Skip to main content

Table 2 C-statistic from logistic regression models predicting six-month mortalitya

From: Comparison of the performance of the CMS Hierarchical Condition Category (CMS-HCC) risk adjuster with the charlson and elixhauser comorbidity measures in predicting mortality

Sample

Analytic file used to code risk adjustersb

Summary score

Condition indicators

  

CMS-HCC

Charlson

Elixhauser

CMS-HCC

Charlson

Elixhauser

All patients c (N = 170,342)

     
 

1) IPindex

0.799

0.728**

0.680**

0.827

0.743**

0.757**

  

(0.796, 0.802)

(0.725, 0.731)

(0.677, 0.684)

(0.824, 0.830)

(0.740, 0.746)

(0.754, 0.760)

 

2) IPpre12mo+index

0.790

0.739**

0.7**

0.825

0.753**

0.766**

  

(0.787, 0.793)

(0.736, 0.742)

(0.696, 0.703)

(0.823, 0.828)

(0.749, 0.756)

(0.763, 0.769)

 

3) IP+OPpre12mo +index

0.774

0.735**

0.719**

0.819

0.754**

0.777**

  

(0.771, 0.777)

(0.732, 0.738)

(0.716, 0.722)

(0.816, 0.822)

(0.751, 0.757)

(0.774, 0.780)

 

4) IP+OPpre12mo

0.716

0.711**

0.696**

0.745

0.726**

0.739**

  

(0.712, 0.719)

(0.717, 0.714)

(0.692, 0.699)

(0.742, 0.748)

(0.722, 0.729)

(0.736, 0.743)

DM (N = 2,339)

      
 

1) IPindex

0.736

0.706**

0.685**

0.796

0.721**

0.753**

  

(0.712, 0.763)

(0.676, 0.734)

(0.654, 0.712)

(0.786, 0.832)

(0.699, 0.753)

(0.736, 0.787)

 

2) IPpre12mo+index

0.74

0.724**

0.7**

0.796

0.739**

0.758**

  

(0.719, 0.773)

(0.701, 0.755)

(0.679, 0.736)

(0.795, 0.84)

(0.723, 0.777)

(0.749, 0.797)

 

3) IP+OPpre12mo +index

0.733

0.72**

0.699**

NA

NA

0.762

  

(0.712, 0.767)

(0.697, 0.749)

(0.678, 0.733)

  

(0.752, 0.799)

 

4) IP+OPpre12mo

0.703

0.698**

0.682**

NA

0.722**

0.737**

  

(0.681, 0.742)

(0.676, 0.731)

(0.66, 0.719)

 

(0.708, 0.761)

(0.729, 0.78)

Stroke (N = 9,828)

     
 

1) IPindex

0.79

0.667**

0.662**

0.842

0.682**

0.728**

  

(0.776, 0.799)

(0.651, 0.679)

(0.647, 0.675)

(0.833, 0.853)

(0.668, 0.695)

(0.717, 0.742)

 

2) IPpre12mo+index

0.774

0.683**

0.684**

0.841

0.701**

0.736**

  

(0.76, 0.784)

(0.67, 0.696)

(0.672, 0.698)

(0.832, 0.853)

(0.691, 0.716)

(0.725, 0.75)

 

3) IP+OPpre12mo +index

0.752

0.679**

0.693**

0.83

0.712**

0.739**

  

(0.739, 0.763)

(0.665, 0.692)

(0.682, 0.708)

(0.821, 0.842)

(0.699, 0.725)

(0.729, 0.754)

 

4) IP+OPpre12mo

0.688

0.675**

0.679**

0.718

0.706**

0.709**

  

(0.677, 0.703)

(0.662, 0.689)

(0.668, 0.695)

(0.712, 0.737)

(0.693, 0.719)

(0.701, 0.726)

AMI (N = 5,749)

      
 

1) IPindex

0.777

0.692**

0.678**

0.811

0.704**

0.749**

  

(0.763, 0.789)

(0.677, 0.707)

(0.661, 0.693)

(0.801, 0.826)

(0.69, 0.721)

(0.736, 0.764)

 

2) IPpre12mo+index

0.776

0.71**

0.705**

0.814

0.719**

0.75**

  

(0.763, 0.789)

(0.695, 0.725)

(0.691, 0.721)

(0.805, 0.83)

(0.706, 0.736)

(0.738, 0.767)

 

3) IP+OPpre12mo +index

0.77

0.713**

0.727**

0.813

0.725**

0.763**

  

(0.757, 0.783)

(0.699, 0.729)

(0.714, 0.742)

(0.805, 0.829)

(0.713, 0.743)

(0.752, 0.78)

 

4) IP+OPpre12mo

0.72

0.707**

0.712**

0.741

0.717**

0.732**

  

(0.706, 0.737)

(0.693, 0.723)

(0.698, 0.728)

(0.736, 0.764)

(0.705, 0.735)

(0.721, 0.751)

CHF(N = 11,287)

 

1) IPindex

0.685

0.653**

0.642**

0.719

0.669**

0.7**

  

(0.675, 0.697)

(0.642, 0.665)

(0.631, 0.654)

(0.712, 0.733)

(0.659, 0.681)

(0.69, 0.712)

 

2) IPpre12mo+index

0.694

0.671**

0.668**

0.725

0.681**

0.709**

  

(0.685, 0.707)

(0.661, 0.684)

(0.658, 0.681)

(0.719, 0.739)

(0.673, 0.695)

(0.701, 0.723)

 

3) IP+OPpre12mo +index

0.69

0.666**

0.671**

0.721

0.679**

0.702**

  

(0.682, 0.704)

(0.656, 0.679)

(0.662, 0.685)

(0.715, 0.736)

(0.671, 0.693)

(0.695, 0.716)

 

4) IP+OPpre12mo

0.674

0.663**

0.664**

0.698

0.675**

0.688**

  

(0.665, 0.689)

(0.654, 0.677)

(0.655, 0.679)

(0.694, 0.715)

(0.667, 0.69)

(0.682, 0.704)

  1. Note: **: The c-statistic is significantly different from the c-statistic of CMS-HCC model at the 5% level;
  2. NA: Models did not converge due to complete (or quasi-complete) separation.
  3. a. 95% confidence intervals of c-statistics are in parentheses.
  4. b. Four analytic files were used to code risk adjusters: "IPindex" only included claims from the index hospitalization for each of the four disease samples. "IPpre12mo+index" included all inpatient claims in the 12 months prior to the index hospitalization including the index admission date. "IP+OPpre12mo+index" included all inpatient and outpatient claims in the 12 months prior to the index hospitalization including the index admission date. "IP+OPpre12mo " included all inpatient and outpatient claims in the 12 months prior to the index hospitalization excluding the index admission date.
  5. c. Patients with any hospital admission between January 1st, 2006 and June 30th, 2006.
Back to article page

BMC Health Services Research

ISSN: 1472-6963

Contact us