SCREENING, IDENTIFICATION, DIAGNOSIS OF MALNUTRITION IN HOSPITALIZED PATIENTS WITH SOLID TUMORS: A RETROSPECTIVE COHORT STUDY

Abstract

Background

Malnutrition is common in hospitalized cancer patients and adversely affects clinical outcomes. We evaluated the prevalence of malnutrition risk, dietitian-identified malnutrition (DIMN), and physician-diagnosed malnutrition (PDMN) at admission.

Methods

This retrospective study included a consecutive cohort of all adults diagnosed with a stage I-IV solid tumor malignancy and admitted to Atrium Health Carolinas Medical Center from January 2016 to May 2019. Data were collected from the first admission closest to the cancer diagnosis date. Malnutrition risk was determined by a score ≥2 on the Malnutrition Screening Tool (MST) administered by a registered nurse during the intake process. Registered dietitian nutritionist (RDN) assessments were reviewed for DIMN and grade (mild, moderate, severe). PDMN included malnutrition ICD-10 codes in the discharge summary. Univariate models were estimated; multivariate logistic regression models identified associations between clinicodemographic factors and malnutrition prevalence with stepwise selection.

Results

5143 patients were included. Median age was 63 (range 18-102) years; 48% female; 70% White and 24% Black. Upper gastrointestinal (21%), thoracic (18%), and genitourinary (18%) cancers were most common. 28% had Stage IV disease. MST scores were available for 4085 (79%); 1005/4085 (25%) were at malnutrition risk. 11% (n=557 of 5143) had malnutrition coded by a physician or documented by an RDN; 4% (n=223 of 5143) of these were identified by both clinicians, 4% (n=197 of 5143) by RDNs only, and 3% (n=137 of 5143) by physicians only.

Conclusions

Malnutrition appears to be underdiagnosed by both RDNs and physicians. Underdiagnosis of malnutrition may have significant clinical, operational, and financial implications in cancer care.

INTRODUCTION

Malnutrition (undernutrition) in patients with cancer leads to unfavorable body composition changes and functional impairment due to inadequate food intake and metabolic dysregulation.¹ In the United States (US), patients with cancer are among the most malnourished of all patient groups, with up to 80% experiencing unintentional weight loss,^{2, 3} and they have the largest number of all-cause hospital readmissions.⁴ Furthermore, once patients with cancer are admitted to a hospital, their nutritional status often worsens due to inadequate food intake, treatment side effects, and suboptimal nutritional management.⁵ Malnutrition is associated with greater treatment toxicity, lower treatment response, more complications, poor quality of life, and shorter survival.⁶ The cost of care for malnourished patients is substantially increased with longer hospital stays and more resource-intensive interventions.^7–9 In one study, malnourished patients had an 8-fold greater risk of mortality and prolonged length of stay with higher hospital costs.¹⁰ Despite the known poor outcomes of malnourished patients, malnutrition remains underrecognized and undertreated¹¹; only 40% of those who are malnourished are identified during chemotherapy, and 50% during hospital stay by clinicians.¹¹

Timely malnutrition screening and referral to a registered dietitian nutritionist (RDN) improves patient outcomes.^{12, 13} Malnutrition screening, assessment, and intervention during hospital stays and discharge planning are important components of the nutrition care process. Although guidelines and positions^12–14 recommend consistent malnutrition screening to identify those who may have or are at risk for malnutrition, screening is not a standardized component of cancer care in the US.¹⁵ For example, nutrition screening within 24 hours of admission is a requirement by the Joint Commission on Accreditation of Healthcare Organizations¹⁶, but implementation of this standard varies across healthcare facilities. Screening is further complicated by infrequent employment of RDNs in cancer centers, and inconsistent inclusion of medical nutritional therapy in multidisciplinary cancer care. Variations in definitions and tools used to screen and diagnose malnutrition continue to be a barrier to process standardization in healthcare settings. These issues can negatively impact patient outcomes and healthcare quality, underscoring the need for improved nutritional care.

