By Bowman, Jennifer J; Keller, Heather H
Abstract
The validity was determined for Minimum Data Set (MDS) 2.0 oral/ nutrition status (Section K) items, used to identify long-term care residents at nutritional risk. A registered dietitian assessed 128 long-term care residents using standardized procedures, and used clinical judgment to provide a nutritional risk rating. Registered nursing staff completed the MDS assessments. Bivariate tests of association were used to assess the relationship between the dietitian rating and each Section K item. The sensitivity (Se) and specificity (Sp) of specific and combinations of variables were also determined. The MDS variables of dietary prescription (diet rx), supplement use, and swallowing problems were significantly associated with nutritional risk rating. Body mass index (BMI), calculated from MDS data, also was significantly associated with nutritional risk rating. The MDS trigger system, however, had poor Se and Sp. The best combination of variables included the presence of one or more of diet rx, supplement use, swallowing problem, or BMI
(Can J Diet Prac Res 2005;66:155-161)
Rsum
On a dtermin la validit des items relatifs l’tat nutritionnel (section K) de l’ensemble minimal de donnes (Minimum Data Set – MDS 2.0) utilis pour reprer les rsidents d’tablissements de soins de longue dure risque nutritionnel. Une dittiste professionnelle a valu 128 rsidents l’aide de mthodes normalises et a utilis un jugement clinique pour fournir une valuation du risque nutritionnel. Des infirmires professionnelles ont aussi effectu les valuations l’aide du MDS. Des tests d’association bivaris ont t utiliss pour valuer la relation entre l’valuation de la dittiste et chaque item de la section K. La sensibilit (Se) et la spcificit (Sp) de chacune des variables et des combinaisons de variables ont galement t dtermines. Les variables du MDS relatives la prescription dittique, l’usage de supplments et aux problmes de dglutition ont t associes significativement au risque nutritionnel. L’indice de masse corporelle (IMC), calcul partir des donnes du MDS, tait galement associ significativement au risque nutritionnel. Le systme de signaux d’alarme du MDS prsentait toutefois de faibles Se et Sp. La meilleure combinaison de variables s’est avre la prsence de l’un ou de plusieurs des lments suivants : prescription dittique, usage de supplments, problme de dglutition ou IMC
(Rev can prat rech ditt 2005;66:155-161)
INTRODUCTION
Since 1996, the Minimum Data Set (MDS) 2.0 has been mandated for use in all Ontario chronic-care facilities. It also is used in Saskatchewan and across the United States. The oral/nutrition status portion, Section K, and specific nutrition trigger variables (Table 1) have been designed to stimulate a referral to a registered dietitian (RD) (1). An RD uses the triggers to guide care planning and interventions. The validity of the entire MDS tool in assessing nutritional risk was evaluated m one other study (2); however, this study did not validate the specific trigger mechanism within Section K. The following were evaluated in the current study: the validity of single items, the recommended trigger system, and combinations of variables from the MDS 2.0 Section K compared with clinical judgment in determining nutritional risk.
METHODS
All residents (n=215) occupying a continuing care unit (CCU) or nursing home (NH) bed in St. Joseph’s Hospital and Home in Guelph, Ontario, were invited to participate in this study. The St. Joseph’s Hospital research ethics board reviewed and accepted the study protocol, and all residents or a designated family member consented to study participation. Participants underwent a comprehensive nutritional assessment, which included a review of their medical chart for weight, medical history, diagnoses, problems, and medication use (3).
Anthropometry and body composition
A trained dietitian completed a standardized anthropometric assessment, including triceps and subscapular skinfold measurements, knee-height measurement, and wrist, calf, and mid-upper-arm circumference measurement. A Lange caliper (Cambridge Instruments, Cambridge, MD) was used for skinfold measurement. Weight was obtained from medical charts, provided that the most recent weight had been recorded within 30 days. Otherwise, weight was measured with ward scales on the assessment day, using standard procedures where minimal clothing was worn (3). Calibration of scales was not controlled. Standing height was estimated from knee-height measurement with a Ross knee-height caliper (3,4), and calculated using recommended formulas (5).
