Quality Improvement Collaboratives and the IHI Breakthrough Series Model

A quality improvement collaborative is a time-limited, multi-team learning structure: a group of hospitals or clinics — usually somewhere between 20 and 40 teams — work on the same focused problem at the same time, sharing data and tactics with each other rather than improving in isolation. The best-documented version of this model is the Breakthrough Series, published by the Institute for Healthcare Improvement (IHI) in a 2003 white paper, “The Breakthrough Series: IHI’s Collaborative Model for Achieving Breakthrough Improvement.” It is not the only collaborative format in use, but its Learning Session / Action Period structure is the one most other hospital and health-system collaboratives borrow from, whether or not IHI is formally running them.

The Structure: Learning Sessions and Action Periods

A Breakthrough Series collaborative runs for roughly six to fifteen months and alternates between two phases:

• Learning Sessions. Multi-day, in-person meetings — typically three across the life of the collaborative (often labeled LS1, LS2, LS3) — where each participating team sends a small delegation (commonly around three members) to hear from subject-matter faculty, review the collaborative’s shared “change package,” and present their own team’s data alongside every other team’s.
• Action Periods. The weeks or months between Learning Sessions, spent back at each team’s own site actually running the changes: local PDSA cycles testing a specific change on a small scale, collecting the agreed measures, and adjusting before the next Learning Session. Teams typically submit a monthly progress report to the collaborative’s coordinating group during this phase, and stay connected between sessions through conference calls, a shared listserv, and occasional peer site visits.

The whole arrangement is designed to compress a normally slow, isolated improvement cycle into something faster and less prone to reinventing the wheel at every site — a team stuck on a specific barrier can see, in near real time, that three other teams in the same collaborative already solved it.

What a Collaborative Actually Requires to Run

The Breakthrough Series model asks for real infrastructure, not just a topic and a mailing list. A collaborative needs:

• A planning group and expert faculty. A small group defines the collaborative’s scope and selects the changes worth testing based on existing evidence, then recruits clinical and improvement-methodology faculty to teach and coach across the Learning Sessions.
• An aim statement and a change package. The aim is a specific, numeric, time-bound target (e.g., “reduce X by Y% within Z months”), and the change package is the pre-vetted menu of specific changes teams are meant to test locally — drawn from the Model for Improvement’s underlying logic of “what are we trying to accomplish, how will we know a change is an improvement, and what changes can we make that will result in improvement,” applied through repeated small-scale PDSA cycles rather than one big rollout.
• A measurement strategy every team can actually sustain. Collaboratives typically ask for monthly data submission plotted as a run chart, not a single before/after snapshot — the whole point is to see whether a team’s change is holding, not just whether one measurement looked better once.
• Sponsor commitment at each site. IHI’s own Assessment Scale for Collaboratives tool — a 0-100 maturity scale IHI built to help a collaborative’s leadership judge how far a given team has actually progressed, beyond attendance alone — treats senior-leader sponsorship and protected staff time as prerequisites for a team to reach the higher bands on the scale, not optional extras.

The Published Evidence: Positive but Genuinely Limited

The most-cited systematic review of this literature is Schouten et al., published in BMJ in 2008. Of 1,104 articles identified studying quality improvement collaboratives, 72 reported outcome or effectiveness data — but 60 of those 72 (82%) used an uncontrolled before-after design, which cannot separate a collaborative’s own effect from secular trends, regression to the mean, or simple Hawthorne-effect measurement artifacts. Only 12 reports, covering nine actually-controlled studies (two of them randomized), let the review draw a real causal comparison. Of those nine: seven showed at least some positive effect on a process or outcome measure, and two showed no effect at all. The review’s own conclusion is worth stating plainly rather than rounding up: the evidence for collaboratives is positive but limited, and the size of the effect a given collaborative will produce “cannot be predicted with great certainty.”

Later work has not resolved this into a cleaner verdict. A subsequent systematic review and meta-analysis focused on low- and middle-income-country settings pooled 29 studies — most using an interrupted-time-series design rather than a randomized or even a concurrently-controlled one — and rated the overall quality of the underlying evidence as low to very low, even while finding a net-positive pooled effect. The pattern across this literature is consistent: collaboratives that get evaluated with a real comparison group tend to show *some* benefit more often than not, but the evaluation designs available are mostly weak, and a collaborative’s success is not something a hospital can assume will happen just by joining one and sending a team to the Learning Sessions.

When Collaboratives Work — and When They Don’t

Reading the Schouten-era literature and the collaborative-methodology tools IHI built in response to it (the Assessment Scale above exists specifically because IHI’s own faculty observed wide variation in team performance within the same collaborative), a few practical distinctions hold up:

• Works better with a narrow, well-defined aim rather than a broad mandate; with real, protected staff time for local PDSA testing between Learning Sessions rather than an unfunded volunteer effort layered onto existing workload; with a measurement infrastructure the team can sustain monthly without heroic manual effort; and with senior sponsor engagement that can remove local barriers a frontline team can’t remove itself.
• Tends to fail when a team treats the collaborative as a reporting obligation — showing up to Learning Sessions and submitting data without running genuine local PDSA cycles in between — or when the change package doesn’t fit local context and no one adapts it; when there is no protected time, the “collaborative” work competes with regular clinical duties every week and quietly stops; and when a team lacks the data infrastructure to produce a real run chart, so progress gets judged on anecdote instead of trend.

None of this is unique to IHI’s specific format — the same distinctions apply to any collaborative-improvement structure built on the same Learning-Session/Action-Period logic, including state-run or specialty-society collaboratives that use IHI’s methodology without IHI running the program directly.

Frequently Asked Questions

How is a quality improvement collaborative different from a hospital just running its own internal QI project?

The core improvement method — the Model for Improvement and PDSA cycles — is identical either way. What a collaborative adds is the shared structure across many organizations at once: a common aim, a common change package drawn from existing evidence, faculty coaching multiple teams simultaneously, and visibility into what other teams tried and whether it worked, which a single hospital working alone doesn’t get.

How long does a Breakthrough Series collaborative typically run?

IHI’s original model specifies roughly six to fifteen months, structured around a fixed number of Learning Sessions (commonly three) with Action Periods in between. Some adaptations run longer or shorter, but the alternating in-person/local-testing rhythm is the defining structural feature, not a specific duration.

Does joining a collaborative guarantee better outcomes?

No. The best available systematic evidence (Schouten et al., BMJ 2008) found a positive average effect across controlled studies, but with only nine genuinely controlled studies to draw on and most of the wider literature using weaker before-after designs, a hospital cannot assume participation alone produces results — the infrastructure and local execution described above are what the evidence actually links to success.

What is the relationship between the Breakthrough Series and the Model for Improvement?

The Breakthrough Series is the multi-organization collaborative structure; the Model for Improvement (the three fundamental questions plus PDSA cycles) is the underlying improvement method every team applies locally during each Action Period. One is the container, the other is the tool used inside it.

Related CASRAI Resources

• PDSA Cycle and the Model for Improvement — the underlying test-of-change method every collaborative team applies during its Action Periods.
• Run Charts for Quality Improvement — the measurement discipline collaboratives rely on to distinguish a real signal from noise.
• SQUIRE 2.0 — the reporting standard for publishing a collaborative’s own results.
• Quality Improvement vs. Human Subjects Research — whether a collaborative’s own data collection needs IRB review.
• Patient Safety & Infection Prevention — the CASRAI hub for this cluster.