ABSTRACT
Rationale
Malnutrition remains systematically undercoded in hospital administrative records, despite its high prevalence among patients with non‐communicable diseases (NCDs). China's Diagnosis‐Related Group (DRG) payment reform links reimbursement to secondary diagnosis coding, potentially incentivising changes in clinical documentation.
Aims and Objectives
This study evaluates the validity of administrative malnutrition coding and its association with clinical outcomes in the context of DRG reform.
Method
A retrospective cross‐sectional study was conducted using administrative data from a tertiary hospital in southwest China (2023–2025), including 31,133 inpatient admissions with a primary NCD diagnosis. Patients were categorised by insurance type: Urban and Rural Residents' Basic Medical Insurance (URRBMI) or Urban Employees' Basic Medical Insurance (UEBMI). Multivariable logistic regression was used to identify predictors of malnutrition coding and estimate associations with in‐hospital mortality and prolonged hospitalisation (LOS > 10 days), adjusting for age, insurance, admission route, and primary diagnosis.
Results
The prevalence of coded malnutrition was 8.5%, surging from 5.3% in 2023% to 11.0% in 2024 before stabilizing at 9.2% in 2025. After case‐mix adjustment, the year of admission remained a significant predictor (2024: OR = 2.23; 2025: OR = 1.71). Coded malnutrition independently predicted in‐hospital mortality (OR = 4.35) and prolonged hospitalisation (OR = 1.32). URRBMI patients consistently demonstrated higher coding rates, while the association between malnutrition coding and mortality was stronger in UEBMI patients.
Conclusion
Malnutrition coding increased markedly following DRG reform. The robust association with adverse clinical outcomes validates these codes as meaningful markers of nutritional risk. Our findings underscore the need for systematic nutritional screening in hospital admission protocols to ensure equitable and clinically accurate reimbursement.
Keywords: China, DRG reform, ICD coding, in‐hospital mortality, insurance disparity, malnutrition, non‐communicable diseases
1. Introduction
Malnutrition among hospitalised patients represents a significant and persistent global health challenge [1]. Clinical screening studies report that 20%–50% of acute care inpatients are malnourished or at nutritional risk on admission [2, 3, 4]. In Chinese tertiary hospitals, approximately 30%–41% of hospitalised patients have been identified as being at nutritional risk using the Nutritional Risk Screening 2002 (NRS‐2002) [5, 6]. Hospitalised malnutrition is associated with a range of adverse outcomes, including prolonged length of stay, increased susceptibility to complications, higher readmission rates, and elevated in‐hospital mortality [7, 8, 9]. The burden is particularly pronounced among patients with non‐communicable diseases (NCDs), who account for a substantial proportion of hospital admissions and healthcare utilisation globally [10]. Despite this, malnutrition remains systematically underdiagnosed and underdocumented in routine clinical practice, resulting in substantial underrepresentation in hospital administrative data and potentially limiting nutrition‐related resource allocation [2, 11].
China has undergone a fundamental transformation in hospital payment policy over the past decade, with the national rollout of Diagnosis‐Related Group (DRG) and Disease‐Diagnosis‐Related Group (DIP) payment systems accelerating significantly from 2021 onwards [12]. Under DRG‐based payment systems, hospital reimbursement is linked to case‐mix classification, in which secondary diagnoses and comorbidities contribute to DRG weighting [13]. As a clinically relevant comorbidity, malnutrition coding may therefore influence hospital reimbursement [14, 15]. This creates incentives for improved documentation of nutritional problems but may also increase the risk of reimbursement‐driven coding behaviour [16]. Evidence from early DRG adopters suggests that diagnosis‐based payment systems can increase comorbidity coding within administrative datasets [17, 18]. Whether such changes reflect genuine improvements in clinical recognition or coding inflation remains an important question for health system oversight and payment policy.
Despite the scale of China's DRG reform and its potential implications for nutritional care, little evidence is currently available regarding how malnutrition coding patterns have changed in Chinese hospitals following payment reform implementation [19, 20, 21]. Existing research on hospital malnutrition in China has relied predominantly on clinical nutritional screening studies using instruments such as the NRS‐2002 or MNA [22, 23]. While these studies provide important prevalence estimates, they offer limited insight into administrative coding behaviour, documentation practices, and the potential influence of DRG‐related reimbursement incentives. In addition, it remains unclear whether administratively coded malnutrition reflects clinically meaningful nutritional risk or primarily changes in coding behaviour under prospective payment systems. The potential differential impact of DRG reform across insurance schemes, particularly between Urban and Rural Residents' Basic Medical Insurance (URRBMI) and Urban Employees' Basic Medical Insurance (UEBMI), has also received limited attention [24, 25].
Therefore, this study aimed to: (1) describe temporal trends in malnutrition coding among hospitalised NCD patients following DRG reform implementation; (2) identify predictors of malnutrition coding, with particular attention to insurance‐related disparities; and (3) evaluate whether coded malnutrition was associated with adverse clinical outcomes, thereby assessing the clinical meaningfulness of administrative coding within a DRG‐based reimbursement context.