Table 1
Minimum Data Set 2.0 trigger variables suggesting malnutrition risk
Biochemical indicators
The biochemical indicators of nutritional status included serum cholesterol level, albumin and hemoglobin testing, hematocrit, and total lymphocyte count (TLC).
Food intake and eating problems
An RD observed each participant during one random meal period. Length of time required to eat, type and degree of assistance required, texture of meal provided, behaviour during the meal, number of adaptive feeding tools required, and percentage of meal consumed were recorded.
The comprehensive nutritional assessment, summarized above, occurred within four weeks of each resident’s quarterly MDS review. The dietitian rated nutritional risk on a ten-point scale where each interval indicated approximately a 10% increase in risk (1=lowest risk, 10=highest risk). To provide standard criteria for the dietitian’s clinical judgment rating, criteria were modified from a validation study on use of clinical judgment of nutritional status in community-dwelling seniors (6) and an assessment of nutritional status in long-term care facilities (7,8) (Table 2).
Completion of Minimum Data Set
Trained nursing staff completed MDS Section K for continuing care and nursing home participants. In the NH section of the facility, one nurse completed all MDS forms. In the CCU, two nurses completed the forms. Inter-rater reliability of MDS data was not assessed. The MDS data, including the most recent weight (within 30 days) and height (measured on admission), were collected from patient charts. The dietitian performing the assessments and rating was blinded to data from the MDS appraisal.
Statistical analysis
Bivariate tests of association (t-tests, correlation and analysis of variance) were used to assess the relationship between the RD’s rating of degree of nutritional risk for malnutrition (1 to 10) and each MDS 2.0 Section K variable. Sensitivity (Se) and specificity (Sp) of each of the top three individual predictor variables, based on strength of association and all combinations of the three variables, were determined at a cut-point at or above five, indicative of a moderate or higher level of nutritional risk (9). To determine the precision of the Se and Sp estimates in the sample population, the 95% confidence interval (CI) based on the standard error of the estimate was considered (10). The Statistical Package for the Social Sciences (SPSS Inc., version 10, Chicago, IL) was used for all statistical analyses, and associations were considered statistically significant at p≤0.05.
RESULTS
Of the 215 residents, 129 (60%) agreed to participate. One resident died during the study period, before assessment was completed. Descriptions of participants stratified by bed location are shown in Table 3; data were not collected for non-participants.
The RD’s assessment revealed a mean overall nutritional risk rating of 6.1 (standard deviation 1.8), indicative of moderate risk; 15.7% of the study participants were classified as being at a low level of nutritional risk (rating 7).
The range m prevalence of Section K items was wide, from 1.6% to 44.5% (Table 4). The most frequent nutritional indicators, as indicated by the RNs’ completion of Section K, were supplement use between meals (44.5%), a diet prescription for parenteral nutrition (PN) or enteral nutrition (EN), a mechanical diet, or syringe feeding (36.7%), a chewing problem (29.7%), and a swallowing problem (22.7%). Associations between the RD’s rating of degree of nutritional risk and each of the Section K variables are shown in Table 4. Independent sample t-tests revealed that the only MDS variables significantly associated with nutritional risk rating were
* a swallowing problem (t=-2.013, p
* a diet prescription for PN, EN, or a mechanically altered or syringe diet (=-4.249, p
* supplement use (a nutritional supplement provided between meals) (t=-2.862, p
Because of the low prevalence of several trigger items, a loss of power was observed in the bivar\iate analysis.
