The role of daily adjustment disorder, depression and anxiety symptoms for the physical activity of cardiac patients

Background Physical activity (PA) is crucial in the treatment of cardiac disease. There is a high prevalence of stress-response and affective disorders among cardiac patients, which might be negatively associated with their PA. This study aimed at investigating daily differential associations of International Classification of Diseases (ICD)-11 adjustment disorder, depression and anxiety symptoms with PA and sedentary behaviour (SB) during and right after inpatient cardiac rehabilitation. Methods The sample included N = 129 inpatients in cardiac rehabilitation, Mage = 62.2, s.d.age = 11.3, 84.5% male, n = 2845 days. Adjustment disorder, depression and anxiety symptoms were measured daily during the last 7 days of rehabilitation and for 3 weeks after discharge. Moderate-to-vigorous PA (MVPA), light PA (LPA) and SB were measured with an accelerometer. Bayesian lagged multilevel regressions including all three symptoms to obtain their unique effects were conducted. Results On days with higher adjustment disorder symptoms than usual, patients engaged in less MVPA, and more SB. Patients with overall higher depression symptoms engaged in less MVPA, less LPA and more SB. On days with higher depression symptoms than usual, there was less MVPA and LPA, and more SB. Patients with higher anxiety symptoms engaged in more LPA and less SB. Conclusions Results highlight the necessity to screen for and treat adjustment disorder and depression symptoms during cardiac rehabilitation.


Inclusion of all control variables for sensitivity analyses
This section of the supplementary material shows different models for sensitivity analysis, for transparency reasons, or for completeness regarding the registration of the hypotheses and statistical plan prior to the analyses of the data (see [link blinded for review] preregistration).
The following control variables were included for sensitivity analyses: 1) Age: centred at the mean.
2) Gender: coded as 0 for female and 1 for male.
3) Weekday: coded as 1 for weekday and 0 for weekend day = 0. 4) Discharge: coded as 0 for days at the clinic and 1 for days at home after discharge.5) Low variance: differentiating people with variance in their questionnaire answers across time (=0) from people without variance = 1.6) Smoking: coded as 0 for non-smokers and as 1 for smokers.
7) Physical quality of life: Score for self-reported physical quality of life at the beginning of cardiac rehabilitation measured with the MacNew Heart Disease Quality of Life Questionnaire (Höfer et al., 2004).Items to form the subscale for physical quality of life were chosen following the proposed factorial structure by Bermudez et al. (2021) and were rated on a 7-point-Likert-scale from 1 (e.g., never) to (e.g., all the time).The score was built as the mean of the items and, thus, could range from 1 to 7. Higher values indicate a higher quality of life.8) Number of diagnoses: This control variables indicates the number of ICD-10 diagnoses in record at the rehabilitation center (maximum of 9) as a proxy for comorbidity.9) BMI: Body-Mass-Index.
Table S1 shows the Bayesian lagged multilevel models for each outcome (MVPA, LPA, and SB) and with all control variables for sensitivity analyses.

Table S1
Bayesian lagged multilevel models with all control variables for sensitivity analyses ).CI = credible interval.LL = lower limit.UL = upper limit.¹ Negative binomial multilevel model.² Gaussian multilevel model.Estimates marked with an "*" represent significant results inferred from the 95% CI excluding zero.Note that significance for the random effects cannot be derived from the CI, given that the regression coefficient is the estimated standard deviation (always positive).

Preregistered same-day multilevel models
The registered statistical analyses also included same-day only models.These models are reported in Table S2 and include interindividual and same-day intraindividual predictors (adjustment disorder, depression, and anxiety symptoms), as well as essential control variables (wear time and time).).CI = credible interval.LL = lower limit.UL = upper limit.¹ Negative binomial multilevel model.² Gaussian multilevel model.Estimates marked with an "*" represent significant results inferred from the 95% CI excluding zero.Note that significance for the random effects cannot be derived from the CI, given that the regression coefficient is the estimated standard deviation (always positive).

