Appendix D: Summary of the statistical criteria for model selection

Smallest class size (%) / Number of converging models / SABIC / Entropy / Clinical interpretation
Supportive Care Needs
Physical and Daily Living
1 / - / 10 / 8159.29 / - / Despite a higher SABIC, the 3-class model was selectedbecause the 4-class model contained two trajectories with low scores providing redundant information
2 / 40% / 10 / 8003.98 / 0.71
3* / 22% / 10 / 7982.07 / 0.70
4 / 14% / 10 / 7973.63 / 0.73
Psychological
1 / - / 10 / 8190.31 / - / Despite a higher SABIC, the 3-class model was selectedbecause the 4-class model contained two trajectories with low scores providing redundant information
2 / 49% / 10 / 8105.06 / 0.67
3* / 18% / 10 / 8064.74 / 0.71
4 / 12% / 10 / 8027.90 / 0.79
Patient Care and Support
1 / - / 10 / 8078.29 / -
2 / 28% / 10 / 7904.64 / 0.67
3 / 18% / 10 / 7854.96 / 0.71
4 / 0.60% / 1 / 7878.45 / 0.76
Health System and Information
1 / - / 10 / 8073.41 / - / Despite a higher SABIC, the 3-class model was selectedbecause the 4-class model contained two trajectories with medium scores providing redundant information
2 / 34% / 10 / 7951.77 / 0.63
3* / 5% / 10 / 7929.39 / 0.73
4 / 5% / 9 / 7921.24 / 0.67
Sexuality
1 / - / 10 / 8535.63 / - / Despite a higherSABIC, the 3-class model was selectedbecause the 4-class model contained two trajectories with medium scores providing redundant information
2 / 32% / 10 / 8436.23 / 0.86
3* / 7% / 10 / 8337.55 / 0.89
4 / 6% / 9 / 8286.60 / 0.89
Anxiety
1 / - / 10 / 4905.41 / - / Despite a higher SABIC, the 3-class model was selectedbecause the 4-class model contained two trajectories with low scores providing redundant information
2 / 42% / 10 / 4845.68 / 0.45
3* / 5% / 10 / 4833.62 / 0.61
4 / 5% / 2 / 4832.18 / 0.66
Depression
1 / - / 10 / 4805.52 / - / Despite a higher SABIC, the 2-class model was chosen because the 3 and 4-class models contained several trajectories with low scores providing redundant information
2* / 47% / 10 / 4661.57 / 0.64
3 / 26% / 8 / 4609.95 / 0.72
4 / 12% / 9 / 4597.62 / 0.72
Social Support
Average number of supporting persons
1 / - / 10 / 3210.66 / - / It is more relevant to consider 2 classes with distinct trajectories than 3 or 4 classes with similar ones
2* / 37% / 10 / 3072.71 / 0.64
3 / 18% / 10 / 3032.86 / 0.63
4 / 10% / 9 / 3032.36 / 0.68
Average satisfaction level
1* / - / 10 / 2076.4 / - / The 2-class model yielded almost identical trajectories
2 / 45% / 10 / 1911.52 / 0.60
3 / 0.30% / 3 / 1911.88 / 0.75
4 / 1.40% / 10 / 1973.28 / 0.75
Posttraumatic Growth
Relating to others
1 / - / 10 / 6110.60 / -
2 / 42% / 10 / 6014.30 / 0.57
3 / 7% / 9 / 5997.43 / 0.66
4 / 2% / 10 / 5997.59 / 0.71
New Possibilities
1 / - / 10 / 5609.35 / -
2 / 10% / 10 / 5547.81 / 0.92
3 / 11% / 5 / 5536.80 / 0.51
4 / 3% / 1 / 5536.98 / 0.63
Personal Strength
1 / - / 10 / 5256.57 / - / Despite a higher SABIC, the 3-class model was selected because the 4-class model contained two trajectories with low scores providing redundant information
2 / 46% / 10 / 5213.78 / 0.61
3* / 10% / 9 / 5189.79 / 0.71
4 / 10% / 8 / 5188.89 / 0.64
Spiritual Change
1 / - / 10 / 4109.93 / - / Despite a higher SABIC, the 3-class model was selected because the 4-class model contained two trajectories with high scores providing redundant information
2 / 43% / 10 / 3943.77 / 0.73
3* / 17% / 10 / 3992.39 / 0.83
4 / 13% / 9 / 3955.16 / 0.85
Appreciation of life
1 / - / 10 / 4875.63 / - / In this model, although the convergence was lower the 3-class model wasselected as it yielded a clinically relevant trajectory with low needs.
2 / 49% / 10 / 4726.59 / 0.58
3** / 4% / 5 / 4684.24 / 0.74
4 / 6% / 3 / 4673.46 / 0.66
SABIC: Sample-Size Adjusted Bayesian Information Criteria. Note: we sought models with 1) a smallest class size > 4% 2) clinical interpretability 3) lower values for the SABIC 4) highest number of converging models and 5) entropy closest to 1. *: In these models,. a higher number of classes yielded trajectories with similar shape so interpretability was favored over SABIC.** : In this model although the convergence was lower,. 3 classes were chosen as it yielded a clinically relevant trajectory with low stable posttraumatic growth.

1