Can routinely collected electronic medical record (EMR) data support hospital resource allocation? A retrospective analysis of 332,711 presentations to a public quaternary teaching hospital in South Australia (2020-2025)

Can routinely collected electronic medical record (EMR) data support hospital resource allocation? A retrospective analysis of 332,711 presentations to a public quaternary teaching hospital in South Australia (2020-2025) | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Research Article Can routinely collected electronic medical record (EMR) data support hospital resource allocation? A retrospective analysis of 332,711 presentations to a public quaternary teaching hospital in South Australia (2020-2025) Madison Bills, Taryn Bessen, Matthew Williams, Annie Conway, Brandon Stretton, and 5 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-7577977/v1 This work is licensed under a CC BY 4.0 License Status: Published Journal Publication published 27 Mar, 2026 Read the published version in BMC Health Services Research → Version 1 posted 11 You are reading this latest preprint version Abstract Background Hospitals face increasing strain from rising clinical complexity and demand. Traditional resource allocation approaches often lack the granularity and timeliness needed for responsive planning. This study evaluates whether routinely collected electronic medical record (EMR) data can be used to classify hospital inpatients into resource-based groups to support real-time planning and hospital-wide operational management. Methods A retrospective analysis was conducted on 332,711 inpatient admissions to a quaternary public hospital in South Australia between January 2020 and January 2025. Patients were classified into one of four flow streams within 72 hours of admission using a resource-based classification framework developed through a modified Delphi process and validated by clinical review. Summary statistics were used to assess differences in resource use across streams and to evaluate classification stability. Data quality limitations and documentation variability were also assessed. Results Flow streams demonstrated distinct differences in length of stay, diagnostic testing, consultations, and allied health input. The model showed strong initial stability, with fewer than 5% of patients changing streams during admission. Key data quality issues included inconsistent consultation documentation, underuse of structured fields, and retrospective overwriting of demographic information, affecting visibility of resource use. Despite these limitations, flow stream classification effectively differentiated patients by resource intensity and care complexity, offering a practical framework to support real-time hospital operations, complementing diagnosis-based groupings. Conclusion A structured classification model using routinely collected EMR data can differentiate inpatient resource needs. Flow stream stratification offers a complementary approach to traditional coding-based systems and may help identify operational bottlenecks. With improved documentation and system integration, this approach could enhance hospital responsiveness, resource planning, and overall system performance. Electronic Medical Records Hospital Resource Management Health Information Systems Data Quality Patient Flow Figures Figure 1 Figure 2 Figure 3 BACKGROUND The performance of Australian public hospitals is increasingly challenged by finite funding, rising patient demand, and growing clinical complexity. These pressures contribute to persistent underperformance and capacity constraints across the hospital system, leading to delays in access to care and reduced operational efficiency. [ 1 , 2 ] Admitted patients are typically assigned to general or subspecialty teams, with all aspects of care documented in the electronic medical record (EMR). Diagnosis-Related Groups (DRGs) are retrospectively assigned post-discharge to support hospital reimbursement. While DRGs are well established for funding, they offer limited operational utility in real-time and lack prospective insight into resource consumption or patient complexity during admission.[ 3 ] EMRs in their current form are largely optimised for documentation, funding reconciliation and regulatory reporting, rather than as dynamic tools that support real-time operational decision making. This retrospective nature hinders timely adjustments to care pathways or resource allocation, as decisions are based on historical data rather than current patient needs. Without active visibility of resource consumption and patient complexity, healthcare providers face significant challenges in delivering timely, efficient and resource-appropriate care. The Theory of Constraints (TOC) provides a structured framework for identifying and managing bottlenecks within complex adaptive systems, such as healthcare.