Real-world data analysis of delirium: examining hospital reporting rates, biomarkers, and health outcomes using routinely collected health data

Abstract

BACKGROUND: Delirium is a serious, often fluctuating neuropsychiatric disorder, mainly triggered by acute illness outside of the brain, and characterised by disturbances in attention, awareness, and cognition. Onset is typically hours to days, with median duration 3-5 days, but it may persist for weeks to months. Delirium is associated with multiple adverse outcomes, including increased mortality risk, and future risk of dementia. It affects 1 in 4 older hospitalised adults but is often not detected. The Four A’s Test (4AT) is a well-validated, rapid screening tool for delirium that evaluates alertness, orientation, attention, and acute change or fluctuating course. The score range is 0-12, and a 4AT score of 0 suggests no cognitive impairment or delirium, a score of 1-3 suggests possible cognitive impairment without delirium, and a score of ≥4 indicates delirium ± cognitive impairment. Several hospital centres have recently begun to integrate the 4AT into their electronic medical record (EMR) systems. This development allows new opportunities for large-scale routine data studies of delirium which, previously, have been scarce in the field. Topics include exploring health outcomes after delirium, and mechanisms underpinning delirium as indexed by routine laboratory inflammatory markers.

This development is significant given the growing recognition of inflammatory processes as important in delirium. The aims of the thesis were therefore to: 1) study the documentation and coding rates of delirium in routine discharge summaries and hospital administrative systems; and using EMR data for exploratory analyses; 2) analyse the relationship between inflammatory markers and delirium, determined by 4AT; and 3) analyse the short- and long-term health outcomes experienced in patients with and without delirium (based on 4AT status), exploring relationships with clinical variables including inflammatory markers.

METHODS: For aim 1, I conducted a systematic review to analyse varying rates of delirium documentation and coding in the published literature. I also used publicly available hospital coding data to examine trends in delirium coding rates in older people using hospital discharge data in two national population samples. For aims 2 and 3, I analysed retrospective data from the University of Edinburgh’s DataLoch EMR repository (www.dataloch.org). The analyses focused on consecutive emergency hospital admissions of patients aged ≥65 years and older in the Lothian region of Scotland spanning from January 1, 2016, to October 3, 2022. These data formed the Delirium Routine Data in Edinburgh (DELRED) cohort.

Delirium status was derived using 4AT cut-off scores (4AT 0, 4AT 1-3, and 4AT ≥4). Patients with missing 4AT were included and coded as 4AT missing. Emergency admissions from NHS Lothian during the study period were linked to a comprehensive array of clinical data, including hospital readmission data, laboratory results, mortality records, and diagnostic information. Descriptive statistics were reported using median with interquartile range (IQR), or mean with standard deviation for normally distributed data, and group-wise comparisons were analysed using Kruskal-Wallis, ANOVA, and Pearson’s chi-squared tests as appropriate. Health outcomes were evaluated using logistic regression models (adjusted for age and sex), and survival probabilities using Kaplan-Meier plots.

RESULTS: The systematic review of 24 studies found that delirium was significantly under-reported in discharge summaries and hospital administrative systems, with delirium reporting rates ranging from 0.1% to 64%. The national delirium coding trends study showed that delirium coding rates significantly increased in older hospital inpatients in England and Scotland from 2012 to 2020, with a 3 to 4 -fold increase observed. A total of 68,424 patients were included in the DELRED cohort. The median age was 77 years (IQR: 71-84 years), and 54% of the cohort were female. White patients comprised 74% of the cohort.

The median time reported to 4AT assessment was approximately 3 hours (IQR: 1.37-9.83 hours) and most (86.63%) assessments were completed within 24 hours of admission (p<0.001). Across the 4AT groups, 4AT ≥4 had the highest prevalence of comorbid dementia (N=2,542, 36.38%). Thirty-day inpatient, one-year and three-year mortality was highest in 4AT ≥4 (18.54%, 43.01% and 59.98%, respectively). Length of stay more than doubled in 4AT 1-3 and 4AT ≥4 compared to 4AT 0 (p<0.001). Over a median follow-up time of 2.4 years, a new dementia diagnosis following index admission was highest in 4AT 1-3 (20.66%), followed by 4AT ≥4 (18.98%, p<0.001), to 4AT 0 (5.89%, p<0.001). The median time to a new dementia diagnosis was three times shorter in 4AT ≥4 (252 days) compared to 4AT 0 (750 days) and roughly 1.5 times shorter in 4AT 1-3 (343 days, p<0.001).

Several inflammatory markers were associated with delirium status, such as ferritin and neutrophil counts (p<0.05). The median neutrophil:lymphocyte ratio (NLR) was higher in those with 4AT ≥4 than those with 4AT 0 (6.32 vs 5.04, respectively, p<0.001). In adjusted subgroup analyses exploring relationships with mortality in 4AT ≥4 (delirium), clinical variables that predicted the odds of 30-day inpatient mortality in this subgroup included an NLR >5 (adjusted hazard ratio [aHR] 13.40, 95% confidence interval [CI]: 1.76-96.6, p=0.010), and abnormal white cell count (aHR 5.46, 95% CI: 1.27-23.4 p=0.022).

DISCUSSION: Despite improvements due to policy initiatives, detection tools, and education, in delirium documentation and coding, there is still a significant gap in the accurate reporting of delirium, suggesting a need for continued efforts in healthcare. Delirium, measured by the 4AT, is strongly associated with adverse patient outcomes. The group of patients in the present dataset who did not receive 4AT assessment likely include very vulnerable and severely ill group of patients, with their severe conditions probably causing both missed 4AT assessments (perhaps due to perceived more urgent medical needs being attended to by healthcare staff) and higher 30-day inpatient mortality. Patients with delirium (4AT ≥4) had elevated inflammatory signatures compared to patients with no delirium. These findings provide support for the role of inflammation in delirium Additionally, in patients with 4AT ≥4, abnormal white cell count and NLR >5 were strong predictors of 30-day inpatient mortality, suggesting the possibility that increased inflammatory markers is linked with worse outcomes in patients with delirium. This finding requires further study both in relation to short-term and longer-term outcomes.

This is first large-scale study using routinely collected health data to explore inflammatory determinants of mortality outcomes in relation to 4AT-ascertained delirium. Future data linkage analyses of this extensive cohort could yield valuable insights, essential for generating new hypotheses and advancing research with significant clinical relevance to delirium risk factors and outcomes. For example, future research could examine other routine laboratory markers beyond inflammation.