Table 5 Classification rules extracted from the decision tree model; only rules with the condition fall = ye s as consequent and which cover a number of at least 100 instances and have a related accuracy of at least 70% were considered; the rules are ordered by their relative accuracy
From: Mining geriatric assessment data for in-patient fall prediction models and high-risk subgroups
no. | rule (consequent: fall=yes) | relative accuracy |
|---|---|---|
1a b | (Barthel index score ≤ 45 pts) and (sex = male) and (age > 75y) and (Lachs depression item = 0) | 84.3% 77.8% |
2 | (Barthel index score > 10 and ≤ 45 pts) and (sex = female) and (number of medications < 14) and (MMSE score ≤ 26 pts) and (institutionalized=yes) and (needs aid for standing) | 80.0% |
3 | (Barthel index score > 45 and ≤ 65 pts) and (MMSE score > 21 pts) and (Timed 'Up& Go' time ≤ 42s) and (number of diagnoses > 11) and (number of medications > 8) | 72.8% |
4 | (Barthel index score > 45 and ≤ 65 pts) and (MMSE score ≤ 18 pts) | 72.1% |
5 | (Barthel index score > 45 and ≤ 65 pts) and (MMSE score > 18 pts) and (Timed 'Up& Go' time > 42s) and (age > 71y) | 71.9% |