1 Introduction

What makes a good life? Many things contribute to how we answer this question. Whereas some influences possess rather individual meaning for a specific individual, others have general importance for most if not all people. Health commonly is considered to be one such crucial component of a good life. Yet, health issues do not necessarily diminish life quality. Asked about general satisfaction with one’s life, chronic diseases or disabilities often are not related to lower satisfaction [1, 2], whereas asking specifically about satisfaction with the current health status typically produces a different outcome [3, 4]. Obviously, responses depend on what is asked about precisely. In addition, responses depend on how questions are posed.

Among other threats for the validity of questionnaires (for example, social desirability or response tendencies), effects of question order prove the importance of careful survey design [5]. An assimilation effect occurs when responding to a specific question influences the response to a subsequent question to become closer to the response to that first question, compared to when questions are not responded in succession but separately. For example, it has been shown that estimations of global life satisfaction were higher after participants had rated their marital satisfaction [6]. Apparently, this assimilation effect [7] reflected that overall higher ratings for marital satisfaction were an anchor for rating global life satisfaction. A contrast effect is the opposite of an assimilation effect: A specific question influences the response to a subsequent question to become more distant to the response to that first question. In particular, assimilation and contrast effects both can occur when subsequent questions ask for specific and global ratings of a certain topic (e.g. life satisfaction in general or regarding one certain domain). An assimilation effect consists in a decreasing difference between a global and a specific rating when the specific question is first, whereas a contrast effect consists in an increasing difference between a global and specific rating. The conversational-logic account by Schwarz et al. [8] assumes assimilation to occur when respondents assume a global rating to ask for summarization of specific aspects, including those aspects asked about in previous questions. On the other hand, a contrast effect occurs when respondents assume they should exclude previously mentioned specific aspects from a global rating, in accordance with conversational norms of non-redundancy. Thus, contrast effects depend on the understanding that a global question refers to something else than preceding specific questions. However, when a global question following questions about specific aspects of one topic is understood as a summarization of all these topics, assimilation occurs. In correspondence with these assumptions, an assimilation effect was found for a global measure of self-reported health (SRH) when several domain-specific health items preceded it [9].

In addition, asking about membership to social groups might influence responses, because indicating whether you belong to a certain group can activate knowledge and beliefs about your identity and role as a group member. For example, a study asking university students with a disability questions about their life as a student primed more autonomy-related responses in a subsequent writing task as compared to asking disability-related questions [10]. The mere indication of gender or race can even entail such extensive consequences as diminished performance in cognitively demanding tasks when stereotypes about lower ability of that gender or race exist [11, 12]. Such findings may imply that responses to questions about well-being or satisfaction also depend on asking about a certain diagnosis that can constitute a social group as well. Priming identity as a member of that group might entail comparing oneself to other members of the group. Beliefs about what is typical for members of the respective group then would be used as a reference when estimating, for example, the own health status. However, if group membership is not primed, the reference might be a different group or the general population.

Examining quality of life, measures can address different life domains. According to the definition by the World Health Organization, quality of life concerns a person’s perception of their own life. In addition to such domains as freedom, education, social belonging or wealth, this also includes mental and physical health, which is the prime focus when health-related quality of life (HRQOL) is regarded. Measures of HRQOL typically assess how strongly current health issues affect one’s life, although measures vary with regard to how many and which life domains are regarded. For example, the Pediatric Quality of Life Inventory asks about physical and mental problems, and how these affect social relations and school [13]. In comparison, the KIDSCREEN assesses psychological well-being, physical well-being, financial resources, autonomy, moods and emotions, self perception, parent relation and home life, peers and social support, school environment, as well as bullying [14]. Despite differences in the number of items and scales, the common theme of these and other HRQOL questionnaires is to ask how good one’s health is currently and some consequences of that health status.

Here, we investigated question-order effects on ratings of global SRH and more specific ratings of HRQOL, as well as a life-satisfaction scale in a sample of participants with diagnosed spina bifida and a comparison group of participants without spina bifida. We manipulated the position of the global SRH item and the position of an item asking about a diagnosis of spina bifida. We expected an assimilation effect for global SRH, that is, higher correspondence with the HRQOL scale when this scale preceded the global SRH item. In addition, we expected a contrast effect regarding the position of the item asking about spina bifida to influence ratings of those participants with spina bifida on the HRQOL scale. In particular, we expected participants with spina bifida to rate their HRQOL higher when the item asking about spina bifida preceded it, assuming that health issues related to this diagnosis would rather be excluded from HRQOL.