Weight change and risk
Participants who had lost weight had a higher mean risk rating than those who had not, and those who had gained weight had a lower mean risk rating than those who had not (Table 4). However, weight loss or gain was documented by the MDS in only 2.3% of participants (n=3). Conversely, the RD documented that 34.4% of participants had experienced some weight loss in the previous three months – 29.7% in the previous six months and 30.5% in the previous 12 months. Because of this discrepancy and suspected problems with the MDS-documented percentage weight change, the association between weight loss and nutritional risk was determined using the RD’s assessment. Any degree of weight loss, as determined by the dietitian, was significantly associated with degree of , nutritional risk (t=- 2.47, p
Table 2
Standardized criteria for nutritional risk rating by dietitian (6)
Table 3
Selected demographic and health characteristics of study participants
Body mass index calculation
Completion of the MDS 2.0 Section K assessment does not require calculation of body mass index (BMI); however, BMI was calculated from height and weight recorded on the form by nursing staff and was included in the bivariate analysis. Typically, the nurse used the admission height and most recent weight recorded on the chart. The dietitian did not use this BMI from the MDS form, as she estimated height from knee-height measurements, rather than relying on the more questionable charted height. However, correlation between the MDS BMI and the dietitian’s calculated BMI was high (Spearman’s rho=0.876, p
According to MDS-recorded height and weight, approximately 51% of the sample had a BMI at or below 24 kg/m^sup 2^ and 28% had a BMI at or below 20 kg/m^sup 2^. A BMI at or below 24 kg/m^sup 2^ (calculated from MDS data) was significantly associated with a higher level of nutritional risk (t=-3.845 and t=-3.751, respectively; p
Documentation suggests that malnutrition risk is indicated through MDS 2.0 assessment if any one of the eight “trigger” variables in Table 1 is present. Seven of these triggers are present in Section K. With the current trigger method, at the moderate risk cut-point (rating ≥5 on the dietitian’s nutritional risk scale), Se and Sp were modest (Se=0.56 and Sp=0.75) (Table 5). This analysis indicates that any one of the seven trigger variables recommended for use by MDS 2.0 in Section K exhibits only modest Se for risk screening.
Swallowing problem, diet prescription, or supplement effects
The next step in the analysis was determining the Se and Sp for the three variables shown in the previous bivariate analysis to be significantly associated with nutritional risk – a swallowing problem, a diet prescription for PN, EN, or mechanical or syringe diet (diet rx), and a nutritional supplement between meals. Individually, the Se and Sp of the variables were poor (data not shown). The highest Se and Sp were seen for the combination including any or all of the three variables (Se=0.69, Sp=0.60, at the moderate [cut-point 5] risk level) (Table 5).
Because the bivariate analysis showed that BMI calculated from MDS-recorded height and weight was significantly associated with nutrition risk, the BMI was included in the Se and Sp analysis. A BMI at or below 24 kg/m^sup 2^ had greater Se than did any other single variable; Se was 0.56 at the moderate nutritional risk cut- point of five. The best overall Se and Sp (0.81 and 0.50, respectively) for BMI at or below 24 kg/m^sup 2^ were observed in combination with a diet rx, supplement use, and/or a swallowing problem.
Weight loss and body mass index analysis
To determine the Se and Sp of Section K further, in an attempt to be as inclusive as possible, MDS-indicated weight loss (either 5% in the previous 30 days or 10% in the previous 180 days) and BMI also were compared with the dietitian rating. Weight loss was included in this analysis despite the nonsignificant association found with the bivariate analysis, as it seemed to be an intuitively important indicator of nutritional risk. On its own, the Se of weight loss as an indicator of nutritional risk was extremely low: 0.03 at the moderate risk cut-point of 5. When evaluated in combination with other variables, the Se improved, ranging from 0.25 to 0.83. The best overall Se and Sp (0.68 and 0.70, respectively) with the weight loss variable were seen in combination with a diet rx and/or a supplement.
Confidence intervals for key variables
The 95% confidence intervals for Se and Sp estimates for key variables also are presented in Table 5. Confidence intervals for these estimates suggest that alternative trigger variable combinations are more useful than the current system for nutritional risk screening. However, Sp confidence intervals are wide, and overlap for all mam indicator combinations assessed, indicating that false-positives are a consistent problem.
