Acute kidney injury (AKI) is a serious condition where the function of your kidneys drops suddenly. AKI has many different causes which make it difficult to find ways to prevent vulnerable people, such as those with chronic kidney disease, from developing AKI. In some countries, hospitalisations with AKI are higher in the winter, a time when there are many infectious diseases being transmitted between people. This may indicate that infectious diseases are a possible driver of increased hospitalisations of AKI in the winter.
To investigate this, we want to first describe the different types of people getting AKI and their characteristics. We want to know if people with certain types of characteristics are more likely to get AKI in the winter and if this could be related to infectious diseases.
Understanding whether AKI hospitalisations are related to infectious diseases in the winter is important so we can understand if interventions like vaccines may have the potential to prevent infections and therefore also reduce the risk of being hospitalised with AKI in vulnerable people.
AKI is a syndrome defined by rapid decline in kidney function from hours to days leading to disruption in metabolic, electrolyte, and fluid homeostasis. Studies have reported between 20-25% of patients admitted to hospital have AKI and is associated with a 4-16 fold increase in odds of death. The large variation, setting dependent, and chronic nature of many of the risk factors makes it difficult to identify important mechanisms which can be modified to reduce the incidence of AKI. Recent studies have shown that AKI hospital admissions in the UK may have a seasonal pattern, indicating potential associations with infectious disease patterns.
Given the complex and multifactorial aetiology of AKI, understanding the associations between this syndrome and other conditions is well suited for unsupervised machine learning (ML) cluster classification methods. Using ML clustering classifications methods we aim to identify groups of people with certain characteristics (phenotypes) of AKI. This may include groups affected by infectious diseases more common in the winter (such as norovirus, influenza) or groups who have particular comorbidities. Using ML clustering classification methods, we aim to describe AKI phenotypes and whether there is a seasonal pattern of AKI phenotypes attending hospital.
To characterise phenotypes of hospitalised AKI patients, we identify a cohort from Hospital Episode Statistics using ICD-10 codes. CPRD records of the AKI cohort will be extracted and k-means clustering will be conducted to classify patients into clusters. Phenotyping deconstructs and categorises patients into distinct groups identifiable by dominant characteristics and risk factors. These clusters will be defined by their dominant characteristics and described over time.
Phenotyping AKI patients over time generates hypotheses of associations with winter increases in AKI, such as infectious diseases, and therefore identifies possible areas for intervention (for example vaccines) to reduce the risk of being hospitalised with AKI in vulnerable people.
Acute kidney injury diagnosed in secondary care (defined by ICD-10 codes). Code list included in appendix.
Rosalind Eggo - Chief Investigator - London School of Hygiene & Tropical Medicine ( LSHTM )
Hikaru Bolt - Corresponding Applicant - London School of Hygiene & Tropical Medicine ( LSHTM )
Frank Sandmann - Collaborator - London School of Hygiene & Tropical Medicine ( LSHTM )
Laurie Tomlinson - Collaborator - London School of Hygiene & Tropical Medicine ( LSHTM )
Frank Sandmann - Collaborator - London School of Hygiene & Tropical Medicine ( LSHTM )
HES Admitted Patient Care;Patient Level Index of Multiple Deprivation