Delirium is an acutely occurred and short-lived syndrome, characterized by fluctuating changes in attention, level of consciousness, and cognitive function (1). It is a common complication after major surgery, especially in older patients (2). According to a recent systematic review, the reported incidence ranged from 4% to 46% in patients aged ≥60 years following noncardiac surgery (3). The occurrence of postoperative delirium is associated with increased morbidity and mortality (4), prolonged stays in the intensive care unit (ICU) and hospital, and worsened functional recovery (5, 6). Studies showed that postoperative delirium mostly occurs in the first few days (7–9). Various diagnostic tools, such as the Confusion Assessment Method (CAM) and the CAM for the intensive care unit (CAM-ICU), have been introduced and validated to facilitate delirium screening (10–12).

The detection of delirium can be challenging due to the fluctuating nature and frequent hypoactive presentation. Interview-based methods, i.e., delirium assessed by trained interviewers during brief encounters, is frequently used in clinical studies and daily practice (13). However, it is still unknown how often should the assessments be done and how long should the evaluation period be. A previous study reported that daily assessment is not enough because cases occurring during night-shift might be missed (14). In a recent study of 788 patients following cardiac surgery, delirium was detected more often in the mornings than evenings, and CAM-ICU assessment twice daily for 4 days detected an estimated 97% of postoperative delirium (15).

In this analysis, we aimed to evaluate the effect of different strategies in detecting delirium during the first 7 days after non-cardiac surgery, and to determine the most favorable strategy that detected delirium with high sensitivity and low effort.

A total of 700 patients were enrolled in the underlying trial. The average age of enrolled patients was 74 years, with 60.4% were male. Patients who experienced delirium were older, had lower body mass index, suffered more previous stroke, had lower preoperative serum albumin, were more frequently admitted to ICU with intubation, and received less low-dose dexmedetomidine when compared with those who did not (Table 1).

Of all enrolled patients, 111 (15.9%; 95% CI: 13.3% to 18.8%) developed at least one episode of delirium during the first 7 postoperative days. Among these, 60.4% (67/111) had their first delirium onset on postoperative day 1, 84.7% (94/111) by the end of day 2, 91.9% (102/111) by the end of day 3, 99.1% (110/111) by the end of day 4, and 100% (111/111) by the end of day 5 (Table 2).

We compared the sensitivity of different strategies in detecting first postoperative delirium compared with twice-daily assessments for 7 days (Table 3). There was no significant difference in the sensitivity of delirium detection with morning only or afternoon only assessments (Figure 1). Both of the above strategies detected 83 (11.9% of 700) delirium patients during the first 7 postoperative days. Twice-daily measurements for 4 days detected 99% (95% CI: 94% to 100%) of delirium patients with 57% efforts. Similarly, twice-daily assessment for 3 days detected 92% (95% CI: 85% to 96%) of delirium patients with only 43% efforts.

In the present study, delirium occurred in 15.9% of our patients during the first seven days after non-cardiac surgery. The rate was relatively lower than previously reported incidence (24.4% to 44.5%) in similar patient populations (20, 21). This could be explained by the fact that half of our patients received low-dose dexmedetomidine infusion during the night after surgery, a regimen that has been proved effective in decreasing delirium in high-risk patients (5, 22, 23). Another possible reason was that multiple non-pharmacological measures have been widely applied in clinical practice to prevent delirium (24, 25). When patients given dexmedetomidine were excluded, the incidence of delirium (22.6%) was close to the reported range.

Our results showed that delirium mainly occurred early after surgery, i.e., 91.9% of new-onset delirium developed within the first three postoperative days in our patients. This is in line with other studies (8, 26). Similar results were also reported in recent trials. For example, 89% of all delirious cases occurred within 3 days after cardiac surgery (7), and that after noncardiac surgery was 88% (27). The phenomenon could be attributed to the fact that surgery-related responses including the degree of inflammation and the secretion of neuroendocrine hormone, which play important roles in the pathogenesis of delirium, peak in the early postoperative period (17, 28–30).

Delirium has a fluctuating nature and is frequently presented in a hypoactive form (31, 32). It is not surprising that many delirious cases, up to 76%, were miss diagnosed in the settings of emergency department, palliative care unit, and ICU (33–35). Delirium detection can be improved with frequent monitoring, and early treatment of the underlying condition can improve outcome (36). However, redundant evaluations require more manpower and can be costly. The commonly used delirium evaluation strategy is twice daily after surgery for several days. Limiting the number of delirium assessments while maintaining high sensitivity would help to make the research work cheaper and more practical.

During the analysis, we designed various evaluation strategies to simplify the original one that assessed delirium twice daily for 7 days after surgery. We found that CAM-ICU assessments twice daily for 5 postoperative days detected 100% delirium but required 29% less evaluation, while CAM-ICU assessment twice daily for 4 days detected 99% delirium but required 43% less evaluation. It is thus reasonable to design studies with no more than 5 days of delirium assessments, and if the personnel and funds of the study are insufficient, 4 days could be sufficient. A previous study reported that delirium was detected more often in the morning than in the evening (15). However, we did not find difference in detecting delirium between morning only and afternoon only assessments, possibly due to different patient population and improved nighttime environment and care in the ICU. Both methods detected only 75% delirium within 7 days and are therefore not recommended.

The advantages of our study included that the investigators who performed delirium assessment had been trained by a psychiatrist and that delirium assessment was performed twice daily for 7 consecutive days. These helped us to detect most delirium cases. There are also some limitations. Firstly, while investigators had been trained for delirium assessment, the inter-investigator differences were hardly avoidable. Secondly, we enrolled patients following various kinds of non-cardiac surgeries. This increases the generalizability of our study but increases the complexity in explaining the results. Thirdly, all patients included in the underlying trial were admitted to the ICU after noncardiac surgery. Whether our results can be extrapolated to non-ICU patients requires further investigation.

In older patients admitted to the ICU after elective noncardiac surgery, 60.4% of all delirium detected over 7 days occurred during the first postoperative day and 91.2% occurred within the first 3 postoperative days. Compared with CAM-ICU assessment twice daily for 7 days, a twice-daily assessment for 5 days detected 100% delirium with 29% less effort, while a twice-daily assessment for 4 days detected 99% delirium with 43% less effort. Delirium assessments should therefore be performed twice daily for at least 4 initial postoperative days.