Our study findings add to the growing body of literature demonstrating that patient personality factors are associated with patients’ self-ratings of their health, and demonstrate the association extends to a preference-based health measure. They also suggest complex interrelationships exist among patient personality factors, socio-demographics, medical conditions, and preference-based health assessments, raising important questions regarding the external validity of clinical research studies including CEAs.
Consistent with a number of prior studies exploring associations between FFM factors and self-rated health as assessed using non-preference-based measures [4
], we found higher Neuroticism was associated with worse preference-based health, as measured by the EQ-5D summary index, in our sample of chronically ill RCT participants. The association appeared robust, persisting after adjustment for socio-demographic factors [17
] and medical conditions/diagnoses [18
] that are recognized powerful correlates of self-rated health.
The effect of a 1 SD change in Neuroticism on EQ-5D summary index scores was comparable to that of having depression or arthritis, greater than the effect of having diabetes, heart failure, or chronic lung disease, and much greater than the effects of socio-demographic variables that are currently almost universally measured and adjusted for in clinical studies. Neuroticism also accounted for more variance in EQ-5D summary index scores than these traditionally measured variables. For a 1 SD increase in Neuroticism, the summary index score was decreased by 0.04, which falls within the MCID range for the EQ-5D of 0.03 [51
] to 0.07 [52
]. Across an entire population, which encompasses in excess of 4 standard deviations in personality scores, the effect of Neuroticism on EQ-5D summary index scores would therefore be around 0.16. Thus, the association between Neuroticism and EQ-5D summary index scores we observed appears to have importance from a population health perspective.
Regarding individual EQ-5D dimensions, higher Neuroticism was associated with worse anxiety/depression scores, a finding consistent with a large body of prior research. We also noted higher Openness was associated with worse anxiety/depression scores, a finding that may seem more puzzling at first glance. However, predicting the effects of relatively high levels of FFM personality factors is not always a straightforward matter, and beyond its potential benefits, high Openness may have some downsides [55
]. For example, a prior study found higher Openness was associated with worse scores on the Social Functioning scale of the non-preference-based SF-36 self-rated health measure [21
], and has also been associated with mood disorders [56
]. Several attributes of high Openness individuals might help to explain such findings. These include a tendency toward high sensitivity to feelings and emotions, as well as a proclivity for unconventional or esoteric interests that might make it difficult for individuals to relate to others who do not share such interests [57
]. Thus, interrelationships between FFM personality factors and self-rated health can be complex, underscoring the need for additional empirical research in this area. Finally, higher Conscientiousness was associated with better usual activities scores, consistent with prior literature concerning non-preference-based self-rated health measures [21
]. These findings again all appeared robust, persisting after full adjustment for covariates.
We found no associations between Agreeableness and Extraversion and any EQ-5D dimension scores, consistent with the findings of a single prior study examining the relationship of personality factors with sub-facets of self-rated health, measured via SF-36 subscales [21
]. We also found no associations of Agreeableness, Conscientiousness, Extraversion, or Openness with EQ-5D summary index scores. The results of prior studies concerning associations between these personality factors and self-rated health assessed via non-preference-based measures have been mixed, with some but not all finding associations, and the specific personality factors considered and/or found to be associated with self-rated health varying among “positive” studies [2
The reasons for the differences in specific associations between FFM personality factors and self-rated health in our study compared with previous studies remain unclear. However, considerable variation among studies in regard to participant and design characteristics is likely to be one important contributor. For example, our analyses employed baseline data from RCT participants, whereas all prior studies exploring personality/self-rated health associations employed data from observational studies. This is salient because accumulating evidence suggests that RCT participants may differ from others, including observational study participants, in ways that may influence their self-assessments of health [58
It also seems likely that different self-rated health measures may tap distinct and/or only partially overlapping facets of the broad subjective health construct. Indeed, a number of prior studies that compared the performance of different preference-based and non-preference-based self-rated health measures found important differences among measures in terms of baseline scores and responsiveness to change [33
]. Thus, different self-rated health measures should not necessarily be expected to yield uniform results. There are also significant correlations among the FFM factors, so that the effects of the less salient factors may have been obscured in this sample. In analyses (not presented) that examined only one of the FFM factors at a time, three of the factors (Neuroticism, Conscientiousness, and Extraversion) exhibited strong adjusted associations in the direction predicted. Finally, all self-rated health measures have specific limitations in their performance that might influence the findings of analyses exploring personality factor/self-rated health associations. For example, the EQ-5D is susceptible to ceiling effects [53
], and its scores tend to be skewed toward better health [64
]. The possibility that different self-rated health measures may be differentially susceptible to contamination by different personality factors warrants further study.
