Abstract
We investigated whether self-ratings of health are affected by a symptom rating. A diary including a one-item self-rating of health (“pre–self-rated health”; 1 = excellent, 7 = very poor), a subsequent 26-item rating of symptoms, and thereafter a second (identical) health rating (“post–self-rated health”) was completed by 820 persons 21 times. Self-rated health worsened significantly (p < .0001) after the symptom rating, from 2.72 pre–self-rated health (95% confidence interval: 2.70–2.74) to 2.77 post–self-rated health (95% confidence interval: 2.75–2.79) and more so in persons who reported more symptoms (b = .058, p < .05). The results support the notion that subjective health perception is influenced by attending to symptoms, especially so in persons with a high symptom burden.
Keywords
Introduction
Bodily signals and how these are interpreted are believed to influence self-ratings of health (Jylhä, 2009). It is thus reasonable that appraisals of health are affected by imminent exposures and disease primes and that subjective health perception is affected by focus of attention. This study investigated this by comparing self-rated health (SRH) provided both after and before a number of symptom ratings.
Self-reports rely on a cognitive process where respondents have to interpret the question, retrieve information, make a judgment, and communicate the response (Tourangeau et al., 2000). This process is sensitive to contextual factors, where wording, format, and order effects have all been suggested to be influential for the responses in surveys (Schwarz, 2007; Tourangeau et al., 2000). For example, how anchor points are defined in questionnaires affect individuals’ ratings (Schwarz, 2007), seemingly giving clues to the respondent in which range a response is reasonable or expected. Similarly, responses can be sensitive to item order (Dahlstrom et al., 1990; Kieffer and Hoogstraten, 2008; Schwarz, 2007; Schwarz et al., 1991). Strack et al. (1988) showed a so-called part-whole assimilation effect as a function of the order in which questions on life satisfaction were asked. Students who reported happiness with dating and thereafter were asked to rate happiness with life-as-whole showed more positive ratings compared to if the questions were asked in the opposite order. From a cognitive perspective, the search process starts by an individual when asked to perform a rating and is truncated when a sufficient amount of information is retrieved, so that the most salient information influences the outcome (Schwarz et al., 1991). Through availability heuristics, the information presented just before an individual performs a judgment affects the responses (Tversky and Kahneman, 1974). The prototypical situation activated in working memory to represent a general condition is thus likely to be affected by contextual factors, as in the case of the students rating life satisfaction in the study by Strack et al. (1988).
Whereas research on survey responses regarding opinions and attitudes is well established (Schwarz, 2007; Tourangeau et al., 2000), studies specifically focusing on the influence of contextual factors in health ratings is relatively sparse. In one study, item order of the Oral Health Impact Profile was shown to influence responses and the scale’s factor structure, depending on whether more general or more specific questions were asked first (Kieffer and Hoogstraten, 2008). If ratings of behaviors and attitudes depend on contextual factors and estimation processes, this opens up for the possibility that also judgments of overall health are influenced by situational factors. Given the fact that health ratings are top predictors of mortality, and that patient-reported outcomes are generally better in this respect compared to lab-based and objective measures (Ganna and Ingelsson, 2015), the question emerges whether even such valid and seemingly robust ratings are influenced by context. Whether mere attention to symptoms influence an individual’s subjective health perception is not known. In this study, we investigated if appraisal of a range of mental and somatic symptoms, made just after an initial health rating, influenced a subsequent identical health rating. The hypothesis was that health would be rated as less favorable in the measurement following the symptom rating. In addition, we hypothesized that the more symptoms an individual would report, the stronger the effect on the second health rating would be.
Method
Participants and procedure
Included in the study are data from 820 employees working in the public sector in Sweden. The age range was 20–65 years (M = 44.39, SD = 10.6) and 79 percent were women. Four working areas were represented where 25 percent worked within Social services, 32 percent in Technical services, 33 percent in Care and welfare, and 10 percent worked within Telecall. A majority of the participants worked full time (78%). Data were collected within a longitudinal study aiming to evaluate a work time reduction of 25 percent as previously described (Schiller et al., 2017). The design included three periods of data collection: baseline, 9-month follow-up, and 18-month follow-up. The outcome variables were measured with a paper diary during 1 week per data collection period. The intervention was implemented in 2005–2006 by the Swedish National Working Life Institute and commissioned by the government (Bildt, 2007) and was approved by the regional ethics board in Stockholm (Dnr 04-1059/5). The study was performed in concordance with the 1964 Helsinki declaration and its later amendments.
The participants completed the diary in the evening, daily for three full weeks, 1 week per data collection period. The diary contained questions on work hours, work load, sleepiness, stress, daily functioning, and health. The health section of the diary included a self-rating of health (“pre-SRH”), followed by a 26-item rating of a range of somatic and mental symptoms and subsequently a second (identical) self-rating of health (“post-SRH”).
SRH
SRH was assessed using the following question: “How would you rate your state of health for the day?” answered by “excellent”(1), “good” (2), “rather good” (3), “neither good nor poor” (4), “rather poor” (5), “poor” (6), and “very poor”(7).
Symptoms score
The symptom score rating adapted from the Symptom Checklist-90 included 26 symptoms (Have you during the day felt or had: fatigue, sleepiness, agitated, full of energy, ability to rewind, tense, difficulties concentrating, nervous, psychological exhaustion, physical exhaustion, headache/migraine, dizziness/malaise, upset stomach/acid regurgitation, flu or cold symptoms, fever, asthma, allergic symptoms, dry skin/mucous membranes, eczema/rash, tensions or stiffness in shoulders or back, tender joints, tender muscles, anxious, depressed/blue, worn out) rated on a 5-graded Likert scale (1 = not at all to 5 = to a great extent). The reliability of the scale (Cronbach’s alpha) between days was .94. The summary score was divided by 26 to produce the average symptom rating (range, 1–5).
