Correction to: Resilience to health challenges is related to different ways of thinking: mediators of physical and emotional quality of life in a heterogeneous rare-disease cohort

Correction to: Qual Life Res (2017) 26:3075–3088 https://doi.org/10.1007/s11136-017-1633-2

It has come to our attention that, due to a statistical programming error, the Resilience score was incorrectly calculated. The score used in the analyses reported in our article “Resilience to health challenges is related to different ways of thinking: Mediators of quality of life in a heterogeneous rare-disease cohort” [1] reflected the predicted values for CDC Health Days Activities of Daily Living (ADL), not the saved residuals multiplied by negative 1, as we had thought and thus reported in the abovementioned article. This error could affect the interpretation of the findings presented in the article that was published. Accordingly, we have re-analyzed the data after correcting this error. Resilience operationalized in the corrected analysis manner reflects the intended score: greater- or less-than-expected ADL days in the past month given physical and mental health problem days.

Results are summarized in Tables 2, 3, and 4 below, and in the pie charts below (Figs. 1 and 2). Tables 2 and 3 highlight the small differences in the regression models from those originally published. Specifically, while physical-functioning models with resilience explain less variance than in the original publication, the p values were unchanged. For mental-health functioning, corrected models with resilience explain less variance than in the original publication and the p values for resilience become less statistically significant in models with appraisal (p = 0.031) and non-significant in models with appraisal and catalysts (p = 0.139). Resilience measured in this manner contributes little unique variance to the explanation of physical- or mental- health functioning (1.1% and 0%). The small total effect of resilience is largely subsumed by appraisal (0.7% and 0.4%). The explanatory power of appraisal alone, however, is not affected by this error, and remains substantial (13.7% and 27.4% for physical and mental health, respectively).

Table 2 Hierarchical series of regression models predicting physical functioning to test mediation hypothesis

Table 3 Hierarchical series of regression models predicting emotional functioning to test mediation hypothesis

Table 4 Summary of explained variance in hierarchical regression model series testing mediation hypothesis

Fig. 1

Decomposition of Explained Variance in Physical Functioning. Proportion of variance in physical health-related QOL explained by appraisal and resilience, after controlling comorbidities and treatment (standard model). Overlap indicates variance explained by both resilience and appraisal, suggesting relatively weak mediation. Specifically, 0.007/(0.007+0.137) = 4.9% of the association of appraisal with physical health is also explained by resilience

Fig. 2

Decomposition of Explained Variance in Emotional Functioning. Proportion of variance in mental health-related QOL explained by appraisal and resilience, after controlling for comorbidities and treatment (standard model). Overlap indicates variance explained by both resilience and appraisal, suggesting almost no role for resilience. Specifically, 0.004/(0.004+0.274) = 1.43% of the association of appraisal with mental health is also explained by resilience

The conclusions of the paper are similar. Appraisal processes differ somewhat for physical and emotional outcomes, and resilient people employ different processes than non-resilient people. Namely, high resilience was associated with a focus on maintaining a calm and healthy lifestyle, self-acceptance, and remaining positive. In contrast, low resilience was associated with a focus on health problems, concern about what their doctors are telling them, and frequent social comparison to others who were better off. In these corrected analyses, resilience was a significant predictor of physical but not emotional functioning in the full model (p < 0.0001 and p = 0.139). All effects involving resilience were much smaller in these corrected analyses, even if significant.