Developing a cost-effective computing model for optimal diets for people living with HIV

Abstract

People living with HIV (PLWHIV) without proper treatment are vulnerable to many kinds of opportunistic infections due to their weaker immune systems than healthy people. Poor nutrition intensifies the progression of HIV into AIDS by further compromising the immune system. Therefore, achieving basic nutritional recommendations is important at all stages of the disease. However, economic limitation (poverty) and lack of knowledge to find adequate amounts with right combinations of different locally available foods hinders them to meet the recommended daily nutrients requirements, leading them to become weak in a very short time and even experience early mortality. In this research, I developed a mathematical model and extended it to a MATLAB based graphical user interface (GUI) that could be used as a computation tool to compute adequate amounts of available foods to achieve the recommended nutrients at a minimum cost compared to an alternative. The mathematical model is the combination of multiple linear regression models and a linear programming model. Multiple linear regression models use the factors of age, weight, height, and gender to predict the nutritional requirements in the body. The results from the multiple linear regression model were used to define the constraints in the linear programming model. The linear programming model was used to compute the adequate amounts of foods that would lead to the achievement of the recommended nutrients taking into consideration practical biological/physical and economic constraints. The Graphical User Interface (GUI) was developed by the Graphical User Interface Development Environment (GUIDE) method in MATLAB. With the incorporated mathematical models it could be used to compute the adequate amount of foods. The GUI has five parts: the first part contains list of foods the user needs to select, the second part is to enter user’s particulars of age, weight, height and gender. The third part is to enter the cost of each selected foods. The fourth part is the computation part, which will initiate the computation. There is a status box, which shows whether the food combinations and financial constraints produce an optimal or non-optimal output and a reset button to enable clearance of previous computations and allowance of new data entrances. The last part is the output section which displays the amounts of foods to be bought and the total cost to be incurred when the computation is optimal. Results show that the multiple linear regression model has high predictive power by suggesting values that are close to the recommended daily/dietary intake (RDI). This was validated by testing the mean difference between paired samples using a t-test. By this analysis we found that there was no statistical ii difference between the means as the p  values were greater than the significance level of 0.05. The cost for optimal diets was less when model predicted values are used to limits the constraints in linear programming compared to when RDI values are used. The GUI developed could serve as the computation tool to compute adequate amount of foods to meet the recommended nutrients at minimum costs.