Table 4
Descriptive and bivariate analysis of MDS characteristics compared with dietitian risk rating
Positive and negative predictive values
The positive predictive value (PPV) and negative predictive value (NPV) were calculated for each of the MDS variables listed in Table 5. For screening purposes, NPV, or the probability of not having the disease when the test result is negative (normal), is the main concern: a high NPV is desirable (9). Calculated NPVs are low to modest, ranging from 0.20 to 0.33, for the variables indicated by Se and Sp analysis to be best for screening. These NPV values confirm Se and Sp analysis-based results that the best set of variables to use as triggers includes diet rx, supplement use, swallowing problems and/or BMI
SUMMARY
When compared with the nutritional risk rating at or above five, which indicates moderate or higher risk, the best set of indicator variables includes the variable representing a BMI below 24 kg/ m^sup 2^, calculated from Section K height and weight, with the current trigger variables diet rx, supplement use, and/or swallowing problem (Se=0.81, Sp=0.50).
DISCUSSION
The Se of individual MDS 2.0 Section K variables was poor; however, greater Se and Sp were observed when combinations of variables were used. This finding is consistent with the current MDS trigger mechanism. The Se and Sp of the variables currently used in this trigger mechanism were poor, at 0.56 and 0.75 respectively. This indicates that the current mechanism is unable to differentiate adequately between individuals who are at nutritional risk and those who are not. Bivariate analyses found that only three Section K items were significantly associated with an RD’s rating of nutritional risk. The combination of diet rx, supplement, swallowing problem, and BMI less than 24 kg/m^sup 2^ provided the best Se for screening purposes (0.81). Only diet rx (which consists of parenteral/intravenous feeding, enteral feeding, and mechanically altered or syringe diet) is currently used as a trigger on the MDS 2.0. Notably, this four-component combination had greater Se than the current seven trigger variables in Section K.
Table 5
Sensitivity, specificity, and positive and negative predictive values (and 95% CI) of key variables and combinations of variables using the moderate risk (≥5) cut-point
In comparison with the RD’s assessment of nutritional risk and specific nutritional status indicators, the current MDS 2.0 Section K did not adequately document nutritional risk elements within the sample population. The prevalence of MDS variables in the study participants ranged from 1.6% for complaints about hunger, to 44.5% for a supplement between meals. The overall prevalence of MDS items differs from the estimation of nutritional risk (84% at moderate to high risk) and the prevalence of similar characteristics assessed by the dietitian. For example, the prevalence of weight loss (either 5% in 30 days or 10% in 180 days) was 2.3% according to the MDS assessment completed by nurses. In contrast, the dietitian assessment documented the prevalence of any weight loss to be 34.4% in the previous three months, and 29.7% in the previous six months. Although they are not a direct comparison, these results suggest a large discrepancy between the nurses’ and the dietitian’s assessments, even though they used the same raw data from the medical chart. In addition, missing information meant that weight change at three, six, and 12 months could not be determined for all residents. Although weight is supposed to be recorded for each resident monthly, apparently this is not done consistently.
As indicated by the high proportion of participants with a BMI below 24 kg/m^sup 2^ (51%), many study participants already are below their optimal body weight. Perhaps the calculation of percent weight change therefore is unnecessary and is contributing to documentation errors. Beck and Ovesen (11) reviewed several studies of weight change in older adults and suggested that even a small weight loss (i.e., 1%) was associated with increased mortality. A simpler and perhaps more effective Sectio\n K indicator than percent weight change would be an indicator detecting any degree of weight loss.
Discrepancies between nurses’ and the RD’s weight loss assessments indicate the importance of training. Previous work has identified accuracy issues with height and weight recorded by nurses on the MDS (12). In the Ontario census MDS data collection, Keller and Hirdes eliminated 638 patients (approximately 5%) with a missing or inconceivable weight or height, a finding that suggests measurement or recording problems (12). Improved implementation and training conceivably may have led to MDS data more consistent with the dietitian’s assessment.
No other researchers have examined the ability of MDS 2.0 Section K trigger variables to identify nutritional risk. Currently, one other group has attempted to establish the validity of the MDS 2.0 for assessing nutritional status in NH residents. Blaum et al. (2) used convergent and construct validity to determine how MDS 2.0 items were associated with biomedical measures of nutritional status, and how the anthropometric measures (weight and BMI) recorded on the MDS were associated with other MDS clinical characteristics, some of which are not included in Section K. Like the current study, the Blaum et al. study showed that BMI was a useful measure of nutritional status because it was significantly correlated with other anthropometric measures not available on the MDS, such as mid-arm muscle area, percent body fat, and fat-free mass (2). In addition, Blaum et al. found that the lowest quartile of BMI measures was significantly associated with poor oral intake, weight loss, advanced cognitive impairment, and pressure ulcers. The best overall combination of MDS variables in the current study included BMI, a variable not currently calculated from the MDS data. However, low Sp and NPVs (Table 5) remained a problem with the “best combination,” which indicates the MDS 2.0 has poor efficiency as a screening tool.
