how to cut the incident AF stroke rate by 15% in 18 months. Turning science in to change.

 

 

 

 

This is a very old story. Its also a very long one.

It is one I have wanted to tell for a time. It is the story of how we increased performance by 33% on a “difficult indicator” that was known to be problematic to change.

It is also the story how we did it very cheaply without investing in lots of services but we used simple techniques from improvement science and a lot of enthusiasm among a large body of GPs.

 The driving force for this one was Maciek – he knows who he is, @fatherofhan and yours truly. 

 

WARNING – this is based on some slides I often use, and an unpublished manuscript. I lost the will to live trying to get it published in a proper science journal and whatnot. Apply your judgement in terms of the caveats re data.

 

 

introduction

Atrial fibrillation, or AF, is a heart rhythm disorder that increases the risk of stroke five-fold. The risk of an AF-related stroke can be substantially reduced through anticoagulation therapy. There are numerous clinical guidelines which outline the case for utilisation of anticoagulation in AF patients however there remains significant under implementation of anticoagulation in AF patients

Within Bradford, the same extent of under treatment rates was found as in the national GRASP AF analysis.

12

 

 

 

6

 

 

Whilst there are many clinical guidelines , little is known about how the best methods to improve population outcomes and systems of care in AF or other clinical conditions. Data distribution alone is insufficient to instil change

 

9

 

What we did

We implemented a model of achieving change and redressing a well documented quality gap in a large population building on previous quality improvement work undertaken in renal care. We used the methods we had previously developed to encourage improvement using a voluntary approach that complemented the existing QOF framework

10

 

 

 

 

11

our aim

The objective of this initiative was to work collaboratively with volunteering Bradford practices to improve uptake of anticoagulation within patients diagnosed with AF.

 

12

 

Methods

The implementation strategy was based on simplicity, the collaborative breakthrough series was used as a model for improvement. This type of model has been demonstrated to have achieved significant health gain in other health care systems.

All 80 practices in Bradford were invited to participate. 56 did, a population of c330,000

 

14

ACTIVE intervention.

ALOT more than just sending out a guideline

15

 

 

Results

56/80 practices participated,

Both of the local hospital providers also participated in this (mainly in terms of anticoagulation monitoring service) participated,

Over a period of 18 months, the number of patients on the practice databases with a diagnosis of AF, a CHADS2 score of 1 or more, being prescribed warfarin increased from 2,274 to 2,978, an absolute increase of 714 additional patients, a relative increase of 31%, with the increase being highest in those with the highest risk scores.

 

 

1617

Impact on primary outcome

The initiative achieved a significant increase in numbers of patients with AF who receive anticoagulation. During the 18 month course of the project an additional 714 additional patients became anticoagulated. This represents a 31% relative improvement on the baseline and is compared to a 4% change in non participating practices.

We conducted similar data queries in neighbouring CCG areas to give a sense of what is happening in a “control group” area. It is possible that due to AF07 being implemented mid way through our project, there will be an increase in anticoagulation in the control group.

Even taking into account the improvement in the non participating practices we contend that this scale of improvement in a difficult to shift indicator is unusual. The improvement in anticoagulation rates was seen across all CHADS2 strata, with the greatest improvement being in those patients at highest risk. This is encouraging, in that the greatest improvement was seen in those most at risk.

18

 

19

Prior to a more detailed analysis some simple calculations, based on NNT from published studies, were undertaken to estimate impact. These indicated that with this scale of improvement we would expect to have prevented 29 strokes and 17 deaths.

 

economics

We have not conducted an economic analysis. The evidence is clear that improving rates of anticoagulation is highly cost effective (with likely, thought not calculated, cost effectiveness ratios that are below £5000 / QALY ore perhaps more likely cost saving in net terms. It is most likely cost neutral intervention and may be cost saving (particularly if targeting those at highest underlying stroke risk). Adding the cost of project implementation and INR monitoring to this equation we estimate that the project is cost neutral at very worst.