The prevalence of malnutrition varies significantly across different cancer types, which has important implications for patient outcomes and treatment approaches. Factors such as tumor location and stage, treatment modalities, and care settings influence this variability.^{3, 17, 18} Conventionally, cancer-related malnutrition was evidenced by unintentional weight loss and experienced in the majority of patients with gastroesophageal, pancreatic, head and neck, and lung cancer.^18–20 These cancers often cause dysphagia (difficulty swallowing), early satiety, and gastrointestinal symptoms like nausea and vomiting, which severely impact nutritional intake¹⁸. Patients who have treatment plans with high symptom burden, like hematopoietic transplantation, are also at greater risk.²¹ As cancer progresses to more advanced stages, the risk of malnutrition increases, such as advanced colorectal cancer.²² Bossi et al.²² showed that malnutrition could affect up to 75% of patients with locally advanced or metastatic cancer, with the highest weight loss among those with tumors of the esophagus (57%), stomach (50%), and larynx (47%). The varying prevalence of malnutrition across cancer subtypes underscores the need for cancer-specific nutritional strategies.

Most evidence on hospital malnutrition risk and prevalence has been derived from convenience or non-representative samples.^{23, 24} This approach limits the generalizability of findings, as these samples may not accurately reflect the broader hospital population. The majority of hospital-based cohort studies have focused on specific subpopulations (e.g., advanced cancer, gastrointestinal, and head and neck cancers) or have been conducted outside the US, where cancer nutrition care differs from the US.^25–31 This geographic variation can lead to differences in malnutrition prevalence and management strategies. Conducting region-specific studies in the US will provide a more accurate understanding of the prevalence and risk factors for malnutrition among cancer patients.

The objectives of this study were to evaluate the prevalence of malnutrition among cancer inpatients as determined by both RDNs and physicians, and to examine the clinical characteristics of these patients at admission. The effect of quality improvement (QI) interventions on malnutrition identification over time was also evaluated.

METHODS

Design

Clinical data from electronic medical records (EMR) were retrospectively reviewed for a consecutive cohort of all adults diagnosed with a stage I-IV solid tumor malignancy and admitted to Atrium Health Carolinas Medical Center between January 2016 and May 2019 over 3 yearThe EMR was queried for inpatient encounters seven months before and five months after the diagnosis period to capture admissions occurring within approximately six months of a cancer diagnosis (Figure 1). People with a primary diagnosis of hematological malignancy or multiple solid tumor malignancies and those who attended only ambulatory visits were excluded. The malnutrition care workflow was used as a framework for implementing malnutrition QI The Institutional Review Board for Atrium Health approved this protocol (IRB #09-20-10E) and waived the requirement for informed consent. The Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) Statement guided this manuscript

FIGURE 1. STUDY FLOW AND DESIGN*.

*The EMR was queried for inpatient encounters seven months before and five months after the diagnosis period to capture admissions within approximately six months of a cancer diagnosis.

Setting

to Atrium Health Carolinas Medical Center is one of the largest health systems in the US southeast region, with 40+ acute care facilities and 61 cancer care locations. to Atrium Health Carolinas Medical Center is a high-volume, not-for-profit, academic medical center with 700 licensed beds and 16 RDNs employed to provide general inpatient nutrition services.

Data Collection and Processing

The study utilized EMR (Cerner Corporation, MO, USA) stored in the enterprise data warehouse (EDW) and the institutional tumor registry to identify unique eligible patients with complete case records. Disease-specific data (cancer diagnosis date, primary cancer diagnosis, disease extent, and stage) were obtained from the institutional tumor registry. EDW was queried to identify hospitalizations (at least 24 hours of inpatient admission). Patient residence ZIP Codes (at diagnosis) were matched to the 2015-2019 American Community Survey³⁴ to generate individual-level proxies for educational attainment (i.e., percent of the population ≥25 years graduating high school) and household income (i.e., estimated median household income in dollars).

All data were collected at first admission closest to cancer diagnosis, including age (years) at cancer diagnosis, sex, race, ethnicity, body mass index (BMI) (kg/m²), and malnutrition-related diagnoses. Cancer diagnostic groups were established, which included breast, genitourinary (GU), gynecologic (GYN), head and neck (H&N), lower gastrointestinal (GI), thoracic, upper GI, and other.

