Order Now  +1 678 648 4277 

Understanding multivariable analysis

Multivariable analysis is used in theinvolvementin the analysis and observation of several variables in a time. Multivariate analysis is used to address situations where different measurements are combined until the relationship among the measurements is important. MVA is always complicated by the need to include physics analyses in calculating the effects of variables.

Table Of Contents
  • Hazards Regression
  • Survival
  • Log-rank test
  • Atopic dermatitis and osteoporosis
  • Survival Analysis
  • Code for Survival
  • Information on Study Data, including Data Dictionary:

Hazards Regression

Objective – Investigating the risk of osteoporosis risk in atopic dermatitis patients.
Exposure – atopic dermatitis (AD).
Outcome – diagnosis with osteoporosis.
Potential confounders – Diabetes mellitus, hypertension, chronic liver disease, chronic kidney disease, depression, chronic obstructive pulmonary disease.

Survival

Survival is determined by time until the occurrence of hospitalization, the end of the study period, or an ambulatory visit for osteoporosis whichever of the three comes first.
The authors adjust the regression models to account for censoring. This ensures that all information of censored people when a patient is censored is utilized.

Log-rank test

The atopic dermatitis incident curve is higher than for the control cohort. In the control cohort, log-rank test P<0.001, which indicates the difference between survival curves of 2 cohorts.
Kaplan-Meier curve in Figure 2 shows the cumulative incidence of osteoporosis between patients with and without dermatitis incident after following up for 15 years.

Atopic dermatitis and osteoporosis

We conclude there is a true relationship between atopic dermatitis and osteoporosis. All general incidence of AD was 4.72 times greater in the atopic dermatitis group after adjusting for possible puzzling factors in the control group.

Survival Analysis

 

Time Intervals

N Alive

NDied

Proportion Died

Proportion Survived

Cumulative Survival

6 months

169

22

0.1152

0.8848

0.8848

12 months

144

17

0.1056

0.8944

0.7914

24 months

87

31

0.2627

0.7373

0.5835

36 months

36

10

0.2174

0.7826

0.4566

Understanding multivariable analysis

Understanding multivariable analysis

Log-rank test

P <0.0001, indicates a significant difference in the patient's survival in the 4 clinical stages.

Clinical Stage Survival at 6 Months Survival at 12 Months Survival at 24 Months Survival at 36 Months Median Survival
Stage 1 0.9327 0.9186 0.8376 0.7908  
Stage 2 0.8980 0.7906 0.5124 0.3933 25.4600
Stage 3 0.8696 0.7391 0.5593 0.4187 31.9700
Stage 4 0.6667 0.4000 0.0667 0 7.2900

Yes, this agrees with the plot and statistical test. The survival rates are different for patients in various clinical stages.
Understanding multivariable analysis

Stage 2 – 3.739. Death risk is 3.739 times higher in Stage 2 compared to Stage 1.

Stage 3 – 3.659. The risk of death is 3.659 times higher in Stage 3 compared to Stage 1.

Stage 4 – 13.851. The risk of death is 13.851 times higher in Stage 4 compared to Stage 1.

This agrees with the previous findings.

Understanding multivariable analysis

Stage 2 – 3.848. The risk of death is 3.848 times higher in Stage 2 compared to Stage 1.

Stage 3 – 4.115. The risk of death is 4.115 times higher in Stage 3 compared to Stage 1.

Stage 4 – 15.893. The risk of death is 15.893 times higher in Stage 4 compared to Stage 1.

The risk of death is 1.070 times higher among patients who don’t have insurance covers.

The risk of death is 1.459 times higher for Black patients.

The risk of death is 1.436 times higher for Hispanic patients.

Increase in the Annual hospital patient volume by increasing the risk of death by 0.981.

The confounders increase the risk ratios between clinical stage and time to mortality.

Code for Survival

*Upload the dataset and run proc import statement*;

*Kaplan-Meier Method for time to mortality; 

ods graphics on;

proclifetest data=work.mortality plots=s(atrisk=0 to 42 by 6);

timedx_lastcontact_death_months*mortality(0); label dx_lastcontact_death_months='Months from Diagnosis';

run;

ods graphics off;

*Kaplan-Meier Method for time to mortality by clinical stage (1, 2, 3, and 4); 

ods graphics on;

proclifetest data=work.mortality plots=s(nocensoratrisk=0 to 36 by 6) maxtime=36;

timedx_lastcontact_death_months*mortality(0); label dx_lastcontact_death_months='Months from Diagnosis';

strataClin_Stage; label Clin_Stage='Clinical Stage';

run;

ods graphics off;

*Univariate Cox Proportional Hazards Model for Clinical Stage*; 

procphreg data=work.mortality; class clin_stage(ref='1') /param=reference;

modeldx_lastcontact_death_months*mortality(0) = clin_stage /rl;

run;

*Multivariable Cox Proportional Hazards Model for Clinical Stage*; 

procphreg data=work.mortality; class clin_stage(ref='1') /param=reference;

modeldx_lastcontact_death_months*mortality(0) = clin_stagenoinsur black hispanic volume /rl;

run;

Information on Study Data, including Data Dictionary:

NAME: Cancer Mortality Data (Mortality.csv)

SIZE: 200 observations, 8 variables

DATA DICTIONARY:

Variable Label Variable
ID Patient ID (1-200)
Dx_LastContact_Death_Months Months from cancer diagnosis to death or date of the last follow-up
CLIN_Stage Clinical Stage (1-4)
Volume Annual hospital patient volume
Noinsur Patient insurance status (0=no insurance, 1=insurance)
Black Black race (0=no, 1=yes)
Hispanic Hispanic ethnicity (0=no, 1=yes)
Mortality Survival status (0=alive, 1=dead)