AKI bundles and measuring AKI outcomes…. What should we be focusing on?

At my hospital I have been collecting inpatient mortality data for non-elective admissions, for over two years now and I’m able to compare mortality rates for patients with and without AKI.  I suspect a similar aim of the UK Renal Registry and Think Kidney’s AKI programme is to provide similar data for hospitals and regions.  One aspect that I have found interesting is that in the whole of 2017, the average (incl. AKI 1-3 patients) AKI inpatient crude mortality rate for non-elective admissions ran at around 16% vs. 1.4% for similar patients who did not have or develop AKI.  In December 2017, the mortality rate peaked from 16.5% in October 2017 to 28.8%! 

This was a huge spike, but not altogether unexpected in seeing first-hand, how busy the NHS was last Winter and the severity of illness that patients were presenting with.  One of my first blogs covered this topic in some detail.  This mortality spike seen in December 2017 persisted way beyond the expected ‘winter-pressures’ with mortality not falling below 22% until May 2018 when it dipped down to early 2017 levels at 15.3%.  Patients without AKI during this time had a mortality consistently around 1.6% with no such fluctuation seen.  The ‘normalised’ AKI mortality of 15.3% remained throughout the summer, but unfortunately spiked again at the end of September 2018 to 21.6% vs. 1.4% without AKI and remained around 20% into the Winter months. 

The reason for highlighting this average mortality fluctuation, is that AKI seems to have huge variation in its crude mortality trends, which I suspect is not limited to just my hospital but UK-wide.   This makes study design extremely difficult as showing any effect (hopefully beneficial) from interventions like AKI bundles is a real problem, particularly when AKI mortality can fall 10% at the drop of a hat!

However, all is not lost!  Over the last year, there has been a number of AKI and Fluid management interventions made by my hospital including the introduction of an electronic-observation system (Vital-PacTM), with ward online fluid management documentation and integrated AKI e-alerts to complement the Trust AKI bundle.  This and daily automated new AKI patient lists, for teams such as outreach and pharmacy to use, allow almost all new AKI cases to be seen in addition to medical team review and acted upon with bundle interventions started within 24h of AKI e-alert.   It is difficult to confidently show a significant reduction in AKI mortality outcomes from crude mortality rates alone.  However, if one is to look at the Hospital Standardised Mortality Ratio (HSMR) for AKI, our hospital has shown a significant reduction in AKI mortality over the last year (September 2017 to October 2018) since the introduction of AKI interventions, with HSMR falling well below UK-average.  

The HSMR scoring system works by taking a hospital’s crude mortality rate and adjusting it for a number of factors including population size, age and poverty.  From this it is possible to calculate two scores, the expected mortality rate for any given hospital and actual observed mortality rate.  It is the difference between expected and observed mortality that is important when considering HSMR.  HSMR for a given condition such as AKI, allows mortality rates between different hospitals and regions to be statistically compared and is used frequently as a mortality comparator between Trusts (see Dr Foster Intelligence).    When comparing similar sized hospitals (regional peers) our AKI mortality is comparable and one of the better performers, which is really reassuring that what we are doing as a Trust is having a positive impact on patient mortality.  I believe that AKI e-alerting and the Trust response to AKI presence through the AKI bundle and staff intervention is key and has really made this difference possible and something my Trust should be proud of!   

It is clear from this, that crude hospital AKI inpatient mortality is probably far too insensitive to be a useful improvement measure.  It is influenced by too many patient and hospital factors!  However, standardised HSMR is potentially a more robust measure of AKI mortality and may be something for Trust’s to look at more closely when trying to benchmark improvement. 

For other AKI outcomes, I believe that some refocus is needed.  For example, when looking at non-elective inpatient mortality in my Trust.  This group of patients due to their emergency presentation are at least 4x at risk of AKI versus elective admissions, they are more likely to be medical patients with a large majority being over the age of 65years.  Consistently 50-60% of these patients develop AKI in the community and present with it on hospital admission.  They are an extremely heterogenous group with many different causes of their AKI, with sepsis and hypovolaemia being most often to blame.

So how do we refocus our outcome aims? 

Patients who develop AKI in the community (e.g. Community-Acquired AKI) before hospital admission are in my view a different cohort of patients to those who did not have AKI on admission but then developed it whilst in hospital (e.g. Hospital-Acquired AKI).  Such groups could then be subdivided further into non-elective and elective admissions.  All groups are important to tackle, but it is possible that those who develop AKI in hospital may have been more preventable than those who developed it in the community or potentially have more learning points to consider and intervene on.  Who knows, this is just a hunch! It is important to investigate and consider whether this is indeed true, as it may involve different preventative strategies and approach for the community versus the hospital.  One size does not always fit all!  Many Critical Care studies in the past have failed when a strategy has been applied to the whole critical care cohort rather than focusing down on specific subgroups of patients, where certain interventions are of benefit, making a true real difference!   If NCEPOD is to be believed, only 30% of all AKI cases in hospital were deemed preventable, with the vast majority of patients developing AKI due to severity of their disease process regardless of what preventative measures were employed.  However, hospital AKI bundles may be more effective in those who have hospital-acquired rather than community-acquired AKI with such interventions improving time taken for AKI resolution and peak severity.  We may not be able to prevent AKI development in the majority, but our interventions could make a difference in bringing about prompt resolution!  Ultimately, it is hoped that such subgroup improvements could influence favourably overall mortality and length of stay. 

There is still a lot of work to do in the face of increasing scrutiny from hospital management of AKI interventions and effectiveness versus the cost of providing such resources.  We constantly need to justify why such initiatives are needed!  AKI HSMR and focusing on AKI subgroup outcomes may indeed add to the growing evidence for AKI bundle use, providing much needed proof, that when employed with AKI e-alerts and other interventions, AKI bundles do really make a difference to our patient’s lives and hospital outcomes! 



1.  Dr Foster Intelligence:  https://www.drfoster.com/service/quality-and-outcomes-measurement/#product-mortality-comparator

Your comments are welcome on this site, why not have your say on our forums?