Malnutrition Risk Screening

Malnutrition risk screening was determined via the Malnutrition Screening Tool (MST) administered by a registered nurse. This assessment could help establish if the patient would benefit from a more thorough nutrition assessment. The MST assesses two primary domains including unintentional weight loss and appetite (Appendix 1). MST scores were extracted from the nursing intake form on admission. MST scores ranged from 0 to 5, with scores ≥2 indicated malnutrition risk.³⁵ The MST threshold for malnutrition risk and the Academy of Nutrition and Dietetics/American Society for Parenteral and Enteral Nutrition (AND/ASPEN)³⁶ (Appendix 2) criteria for malnutrition diagnosis were constant throughout the study.

Malnutrition Identification by Registered Dietitian Nutritionist and Diagnosis by Physician

Individuals who were deemed at risk for malnutrition were further assessed by RDN to confirm the presence of malnutrition and determine its severity. Malnutrition assessment contained both anthropometric and historical measurements of nutrition status. The clinical criteria for malnutrition identification were based on the AND/ASPEN consensus statement.³⁶ These included 6 key characteristics: 1) diminished functional status as measured by handgrip strength, 2) evidence of inadequate intake, 3) fluid accumulation, 4) muscle loss, 5) subcutaneous fat loss, and 6) unintentional weight loss. The presence of at least 2 out of 6 AND/ASPEN criteria were recommended for a malnutrition diagnosis. While AND/ASPEN does not specifically define mild malnutrition, RDN can infer mild malnutrition by applying less severe thresholds to the same indicators used for moderate and severe malnutrition.

Dietitian-identified malnutrition (DIMN) and malnutrition severity (mild, moderate, severe) were obtained from the RDN documentation when malnutrition was recognized Malnutrition severity data were extracted from structured data fields (checklists) in the RDN nutrition assessment note, while the data fields for the AND/ASPEN criteria for malnutrition identification were unstructured (free text).

Upon reviewing the RDN nutrition assessment, the physician documents malnutrition in the progress notes and assigns severity (mild, moderate, severe) in the discharge summary (Appendix 3). Mild malnutrition was diagnosed by applying less severe thresholds to the same indicators used for moderate and severe malnutrition. Physician-diagnosed malnutrition (PDMN) was obtained from the International Classification of Diseases-Tenth Revision (ICD-10) malnutrition diagnoses codes in the discharge summary. These codes included: 1) mild protein-calorie malnutrition (E44.1), 2) moderate protein-calorie malnutrition (E44), 3) unspecified protein-calorie malnutrition (E46), and 4) unspecified severe protein-calorie malnutrition (E43).

Institutional Documentation of Malnutrition

In January 2016, Atrium Health adopted the AND/ASPEN criteria to assess nutrition status of inpatients and outpatients. In May 2016, Atrium Health implemented the validated MST (scores 0-5) in the acute care inpatient setting. Upon admission, those with an MST score ≥2 automatically receive an oral nutrition supplement. MST scores ≥3 triggered a comprehensive inpatient nutrition assessment by an RDN. Some patients were diagnosed in the absence of screening at the discretion of the clinician.

The Atrium Health Nutrition Care Plan and Quality Improvement Projects (QIPs) are presented in Appendix 3. Two QIPs were implemented during the study period to enhance the diagnosis and documentation of malnutrition in patients:

1. August 2016: The first strategy involved RDN sending messages to physician via the EMR system. This communication was intended to confirm the malnutrition diagnosis initially made by the RDN. The goal was to ensure that the physician reviewed and agreed with the RDN’s assessment of the patient’s nutritional status. (Appendix 4).

2. April 2018: In the second strategy, the role of direct messaging from RDN to physician was replaced by the Clinical Documentation Integrity (CDI) team. The CDI team notifies physicians directly through an integrated coding EMR system when a patient meets the criteria for a malnutrition diagnosis. The physician then determines if they agree with the presence of malnutrition (yes/no) and severity (mild/moderate/severe) and adds it to the discharge summary. This change streamlines the workflow and ensures that physicians receive standardized, timely, and accurate information about patients’ nutritional assessments directly within their existing workflow. (Appendix 5).

