TITLE:
A novel algorithm for describing population level trends in body weight
AUTHORS:
Azadeh Alimadad, Carrie Matteson, Warren L. Hare, Ozge Karanfil, Diane T. Finegood
KEYWORDS:
Body Weight; Modeling; Obesity
JOURNAL NAME:
Health,
Vol.4 No.12A,
December
31,
2012
ABSTRACT:
Modeling population
trends and predicting the impact of interventions to address obesity requires
algorithms for predicting body weight status in the future. Predictions can be
based on statistical consideration of different risk factors, or be an
extrapolation of past and current trends. Despite the well known correlation
between previous and future weight, individual weight history has not been used
to predict future trends. We developed a novel population-level model to
examine trends of different classes of body weight considering individual body
weight histories from the National Longitudinal Survey of Youth
(NLSY79). A subset of data used to assess
the predictive ability of our proposed model with actual data. Our
results confirm the importance of weight history in determining future weight
status. Over 80% of individuals in a specific weight category (normal,
overweight, obese) will stay in the same weight category after two years
(except overweight females). The length of body weight stability was also found
to be important. The probability of remaining normal weight increased with
longer prior periods of being at a normal weight over 18 years (0.834 to
0.893). We demonstrate that an individual’s most probable weight class in the
future is consistent with their maximal historical weight class.