Improving Care Outcomes Tool (ICOT) 

   

Background

The iCOT is a region wide, automated behavioral health outcomes assessment which was developed by the Behavioral Health (BH) Department at KPCO in partnership with labor and with support from the Kaiser Permanente Institute for Health Research (IHR). The iCOT is a composite of validated and widely used instruments and includes the PHQ-9, GAD-7, ECHO, AUDIT and DAST. These instruments survey the conditions which make up the majority of costly behavioral health issues in the United States: depression, anxiety, general functioning, and alcohol and drug usage. Implemented in December 2011, all BH patients are asked to complete the iCOT at intake and follow-up visits. Clinicians can access reports generated from the iCOT prior to their patients’ appointments in order to guide treatment planning and monitor progress. Recovery curves based on acuity at intake are also displayed showing change in overall iCOT scores from intake through four visits. A sample of the report and recovery curve is presented below:

[image coming soon]
   

 

iCOT data have not yet  been linked to VDW data for a more detailed analysis of factors associated with recovery trajectories.

Additionally, a widespread and significant problem for assessing psychotherapy outcomes is early treatment discontinuation. Approximately 40% of KPCO BH patients completing the iCOT at intake have only one visit with a BH clinician. Linked iCOT/VDW data may be used to assess factors associated with and predict early treatment discontinuation.

 
   

Project Objective

The objectives of this project were to link iCOT data with VDW data to predict recovery trajectories and provide data on early treatment discontinuation. This project will increase the clinical utility of the iCOT for the BH department and prepare for an NIMH-funded Mental Health Research Network (MHRN) project on this topic.

   

 

Results

We applied multivariate models to 10,533 eligible cases to examine the contributions of clinical, demographic, and utilization variables to the total iCOT score from intake to visit 4. The most significant variables in the multivariate model were a diagnosis of PTSD, dispensing of anticonvulsants or benzodiazepines, and suicidal ideation as measured by item 9 on the PHQ9.

The multivariate model was used in bootstrap samples applied to the original data set and it fit these samples with a high degree of accuracy.

A simulation model (see simulation tab of attached excel spreadsheet) was created in excel that allows one to vary the predictors in the multivariate model in order to generate different recovery curve plots. The algorithm for creating recovery curve plots will be simplified and made available to behavioral health clinicians for use in clinical decision support when treating their patients.

 Regarding early treatment discontinuation, 47% of the patients had no follow up visits to behavioral health after the intake visit. The strongest predictors of early treatment discontinuation in the multivariate models were a diagnosis of eating disorder (thought the numbers were small), ADD, substance use disorder, and dementia.

    If you have questions, please contact Dr. Arne Beck, MHRN researcher, at Kaiser Permanente Colorado.