1 of 12

TMLE Analysis Using MIMIC

​

Yiren Cao

2 of 12

Ventilation and RRT increased mortality in low SOFA group patients; �no such effect was observed for vasopressor

0.035

(0.025, 0.045)

0.015

(0.001, 0.30)

-0.106

(-0.149, -0.064)

-0.136

(-0.259, -0.013)

0.095

(0.030, 0.160)

0.077

(0.041, 0.112)

0.037

(0.001, 0.074)

-0.072

(-0.140, -0.004)

0.003

(-0.007, 0.013)

-0.003

(-0.019, 0.012)

0.077

(0.024, 0.129)

-0.184

(-0.364, -0.003)

Note: Used TMLE for counterfactual average treatment effect analysis;

Controlling for age, gender, race (white/non-white), SOFA score, CCI group, year of admission and the other two invasive interventions

SOFA

3 of 12

By Year Analysis

4 of 12

Though mortality and ventilation usage increased, we did not find ventilation has different impact on mortality over time among less severe patients

SOFA

2008 - 2010

2011 - 2013

2014 - 2016

2017 - 2019

0-5

0.035

(0.020, 0.050)

0.032

(0.011, 0.052)

0.042

(0.021, 0.063)

0.036

(0.004, 0.067)

​

2008 - 2010

2011 - 2013

2014 - 2016

2017 - 2019

Ventilation Rate

29.1%

31.3%

36.1%

30.4%

Mortality Rate

7.2%

8.0%

8.3%

11.6%

+ Used TMLE for counterfactual average treatment effect analysis; controlling for age, gender, race (white/non-white), SOFA score, CCI group, RRT, and vasopressor

5 of 12

We derived similar findings in other SOFA groups and other invasive interventions including RRT and vasopressor

Ventilation

SOFA 0-5

6-10

11-15

>15

RRT

Vasopressor

Note: Used TMLE for counterfactual average treatment effect analysis; controlling for age, gender, race (white/non-white), SOFA score, CCI group, and the other two invasive interventions

6 of 12

By Race Analysis

7 of 12

White patients were less likely to receive ventilation and RRT, but more likely to receive vasopressor in less severe SOFA groups; no race difference was found in the most severe patients

Odds Ratio

0 - 5

6 - 10

11 - 15

>15

Across SOFA Groups

Ventilation

0.788*

0.863*

0.708*

0.767

0.814*

RRT

0.542*

0.583*

0.845*

1.079

0.656*

Vasopressor

1.323*

1.364*

1.090

1.098

1.321*

Likelihood of Receiving Invasive Intervention as a White Patient

*Significant at 0.05 level; Logistic regression analysis, controlling for age, gender, race (white/non-white), SOFA score, CCI group, year of admission and the other two invasive interventions

8 of 12

Being white or non-white, ventilation did not have significantly different impact on mortality across all SOFA groups

0.051

(0.032, 0.070)

0.029

(0.017, 0.041)

0.044

(0.018, 0.070)

0.005

(-0.012, 0.023)

-0.099

(-0.171, -0.026)

-0.117

(-0.168, -0.065)

0.019

(-0.169, 0.208)

-0.164

(-0.329, 0.001)

Note: Used TMLE for counterfactual average treatment effect analysis;

Controlling for age, gender, race (white/non-white), SOFA score, CCI group, year of admission, RRT, and vasopressor

9 of 12

We derived similar findings in other invasive methods including RRT and vasopressor

Ventilation

SOFA 0-5

6-10

11-15

>15

RRT

Vasopressor

Note: Used TMLE for counterfactual average treatment effect analysis; controlling for age, gender, race (white/non-white), SOFA score, CCI group, year of admission, and other two interventions

10 of 12

The average treatment effect of ventilation on mortality was similar among white and non-white patients; Ventilation had a significant impact on mortality had everyone been non-white (or white)

ATE

Non-Ventilation

Ventilation

Non-white

0.136

(0.127, 0.146)

0.283

(0.271, 0.295)

White

0.129

(0.123, 0.135)

0.217

(0.209, 0.226)

Non-white

White

ATE of Ventilation on Mortality

by Race and Ventilation+

+ Among all patients

Note: Used LTMLE for counterfactual average treatment effect analysis; controlling for age, gender, SOFA score, CCI group, year of admission, RRT, and vasopressor

11 of 12

We derived similar findings among SOFA 0-5 patients: Ventilation had a significant impact on mortality had everyone been non-white (or white)

ATE

Non-Ventilation

Ventilation

Non-white

0.078 

(0.068, 0.087)

0.123 

(0.107, 0.140)

White

0.071 

(0.065, 0.076)

0.092 

(0.082, 0.102) 

Non-white

White

ATE of Ventilation on Mortality

by Race and Ventilation (SOFA 0-5)

Note: Used LTMLE for counterfactual average treatment effect analysis; controlling for age, gender, SOFA score, CCI group, year of admission, RRT, and vasopressor

12 of 12

The effect of ventilation on mortality was less different among severe patients (SOFA >10)

​

�

Non-Ventilation

Ventilation

SOFA 0-5

Non-white

0.078 

(0.068, 0.087)

0.123 

(0.107, 0.140)

White

0.071 

(0.065, 0.076)

0.092 

(0.082, 0.102) 

�

​

�

Non-Ventilation

Ventilation

SOFA 10-15

Non-white

0.512

 (0.444, 0.581) 

0.447

 (0.420, 0.473)  

White

0.500  

(0.453, 0.547)

0.391 

 (0.370, 0.413) 

​

�

Non-Ventilation

Ventilation

SOFA 6-10

Non-white

0.188

 (0.169, 0.206) 

0.239 

(0.221, 0.257) 

White

0.200 

(0.187, 0.212)

0.193 

(0.181, 0.205) 

�

​

�

Non-Ventilation

Ventilation

SOFA >15

Non-white

0.667 

(0.428, 0.905)  

0.663  

(0.612, 0.714)  

White

0.800 

(0.6431, 0.957)   

0.622 

(0.574, 0.670) 

ATE of Ventilation on Mortality by Race and Ventilation

Note: Used LTMLE for counterfactual average treatment effect analysis; controlling for age, gender, SOFA score, CCI group, year of admission, RRT, and vasopressor