Skip to main content

Table 3 Most prominent feature subset selected using multivariate features analysis based on Weka correlation-based feature selection method with best fit search method for classification of not frail and pre-frail (FFP related features removed, 63 features remain)

From: Machine learning models for identifying pre-frailty in community dwelling older adults

Fried Frailty Phenotype

Clinical Frailty Scale

Features

Scorea

Features

Score

Higher Body Mass Index (BMI)

0.157

Higher BMI

0.141

Lower Purdue Dexterity Test score (both hands)

0.089

Lower muscle mass

0.116

Higher Kessler Psychological Distress Scale (K10) score

0.075

Higher fat mass

0.090

Higher Pittsburgh Sleep Quality Index (PSQI) Global score

0.072

Lower FMS score (raised right leg/left arm)

0.087

Lower Functional movement screen (FMS) score (hurdle, left leg first)

0.059

Higher K10 score

0.077

Lower FMS score (raised right leg/left arm)

0.056

Higher PSQI Global score

0.074

Poorer balance assessment (eyes closed)

0.055

Lower FMS score (hurdle, right leg first)

0.059

Poorer balance assessment (eyes open)

0.054

Poorer balance assessment (eyes open)

0.054

Lower FMS score (lunge with right leg)

0.053

Shortness of breath

0.047

Bother with pelvic floor problems

0.050

Lower FMS score (hurdle, left leg first)

0.047

Shortness of breath

0.047

Bother with pelvic floor problems

0.046

Lower FMS score (raised left leg/right arm)

0.044

Poor lung function test scores

0.031

Lower lung function test scores

0.033

Avoiding/unable to eat some foods because of your mouth, teeth or dentures

0.024

Housing (more likely to live in a house)

0.032

Lower GPCOG test score

0.023

Income (more likely to be sourced from a pension)

0.032

Less able to walk up and down 15 stairs without rest

0.021

Lower General Practitioner assessment of Cognition (GPCOG) test score

0.024

Poorer balance assessment (heel toe steps backwards)

0.019

Self-conscious about their mouth, teeth or dentures

0.021

Mode of transportation (less likely to walk or ride a bike)

0.019

Poorer balance assessment (heel toe steps backwards)

0.018

Poorer balance assessment (standing, right leg, eyes open)

0.012

Current pain

0.015

Diagnosed with any health conditions

0.011

Higher consumption of alcohol

0.013

  1. Bold text indicates features found in both the FFP and CFS feature ranking
  2. aScore: variable ranking by their contribution to the prediction using Random forest algorithm. Higher scores indicate greater predictive value