2. Outline
incidence of underfeeding in the ICU
nutritional screening tools available for use in ICU
familiar with the novel approach used to assess the
nutritional risk of critically ill patients and implications
of this risk assessment for clinical practice
3. Does underfeeding in ICUs
exist?
Mean intake 56% International Nutrition Survey, n =211 ICUs
4. Purpose of Nutrition Screening
Predict the probability of a better or worse
outcome due to nutrition
SCREENING
Malnutrition
goes
undetected
6. Screening leads to Nutritional Care
Hospitals & healthcare organizations should have a policy and a
specific set of protocols for identifying patients at nutritional risk.
The following process is suggested:
Screening
Assessment
Monitoring & Outcome
Communication
Audit
Kondrup et al. Clin Nutr 22(4):415-421;2003.
7. • Underfeeding does occur in ICUs
• Malnutrition: 30% ICU patients (SGA)
• Existing tools for nutrition screening
8. Malnutrition Universal Screening Tool (MUST)
Nutritional Risk Screening (NRS 2002)Nutritional Risk Screening (NRS 2002)
Mini Nutritional Assessment (MNA)
Short Nutritional Assessment Questionnaire (SNAQ)
Malnutrition Screening Tool (MST)
Subjective Global Assessment (SGA)Subjective Global Assessment (SGA)
Anthony NCP 2008
11. When training provided in
advance, SGA can produce
reliable estimates of malnutrition
Note rates of missing data
(7-34%)
12. n = 119, > 65 yrs, mostly medical patients, not all ICU
no difference between well-nourished and malnourished patients with
regard to the serum protein values on admission, LOS, and mortality
rate
13. n = 124, mostly surgical patients
100% data available for SGA
SGA predicted mortality
14. Quantify Lean Muscle Mass: CT Scan
• Body composition tools:
– BIA, skin fold: low precision , DEXA: not specific, $$
• CTs becoming common research tool
– Measures tissue mass and changes over time
50 geriatric trauma pts
prevalence of sarcopenia (low
muscularity) on admission
=78%
Despite the majority being
overweight!
M. Mourtzakis et al
15. ICU patients are not all created equal…should we
expect the impact of nutrition therapy to be the
same across all patients?
16. Malnutrition should be diagnosed on the
basis of etiology…. inflammation acute vs
chronic
17. How do we figure out who will benefit
the most from Nutrition Therapy?
In the ICU…..
Caloric debt/underfeeding
Malnutrition exists 34% or >
Historical nutrition data n/a
Not all patients equal
Consider
Inflammation
Acute diseases
Chronic diseases
18. Nutrition Status
micronutrient levels - immune markers - muscle mass
Starvation
Acute
-Reduced po intake
-pre ICU hospital stay
Chronic
-Recent weight loss
-BMI?
Inflammation
Acute
-IL-6
-CRP
-PCT
Chronic
-Comorbid illness
A Conceptual Model for Nutrition Risk
Assessment in the Critically ill
19. Objective
Develop a score using the variables in the model to
Quantify the risk of ICU pts developing adverse
events that may be modified by nutrition
20. The Development of the NUTrition Risk in the
Critically ill Score (NUTRIC Score)
• When adjusting for age, APACHE II, and SOFA, what
effect of nutritional risk factors on clinical outcomes?
• Multi institutional data base of 598 patients (3 ICUs)
• Historical po intake and weight loss only available in
171 patients
• Outcome: 28 day vent-free days and mortality
21. What are the nutritional risk factors
associated with mortality?