Preregistered exploratory analyses
Table S3 reports the lagged multilevel model with MVPA as outcome and the inclusion of all exploratory analyses, that is, the interaction of the discharge variable with each intraindividual predictor.Tables S4 and S5 show the same exploratory analyses, but for the outcomes light physical activity (LPA) and sedentary behaviour (SB), accordingly.).CI = credible interval.LL = lower limit.UL = upper limit.¹ Negative binomial multilevel model.² Gaussian multilevel model.Estimates marked with an "*" represent significant results inferred from the 95% CI excluding zero.Note that significance for the random effects cannot be derived from the CI, given that the regression coefficient is the estimated standard deviation (always positive).458 days.LPA = light physical activity.b = unstandardized regression coefficient.SE = standard error.CI = credible interval.LL = lower limit.UL = upper limit.¹ Negative binomial multilevel model.² Gaussian multilevel model.Estimates marked with an "*" represent significant results inferred from the 95% CI excluding zero.Note that significance for the random effects cannot be derived from the CI, given that the regression coefficient is the estimated standard deviation (always positive).SE = standard error.CI = credible interval.LL = lower limit.UL = upper limit.¹ Negative binomial multilevel model.² Gaussian multilevel model.Estimates marked with an "*" represent significant results inferred from the 95% CI excluding zero.Note that significance for the random effects cannot be derived from the CI, given that the regression coefficient is the estimated standard deviation (always positive).

Multilevel models with only anxiety symptoms
Given the surprising direction of the associations of anxiety symptoms found in the models reported in the main article (see Table 3), we decided to run the models including anxiety symptoms only.In Table S6, we report three multilevel models with anxiety symptoms (without adjustment disorder and depression symptoms) and essential control variables only.A model is reported for each type of activity (MVPA, LPA, and SB) as outcome.All associations of anxiety symptoms were rendered non-significant.Multicollinearity is a likely explanation, given the very high icorrelations between the different types of symptoms (see Table 2 in the main article).This would only apply to the interindividual level where correlations between the different types of symptoms were >.70.At the intraindividual level, correlations were not as high to expect multicollinearity.A second explanation might still be that the strong overlap between depression symptoms, adjustment disorder symptoms and anxiety symptoms might mask the unique effects that only show when the other constructs are controlled for.).CI = credible interval.LL = lower limit.UL = upper limit.¹ Negative binomial multilevel model.² Gaussian multilevel model.Estimates marked with an "*" represent significant results inferred from the 95% CI excluding zero.Note that significance for the random effects cannot be derived from the CI, given that the regression coefficient is the estimated standard deviation (always positive).).CI = credible interval.LL = lower limit.UL = upper limit.¹ Negative binomial multilevel model.² Gaussian multilevel model.Estimates marked with an "*" represent significant results inferred from the 95% CI excluding zero.Note that significance for the random effects cannot be derived from the CI, given that the regression coefficient is the estimated standard deviation (always positive).).CI = credible interval.LL = lower limit.UL = upper limit.¹ Negative binomial multilevel model.² Gaussian multilevel model.Estimates marked with an "*" represent significant results inferred from the 95% CI excluding zero.Note that significance for the random effects cannot be derived from the CI, given that the regression coefficient is the estimated standard deviation (always positive).

Table S2
Bayesian same-day multilevel models

Table S3
Bayesian lagged multilevel model with MVPA as outcome and all exploratory interactions with after rehab = 129 patients.n = 2'458 days.MVPA = moderate-vigorous physical activity.b = unstandardized regression coefficient.SE = standard error.IRR = incidence rate ratio (i.e., e b

Table S4
Bayesian lagged multilevel model with LPA as outcome and all exploratory interactions with after rehab

Table S5
Bayesian lagged multilevel model with SB as outcome and all exploratory interactions with after rehab

Table S7
Bayesian lagged multilevel models with only adjustment disorder symptoms

Table S8
Bayesian lagged multilevel models with only depression symptoms