Statistical Analysis

Data are presented as frequencies (percentages) and medians (ranges). The proportion of subjects at risk of malnutrition by MST and the prevalence and severity of DIMN and PDMN were summarized as frequencies (percentages). The associations of MN risk by MST and DIMN/PDMN were evaluated by Fisher’s Exact tests. Agreement in malnutrition identification was defined as the absence or presence of both DIMN and PDMN; agreement in severity of malnutrition was defined as the absence of DIMN and PDMN or the agreement in presence and severity of DIMN and PDMN. Cochran-Armitage tests for trend assessed prevalence and agreement across the 3 periods defined by the 2 sequential QI interventions. Logistic regression models were estimated to describe the associations of DIMN and PDMN and patient and disease characteristics. Univariate and multivariable models were estimated. Multivariable model building was accomplished by first performing backward elimination with all covariates significant in univariate models at P<0.05 significance level and then completing forward selection with all remaining eligible covariates in the presence of the covariates surviving backward elimination. Analyses were performed in SAS 9.4 (SAS Institute, Cary, North Carolina); P<0.05 was considered statistically significant.

RESULTS

Demographic and Disease Characteristics

A total of 27144 unique patients were retrieved from the tumor registry (Figure 1). Of these, 5143 patients had a single primary cancer diagnosis and at least one hospital admission and were included in the analyses. Demographic data are shown in Table 1. Median age was 63 years (range, 18-102). Approximately half (48%) of patients were female, and most (70%) were White. Upper GI, thoracic, GU, and lower GI cancer diagnostic groups comprised over half (51%) of the population. Nearly a third (28%) had known Stage IV disease. BMI was available in the majority (94%). Among those with known BMI, 67% (n=3092) were overweight or obese.

TABLE 1.

DEMOGRAPHIC AND DISEASE CHARACTERISTICS

Variable	N=5143
Age (years)
Median (range)	63 (18-102)
Gender, n (%)
Female	2486 (48)
Male	2657 (52)
Race, n (%)
White	3607 (70)
Black	1210 (23)
Other	240 (5)
Unknown*	86 (2)
Ethnicity, n (%)
Latinx	157 (3)
Educational attainment (% high school graduate or higher), median (range)	88 (64-100)
Median income (USD thousands), median (range)	55 (14-136)
Cancer diagnostic group, n (%)
Upper gastrointestinala	1106 (21)
Thoracicb	952 (18)
Genitourinaryc	913 (18)
Lower gastrointestinald	598 (12)
Breast	491 (9)
Gynecologice	403 (8)
Otherf	341 (7)
Head & neck	339 (7)
Disease stage, n (%)
I	394 (27)
II	1149 (22)
III	1158 (23)
IV	1442 (28)
BMI
<18.5	217 (4)
≥18.5, <25	1528 (32)
≥25, <30	1561 (32)
≥30	1531 (32)
Unknown	306 (6)

^*The patient chose not to give their race or ethnicity; missing data

^aEsophagus, gallbladder, liver and bile ducts, pancreas, stomach

^bLung and bronchus

^cProstate, penis, testis, bladder, kidney, renal pelvis, ureter

^dAnus, anal canal, colon, rectosigmoid junction, rectum, small intestine

^eCervix, ovaries, uterus, vagina, vulva

^fCancer unknown primary, sarcoma, central and peripheral nervous system, bones, peritoneum, thyroid, malignant melanoma bones, joints, soft tissue, heart, mediastinum, pleura, orbit, other endocrine glands

Note: Percentages may not add to 100% as values are rounded to the nearest whole number

Malnutrition Risk Screening

MST scores were available for 79% of patients (n=4085). Of those with a known MST score, 25% (n=1005) were at risk of malnutrition. The proportions at risk of malnutrition across cancer diagnostic groups were 39% upper GI, 31% thoracic, 27% head and neck, 26% breast, and 11% GU (Figure 2).

FIGURE 2. MALNUTRITION RISK BY MST ACROSS CANCER DIAGNOSTIC GROUPS (N=4085).

At risk is defined as a malnutrition screening tool score ≥ 2

GI, Gastrointestinal; GU, Genitourinary

Percentages may not add to 100% as values are rounded to the nearest whole number.