2. Methods
2.1. Study Design and Data Source
This retrospective cross‐sectional study utilised administrative discharge data from a tertiary hospital in southwest China, covering inpatient admissions from 1 January 2023 to 31 December 2025. The dataset included demographic characteristics, insurance type, admission and discharge information, up to 16 ICD‐10 diagnostic codes, discharge disposition, length of stay, and hospitalisation costs.
2.2. Study Population
Patients were eligible for inclusion if they were aged 18 years or older and had a primary diagnosis corresponding to one of the four major NCD categories: cardiovascular diseases (ICD‐10 I00–I99), cancers (C00–C97), chronic respiratory diseases (J40–J47 and J96), and diabetes (E10–E14). Only patients insured under URRBMI or UEBMI were included. Patients with other or unspecified insurance types, those aged under 18 years, and those with zero or missing total hospitalisation costs were excluded. The initial dataset comprised 31,470 admissions. After applying these criteria, the final analytical sample comprised 31,133 admissions (99.1% of the starting dataset).
2.3. Variables
The primary exposure variable was coded malnutrition, defined as the presence of at least one of the following ICD‐10 codes in any of the 16 diagnostic fields: E41 (nutritional marasmus), E43 (unspecified severe protein‐energy malnutrition), or E46 (unspecified protein‐energy malnutrition). E44 (protein‐energy malnutrition of moderate and mild degree) was not included in the primary definition, consistent with a conservative focus on codes indicative of clinically significant malnutrition.
Two outcome variables were examined. In‐hospital mortality was defined as a discharge disposition recorded as ‘death.’ Prolonged hospitalisation was defined as a length of stay (LOS) exceeding 10 days, corresponding to the 75th percentile of the LOS distribution in the study sample (P75 = 10 days), consistent with thresholds used in comparable studies.
Covariates included: age (standardised); sex (male vs female); insurance type (URRBMI vs UEBMI); admission route (emergency vs non‐emergency); place of residence (rural, defined as a registered residential address containing the Chinese character for village; urban otherwise); primary diagnosis category (cardiovascular disease as reference; cancer, chronic respiratory disease, and diabetes as indicator variables); and calendar year (2023 as reference; 2024 and 2025 as indicator variables). Total hospitalisation costs and LOS were log‐transformed prior to inclusion as continuous covariates.
2.4. Statistical Analysis
Descriptive statistics were reported as median (interquartile range) for continuous variables and as frequency (percentage) for categorical variables. Between‐group differences were assessed using the Mann–Whitney U test for continuous variables and Pearson's chi‐squared test for categorical variables.
To identify predictors of malnutrition coding, multivariable logistic regression was performed with coded malnutrition as the dependent variable. All covariates were entered simultaneously without stepwise selection. Model fit was assessed using the Nagelkerke R 2 and Hosmer–Lemeshow test, and multicollinearity was evaluated using variance inflation factors (VIF; threshold VIF < 5).
To examine associations between coded malnutrition and clinical outcomes, two separate multivariable models were fitted. In‐hospital mortality was analysed using Firth's penalised logistic regression because of complete separation arising from sparse deaths in the diabetes subgroup [26]; this model additionally adjusted for log‐transformed LOS, as patients who die in hospital tend to have shorter admissions. Prolonged hospitalisation (LOS > 10 days) was analysed using standard logistic regression, without LOS as a covariate. Both models adjusted for age, insurance type, admission route, primary diagnosis category, and calendar year.
Three sensitivity analyses were conducted (full results are provided in Table S1): (1) broader malnutrition coding definitions including E77.8 (hypoproteinaemia) and R63.8 (nutritional risk); (2) exclusion of admissions with LOS > 30 days; and (3) insurance‐stratified mortality models. Analyses were performed using R version 4.5.3, with two‐sided α = 0.05. Results are reported as odds ratios (OR) with 95% confidence intervals (CI).
3. Results
3.1. Sample Characteristics
A total of 31,133 hospitalised patients with NCDs were included (2023: n = 10,736; 2024: n = 10,292; 2025: n = 10,105). Median age was 69.0 years (IQR: 59.0–76.0), 53% were male, 84.4% were insured under URRBMI, and 60.3% resided in rural areas. Cardiovascular disease was the most common primary diagnosis (62.7%), followed by chronic respiratory disease (20.8%), diabetes (8.9%), and cancer (7.6%).
3.2. Baseline Characteristics by Malnutrition Status
Compared with patients without coded malnutrition, those with malnutrition were significantly older (median 75.0 vs 69.0 years), more frequently admitted via emergency (44% vs 30%), more likely to reside in rural areas (67% vs 60%), and more often insured under URRBMI (91% vs 84%; all p < 0.001). In‐hospital mortality was markedly higher in the malnutrition group (1.8% vs 0.4%), as was prolonged hospitalisation (29% vs 22%; both p < 0.001). Sex distribution did not differ significantly between groups (Table 1).
