Automated alerts for acute kidney injury warrant caution
BMJ 2015; 350 doi: https://doi.org/10.1136/bmj.h19 (Published 07 January 2015) Cite this as: BMJ 2015;350:h19
- Simon Sawhney, clinical lecturer in nephrology, Division of Applied Health Sciences, University of Aberdeen, Aberdeen AB25 2ZD
- simon.sawhney{at}abdn.ac.uk
As an academic renal trainee and apprentice of acute kidney injury (AKI), I have had the privilege of learning my art in parallel with a paradigm shift in diagnosis. In a decade the new term AKI has transformed our awareness of a serious condition that can happen anywhere in the NHS.1 2
AKI is defined by changes in serum creatinine from a previous baseline over a short period (28 μmol/L in 48 hours, or 50% in one week), and it is measured in three stages of severity depending on the magnitude of change. Even small changes in vulnerable patients can signal increased mortality.3 A national inquiry, Adding Insult to Injury, reminds us that late intervention can lead to preventable death, and we could all do better in our efforts to provide prompt and safe care.4 But our desire to recognise and treat patients early must be balanced with the harms of overdiagnosis and overtreatment.
A promising new development is the imminent mandated introduction to all hospitals in England of an automatic system to detect AKI.5 Using an algorithm modified from international criteria,1 a warning that AKI is suspected can appear automatically with a serum creatinine result. This may prompt early reassessment, close monitoring, or hospital admission. It may also be linked with clinical guidance and a national AKI registry. This could improve decision making and early management in a condition where most patients are cared for by non-specialists. But a new gadget should be stress tested for unintended effects and should come with a disclaimer against incorrect use.
An umbrella of pathophysiological states
AKI is not a single condition but rather an umbrella of pathophysiological states and circumstances that have led a person to cross a threshold and acquire new labels. Just as one extra grain of sand does not make a heap, logically one extra micromole of creatinine in each litre of serum does not make an AKI.6
Nevertheless, some creatinine changes will in time lead to some patients being judged as lying uncomfortably on the wrong side of a cut-off. This betrayal, without discernment, may risk misinterpretation or may be coupled with a diverging road in clinical decision making and unintended consequences. To add a further challenge, it is not even possible to apply the AKI diagnostic rules to a change in creatinine without some knowledge of the previous baseline, yet this is often unknown and assumed. The result is an abundance of definitions for AKI in clinical research, stemming from attempts to traverse a void of missing data.
So, we know in broad terms that patients with AKI have poor health, but we remain uncertain about how to identify those patients in whom the diagnosis matters most and where the prognostic influence is the greatest or most modifiable. We also remain unclear about the patients in whom we might be liable to overdiagnose disease and those whose disease we might underdiagnose, by using different definitions.
In the Grampian region of Scotland (population 530 000) our renal research group has tested variations of the intended NHS England algorithm for automated detection in a one year extract from our biochemistry database. Just a small change in the baseline definition—comparing creatinine concentration with the lowest rather than the median result from the past year (both are common definitions in AKI clinical research7 8)—increased the number of patients we labelled as having AKI to 5758 from 2546. We are left to speculate who the extra 3212 patients are and what the implications could have been of sending an automated alert or including them in a clinical research study.
Modicum of uncertainty
The widespread and successful implementation of an automated AKI detection system represents a great opportunity to improve further the early recognition and communication of AKI throughout the NHS. New tools, however, can come with unforeseen harms, and we must be careful to lay down clear boundaries as to how a new label of AKI is interpreted by specialists, non-specialists, patients, policy makers, and researchers. For instance, when I carry the on-call renal pager I will try to be proactive and to act early when called about the next patient with AKI—but with a modicum of uncertainty over what the diagnosis really is. Also, I will not reassure myself just because a patient’s results have not yet crossed the diagnostic threshold.
What is the diagnosis? AKI is what we choose to define it as, but we still have a responsibility to understand what it is that we are testing, whether we are focusing on the right people, and whether it is always clinically relevant.
Notes
Cite this as: BMJ 2015;350:h19
Footnotes
Competing interests: I have read and understood the BMJ policy on declaration of interests and declare the following interests: I am supported by a Clinical Research Training Fellowship from the Wellcome Trust.
I thank Corri Black and Angharad Marks for their comments on a draft manuscript.
Provenance and peer review: Not commissioned; externally peer reviewed.