Regardless of the specific connections between different FFM personality factors and self-rated health measures, our findings have potential ramifications for clinical research studies. For example, since it appears individuals with relatively high levels of Neuroticism have worse self-rated health as assessed by both preference-based and non-preference-based measures, one might hypothesize such individuals may have (and/or perceive they have) “more to gain” from interventions aimed at improving self-rated health than those with lower Neuroticism. Personality factors might also moderate the effects of interventions on preference-based health [65
]. For example, one might hypothesize individuals higher in Neuroticism might be more (or less) responsive to a given intervention than those lower in Neuroticism. Indeed, in prior analyses, we found that the self-efficacy enhancing effects of our study experimental intervention were confined to participants who were higher in Neuroticism and/or lower in Conscientiousness, Agreeableness, and Extraversion [66
Such personality-driven effects on intervention receptiveness and/or response may reduce the external validity or applicability of RCTs as well as observational studies if, as some research evidence suggests [38
], the status of psychological variables such as FFM personality factors is different in those who enroll and remain in clinical studies than for the general population. Personality effects may additionally threaten internal validity in observational studies. Subjects are not randomly assigned to groups in such studies, so it is unlikely the status of various personality factors is equally distributed among groups.
Finally, it follows from these examples that unmeasured personality effects might bias the findings of CEAs, since preference-based health assessments are used to calculate QALYs for use in cost-effectiveness ratios. The key point is that routinely assessing the status of personality factors in clinical studies, along with the usual socio-demographic variables, would permit detection and, when indicated, statistical control for such effects. These examples, provided here for illustrative purposes, remain somewhat speculative given the relative paucity of supporting empirical research. Clearly, additional studies examining what are likely to be complex interrelationships among patient personality, changes in preference-based health in response to interventions, and intervention cost-effectiveness estimates appear warranted.
Our study had some limitations. As noted previously, we examined a sample of chronically ill outpatients who volunteered for a RCT, which may limit the generalizability of our findings to other groups and settings. For example, mean EQ-5D summary index and dimension scores were somewhat lower than in the general population [29
]. Likewise, mean Neuroticism scores were somewhat higher and mean scores for the other four FFM factors somewhat lower than reported in the general population [44
]. Additionally, because our analyses were cross-sectional, causality cannot be inferred from the personality/preference-based health associations we observed. In other words, while it seems likely that the status of certain personality factors contributed to poorer health in some subjects, it may also be true that subjects’ overall health and/or specific medical conditions helped shape the status of their personality factors over time. Indeed, recent research indicates that, rather than being viewed as intransigent “traits,” FFM factors may best be conceptualized as general tendencies that are subject to significant change over time within some individuals [67
]. It is also unclear to what extent differences in self-rated health among persons with varying levels of the FFM personality traits reflect differences in their reporting of subjective health and/or differences in “actual” underlying health. On the other hand, the apparently robust relationship between personality and mortality in a number of prior longitudinal studies strongly suggests a more fundamental relationship also exists [27
In conclusion, our study demonstrated that personality factors in the FFM were associated with preference-based self-rated scores in a sample of outpatients participating in a RCT of a chronic illness self-management intervention. These associations remained significant even after adjusting for socio-demographic factors and medical conditions, which are known to influence preference-based health ratings. Furthermore, among these variables, FFM personality factors were the most powerful correlates of preference-based health. Our findings underscore the need for additional studies conducted with a wide array of samples, at least some of which are followed longitudinally, to further explore the potential for unmeasured effects due to patient personality factors in clinical research.