点击显示 收起
1 From the Department of Clinical Dietetics and Division of Nephrology, The Hospital for Sick Children, Toronto, Canada (DJS), and the Division of Gastroenterology, St Michael's Hospital and the Institute of Medical Sciences, University of Toronto, Toronto, Canada (KNJ)
2 Supported by the Canadian Foundation for Dietetic Research, the Canadian Institutes of Health Research (CIHR) Doctoral Research Award (DJS), the CIHR Clinician Scientist Training Program in Clinical Nutrition (DJS), and the Hospital for Sick Children Research Institute Research Training Centre (Restracomp) (DJS). CDC growth charts were provided by Abbott Laboratories, Ltd. 3 Reprints not available. Address correspondence to DJ Secker, Department of Clinical Dietetics, The Hospital for Sick Children, 555 University Avenue, Toronto, ON, Canada M5G 1X8. E-mail: donna.secker{at}sickkids.ca.
ABSTRACT
Background:Subjective Global Assessment (SGA), a method of nutritional assessment based on clinical judgment, has been widely used to assess the nutritional status of adults for both clinical and research purposes.
Objective:Foreseeing benefits of its use in children, we chose to adapt SGA and test its validity and reproducibility in the pediatric population.
Design:We prospectively evaluated the preoperative nutritional status of 175 children (aged 31 d to 17.9 y) having major thoracic or abdominal surgery with the use of Subjective Global Nutritional Assessment (SGNA) and commonly used objective measurements. Each child underwent nutritional assessment by 2 independent assessors, one performing measurements of anthropometrics and handgrip strength and one performing SGNA. To test interrater reproducibility, 78 children had SGNA performed by a third assessor. Occurrence of nutrition-associated complications was documented for 30 d postoperatively.
Results:SGNA successfully divided children into 3 groups (well nourished, moderately malnourished, severely malnourished) with different mean values for various anthropometric and biochemical measures (P < 0.05). Malnourished children had higher rates of infectious complications than did well-nourished children (P = 0.042). Postoperative length of stay was longer for malnourished children (8.2 ± 10 d) than for well-nourished children (5.3 ± 5.4 d) (P = 0.002). No objective nutritional measures showed association with outcomes, with the exception of serum albumin, which was not clinically predictive because mean concentrations were in the normal range irrespective of the presence or absence of complications.
Conclusion:SGNA is a valid tool for assessing nutritional status in children and identifying those at higher risk of nutrition-associated complications and prolonged hospitalizations.
Key Words: Malnutrition • children • surgery • nutritional assessment • Subjective Global Assessment • SGA • outcomes
INTRODUCTION
Current methods of assessing nutritional status in children rely on a combination of objective anthropometric, dietary, biochemical, and immunologic measures. Although epidemiologically useful, these measures have several shortcomings that hamper their effectiveness in clinical practice, and no single objective marker has the sensitivity and specificity to be a reliable index of protein-energy malnutrition or predictive of nutrition-related complications (1-5).
Almost 25 y ago, Baker et al (6) published a study showing that nutritional assessment using clinical judgment was a reproducible and valid technique for evaluating the nutritional status of adults. With the use of a Bayesian analysis of pretest and posttest probabilities they showed that a structured clinical approach, called Subjective Global Assessment (SGA), was able to predict nutrition-associated complications (NACs; ie, infections, use of antibiotics, length of stay) better than serum albumin, transferrin, delayed cutaneous hypersensitivity, the Prognostic Nutrition Index, creatinine-height index, triceps skinfold thickness, and various combinations of these measures (7-10). The method of SGA was since refined (11, 12) and was applied and validated internationally for clinical, epidemiologic, and research purposes (13-20).
We hypothesized that SGA could be adapted to identify presurgical malnutrition and to predict postsurgical nutrition-associated morbidities that lead to prolonged hospital stay in pediatric patients. The adapted version was called Subjective Global Nutritional Assessment (SGNA). We then performed a prospective cohort study to determine whether SGNA could better predict NACs and length of hospital stay than current objective methods of pediatric nutritional assessment.
