Problem Description
This SAS assignment explores the associations between diseases and country of origin, using data analysis techniques. We investigate two distinct medical conditions, Extrapulmonary Tuberculosis (EPTB) and Tuberculosis (TB) diagnosis in prison. The study aims to identify key factors influencing these conditions and understand the impact of various covariates.
Table 1: Sites of EPTB Association with Covariates
| Characteristics | CNS/Meningeal (N = 67) | Pleural (N = 42) | OR (95% CI) |
|---|---|---|---|
| Length of Hospital Stay (LOS) Median (IQR) | 14 (6 – 30) | 8.5 (8.63 – 20.08) | 20.20 (15.55 – 24.85) |
| Drug Use | No: 41 (62) Yes: 21 (62) | No: 24 (37) Yes: 13 (37) | REF: 0.946 (0.402 – 2.225) |
| DM (Diabetes Mellitus) | No: 64 (65) Yes: 1 (65) | No: 38 (39) Yes: 1 (39) | REF: 0.745 (0.182 – 3.048) |
| ARV (Anti-retroviral) | No: 44 (51) Yes: 7 (51) | No: 8 (8) Yes: 0 (8) | REF: 0.846 (0.754 – 0.950) |
| Homeless | No: 53 (65) Yes: 12 (65) | No: 28 (38) Yes: 10 (38) | REF: 0.761 (0.440 – 1.314) |
| HIV | Negative: 25 (67) Positive: 42 (67) | Negative: 33 (42) Positive: 9 (42) | REF: 6.16 (2.535 – 14.969) |
Key Findings:
- The analysis explores the association between EPTB sites and covariates, emphasizing the impact of factors like drug use, diabetes mellitus, and ARV treatment.
- Notably, drug use has the strongest association with EPTB sites, with an odds ratio of 0.946.
- Diabetes mellitus has the lowest association, suggesting a protective effect on EPTB.
SAS CODE:
proc freq data=WORK.TABLE1 order=data;
tables (LOS DrugUse DM ARV HIV Homeless)*XPSITE/nopercent norow nocol oddsratio;
run;
proc means data=WORK.TABLE1 chartype median vardef=df clm alpha=0.05 q1 q3 qmethod=os;
var LOS ARV DM Homeless HIV DrugUse;
class XPSITE;
run;
Table 2: TB Diagnosis in Prison and Association with Covariates
| Characteristics | No (N = 85) | Yes (N = 19) | Assumption (Diagnosed) | OR (95% CI) |
|---|---|---|---|---|
| CD4ADM Median (IQR) | 105 (35 – 244) | 131 (25.5 – 192) | - | 155.84 (99.45 – 212.23) |
| Country | US: 43 (55) Foreign: 12 (55) | US: 10 (0) Foreign: 1 (2) | - | - |
| Drug Use | No: 39 (54) Yes: 15 (54) | No: 2 (8) Yes: 6 (8) | REF: 0.1282 (0.0232 – 0.7071) | - |
| Alcohol Abuse | No: 32 (55) Yes: 23 (55) | No: 1 (8) Yes: 7 (8) | REF: 0.1027 (0.0118 – 0.8928) | - |
| Previous TB | No: 40 (47) Yes: 7 (47) | No: 7 (9) Yes: 2 (9) | REF: 1.6327 (0.2796 – 9.5348) | - |
Key Findings:
- This analysis focuses on the association between TB diagnosis in prison and various covariates.
- It highlights the impact of factors like drug use, alcohol abuse, and previous TB history.
- Notably, individuals born in the United States are more likely to have TB diagnosis in prison.
SAS CODE:
proc freq data=WORK.TABLE2 order=data;
tables (DrugUse AlcoholAbuse CD4ADM PreviousTB COUNTRY)*PRISON/nopercent norow nocol oddsratio;
run;
proc means data=WORK.TABLE1 chartype median vardef=df clm alpha=0.05 q1 q3 qmethod=os;
var DrugUse AlcoholAbuse CD4ADM PreviousTB COUNTRY;
class PRISON;
run;
Table 3: DAG Model Analysis - HIV and Other Covariates
| Participant Characteristics | Crude OR | Adjusted OR* |
|---|---|---|
| HIV Status | Negative: 0.1020 Positive: 0.1491 | Negative: 0.0891 Positive: 0.1428 |
| HIV Status** | Negative: 0.0971 Positive: 0.2500 | Negative: 0.0813 Positive: 0.1821 |
Key Findings:
- The study analyzes Directed Acyclic Graph (DAG) models to understand the relationship between HIV status and other covariates.
- It discusses the presence of colliders and their impact on the association between HIV and EPTB.
- Notably, the analysis indicates a weak and nonsignificant association between EPTB and HIV.
SAS CODE:
proc causalgraph
model “HIV”
HIV ==> ARV LOS DrugUse DM Homeless XPSITE,
LOS ==> DM Homeless,
ARV ==> HIV;
unmeasured AGE AlcoholAbuse CD4ADM COPU COUNTRY DEATH;
identify HIV ==> XPSITE;
run;
Table 4: DAG Model Analysis - Prison and Country of Origin
| Participant Characteristics | Crude OR | Adjusted OR* |
|---|---|---|
| Country | US: 0.2421 FOREIGN: 0.5130 | US: 0.2100 FOREIGN: 0.5021 |
| Country** | US: 0.124 FOREIGN: 0.1202 | US: 0.1018 FOREIGN: 0.1187 |
Key Findings:
- The analysis continues to use DAG models to study the association between prison and country of origin.
- It discusses strategies for model building and the impact of assumptions on the results.
SAS CODE:
proc causalgraph
model “PRISON”
PRISON ==> COUNTRY DrugUse CD4ADM,
DrugUSE ==> PRISON AlcoholAbuse;
unmeasured AGE ARV LOS DrugUse DM Homeless XPSITE;
identify HIV ==> XPSITE;
run;
This assignment exemplifies the use of SAS for in-depth data analysis and presents key findings related to disease associations and the impact of covariates on the studied outcomes.