[ 4 ] By focusing on performance limiting constraints, TOC enables targeted interventions to improve throughput and efficiency contributing to workflow redesign by redirecting patient pathways away from identified bottlenecks resulting from workforce shortages, increased demand for specific services or rigid legacy rule-based models of care. Applied to hospital operations, TOC can facilitate real-time decision-making and resource optimisation, addressing the limitations posed by retrospective data systems like DRGs. Despite claims that EMR systems can enhance clinical care and support proactive resource planning,[ 5 ], these benefits have yet to be fully realised in practice.[ 6 , 7 ] Outcomes remain highly dependent on the operational characteristics and workflow alignment within each healthcare organisation. A disconnect remains between EMR system design and clinical workflows, resulting in limited visibility of system-wide operations and actual resource use.[ 8 , 9 ] We hypothesise that stratifying patients into flow streams based on care demands early in their admission, using routinely collected EMR data, could unlock potential to understand service demand in real-time and provide actionable intelligence to guide hospital-wide resource management and improve system efficiency. The aim of this study was to evaluate whether routinely collected data in the EMR could feasibly be used to classify admitted patients into cohorts based on resource use patterns during early admission. METHODOLOGY The study is a retrospective analysis of EMR data of all patients admitted to a quaternary hospital in South Australia between January 2020 and January 2025. The study consists of three key components: data extraction, creation of new patient flow streams, and descriptive analysis. This study is reported in accordance with the RECORD (Reporting of studies Conducted using Observational Routinely-collected health Data) guidelines.[10] Data Ethics Approval was granted by the Central Adelaide Local Health Network (CALHN) Human Research Ethics Review Committee (2024/HRE00225), with a waiver of individual consent due to the retrospective nature and scale of the cohort. Access to EMR data was authorised by the data custodian, and data handling complied with relevant governance and privacy legislation. Data source, extraction and internal linkage to capture encounters Data were extracted from the hospital’s fully integrated Altera Digital Health Sunrise EMR system, using the SA Health Data Analytics Platform, an integrated data infrastructure capable of large-scale interrogation and extraction from the EMR. To accurately represent complete admission episodes, internal deterministic linkage was performed using medical record numbers (MRNs) and time-stamped metadata to consolidate care events (e.g., emergency department visits and inpatient stays) into a unified patient admission journey for each presentation. A time-based threshold associated related care events. MRNs were subsequently mapped to a study-specific unique patient identifier in a separate, secure mapping file maintained outside the analytical dataset. This approach enabled the dataset to remain fully de-identified during analysis, while preserving the capacity for controlled re-identification. Manual review was conducted for a small number of cases with documentation inconsistencies. Linkage accuracy was verified through sample checks and cross-validation with official admission counts from the health network’s Business Intelligence Unit. Data variables No sampling or exclusion criteria were applied to the admission cohort in this study, to capture the complex operational reality of diverse, dynamic, and evolving nature of hospital resource demands. This comprehensive approach ensures that infrequent but high-impact events, such as seasonal demand surges, unplanned service disruptions, or escalation responses are represented in the dataset. Such inclusivity enables a more accurate understanding of system pressures across routine and exceptional conditions. For example, instances where a “Code White” is declared, signalling severe and sustained access block under the CALHN Demand Escalation Framework (unpublished internal document, Central Adelaide Local Health Network, 2025), can now be examined in relation to real-time admission volumes, resource consumption patterns, and flow stream distribution. A comprehensive range of variables were extracted from the EMR including demographic details, clinical information such as comorbidities, diagnostics (pathology and imaging) and treatments (medications, procedures, operations, and therapies). Risk assessment data were captured using validated tools for skin integrity,[11] cognition,[12] malnutrition,[13] and mobility.