Page 65 - ISMCON souvenir 2021

ISMSCON - 2021

foods must be judiciously chosen to provide adequate nutrients to meet human requirements based on
age, body weight and physiological status, but also with an eye on feasibility and economics.
Aims & Objective:

To create an optimisation model for recommending the quantity of selected raw foods to meet key nutrient
requirements for a healthy population with optimal cost, and to create an interactive user-app for this
purpose.
Methods:

A linear programming optimization technique was used to derive the intake quantities of different raw food
items. The objective function was to minimize the total cost incurred, subject to the constraints of meeting
the nutrient requirements set by the Indian Council of Medical Research- National Institute of Nutrition
(ICMR NIN) in 2020. Since the estimated average requirement of nutrients (EAR) and the upper limit of
nutrient intake (TUL) were different across age and sex groups, we considered a separate optimization
for each of these. EAR and TUL are used as the minimum and maximum requirements respectively in
the optimization. Additional user specific and nutrition expert-advised constraints were also considered.
Results:
The model provides the quantity of selected raw foods to be consumed by an individual or a household
based on per day, week, or month model, to meet 17 different nutrient requirements with minimal cost.
The costs will vary depending on characteristics of the individuals and the type of food selection. An
interactive app for public use was developed and is hosted in https://nutrition-optimization.herokuapp.
com/.
Conclusions :

The present model shows that diverse diets that meet multiple nutrient requirements can be accessed at
low-cost at a population level. More features of the model are currently being developed.

OS30: BAYESIAN LATENT CLASS MODEL ANALYSIS FOR
DIAGNOSTIC TEST EVALUATION

Jeswin Baby , Sriram Sampath , Bhuvana Krishna , Nandini Dendukuri , Tinku Thomas d
a
b
b
c
Division of Epidemiology & Biostatistics, St John’s Research Institute, Bangalore
Department of Critical Care Medicine, St John’s Medical College, Bangalore
Centre for Outcomes Research, McGill University, Montreal, Canada
Department of Biostatistics, St John’s Medical College, Bangalore
Email: jeswin.b@sjri.res.in
Keywords: Bayesian analysis, Sepsis, Blood Culture, Molecular Diagnostics, Intensive care unit
Background:
Evaluation of test properties of a newly developed diagnostic method in the absence of a gold standard test
is an insolvable problem. The researchers usually compare the new test with the existing tests in terms of
positivity or wrongly assume an existing study to be the gold standard. Studying the test properties of the
new test considering the uncertainties in the existing tests is possible with Bayesian Latent class models
(LCMs). Bayesian LCMs combine any prior information on the prevalence and diagnostic characteristics
to estimate the posterior prevalence and the diagnostic characteristics of the tests. We examined the
utility of Bayesian LCM in the confirmation of sepsis by Standard blood cultures (STD) and Molecular
diagnostics (MOL) where STDs are often inconclusive and MOL have higher positivity.
Methods:
Results from an ICU sepsis study that used both STD and MOL tests simultaneously were analyzed, and

CONFERENCE SOUVENIR 63