Table 1.
Baseline characteristics of hospitalised NCD patients by coded malnutrition status.
| Variable | Overall N = 31,133 1 | Coded malnutrition | p‐value 2 | |
|---|---|---|---|---|
| No N = 28,495 1 | Yes N = 2638 1 | |||
| Age (years) | 69.0 (59.0, 76.0) | 69.0 (59.0, 75.0) | 75.0 (69.0, 82.0) | < 0.001 |
| Sex | 0.145 | |||
| Female | 14,549 (47%) | 13,280 (47%) | 1269 (48%) | |
| Male | 16,584 (53%) | 15,215 (53%) | 1369 (52%) | |
| Insurance type | < 0.001 | |||
| UEBMI | 4846 (16%) | 4,597 (16%) | 249 (9.4%) | |
| URRBMI | 26,287 (84%) | 23,898 (84%) | 2389 (91%) | |
| Emergency admission | 9751 (31%) | 8586 (30%) | 1165 (44%) | < 0.001 |
| Residence | < 0.001 | |||
| Urban | 12,354 (40%) | 11,480 (40%) | 874 (33%) | |
| Rural | 18,779 (60%) | 17,015 (60%) | 1764 (67%) | |
| Length of stay (days) | 8.0 (6.0, 10.0) | 7.0 (6.0, 10.0) | 8.0 (7.0, 11.0) | < 0.001 |
| In‐hospital mortality | 169 (0.5%) | 121 (0.4%) | 48 (1.8%) | < 0.001 |
| Prolonged LOS (> 10 days) | 6949 (22%) | 6192 (22%) | 757 (29%) | < 0.001 |
| Primary diagnosis | < 0.001 | |||
| Cardiovascular | 19,527 (63%) | 17,481 (61%) | 2046 (78%) | |
| Cancer | 2355 (7.6%) | 2185 (7.7%) | 170 (6.4%) | |
| Respiratory | 6470 (21%) | 6087 (21%) | 383 (15%) | |
| Diabetes | 2781 (8.9%) | 2742 (9.6%) | 39 (1.5%) | |
| Year | < 0.001 | |||
| 2023 | 10,736 (34%) | 10,165 (36%) | 571 (22%) | |
| 2024 | 10,292 (33%) | 9157 (32%) | 1135 (43%) | |
| 2025 | 10,105 (32%) | 9173 (32%) | 932 (35%) | |
Abbreviations: UEBMI, Urban Employees' Basic Medical Insurance; URRBMI, Urban and Rural Residents' Basic Medical Insurance.
Median (Q1, Q3); n (%).
Wilcoxon rank sum test; Pearson's Chi‐squared test.
3.3. Temporal Trends in Malnutrition Coding
The overall prevalence of coded malnutrition was 8.5% (n = 2638), increasing from 5.3% in 2023% to 11.0% in 2024 before declining to 9.2% in 2025 (Figure 1). The increase was most pronounced among patients with cardiovascular disease, where rates rose from 6.0% to 14.7% between 2023 and 2024 before declining to 11.0% in 2025. Among patients aged ≥ 75 years, coding rates approached 20% in 2024 (9.8% in 2023, 19.6% in 2024, 16.8% in 2025). URRBMI patients consistently demonstrated higher coding rates than UEBMI patients across all 3 years (2023: 5.7% vs 3.3%; 2024: 12.0% vs 5.5%; 2025: 9.7% vs 6.8%). By contrast, coding prevalence remained comparatively low among patients with diabetes (range: 0.4%–2.3% across the three study years). Coding rates by diagnosis category are shown in Figure 1.
Figure 1.
Annual prevalence of coded malnutrition among hospitalised patients with non‐communicable diseases (NCDs), 2023–2025. The figure illustrates the temporal trend of malnutrition coding prevalence overall and stratified by primary diagnosis (cardiovascular disease, cancer, and chronic respiratory disease). The shaded area represents the 95% confidence interval for the overall prevalence trend. Coding rates are expressed as percentages of the total NCD patient population within each diagnosis category for each calendar year.
3.4. Predictors of Malnutrition Coding
In multivariable analysis, malnutrition coding remained significantly more likely in both 2024 (OR = 2.23, 95% CI: 2.00–2.48) and 2025 (OR = 1.71, 95% CI: 1.53–1.91) relative to 2023 after adjustment for patient case‐mix (Table 2). URRBMI insurance (OR = 1.70, 95% CI: 1.47–1.97), emergency admission, older age, rural residence, and male sex were each independently associated with higher coding odds. Compared with cardiovascular disease, chronic respiratory disease and diabetes were associated with substantially lower odds of malnutrition coding. Model fit was acceptable (Nagelkerke R 2 = 0.132; Hosmer–Lemeshow p = 0.096), and no evidence of multicollinearity was detected (all VIF < 2).
Table 2.