Malnutrition Assessment

Among the entire study population, 11% (557 of 5143) had malnutrition identified by at least one clinician (documented by an RDN or coded by a physician) (Figure 3). Malnutrition was identified by both the RDN and physician in 4% (223 of 5143), by only an RDN in 4% (197 of 5143), and by only a physician in 3% (137 of 5143). The overall prevalence of DIMN and PDMN was 8% (420 of 5143) and 7% (360 of 5143), respectively (Figure 3). The RDNs and physicians agreed on the presence/absence of MN (irrespective of whether MST was administered) in 94% (4809 of 5143) of cases. Table 2 shows the malnutrition prevalence by MST score.

FIGURE 3. MALNUTRITION DOCUMENTATION BY BOTH REGISTERED DIETITIAN NUTRITIONIST AND PHYSICIAN (N=5143).

*Among the entire study population, 4% of patients had malnutrition documented by both RDN and physician (overlap).

TABLE 2.

MALNUTRITION PREVELANCE BY MALNUTRITION SCREENING TOOL SCORE AND CLINICIAN (N=4085)

MST Score	N (%)	No identification of MN	DIMN Only (N, %)	PDMN Only (N, %)	DIMN and PDMN (N, %)
<2	3080 (75.4)	2952 (95.8)	29 (0.9)	57 (1.9)	42 (1.4)
≥2	1005 (24.6)	643 (64.0)	138 (13.7)	62 (6.2)	162 (16.1)

MST – malnutrition screening tool, DIMN – dietitian identified malnutrition, PDMN – physician diagnosed malnutrition

DIMN and PDMN varied from 3% to 13% and 2% to 14%, respectively, across cancer diagnostic groups (Figure 4). For both DIMN and PDMN, the prevalence of malnutrition was lowest in breast cancer and highest in upper GI and head/neck cancers. The prevalence of malnutrition in lower GI cancers ranged from 6% for PDMN to 9% for DIMN.

FIGURE 4. PREVALENCE OF MALNUTRITION BY CLINICIAN ACROSS CANCER DIAGNOSTIC GROUPS (N=5143).

DIMN, Dietitian-identified malnutrition; GI, Gastrointestinal; GU, Genitourinary; PDMN, Physician-diagnosed malnutrition

Malnutrition Severity

Malnutrition severity was most often graded severe (RDN 73% [n=332]; physician 69% [n=247]) and least often mild (RDN 2% [n=7]; physician 10% [n=37]). The physician and RDN agreed on the severity in 79% of cases (177 of 223).

MST score was available for 371 of 420 (88%) RDN-identified patients. In those with DIMN and an available MST, 300 (81%) were at malnutrition risk. Malnutrition severity was similar between those at malnutrition risk (2% mild, 20% moderate, and 79% severe) and those not at malnutrition risk (3% mild, 18% moderate, and 79% severe) (P=0.78). In each of the DIMN severity groups, over 70% had MST ≥2 (mild 71%, moderate 82%, and severe 81%). In every diagnostic group, over 70% of DIMN were graded severe and <4% mild. The proportions of severe DIMN were highest in other cancers (83%) and gynecologic (81%) cancers and lowest in breast cancer (71%).

MST score was available for 323 of 360 (90%) physician-diagnosed patients. In those with PDMN and an available MST, 224 (69%) were at malnutrition risk. Differences in the proportions of PDMN grades were observed between those at MN risk (9% mild, 17% moderate, 74% severe) and those not at malnutrition risk (11% mild, 31% moderate, and 58% severe) (P=0.007). Numerically, higher proportions of mild MN were observed in PDMN versus DIMN groups (22% of PDMN in GU and 18% in lower GI cancers). In those with a low BMI, 54% had DIMN (moderate or severe) and 60% PDMN (mild, moderate, or severe).

Multivariable Analysis of Associations with High Malnutrition Prevalence

In univariate models, age, gender, race, median income, advanced disease, diagnostic group, and BMI were associated with odds of both DIMN and PDMN. In multivariable models, age (P=0.017), gender (P=0.032), race (P=0.004), advanced disease (P<0.001), diagnostic group (P<0.001), and BMI (P<0.001) were independent prognosticators of DIMN; male gender and Black race were associated with greater odds of DIMN (Figure 5A and 5B). Neither gender nor race was associated with PDMN in multivariable models; age (P=0.022), stage (P<0.001), diagnostic group (P<0.001), and BMI (P<0.001) were also independently prognostic of PDMN. The prevalence of malnutrition, as documented by any clinician, was associated with the following variables in a multivariable model: age (P=0.01), median income (P=0.02), stage four disease (P<0.001), diagnostic group (P<0.001), and BMI (P<0.001). Education level and median income in residence zip code were not associated with malnutrition prevalence in any multivariable models.

