There are many reasons for false positive AKI alerts, post-partum being just one of them.

If you were already reporting AKI alerts to primary care your GPs should be used to seeing them, otherwise the advice was not to report to primary care until given the go-ahead, likely 2016.

The current experiences should help inform  where cutoffs are set in future to achieve the best compromise between sensitivity and specificity - just as in any screening programme - it is a flag to review the patient as discussed in the documentation issued with the algorithm and in the NICE guideline. Telephoned alerts should make this clear.

There are also false negatives and the absence of an alert cannot be read as absence of acute deterioration in renal function.

These issues have been discussed at NHS England AKI detection workstream meetings and at the Think Kidneys Programme Team meetings - ACB members all received a letter alluding to this from the detection workstream chairmen on May 11th.

The false positive problem lies mainly in the automated assignment of baseline creatinine - our eye and/or clinical review can weed out the obviously erroneous but our IT currently cannot.

We must be open about the shortcomings and work to improve.

The meeting statement following the algorithm derivation

http://www.acb.org.uk/docs/E-Alerts_for_AKI_meeting_statement

recommended 7c) audit of algorithm performance in live systems at pilot sites to determine the false positive rate.

The subsequent level 3 patient safety alert and delayed delivery of the algorithm to many labs precluded this from happening before going live.

The renal registry has acknowledged that they cannot use the feed data directly as an indicator of AKI incidence and that they are devising means to weed out the obvious false positives.

I imagine few of you were at the recent European Renal Association meeting where Perry Wilson gave us some insights into follow-up studies from his paper in the Lancet 2015;385:1966-1974 showing that AKI e-alerts did not improve outcomes - he believes confounding by false positives in electronic detection was probably responsible for the negative outcome.

Best wishes

Anne

Dr Anne Dawnay PhD FRCPath

Consultant Biochemist

Clinical Lead UCLH Clinical Biochemistry & Chair UCLH POCT Committee

Tel 020 344 72954 direct dial

Date:    Thu, 18 Jun 2015 14:34:18 +0100

From:    "Colley, Michael" <[log in to unmask]>

Subject: AKI Algorithm Again

Just had a GP phone me.

Same scenario - the bloods the patient had over the past year were only while she was pregnant and therefore had a low creatinine.

Her recent creatinine is 86 - more than 1.5 her median

Am I going to be phoned about all 3000 women in Swindon who've had babies in the past year?   'Cos I'm going to be creatinined off...

Michael

Date:    Thu, 18 Jun 2015 13:58:50 +0000

From:    Anders Kallner <[log in to unmask]>

Subject: Michael Colley

This is not surprising. An algorithm essentially describes a 'central tendency' and is therefore irreversible in nature. It is impossible to apply an algorithm to individuals.

All the EGFR algorithms suffer from this undisputable fallacy.

In addition it separates the physician from the patient - diagnosis becomes a number. Aesculapius<https://en.wikipedia.org/wiki/Asclepius> and Hippocrates<https://en.wikipedia.org/wiki/Hippocrates> would turn in their tombs if he knew what is going on in the Queen's England!

Anders Kallner, MD, PhD (R)

Dept Clin Chem

Karolinska University Hospital

SE 176 71 Stockholm, SWEDEN

Phone: +46 8 51774943

[log in to unmask]

Date:    Thu, 18 Jun 2015 15:07:43 +0100

From:    "Salter Simon (ROYAL FREE LONDON NHS FOUNDATION TRUST)"

         <[log in to unmask]>

Subject: Re: Michael Colley

I think the point is that the AKI algorithm is designed to flag up possible AKI and should make a Doctor look at the patient and think for themselves "could this person have AKI?". It is not a diagnosis and never will be. Lab medicine is not trying to replace a Doctor with this algorithm, just help them spot something sooner than if they relied solely on clinical judgement.

Simon

Date:    Thu, 18 Jun 2015 15:15:36 +0100

From:    "Colley, Michael" <[log in to unmask]>

Subject: Re: Michael Colley

I take Simon’s point about its being an alert for the doctor to look at the patient and think.

But it is common experience that GP systems flag up any result the slightest outside a reference range as “abnormal” and the GP believes that to be the case.

I think that Nationwide we’re going to have a lot of phone calls.

And if we get automated links to the Renal registry they are going to get a lot of incorrect data.

Michael

Date:    Thu, 18 Jun 2015 16:14:16 +0100

From:    "Reynolds Tim (RJF) BHFT" <[log in to unmask]>

Subject: Re: AKI Algorithm Again

Our solution has been to have a comment on all reports that all contact about AKI alerts should be made to the AKI alert team, not the lab!

Date:    Thu, 18 Jun 2015 16:15:17 +0100

From:    "Bosomworth Mike (LEEDS TEACHING HOSPITALS NHS TRUST)"

         <[log in to unmask]>

Subject: Re: Michael Colley

I absolutely agree with Simon.

This is early days yet. The GPs will become used to the alert and as they gain experience will be able to interpret it in terms of the patients history etc. If there prove to be too many false positives, then the alert will have to be reviewed, which is the intention in any case.

Best wishes

Mike

Dr Mike Bosomworth

Clinical Service Lead for Blood Sciences and Specialist Laboratory Medicine

Tel: 0113 3922340

Mobile: 07789174344