Multivariable logistic regression: predictors of coded malnutrition (n = 31,133).
| Variable | OR 1 | 95% CI | p‐value |
|---|---|---|---|
| Age (standardised) | 2.11 | 2.00, 2.22 | < 0.001 |
| Sex (Male vs Female) | 1.12 | 1.03, 1.22 | 0.008 |
| Insurance (URRBMI vs UEBMI) | 1.70 | 1.47, 1.97 | < 0.001 |
| Emergency admission | 1.46 | 1.34, 1.59 | < 0.001 |
| Residence (Rural vs Urban) | 1.22 | 1.11, 1.34 | < 0.001 |
| Year 2024 (vs 2023) | 2.23 | 2.00, 2.48 | < 0.001 |
| Year 2025 (vs 2023) | 1.71 | 1.53, 1.91 | < 0.001 |
| Cancer (vs Cardiovascular) | 0.83 | 0.70, 0.98 | 0.024 |
| Respiratory (vs Cardiovascular) | 0.47 | 0.42, 0.52 | < 0.001 |
| Diabetes (vs Cardiovascular) | 0.21 | 0.15, 0.28 | < 0.001 |
Abbreviations: CI, Confidence Interval; OR, Odds Ratio.
OR = Odds Ratio.
3.5. Association With Clinical Outcomes
Coded malnutrition was independently associated with in‐hospital mortality (OR = 4.35, 95% CI: 3.00–6.23) and prolonged hospitalisation exceeding 10 days (OR = 1.32, 95% CI: 1.20–1.45) after multivariable adjustment (Table 3). In the prolonged hospitalisation model, the odds of LOS > 10 days were lower in 2024 (OR = 0.84) and 2025 (OR = 0.66) relative to 2023, suggesting a concurrent LOS compression effect following DRG reform implementation. Model fit was acceptable for the mortality model (likelihood ratio test: χ 2 = 385.14, df = 11, p < 0.001) and the prolonged hospitalisation model (Nagelkerke R 2 = 0.122); the Hosmer–Lemeshow test for the latter was statistically significant (χ 2 = 21.14, df = 8, p = 0.007), likely reflecting the sensitivity of this test to large sample sizes rather than substantive lack of fit. Findings remained materially unchanged across all three sensitivity analyses (Table S1), supporting the robustness of the primary results.
Table 3.
Multivariable regression: Association between coded malnutrition and clinical outcomes (n = 31,133).
| Variable | Model A: In‐hospital mortality | Model B: Prolonged LOS | ||||
|---|---|---|---|---|---|---|
| OR | 95% CI | p‐value | OR | 95% CI | p‐value | |
| Coded malnutrition | 4.35 | 3.00–6.23 | < 0.001 | 1.31 | 1.19–1.44 | < 0.001 |
| Age (standardised) | 1.16 | 0.99–1.36 | 0.069 | 1.09 | 1.06–1.12 | < 0.001 |
| Sex (Male vs Female) | 1.64 | 1.18–2.31 | 0.003 | 1.18 | 1.11–1.24 | < 0.001 |
| Insurance (URRBMI vs UEBMI) | 0.26 | 0.19–0.37 | < 0.001 | 0.81 | 0.74–0.88 | < 0.001 |
| Emergency admission | 2.95 | 2.13–4.11 | < 0.001 | 1.71 | 1.61–1.81 | < 0.001 |
| Rural residence | — | — | — | 0.98 | 0.92–1.04 | 0.454 |
| Cancer (vs Cardiovascular) | 12.23 | 8.41–17.70 | < 0.001 | 6.08 | 5.54–6.66 | < 0.001 |
| Chronic respiratory (vs Cardiovascular) | 1.23 | 0.79–1.87 | 0.357 | 0.85 | 0.79–0.91 | < 0.001 |
| Diabetes (vs Cardiovascular) | 0.06 | 0.00–0.39 | < 0.001 | 0.27 | 0.23–0.32 | < 0.001 |
| Year 2024 (vs 2023) | 1.00 | 0.66–1.51 | 0.993 | 0.84 | 0.78–0.90 | < 0.001 |
| Year 2025 (vs 2023) | 1.32 | 0.91–1.94 | 0.146 | 0.66 | 0.61–0.71 | < 0.001 |
| Log (length of stay) | 0.28 | 0.22–0.36 | < 0.001 | — | — | — |
Note: Model A (in‐hospital mortality) was estimated using Firth's penalised logistic regression and additionally adjusted for log‐transformed length of stay; rural residence was not included as a covariate in this model. Model B (prolonged hospitalisation, LOS > 10 days) was estimated using standard logistic regression and additionally adjusted for rural residence; length of stay was not included as a covariate in this model, as it forms the basis of the outcome definition. Model fit: Model A, likelihood ratio test χ 2 = 385.14, df = 11, p < 0.001. Model B, Nagelkerke R 2 = 0.122; Hosmer–Lemeshow test χ 2 = 21.14, df = 8, p = 0.007 (likely reflecting the sensitivity of this test to large sample sizes rather than substantive lack of fit).
