(16) ASAM Criteria-based Standardized Recommendations and Discrepancies with Actual Referrals

Background & Introduction: Amidst the national drug epidemic, there is a growing need for quality improvement in addiction care. Matching patients to the appropriate level of care for their multidimensional needs can improve clinical outcomes. However, few large datasets enable comparison of recommended versus received services for addiction treatment. Examining discrepancies between ASAM Level of Care (LOC) recommendations from a standardized, automated tool versus actual LOC placement could help identify service gaps and guide efforts to improve treatment system design and quality of care.

A 24-item structured interview and branched-tree, automated decision assistance software tool has been used by intake workers and counselors around the United States since 2017 to assess patients with substance use disorders. Assessment settings include central intakes, telephone call centers, and crisis centers, among others. The tool generates provisional recommendations for what LOCs a patient should be referred to, where the patient would then undergo comprehensive evaluation and treatment planning. Commissioned by the Los Angeles County Department of Public Health, Substance Abuse Prevention and Control, the tool was developed with expert consensus based on The ASAM Criteria 3rd edition.

Methods: We analyzed clinical data from all computer-assisted brief assessments (ASAM CO-TriageTM) completed nationwide in September 2023 to explore (1) proportions of computer-generated provisional LOC recommendations and (2) the extent to which actual or planned placements agreed with recommended LOCs.

Analyses consisted of counts and percentages of raw, de-identified data from servers centrally maintained by FEi Systems under license from ASAM. The tool prompted brief assessment of all six dimensions of The ASAM Criteria 2013 Edition: Intoxication and Withdrawal; Biomedical; Emotional/Cognitive/Behavioral; Readiness, Relapse Potential; and Recovery Environment.

For analytic purposes, recommendations and referrals were aggregated into the following major ASAM LOCs: L0.5-Prevention/Early Intervention, L1-Outpatient, L2-Intensive Outpatient/Partial Hospitalization, L3-Residential/Inpatient, L4-Intensive Inpatient (e.g., hospital) care, and LOTS-Opioid Treatment Services (including all Medications for Opioid Use Disorder (MOUD)). In some cases, total percentages of all LOCs exceeded 100% because, for example, recommendations for LOTS can occur in combination with any other LOC.

Results: The one-month sample consisted of 5,025 unique assessments.

A total of 376 clinical interviewers submitted assessments from 43 treatment systems or networks using 18 electronic health record (EHR) systems across 34 states and the District of Columbia. Of all submitted assessments, 95.9% (4,817) were completed, with a mean duration of administration of 7.80 minutes. In less than 1% (42) of assessments, the data did not resolve to a particular LOC recommendation. This may occur, for example, when an individual may not require any LOC or does not meet provisional criteria for any LOC.

Of 4,807 computer-recommended LOC instances, interviewers also input actual or referred LOC data for 3,777 assessments (79%).
(1) The proportions of computer-generated ASAM LOC recommendations were:
L0.5=19.0%, L1=27.2%, L2=24.3%, L3=24.0%, L4= 0.9%, and LOTS=29.7%
(2) Actual or planned referrals, however, differed considerably in some categories:
L0.5= 1.3%, L1=29.9%, L2= 6.2%, L3=23.1%, L4= 0.3%, and LOTS=35.7%.

In 56.6% of cases, the interviewer expected the final referral for full evaluation to be the same as recommended by the ASAM Criteria. Interviewers disagreed with the computer-generated provisional recommendation in only 2.7% of cases.

The leading reasons for discrepant referrals were “patient rejects any treatment at this time” (13.5%), followed by “different treatment selected due to patient choice” (9.8%), and “treatment unavailable in geographic region” (6.6%).

Among the patients with discrepant referrals attributed to patient choice, half (50.5%) had been referred by the software to L2 care. Similarly, among patients whose interviewer disagreed with the ASAM CO-Triage recommendation, the largest proportion (35.6%) disagreed with L2 recommendations. Of patients recommended to L2, 52.5% were actually or expected to be referred to a less intensive LOC, and 20.5% to a more intensive LOC.

Conclusion & Discussion: In these registry data, the pattern of computer-recommended LOCs was evenly distributed across all LOCs except L4. Fewer L4 recommendations might be expected as patients needing hospitalization might be disinclined to visit a central intake or screening center.

Respondents disagreed with the ASAM Criteria recommendations in less than 3% of cases. Relative to ASAM-based recommendations, however, actual or planned referrals under-utilized L2. Much of the L2 discrepancy was due to patient preference and poor access. Further research might examine retention rates and clinical outcomes when higher intensity/acuity patients pursue assessment at lower intensity LOC (i.e., L1-outpatient).

These large-scale results may signal needs both for system change (i.e., increasing geographic distribution of L2 services), and for treatment planning (e.g., motivational interviewing to improve patients’ understanding of and readiness for more intensive L2 service needs, as well as case navigation to help address treatment barriers such as caregiver responsibilities or unreliable transportation access). Although real-world data such as these necessarily involve a considerable degree of noise, this type of service discrepancy analysis may offer meaningful signals that can help guide health care resource allocation and delivery of care.

References: 1. Gastfriend DR. Addiction Treatment Matching: Research Foundations of the American Society of Addiction Medicine (ASAM) Criteria. Binghamton NY: Haworth Medical Press, 2004, pp. 1-170
2. Magura S, Staines G, Kosanke N, Rosenblum A, Foote J, DeLuca A, Bali P (2003). Predictive validity of the ASAM Patient Placement Criteria for naturalistically matched vs. mismatched alcoholism patients. American Journal on Addictions, 12(5) 386-397
3. Sharon E, Krebs C, Turner W, Desai N, Binus G, Penk W, Gastfriend DR (2003). Predictive Validity of the ASAM Patient Placement Criteria for Hospital Utilization. Journal of Addictive Disease 22(supplement 1):79-93
4. Stallvik M, Gastfriend DR, Nordahl HM. Matching patients with substance use disorder to optimal level of care with the ASAM Criteria software. J of Substance Use 2015, 20:389-98 (DOI:10.3109/14659891.2014.934305)
5. Drymalski WM, Nunley MR (2018). Sensitivity of the ASAM Criteria to psychiatric need. Int J Ment Health Addiction 16, 617–629 (2018). https://doi.org/10.1007/s11469-017-9801-8