(validation of our candidate variables)
Non-survivors by day 28
(n=138)
Survivors by day 28
(n=460)
p values
Age 71.7 [60.8 to 77.2] 61.7 [49.7 to 71.5] <.001
Baseline APACHE II score 26.0 [21.0 to 31.0] 20.0 [15.0 to 25.0] <.001
Baseline SOFA 9.0 [6.0 to 11.0] 6.0 [4.0 to 8.5] <.001
# of days in hospital prior to ICU admission 0.9 [0.1 to 4.5] 0.3 [0.0 to 2.2] <.001
Baseline Body Mass Index 26.0 [22.6 to 29.9] 26.8 [23.4 to 31.5] 0.13
Body Mass Index 0.66
<20 6 ( 4.3%) 25 ( 5.4%)
≥20 122 ( 88.4%) 414 ( 90.0%)
# of co-morbidities at baseline 3.0 [2.0 to 4.0] 3.0 [1.0 to 4.0] <0.001
Co-morbidity <0.001
Patients with 0-1 co-morbidity 20 (14.5%) 140 (30.5%)
Patients with 2 or more co-morbidities 118 (85.5%) 319 (69.5%)
C-reactive protein¶ 135.0 [73.0 to 214.0] 108.0 [59.0 to 192.0] 0.07
Procalcitionin¶ 4.1 [1.2 to 21.3] 1.0 [0.3 to 5.1] <.001
Interleukin-6¶ 158.4 [39.2 to 1034.4] 72.0 [30.2 to 189.9] <.001
171 patients had data of recent oral intake and weight loss
Non-survivors by day 28
(n=32)
Survivors by day 28
(n=139)
p values
% Oral intake (food) in the week prior to enrolment 4.0[ 1.0 to 70.0] 50.0[ 1.0 to 100.0] 0.10
% of weight loss in the last 3 month 0.0[ 0.0 to 2.5] 0.0[ 0.0 to 0.0] 0.06
22. Variable
Spearman
correlation with
VFD within 28
days
p values
Number of
observations
Age -0.1891 <.0001 598
Baseline APACHE II score -0.3914 <.0001 598
Baseline SOFA -0.3857 <.0001 594
% Oral intake (food) in the week prior to enrollment 0.1676 0.0234 183
number of days in hospital prior to ICU admission -0.1387 0.0007 598
% of weight loss in the last 3 month -0.1828 0.0130 184
Baseline BMI 0.0581 0.1671 567
# of co-morbidities at baseline -0.0832 0.0420 598
Baseline CRP -0.1539 0.0002 589
Baseline Procalcitionin -0.3189 <.0001 582
Baseline IL-6 -0.2908 <.0001 581
What are the nutritional risk factors
associated with Vent Free days?
(validation of our candidate variables)
BMI: no effect on Vent free days
23. The Development of the NUTrition Risk in the
Critically ill Score (NUTRIC Score)
• % oral intake in the week prior dichotomized into
– patients who reported less than 100%
– all other patients
• Weight loss was dichotomized as
– patients who reported any weight loss
– all other patients
• BMI was dichotomized as
– <20
– all others
• Comorbidities was left as integer values range 0-5
24. The Development of the NUTrition Risk in the
Critically ill Score (NUTRIC Score)
All other variables (Age, APACHE 2, SOFA, Comorbidities, LOS pre ICU, IL 6)
were categorized into five equal sized groups (quintiles)
Exact quintiles and logistic parameters for age
Exact Quintile Parameter Points
19.3-48.8 referent 0
48.9-59.7 0.780 1
59.7-67.4 0.949 1
67.5-75.3 1.272 1
75.4-89.4 1.907 2
Logistic regression analyses
Each quintile compared to lowest risk
category
Rounded off to the nearest whole # to
provide points for the scoring system
25. The Development of the NUTrition Risk in the
Critically ill Score (NUTRIC Score)
Variable Range Points
Age <50 0
50-<75 1
>=75 2
APACHE II <15 0
15-<20 1
20-28 2
>=28 3
SOFA <6 0
6-<10 1
>=10 2
# Comorbidities 0-1 0
2+ 1
Days from hospital to ICU admit 0-<1 0
1+ 1
IL6 0-<400 0
400+ 1
AUC 0.783
Gen R-Squared 0.169
Gen Max-rescaled R-Squared 0.256
BMI, CRP, PCT, weight loss, and oral intake were excluded because they were not significantly
associated with mortality or their inclusion did not improve the fit of the final model.
