Skip to main content

Table 4 Predictive accuracy for predicting composite of negative health outcomes, mortality, ADL decline, and institutionalisation after 12 months of follow-up

From: Predictive validity of a two-step tool to map frailty in primary care

Models

Composite of negative health outcomesa

p

ADL decline

p

Institutionalisation

p

Died

p

Age, sex, GP practice; AUC (reference model)

0.70

[ref]

0.65

[ref]

0.73

[ref]

0.77

[ref]

Age, sex, GP practice, number diseases, number medications; AUC (change versus reference model)

0.73 (+0.03)

0.7

0.69 (+0.03b)

0.23

0.74 (+0.01)

0.82

0.80 (+0.03)

0.38

Age, sex, GP practice, EASY-Care TOS judgement; AUC (change versus reference model)

0.75 (+0.05)

0.02

0.70 (+0.05)

0.07

0.76 (+0.03)

0.57

0.85 (+0.08)

0.13

Age, sex, GP practice, CGA judgement; AUC (change versus reference model)

0.76 (+0.07b)

0.005

0.72 (+0.07)

0.009

0.77 (+0.04)

0.43

0.87 (+0.11b)

0.009

  1. Predictive accuracy of different alternative models reported in the rows as the Area Under the Receiver Operating Curve (AUC). This table describes the additional predictive value of different alternatives, when age, sex and GP practice are the reference model
  2. ADL activities of daily living, GP general practitioner, EASY-Care TOS EASY-Care Two step Older people Screening
  3. aComposite of ADL decline, institutionalisation, too ill to be assessed with EASY-Care TOS at follow up, and died
  4. bDifference between the result for the change score and extracting the reported AUC of the reference model from reported AUC of larger model (e.g. +0.07 instead of 0.76 − 0.70 = +0.06) is due to rounding off the results to the second decimal