Skip to main content

Table 3 Diagnostic properties of clinical prediction rule applied to validation cohort

From: Risk factors, costs and complications of delayed hospital discharge from internal medicine wards at a Canadian academic medical centre: retrospective cohort study

Score

ALC patients / Patients with same number of points

Positive predictive value

ALC patients

(N = 124)

Non-ALC patients

(N = 2032)

Likelihood ratio

(95% CI)

Point system

 0 point

2 / 459 (0.4%)

2 (1.6%)

457 (22.5%)

0.07 (0.02–0.28)

 1 point

11 / 742 (1.5%)

11 (8.9%)

731 (36.0%)

0.25 (0.14–0.44)

 2 points

17 / 440 (3.9%)

17 (13.7%)

423 (20.8%)

0.66 (0.42–1.03)

 3 points

26 / 314 (8.3%)

26 (21.0%)

288 (14.2%)

1.48 (1.03–2.12)

 4 points

40 / 148 (27.0%)

40 (32.3%)

108 (5.3%)

6.07 (4.43–8.31)

 5 points

23 / 45 (51.1%)

23 (18.6%)

22 (1.1%)

17.13 (9.83–29.86)

 6 points

4 / 7 (57.1%)

4 (3.2%)

3 (0.2%)

21.85 (4.94–96.56)

 7 points

1 / 1 (100%)

1 (0.8%)

0 (0%)

N/A

Risk category

 Low risk (0–1 points)

13 / 1201 (1.1%)

13 (10.5%)

1188 (58.5%)

0.18 (0.11–0.30)

 Medium risk (2–3 points)

43 / 754 (5.7%)

43 (34.7%)

711 (35.0%)

0.99 (0.77–1.27)

 High risk (> = 4 points)

68 / 201 (33.8%)

68 (54.8%)

133 (6.6%)

8.38 (6.66–10.54)

  1. In the clinical prediction rule, score for each patient is calculated where each of the predictors in the multivariate model in Table 3 is worth 1 point (Age > =80 years = 1 point; female = 1 point; dementia = 1 point; diabetes with complications = 1 point; physiotherapy = 1 point; occupational therapy = 1 point; speech language pathologist = 1 point). N/A = not applicable as too few patients to accurately calculate likelihood ratios