SUBJECTS AND METHODS
Subjects
Consecutive children scheduled for surgery by surgeons in the Division of General Surgery at The Hospital for Sick Children, a large pediatric academic health science center located in Toronto, Canada, were considered for the study. Children were deemed eligible if they were aged 31 d to 17.9 y, required major abdominal or noncardiac thoracic surgery on a nonemergency basis, and had not undergone surgery in the 30 d before screening. Surgery was deemed major if it was performed in the operating room, involved general anesthesia, and required at least one night's stay in the hospital after the surgical procedure. Subjects were precluded from study entry if they were preterm infants who had not reached 31 d corrected age, were clinically unstable, or, along with their caregivers, were non-English speaking (because SGNA depended on clear verbal communication with the children and caregivers). Written consent was obtained from the responsible caregivers of children aged <16 y or directly from adolescents 16 y. Informed assent was obtained from children aged 7-15.9 y. The study was approved by the Research Ethics Board of The Hospital for Sick Children.
Study design
Five staff dietitians served as assessors for the study after training for standardized performance of objective measurements and the SGNA. Study results are not based on a single assessor but are a composite of the data from the 5 different assessors, who performed the assessments in random order. Each child underwent nutritional assessment by 2 independent assessors blinded to the results of the other, one performing measurements of anthropometrics and handgrip strength, and one performing the SGNA. To test interrater reproducibility of the SGNA, a subset of 78 children were seen by a third assessor who also performed the SGNA. Nutritional assessments were performed the day before surgery for prior-day admissions or in-patients and the morning of surgery for same-day admissions; assessments took place in the Clinical Investigation Unit, the patient's individual hospital room, or the General Surgery Same-Day Admission Unit. A blood sample for measurement of serum albumin and transferrin was obtained coincidental with routine preoperative blood work (ie, complete blood count) or by the anesthetist at the time of intravenous induction. Children were followed in a purely observational fashion, from the day of surgery until 30 d after surgery, for the development of nutrition-associated infectious and noninfectious morbidities related to their surgical procedure. Medical and surgical risk factors that might influence the occurrence of complications and outweigh nutritional risk were documented. Data were accrued prospectively and entered into a database (MICROSOFT ACCESS 2000, version 9.0; Microsoft Corp, Redmond, WA) and then partially exported to NUTSTAT (Epi Info, version 3.3.2; Centers for Disease Control and Prevention, Atlanta, GA) for calculation of age, body mass index (in kg/m2), and anthropometric z scores. Analyses compared the ability of SGNA and anthropometric and biochemical measures to identify malnourished children and those at increased risk of developing complications, as well as the rate of agreement between the pairs of dietitians who performed SGNA on the same child.
Subjective Global Nutrition Assessment
Assessors used a SGNA questionnaire to guide their interview and to gather information on the child's recent and current height and weight history as well as parental heights, dietary intake (type, volume, and frequency of liquid and solid feedings for infants and frequency of eating and a brief description of a typical day's intake for children; rating of appetite and recent changes; feeding or eating problems; diet restrictions), frequency and duration of gastrointestinal symptoms (loss of appetite, vomiting, diarrhea, and constipation for all ages plus stomach pain and nausea for children), and current functional capacity and recent changes (alertness, amount of energy or activity for infants and school attendance, ability to run and play games or sports with friends, and time sleeping for children). They then performed the nutrition-related physical examination, looking at specific sites on the body for signs of fat and muscle wasting (combined as "general wasting" for infants and young toddlers), as well as edema. Finally, they considered the presence or absence of the specified historical features, metabolic demands of the underlying condition, and physical signs associated with malnutrition to come to a global assessment of the patient's nutritional status, assigning a global rating of well-nourished, moderately malnourished, or severely malnourished. A rigid scoring system based on specific criteria was not used.
Objective nutritional assessment
We examined the following objective markers of nutritional status: length or height; weight; percentage of ideal body weight for height; body mass index-for-age; midarm circumference; triceps skinfold thickness; midarm muscle area; handgrip strength; concentrations of serum albumin, transferrin, and hemoglobin; and total lymphocyte count. Prealbumin measurement was unavailable at our center and therefore was not included. Measurements were performed with the use of calibrated equipment and standardized techniques (21-23). Anthropometric z scores were calculated with the use of NUTSTAT anthropometric software package (Epi Info, version 3.2.2; Centers for Disease Control and Prevention) (24) or common reference data (25, 26). Preterm infants (n = 22) were assessed with the use of corrected age up to age 2 y.
Biochemical analysis was performed by the hospital's Department of Pediatric Laboratory Medicine with the use of the bromocresol green dye–binding method for serum albumin and immunonephelometry for serum transferrin. The z scores for hemoglobin and transferrin were calculated from data from the American National Health and Nutrition Examination Survey (27).
