Doren Walker

Forecasting Substance Abuse Treatment Admissions

Time Series Analysis in TEDS (Treatment Episode Dataset)

by: DJ (Doren) Walker

Overview of Project

This project aims to answer three specific questions:

Can a model accurately predict the number of substance abuse treatment admissions for a given day, week, month?

Are treatment admissions cyclical or do they display any seasonal characteristics? (Maybe binge drug use leading to an admission to rehab has some seasonal patterns)

Are there different cyclical behaviors between different substances?

Why Is this Important?

Substance abuse is a public health crisis and If cyclical trends are observed, and a model can accurately predict when there will be spikes in treatment, this information can help policy makers invest in unique, targeted intervention programs that can change in intensity/funding throughout the year. Research suggests that treatment of substance abuse disorders saves the country and tax-payers money, so, this project has financial implications that may put a dent in the $600 billion spent, annually on substance abuse.

Heat map made with plotly using TEDS, illustrating the density of treatment admissions per state (raw data, not normalized by population).

What is TEDS?

TEDS is a national data system of annual admissions and discharges to substance abuse treatment facilities. State laws require substance abuse treatment programs to report their publicly-funded admissions to the state. States then report these data from their state administrative systems to Substance Abuse Mental Health Services Administration (SAMHSA). The data has been collected since 2006 and publicly available but has been re-coded to protect privacy.

TEDS contains records on admissions aged 12 or older, and includes information on admission demographics such as:

•age

•sex

•race

•ethnicity

•employment status

•substance abuse characteristics.

•Date of Admission (Not in public use file)

eXPLORATORY dATA aNALYSIS

In this presentation, I go through the data to ensure it's ready for advanced analytics. Some preliminary forecasting was also performed.

Findings

github repository: https://github.com/dorenw/Time_Series_Analysis_TEDS

References

Grassly, N. C., & Fraser, C. (2006). Seasonal infectious disease epidemiology. Proceedings. Biological sciences, 273(1600), 2541–2550. https://doi.org/10.1098/rspb.2006.3604

Nesreen K. Ahmed, Amir F. Atiya, Neamat El Gayar & Hisham El-Shishiny (2010) An Empirical Comparison of Machine Learning Models for Time Series Forecasting, Econometric Reviews, 29:5-6, 594-621, DOI: 10.1080/07474938.2010.481556

S. Ben Taieb and G. Bontempi, "Recursive Multi-step Time Series Forecasting by Perturbing Data," 2011 IEEE 11th International Conference on Data Mining, Vancouver,BC, 2011, pp. 695-704, doi: 10.1109/ICDM.2011.123.

Page updated

Report abuse