Statistical analyses
All analyses were performed as mixed effect regression analysis, a method that allows taking into account data from the 21 repeated diary entries by including the participant identity as a random effect. In the first model, the difference between Pre-SRH and Post-SRH was analyzed by testing the effect of repeated ratings on SRH (30428 SRH ratings in 820 participants) in a model that used SRH as outcome and time (referring to the first and the second SRH-rating coded 0 and 1, respectively) as explanatory variable. A second model included time, symptom score, and interaction between time and symptom score and was adjusted for age and gender. This was done in order to test if the reported symptom score would influence the change in SRH (27,584 observations, 812 participants). All statistical analyses were performed in STATA 14 (StataCorp, Texas, USA) and a two-sided alpha level of .05 was used to test for statistical significance.
Results
SRH worsened slightly but significantly between the first and the second rating, from 2.72 pre-SRH to 2.77 post-SRH (Table 1), corresponding to a change with 0.05 steps on the SRH scale (b = 0.051; 95% confidence interval (CI): .029 to .074, p < .001, Cohen’s D = .039).
Characteristics of participants.
SRH: self-rated health.
Sample demographics, symptom score, and self-rated health presented in three symptom categories and overall.
Continuous variables represented with mean (95% confidence interval). Categorical variables represented with % (95% confidence interval).
Self-rated health (1, excellent; 7, very poor). Symptom score: 1 = not at all; 5 = to a great extent.
The symptom score was associated with SRH, independent of the effect of repeated ratings, so that a higher symptom score was significantly associated with poorer health (b = 2.29; 95% CI: 2.21 to 2.36, p < .001). There was no significant effect of time that was independent of symptoms (b = .040; 95% CI: −.112 to .032); hence, the repeated rating in itself did not affect SRH. Indeed, the change between pre-SRH and post-SRH was accounted for by the symptom score; the higher the symptom score, the larger the reduction in SRH corresponding to an additional worsening of SRH with 0.06 points per unit symptom score (interaction between time and symptom score, b = 0.058; 95% CI: .011 to .104, p < .05). There were no significant associations between gender (p = .93) or age (p = .06) and SRH.
Discussion
In this study, the aim was to investigate if the rating of a range of mental and somatic symptoms made just after a health rating influenced a subsequent identical health rating. In accordance with our hypotheses, SRH was rated as less favorable in the measurement following the symptom rating, although the effect size was small. Of note, the more symptoms the individual reported, the stronger was the effect on the second SRH rating. The repeated rating in itself, independent of symptom score, did not affect SRH.
Schwarz et al. (1991) concluded that a subsequent general question after several specific questions result in an assimilation effect, while a contrast effect was seen if only one specific question was asked before the general question. The result of this study is in line with their findings, as the symptom rating consisting of a number of specific questions was taken into account in the subsequent health rating (assimilation effect).
It has previously been hypothesized that the interoceptive pathway, providing information of bodily symptoms and signs (Craig, 2003), may be tonically activated in certain conditions, resulting in a change in subjective health perception (Andréasson et al., 2007). Although the effect of symptom ratings in this group of relatively healthy working people is small, the results suggest that attention to symptoms might inadvertently worsen the subjective health status of the individual, an effect which may be clinically relevant in patient populations. Based on the results from this study, we hypothesize that interventions breaking off an interoceptive body-oriented focus might be useful in patients with poor SRH where no objective indices of poor health are detected. This is similar to what is performed in Cognitive Behavioral Therapy for severe health anxiety (cf. hypochondriasis), where reduced health-related safety behaviors such as symptom checking or seeing doctors lead to decreased health anxiety (Hedman et al., 2014). Correspondingly, healthy subjects put on a regime to perform health safety behaviors increase disgust sensitivity, a trait measure of disgust related to health anxiety (Olatunji et al., 2014) believed to increase body vigilance and perception of bodily symptoms. On a related note, it has been shown that even false information on symptoms affect how symptoms are reported. Thus, non-detected misinformation that two target symptoms related to emotional distress in the Symptom Checklist-90 were rated relatively high lead participants to revise their ratings in the direction of misinformation at a 1-week follow-up test (Merckelbach et al., 2011).
The repeated use of the SRH question might have caused a bias in the answers to the second SRH question. The fact that each person responded to the diary 21 times in total (over a period of 18 months) should have cancelled out this effect. This notion is supported by the lack of an overall effect of repeated ratings when symptoms were included in the mixed effect regression model. Nevertheless, future studies should randomize participants to perform health estimations either after or before rating symptoms in order to reduce bias from repeated ratings. In addition, future studies should examine the effect on symptom ratings on general SRH, as this study used a state version of SRH referring to the present day rather than to health in general.
Presentation order and other contextual factors are likely to influence overall ratings of health. Increased knowledge of such factors can be used to increase precision in subjective health measurements and thereby further improve predictive capacity. Given the importance of self-ratings of global health as predictor of adverse health outcomes (Ganna and Ingelsson, 2015), and the movement to include patient-reported outcomes to improve health care processes (Basch et al., 2013; World Health Organization (WHO), 2007), it is suggested that SRH assessment should be performed first in surveys. In common instruments such as the Short Form (36) Health Survey (Sullivan et al., 1995), the question on general health is indeed presented first, preceding questions regarding physical and mental health symptoms.
In conclusion, prompting individuals to consider a range of somatic and mental symptoms negatively affects self-reported health. The results support the notion that subjective health perception is affected by attention (Andréasson et al., 2007), and this may be especially true in persons with a high symptom burden.
Footnotes
Declaration of Conflicting Interests
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The author(s) received no financial support for the research, authorship, and/or publication of this article.
