Background: Quality improvement collaboratives (QICs) are widely used to improve healthcare, but there are few studies of long-term sustained improved outcomes, and inconsistent evidence about what factors contribute to success. The aim of the study was to open the black box of QICs and compare characteristics and activities in detail of two differing QICs in relation to their changed outcomes from baseline and the following 3 years.

Methods: Final reports of two QICs - one on heart failure care with five teams, and one on osteoarthritis care with seven teams, including detailed descriptions of improvement projects from each QIC's team, were analysed and coded by 18 QIC characteristics and four team characteristics. Goal variables from each team routinely collected within the Swedish Heart Failure Registry (SwedeHF) and the Better Management of Patients with OsteoArthritis Registry (BOA) at year 2013 (baseline), 2014, 2015 and 2016 were analysed with univariate statistics.

Results: The two QICs differed greatly in design. The SwedeHF-QIC involved eight experts and ran for 12 months, whereas the BOA-QIC engaged three experts and ran for 6 months. There were about twice as many activities in the SwedeHF-QIC as in the BOA-QIC and they ranged from standardisation of team coordination to better information and structured follow-ups. The outcome results were heterogeneous within teams and across teams and QICs. Both QICs were highly appreciated by the participants and contributed to their learning, e.g. of improvement methods; however, several teams had already reached goal values when the QICs were launched in 2013.

Conclusions: Even though many QI activities were carried out, it was difficult to see sustained improvements on outcomes. Outcomes as specific measurable aspects of care in need of improvement should be chosen carefully. Activities focusing on adherence to standard care programmes and on increased follow-up of patients seemed to lead to more long-lasting improvements. Although earlier studies showed that data follow-up and measurement skills as well as well-functioning data warehouses contribute to sustained improvements, the present registries' functionality and QICs at this time did not support those aspects sufficiently. Further studies on QICs and their impact on improvement beyond the project time should investigate the effect of those elements in particular.