21.1 Outcome Measurement & Continuous Quality Improvement in Addictions Care
Key Takeaways
- Area V, covering ethics, evidence-based practice, continuous quality improvement and policy advocacy, carries 5% of the CARN-AP blueprint, about 7 to 8 items.
- Quality improvement seeks to improve care in a local setting and does not require IRB review, while research seeks generalizable knowledge and does.
- The Plan-Do-Study-Act cycle drives small, rapid, repeated tests of change rather than a single large implementation.
- Structure, process and outcome measures answer different questions, and balancing measures detect harm caused by the improvement itself.
- Retention in treatment and medication continuity are more actionable addiction quality measures than abstinence rates, which are heavily influenced by case mix.
21.1 Outcome Measurement & Continuous Quality Improvement in Addictions Care
Quick Answer: Activity Area V — incorporate ethical principles and evidence-based practice to promote clinical excellence, optimal outcomes, continuous quality improvement and policy advocacy — is 5% of the CARN-AP (7 to 8 items) but contains 26 activity statements, the most of any area. Quality improvement improves care in a local setting and generally does not require IRB review; research generates generalizable knowledge and does. Improvement work runs on Plan-Do-Study-Act (PDSA) cycles and uses structure, process, outcome and balancing measures.
1. Quality Improvement Versus Research
| Dimension | Quality improvement | Research |
|---|---|---|
| Purpose | Improve a process in this setting | Produce generalizable knowledge |
| Design | Iterative; the intervention changes as you learn | Fixed protocol |
| Participants | Everyone receiving care in the setting | Recruited and consented subjects |
| Risk | Usually no more than standard care | May exceed standard care |
| Oversight | Institutional QI structure; IRB review generally not required | IRB review required |
| Publication | May be published; SQUIRE reporting guideline applies | CONSORT, STROBE, PRISMA as applicable |
The boundary blurs when a QI project randomizes patients, withholds standard care, or is designed from the outset to produce findings intended for other institutions. When in doubt, ask the IRB for a determination rather than deciding unilaterally — that consultation is itself the defensible answer on an exam item.
2. The Model for Improvement and PDSA
Three questions precede any cycle:
- What are we trying to accomplish? A specific, measurable, time-bound aim. "Increase the proportion of patients with opioid use disorder who leave the hospital with a buprenorphine prescription from 22% to 60% within six months."
- How will we know a change is an improvement? The measures.
- What changes can we make that will result in improvement? The change ideas.
Then PDSA:
| Phase | Content |
|---|---|
| Plan | Define the test, who does what by when, what data will be collected, and the prediction |
| Do | Run the test on a deliberately small scale — one clinician, one shift, five patients |
| Study | Compare results with the prediction; ask what was learned, including from failure |
| Act | Adopt, adapt, or abandon; then start the next cycle |
The most common error is scale. Teams launch a unit-wide protocol as the first test and discover a fatal flaw after a month of wasted effort. A first PDSA should be small enough to run tomorrow.
3. Choosing Measures
| Type | Question it answers | Addiction example |
|---|---|---|
| Structure | Do we have the capacity? | Number of clinicians credentialed to prescribe buprenorphine; naloxone stocked on every unit |
| Process | Are we doing the right things? | Percentage of admitted patients screened with a validated tool; percentage of patients with OUD offered medication |
| Outcome | Did the patient do better? | 30-day treatment retention; 90-day medication continuity; readmission; non-fatal overdose |
| Balancing | Did we cause harm elsewhere? | Emergency department length of stay after adding a screening protocol; nursing time displaced from other tasks |
Process measures move first. An improvement team that tracks only overdose mortality will see nothing for a year and lose momentum. Track the process measure you control weekly and the outcome measure quarterly.
Addiction-specific measures worth knowing
- Initiation and Engagement of Substance Use Disorder Treatment (IET): a widely used measure of whether a patient initiates treatment within 14 days of a new diagnosis and engages with two or more further encounters within 34 days of initiation.
- Continuity of pharmacotherapy for opioid use disorder: the proportion of patients with at least 180 days of continuous medication — a strong, actionable measure.
- Follow-up after emergency department visit for substance use.
- Naloxone co-prescribing rate for patients at elevated overdose risk.
Why abstinence is a poor primary measure: abstinence rates are dominated by case mix. A program that admits the sickest patients will look worst on abstinence and may be delivering the best care. Retention and medication continuity are less gameable and more directly modifiable.
4. Displaying and Interpreting Data
- Run charts and statistical process control charts distinguish real signal from normal variation. A single month's change is almost always noise.
- Common cause variation is inherent to the process and requires a process change. Special cause variation has an identifiable external explanation and requires investigating that cause. Reacting to common cause variation as though it were special cause — "why were numbers down in March?" — wastes effort and demoralizes teams.
- Stratify by race, ethnicity, language, insurance and sex. An aggregate improvement can conceal a widening disparity, and in addiction care, disparities in who is offered medication are well documented.
5. From Individual Frustration to System Change
The practice analysis lists "uses the results of quality improvement to initiate changes in nursing practice and the health care delivery system" and "differentiates outcomes that require care process interventions from those that require system-level interventions." The distinction is practical:
| Problem | Level | Fix |
|---|---|---|
| One nurse does not know how to score CIWA-Ar | Care process | Individual education |
| Most nurses score CIWA-Ar inconsistently | System | Standardized training, an embedded scoring tool, and audit with feedback |
| A patient left before buprenorphine induction | Care process | Review that case |
| A third of patients leave before induction because pharmacy takes four hours to deliver the first dose | System | Stock buprenorphine as a floor medication or use an override pathway |
The recurring exam pattern: when a problem repeats across clinicians and patients, the correct answer is a system-level intervention, not more individual education.
An addiction consult service wants to increase the proportion of hospitalized patients with opioid use disorder who are discharged with a buprenorphine prescription. Which first step best reflects the Model for Improvement?
A program compares its abstinence rate to a neighboring program's and finds it substantially lower, despite serving a population with more severe disease, homelessness and co-occurring psychiatric illness. What is the most appropriate interpretation?
An audit finds that a third of patients with opioid use disorder leave the hospital before buprenorphine induction because pharmacy delivery of the first dose takes about four hours. What does the CARN-AP practice-analysis distinction between care-process and system-level interventions indicate?