Diagnosis of alcoholism based on neural network analysis of phenotypic risk factors
2005

Using Neural Networks to Diagnose Alcoholism

Sample size: 650 publication Evidence: high

Author Information

Author(s): Catherine T Falk, Joan E Bailey-Wilson, Laura Almasy, Mariza de Andrade, Julia Bailey, Heike Bickeböller, Heather J Cordell, E Warwick Daw, Lynn Goldin, Ellen L Goode, Courtney Gray-McGuire, Wayne Hening, Gail Jarvik, Brion S Maher, Nancy Mendell, Andrew D Paterson, John Rice, Glen Satten, Brian Suarez, Veronica Vieland, Marsha Wilcox, Heping Zhang, Andreas Ziegler, Jean W MacCluer

Primary Institution: CCNY of the City University of New York

Hypothesis

Can neural network analysis of phenotypic risk factors improve the diagnosis of alcoholism?

Conclusion

Neural network analysis can reliably predict the presence or absence of alcoholism about 94–95% of the time.

Supporting Evidence

  • The neural network achieved about 94% reliability in predicting alcoholism using 36 input factors.
  • Pruning the network to 14 input factors maintained a reliability of 95%.
  • The maximum number of drinks consumed in a day was a significant predictor of alcoholism.

Takeaway

Scientists used a computer program to help figure out if people have alcoholism by looking at their behaviors and traits, and it worked really well.

Methodology

A back-propagation, feed-forward neural network was trained using 36 input phenotypic risk factors to classify individuals as affected or normal.

Limitations

The study did not include individuals with missing data and did not account for family relationships.

Participant Demographics

Of the 650 individuals analyzed, 44 were 'pure unaffected', 293 were 'unaffected with some symptoms', and 313 were 'affected'.

Digital Object Identifier (DOI)

10.1186/1471-2156-6-S1-S131

Want to read the original?

Access the complete publication on the publisher's website

View Original Publication