Agreement between NU-DESC and CAM-ICU for delirium in high-complexity hospital settings

Abstract

Objective

To analyse the agreement between NU-DESC and CAM-ICU scales in detecting hyperactive and hypoactive delirium in adult patients at a tertiary hospital, testing the alternative hypothesis (HA) of significant concordance.

Method

Cross-sectional study of 105 patients (18–85 years) under moderate sedation and ventilatory weaning, assessed with NU-DESC and CAM-ICU at three time points after staff training. Concordance was evaluated using the Kappa index.

Results

Mean age: 45 years (52% male, 48% female). Initial agreement: Kappa = .758 (good, p < .001); second measurement: Kappa = .448 (moderate, p < .001); third measurement: Kappa = .848 (very good, p < .001). Variability reflects differences in sensitivity across delirium phases.

Discussion

NU-DESC was optimal for initial screening (speed), while CAM-ICU showed higher accuracy in advanced stages. Their combined use captures dynamic delirium manifestations.

Conclusion

HA was confirmed, proving that integrating NU-DESC and CAM-ICU enhances delirium detection and management in critically ill patients, adapting to their evolving clinical needs.

Resumen

Objetivo

Analizar la concordancia entre las escalas NU-DESC y CAM-ICU en la detección de delirio hiperactivo e hipoactivo en pacientes adultos de un hospital de tercer nivel, validando la hipótesis de concordancia significativa (HA).

Método

Estudio transversal en 105 pacientes (18–85 años) bajo sedoanalgesia moderada y en destete ventilatorio, evaluados con NU-DESC y CAM-ICU en tres momentos tras capacitación del personal. La concordancia se analizó con el índice Kappa.

Resultados

Edad promedio: 45 años (52% hombres, 48% mujeres). Concordancia inicial: Kappa = 0.758 (buena, p < 0.001); segunda medición: Kappa = 0.448 (moderada, p < 0.001); tercera medición: Kappa = 0.848 (muy buena, p < 0.001). La variabilidad refleja diferencias en sensibilidad según la fase del delirio.

Discusión

La NU-DESC mostró mayor utilidad en tamizaje inicial (rapidez), mientras la CAM-ICU fue más precisa en etapas avanzadas. La complementariedad de ambas escalas optimiza la detección dinámica del delirio.

Conclusión

Se confirma la HA, demostrando que la integración de NU-DESC y CAM-ICU mejora la identificación y manejo del delirio en pacientes críticos, adaptándose a sus necesidades clínicas evolutivas.

What is known?

Delirium is an acute neuropsychiatric disorder that frequently affects patients hospitalised in intensive care units (ICUs), especially those receiving mechanical ventilation or sedation. Its timely detection is crucial for preventing complications, reducing hospital stays, and improving patient recovery. Several validated tools exist for delirium assessment, including the CAM-ICU and the NU-DESC, both widely used in clinical settings. The CAM-ICU is recognised for its high sensitivity and specificity, while the NU-DESC provides a rapid assessment focused on behavioural symptoms. However, little is known about the degree of concordance between these two scales, particularly in differentiating between delirium subtypes (hyperactive and hypoactive). This study provides evidence on how these tools relate to each other when applied at different clinical times and under specific conditions. The study shows good concordance in some measurements but less in others, reinforcing the idea that delirium is a dynamic and diverse phenomenon requiring multiple approaches for its proper assessment.

Study implications

The findings of this study have significant practical and clinical implications for healthcare personnel, especially nurses, who frequently are the people who are the first to identify delirium. The evidence suggests that the combined use of the CAM-ICU and NU-DESC scales can improve the detection of different types of delirium over time, enabling more precise and timely interventions. This is particularly important in patients under moderate sedation, where signs can be subtle or fluctuate rapidly. From an institutional perspective, implementing both tools as part of the daily assessment protocol could strengthen monitoring of the patient's neurological status, optimise resources, and improve clinical outcomes. Furthermore, this study encourages further research into the applicability of these scales in different hospital settings and levels of care, promoting a more evidence-based and patient-safety-focused clinical practice.

