Table 6

Model comparisons for death in age-specific diabetes cohorts
Cohort 1a (n = 29,058) Cohort 2 (n = 41,925)
Model (df)b c-statisticc(95% CI) cd(%) Brier score LRTf c-statisticc(95% CI) c (%) Brier score LRTf
20-44 years
Base (7) 0.65 (0.56, 0.74) . 0.01 . 0.64 (0.57, 0.71) . 0.01 .
# diagnoses (10) 0.74 (0.65, 0.83) 0.09 (13.02) 0.01 14.1* 0.75 (0.69, 0.81)* 0.11 (16.82) 0.01 43.1*
Charlson (10) 0.81 (0.74, 0.89)* 0.16 (24.20) 0.01 50.1* 0.80 (0.74, 0.86)* 0.16 (24.61) 0.01 87.9*
Elixhausere -- -- -- -- 0.72 (0.66, 0.79)* 0.08 (12.77) 0.01 36.2*
# drugs (10) 0.72 (0.64, 0.80) 0.07 (10.26) 0.01 9.0 0.71 (0.65, 0.78)* 0.07 (11.06) 0.01 21.1*
CDS -- -- -- -- 0.70 (0.63, 0.77) 0.06 (9.35) 0.01 19.0*
45-64 years
Base (7) 0.62 (0.59, 0.65) . 0.04 . 0.62 (0.59, 0.64) . 0.03 .
# diagnoses (10) 0.72 (0.69, 0.74)* 0.10 (15.81) 0.04 149.0* 0.69 (0.67, 0.72)* 0.07 (12.85) 0.03 149.0*
Charlson (10) 0.76 (0.74, 0.79)* 0.14 (22.74) 0.04 345.0* 0.74 (0.72, 0.76)* 0.12 (20.33) 0.03 368.2*
Elixhauser (35) 0.77 (0.74, 0.79)* 0.15 (23.71) 0.04 416.7* 0.75 (0.73, 0.77)* 0.13 (21.95) 0.03 416.7*
# drugs (10) 0.60 (0.66, 0.71)* 0.08 (10.97) 0.04 103.5* 0.68 (0.65, 0.70)* 0.06 (9.76) 0.03 103.5*
CDS (11) 0.66 (0.63, 0.69) * 0.04 (6.29) 0.04 60.1* 0.65 (0.62, 0.68)* 0.03 (5.69) 0.03 60.1*
65-74 years
Base (7) 0.60 (0.58, 0.62) . 0.09 . 0.59 (0.57, 0.61) . 0.08 .
# diagnoses (10) 0.68 (0.66, 0.70)* 0.08 (13.88) 0.09 186.8* 0.66 (0.64, 0.68)* 0.07 (12.05) 0.08 186.8*
Charlson (10) 0.73 (0.71, 0.75)* 0.13 (22.24) 0.08 407.0* 0.71 (0.69, 0.73)* 0.12 (20.03) 0.07 407.0*
Elixhauser (35) 0.75 (0.73, 0.76)* 0.15 (24.58) 0.08 548.8* 0.72 (0.70, 0.74)* 0.13 (22.24) 0.07 548.8*
# drugs (10) 0.67 (0.65, 0.69)* 0.07 (11.71) 0.09 131.1* 0.65 (0.63, 0.67)* 0.06 (9.68) 0.08 131.1*
CDS (11) 0.66 (0.64, 0.68)* 0.06 (9.70) 0.09 63.0* 0.62 (0.60, 0.64)* 0.03 (4.92) 0.08 63.0*
75+ years
Base (7) 0.64 (0.62, 0.66) . 0.16 . 0.67 (0.65, 0.68) . 0.16 .
# diagnoses (10) 0.67 (0.66, 0.69)* 0.03 (4.84) 0.16 159.7* 0.69 (0.68, 0.70)* 0.02 (3.60) 0.16 159.7*
Charlson (10) 0.70 (0.68, 0.71)* 0.06 (8.91) 0.16 315.3* 0.71 (0.70, 0.72)* 0.04 (6.76) 0.15 318.0*
Elixhauser (35) 0.72 (0.70, 0.73)* 0.08 (12.34) 0.15 633.5* 0.74 (0.73, 0.75)* 0.07 (11.11) 0.15 633.5*
# drugs (10) 0.68 (0.66, 0.69)* 0.04 (5.47) 0.16 161.2* 0.69 (0.68, 0.70)* 0.02 (3.45) 0.16 161.2*
CDS (11) 0.67 (0.65, 0.68)* 0.03 (4.22) 0.16 95.1* 0.68 (0.67, 0.69)* 0.01 (1.95) 0.16 95.1*

aCohort 1 was defined using data from fiscal years 1996/97 to 1997/98 and Cohort 2 was defined using data from 1996/97 to 2001/02; bBase model includes age, age2, sex, income quintile, geography, and recent diabetes diagnosis, and full models include these variables in addition to the specified comorbidity measure; cA c-statistic with * is significantly different (p < .01) from the c-statistic for the base model; dc = c-statistic for full model minus c-statistic for base model; eSome full models failed to converge due to the small number of deaths; fA likelihood ratio test (LRT) with * is statistically significant at α = .01; df=degrees of freedom; CDS Chronic Disease Score.

Lix et al.

Lix et al. BMC Health Services Research 2013 13:340   doi:10.1186/1472-6963-13-340

Open Data