[14] In addition, resource utilisation metrics were collected, including allied health and pharmacy input, specialty consultations, hospital length of stay, one to one nursing care, intra-hospital care transfers, ward locations, clinical coding data (Table 1). To explore the relationship between evolving care needs and classification stability, a convenience sample of cases with clearly observed changes in flow stream allocation were reviewed to assess corresponding shifts in patients care requirements and resource utilisation. Table 1. Summary of Variables Extracted from the Electronic Medical Record (EMR) Variable Description Demographic Details Age Gender Postcode Clinical Information Comorbidities Treatment received Clinical course Discharge disposition Risk Assessments Validated tools for skin integrity (Braden Score[11]), cognition (4AT[12]), malnutrition (MUST[13]), and mobility[14]). Diagnostic Information Blood Tests CT and MR orders Interventional Radiology orders Resource Utilisation Metrics Allied health* and pharmacy consultation Sub-specialty consultations Hospital length of stay One to one nursing care Intra-hospital transfers Ward type (home vs. outlier) ICU admission Coding Data International Classification of Diseases (ICD)-10 codes** DRG Procedure codes [15] DRG Severity codes [15] *Allied Health Professional (AHP) considered as Physiotherapy, Occupational Therapy, Speech Pathology, Dietetics and Social Work. ** ICD-10 is the 10th revision of the International Classification of Diseases, a medical classification list by the World Health Organization.[16] Data de-identification and cleaning All patient data were de-identified using a unique study identifier. The dataset was cleaned to ensure consistency, including removal of duplicate entries, standardising formats (e.g., date/time fields), correction of errors, and exclusion of clinically implausible outliers (e.g., negative timestamped data). Missing data were addressed pragmatically: if a variable was not recorded in structured EMR fields, it was assumed not collected. This assumption reflects routine clinical documentation practices and the study’s focus on operational visibility from real-time data. However, this may underestimate some clinical characteristics or care events captured only in free text or omitted due to documentation variability. Data Quality Assessment Data quality was assessed to evaluate the completeness, consistency, and operational reliability of key variables prior to analysis. This included verifying demographic information, inconsistencies in service use documentation, and limitations in patient journey mapping across care transitions. Particular attention was given to data elements central to modelling patient flow and resource utilisation such as time-stamped transitions of care, moves to critical service delivery areas for urgent interventions and diagnostics. Front-end clinical documentation was compared with analytics extracts to identify discrepancies and assess the extent to which informal workflows (often recorded in unstructured free-text fields) were missing from the structured EMR data that is used for analysis and planning. These insights informed downstream modelling decisions and guided interpretation of care delivery patterns. Creation of new patient flow streams A novel classification framework was developed to stratify patients into four distinct flow streams based on resource use. The initial framework was established through a modified Delphi Technique, involving structured consensus from a multidisciplinary panel of expert clinicians with operational and frontline insight. This iterative, expert informed approach enabled the development of clinically meaningful flow streams grounded in real world care delivery. Stream definitions were subsequently refined throughout the analysis to better reflect observed variations in patient care profiles resulting in four discrete resource-utilisation categories as shown in Table 2. Table 2. Flow Stream Definitions and Classification Criteria Flow Stream Definition Criteria Operational Considerations Opportunities for Future Improvement 1 Short-stay or ambulatory care patients Day Procedure or Treatment LOS <24 hours Operational interest when expected discharge is delayed or procedural outcomes are not achieved. Refine patient selection protocols to maximise efficiency and minimise avoidable overnight stays. 2 Admissions for a primary condition under a single clinical specialty ≤1 medical consult Length of stay < 5 days Well-suited to care standardisation through clinical pathways. Optimise clinical workflow and care timeliness through pathway adherence and streamlined delivery. 3 Patients of intermediate complexity requiring multidisciplinary input Do not meet criteria for FS1, FS2, or FS4 Benefit from early senior clinical decision-making to determine care pathways. Implement targeted risk stratification and escalation triggers to reduce care delays and prevent deterioration. 