Skip to main content

Table 3 Multistate models for the transition to cluster 3 (“Likely Dementia”)

From: Identification of high likelihood of dementia in population-based surveys using unsupervised clustering: a longitudinal analysis

 

SHARE

Main analysis

Sensitivity analyses

Model 1 (n = 15,278)

Model 2 (n = 15,173)

Model 3 (n = 14,190)

HR (95% CI) (1 → 3)

HR (95% CI) (2 → 3)

HR (95% CI) (1 → 3)

HR (95% CI) (2 → 3)

HR (95% CI) (1 → 3)

HR (95% CI) (2 → 3)

Low education

1.92 (1.58–2.33)

1.32 (0.91–1.9)

1.86 (1.6–2.17)

1.18 (0.86–1.61)

1.77 (1.52–2.07)

0.97 (0.67–1.4)

Hearing loss

1.74 (1.45–2.09)

1.23 (0.85–1.79)

1.38 (1.2–1.59)

1.03 (0.75–1.42)

1.2 (1.03–1.4)

0.88 (0.59–1.32)

Hypertension

1.35 (1.14–1.16)

1.24 (0.9–1.72)

1.36 (1.2–1.55)

1.14 (0.86–1.5)

1.34 (1.17–1.53)

1.09 (0.76–1.55)

Drinking (> 21 units)

0.79 (0.55–1.14)

0.42 (0.16–1.09)

1.25 (0.99–1.58)

0.36 (0.13–1.04)

1.37 (1.09–1.73)

0.54 (0.21–1.41)

Smoking

1.45 (1.13–1.87)

1.29 (0.73–2.29)

1.64 (1.36–1.99)

2.23 (1.57–3.16)

1.7 (1.39–2.07)

2.68 (1.79–4.03)

Depression

2.51 (1.06–3.07)

2.39 (1.62–3.53)

2.05 (1.76–2.4)

1.98 (1.42–2.77)

1.78 (1.51–2.11)

1.65 (1.07–2.54)

Social isolation

1.66 (1.39–1.98)

2.31 (1.51–3.53)

1.61 (1.4–1.86)

1.6 (1.15–2.24)

1.56 (1.35–1.81)

1.15 (0.79–1.68)

Physical inactivity

3.66 (2.97–4.51)

3.21 (2.12–4.87)

2.48 (2.07–2.97)

2.89 (2–4.17)

2.09 (1.72–2.54)

1.33 (0.68–2.6)

Diabetes

2.4 (1.94–2.96)

1.32 (0.79–2.21)

2.15 (1.82–2.54)

1.32 (0.85–2.05)

2.22 (1.88–2.62)

0.88 (0.45–1.73)

Obesity

1.7 (1.39–2.07)

1.58 (1.08–2.32)

1.65 (1.41–1.93)

1.43 (1.01–2.01)

1.76 (1.5–2.06)

1.32 (0.83–2.1)

Air Pollution

0.84 (0.7–1.02)

1.26 (0.83–1.9)

0.92 (0.79–1.07)

1.32 (0.92–1.89)

0.94 (0.81–1.1)

1.32 (0.83–2.08)

  1. Analyses using age as time-scale. All transitions were adjusted for sex. Transition towards the third cluster (“Likely Dementia”) was further adjusted for age and each risk factor individually. All risk factors were taken at baseline. Main analysis was based on a multistate model (Model 1). Sensitivity analyses were based on a multistate survival model with death as an absorbing state. First, 105 participants were removed because of inconsistencies of dates (Model 2). Second, cases identified either at the first or the second waves were removed (Model 3)
  2. HR hazard ratio, CI Confidence interval