Prediction of poor outcomes six months following total knee arthroplasty in patients awaiting surgery

Table 4 Validity measures of the predictive rule

Measure	Estimates in training sample	Estimates with 1,000 bootstrap resamples
*Sensitivity% (95% CI)*	82.1 (64.4-92.1)	82.1 (66.7-95.8*)
*Specificity% (95% CI)*	71.7 (62.8-79.2)	71.7 (62.8-79.8*)
*Positive predictive value% (95% CI)*	41.8 (29.7-55.0)	41.8 (29.1-55.8*)
*Negative predictive value% (95% CI)*	94.2 (87.1-97.5)	94.2 (88.8-98.8*)
*Positive likelihood ratio (95% CI)*	2.90 (2.06-4.08)	2.90 (1.81-4.74*)
*Negative likelihood ratio (95% CI)*	0.25 (0.11-0.57)	0.25 (0.11-0.58*)
*Area under ROC curve (95% CI)*	0.77 (0.69-0.85)	0.77 (0.69-0.85*)

• *95% asymptotic confidence intervals.
Sensitivity: number of participants classified at risk both by the PR and the post-operative WOMAC score divided by all participants classified at risk by the post-operative WOMAC score (actual outcome).
Specificity: number of participants classified not at risk by the PR and the post-operative WOMAC score divided by all participants classified not at risk by the post-operative WOMAC score (actual outcome).
Positive predictive value: number of participants classified at risk by the PR and the post-operative WOMAC score divided by all participants classified at risk by the PR (predicted outcome).
Negative predictive value: number of participants classified not at risk by the PR and the post-operative WOMAC score divided by all participants classified not at risk by the PR (predicted outcome).
Positive likelihood ratio: sensitivity/(1-specificity).
Negative likelihood ratio: (1-sensitivity)/specificity.
Area under the ROC curve is defined as the area under the sensitivity vs. 1-specificity curve.

ISSN: 1471-2474