Background: Previous population-based, longitudinal studies have shown that delirium is associated with an increased risk of dementia and cognitive decline. However, the underlying biological mechanisms are largely unknown. We aimed to assess the effects of delirium on both cognitive trajectories and any neuronal injury, measured via neurofilament light chain (NfL).

Methods: In this analysis of a prospective, 2-year follow-up, cohort study of participants aged 65 years or older living in Sandefjord municipality, Norway, we included cohort participants who were receiving domiciliary care services at least once per week between May 12, 2015, and July 8, 2016. Individuals with a life expectancy of less than 1 week, with Lewy body dementia, with psychiatric illness (except dementia), or for whom substance misuse was the principal indication for domiciliary services were excluded. Participants had a comprehensive assessment at 6-month intervals for 2 years, which included the Montreal Cognitive Assessment (MoCA) and a blood sample for NfL to measure neuronal injury. All information on clinical diagnoses and medications were cross-referenced with medical records. During any acute change in mental status or hospitalisation (ie, admission to hospital), participants were assessed once per day for delirium with Diagnostic and Statistical Manual of Mental Disorders, fifth edition criteria. We also measured NfL from blood samples taken from participants who were acutely hospitalised.

Findings: Between May 12, 2015, and July 8, 2016, 210 participants were eligible for inclusion and assessed at baseline (138 [66%] of whom were female and 72 [34%] of whom were male), 203 completed cognitive assessment, and 141 were followed up for 2 years. 160 (76%) of 210 had moderate or severe frailty and 112 (53%) were living with dementia. During the 2-year follow-up, 89 (42%) of 210 participants were diagnosed with one or more episodes of delirium. Incident delirium was independently associated with a decrease in MoCA score at the next 6-month follow-up, even after adjustment for age, sex, education, previous MoCA score, and frailty (adjusted mean difference –1·5, 95% CI –2·9 to –0·1). We found an interaction between previous MoCA score and delirium (β –0.254, 95% CI –0·441 to –0·066, p=0·010.), with the largest decline being observed in people with better baseline cognition. Participants with delirium and good previous cognitive function and participants with a high peak concentration of NfL during any hospitalisation had increased NfL at the next 6-month follow-up. Mediation analyses showed independent pathways from previous MoCA score to follow-up MoCA score with contributions from incident delirium (–1·7, 95% CI –2·8 to –0·6) and from previous NfL to follow-up MoCA score with contributions from acute NfL concentrations (–1·8, –2·5 to –1·1). Delirium was directly linked with a predicted value of 1·2 pg/mL (95% CI 1·02 to 1·40, p=0·029) increase in NfL.

Interpretation: In people aged 65 years or older, an episode of delirium was associated with a decline in MoCA score. Greater neuronal injury during acute illness and delirium, measured by NfL, was associated with greater cognitive decline. For clinicians, our finding of delirium associated with both signs of acute neuronal injury, measured via NfL, and cognitive decline is important regarding the risk of long-term cognitive deterioration and to acknowledge that delirium is harmful for the brain.