Skip to main page content
U.S. flag

An official website of the United States government

Dot gov

The .gov means it’s official.
Federal government websites often end in .gov or .mil. Before sharing sensitive information, make sure you’re on a federal government site.

Https

The site is secure.
The https:// ensures that you are connecting to the official website and that any information you provide is encrypted and transmitted securely.

Access keys NCBI Homepage MyNCBI Homepage Main Content Main Navigation
. 2021 Jul:2021:10.1109/bhi50953.2021.9508614.
doi: 10.1109/bhi50953.2021.9508614. Epub 2021 Aug 10.

Theory-Guided Randomized Neural Networks for Decoding Medication-Taking Behavior

Navreet Kaur  1 , Manuel Gonzales 4th  2 , Cristian Garcia Alcaraz  2 , Laura E Barnes  1 , Kristen J Wells  2   3 , Jiaqi Gong  4
Affiliations

Theory-Guided Randomized Neural Networks for Decoding Medication-Taking Behavior

Navreet Kaur et al. IEEE EMBS Int Conf Biomed Health Inform. 2021 Jul.

Abstract

Long-term endocrine therapy (e.g. Tamoxifen, aromatase inhibitors) is crucial to prevent breast cancer recurrence, yet rates of adherence to these medications are low. To develop, evaluate, and sustain future interventions, individual-level modeling can be used to understand breast cancer survivors' behavioral mechanisms of medication-taking. This paper presents interdisciplinary research, wherein a model employing randomized neural networks was developed to predict breast cancer survivors' daily medication-taking behavior based on their survey data over three time periods (baseline, 4 months, 8 months). The neural network structure was guided by random utility theory developed in psychology and behavioral economics. Comparative analysis indicates that the proposed model outperforms existing computational models in terms of prediction accuracy under conditions of randomness.

Keywords: choice model; medication adherence; random utility maximization; randomized neural networks.

PubMed Disclaimer

Figures

Figure 1.
Figure 1.
When gaussian noise with SNR=20 is added in the data, probability distribution of data remains unchanged.
Figure 2.
Figure 2.
The proposed RNN model with Win random untrained weights between input u and hidden neurons (in the box); Wout is the matrix of adaptive weights used to compute network output. The output is added with another random probability to obtain the final predicted probability.
Figure 3.
Figure 3.
The randomized neural networks outperform in various conditions of Signal-Noise Ratio (SNR) of the subjective values.
Figure 4.
Figure 4.
When SNR=20 in the subjective values, the architecture of neural networks will slightly impact the prediction performance.
Figure 5.
Figure 5.
Individual accuracy in two types of neural networks. Traditional neural networks cannot capture the dynamic patterns in patients’ medication behavior, such as subject 1 and 4.
See this image and copyright information in PMC

References

    1. Multack Megan. “Use of clinical preventive services and prevalence of health risk factors among adults aged 50–64.” AARP Public Policy Institute: Washington, DC, USA (2013).
    1. World Health Organization, “Adherence to long-term therapies: Evidence for action,” Ann Saudi Med., vol. 24, pp. 221–222, 2003.
    1. Wells KJ et al., “Barriers and facilitators to endocrine therapy adherence among underserved hormone receptor-positive breast cancer survivors: A qualitative study,” Supportive Care Cancer, vol. 24, pp. 4123–4130, 2016 - PMC - PubMed
    1. de Bekker-Grob EW, Donkers Bas, Bliemer MCJ, Veldwijk Jorien, and Swait JD. “Can healthcare choice be predicted using stated preference data?.” Social Science & Medicine 246 (2020): 112736. - PubMed
    1. BANDURA A, Social foundations of thought and action: A social cognitive theory, Prentice-Hall series in social learning theory., Prentice-Hall, Inc, Englewood Cliffs, NJ, US, 1986.

LinkOut - more resources