Nutritional status of Fishermen Communities: validation of conventional methods with discriminant function analysis

Anthropological methods of assessing nutritional status of adults have been reinvestigated. Objective of the study is to detect the predictor variables that discriminate for under nutrition or Chronic Energy Deficiency (CED) by two conventional methods e.g. Body Mass Index (BMI) and Mid Upper Arm Circumference (MUAC). Discriminant function analysis was used to build valid and accurate predictive model for evaluating nutritional status. Anthropometric measurements were collected using standard techniques and used as independent variables. Recommended cut-off values of BMI and MUAC was used for evaluating nutritional status. The extent of CED (BMI < 18.5) was found 43.50% and prevalence of under-nutrition in terms of MUAC (MUAC < 23.0 cm for Male and < 22.0 cm for Female) was 21.7%. Discriminant function analysis reveals that 85.7% and 72.0% individuals were classified correctly in terms of nutritional status. Therefore, BMI is the good indicator for detecting malnutrition. Fat mass discriminates between groups.


Introduction
Under-nutrition or malnutrition is considered as a major public health problem and is a significant area of concern in developing countries (Schofield and Ashworth 1996;Antwi 2008;Khor 2008) like India. India is still one of the poor countries in the world with a total population of more than a billion (World Bank 2000). Krishnaswami (2000) observed that, more than half of the world's undernourished individuals live in India.
It is estimated that approximately eight hundred million individuals worldwide are under-nourished, of which 258 million individuals (a little under one third) are concentrated in south Asia (Gaiha 1997). More than 3.6 million mothers and children die each year as a result of malnutrition (Blake et al. 2008).
It has been recognized that the techniques of anthropometry has a long tradition to assess nutritional and health status of adults (Misra et al. 2001; Rao et al. 2006;Bharati et al. 2007; Bisai et al. 2008) because it is universally applicable, inexpensive and non-invasive in nature (WHO, 1995;Ferro-Luzzi et al., 1992;James et al., 1994;Pirlich and Lochs, 2001). Though nutritional status of adults can be evaluated in many ways, one important measure is the calculation of Quetelet index, popularly known as Body Mass Index (BMI), which is the body weight (in kg) divided by square of stature (in m) (Keys et al., 1972),. BMI is generally considered as a good indicator and used for the assessment of adult nutritional status (Lee & Nieman, 2003;Khongsdier, 2002), especially in the developing countries (Khongsdier, 2002;Ferro-Luzzi et al., 1992;. Nowadays BMI has been widely used for assessing Chronic Energy Deficiency (CED) (Khongsdier, 2002). On the other, Mid Upper Arm Circumference (MUAC) is another important indicator for simple screening of adult nutritional status specifically in developing countries , but its cut-points have not been standardized.
Many studies on Indian populations used single method for evaluating malnutrition in different populations living in different regions, which shows significant association between malnutrition and anthropometric measurements. To the best of our knowledge, till date we have not encountered any evidence which have established a suitable method for evaluating malnutrition by comparing two or more methods using common predictors of the same population data. Therefore, we have considered two existing methods (BMI and MUAC) to find out the most effective variables for detecting malnutrition. In view of the above the objective of the present study are as follows: 1. To determine the prevalence of malnutrition on the basis of conventional anthropological methods. 2. To evaluate the best anthropological method 3. To find out the important discrimating variables out of so many anthropometric measurements. 4. To perform a significant statistical model for predicting malnutrition Studies on nutritional status among adults of fishing community in India more specifically in West Bengal and Orissa are scanty. Thus, the present anthropometric data was collected on the adults belonging to fishing community living at Digha (West Bengal) and Udaypur (Odissa). With this background information, the present study is carried out.

Study population:
The data of the present study was part of an larger ongoing project on health and nutrition. The fishing community was chosen for its numerical dominance in the area. It had been observed that the children and women of the community had visible symptoms of under nutrition. Villages were selected on the basis of the occupational homogeneity of the population (the adult males generally practice deep sea fishing). Populations are two rural settlements namely Duttapur and Udaypur were selected purposively. However, other nearby settlements was dominated by other occupational groups. A total of 736 individuals were included in the study aged 18 years and above in both sexes. Explaining the purpose of the study, a complete enumeration of both the villages consisting of 285 households was done. The vast majority of the study populations were illiterate (can not read and write) and belonged to lower economic group (family income less than Rs.1000/= per month equivalent to less than 20$).

Anthropometric measurements and indices:
The following measurements were obtained from the adult individuals (18 to 65 years of age) of both sexes using the standard protocol and instrument (Weiner and Lourie, 1981)  Receiver operating characteristic (ROC) analysis was performed to examine the concordance among predictors using the status of malnutrition of dependent variable. ROC analysis for predictors were not better than chance of discriminating between groups of dependent variable. For this reason stepwise discriminant analysis was adopted. It is useful for situation in which to build a predictive model of group membership based on observed characteristics of each case. The target of discriminant analysis is to identify the independent variables that have a strong relationship to the group membership in the category of dependent variable. The technique provides a linear combination of the predictors. The discriminant function creates the maximum difference between group memberships in the categorical dependant variable.

Assumptions:
1. Each predictor variables is normally distributed. 2. Each group or category is well defined, clearly differentiated from any other group(s) and natural.
3. There exists a linear relationship between pairs of independent variable. 4. Homogeneity of variance among the groups.  Table 7 shows the correlation of each independent variable with each discriminate function. In this table BMI have the highest correlation than Cc with discriminant function 1 in MUAC. On the other hand Fm has the highest value than other independent variables with function 1 in BMI. These results provide another way of indicating the relative importance of the predictors and nutritional variables. Generally with a loading of 0.30 or more is considered to be important in define the discriminate dimension. Table 8 the function of the group centroid gives the average discriminant score of the group. It is observe that Function 1 separates group mean in dependant variables and group 1 (least value) difference between the other canonical group means in classifying observable. Table 9 is used to assess how well the Fisher's classification function coefficients are classified between the groups and useful in deciding which variable affects more in the classification. Result shows that Sex, Bad Fm, and Fp have most contributions for malnutrition classification. The coefficients of the independent variables are used to construct a discriminant function for each group, i.e. Malnutrition (CED), Normal and Over Weight where,  (3) is highest value then the case is classified into malnutrition.
It has been observed from table 10 that 72.0% and 85.7% of overall data was correctly classified by the MUAC and BMI respectively. It has also been noticed that 80.0% and 69.8% were correctly classified in two groups of MUAC and 90.0%, 81.8% and 95.0% were correctly classified in three groups of BMI respectively. This result is rather satisfactory for MUAC. Thus the accuracy of the lassification may hence be considered better BMI categories than MUAC categories.

Conclusion:
This paper is concerned with anthropometric characteristics and nutritional status of the fishing community of West Bengal and Odisha. It was observed that malnutrition has the highest proportion of correctly classifying than normal in both nutritional variables. Furthermore, BMI has highest percentage of correctly classification than MUAC. Thus BMI should be preferred than MUAC method for evaluating the malnutrition. Result suggest that fat mass has portent discriminator power for evaluate malnutrition. As per discriminant analysis results show that the condition of the population causes severe health problem. An appropriate intervention for improving the nutritional status should be initiated at the local level.