-
A review on some techniques to control the effects from confounding factor
-
Seung Wook Lee
-
Korean J Epidemiol. 1990;12(1):9-23.
-
-
-
 Abstract
 PDF
-
Abstract
This paper is to review some methods available to control the effect from confounding factors when we conduct epidemiological studies.
Basically, there are two approaches: 1) matching before data collection, and 2) statistical analysis after data collection.
For matching, caliper matching, nearest available matching, stratified matching, frequency matching, and mean matching as univariate matching method are reviewed. Also, multivariate caliper matching, multivariate stratified matching, minimum distance matching, discriminant matching and linear adjustment matching as multivariate matching methods are explained.
The standardization and stratification are introduced as methods for after-data-collection. Also various statistical techniques including log-linear method, logistic regression method, multiple regression method, and covariance analysis are illustrated for possible use in data analysis with confounding variables.
Finally, analysis techniques for matched data with controls more than single are explained.
We reviewed all theses techniques briefly, mainly on the logical meaning. Also, we reviewed the statistical techniques that will be used for analysis after the matched data are obtained.
-
Summary
-
A study on health impact of a river pollution neighboring industrialized big city on the rural people residing along the river
-
Joung Soon Kim, Seung Wook Lee, H.S Yoon, Y Heo, S.J Yang, T.W Ha, H.K Hong, Y.W Lee, H.H Kwon, D.H Lee, H.C Kim
-
Korean J Epidemiol. 1986;8(1):37-95.
-
-
-
Abstract
PDF
-
Abstract
This study is conducted to identify any adverse health impact for the rural residents living along a certain river in Korea. The river actually accomodates all the industriral and domestic waste-water discharged from its neighboring big city. Unfortunately, the river is being served for farming water residents, while the current water cleaning system does not function satisfactorily. It means a possibility that the polluted water can contaminate underground water which is the main source of drinking water in the area.
The aim of the study is to determine the contamination level of heavy metals and evaluate the health impact of the residents with a linkage to the river pollution. The study is proceeded under the following steps.
1) The area is divided into two parts where the water is used for the farming or not. They formed study and control area. For the residents within the sites, study and control groups were formed.
2) Between the two, any statistical significance was pursued for the following items,
(1) the contamination levels by the metals in drinking water source, soil, farming products, farming products, and fishes which live in the water where the river joins the main river,
(2) the contamination levels of rural residents, blood samples, and
(3) by physicians the health evaluation results by clinical and pathological tests for the residents.
3) the correlation between the blood contamination levels and the health results,
4) the correlation between the blood and environmental pollution levels were studied.
In spite of several limitations from the nature of the study,
The findings were summarized as follows：
1) No statistical significance was found in the differences between the groups with regard to age, sex, occupation, economic status, education levels, pregnancy history, number of household members, and medical insurance status.
2) No statistical significance were found with regard to smoking habit, proportions of drinking, boiled water, number of days using pesticides and number of pregnancy poisoning, frequency of monthly meat intake. But a significance was found in the number of fish intake, while almost nobody took fishes from the river.
3) Most of the residents(93%) think the river is harmful to their health and 34% among them addressed that the river contamination is the cause of various dermatitis.
4) Stillbirth rate was significantly higher in the study group, but no correlation was found with the metal contamination.
5) The study group showed significantly higher complaints for 24 symptoms believed to appear with metal contamination, but no correlation was found. No correlation was found when the symptoms are grouped into 17 disease classification, either.
6) Neurological symptoms and clinical screening tests did not show any significant difference between the groups. Past disease history showed the same thing.
7) The results of diagnoses by physicians are classified into three groups from normal to abnormal. The control groups had higher proportion of the normal, but conjunctiva abnormaly, unspecified digestive ulcer, symptoms of abdomen and pelvic area, gasritis and duodenum inflammation were more frequent in the control group. Neurologic findings by a specialist did not show any difference.
8) The values of laboratory tests for hematopoietic system, kidney, and liver functions as indices of health measurements didn’t show any significance.
9) For the level of heavy metal contamination in blood, those of Cu, Cd, Pb but Zn were statistically higher than those determined by the National Envoronmental Protection Institute for the Yang-pyong-kun residents. For Pb, Zn, Ni at 0.1% and Hg at 5%, the values from the control groups was rather higher than those in the study group. However, at was seen that the level of Cd of the study group was higher than the control groups at 5%. They failed to show any consistent pattern, but the males has higher value than the females for Cd, Zn, Hg and lower value for Cu, Pb with statistical significance.