FIGURE 5. ODDS OF (A) DIETITIAN IDENTIFIED MALNUTRITION (DIMN) PREVALENCE AND (B) PHYSICIAN DIAGNOSED MALNUTRITION (PDMN) PREVALENCE BY MULTIVARIATE LOGISTIC REGRESSION (N=5143).

BMI, body mass index; CI, Confidence Interval; DIMN, Dietician-identified malnutrition; GI, Gastrointestinal; GU, Genitourinary; OR, odds ratio; PDMN, physician-diagnosed malnutrition

Quality Improvement Projects and Longitudinal Changes in Malnutrition Prevalence and Severity

Significant trends in prevalence of DIMN were observed as DIMN increased from 3.1%, 8.1%, to 10.3% (P<.001), and PDMN from 0.5%, 7.8%, to 8.2% (P<.001) over 3 years. While rates of mild, moderate, and severe malnutrition varied across time periods, statistically significant changes in these distributions were not identified in DIMN (P=0.62) or PDMN (P=0.20) after the second QIP.

DISCUSSION

This study provides valuable information on malnutrition among more than 5000 consecutive hospitalized patients with cancer in a representative clinical practice within the US. A total of 79% of patients were screened for malnutrition during their first hospital admission post-diagnosis. Notably, 25% of those were at risk for malnutrition. Overall, malnutrition was present in 11% patients. When identified, malnutrition was usually graded as severe. While nutrition support is recognized as critical for oncology care, our data show alarmingly low rates of malnutrition diagnosis.

Our data support the previous findings of a multi-institutional admissions database,³⁷ with a median malnutrition prevalence of 4% (range 0.6–18%). Unlike our study, severe MN was observed in only 25% of patients. According to the Healthcare Cost and Utilization Project, a malnutrition diagnosis was coded in 3.2% of all US hospital discharges in 2010, 4.5% in 2013, and 9% in 2018,^{4, 8, 38} a modest increase over that decade. Notably, these prevalence estimates are related to coded diagnoses rather than actual numbers of malnourished patients. Therefore, it is difficult to determine whether the absence of a malnutrition code is from poor documentation and/or lack of a structured process for identification.

There was a large discrepancy in the malnutrition prevalence based on real-world health administrative data and studies that performed a structured nutrition status assessment. When the Subjective Global Assessment Tool was used, malnutrition was diagnosed in up to 78% of hospitalized cancer patients.³⁹ Similarly, other studies found that 20-60% were malnourished.^{17, 36, 38, 40, 41} Notably, most of this information was derived from convenience samples and potentially biased populations with inconsistent methodologies. The large variation between studies may be due to poor performance of a malnutrition risk screening tool or diagnostic criteria. The differences may also be related to variations in patient age, cancer type and stage, geographical region, and socioeconomic status.

In our study, malnutrition prevalence varied based on the location of the tumor. Malnutrition was more common in patients with upper GI and head/neck cancers than with other cancers. However, it is important to note that lower GI cancers only accounted for 12% of our cohort, which might affect our analysis of malnutrition prevalence by cancer site. Previous studies have been conducted to examine the overall prevalence of malnutrition across various cancer types in different countries and environments. A significant challenge in interpreting these studies is the diverse methodologies employed to define malnutrition. Despite these methodological variations, the majority of studies consistently indicate that gastroesophageal, pancreatic, and head and neck cancers are associated with the highest risk of malnutrition, while malnutrition is less common in individuals with breast and prostate cancer.⁴² Advanced cancer patients typically have a higher rate of malnutrition as compared to patients with early stage disease.²³ Consequently, approximately one-third of the patients with advanced cancer in our study likely contributed disproportionately to the overall malnutrition prevalence.

