Skip to main content

Table 2 Features retained in Logistic Regression with Lasso Penalisation, 1 year and 2–5 years separated

From: Identifying undetected dementia in UK primary care patients: a retrospective case-control study comparing machine-learning and standard epidemiological approaches

Feature name

Logistic regression parameter

1 year prior to diagnosis/matched date

1 year prior to diagnosis/matched date

2–5 year predictors

Disorientation and Wandering

2.31

0.88

Behaviour change

1.99

0.65

Schizophrenia

1.53

–

Self-neglect

1.45

–

Difficulty managing

1.38

–

Personality change

1.18

0.58

Family history of dementia

1.14

–

Third party consultation

0.85

–

Antidepressant

0.81

–

Antipsychotic medication

0.76

−0.11

Cerebrovascular disease

0.58

0.14

Did not attend

0.56

0.22

GP home visit

0.55

−0.11

Bipolar disorder

0.51

− 0.11

Interaction with social services

0.51

–

Possible Fall

0.47

0.22

Alcohol

0.42

–

Unable to cope

0.41

0.21

Attended Emergency Department

0.39

–

Depression

0.34

–

Living in a nursing home

0.31

–

Receiving care in home

0.28

–

Epilepsy or Seizures

0.23

0.25

Blood pressure measurement

0.16

–

Stroke

0.15

–

Routine hospital admission

0.15

−0.14

Z-drugs

0.13

−0.11

Lower limb fracture

0.12

–

Receiving care in home

0.11

–

Anxiety

0.10

–

Impaired mobility

0.10

–

Needs help with activities of daily living

−0.11

−0.30

Dressing of wound, burn or ulcer

−0.13

–

Family bereavement

−0.16

–

Hypertension

−0.20

−0.16

Infections

−0.21

−0.16

Angina

−0.22

–

Vertebral collapse

−0.27

–

Lithium

−0.28

–

PTSD reaction

−0.46

–

Cancer

−1.06

–

Psychotic Depression

−1.11

–

Personality disorder

–

0.21

Constipation

–

0.10

Coronary Heart Disease

–

−0.12

Obesity

–

−0.16

Benzodiazepines

–

−0.22