Introduction

Delirium is an acute and fluctuating neuropsychiatric disorder that significantly affects patients admitted to intensive care units (ICUs).¹ Its prevalence ranges from 40% to 80%, depending on the severity of the patient's condition,² and is particularly common in those requiring mechanical ventilation.³ This syndrome is defined by alterations in consciousness, attention, and other cognitive functions, and is associated with poorer clinical outcomes, such as higher mortality rates, prolonged hospital stays, functional impairment, and long-term cognitive decline in survivors.^4,5

Despite its clinical impact, diagnosing delirium remains a challenge in clinical practice. Its variability in manifestations, along with the lack of universal and standardised assessment tools, complicates its accurate identification.⁶ This situation underscores the importance of having effective and accessible methods for its detection, which can improve clinical outcomes and reduce associated complications.⁷ Among the most widely used tools are the Confusion Assessment Method for the Intensive Care Unit (CAM-ICU) and the Nursing Delirium-Screening Scale (NU-DESC), both designed to facilitate the detection of delirium in critical care settings.⁸

The CAM-ICU is considered the gold standard in ICUs due to its high sensitivity (85.7%) and specificity (86.8%). However, its implementation requires specialised training, which may limit its use in high-workload or resource-constrained settings.³ The CAM-ICU is a tool recommended by clinical guidelines and is widely used in clinical and research settings.³ The Spanish-language version of the CAM-ICU was developed in accordance with the recommendations of ISPOR (International Society for Pharmacoeconomics and Outcomes Research).⁹

On the other hand, the NU-DESC8 stands out for its simplicity and speed,¹⁰ being especially effective in the initial stages of delirium.¹¹ This tool assesses key symptoms such as changes in the level of consciousness, disorientation, hallucinations, and behavioural disturbances. However, its sensitivity decreases over time, which could affect its effectiveness in advanced stages of the disorder.¹²

The NU-DESC scale was developed by Gaudreu et al. in 2002 and originally validated in 146 hospitalised cancer patients.⁸ It has also been validated for use in postoperative services, emergency departments, oncology units, palliative care units, neurological units, rehabilitation units, and geriatric units.¹¹ This scale was designed for use by nurses and other healthcare professionals to rapidly assess delirium.¹¹ The NUD-ESC scale has been validated in several languages, including German,^8,13 English, and Spanish,¹⁴ and has been previously validated by researchers.^8,10,15

In this context, this study aims to analyse the agreement between the CAM-ICU and the NU-DESC in identifying delirium, with particular emphasis on differentiating between hyperactive and hypoactive subtypes. The alternative hypothesis (HA) is that there is significant agreement between both scales for this differentiation in adults hospitalised in a tertiary care centre, while the null hypothesis (H0) postulates the absence of such agreement.

Method

A cross-sectional, observational, quantitative, and comparative study was conducted between January and December 2024 in different departments of a tertiary care hospital in Mexico City. Its main objective was to analyse the agreement between the NU-DESC and CAM-ICU scales for delirium detection, evaluating the differentiation between hyperactive and hypoactive delirium in adult patients. The study population included adult patients, aged 18–85 years, hospitalised in clinical areas such as intensive care, intermediate care, emergency department, and recovery, because delirium can manifest in all of these hospital settings. Patient selection was performed using consecutive non-probability sampling. Inclusion criteria required that individuals had a sedation/analgesia level ranging from moderate (RASS-3)¹⁶ to combative agitation (RASS + 4) and that they were in the process of ventilator weaning at the time of the selected study shifts. Although sedation and analgesia procedures and ventilator weaning are more prevalent in ICUs and intermediate care units (ICUs), the possibility of including isolated cases from other hospital areas where these clinical scenarios might occur was considered. The final sample included 105 patients, randomly selected by convenience sampling. The formula for calculating the sample size for quantitative studies of finite populations was used, based on written inclusion/exclusion criteria to homogenise the sample and the need to analyse the agreement between the two scales. The inclusion criteria were: adult patients hospitalised in clinical areas of the hospital, under moderate sedation and analgesia, undergoing ventilator weaning, and available for evaluation in 12 -h shifts (day and night). Patients treated with benzodiazepines or antipsychotics in the previous 12 h, as well as those with metabolic or endocrine disorders that affected their cognitive state, were excluded. The NU-DESC and CAM-ICU scales, both previously validated in a Mexican population, were used. Before implementation, selected clinical staff were trained to ensure the correct application of the instruments. Assessments were conducted during 12-h shifts at three specific times: 8:00, 12:00, and 20:00 during the day shift; and 21:00, 3:00, and 7:00 during the night shift. Data collection was carried out using digital and physical forms, and the collected data were imported into SPSS version 25 software for analysis. The data were verified and structured to identify missing values ​​and ensure their integrity. Subsequently, a descriptive analysis was performed, including obtaining measures such as mean, median, and standard deviation for the scores of both scales. The agreement between NU-DESC and CAM-ICU was assessed using the Kappa coefficient, supplemented by statistical tests to identify significant differences between the results.

