A practical dementia risk estimator for routine health data.

Enter the 38 non-cognitive variables used by the published Cog-Free model. The result is a short-term statistical screening estimate for adults 65 and older, not a diagnosis or a substitute for clinical evaluation.

Inputs 38

Routine questionnaire and basic physical measures

Model LR

Published logistic regression calculator

Internal AUC 0.86

Reported 95% CI: 0.82 to 0.89

High-risk threshold 51.37%

Threshold used in the public calculator

Risk inputs

Group 01Sociodemographics Age, education, social and coverage variables

Age, years

Education, years

Living arrangement

Medical insurance

Satisfied with income

Family emotional support

Living siblings

Close friends

Group 02Physical measures Basic measures available in routine health checks

Height, cm

Weight, kg

Waistline, cm

Right-hand grip, kg

Heart rate, bpm

Group 03Medical history Personal and family history indicators

Hypertension

Anemia

History of head trauma

Facial infection

Thyroid dysfunction

Cervical spondylosis

Depression

Parkinson's disease

Family history of diabetes

Family history of stroke

Group 04Behavior and lifestyle Diet, cognitive activity and sleep interruption

Fish consumption

Fruit consumption

Eating behavior

Cognitive activity

Nighttime awakenings

Group 05Peripheral function Sensory, gait and bowel variables

Hearing compared with before

Sense of smell compared with before

Vision compared with before

Walking pace

Gait instability

Defecation habit

Defecation duration, minutes

Group 06Subjective memory Self-reported memory status, without cognitive testing

Memory concerns

Worse memory than peers

Saw a doctor for memory concerns

Risk result

0 of 38 inputs

Waiting for inputs

Lower estimate High risk at 51.37%

Complete the form or load the demo profile to calculate the published logistic regression estimate.

Logit score --

Missing inputs 38

Largest model terms for this profile

No calculation yet--

This is a screening estimator. It does not diagnose dementia, Alzheimer's disease, mild cognitive impairment or any medical condition. High estimates should lead to professional assessment, not self-diagnosis.

Model limits and sign warnings

The formula reproduces the coefficients exposed in the authors' public calculator. The paper warns that several inverse associations, including hearing loss, head trauma, nighttime awakenings and anemia, likely reflect cohort and measurement effects. Do not treat a negative coefficient as advice to acquire or ignore a health problem.

The development cohort was community-dwelling adults aged 65 and older in central China.
The follow-up window was short, with mean follow-up of about 2.18 years.
Several inputs are self-reported and can be affected by recall or coding differences.
The model is useful for triage and discussion, not individual clinical certainty.