2 Methods

2.1 Participants

The sample comprised 244 participants, of whom 68 indicated to have a diagnosis of spina bifida. This sample size ensured to be able to detect small effect sizes of question order for the whole sample and medium effect sizes for the spina-bifida sub-sample. Participants were contacted via email distribution lists of German universities and spina-bifida advocacy groups. All participants provided informed consent prior to participation.

2.2 Design

Two factors of question order were manipulated between participants. One factor concerned the position of global SRH, the other factor concerned the position of asking about a diagnosis of spina bifida (before vs. after rating life satisfaction and HRQOL). The two groups (spina-bifida group, comparison group) constituted a third factor.

2.3 Material

Participants indicated whether they had a diagnosis of spina bifida or not (forced choice) and if applicable, which kind of spina bifida. For global SRH, they gave a rating on a five-point scale (1 = excellent, 5 = poor). This question corresponded to the first item of the 36-item short-form health survey (SF-36; [15]).

The remaining 35 items of the SF-36 concern various health-related aspects. We used these 35 items as a measure of HRQOL here. First, participants rated their current health status in comparison to the previous year on a five-point scale (1 = currently much better; 5 = currently much worse). Then, ten activities were rated on a three-point scale (1 = strongly affected, 3 = not affected at all) with regard to how much these were affected by the current health status. These were: exhausting activities, for example, running or lifting heavy objects; moderately demanding activities, for example, moving a table or vacuuming; carrying shopping bags; climbing the stairs; kneeling or bending over; walking more than one kilometer; walking several hundred meter; walking one hundred meter; taking a bath or getting dressed). Four items asked to rate problems at work or daily chores because of physical health during the last 4 weeks on a five-point scale (1 = always, 5 = never): not being able to remain as active as usually; getting done less than usually; only being able to do certain things; having problems at execution. Three items asked to rate problems at work or daily chores because of mental health during the last 4 weeks on a five-point scale (1 = always, 5 = never): not being able to remain as active as usually; achieving less than desired; not being able to work as diligently as usually. One item asked to rate on a five-point scale (1 = not at all; 5 = very much) how strongly normal contact with family, friends, neighbors, or acquaintances had been affected by physical or mental health. Additionally, another item asked to rate on a five-point scale (1 = always; 5 = never) how frequently normal contact with family, friends, neighbors, or acquaintances had been affected by physical or mental health. One item asked to rate pain during the past 4 weeks on a six-point scale (1 = no pain; 6 = very strong pain). One item asked to rate how strongly daily activities at work or at home had been affected by pain on a five-point scale (1 = not at all, 5 = very much). Mood and feelings during the past 4 weeks were assessed by nine items, rated on a five-point scale (1 = always, 5 = never): full of life; nervous; feeling down; calm and peaceful; full of energy; downhearted and depressed; worn out; happy; tired. Finally, four statements on the current health status were rated on a five-point scale (1 = true, 5 = not true): getting sick more easily than others; being as healthy as others; expecting declining health; being at excellent health. After reverse coding where necessary, these items were summed for the HRQOL score with higher values representing lower HRQOL.

Life satisfaction was rated on a five-point scale (1 = very satisfied, 5 = very unsatisfied). Seven items asked about satisfaction with abilities, the way of life so far, physical appearance, self-confidence, personality, happiness, and getting along with others. The mean value served as the life-satisfaction score.

Demographic data comprised gender, age, highest educational degree, and occupation/educational stage (see Table 1).

Table 1 Demographic characteristics of participants
Full size table

2.4 Procedure

Participants responded to all items online. Upon arriving on the respective website, instructions were displayed explaining that all data were collected in completely anonymous manner. Participants were asked to carefully consider their responses to all items. Participants proceeded when they confirmed a respective agreement. By random assignment, then the items appeared in one of four sequences that manipulated the position of the items asking about global SRH and spina bifida. The items appeared subsequently on a series of websites, with items referring to judgements of separate aspects of one question appearing on the same page, however (e.g., the ten items referring to how much different activities were affected by the current health status).

3 Results

Three dependent variables were examined: global SRH, life satisfaction, and HRQOL. In order to compare effects of the independent variables on them, these three variables were z-standardized and entered as three levels of a repeated-measures factor in a mixed-model 3 × 2 × 2 × 2 ANOVA with the between-participants factors positions of the spina-bifida item and global-SRH item, as well as group (with spina bifida, no spina bifida).

