Table 1. Variables used in the analysis and variables selected by TWIST system in the subsequent analysis: SDI = 0 vs SDI≥1 (SDI≥1) and SDI = 0 vs SDI≥5 (SDI≥5).
| SDI≥1 | SDI≥5 | |
| Age | x | |
| Age<68 years | x | x |
| Age≥68 years | x | |
| Body Mass Index (BMI) Kg/m2 | x | |
| BMI≤21 | x | x |
| BMI>21<30 | ||
| BMI≥30 | x | x |
| Years since menopause (YSM) | x | x |
| YSM<18 | x | |
| YSM≥18 | x | |
| Number of pregnancies | x | |
| Months of breast feeding | ||
| Current smoking yes | x | x |
| Current smoking no | x | |
| Previous smoking yes | x | |
| Previous smoking no | ||
| Alcohol yes | x | |
| Alcohol no | x | |
| Bone mineral density T-score ≤−2.5 yes | ||
| Bone mineral density T-score ≤−2.5 no | ||
| Previous fragility fracture yes | x | x |
| Previous fragility fracture no | x | |
| Familiar history of femoral fracture yes | x | |
| Familiar history of femoral fracture no | ||
| Calcium intake mg/day | x | x |
| Calcium intake ≤300 mg/day yes | x | |
| Calcium intake ≤300 mg/day no | x | x |
| Arterial hypertension yes | x | |
| Arterial hypertension no | x | |
| Dyslipidemia yes | ||
| Dyslipidemia no | x | x |
| Gastric/oesophagus disease yes | x | |
| Gastric/oesophagus disease no | x | |
| Anxiety/depression yes | x | x |
| Anxiety/depression no | x | |
| Chronic pulmonary obstructive disease (COPD) yes | ||
| Chronic pulmonary obstructive disease (COPD) no | x | |
| Osteoarthritis yes | ||
| Osteoarthritis no | ||
| History of kidney stones yes | x | |
| History of kidney stones no | ||
| Type 2 Diabetes Mellitus (T2D) yes | x | x |
| Type 2 Diabetes Mellitus (T2D) no | ||
| SDI = 0 | ||
| SDI≥1 | ||
| SDI≥5 |
SDI≥1: Variables selected by TWIST system in the analysis aimed to differentiate patients with SDI≥1 from those with SDI = 0 (the number 17, reported in Table 4a, refers to a maximisation of these variables); SDI≥5: Variables selected by TWIST system in the analysis aimed to differentiate patients with SDI≥5 from those with SDI = 0 (the number 25, reported in Table 4b, refers to a maximisation of these variables).
Twist system can easily select just one of the two binary forms of the variables since that choosing one option implies also the information of its complement.