Abbreviations: CI, confidence interval; LOS, length of stay; OR, odds ratio; UEBMI, Urban Employees' Basic Medical Insurance; URRBMI, Urban and Rural Residents' Basic Medical Insurance.
4. Discussion
This study found an overall coded malnutrition prevalence of 8.5% among hospitalised NCD patients over a 3‐year period, which is substantially lower than rates reported using clinical screening tools in comparable populations [22, 23]. This discrepancy reflects the well‐documented systematic underestimation inherent to administrative coding, whereby malnutrition is recorded only when clinically identified and actively documented by the attending physician, rather than through structured screening [27]. The modest Nagelkerke R 2 of 0.132 in our predictor model is consistent with this interpretation: coding decisions are influenced by a range of unmeasured institutional, departmental, and individual documentation factors that administrative data cannot fully capture. These patterns have important implications for reimbursement accuracy, resource allocation, and the quality of administrative documentation under DRG‐based payment systems [11].
The most notable finding of this study was the sharp increase in malnutrition coding rates from 5.3% in 2023 to 11.0% in 2024, coinciding with the full‐scale implementation of DRG payment reform and its consolidation during this period. After adjustment for patient case‐mix, year remained a significant independent predictor of malnutrition coding (Year 2024: OR = 2.23, 95% CI: 2.00–2.48; Year 2025: OR = 1.71, 95% CI: 1.53–1.91), indicating that the elevated coding rates cannot be attributed solely to changes in patient demographics or disease severity. Two interpretations merit consideration. First, DRG reform may have improved the hospital identification and documentation practices of malnutrition by incentivising hospitals to capture comorbidities that influence DRG weight and reimbursement. Similar increases in administrative comorbidity coding depth following DRG implementation have been reported in early DRG‐adopting health systems [18]. From this perspective, the rise in coding rates may partly reflect improved clinical recognition and documentation of nutritional risk. Second, the financial incentives inherent to DRG systems may also encourage strategic upcoding, whereby diagnoses are preferentially recorded to optimise case‐mix reimbursement without a commensurate improvement in clinical care [28]. The partial attenuation of the year effect in 2025 (OR = 1.71 vs 2.23 in 2024) may reflect stabilisation following an initial period of coding adjustment, though it does not resolve the question of mechanism. It should also be noted that 2023 represents the first year of DRG implementation at this hospital rather than a pre‐reform reference; the observed coding increase may therefore underestimate the full extent of reform‐related change. The clinical outcome analyses below provide evidence relevant to this distinction.
Coded malnutrition was independently associated with a more than fourfold increase in the odds of in‐hospital mortality (OR = 4.35, 95% CI: 3.00–6.23) after adjustment for age, insurance type, admission route, primary diagnosis, calendar year, and LOS. This magnitude of association is consistent with international evidence linking administratively coded malnutrition to adverse clinical outcomes, and suggests that the observed coding patterns are clinically meaningful rather than purely administrative phenomena [29, 30, 31]. If the observed increase in coding rates were driven primarily by strategic upcoding among clinically lower‐risk patients, one might expect weaker associations between coded malnutrition and adverse clinical outcomes. Importantly, the association remained consistent across all three sensitivity analyses (Table S1), including a broader malnutrition definition, exclusion of long‐stay admissions, and insurance‐stratified models. Together, these findings suggest that administratively coded malnutrition captures clinically meaningful patient severity rather than representing coding inflation alone. Coded malnutrition was also independently associated with prolonged hospitalisation (OR = 1.32, 95% CI: 1.20–1.45), further reinforcing its clinical significance. Overall, these results suggest that, although strategic coding incentives cannot be excluded, the malnutrition codes captured in this administrative dataset identify a patient population with substantially greater clinical complexity and adverse outcome risk.
URRBMI patients demonstrated consistently higher malnutrition coding rates than UEBMI patients across all 3 years, and were significantly more likely to receive a malnutrition code after full adjustment (OR = 1.70, 95% CI: 1.47–1.97). This finding extends previous evidence of insurance‐related health disparities in China, whereby URRBMI patients experience higher healthcare vulnerability and greater financial burden compared with UEBMI counterparts [32, 33]. The convergence of higher malnutrition coding rates, lower reimbursement levels, and greater financial vulnerability among URRBMI patients may represent a compounding structural disadvantage that warrants targeted attention within DRG‐related reimbursement and hospital management policies. Under DRG‐based payment systems, accurate malnutrition coding may influence reimbursement because nutritional comorbidities can contribute to DRG severity weighting and case‐mix classification [34, 35]. These findings support the potential value of systematically integrating nutritional screening and standardised documentation into hospital admission workflows, particularly for URRBMI patients, older adults, and patients with cardiovascular disease. Insurance‐stratified analyses showed a counterintuitive finding: despite lower coding rates, UEBMI patients had a stronger association between coded malnutrition and in‐hospital mortality (OR = 7.98, 95% CI: 4.31–14.62) compared with URRBMI patients (OR = 3.05, 95% CI: 1.86–4.83). One possible explanation is differential case severity selection, as malnutrition may only be coded in UEBMI patients with the most critical presentations (n = 249 coded). However, this interpretation should be treated cautiously, given the relatively small subgroup size. Overall, these findings suggest that more systematic nutritional screening and documentation practices may improve the visibility and consistency of nutritional documentation within administrative datasets while supporting more equitable case‐mix‐adjusted reimbursement [36].
