Skip to main content

Table 2 Test characteristics of Medicare claims-based models for identifying high risk conditions or uncontrolled LDL cholesterol conditions defined in REGARDS study data

From: Claims-based algorithms for identifying Medicare beneficiaries at high estimated risk for coronary heart disease events: a cross-sectional study

Condition and model*

N

Prevalence of condition

Predicted probability threshold

Sensitivity (95% CI)

Specificity (95% CI)

PPV (95% CI)

NPV (95% CI)

C statistic (95% CI)

Among all eligible participants

        

 Condition 1: High risk for CHD events

6,615

49%

      

  Pre-specified

  

0.55

0.69 (0.67, 0.70)

0.90 (0.89, 0.91)

0.87 (0.85, 0.88)

0.75 (0.73, 0.77)

0.86 (0.85, 0.86)

  Pre-specified + data mining

  

0.52

0.71 (0.69, 0.72)

0.89 (0.88, 0.90)

0.86 (0.84, 0.87)

0.76 (0.74, 0.77)

0.87 (0.86, 0.88)

 Condition 2: Very high risk for CHD events

6,615

14%

      

  Pre-specified

  

0.28

0.63 (0.59, 0.66)

0.90 (0.89, 0.90)

0.52 (0.49, 0.54)

0.94 (0.93, 0.95)

0.84 (0.83, 0.86)

  Pre-specified + data mining

  

0.26

0.66 (0.63, 0.70)

0.90 (0.89, 0.90)

0.52 (0.49, 0.55)

0.94 (0.93, 0.94)

0.86 (0.84, 0.87)

Among eligible participants without history of CHD or risk equivalents

        

 Condition 3: Framingham risk score >20%

3,720

9%

      

  Pre-specified

  

0.20

0.47 (0.43, 0.54)

0.90 (0.89, 0.91)

0.31 (0.27, 0.36)

0.95 (0.94, 0.96)

0.82 (0.80, 0.84)

  Pre-specified + data mining

  

0.20

0.46 (0.40, 0.52)

0.90 (0.89, 0.91)

0.31 (0.27, 0.35)

0.95 (0.94, 0.95)

0.81 (0.79, 0.84)

Among eligible participants at high risk for CHD events who were using statins

        

 Condition 4: LDL cholesterol ≥100 mg/dL

1,583

30%

      

  Pre-specified

  

0.43

0.19 (0.15, 0.22)

0.89 (0.86, 0.90)

0.43 (0.38, 0.49)

0.72 (0.70, 0.74)

0.60 (0.57, 0.62)

  Pre-specified + data mining

  

0.45

0.20 (0.15,0.24)

0.88 (0.86, 0.89)

0.43 (0.39, 0.48)

0.71 (0.69, 0.73)

0.62 (0.60, 0.65)

 Condition 5: LDL cholesterol ≥70 mg/dL

1,583

80%

      

  Pre-specified

  

0.88

0.19 (0.17, 0.23)

0.86 (0.81, 0.90)

0.86 (0.83, 0.89)

0.21 (0.19, 0.23)

0.59 (0.56, 0.62)

  Pre-specified + data mining

  

0.88

0.25 (0.23, 0.27)

0.85 (0.81, 0.88)

0.91 (0.85, 0.91)

0.23 (0.20, 0.24)

0.60 (0.57, 0.63)

  1. Abbreviations: NPV negative predictive value, PPV positive predictive value.
  2. *Pre-specified Medicare variables: age, sex, race, Medicaid eligible, area-level income, geographic region, evidence of tobacco use, history of hyperlipidemia, history of hypertension, history of diabetes, acute MI, coronary revascularization, history of CHD, history of stroke, history of abdominal aortic aneurysm, history of peripheral arterial disease, history of carotid artery disease, cardiologist care, endocrinologist care, neurologist care, number of evaluation and management visits, hospitalization for any cause, cardiac stress test, echocardiogram, electrocardiogram. For pre-specified variable definitions and a description of the methods used to obtain variables through a data mining procedure see Additional file 1.
  3. For each model, test characteristics are reported for the predicted probability threshold corresponding to an uncorrected specificity of 0.90.
  4. Corrected for optimism using bootstrap resampling [22].