Risk Prediction for Acute Kidney Injury in Acute Medical Admissions in the UK (2019)

Type of publication:
Journal article

Author(s):
The RISK investigators [including *Chand, S]

Citation:
QJM: An International Journal of Medicine, Volume 112, Issue 3, March 2019, Pages 197–205

Abstract:
Background
Acute Kidney Injury (AKI) is associated with adverse outcomes; therefore identifying patients who are at risk of developing AKI in hospital may lead to targeted prevention.

Aim
We undertook a UK-wide study in acute medical units (AMUs) to define those who develop hospital-acquired AKI (hAKI); to determine risk factors associated with hAKI and to assess the feasibility of developing a risk prediction score.

Design
Prospective multi-centre cohort study across 72 AMUs in the UK.

Methods
Data collected from all patients who presented over a 24-h period. Chronic dialysis, community-acquired AKI (cAKI) and those with fewer than two creatinine measurements were excluded. Primary outcome was the development of h-AKI.

Results
Two thousand four hundred and fourty-six individuals were admitted to the seventy-two participating centres. Three hundred and eighty-four patients (16%) sustained AKI of whom two hundred and eighty-seven (75%) were cAKI and ninety-seven (25%) were hAKI. After exclusions, chronic kidney disease [Odds Ratio (OR) 3.08, 95% Confidence Interval (CI) 1.96–4.83], diuretic prescription (OR 2.33, 95% CI 1.5–3.65), a lower haemoglobin concentration and elevated serum bilirubin were independently associated with development of hAKI. Multi-variable model discrimination was only moderate (c-statistic 0.75).

Conclusions
AKI in AMUs is common and associated with worse outcomes, with the majority of cases community acquired. Only a small proportion of patients develop hAKI. Prognostic risk factor modelling demonstrated only moderate discrimination implying that widespread adoption of such an AKI clinical risk score across all AMU admissions is not currently justified. More targeted risk assessment or automated methods of calculating individual risk may be more appropriate alternatives.

Altmetrics