20

One of the concerns prior to commencement was that adding a large volume of patients into INR monitoring relatively quickly might have overwhelmed those services, and had a detrimental impact on INR control in the population of patients anticoagulated. This proved to be unfounded. Data extracts throughout the project demonstrate no change in the proportion of tests within range, or if anything a minor increase

 

 

Equity

Equity is an important consideration. There is often a tendency to assume that providers in the most affluent parts of a district are most likely to participate, thus exacerbating inequalities. There is no evidence of that in Bradford. Participation was from practices across the district. Furthermore, in our other QI projects the absolute improvement in the project indicators was greatest in those with more deprived populations was higher than those with affluent populations, thus indicating an improvement of inequality.

21

 

22

 

 

23

 

Learning

Learning points

We have demonstrated an approach to a process and system innovation that has allowed us to make a significant difference in quality in a large population, and has real traction with primary care. We have learned an approach to make a substantial difference to an intervention that is based on both sound clinical and epidemiological evidence of the under implementation of this effective intervention. We have adopted the best available methods of improvement and applied them to 64 separately operating practice teams, representing a population of c350,000.

Collaboration was supported by a facilitator who supported the spread of good practices and sharing of ideas between practices. The underpinning philosophy is that no single person or discipline has all the answers or knows how to address the problem, but collectively as a system we can make substantial progress. The offer as entirely voluntary, there was no financial incentivisation for GPs (beyond that in the QOF) and participation was based on strong clinical leadership. It has enabled us to engage a large General Practice population on the quality agenda in a short period, and we have achieved a significant shift in outcomes in that time.

Also critical to our approach is the notion of population impact, large scale change does NOT happen through engagement of small numbers of practices, it is necessary for us to work with large number of practices (and hence population) to achieve a population shift.

Fundamental to this shift was a high quality, locally accessible and high performing INR monitoring service. This services needed to have the capacity to take on large volumes of new patients. Our observation is that a locality / community based model seems more likely to enable us to achieve the shift in anticoagulation

 

 

Success factors in implementation

  • strong clinical and PH Leadership. visible and LOCALLY credible opinion formers and leaders to lead
  • Ruthless and meticulous implementation
  • A small number of locally agreed high impact and measurable indicators
  • a clear approach to peer facilitation, recognising that practices had as much to teach each other as “experts” had to teach them
  • Benchmark live data on achievements against those indicators across all participating practices. This encourages competition within a system on quality metrics – striving to be the best.
  • Single side guidance for clinicians, broader suite of tools embedded in primary care IT system to enable better and more standardised practice. Applied to large population over long time period.
  • Regular feedback on achievement – with data and softer messages. Active evidence based strategies were consistently applied to the practices that were participating to encourage improvement
  • Bespoke support and advice to practice and more widely – Q&A / Expert events / training / Practice visits / IT tools

24

 

 

26

27

We learned that the leadership of the innovation relies on a multi disciplinary team with clinical and public health leadership but with active support from a wide range of disciplines. It is not, however, rocket science – simply hard work and sustained implementation of evidence based clinical behaviour change strategies. This model does rely on enthusiastic individuals with a common goal. We have developed a simple and effective model for QI in primary care that primary care really engages with. There has been consistently positive feedback from practices and those that didnt initially participate are now requesting to do so. The work has enabled practices to make a substantial shift in an important and highly clinically relevant process indicator in a short period, this is an indicator that has been historically difficult to shift.

28

These data demonstrate that it is possible to increase the utilisation of effective anticoagulation therapy in patients at risk of stroke due to atrial fibrillation. Based on Numbers Needed to Treat from published studies the scale of improvement demonstrated would prevent 29 strokes and 17 deaths. We also noted, c18months after implementation there WAS a c15% fall in the incident AF stroke rate.

 

 

 

Relevance

Applying the results achieved in Bradford to national efforts would make a remarkable and substantial shift in outcomes. We have demonstrated that there is a credible, cheap, simple and easy to implement methodology for primary care to improve quality and outcomes.

We believe we have demonstrated this approach to improving quality has made a difference, in line with similar approaches based on data, benchmarking and active support to frontline clinicians. However, spread it critical. Spread is the greatest challenge. Without sustainability and spread, both to broader geography and to other clinical areas, this will be “just another pilot study”.

