Does the type of disease and diet influence hospital weight loss? An analysis using the Generalized Estimating Equation method

— Investigate weight evolution during hospitalization and its association with the type of disease and diet, over time. Retrospective study with 133 in-patients, investigating indicators of nutritional status, diet, disease, surgery and the evolution of body weight during hospitalization. The chi-square, Kruskal-Wallis tests and the method of Generalized Estimating Equation (GEE) were used for the study of weight evolution over time, depending on the diet and type of disease. Underweight according to body mass index, malnutrition by subjective global assessment and nutritional risk according to nutritional risk screening (NRS), were found in 19.5%, 27.8% and 46.6% patients, respectively. There was no influence of the type of diet on weight evolution over time (p=0.0779),


INTRODUCTION
Many studies have shown that hospitalized patients endure malnutrition, and ESPEN (European Society of Clinical Nutrition and Metabolism) published a consensus [1] on malnutrition in hospitalized patients, emphasizing the importance of some anthropometric indicators, such as the reduction of the body mass index (BMI) and this reduction was also considered for the evaluation of malnutrition by the World Health Organization (WHO) [2].
Other instruments have been used to determine nutritional risk in hospitalized patients, such as nutritional risk screening (NRS), subjective global assessment, among others [3,4,5,6,7]; the relevance of NRS was highlighted in a recent study [8] that showed that the reduction of BMI and recent weight loss were the factors that most contributed to the diagnosis of nutritional risk and the joint use of NRS with serum albumin [9] also proved to be useful in patients' nutritional screening A reduced risk of hospitalization was associated with weight gain while weight loss was associated with a higher rate of hospitalization, especially of low weight patients [10]. Another interesting recent study [11] that investigated characteristics and outcomes of malnourished patients, pointed out that malnutrition is still underestimated, and that among the criteria used to identify malnourished patients the most important were the reduction of energy consumption and weight loss. The objective of this study was to investigate the weight evolution during hospitalization and its association with the type of disease and diet, analyzed over time using the Generalized Estimating Equations (GEE) method.

Study design and ethical aspects:-
A retrospective study was carried out with data collection in medical records of adult patients hospitalized by the Unified Health System (SUS), in a surgery ward of a university hospital in the State of São Paulo, Brazil. The indicators of nutritional status, type of prescribed diet, type of disease and surgery, and body weight records during hospitalization were surveyed. The study was approved by the Research Ethics Committee (CEP) of the institution.

Study participants:-
We included hospitalized patients of both genders over the age of 20 years, without terminal illnesses, with at least three records of measurement of body weight on different days of hospitalization and with all clinical, nutritional data and type of prescribed diet, duly registered in the medical records. Patients admitted only for clinical investigation and those who had not undergone nutritional assessment in the first 48 hours of hospitalization were excluded. Thus, the sample size included 133 in-patients.

Procedures:-
All screening and risk assessment and nutritional status instruments were performed within the first 48 hours of hospitalization, using the nutritional risk screening (NRS), subjective global assessment (SGA), anthropometry and laboratory tests (hemoglobin, hematocrit and lymphocyte count). Anthropometric measurements included: body weight, body mass index (BMI), arm circumference (AC), triceps skinfold (TSF) and arm muscle circumference (AMC). The diagnosis at admission (type of disease), type of surgery, type of prescribed diet (whether oral, enteral or parenteral diet and fasting), prescription of hypercaloric and hyperproteic nutritional supplement, length of stay, age and gender were investigated. All this information was properly documented in the medical records of the hospital.
For the classification and analysis of the evaluated indicators, the cutoff points standardized by the relevant literature for the NRS (with and without nutritional risk) were adopted [12,13]; SGA (well nourished and malnourished) [14]; BMI (normal, overweight and underweight, for adults and the elderly) [2,15]; AC, TSF and AMC (percentile classification ≤ P15, P15-P85 and ≥ P85) [16,17] and laboratory tests [18]. Finally, body weight measurements on 3 different days of hospitalization were recorded among all the variables studied.