Outcome measures
Patients were followed for 30 d after surgery for the development of NACs related to their surgical procedure. Complications were subdivided into infectious and noninfectious complications, as well as major and minor complications, according to published objective criteria (28-31). Secondary outcomes were length of postoperative hospitalization, use of nonprophylactic antibiotics, and unplanned rates of reoperation and readmission. Monitoring after discharge continued until 30 d after surgery and occurred by either a routine follow-up clinic visit or telephone interview shortly thereafter.
Statistical analyses
According to previous observations (6, 32), we planned to enroll 175 children for the study to have the capacity to detect a 3-fold difference in incidence of postoperative complications between the malnourished and well-nourished groups (two-sided level: <0.05; ß level: 0.8). To examine the relation among SGNA and individual objective measurements, means of the objective measures for the 3 groups defined by SGNA were tested with analysis of variance or the Kruskal-Wallis nonparametric test and Tukey's honest significant difference or the Mann-Whitney nonparametric test for post hoc analysis. Associations among numerical equivalents of the SGNA groups with objective measures were tested by correlation analysis (Kendall's ). Correlation analysis (Kendall's ) and multinomial logistic regression were used to explore the relation of individual components of SGNA with the final SGNA rating. The association between SGNA and clinical morbidity was tested with the use of the chi-square test or Kendall's test for categorical variables and the Kruskal-Wallis and Mann-Whitney U nonparametric tests for length of stay. Relations between objective measures of nutritional status and morbidities were examined by the Student's t test, Mann-Whitney U nonparametric test, or correlation analysis (Kendall's ).
Analysis was performed with the use of SPSS 14.0 for WINDOWS GRADPACK (SPSS Inc, Chicago, IL). Data collection, entry, and analysis were performed by the principal author (DJS). Results are reported as means ± SDs, unless stated otherwise. Results were considered significant at P < 0.05.
RESULTS
Between February 2003 and August 2004, 180 children were enrolled and completed the study. One hundred seventy-five children (99 males, 76 females) ranging in age from 31 d to 17.9 y (
View this table:
TABLE 1. Characteristics of the study population (n = 175)
Almost half of the children (n = 80, 46%) had one or more disease processes or anomalies [eg, inflammatory bowel disease; Hirschsprung disease; vertebral, anal, cardiac, tracheal, esophageal, renal, and limb anomalies (VACTERL association; a random association of birth defects); cerebral palsy; Down syndrome; cystic fibrosis; sickle cell disease]. Sixteen children (9%) had cancer. On a 6-point scale of medical complexity [American Society of Anesthesiologists (ASA) classification of physical status score (33)], with 1 being the lowest complexity and 6 being the highest, the majority of children were rated preoperatively by the anesthetist as either ASA 2 (47%) or ASA 3 (32%). All but 8 children (5%) received prophylactic antibiotics before surgery. The most commonly performed surgical procedures are listed in Table 1. Distribution of surgical procedures classified by risk of microbial infection (clean, clean-contaminated, contaminated, dirty) did not differ across the 3 SGNA groups (P = 0.88) nor did the proportion of children with cancer (P = 1.0).
Concurrent validity of SGNA
Nutritional status, as determined by SGNA, was compared with objective measurements. With the use of SGNA, 85 children (49%) were classified as well-nourished, 64 children (36%) were classified as moderately malnourished, and 26 children (15%) as severely malnourished. The 3 groups of children had mean ages that did not differ, but they had significantly different mean values for 7 of the 8 anthropometric measures, including handgrip strength and serum albumin, but not transferrin, hemoglobin, or total lymphocyte count (Table 2). It should be noted that, although the mean serum albumin concentrations of the 3 groups of children classified by SGNA were statistically different, they were not clinically different in that even the lowest mean serum albumin (ie, the severely malnourished group, 38 ± 7 g/L) fell well within the hospital's normal reference range. SGNA, with the use of numerical equivalents of the clinical groups, showed moderate-to-fair correlation with all of the objective anthropometric and biochemical measurements of nutritional status, except transferrin z score and total lymphocyte count (Table 2). Handgrip strength, as a surrogate marker of muscle function and impaired functional status, decreased with worsening nutritional status (P = 0.001).