The study complied with international and national regulations on research ethics, including the Declaration of Helsinki,¹⁶ the Belmont Report,¹⁷ the Nuremberg code¹⁸ and Mexican Official Standard NOM-012-SSA3-2012.¹⁹ The protocol was approved by the hospital's Research Ethics Committee, guaranteeing the protection of participants' rights and well-being. Informed consent was obtained by proxy (family member/legal guardian), ensuring that participants fully understood the study's objectives, procedures, risks, and benefits. Data confidentiality was safeguarded, and anonymity was guaranteed. Participation was entirely voluntary, with family members/legal guardians having the right to withdraw at any time without repercussions or needing to justify their decision. Family members were encouraged to communicate any doubts or concerns to the research team to ensure a transparent and ethical process.

Results

A population of 105 patients at a tertiary care institution was analysed, and data related to sociodemographic characteristics were presented. The average age obtained was 45 years, indicating a wide variability in participants’ ages. After analysing the groups, it was observed that 24.3% of the patients were over 60 years old, 52.3% were between 27 and 59 years old, 12.1% were between 19 and 26 years old, and 9.3% were teenagers. Regarding sex, the population was relatively balanced: 52% were men and 48% were women, as shown in Table 1.

Table 1.

Age and sex.

Variable	f	%
	N = 105
Age (years)	45 ± 45.5
>60 Years	26	24.3
27−59 Years	56	52.3
19−26 Years	13	12.1
13−18 Years	10	9.3
Sex
Male	55	52
Female	50	48

The proportion of patients observed during different 12-h shifts is also detailed. The distribution was relatively uniform across the evaluated shifts. The "evening A" and "night B" shifts each attended to 24.2% of the patients, while the "special daytime" shift had a slight majority at 28.4% (Table 2). The special night shift covered 23.2% of the patients, indicating a balance in the care provided during the different periods. Of the 105 patients, 44.2% were intubated, while 54.8% were not. Furthermore, 24.5% of the patients were in the process of ventilator weaning while 75.5% were not. This demonstrates that the majority of patients do not require invasive ventilation or withdrawal from it (Table 3).

Table 2.

Proportion of patients observed during the 12-h shifts.

Turno	f	%
Evening A	26	24.2
Night-time B	26	24.2
Daytime special	29	284
Night-time special	24	23.2

Table 3.

Percentage of patients who were intubated, and patients who were not intubated. The proportion of patients who were not experiencing the weaning process, and patients who were, is also presented.

P. intubated	Yes		No		Total
	f	%	f	%	105
	48	44.2	57	54.8
Weaning process	Yes		No
	f	%	f	%
	25	24.5%	80	75.5%

Table 4 examines the agreement between two delirium detection methods: NU-DESC and CAM-ICU. The results indicate that both methods correctly classified 55.4% of patients as positive, while 6.6% were positive only by CAM-ICU and another 6.6% were positive only by NU-DESC. On the other hand, 31.4% of patients were classified as negative by both methods. In total, 62% of patients were positive and 38% were negative. Statistical analysis reveals a p-value < .001, indicating statistical significance, and a Kappa value of .758, reflecting good agreement between the two methods.

Table 4.

Concordance between two delirium detection methods, CAM-ICU and NU-DESC, during the first measurement.

NU-DESC 1	CAM-ICU1				Total
	Positive		Negative
	f	%	f	%	f	%
Positive	59	55.4	6	6.6	65	62
Negative	6	6.6	34	31.4	40	38
Total	65	62	40	38	105	100

P value < .001.

Kappa value: .758 (good concordance).

Table 5 presents another assessment of the agreement between the NU-DESC and CAM-ICU methods. In this case, 36% of patients were classified as positive by both methods, 14.3% were positive only by CAM-ICU, and 13.3% were positive only by NU-DESC. The remaining 35% were classified as negative by both methods. Overall, 51% of patients were positive and 49% were negative. Statistical analysis shows a p-value < .001, with a Kappa value of .448, suggesting moderate agreement between the methods.