Statistical Analysis:-
Initially, the characteristics of the studied population were described and compared in relation to the type of diet prescribed. Subsequently, weight gain was related to the type of diet prescribed and the type of disease.
Categorical variables were expressed as frequency and percentage, and continuous variables were expressed as mean and standard deviation. To compare proportions, the Chi-square test or Fisher's exact test was used; when necessary. To compare the numerical measures summarized between the groups on the prescribed diet, the Kruskal-Wallis test was used, followed by the Dunn test to locate the differences; when necessary. For the study of weight evolution during the hospitalization period, depending on the prescribed diet and the type of disease, the Generalized Estimating Equations (GEE) method was used in this study. The level of significance adopted for the statistical tests was 5% [19,20].
Regarding nutritional indicators at the beginning of hospitalization, 19.5% patients (n=26) were underweight according to BMI and 39.1% (n=52); 20.3% (n=27) and 42.1% (n=56), were classified as equal to or below the 15 th percentile (≤P15) for AC, TSF and AMC, respectively. According to the SGA, 27.8% (n=37) patients were malnourished and with the NRS it was found that 46.6% (n=62) patients were at nutritional risk. Regarding the type of diet at the beginning of hospitalization, 66.2% (n=88) of the patients were on oral diet; 16.5% (n=22) on enteral and parenteral diet and 17.3% (23) on oral fasting. The use of oral nutritional supplementation was observed in 6.8% (n=9) of the patients. No patient died.
A comparison was made between all the variables studied (numerical, categorical variables, body composition and nutritional status at the beginning of hospitalization) and the types of diet in the first dietary prescription, with the aim of verifying whether any characterization variable at the beginning of hospital stay, was associated with the type of diet. When the relationship between the three types of diets prescribed in the first evaluation and all the variables was assessed, there was a significant difference only in body weight (p=0.0470), BMI (p=0.0147) and hemoglobin (p=0.0377) (analyzed by the Kruskal-Wallis test, followed by the Dunn test to find the differences); in gender (p=0.0152), type of disease (<0.0001), surgery (p=0.0007), in AC (p=0.0094) and in SGA (p=0.0278) (analyzed by the chi-square test).
In the relationship between the evolution of body weight during hospitalization and the type of diet prescribed, analyzed by the GEE method, there was no significant influence of the type of diet on weight evolution over time (hospital stay) (p=0.0779) ( Table 1). In relation to the evolution of body weight over time (hospital stay) and the type of disease, there was a statistically significant difference (p = 0.0100), using the GEE method (Table 2). It was found that in digestive neoplasms, the reduction in body weight over time was greater in relation to other diseases. A:-first prescription; B:-second prescription; C:-third prescription.
p=0,0100 value (GEE); significant difference in body weight for diseases. Differences between: -digestive tract disease and digestive neoplasia; fracture and digestive neoplasia; colon rectal cancer and digestive neoplasm.

IV. DISCUSSION
This study investigated the evolution of body weight during hospitalization and its relationship with adult in-patients' clinical and nutritional status variables. As evidenced by the findings, NRS was more sensitive than the other indicators in the diagnosis of malnutrition. The initial hypothesis of this investigation was to assess any influence of a few indicators of nutritional status and the type of diet prescribed on the weight evolution during hospitalization, and no influence was observed here. However, the type of disease on the evolution of body weight over time was observed, with a greater reduction in weight over time in digestive neoplasm; in relation to other diseases. In the present study, the GEE method was adopted as the method of analysis for the study of body weight over time, as a function of the type of diet prescribed and the type of disease. The selection of this method of analysis allowed better adjustments in relation to the distribution of variables. With this method of analysis (GEE), it was possible to model both the response (body weight) and the factor under study (diet prescribed and type of disease), with repeated measurements (assessments over time).
In a retrospective study [11] with adult patients, it was observed that 62% of the patients exhibited malnutrition due to chronic disease, and the most common criteria for identifying malnourished patients were weight loss and reduced energy consumption. It has been reported that during hospitalization, significant changes in nutritional status can occur [21], with important changes in anthropometric and laboratory indicators.
As indicated by ESPEN [1], low BMI in conjunction with weight loss can constitute a diagnosis of malnutrition in patients at nutritional risk; in addition, the Malnutrition Universal Screening Tool (MUST) [22] related to the ESPEN criteria for the definition of malnutrition should also be considered.
Multivariate models showed a decline in nutritional status [23] by SGA and loss of body weight, being significantly associated with longer hospital stays. In an investigation of the prevalence of malnutrition with the criteria of weight loss and BMI [24], preoperative malnutrition was associated with an increased risk of serious complications. And an additional study [25] further pointed out that self-reported weight loss can be considered an important prognostic factor in hospitalized patients. Another study that investigated the effects of nutritional support in hospitalized patients with malnutrition [26], reported increased energy and protein consumption, and body weight. There are also studies showing that there was no influence of the type of diet on weight loss in hospitalized patients [27] and that weight loss during hospitalization was related only to sex and type of disease [28].
The relevance of assessing body weight loss has also been reported in several studies with different clinical situations [29], which can also be associated with other markers of malnutrition, such as laboratory tests, body composition and muscle strength assessment, energy consumption and nutritional screening instruments assessment [30].
The limitations of this study include the loss of some data, inherent to the type of retrospective study and data collection in medical records. Another limitation refers to the fact that acceptance of the hospital diet was not evaluated, but only the dietary prescription adopted at the beginning of hospitalization. However, this type of retrospective observational study has the advantage of being able to show the findings and results of the hospital routine, as it actually occurs in the hospital, and not in a controlled manner.

V. CONCLUSION
The NRS was more sensitive in the diagnosis of malnutrition. The type of disease influenced weight loss, but the type of diet did not influence the evolution of body weight over time.

ACKNOWLEDGMENTS
We are grateful to the Pontifical Catholic University of Campinas-SP-Brazil.