Skip to main content

Table 2 Fixed effect estimates, summarized by the mean (standard error) from Bayesian and h-lik (as implemented in this paper and the iGLM algorithm) analyses for the 1% dataset with grouping at the facility level

From: High performance implementation of the hierarchical likelihood for generalized linear mixed models: an application to estimate the potassium reference range in massive electronic health records datasets

Fixed Effects

Bayesian (MCMC)

h-lik

iGLM

Intercept

−6.214 (0.961)

−6.218 (0.994)

− 6.168 (0.992)

Potassium (natural spline coefficient 1)

8.062 (1.979)

6.808 (2.462)

6.805 (2.461)

Potassium (natural spline coefficient 2)

−38.03 (31.12)

−9.491 (40.24)

−10.00 (40.23)

Potassium (natural spline coefficient 3)

−75.33 (61.7)

−18.50 (79.78)

−19.51 (79.76)

eGFR (natural spline coefficient 1)

−0.406 (0.255)

− 0.43 (0.251)

− 0.435 (0.251)

eGFR (natural spline coefficient 2)

−1.872 (0.492)

−1.839 (0.505)

− 1.839 (0.505)

eGFR (natural spline coefficient 3)

−1.141 (0.502)

− 1.091 (0.509)

−1.089 (0.509)

Age (natural spline coefficient 1)

0.305 (0.332)

0.309 (0.329)

0.306 (0.329)

Age (natural spline coefficient 2)

−0.086 (0.938)

−0.227 (0.913)

− 0.227 (0.913)

Age (natural spline coefficient 3)

0.924 (0.199)

0.906 (0.198)

0.902 (0.198)

Male Gender

0.214 (0.093)

0.214 (0.093)

0.214 (0.093)

White Race

0.155 (0.143)

0.158 (0.146)

0.160 (0.145)

Hispanic Race

−0.745 (0.853)

− 0.457 (0.723)

−0.452 (0.723)

Native American

−0.233 (0.829)

0.031 (0.731)

0.040 (0.730)

Other Race

−0.618 (0.321)

−0.580 (0.323)

− 0.571 (0.322)

Unknown Race

−0.347 (0.488)

− 0.242 (0.472)

−0.238 (0.471)

Inpatient status

2.404 (0.198)

2.390 (0.194)

2.397 (0.194)

Charlson Comorbidity Score

0.098 (0.014)

0.099 (0.014)

0.099 (0.014)