E-alerts for AKI: How can they improve the quality of care? Dr Nick Selby Consultant Nephrologist and Honorary Associate Professor
National groups ACB scientific committee Met July 2013 Biochemists, nephrologists and software providers Algorithm and minutes available online Renal Association guidelines committee Met October 2013 Nephrologists, biochemists, acute physicians, ICU, patients Ratified algorithm Guidelines to be produced http://www.england.nhs.uk/ourwork/patientsafety/akipr ogramme/aki-algorithm/ British Association Paediatric Nephrologists Met Sept 2013 Paediatric nephrologists, biochemists Ratified algorithm with one adaptation for paeds
Comparisons with existing systems RLH algorithm Number of alerts in patients with baseline creatinine available AKI stage 1 207 212 AKI stage 2 69 68 AKI stage 3 28 49 NHS England algorithm Number of patients with AKI AKI incidence 6.9% 4.5% for stage 1 1.4% for stage 2 1.0% for stage 3 60 50 40 30 20 10 0 30 day mortality AKI 1 AKI 2 AKI 3 Trevor Hine, Shahed Ahmed: personal communication
Terminology e-alerts Detection Alerts
One in five emergency admissions to hospital will have AKI AKI is 100 times more deadly than MRSA infection Around 20 per cent of AKI cases are preventable costs of AKI to the NHS are 434-620m pa
NCEPOD report published in 2009 Poor assessment of risk factors for AKI and acute illness Delays in recognising AKI Most patients with AKI are not cared for by nephrologists Post admission AKI avoidable in 21% Good care in <50% cases
Aims Electronic detection Systematically identify cases of AKI Hospital wide Rapidly report cases to responsible clinician
Can alerting systems effective? ICU setting 2593 alert messages Alert sent by mobile phone Increased resolution of AKI (RIFLE class R) after alert Colpaert et al. Crit Care Med 2012; 40: 1164-70
Electronic detection Aims Systematically identify cases of AKI Challenges No immediate IT solution to apply AKIN or KDIGO criteria Hospital wide Rapidly report cases to responsible clinician No clear definition of baseline creatinine Joint working needed with nephrology/biochemistry/it Funding
AKI detection in Derby In clinical practice at RDH since April 2010 Combination of human and IT algorithms Based on AKIN serum creatinine criteria; modified Jan 2013 to KDIGO Winner of BUPA Foundation Technology for Healthy Outcomes Award 2012 Selby NM et al. CJASN 2012; 7(4): 533
AKI detection Education programme Intranet Guidelines Care bundles Streamlined nephrology referral
Educational programme Collaboration between Royal Derby Hospital and University Hospitals Leicester Joint funding from East Midlands HIEC Initiated April 2011 Project team Prof Sue Carr (project lead) Dr Nick Selby (project lead) Dr Rachel Westacott Dr Richard Baines Dr Nitin Kolhe Dr Gang Xu Dr Salman Riaz Joanne Kirtley James Trew Winner of BMJ Excellence in Education Award 2013
Components of the Educational Toolkit Large group Teaching Ward based E- learning http://www.uhl-library.nhs.uk/aki
Educational outcomes 457 clinicians surveyed (319 at baseline, 138 post intervention) Improvements seen in self-reported indicators: Confidence levels 50% vs. 68%, p<0.001 Independent initiation of investigation and treatment 48% vs. 64%, p=0.002 Awareness of local AKI guidelines 25% vs. 64%, p<0.001 (Combined data from RDH and UHL) Improvements in knowledge scores in junior doctors (F1/F2) Xu G, Westacott R, Baines R, Selby NM, Carr S. BMJ Open 2014;4(3)
AKI care bundle Introduced to assessment units in 2011 Targets systematic improvements in basic elements of care Consistent with intranet guidelines
% completion within 24hrs Impact of interruptive alert Increase in bundle compliance Patients with completed bundles had better outcomes 25 20 completion completion <24 hrs >24 hrs No. 15 of patients 306 (12.2%) 2194 AKI progression 2.9% 6.8% 0.01 10 p-value In-hospital 18% 23.1% 0.046 mortality 5 30-day mortality 25.2% 28.5% 0.219 0 Pre-interuptive alert p<0.001 Post-interuptive alert Kolhe et al, submitted NDT 2014
Impact on standards of basic care * * * Cases note audit of 306 pts. 132 cases baseline 156 cases post intervention 77 in 2012 audit, 79 in 2013 audit Equal numbers in each AKI stage *p<0.001 Baseline 2012 2013 p value Fluid balance assessed 36.4% 66.2% 79.7% p<0.001 Medication review 71.1% - 88.4% p<0.001 Renal imaging (AKI 2 & 3) 45.3% 54.2% 71.0% p<0.001 Nephrology referral (AKI 3) 37.8% 56.5% 78.9% p<0.001 Urinalysis performed 40.3% 57.1% 35.5% p=0.177
Impact on patient outcomes n=8411 Unadjusted 30-day mortality: Sep10-Feb11: 23.7% Mar11-Aug11: 20.8% Sep11-Feb12: 20.8% Mar12-Aug12: 19.5% Chi square for trend p=0.006 No differences in LoS or rate of renal recovery Selby NM. Curr Opin Nephrol Hypertens. 2013; 22(6): 637-42 Cox regression Hazard ratio 95% CI Sep10-Feb11 Reference Mar11-Aug11 0.9 0.79-1.0 Sep11-Feb12 0.87 0.77-0.99 Mar12-Aug12 0.81 0.71-0.93
Impact on progression to higher stages of AKI p=0.03 vs. Q1 p=0.07 vs. Q1 * p=0.016 vs. Q2 ** p=0.031 vs. Q2
AKI risk score Unselected medical admissions caki excluded Risk score to ID those at risk of developing AKI within 7 days of admission Forni L et al. Nephron Clin Prac 2013; 123(3-4):143-50
Alerting for risk? Acknowledgements: Sian Finlay and Mike Jones
Siting the algorithm in LIMS: Standardisation Data transfer to UKRR Not restricted to secondary care in future developments Laboratory quality assurance Sustainability
Measurement can drive improvement Message Master patient index Regional National Research Patient management system LIMS level result AKI Registry QI Local systems Other data systems Alert Response
Conclusions Electronic detection of AKI addresses an obvious clinical need Configure other interventions and method of alert messaging for maximum benefit There are challenges but also huge potential to national approach