A limitation of the study design is the introduction of criterion combination bias, which is inevitable, secondary to the nature of the gold standard used for comparison. The best available gold standard is a comprehensive nutritional assessment, which inherently includes evaluation of some items that may be familiar to both instruments.
RELEVANCE TO PRACTICE
These results suggest that the addition of BMI to Section K, as well as the use of BMI, supplement use, and swallowing problems with the current trigger variables, would improve the MDS as a nutritional risk-screening tool. The number of variables used as triggers could possibly be reduced to those shown to be significantly associated with nutritional risk. Because of the low prevalence of some trigger items in the sample population, this study lacks sufficient evidence to indicate precisely which trigger variables should be removed. In addition, other research has suggested that cognitive impairment is related to lower nutritional status (2,12-14); however, cognitive impairment is not included in Section K. To determine conclusively the usefulness of Section K as a nutritional risk screening tool, continued research and development are necessary.
Acknowledgement
The St. Joseph’s Hospital Foundation and InterRAI provided financial support for this research.
References
1. Canadian Institute for Health Information. Minimum Data Set v 2.0 user’s manual. Toronto; 1996.
2. Blaum CS, O’Neil EF, Clements KM, et al. Validity of the Minimum Data Set for assessing nutritional status in nursing home residents. J Clin Nutr 1997;66:787-94.
3. Gibson RS. Principles of nutritional assessment. New York: Oxford University Press; 1990.
4. Chumlea WC, Roche AF, Steinbaugh ML. Estimating stature from knee height for persons 60-90 years of age. J Am Geriatr Soc 1985;33:116-20.
5. Chumlea WC, Guo S, Wholihan K, et al. Stature prediction equations for elderly non-Hispanic white, non-Hispanic black, and Mexican-American persons developed from NHANES III data. J Am Diet Assoc 1998;98:137-42.
6. Keller HH, McKenzie JD, Goy R. Construct validation and test- retest reliability of SCREEN (Seniors in the Community: Risk Evaluation for Eating and Nutrition). J Gerontol 2001;56A(9):M552- 8.
7. Kerstetter JE, Holthausen BA, Fitz PA. Malnutrition in the institutionalized older adult. J Am Diet Assoc 1992;92:1109-16.
8. Keller HH. Use of serum albumin for diagnosing nutritional status in the elderly – is it worth it? Clin Biochem 1993;26:435-7.
9. Sackett DL, Haynes RB, Tugwell P. Clinical epidemiology. A basic science for clinical medicine. Boston: Little, Brown and Co; 1985.
10. Fletcher RH, Fletcher SW, Wagner EH. Clinical epidemiology. The essentials. 2nd ed. Baltimore: Williams & Wilkins; 1988.
11. Beck AM, Ovesen L. At which body mass index and degree of weight loss should hospitalized elderly patients be considered at nutritional risk? Clin Nutr 1998;17:195-8.
12. Keller HH, Hirdes JP. Using the Minimum Data Set to determine the prevalence of nutrition problems in an Ontario population of chronic care patients. Can J Diet Prac Res 2000;61:165-71.
13. Keller HH. Malnutrition in institutionalized elderly: how and why? J Am Geriatr Soc 1993;41:1212-8.
14. Ortega RM, Requejo AM, Andres P, et al. Dietary intake and cognitive function in a group of elderly people. Am J Clin Nutr 1997;66:803-9.
JENNIFER J. BOWMAN, MSc, RD, HEATHER H. KELLER, PhD, RD, Department of Family Relations and Applied Human Nutrition, University of Guelph, Guelph, ON
Copyright Dietitians of Canada Fall 2005
Comments