4.1. Limitations
Several limitations should be acknowledged. First, this was a single‐centre study conducted at a tertiary hospital in southwest China, which may limit generalisability to other settings. Second, the absence of clinical nutritional assessment data (e.g., NRS‐2002 scores or anthropometric measures) prevented direct validation of administrative malnutrition coding against clinical standards. Third, the retrospective observational design precludes causal inference regarding the relationship between DRG reform and coding behaviour. Fourth, coding quality is inherently dependent on individual physician documentation practices, and systematic variation in recording behaviour across departments or years cannot be fully excluded. Finally, because administrative data were only available from the year of DRG implementation onwards, no pre‐reform baseline was available for interrupted time series analysis, and the observed year effects should therefore be interpreted cautiously.
5. Conclusions
This study demonstrated that malnutrition coding increased substantially following DRG reform implementation and remained strongly associated with adverse clinical outcomes. These findings suggest that administrative malnutrition coding captures clinically meaningful patient complexity while highlighting the importance of standardised nutritional documentation and strengthened coding governance under DRG‐based payment systems.
Ethics Statement
This study was approved by the Institutional Review Board of The Third People's Hospital of Yibin (Approval No: 2025042) and the Health Research Ethics Committee of the University of Waikato (Approval No: HREC(Health)2025#83).
Consent
As the study utilised de‐identified retrospective administrative data, the requirement for individual informed consent was waived.
Conflicts of Interest
The authors declare no conflicts of interest.
Supporting information
Table S1: Sensitivity analyses: predictors of malnutrition coding and association with clinical outcomes.
Acknowledgements
The authors thank the staff at the participating tertiary hospital in southwest China for providing the administrative data necessary for this study. This research is supported by the China Scholarship Council (Grant No. 202500560013). The funders had no role in the study design, data collection, data analysis, data interpretation, or manuscript preparation. Open access publishing facilitated by The University of Waikato, as part of the Wiley ‐ The University of Waikato agreement via the Council of Australasian University Librarians.
Data Availability Statement
The data that support the findings of this study are held by the hospital's administrative department. Restrictions apply to the availability of these data, which were used under license for the current study and are not publicly available due to patient confidentiality and institutional data governance policies. However, de‐identified data may be made available from the corresponding author upon reasonable request.
References
- 1. Inciong J. F. B., Chaudhary A., Hsu H. S., et al., “Hospital Malnutrition in Northeast and Southeast Asia: A Systematic Literature Review,” Clinical Nutrition ESPEN 39, (October 2020): 30–45, 10.1016/j.clnesp.2020.06.001. [DOI] [PubMed] [Google Scholar]
- 2. Barker L. A., Gout B. S., and Crowe T. C., “Hospital Malnutrition: Prevalence, Identification and Impact on Patients and the Healthcare System,” International Journal of Environmental Research and Public Health 8, no. 2 (February 2011): 514–527, 10.3390/ijerph8020514. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3. Jensen G. L., Compher C., Sullivan D. H., and Mullin G. E., “Recognizing Malnutrition in Adults: Definitions and Characteristics, Screening, Assessment, and Team Approach,” Journal of Parenteral and Enteral Nutrition 37, no. 6 (November 2013): 802–807, 10.1177/0148607113492338. [DOI] [PubMed] [Google Scholar]
- 4. Reber E., Gomes F., Vasiloglou M. F., Schuetz P., and Stanga Z., “Nutritional Risk Screening and Assessment,” Journal of Clinical Medicine 8, no. 7 (July 2019): 1065, 10.3390/jcm8071065. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5. Miao J. P., Quan X. Q., Zhang C. T., et al., “Comparison of Two Malnutrition Risk Screening Tools With Nutritional Biochemical Parameters, BMI and Length of Stay in Chinese Geriatric Inpatients: A Multicenter, Cross‐Sectional Study,” BMJ Open 9, no. 2 (February 2019): e022993, 10.1136/bmjopen-2018-022993. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6. Lin Y. M., Wang M., Sun N. X., Liu Y. Y., Yin T. F., and Chen C., “Screening and Application of Nutritional Support in Elderly Hospitalized Patients of a Tertiary Care Hospital in China,” PLoS One 14, no. 3 (2019): e0213076, 10.1371/journal.pone.0213076. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7. Allard J. P., Keller H., Jeejeebhoy K. N., et al., “Decline in Nutritional Status Is Associated With Prolonged Length of Stay in Hospitalized Patients Admitted for 7 Days or More: A Prospective Cohort Study,” Clinical Nutrition 35, no. 1 (February 2016): 144–152, 10.1016/j.clnu.2015.01.009. [DOI] [PubMed] [Google Scholar]
