 2.92, p < .01), suggesting that there were significant individual differences in the association of actual to estimated SBP. The square root of this variance is the estimated SD (among participants) of the slope when predicting estimated BP from actual BP. With a mean slope of 0.18 and an SD of 0.09, the range of estimates (±2 SD) of the relationship of estimated to actual SBP extends from 0 to 0.36. This indicates that some individuals show no consistent relationship between estimated and actual SBP, whereas others show a much closer relationship, about twice the average.
To evaluate possible predictors of individual differences in the relationship of estimated to actual BP, we examined the age of the participants, level of BP (ie, above or below the median for the group), frequency of visits to the clinic (ie, above or below the median number of visits), use of home BP monitoring, and type of treatment. Each variable was entered into equations examining the relationship of estimated to actual BP as a main effect and as an interaction term with actual SBP (or DBP). Actual BP was treated as a random effect. None of the interaction terms was significant, indicating that these variables did not account for the individual differences in the ability to estimate BP.
In summary, these results demonstrated that participants were able to perceive and quantify changes in actual SBP at a level better than chance, but the relationships were small. Overall, participants tended to underestimate changes in BP. The association of actual to predicted SBP persisted despite controlling for a host of cognitive and situational factors that may have allowed participants to make inferences about their BP. There were significant individual differences in the ability to estimate BP.
Symptom and Mood Correlates of Predicted and Actual BP
The next set of analyses examined the degree to which symptoms and mood states were associated with estimated and actual BP. To test these effects, PROC Mixed analyses were used, with estimated or actual BP serving as the outcome criterion variable. In these analyses, individuals’ observation-level and average scores on the negative mood and physical symptoms scales served as independent variables. Covariates included treatment status, beliefs about medication status, and use of home BP monitoring.
The within-person effects of negative mood were significant for both estimated SBP (B = 3.72, SE of B = 1.49, df = 297, t = 2.5, p < .02) and estimated DBP (B = 2.45, SE of B = 1.15, df = 297, t = 2.13, p < .04). As negative mood increased, so did the level of estimated BP. The estimated SBP associated with the highest level of negative mood was about 15 mm Hg higher than that associated with the lowest level of negative mood. The within-person effects for physical symptoms did not reach statistical significance. Between-persons analyses indicated that the average levels of symptoms and moods were not associated with estimated BP. When these analyses were repeated with actual BP serving as the criterion variable, neither the symptom nor the mood scales were associated with actual SBP or DBP.
Treatment Effectiveness
Compliance with antihypertensive treatment may also depend on the patient’s perceptions of the treatment’s effectiveness (rated on a five-point scale). In this sample, both propranolol and clonidine were effective treatments, associated with significantly lower SBP and DBP in comparison to placebo. To evaluate perceived effectiveness, this analysis included treatment (ie, propranolol, clonidine, or placebo) as the predictor variable and perceived effectiveness as the criterion variable. Beliefs about medication status and the use of home monitoring of BP served as covariates. These analyses included only observations obtained when participants were taking some form of treatment and excluded baseline or wash-out readings. There was a small but significant main effect of treatment on ratings of effectiveness (F(2,258) = 3.98, p < .02). When participants were receiving placebo (adjusted mean = 2.64), their ratings of effectiveness were significantly lower than those obtained when they were receiving either clonidine (B = 0.31, SE of B = 0.14, t = 2.25, p < .05, adjusted mean = 2.95) or propranolol (B = 0.25, SE of B = 0.10, t = 2.40, p < .05, adjusted mean = 2.88).
To evaluate the contributions of mood, symptoms, and side effects to the perceptions of medication effectiveness, a PROC Mixed analysis was performed with effectiveness as the criterion variable; observation-level and average scores for the negative mood and physical symptoms scales, as well as the two side effects scales (physical side effects and positive benefits), served as predictor variables. Covariates included home monitoring and beliefs about medication status. Both within-person measures of physical symptoms (B = -0.53, SE of B = 0.19, df = 252, t = -2.77, p < .01) and positive benefits (B = 0.28, SE of B = 0.11, df = 252, t = 2.62, p < .01) were significantly associated with rating of effectiveness. Relatively lower levels of physical symptoms and relatively higher levels of well-being were associated with higher ratings of medication effectiveness. The within-person effects of negative mood and physical side effects as well as the between-persons effects of the participants’ average levels on each of the four scales were not significant.
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONACKNOWLEDGMENTSREFERENCES |
|---|
Patients often claim that they can detect fluctuations in their BP, despite information from physicians and the media to the contrary. Results of withinperson analyses provide limited support of this common belief. The findings indicate that, given some information about their previous BP, participants displayed a limited but reliable relationship between estimated and actual SBP. Estimated and actual SBP were related, even after controlling for a variety of cognitive and situational factors that may have influenced judgments about BP. There were significant individual differences in the ability to estimate BP fluctuations, but we were unable to identify predictors of these differences.