26. The Validation of the NUTrition Risk in the
Critically ill Score (NUTRIC Score)
0 1 2 3 4 5 6 7 8 9 10
Nutrition Risk Score
MortalityRate(%)
020406080
Observed
Model-based
n=12 n=33 n=55 n=75 n=90 n=114 n=82 n=72 n=46 n=17 n=2
Statistical
modeling
higher
score =
higher
mortality
27. The Validation of the NUTrition Risk in the
Critically ill Score (NUTRIC Score)
0 1 2 3 4 5 6 7 8 9 10
Nutrition Risk Score
DaysonMechanicalVentilator
02468101214
Observed
Model-based
n=12 n=33 n=55 n=75 n=90 n=114 n=82 n=72 n=46 n=17 n=2
high score
= longer
ventilation
28. The Validation of the NUTrition Risk in the
Critically ill Score (NUTRIC Score)
Can NUTRIC score modify the association between nutritional adequacy
and mortality? (n=211)
P value for the
interaction=0.01
0 50 100 150
0.00.20.40.60.81.0
Nutrition Adequacy Levles (%)
28DayMortality
NUTRIC 0-3
NUTRIC 4-6
NUTRIC 7-8
NUTRIC 9-10
P value for the
interaction=0.01
Highest score pts,
low nutrition is
associated with
higher mortality!!
Lowest score pts,
more nutrition may
be associated with
higher mortality ?
29. Summarize: NUTRIC Score
• NUTRIC Score (0-10) based on
– Age
– APACHE II
– SOFA
– # comorbidities
– Days in hospital pre ICU
– IL 6
• High NUTRIC Score associated worse outcomes
(mortality, ventilation)
• High NUTRIC Score benefit the most from nutrition
• Low NUTRIC Score : harmful?
30. Applications of NUTRIC Score
• Help determine which patients will benefit more from
nutrition
– Supplemental PN
– Aggressive feeding
– Small bowel feeding
• Design & interpretation of future studies
– Negative studies, non high risk, heterogenous patients
31. Limitations
• Applies only to macronutrients
• Does not apply to pharmaconutrients
• Nutritional history is suboptimal
• Requires IL-6
32.
33. Conclusion
• Iatrogenic underfeeding in ICUs exist
• Nutrition Screening/audits* detect underfeeding
• Existing Screening tools not helpful in ICU
• Not all ICU patients are the same in terms of ‘risk’
• NUTRIC Score is one way to quantify that risk and can
be used in your ICU
• Further refinement of this tool will ensure that the right
patient gets nutrition
- Several modalities have been used to study body composition – including BIA, skin-fold – most have poor precision – DXA – high precision, not as specific and not as accessible
CT imaging is becoming a common research tool - powerful in measuring different tissues and their changes over time
Malnutrition should be diagnosed on the basis of etiology…inflammation vs acute vs chronic
Need picture of malnourshed child
Linked starvation inflammation and nutritional status to outcomes
3 ICUs med/sx mix
Based on the conceptual model, we identified variables that predicted mortality. Looked at their affect by survivors vs non surviviors….validated the variables. ALL sign different hence predicted mortality EXCEPT BMI, CRP, oral intake and % wt loss
Variables on Vfdays…higher the age, lower the VFdays
Step 2:variables defined and validated, how to move to next stage:
Categorized the groups up and do regression analyses
Logistoc regression analyses qunitiles were compared to lowest risk (reference)
The parameters for each logistic regression model estimate the log of the odds ratio (logit) for each category (usually quintile) of the variable compared to the lowest risk (reference) category.
These parameters were rounded to whole numbers to provide the points used in the NUTRIC risk score. Equal point categories were collapsed, and the exact quintile ranges were subsequently rounded to convenient values.
Similary techniques were done for each of the variables
Further statisitcal modelling to see if the SCORING predicted mortality
Predicted and observed model
HIGHER SCORE = HIGHER MORTALITY