 

Strengths

This was a real world quality improvement project, run in a large population in the context of a large number of competing initiatives and change. It is something that is highly replicable, and has a strong body of support and good basis in evidence[i] [ii]. It is an intervention that is cheap to implement and can be readily applied to many other clinical areas.

 

 

Limitations

This project was not run as research project. There are obvious statistical issues in the analysis and the comparators.

The secondary outcome data is limited by what can and cant be concluded from administrative data. There are a number of well documented underlying data quality issues.

For reasons set out above, we did not try to quantify bleed rate, though we acknowledge it is an important metric.

Furthermore, there are issues in trying to quantify the impact of preventive interventions (counting what doesn’t happen, the importance of the baseline – was it is high or low year, the time between risk reducing exposure – anticoagulation – and impact – stroke reduction), and issues about the durability of the intervention (is the risk reduction permanent, the durability and survival of anticoagulation in this cohort of patients). Finally there may be limitations in terms of the difference between absolute risk reduction seen in trials versus what might be achievable in real life. This has been commented on by others. It is reassuring to note that our expectations of “number of strokes reduced (based on published NNTs) and that seems to have been measured for real is reasonably close.

What we did is a reflection of what can be done in improving a system and comparing our system to other areas

 

Spread. Spread remains one of the greatest challenges for Quality Improvement. Without sustainability and spread, both to broader geography and to other clinical areas, this will be “just another pilot study”. Constancy of purpose is important. The NHS needs to be clear in their expectations as to this improvement being the norm and that it cannot wait out this “flavour of the month”. It is important to have a realistic understanding of change fatigue and how much process improvement the organization can do at once.

Here we deliberately focused on “the masses” rather than the “best performers”. Often an assumption is made that   “if you improve the leading edge, the rest will follow”. We propose that whilst this might be true this approach will not achieve population shift at the same level as setting achievable targets for mass improvement.
We think that a visual display of performance of the system really helped motivate change, especially where there is real time shift that can spur further action.
We would propose that the creation of half-life type goals rather than finite targets will be important in sustaining long term improvement. This will embed the notion that the system  does not become complacent once a target has been achieved.

 

Conclusions

The NHS outcomes framework, and the CVD outcomes framework has given a clear mandate to the NHS to make improvements. There remains a significant gap in this area between current and best practice, this is an area that can have impact on CVD outcomes in a short timeframe. More RCTs and more guidelines will clearly be insufficient to further close this care gap. Other strategies are necessary. The effectiveness of those strategies remains under researched, however there is increasing consensus in the quality improvement literature on likely targets.

Our approach was similar to that advocated by world leaders in quality and safety, we explicitly focused on some of the reasons why existing and well publicised guidelines are under implemented. We directly addressed areas where there is disagreement, we simplified guidelines so as they influence decisions at the point of care, we also disruprted the status quo by providing comparative performance data. We relentlessly focused on population based care, as opposed to focusing on individual clinicians and the patient / clinician interaction. This is the first time to our knowledge that such a large scale improvement has been reported in this clinical area anywhere in the world. We have tried to carefully document how we achieved this, in order that others will follow and achieve similar gains.

29

 

 

 

postrcipt 1 – sustaining change is hard

We were happy we had done a good job. We then went off to do other things and took our eye off this ball. Looking back 2 years after we finished, AC levels had slipped significantly back.

Lesson = constant attention needed.

postscript2  – Comparison between the clinical search query and QOF AF04 and AF07

At end of project, in participating practices the final data extract indicated that 65% of patients with CHADS2 ≥1 on Warfarin.

There were some changes to the QOF framework within this time period. AF07 was introduced.. Not sure this will have been powerful enough to have had too much influence. We do have data on what was going on with those practices that chose NOT to participate. From memory there was a 31% relative change in practices that DID participate and a 4% change in those that didn’t.

Direct comparisons with QOF are thus tricky. Also we didn’t allow exceptions.

 

Advertisements

4 thoughts on “how to cut the incident AF stroke rate by 15% in 18 months. Turning science in to change.

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s