Page 1 of 1

Figure 1 -2

Triggers and Targeted Injury Detection Systems (TIDS)

Figure 1. Relationship of positive predictive value (PPV) and prevalence shown for three combinations of sensitivity (Sn) and specificity (Sp)

This figure displays positive predictive value (PPV) on the y axis as a function of prevalence on the x axis. PPV ranges from 0 to 25%. Prevalence ranges from 1-2% There are three lines for three levels of sensitivity: 25%., 50%, and 75%. In each case specificity is 95%. Each line starts at the origin and is slightly convex. The line for the three levels of sensitivity slope upward to about 9%, 17%, and 25% respectively.

Return to Document

Figure 2. 95-percent confidence limits around sensitivity and positive predictive value (PPV) for various levels of prevalence and sample sizes.

This figure displays two graphs: sensitivity and positive predictive value (PPV) as a function of prevalence, Each y-axis ranges from 0-100%. Prevalence ranges from 0-10%. The graphs display a line for central tendency and lines for 95% confidence intervals at n=100, 250, 500, 750, and 1000. In the sensitivity graph the central tendency line is horizontal for sensitivity=50%. 95% confidence lines are increasingly convex towards the central tendency line as n increases. The n=100 line is almost flat from 0% to 20% (lower bound) and 100% to 85% (upper bound) The lines for other n all start below 5% and above 95% at low prevalence and converge towards the central tendency, with the n=1,000 line at +/-10% or so at prevalence = 10%. As described in the text, the lines are much steep er up to about 2% prevalence. The graph for PPV shows a central tendency line that starts near 4% and increases to 50% at prevalence = 10%. The lines for confidence intervals generally follow the slope of the central tendency line with the confidence width interval remaining fairly constant with increasing prevalence.

Note: Sensitivity and specificity were set at 50 percent and 95 percent, respectively. Waves in lines are caused by rounding values to integers for computation of confidence intervals. Thick lines are the point estimates of each characteristic.

Return to Document


Current as of February 2009
Internet Citation: Figure 1 -2: Triggers and Targeted Injury Detection Systems (TIDS). February 2009. Agency for Healthcare Research and Quality, Rockville, MD.