Screening for malnutrition is the first step in the nutritional care process. However, there is no consensus on the best nutritional screening method for hospitalized patients.⁴³ In 2020, the AND published a position statement endorsing the use of the MST to screen all adults in all settings, including hospitals.⁴⁴ In our cancer center, the MST has been adopted in both outpatient and inpatient settings because of its ease and simplicity.⁴³ Nutritional screening barriers for hospitalized cancer patients may include difficulty obtaining accurate weight history, time constraints for nursing staff to complete additional assessments, and patients’ inability to participate due to their critical condition. However, our study did not evaluate the characteristics of patients who did not complete the MST. Patients in more severe conditions may be at high risk for malnutrition but less likely to have the MST completed. Future studies are needed to identify patterns of MST completion accounting for disease severity.

In the present study, we found 25% of inpatients were at risk for malnutrition, as compared to 28% of outpatients from our previous report.⁴⁵ The prevalence of malnutrition risk in our study was similar to results of the US-based nutritionDay study, which also used the MST.⁴⁶ These findings collectively stress the need for consistent malnutrition screening in both general and cancer-specific hospital settings within the US.

Our study identified important patient and disease characteristics that were associated with greater malnutrition prevalence. Similar to a study of older cancer patients⁴⁷, we found that malnutrition prevalence was greater in males than females. Older age was correlated with high malnutrition prevalence. There was significant variation in malnutrition prevalence in upper GI and head/neck cancers, which are commonly associated with weight loss. Unique to this analysis, Black patients accounted for almost a quarter of those studied and had a higher malnutrition prevalence than White patients. Further study would be needed to determine whether clinician bias may account for this difference in malnutrition prevalence based on race and ethnicity as the Latinx population was underrepresented in our study. Interestingly, malnutrition prevalence was not associated with education level and median income. The greatest odds of malnutrition were in those with a low BMI (<18.5). Notably, 67% of our population were overweight (BMI ≥25). This suggests that a single nutrition status indicator, like BMI, may be misleading as it does not account for preexisting obesity with subsequent weight loss.

The AND/ASPEN consensus criteria provides a new standard for diagnosing malnutrition with data on associated clinical outcomes.³⁶ In our study, the institutional documentation of malnutrition is consistent with practice standards in nutrition support by nurses, RDNs, and physicians.^48–50 While nurses completed nutrition screening as part of an admission assessment, the RDN is the clinician primarily responsible for conducting the nutrition assessment. However, malnutrition screening is not universally performed, which is an important area for future quality improvement. In addition, the EMR provides clinical decision support. If a patient meets the criteria for malnutrition, the physician is prompted by the CDI team to document malnutrition in their notes. The CDI team plays a crucial role in ensuring that clinical documentation accurately reflects the patient’s condition, which is essential for appropriate treatment and billing purposes.

Malnutrition screening in the hospital is the first step in the overall workflow for successful coding, billing, and reimbursement. Our study identified two relevant QI interventions that can be replicated in similar settings. Small yet clinically meaningful improvements in malnutrition documentation and coding were observed over the 3-year study period. The increase in the proportion of individuals identified as malnourished may be due to the use of a well-established nutritional screening method, the MST, and the diagnostic criteria for malnutrition by the AND/ASPEN.³⁶ Following these interventions, dietitians and physicians were congruent for identifying malnutrition and assigning severity. Increased patient acuity, an aging population, or improved clinician awareness may have also contributed. Further QI initiatives are needed to improve outcomes for malnourished patients and reduce overall hospital costs due to malnutrition.

Limitations of our study include retrospective data collection; thus, a cause-and-effect relationship cannot be determined. We also are unable to determine which nutrition domains were most influential in malnutrition identification/grading. Additionally, prior cancer treatments and medications for other comorbidities were not recorded. The nutrition assessment tool was integrated into the EMR but used both structured and free-text documentation. Due to variations in clinical data documentation, likely, some malnourished individuals were not captured. Likewise, some patients were documented as malnourished in the absence of screening. Grading mild malnutrition may be more subjective, as it is not part of the AND/ASPEN consensus. We did not evaluate whether malnourished individuals had a nutrition care plan established prior to hospital discharge. Our study sample, although reflective of our southeastern geographic region, may not represent other parts of the US because of differences in racial distribution.

