How to Detect and Handle Confounding Factors

Abstract How to Detect and Handle Confounding Factors Sirima         Mongkolsomlit     BSc, MSc Epidemiology* Jayanton    Patumanond       MD, DTM\&H, MSc Clin Trop Med, DSc Clinical Epidemiology** Petch           Rawdaree             MD, MSc Epidemiology, DLSHTM*** * Faculty of Public Health, Thammasart University. ** Faculty of Medicine, Chiangmai University. *** Department of Medicine, BMA Medical College and Vajira Hospital.                 When a researcher investigates the association between an exposure of factors and an occurrence of outcome, there are other influencing factors to the test result so called "confounding factors". Confounding factors may cause indistinct or inaccurate results of the study. Since statistical analysis cannot clear out the effect of confounders which have not been collected or being unidentified, it is crucial for the researcher to identify and control the impact of these confounding factors. In order to collect all essential data including any potential confounding factors, a thorough literature review prior to conducting a research is the first and important step. If confounders are identified, there are many approaches to deal with confounders: by randomization, restriction, or matching in the research design or methodology process; or by stratification or multivariable analysis in the statistical analysis process.   Keywords: confounding factors, randomization, restriction, matching, stratification   Vajira Med J 2010 ; 54 : 223-235

Predictors of Early Stages of Smoking Uptake among Thai Male Adolescents

Smoking uptake is a complex behavioral process comprised of several stages and remains a major public health problem, especially among Thai adolescents. Specific intrapersonal, attitudinal and social factors may function differently at various stages of smoking uptake. Thus, this study of 1,012 predominantly Buddhist Thai male secondary school students, who were living with their parents and had an average age 12.72 years, aimed to: identify the prevalence of various early stages of smoking uptake , examine predictors of various early stages of smoking uptake, and examine predictors of progression from one early stage of smoking uptake to another . Most were in the non-susceptible pre-contemplation stage, followed by the stages of initiation/tried, susceptible pre-contemplation, experimentation/addiction, and contemplation/preparation. Predictors of the susceptible pre-contemplation stage were: prevalence estimate, attitude toward smoking, parental approval of smoking and parental smoking. Tried stage predictors included: offers of smoking, attitude towards smoking, peer smoking and level of academic success. Predictors of the experimentation/addiction stage involved: attitude toward smoking, offers of smoking, peer smoking, parental smoking and level of academic success. Offers of smoking and parental approval of smoking were factors influencing advancement from the susceptible pre-contemplation stage to the initiation/tried stage, while peer smoking and attitude toward smoking predicted transition from the initiation/tried stage to the experimentation stage. Since only two students were in the contemplation/preparation stage, made the number was too small to demonstrate any significant findings, no predictors of this stage were calculated. The findings may prove useful in developing primary prevention smoking programs for Thai male adolescents.

GESTATIONAL DIABETES IN LUMPHUN HOSPITAL: PREVALENCE, CLINICAL RISK FACTORS AND PREGNANCY OUTCOMES

Objective: To determine prevalence of gestational diabetes mellitus (GDM), clinical risk factors and pregnancy outcomes of pregnant women at Lumphun hospital. Study design: Retrospective descriptive study. Materials and methods: Part 1; Medical records of 637 women attending antenatal care at Lumphun hospital between July 2006 and May 2007 were reviewed. These were divided into 2 groups as with and without GDM according to individual risk factors, 50 gm glucose challenge test (GCT) and 100 gm oral glucose tolerance test (OGTT). Clinical risk factors and plasma glucose of GCT were compared between two groups. Part 2; 377 women who delivered during July 2006 to May 2007 were recruited and classified into 4 groups. Those without risk factors, those with risk factors and normal GCT, those with abnormal GCT and normal OGTT, and those with abnormal GCT and OGTT were defined as group 1, 2, 3 and 4 respectively. Pregnancy outcomes were compared among these groups. Results: Part1; The prevalence of GDM in Lumphun hospital is 1.5% (10 from 637 cases). False positive rate of GCT is high as 75% (31 from 41 cases). Pregnant women with GDM are more likely to have higher plasma glucose from GCT than those without GDM. If the result of GCT is higher than 180 mg/dL, risk of GDM is increased significantly (P-value 0.006) and pregnant women with GDM are more likely to be obese (P-value 0.047). Moreover, if the pregnant women had at least 3 risk factors, they are more likely to be GDM (P-value 0.003) Part2; When compared with group 1, risk of cesarean delivery, infant’s head circumference, maternal and neonatal complications are increased significantly in group 2, 3 and 4 pregnant women (P-value 0.015, 0.018, 0.001 and 0.001 respectively). Conclusion: Pregnancy with GCT higher than 180 mg/dL, obesity or has at least 3 risk factors has high possibility to be GDM. Moreover, pregnancy with glucose intolerance or GDM has increase risk of adverse pregnancy outcomes.