Table 5.

Concordance between two delirium detection methods, CAM-ICU and NU-DESC, during the second measurement.

NU-DESC 2	CAM-ICU2				Total
	Positive		Negative
	f	%	f	%	f	%
Positive	38	36.7	16	14.3	54	51
Negative	13	13.3	38	35	51	49
Total	54	51	51	49	105	100

P value < .001.

Kappa value: .448 (moderate concordance).

Finally, Table 6 shows a further assessment of the agreement between NU-DESC and CAM-ICU. In this case, 35.4% of patients were classified as positive by both methods, while 5.6% were positive only by CAM-ICU. There were no cases that were positive by NU-DESC and negative by CAM-ICU. The remaining 59.4% were classified as negative by both methods. In total, 35.4% of patients were positive and 65% were negative. Statistical analysis shows a p-value <.001 and a Kappa value of .848, indicating very good agreement between the two methods.

Table 6.

Concordance between two delirium detection methods, CAM-ICU and NU-DESC, during the third measurement.

NU-DESC 3	CAM-ICU3				Total
	Positive		Negative
	f	%	f	%	f	%
Positive	37	35,4	6	5,6	43	41
Negative	0	0	62	59,4	62	59
Total	37	35,4	68	65	105	100

P value < .001.

Kappa value: .848 (very good concordance).

Discussion

In complex care settings, especially in intensive care units (ICUs), it is common for nursing staff to omit or delay essential care, increasing the risk of delirium in patients. This situation is even more critical in patients requiring intensive care, who are in a more vulnerable state. Lack of timely attention can increase the incidence of delirium, negatively impacting patient health. Identifying delirium, particularly through the use of assessment scales, is essential for improving patient health and prognosis. The 2018 Society of Critical Care Medicine (SCCM) best practice guidelines recommend that delirium assessment be performed systematically and periodically in all patients admitted to ICUs.²⁰

The results indicate that the NU-DESC scale requires less application time, with a minimum of 2 ± 1.256 min, compared to the CAM-ICU standard, which requires between 2 ± .740 and 3 ± .809 min. In specific contexts, such as the Czech Republic,¹⁰ it has been observed that the application of the NU-DESC scale requires only 3 min without the need for previously trained personnel, suggesting good adherence to professional nursing practice in various healthcare institutions. However, the results of Henao-Castaño al.²¹ show a minimum application time of only 2 min, highlighting the variability and consistency in the implementation of the scale in different clinical settings. Although CAM-ICU is widely used, its longer application time could affect the efficiency of clinical decision-making,⁹ while the consistency of NU-DESC, despite its shorter time, may be a relevant factor when choosing an assessment method in daily practice.

In particular, the CAM-ICU scale has proven effective in identifying delirium in intubated patients and those undergoing weaning. The results of this study show a progressive decrease in positive cases across three measurements using the CAM-ICU scale in patients undergoing weaning: 13 (52%)–8 (32%) and 30 (63%)–16 (33%), suggesting a favourable trend in the identification and potential management of delirium. The CAM-ICU has become the most widely used method in ICUs for delirium assessment due to its accessibility and ease of bedside administration. Its widespread use is supported by over 4000 studies, and its translation into 12 languages ​​reinforces its validity and reliability in various established clinical settings.²⁰ The scale has demonstrated a sensitivity of 94%–95%, a specificity of 90%–95%, and a positive predictive value of 91%–94%. In Brazil, similar findings were observed, with a sensitivity of 95% and a specificity of 93%.^22,23

Furthermore, the NU-DESC scale shows a notable decrease in positive cases, from 63% to 38%, with similar results in patients undergoing weaning, where positive cases decreased from 56% to 32%. The sensitivity of the NU-DESC in this study ranged from 90% to 100%, while the specificity ranged from 85% to 86%, with an overall accuracy of 86% to 90%. These results are consistent with those reported in previous studies, such as one conducted in a hospital in the Czech Republic, which reported a sensitivity of 92%, a specificity of 83%, and a positive prediction rate of 62%.¹² A similar study conducted in Portugal²⁴ achieved a sensitivity of 100% and a specificity of 86%, further supporting the reliability and accuracy of the NU-DESC scale as a robust tool for assessing delirium in diverse clinical contexts, including patients in intensive care, general hospital wards, and surgical settings.²¹ Importantly, both nurses and physicians can apply the NU-DESC scale with a high degree of accuracy in detecting delirium.^21,25

