This is an analysis of the change in frequency of use from admission to discharge for ~4 million completed treatment episodes from the TEDS-D Dataset from 2015-2019

In this analysis, we aggregate the change of frequency of use between admission and discharge into 4 categories listed below:

  1. Improvement

    • Daily use at admission -> Some use at discharge

    • Daily use at admission -> No use at discharge

    • Some use at admission -> No use at discharge

  2. Stagnant Positive
    • No use at admission -> No use at discharge
  3. Stagnant Negative

    • Some use at admission -> Some use at discharge

    • Daily use at admission -> Daily use at discharge

  4. Unknown

To identify the most significant drivers of this aggregated frequency of use change variable, we utilized a random forest model. With frequency of use change as the dependent variable, the model returned these variables as the top 10 most important in making predictions

The top two variables are Services and Length of Stay LOS

  1. SERVICES: Type of treatment/service setting at admission This field describes the type of treatment service or treatment setting in which the client is placed at the time of admission or transfer.

    • Detoxification, 24-hour service, hospital inpatient: 24 hours per day medical acute care services in hospital setting for detoxification of persons with severe medical complications associated with withdrawal.

    • Detoxification, 24-hour service, free-standing residential: 24 hours per day services in non-hospital setting providing for safe withdrawal and transition to ongoing treatment.

    • Rehabilitation/Residential – hospital (other than detoxification): 24 hours per day medical care in a hospital facility in conjunction with treatment services for alcohol and other drug use and dependency.

    • Rehabilitation/Residential – short term (30 days or fewer): Typically, 30 days or fewer of non-acute care in a setting with treatment services for alcohol and other drug use and dependency.

    • Rehabilitation/Residential – long term (more than 30 days): Typically, more than 30 days of non-acute care in a setting with treatment services for alcohol and other drug use and dependency; may include transitional living arrangements such as halfway houses.

    • Ambulatory - intensive outpatient: At a minimum, treatment lasting two or more hours per day for 3 or more days per week.

    • Ambulatory - non-intensive outpatient: Ambulatory treatment services including individual, family and/or group services; may include pharmacological therapies.

    • Ambulatory - detoxification: Outpatient treatment services providing for safe withdrawal in an ambulatory setting (pharmacological or non-pharmacological).

  2. LOS: Length of stay in treatment (days) Describes the length of the treatment episode (in days). Length of stay was computed using the date of admission and the date of last contact. One day is added to all outpatient discharges, so that the first day and last day of outpatient treatment are counted.

Here, we visualize the relationship between frequency of use change and length of stay by service

The next two most important variables are stfips (State) and division

This prompted us to visualize the frequency of use change by state

Frequency of Use Change by State

Since we saw service is highly correlated with frequency of use change, we felt it was necessary to also visualize proportion of treatment episodes per service by state

Percentage of Treatment Episodes per Service by State

Chi-Square test

Living Arrangement Change and Employment Change are also in the top 10 most important variables in prediciting frequency of use change.

## [1] "Result for frequency_of_use_improvement_vs_employment_change_improvement"
## 
##  Pearson's Chi-squared test
## 
## data:  table_data
## X-squared = 1589316, df = 16, p-value < 0.00000000000000022
## 
## [1] "Result for frequency_of_use_improvement_vs_living_arrangement_improvement"
## 
##  Pearson's Chi-squared test
## 
## data:  table_data
## X-squared = 1656229, df = 20, p-value < 0.00000000000000022
## 
## [1] "Result for frequency_of_use_improvement_vs_self_help_improvement"
## 
##  Pearson's Chi-squared test
## 
## data:  table_data
## X-squared = 1249728, df = 16, p-value < 0.00000000000000022
## 
## [1] "Result for employment_change_improvement_vs_living_arrangement_improvement"
## 
##  Pearson's Chi-squared test
## 
## data:  table_data
## X-squared = 3477907, df = 20, p-value < 0.00000000000000022
## 
## [1] "Result for employment_change_improvement_vs_self_help_improvement"
## 
##  Pearson's Chi-squared test
## 
## data:  table_data
## X-squared = 1513371, df = 16, p-value < 0.00000000000000022
## 
## [1] "Result for living_arrangement_improvement_vs_self_help_improvement"
## 
##  Pearson's Chi-squared test
## 
## data:  table_data
## X-squared = 1428017, df = 20, p-value < 0.00000000000000022

The highly significant associations between these variables prompted the visualization of frequency of use by: - Employment Change - Living Arrangement Change