Participants were also somewhat accurate in evaluating the effectiveness of their medication in reducing or controlling their BP (ie, they gave the treatment higher ratings of effectiveness when they were receiving an active drug vs. placebo). Judgments about the medication’s effectiveness were associated with perceived physical symptoms and a sense of well-being that patients attributed to the medication. They did not appear to be associated with negative side effects of the medications despite the salient side effect profiles associated with both clonidine and propranolol.
Although the results lend limited credence to patients’ claims about their ability to detect changes in BP, the findings also support physicians’ concerns about patients’ inability to independently monitor their own BP. Although some participants made reasonable estimates of the direction of change in BP, they were not accurate about the magnitude of the difference and tended to underreport the degree of change. About one-third of the participants made estimates of their SBP that were within 5 mm Hg of their actual SBP. However, others made substantial over- and underestimations, and the strength of the relationship is small. The results also confirm reports that hypertension has no consistent set of symptoms, because between-persons analyses of the relationship of symptoms and moods to actual BP were not significant.
Others have shown relationships of negative mood to BP (26), but in this sample, within-person analyses revealed that mood was related only to predicted, not actual, BP. This suggests that participants may use their hypotheses about the relation of mood to BP to justify their estimates of BP (14, 27). Different situational variables and participants’ hypotheses about the relationships of these situations may also have influenced BP estimation. Some of the situational variables were controlled in this study, because BP estimations were made under the same clinical conditions each visit. However, we did not ask participants to describe the activities they were engaged in before the visit. It is reasonable to suppose that these activities, and the participant’s hypotheses about the health effects of these activities, could affect their BP estimation. Future research should consider a broader range of situational variables.
We may have underestimated the relationship of symptoms and moods to actual BP because we did not inquire about the correct symptoms. However, this seems unlikely because relevant symptoms were assessed, and the expected relations between predicted BP and symptoms emerged. Alternatively, there may be little consistency across subjects in the type of symptoms associated with BP fluctuations (ie, headaches may vary with BP for one subject, and shortness of breath may vary with BP for another). It may be more productive in future work to ascertain which particular symptoms the participant believes are likely to fluctuate with BP and then determine whether there are reliable associations between fluctuations in these symptoms and changes in BP.
All participants in this study were concerned about BP and health and were sufficiently motivated to participate in a complex treatment protocol. These selected individuals may have been unusually accurate about or observant of their own condition. Many were engaged in home monitoring of BP, although home monitoring itself was not significantly related to improved BP estimation. Future research needs to replicate these findings in a more diverse sample. It may also be useful to investigate predictors of individual differences in the ability to make accurate judgments about BP or to identify symptoms or moods that reliably covary with BP.
There are clinical implications of the present findings. Overall, the data suggest that some patients have some ability to detect changes in their BP and to evaluate the effectiveness of their antihypertensive treatment. Although the association of estimated to actual BP is small, it may be sufficient in some cases to convince patients that they are reliable monitors of their own BP status.
Therefore, it may not be completely accurate or helpful to tell these patients that they cannot detect changes in BP or perceive symptoms associated with hypertension. Instead, it may improve physician–patient communication, if the physician discusses the patient’s beliefs about his or her ability to detect changes in BP. These discussions could offer an opportunity to acknowledge the patients’ perceptions and to show the patient that these subjective perceptions are not necessarily accurate in magnitude or direction and need to be validated by objective measures (eg, regular BP monitoring). This may help patients think more carefully about the types of information they use to guide treatment decisions.
Future research may shed light on the interactions between patient’s cognitive representations of their illnesses and their awareness of and response to specific symptoms. In turn, further understanding of patients’ theories about their health can guide the development of programs to improve patient compliance with antihypertensive regimens.
|
ACKNOWLEDGMENTS |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONACKNOWLEDGMENTSREFERENCES |
|---|
Received for publication October 23, 1997.
Revision received December 11, 1998.
|
REFERENCES |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONACKNOWLEDGMENTSREFERENCES |
|---|
This article has been cited by other articles:
 |
|
 |
|
  E. Brondolo, R. Rieppi, S. A. Erickson, E. Bagiella, P. A. Shapiro, P. McKinley, and R. P. Sloan Hostility, Interpersonal Interactions, and Ambulatory Blood Pressure Psychosom Med, November 1, 2003; 65(6): 1003 - 1011. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
W. A. Karlin, E. Brondolo, and J. Schwartz Workplace Social Support and Ambulatory Cardiovascular Activity in New York City Traffic Agents Psychosom Med, March 1, 2003; 65(2): 167 - 176. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
