Short- and long-term effects of a quality improvement collaborative on diabetes management

Good clinical care for patients with type 2 diabetes requires increasingly complicated drug regimens, close monitoring, and ongoing self-management support [1, 2]. For best results, diabetes care also requires effective bridging of primary and specialist care with providers crossing practice and organizational boundaries [3]. Cooperation between hospitals and general practices that focus on integrated, patient-centered care is vital [4, 5].

Despite a wealth of evidence and clinical practice guidelines, treatment gaps in diabetes care are visible when 'best practice' encompassing chronic care management, professional performance, and patient outcomes is compared with usual care [6, 7]. Diabetes is a complex multi-systemic chronic disease and is difficult to fit into a healthcare delivery system designed to deal with acute and episodic illness. Despite reports of interventions designed to improve diabetes care, we do not know which strategies are most effective [8, 9]. In a recent meta-analysis, team changes and case management showed the most robust improvements, although estimates of effectiveness of other specific quality improvement strategies may be limited by difficulty in classifying complex interventions, too few studies, and publication bias [10]. We hypothesized that a multifaceted implementation approach emphasizing collaborative learning and exchange of insights and support among a set of healthcare organizations, like a quality improvement collaborative (QIC), may be helpful to improve diabetes management and reduce risks of complications. QICs bring together a group of healthcare providers from different sites who meet periodically to learn, to exchange ideas and quality methods, and to share experiences with making changes. The aim of a QIC is to close the gap between potential and actual performance by testing and implementing changes quickly across many organizations [11]. QICs are frequently used to improve care systems in multiprofessional settings [12]. The strength of QICs is apparently the relatively efficient use of experts and peers, as well as exchange of evidence and best practices, to guide multiprofessional teams in improvement. However, the widespread acceptance and use of QICs are not based on a systematic assessment of effectiveness [13]. A recent systematic review of QICs shows varying success in achieving collaborative goals; none of the included studies provided information on the sustainability of effect [14].

To improve chronic care in an integrated care setting, a national QIC was set up to encourage high-quality, integrated diabetes care in the Netherlands. This voluntary QIC was designed to bring together and support multiprofessional diabetes teams from primary care and outpatient hospital clinics in applying evidence-based clinical practice. We hypothesized that this intervention could facilitate and support multiprofessional teamwork, integration of primary and specialist care to improve care systems, professional performance (guideline adherence), and patient outcomes.

The aim of our study was to assess the impact of this multi-institutional QIC by measuring patient outcomes, professional performance measures, and structural aspects of chronic care management. Because little is known about the impact and sustainability of QICs, the results are also meant to give insight into the short- and long-term effectiveness of this QIC in diabetes care compared to usual care.

Our study showed modest but significant long-term effects in mean systolic blood pressure, HDL levels, scores for decision support, and less worsening in the percentage of patients visiting a dietician annually. Short term, the percentage of patients receiving advice and instruction to examine feet and scores for self-management improved significantly and less worsening was seen in the percentage of patients receiving instruction to monitor blood glucose. These changes persisted for at least 12 months after the intervention completed. We also found encouraging results on cardiovascular disease risk control at two years follow up (i.e. significant improvement of systolic blood pressure and HDL levels at two years follow up) [24]. The declining patient information scores (received advice about...) are intriguing. Perhaps some patient information is not provided repeatedly every year. Although we cannot exclude that patients may not remember having received an update, implicating some recall bias.

The baseline figures collected in our study are in line with the national figures in other studies [15] and our findings are consistent with the outcomes of a systematic review of QICs [14]. The results also reflect findings from other uncontrolled QIC intervention studies [25, 26] and intervention studies in diabetes care [27] that show that most improvement projects produce small to modest improvement. The 0.3% difference in mean HbA1c levels in both arms is compatible with a recent meta-regression analysis [10].

Examining within group comparisons (difference in differences across intervention and control sites), the mean change (Δ) in proportion of patients with HbA1c, blood pressure, cholesterol, creatinine, and BMI checked annually increased significantly over the years at the intervention sites compared to the control sites (10 to 21% versus 1.5 to 6%; data not shown). Although some key effect parameters clearly improved in the intervention group, the observed difference between intervention and control sites (correcting for baseline scores) was modest. There are several possible reasons. First, significant baseline differences did exist for adherence rates for annual examination of HbA1c, blood pressure, eye examinations, and visits to a dietician. Given the baseline differences, both groups ultimately performed to an equivalent degree. This suggests that especially low-scoring sites were engaged to participate in the QIC, using the intervention to improve their quality of care. It also may suggest that lower baseline status facilitated greater improvement in professional performance at participating sites. Second, for some effect measures, the quality of care might be reasonably good, leading to a ceiling effect at the clinics, with little room for improvement. Third, changing diabetes management may be complex and improving patient care, particularly in a QIC context for 12 months, may generate insufficient robustness to overcome difficult organizational bottlenecks or routines [12]. Perhaps the critical mass of data and sites needed to cultivate useful exchanges of ideas, experience, and learning in the QIC was not reached [13, 28]. Fourth, the specific team or organizational characteristics of the sites may have influenced the effectiveness of the QIC [12, 13, 28]. Although all the intervention sites improved to a certain degree, the specific interventions and the results among sites were heterogeneous. This is no surprise, given the many factors contributing to successful improvement and the likelihood that commitment to improvements, motivations, and mechanisms vary among sites [11, 28, 29]. We assessed key characteristics of teamwork (specific aims, type of changes initiated, degree of participation in our QIC, time, resources, composition, and climate) and organization (size, learning affiliation, and culture), but our study lacks the statistical power to justify a site-by-site analysis and only facilitates an evaluation of the QIC as a general implementation strategy. The influence of these variables therefore remained unclear. Fifth, during the collaborative period, diabetes became a national priority high on the public agenda and received much attention in professional and public media. Although the control sites remained uninvolved in organized quality improvement activities, some individual physicians or provider teams may have implemented small changes independently. Sixth, although we include a wide range of measures and based our measures on internationally accepted indicators of diabetes care, it might be possible that improvements were made, outside the scope of these measurements (e.g. knowledge, skills, teamwork, collaboration). In addition, QIC may have produced changes in care systems that were not large enough to significantly alter clinical processes or outcomes assessed during the evaluation period. Finally, our quasi-experimental study design has some limitations. Although we assessed the outcomes in a before/after design with concurrent controls, this was not a randomized trial. The participating sites volunteered to improve their care, not to be in a trial, so we could not randomize them to participation. Instead, we purposely selected control sites that were comparable to the participating sites. Although some significant baseline differences did exist, study sites characteristics, patient characteristics, structural aspects of diabetes care and patient outcomes did not differ significantly at baseline.

Despite these limitations, this study represents the first controlled evaluation of the collaborative methodology in the Netherlands. Even though the observed difference was modest, considering that a QIC may only be cost effective if the results are maintained, our findings on cardiovascular disease risk control and sustained effects of professional performance measures at two years follow up are promising. A concomitant paper gives an extended description of the cost effectiveness results of this trial [30].

Future research should identify collaborative, organizational and team factors associated with successful improvement to help individual teams and organizations increase the magnitude and pace of improvement.