The main effect of group was significant, F (1, 236) = 15.49, p < 0.001, ηp2 = 0.06, as were the interactions between the repeated-measures factor and group, F (2, 235) = 14.57, p < 0.001, ηp2 = 0.11, as well as with the position of the spina-bifida item, F (1, 235) = 4.98, p = 0.008, ηp2 = 0.04. Simple effects analyses showed that participants with spina bifida rated global SRH lower than participants without spina bifida, F (1, 236) = 11.50, p < 0.001, ηp2 = 0.05, and also HRQOL, F (1, 236) = 35.63, p < 0.001, ηp2 = 0.13, whereas life satisfaction did not differ reliably, F < 1. The position of the spina-bifida item influenced only ratings of HRQOL significantly, F (1, 236) = 9.90, p = 0.002, ηp2 = 0.04, but not global SRH, F (1, 236) = 1.27, p = 0.26, or life satisfaction, F < 1.

Moreover, the three-way interaction of these factors was also significant, F (1, 235) = 3.32, p = 0.038, ηp2 = 0.03. Simple-simple effects analyses showed that the position of the spina-bifida item influenced only ratings of HRQOL of participants with spina bifida significantly, F (1, 236) = 8.14, p = 0.005, ηp2 = 0.03 (see Fig. 1). Ratings of HRQOL of participants without spina bifida were not significantly influenced, F (1, 236) = 1.86, p = 0.174, nor were ratings of global SRH or life satisfaction of any of the two groups, F < 1.831, p > 0.177.

Fig. 1
figure 1

HRQOL values are shown as a function of group and position of the item asking about spina bifida (SB). Higher values represent lower HRQOL. Error bars represent SEM

Full size image

No other interactions or main effects were significant, F < 3.01, p > 0.084. Thus, no reliable effects of the position of the global-SRH item occurred. Moreover, correlations between global SRH and HRQOL (r = 0.57) or life satisfaction (r = 0.44) did not depend on the position, z < 0.84, p > 0.403.

4 Discussion

The present study replicated findings of lower global SRH and HRQOL of people with spina bifida, whereas their life satisfaction did not differ from the comparison group [1, 2]. In addition, a question order effect occurred. Participants with spina bifida rated their HRQOL to be better when they were asked about spina bifida before. That is, there was a significant difference between ratings of HRQOL. Those participants who were asked about having spina bifida before rating HRQOL rated it to be higher than participants who were asked about having spina bifida after rating HRQOL.

According to a conversational-logic account [8], contrast effects result when the respondent excludes previously rated aspects from the present rating. We assume that this occurred here as well. Asking about spina bifida first prompted an understanding of HRQOL questions as not to refer to health conditions that were linked to having spina bifida. Instead, HRQOL questions rather were answered with regard to any health-related aspects aside from issues associated with that diagnosis. Accordingly, the ratings of the comparison group were not influenced because there was nothing to exclude from HRQOL for participants without spina bifida. Moreover, it could be that after indicating the diagnosis, participants responded to the HRQOL items in reference to other people with spina bifida, not in reference to the general population. A different frame of reference then might imply to rate the own HRQOL as relatively better in a comparison to other people with spina bifida as in a comparison to the general population. Question-order effects have been demonstrated in many domains, for example when assessing attitudes or political beliefs (e.g. [16, 17]). Asking about quality of life may have particularly important personal consequences, however. When question order influences quality-of-life assessments, medical decisions or individual decisions on changing life styles or habits could be impacted. Such effects are seldomly obvious, often overlooked, and must receive more attention.

We can only speculate why the position of the global-SRH items did not influence ratings. Although participants with spina bifida rated global health considerably lower than the comparison group, this difference did not depend on whether it was rated before or after asking about life satisfaction and more specifically about HRQOL. Moreover, correlations between these variables did not differ reliably from each other. This pattern of results might indicate that global SRH and HRQOL are in fact separable yet related constructs. Participants rated both independently and did not blend one into the other. Global health per se was considered something else than an aspect of quality of life.

Taken together, we document evidence on how asking about a specific diagnosis can shape responses to HRQOL items. Where to place such questions in a survey or other situations of asking about HRQOL (e.g. an interview) should be considered carefully because it can change the way questions are understood. A limitation of the present investigation is that participants responded anonymously to an online questionnaire. Practical implications of the present findings deserve more research efforts. For example, future studies should examine question-order effects in clinical contexts as well. When physicians ask their participants how well they are, they might reply differently when asked about a specific diagnosis or specific symptoms before. Subtle variations on what to talk about first might influence the assessment of health status and, as a consequence, contribute to treatment decisions.