View this table:
TABLE 2. Relation between Subjective Global Nutritional Assessment (SGNA) and anthropometric measures of nutritional status1
Predictive validity of SGNA
Seventy children had at least one complication, experiencing 97 complications in total; no deaths occurred. The majority of complications were infectious in origin (80%), with an equal percentage being major (49%) or minor (51%) in nature. Occurrence of complications did not differ by sex or by age, with the exception of minor complications, which occurred more often in younger children. A breakdown of complications is provided in Table 3. Children identified by SGNA as malnourished (moderately or severely) experienced a higher rate of infectious complications (P = 0.04) and minor (infectious and noninfectious) complications (P = 0.03) compared with children classified as well-nourished (Table 4).
View this table:
TABLE 3. Breakdown of postoperative complications
View this table:
TABLE 4. Relation between malnutrition and outcomes
Postoperative length of stay was more than twice as long for severely malnourished children (19.0 ± 58.8 d) compared with well-nourished (5.3 ± 5.4 d) and moderately malnourished (8.4 ± 11.1 d) children (P = 0.002, Kruskal-Wallis nonparametric test). Mean postoperative hospitalization for severely malnourished children was skewed by 1 child whose postoperative stay was 306 d. Although unusual, an extreme length of stay such as this does occur in malnourished patients and was not seen in our well-nourished population. When this child was excluded as an extreme outlier, the overall difference in mean stay across the 3 groups remained significant (P = 0.02, Kruskal-Wallis). When the moderately and severely malnourished children were grouped together, mean length of postoperative hospitalization was 2.9 d (55%) longer than for the well-nourished children (P = 0.001) (Table 4). The percentages of children receiving nonprophylactic antibiotics or requiring unplanned reoperation or readmission were not significantly different between the well-nourished and malnourished children (Table 4).
No association was observed among any of the objective measures of nutritional status and outcomes, with the exception of height-for-age and postoperative length of stay (r = –0.18, P = 0.001), and serum albumin, which was associated with infectious (no complication, 41.4 ± 4.5 g/L; complication, 38.2 ± 5.9 g/L; P < 0.001) and minor (no complication, 40.8 ± 5.2 g/L; complication, 38.6 ± 5.3 g/L; P = 0.01) complications, postoperative length of stay (r = –0.20, P < 0.001), and use of nonprophylactic antibiotics (no antibiotic, 41.4 ± 4.5 g/L; antibiotic, 38.2 ± 5.9 g/L; P < 0.001). However, it must be emphasized that mean serum albumin concentrations were within the normal reference range for all 3 nutritional groups; therefore, serum albumin was clinically not useful in identifying malnutrition or predicting outcomes.
Interobserver reproducibility of SGNA
Interrater agreement occurred in 44 (56%) of 78 children having duplicate SGNA assessments (unweighted : 0.28). This represents only fair agreement between the assessors (34, 35). More discrepancies occurred between classifications of normal-to-moderate malnutrition (22 of 34; 65%) than moderate-to-severe malnutrition (11 of 34; 32%).
Individual components of SGNA and their contribution to the overall SGNA rating
Each of the individual historical and physical characteristics was univariately correlated with the overall SGNA ratings. For infants and toddlers, the individual SGNA components that most influenced the overall SGNA rating were physical signs of wasting, gastrointestinal symptoms, and the metabolic stress of the underlying disease. For the older age groups, controlling for age, the most influential components were physical evidence of fat wasting, serial weight loss, gastrointestinal symptoms, and stunting of height.
DISCUSSION
This study shows that SGNA is capable of identifying malnutrition, NACs, and increased length of stay in hospitalized children. As such, it is a useful clinical tool that links nutritional status to outcome and distinguishes children at risk of an adverse outcome and prolonged hospitalization. Although serum albumin showed a significant relation to the occurrence of complications, it was abnormal in only 11 children and was clinically nondiscriminatory because the mean concentrations for the 3 nutritional groups were in the normal range.
The relation of SGNA with commonly used objective measurements of nutritional status was as good as that reported in studies of SGA in adult surgical patients (6, 10, 36). A commonly overlooked marker of nutritional status is functional capacity and muscle power (37). With the use of maximal handgrip strength as a surrogate measure of muscle function, we showed that malnourished children had physical evidence of lower functional capacity than did children who were well nourished, a relation not previously shown. Hand dynamometry was easy to perform and the children enjoyed testing their strength.
Children classified by SGNA as malnourished experienced more frequent infectious complications than well-nourished children as well as more minor infectious and noninfectious complications combined. None of the objective anthropometric or biochemical markers of nutritional status, aside from albumin, showed association with complications or other morbidity. Others have also found a lack of association with biochemical markers of visceral proteins or immune function that have been said to be useful techniques of assessing nutritional status (6, 36) but are now thought to be indicators of inflammation or infection rather than malnutrition (1, 38, 39).