- 8. Khalatbari‐Soltani S. and Marques‐Vidal P., “The Economic Cost of Hospital Malnutrition in Europe: A Narrative Review,” Clinical Nutrition ESPEN 10, no. 3 (June 2015): e89–e94, 10.1016/j.clnesp.2015.04.003. [DOI] [PubMed] [Google Scholar]
- 9. Felder S., Lechtenboehmer C., Bally M., et al., “Association of Nutritional Risk and Adverse Medical Outcomes Across Different Medical Inpatient Populations,” Nutrition 31, no. 11–12 (November/December 2015): 1385–1393, 10.1016/j.nut.2015.06.007. [DOI] [PubMed] [Google Scholar]
- 10. World Health Organization , Non‐Communicable Diseases. World Health Organization, accessed May 17, 2026. [Google Scholar]
- 11. Khalatbari‐Soltani S., Waeber G., and Marques‐Vidal P., “Diagnostic Accuracy of Undernutrition Codes in Hospital Administrative Discharge Database: Improvements Needed,” Nutrition 55–56 (November 2018): 111–115, 10.1016/j.nut.2018.03.051. [DOI] [PubMed] [Google Scholar]
- 12. He A. J., “Scaling‐Up Through Piloting: Dual‐Track Provider Payment Reforms in China's Health System,” Health Policy and Planning 38, no. 2 (February 2023): 218–227, 10.1093/heapol/czac080. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13. Kim S., Choi B., Lee K., Lee S., and Kim S., “Assessing the Performance of a Method for Case‐Mix Adjustment in the Korean Diagnosis‐Related Groups (KDRG) System and Its Policy Implications,” Health Research Policy and Systems 19, no. 1 (June 2021): 98, 10.1186/s12961-021-00739-5. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14. Oblein J. and Doley J., “Malnutrition Coding Denials in the Acute Care Setting,” supplement, Journal of the Academy of Nutrition and Dietetics 125, no. 9S (September 2025): S48–S52, 10.1016/j.jand.2025.05.014. [DOI] [PubMed] [Google Scholar]
- 15. Shepherd E., “Malnutrition Coding and Reimbursement in the Hospital Setting,” Nutrition in Clinical Practice 37, no. 1 (February 2022): 35–40, 10.1002/ncp.10779. [DOI] [PubMed] [Google Scholar]
- 16. Chang W. F., Yan X. Y., Ling H., Liu T., and Luo A. J., “A Study of the Types and Manifestations of Physicians' Unintended Behaviors in the DRG Payment System,” Frontiers in Public Health 11 (2023): 1141981, 10.3389/fpubh.2023.1141981. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17. Joiner K. A., Lin J., and Pantano J., “Upcoding in Medicare: Where Does It Matter Most?,” Health Economics Review 14, no. 1 (January 2024): 1, 10.1186/s13561-023-00465-4. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18. Nimptsch U., “Disease‐Specific Trends of Comorbidity Coding and Implications for Risk Adjustment in Hospital Administrative Data,” Health Services Research 51, no. 3 (June 2016): 981–1001, 10.1111/1475-6773.12398. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19. Dong X., He X., and Wu J., “Can DRG Payment Under Global Budget With Price Adjustment Curb the Soaring Healthcare Costs? Insights From China's Reform and Its Underlying Mechanism,” BMC Public Health 25, no. 1 (November 2025): 3891, 10.1186/s12889-025-25236-0. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20. Zhang H., Fu X., Wu Y., Tang Y., Jin H., and Xie B., “The Impact of DRG‐Based Payment Reform on Inpatient Healthcare Utilization: Evidence From a Natural Experiment in China,” Healthcare (Basel, Switzerland) 13, no. 19 (September 2025): 2424, 10.3390/healthcare13192424. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21. Zhou W., Xu C., Zhang L., Fu H., and Jian W., “Behaviours and Drivers of Diagnosis‐Related Group Upcoding in China: A Mixed‐Methods Study,” Social Science & Medicine (1982) 366 (February 2025): 117660, 10.1016/j.socscimed.2024.117660. [DOI] [PubMed] [Google Scholar]
- 22. Liu H., Jiao J., Zhu M., et al., “Nutritional Status According to the Short‐Form Mini Nutritional Assessment (MNA‐SF) and Clinical Characteristics as Predictors of Length of Stay, Mortality, and Readmissions Among Older Inpatients in China: A National Study,” Frontiers in Nutrition 9 (2022): 815578, 10.3389/fnut.2022.815578. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23. Liu Y., Song C., Wang X., et al., “Prevalence of Malnutrition Among Adult Inpatients in China: A Nationwide Cross‐Sectional Study,” Science China Life Sciences 68, no. 5 (May 2025): 1487–1497, 10.1007/s11427-023-2619-7. [DOI] [PubMed] [Google Scholar]