Our study shows that multiple patient characteristics are associated with malnutrition prevalence in a large cancer inpatient population. Importantly, this study population was representative of a hospitalized US cancer population by age, gender, race, and primary tumor site.⁵¹ Additionally, we show that multidisciplinary involvement in patient care and hospital billing can improve accurate documentation while communication between physicians and RDNs can optimize nutrition management. When malnutrition is identified and coded, resource requirements for provision of care for malnourished patients can be more appropriately allocated.³⁶

CONCLUSIONS

Our study highlights several critical findings regarding malnutrition in hospitalized cancer patients with solid tumors. One-fifth of patients were not screened for malnutrition on their first admission post-diagnosis. Of those screened, 25% were at risk for malnutrition. Malnutrition was identified in 11% of patients. Malnutrition appears to be a significant yet under-documented and underdiagnosed issue in hospital settings, representing unmet medical need. Our data have important clinical, operational, and financial implications in cancer care. There is an urgent need for consistent and systematic malnutrition screening for all cancer patients within 24 hours of hospital admission. Ineffective screening and diagnosis of malnutrition are financial risks for healthcare systems. These findings should inform the development of national public health policies to address malnutrition in cancer care and stimulate advocacy efforts in the US. Future considerations for improvement should include evaluating the efficacy of malnutrition screening, determining the necessary RDN workforce, and performing a cost-benefit analysis.

TABLE 3.

PREVELANCE AND SEVERITY OF MALNUTRITION ACROSS CANCER DIAGNOSTIC GROUPS

	N=4085		N=5143
			PDMN Prevalence	PDMN Severity			DIMN Prevalence	DIMN Severity
	MST <2	MST ≥2		Mild	Moderate	Severe		Mild	Moderate	Severe
N (%)	3080 (79)	1005 (21)	360/5143 (7)	37/360 (10)	76/360 (21)	247/360 (69)	420/5143 (8)	7/420 (2)	81/420 (19)	332/420 (79)
Diagnostic Group, N (%)
Breast	356 (87)	53 (13)	12/491 (2)	0/12 (0)	4/12 (33)	8/12 (67)	14/491 (3)	0/14 (0)	4/14 (29)	10/14 (71)
GU	676 (89)	87 (11)	23/913 (3)	5/23 (22)	9/23 (39)	9/23 (39)	26/913 (3)	0/26 (0)	6/26 (23)	20/26 (77)
Gynecologic	213 (74)	76 (26)	16/403 (4)	1/16 (6)	5/16 (31)	10/16 (63)	21/403 (5)	0/21 (0)	4/21 (19)	17/21 (81)
Head/Neck	206 (73)	78 (27)	46/339 (14)	4/46 (9)	14/46 (30)	28/46 (61)	44/339 (13)	1/44 (2)	8/44 (18)	35/44 (80)
Lower GI	396 (79)	103 (21)	38/598 (6)	7/38 (18)	3/38 (8)	28/38 (74)	55/598 (9)	2/55 (4)	10/55 (18)	43/55 (78)
Thoracic	513 (69)	235 (31)	76/952 (8)	9/76 (12)	13/76 (17)	54/76 (71)	90/952 (9)	1/90 (1)	18/90 (20)	71/90 (79)
Upper GI	523 (61)	332 (39)	138/1106 (12)	11/138 (8)	25/138 (18)	102/138 (74)	158/1106 (14)	3/158 (2)	29/158 (18)	126/158 (80)
Other	197 (83)	41 (17)	11/341 (3)	0/11 (0)	3/11 (27)	8/11 (73)	12/341 (4)	0/12 (0)	2/12 (17)	10/13 (83)
BMI, N (%) *
<18.5	67 (39)	106 (61)	80/217 (37)	5/80 (6)	15/80 (19)	60/80 (75)	79/217 (36)	1/79 (1)	7/79 (9)	71/79 (90)
≥18.5, <25	805 (67)	401 (33)	158/1528 (10)	17/158 (11)	26/158 (16)	115/158 (73)	137/1528 (13)	4/197 (2)	41/197 (21)	152/197 (77)
≥25, <30	1039 (82)	223 (18)	46/1561 (3)	7/46 (15)	15/46 (33)	24/46 (52)	55/1561 (4)	1/55 (2)	16/55 (29)	38/55 (69)
≥30	1048 (85)	187 (15)	42/1531 (3)	7/42 (17)	12/42 (29)	23/42 (55)	47/1534 (3)	1/47 (2)	11/47 (23)	35/47 (74)

^*BMI unknown for 1267