The finding of longer postoperative stays for children identified by SGNA as malnourished is particularly important because extended hospitalization translates into increased hospital costs where, at our center, the average cost of an inpatient day on the general surgery ward is $587. On the basis of the lengths of postoperative stay observed in this study, the average malnourished child would stay 2.9 d (55%) longer than a well-nourished child and generate added costs of $1702. Adult studies have also reported significantly longer postoperative hospitalization for malnourished persons (6, 40-42) and associated increases in hospital and home-care costs (42, 43).
In contrast with the other objective measurements, serum albumin was associated with infectious and minor complications and 2 of the 4 markers of morbidity. Although correlated with outcomes, serum albumin concentrations were well within the normal reference range for even the severely malnourished children and could not have been used clinically to identify malnutrition. Only 11 (6%) of 171 children were hypoalbuminemic. All 11 children were identified by SGNA as malnourished; hence, the addition of albumin to SGNA would not have identified any additional children as malnourished.
Concordance between assessors performing duplicate SGNA in this study was lower than desired and lower than reported in original studies of SGA in adults (agreement range: 79–91%; : 0.66–0.78) (6, 10, 36). Considerable variability was observed among the 5 dietitian assessors in terms of their years of work and pediatric experience. Unlike the original SGA studies, in which assessors were members of the gastroenterology team who frequently performed rounds and discussed patients and therapy together, none of the dietitians had previous experience with the general surgery population. Training has perhaps the most significant impact on concordance or reproducibility. After reviewing all factors, we believe that more hands-on refresher training several times throughout the study would have resulted in improved rates of interrater agreement. Despite reduced concordance based on formal testing, SGNA performed by 5 different assessors nevertheless identified children likely to have complications and prolonged hospital stay and therefore is likely to be clinically useful for this purpose. In contrast, currently used techniques for identifying malnutrition were not able to detect the risk of complications or longer length of stay.
Whereas objective measures of nutritional status cannot take into account all of the variables that a clinician should consider to identify malnutrition, SGNA represents an explicit model of the thought process to be used in assessing nutritional status. It directs the clinician to piece together the nutritional components to make a complete picture; that is, Is this child short or thin? Is it by nature or because of inadequate intake? If intake is inadequate, what is the cause? Is there physical evidence of wasting to support the historical findings? Are losses of body mass affecting the child's ability to perform? Is the child's nutritional status and functioning worsening or beginning to improve? Is the child likely to continue to have problems? What needs to be treated? As a result, after completing the SGNA and identifying the malnourished child, the clinician has determined whether the child's nutritional status is likely to improve or worsen, has established the potential cause or causes of malnutrition, and identified where to target intervention.
A possible source of bias in our study is our exclusion of non-English–speaking children and caregivers; however, it is unlikely that this would have altered our conclusions about the ability of SGNA to predict the risk of morbidity, because the bias would potentially reduce the numbers of severely malnourished children that we saw. Because such a reduction would make it more difficult to show statistical differences in outcomes, the inclusion of such patients would only reinforce our findings.
In conclusion, SGNA and its use of clinical judgment is a valid method for assessing nutritional status in children. SGNA showed good correlation with currently used objective nutritional measures of nutritional status and was the only true nutritional measure that identified children with higher incidence of infectious and minor complications and longer postoperative hospital stays, undesirable consequences for children and families that also increase hospital costs. Is SGNA valid in all children? We suggest that SGNA will be a useful tool in a wide variety of children with chronic and systemic conditions, given that our study included children with mild-to-moderate medical complexity and a wide assortment of comorbidities and underlying illnesses, most of which involved the gastrointestinal tract. Further study is needed to determine whether the fair degree of reproducibility that we observed in this study was a function of the heterogeneity of the patients, the assessors in our study, or both and whether ongoing training throughout the study period can effectively improve interrater agreement.
ACKNOWLEDGMENTS
We thank the children and their families for their participation in this research. In addition, we thank Joann Herridge, Laura Stefanizzi, Sarah Farmer, and Veronika Langos for participating as assessors for the study and the staff of the Divisions of General Surgery and Anaesthesiology of The Hospital for Sick Children for their cooperation throughout.
DJS conceived the hypothesis and study design, conducted the statistical analyses, and drafted the manuscript. KNJ contributed to the concept and design of the study and interpretation and discussion of the results. Both authors critically revised the drafted manuscript. Neither author had any conflict of interest.
REFERENCES