Regarding the agreement between the NU-DESC and CAM-ICU methods, variability in agreement was observed across the measurements, represented by the Kappa (K) value. In the first measurement, the K coefficient = .758 indicates good agreement between the two methods, suggesting a high degree of agreement and statistical significance (p < .001). At this stage, both NU-DESC and CAM-ICU correctly identified the majority of patients with delirium (59 positive patients), supporting the reliability of both methods. However, in the second measurement, the agreement between NU-DESC and CAM-ICU was lower, with a K coefficient = .448, indicating moderate agreement. Although this agreement remains statistically significant with a p-value < .001, the lower Kappa value indicates a greater discrepancy between the two methods for detecting delirium at this stage. This discrepancy is evidenced by the increase in false positives and negatives, with 16 patients classified as negative by NU-DESC but positive by CAM-ICU.

In the third measurement, the agreement between the two methods improved, with a K-value of .848, indicating very good agreement, and the p-value < .001 confirms statistical significance. In this case, there were no false negatives, meaning that NU-DESC and CAM-ICU consistently identified all patients with delirium. This improvement in agreement suggests that, with time and patient follow-up, the two methods become more effective at identifying delirium, or that the assessment criteria align better in later stages of the screening process. Therefore, the correct selection and use of an appropriate screening tool is critical for managing delirium.

The improvement in agreement observed in the third measurement could indicate that, as patient follow-up progresses, the two screening methods align better in their ability to identify cases of delirium. This highlights the importance of using more than one method in delirium assessment, especially in situations where agreement is not as strong, as shown in the second measurement. However, with a Kappa index >.6, and contrary to the literature, the use of the NU-DESC instead of the CAM-ICU would not be indicated. It is noteworthy that the agreement between NU-DESC and CAM-ICU varies across measurements, which could be related to changes in the severity or characteristics of delirium in the patients assessed, as well as to inherent differences between the two assessment tools.

Conclusions

The analysis of the agreement between the NU-DESC and CAM-ICU methods reveals crucial information about the variability in the diagnostic accuracy of both tools over time. Although both methods show acceptable effectiveness in identifying delirium, the agreement between them is not constant, reflecting the dynamic nature of delirium, influenced by factors such as the patient's clinical course and the specific characteristics of each scale.

In the first measurement, the agreement between the NU-DESC and CAM-ICU scales was high, with a coefficient K = .758, suggesting that, in the initial phases of delirium, both tools can be used in a complementary manner to provide a reliable diagnosis. The statistical significance (p < .001) reinforces the robust diagnostic capacity of both scales in this early phase. However, the discrepancy observed in the second measurement (K = .448) reflects lower agreement, indicating that, in intermediate stages of delirium, the sensitivity of each tool may differ. This variability could be due to the diversity of symptoms, fluctuations in the patient's mental state, or differences in the criteria assessed by each method.

This finding is relevant to clinical practice, as it suggests that, in more complex stages of delirium, the use of multiple assessment tools may be essential to improve diagnostic accuracy and avoid errors that could negatively impact patient management. The improved agreement observed in the third measurement (K = .848) indicates that, over time and with the stabilisation of the clinical manifestations of delirium, both scales align better in their ability to identify patients with delirium. These results suggest that the combined use of both scales may be more accurate and reliable as patient follow-up progresses, facilitating the implementation of more appropriate and timely interventions.

This study met its objective by demonstrating significant agreement between the NU-DESC and CAM-ICU scales in detecting hyperactive and hypoactive delirium (initial Kappa = .758, p < .001), thus validating the alternative hypothesis (HA). The variability observed in agreement (Kappa = .448 in the second measurement, recovering to .848 in the third) reflects the dynamic nature of delirium and highlights the complementarity of both tools:

CAM-ICU: Greater accuracy in complex stages, ideal for contexts requiring rigorous evaluation. NU-DESC: Speed and accessibility, optimal for initial screening in areas with high healthcare demand. The joint implementation of these scales improves early detection and management of delirium, optimising clinical outcomes.

Funding

Fully funded by the researchers.