10) The proportions of residents above the permissible levels for the metal contamination were determined. The proportions for Pb and Zn showed that those for the control group were statistically higher than those for the study group. They were 42.3% for Pb and 6.3% for Zn, respectively, in the control group. For the case group, they were 19.3% for Pb and 4.6% for Mn, respectively. The proportions of residents who had abnormal level for more than 2 kinds of heavy meatals was 10.2% and for more than 3 kinds 0.8%.
11) The health evaluation results were linked with the classification of metal levels into the normal and abnormal groups. The abnormal group in Mn level seemed to have high proportion of lassitude, but its reliability was very low because of the sample size of 1. Also, the proportion of no symptoms was significantly higher in the group abnormally contaminated by Pb and Zn rather in the normal group.
The clinical evaluation results showed that the Pb-and Zn-high groups had more frequently hearing difficulty, the Pb-high conjuntictiva problem, the Zn-high group neurotoxic disturbance and hearing difficulty. The neurological findings showed that inflammation, neurotoxic disturbance and hearing difficulty in the Mn-and Pb-high group, respectively. Also, various pathological tests did not show any significance.
12) The water quality was examined for the river and drinking water. The river seemed to be polluted by not heavy metal but various organic materials. This can be backed up by high levels of COD, BOD, and DO. The drinking water classified into underground water, well and simple-lined pump didn’t bear any significant difference from that in the control area and was believed to be generally appropriate for drinking. In the simple-lined pump and well water, the levels of Cu, Pb, and Ni were slightly higher than in the foot site river. It needs further attention to find Hg, As, Cr in the well water which did not exist in the river.
13) The sample of fishes to determine heavy metal contamination level was taken in the main river around the various sites where the river joins. While the site difference couldn’t be identified because of the fish morbidity, the levels were below being serious. However, the levels of Pb and Hg were slightly higher but not significant.
14) For the farmig product, nine crops sampled. The levels of Pb and Zn were significantly higher in brown rice, corns, garlics, while the others were below the permissible level. This results are consistent with the residents results But no significance were found when compared to the control area, while some metals show higher levels in the contral area.
15) The metal contaminatin levels in the soil of the study area were significantly higher for Pb, Zn and Mn. Although the Mn level in the residents blood is very low, the high levels for of Pb and Zn need attention beeanse they were also high in the crops.
16) The correlation between the soil and crops did not show any consistent pattern. However, in certain sites, the high correlations were found and suspected due to the confounding factors such as the regional nature of soil contamination, pesticides and fertilizer.
17) For the corelation between the blood and environmental contaminations, the levels of metals except Mn which showed the high levels in the soil and crops were also high in the blood. The relatively great portion of the residetns with high levels can cause serious health impact. However, the origin of their contamination seems to come from the soil not from the river. It deserves a further investigation on this matter.
-
Summary
-
A study on the theoretical property and applied range of odds ratio
-
Kyung Hee Park, Seung Wook Lee
-
Korean J Epidemiol. 1985;7(2):202-227.
-
-
-
Abstract
PDF
-
Abstract
The odds ratio, termed the cross-product ratio and the relative odds, has been used as the measure of degree of association in 2X2 contingency table. The literature on odds ratio is vast, but it is poorly integrated. This paper presents to try to get sharp understanding through the study for whole integration on the basic cencept and theoretical properties and applied range of odds ratio by means of literature consideration.
The content of study is summarized as follows:
In 1887, the odds ratio originally was proposed by Korosy as a natural measure of association on the efficacy of small pox vaccination, but, in his paper, was not given operational interpretation and it was difficult to compare meaningfully its values for 2X2 contingency tables. It was not used for a long time from then on. Afterwards, many statisticians have studied on the statistical property of odds ratio. For rare diseases, the odds ratio is a close approximation to the relative risk. Further, the odds ratio is directly estimable from case-control study data, even though the relative risk is not. Odds ratio has its proper invariance and nice mathematical property of log-odds transformation, called the logit. Because of these property, much attention has given to its use by epidemiologists, statisticians and a number of study has been presented.
The methods of estimation of odds ratio has been studied according to statistical technique for controlling bias due to potential confounding factors, and the scope of its practical use has been very wide. In this paper, their methods are introduced, and illustrated.
-
Summary
|
KSE