- 24. Xu Zheng E., Zhu X., Zhu Y., Qin Z., Zhang J., and Huang Y., “Impact of Insurance on Readmission Rates, Healthcare Expenditures, and Length of Hospital Stay Among Patients With Chronic Ambulatory Care Sensitive Conditions in China,” Healthcare (Basel) 12, no. 17 (September 2024): 1798, 10.3390/healthcare12171798. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25. Lin K., Li Y., Yao Y., Xiong Y., and Xiang L., “The Impact of an Innovative Payment Method on Medical Expenditure, Efficiency, and Quality for Inpatients With Different Types of Medical Insurance: Evidence From a Pilot City, China,” International Journal for Equity in Health 23, no. 1 (June 2024): 115, 10.1186/s12939-024-02196-2. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26. Heinze G. and Schemper M., “A Solution to the Problem of Separation in Logistic Regression,” Statistics in Medicine 21, no. 16 (August 2002): 2409–2419, 10.1002/sim.1047. [DOI] [PubMed] [Google Scholar]
- 27. Rasmussen N. H., Thomsen R. W., Rasmussen H. H., and Søgaard M., “Validity of Diagnostic Coding for Undernutrition in Hospitals,” Clinical Nutrition 35, no. 2 (April 2016): 491–495, 10.1016/j.clnu.2015.03.017. [DOI] [PubMed] [Google Scholar]
- 28. Spika S. B. and Zweifel P., “Buying Efficiency: Optimal Hospital Payment in the Presence of Double Upcoding,” Health Economics Review 9, no. 1 (December 2019): 38, 10.1186/s13561-019-0256-4. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29. Kaegi‐Braun N., Mueller M., Schuetz P., Mueller B., and Kutz A., “Evaluation of Nutritional Support and In‐Hospital Mortality in Patients With Malnutrition,” JAMA Network Open 4, no. 1 (January 2021): e2033433, 10.1001/jamanetworkopen.2020.33433. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30. Ponce J., Anzalone A. J., Bailey K., et al., “Impact of Malnutrition on Clinical Outcomes in Patients Diagnosed With COVID‐19,” Journal of Parenteral and Enteral Nutrition 46, no. 8 (November 2022): 1797–1807, 10.1002/jpen.2418. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 31. Bansal N., Alharbi A., Shah M., Altorok I., Assaly R., and Altorok N., “Impact of Malnutrition on the Outcomes in Patients Admitted With Heart Failure,” Journal of Clinical Medicine 13, no. 14 (July 2024): 4215, 10.3390/jcm13144215. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 32. Jian W., Chan K. Y., Reidpath D. D., and Xu L., “China's Rural‐Urban Care Gap Shrank for Chronic Disease Patients, but Inequities Persist,” Health Affairs 29, no. 12 (December 2010): 2189–2196, 10.1377/hlthaff.2009.0989. [DOI] [PubMed] [Google Scholar]
- 33. Fu X., Wang L., Sun C., et al., “Inequity in Inpatient Services Utilization: A Longitudinal Comparative Analysis of Middle‐Aged and Elderly Patients With the Chronic Non‐Communicable Diseases in China,” International Journal for Equity in Health 19, no. 1 (January 2020): 6, 10.1186/s12939-019-1117-9. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 34. Fernandes A. C., Pessoa A., Vigário M. A., Jager‐Wittenaar H., and Pinho J., “Does Malnutrition Influence Hospital Reimbursement? A Call for Malnutrition Diagnosis and Coding,” Nutrition (Burbank, Los Angeles County, Calif.) 74, (June 2020): 110750, 10.1016/j.nut.2020.110750. [DOI] [PubMed] [Google Scholar]
- 35. Ockenga J., Freudenreich M., Zakonsky R., Norman K., Pirlich M., and Lochs H., “Nutritional Assessment and Management in Hospitalised Patients: Implication for DRG‐Based Reimbursement and Health Care Quality,” Clinical Nutrition 24, no. 6 (December 2005): 913–919, 10.1016/j.clnu.2005.05.019. [DOI] [PubMed] [Google Scholar]
- 36. Schuetz P., Fehr R., Baechli V., et al., “Individualised Nutritional Support in Medical Inpatients at Nutritional Risk: A Randomised Clinical Trial,” Lancet 393, no. 10188 (June 2019): 2312–2321, 10.1016/S0140-6736(18)32776-4. [DOI] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Table S1: Sensitivity analyses: predictors of malnutrition coding and association with clinical outcomes.
Data Availability Statement
The data that support the findings of this study are held by the hospital's administrative department. Restrictions apply to the availability of these data, which were used under license for the current study and are not publicly available due to patient confidentiality and institutional data governance policies. However, de‐identified data may be made available from the corresponding author upon reasonable request.