4 Hypercomplex, high-acuity admissions requiring intensive resource mobilisation ICU admission ≥ 24 hours OR Critical Location – ED Resus, Theatre, ICU, Spinal Emergency surgery or theatre listed Treatment Intensity (CT or MR) and ≥ 5 bloods Group + Hold ordered Within 24hours: ≥3 specialist consults ≥4 intra-hospital transfers Benefit from proactive care coordination to manage complexity and mitigate system strain. Design robust models of care with embedded coordination, workforce agility, and real-time monitoring to manage surge and mitigate systemic disruption. Note: Some criteria (e.g. length of stay < 5 days are assessed dynamically during admission. Patients may be reclassified as their care evolves. For instance, a patient initially meeting Stream 2 criteria may transition to Stream 3 or 4, if their admission exceeds five days or requires additional resources. This dynamic assignment supports real-time planning and operational responsiveness. To minimise misclassification and reduce potential confounding, a structured hierarchical allocation was used to classify patients into flow streams based on observed care delivery patterns. Each admission was retrospectively assessed for eligibility beginning with Flow Stream 1 (FS1), which reflects short-stay, low-resource admissions. If criteria were not met, patients were sequentially evaluated for assignment to Flow Stream 4 (FS4), 2 (FS2), and finally 3 (FS3), based on observed resource intensity and care characteristics. This approach ensured assignment to the highest appropriate resource-intensity stream and prioritised early identification of high-demand cases for operational planning. While this study focused on retrospective classification using complete admission data, the same logic may be applied at key early points in the patient journey (e.g. 24 or 72 hours) to predict likely stream assignment based on partial data. Patients who did not meet defined criteria for Streams 1, 2, or 4 were allocated to FS3, a heterogeneous group retained for future refinement through machine learning-based classification models. Clinical Validation of Flow Stream Allocation Clinical validation was undertaken on a stratified sample across all flow streams, using full EMR records accessed through the clinical interface. Classification accuracy was assessed by comparing assigned stream categories against actual care pathways and resource use. Discrepant cases were reviewed in multidisciplinary consensus discussions, and findings were used to refine stream definitions and allocation logic. Full methodology and examples are provided in Appendix 1. Statistical analysis Descriptive statistics examined hospital resource utilisation trends at three time points: 24 hours, 72 hours, and at discharge. This enabled assessment of transitions between flow streams or shifts in resource-use profiles over the course of an admission. Summary measures included medians and interquartile ranges of key resource variables, as these are more appropriate than means for data with skewed distributions and operational outliers, such as length of stay and consult counts. Comparative analysis across flow streams was undertaken to evaluate the alignment between anticipated and actual resource intensity, informing iterative refinement of stream definitions and identification of potentially misclassified or transitional cases. For example, the original FS2 criteria limited patients to a single sub-specialty consult; however, analysis identified a subset of short-stay patients who received a second consult from their usual treating team to support continuity of care. In response, the classification rule was modified to allow up to two consults in FS2. Statistical analyses were performed using R, version 4.4.3 and STATA, version 18.5. The summary statistics in Table 3 were generated with the R package, gtsummary.[17] RESULTS Between January 2020 and January 2025, a total of 332,711 inpatient admissions were recorded. The distribution of these admissions by source and final flow stream classification is shown in Fig. 1 . Descriptive characteristics, including age, gender, medication burden, and list of investigations, were analysed and stratified across flow streams to identify differences in patient complexity and care patterns. A high-level summary is presented in Table 3 , with the full set of descriptive statistics provided in Appendix 2. Patterns of hospital length of stay by flow stream, also shown in Table 3 , are visualised in Fig. 2 to highlight distributional differences. These variables were used to inform flow stream classification and support early assessment of hospital resource distribution. All findings presented in this section are descriptive and unadjusted, consistent with the exploratory objectives of this study’s initial phase. No statistical modelling or hypothesis testing was performed, as the focus was on classification accuracy and resource profiling. Table 3 Summary Characteristics of the Total Study Population (N = 332,711), Stratified by Flow Stream Variable Stream 1 N = 149,682 Stream 2 N = 108,840 Stream 3 N = 61,936 Stream 4 N = 12,253 Age, years 1 61 (48, 72) 62 (44, 77) 70 (56, 81) 62 (49, 73) Female, n (%) 59,799 (40.0) 47,750 (44.0) 27,548 (44.0) 4,636 (38.0) Length of Stay, days 1 0.2 (0.2, 0.3) 2.0 (1.1, 3.1) 8.4 (6.2, 13.3) 11.2 (6.4, 20.3) Admission Type 2 , n (%) Medical 108,437 (72.4) 53,338 (49.0) 37,144 (60.0) 5,367 (43.8) Surgical 28,027 (18.7) 54,471 (50.0) 24,607 (39.7) 6,877 (56.2) Other 3 13,218 (8.8) 1,031 (0.9) 185 (0.3) 9 (< 0.1) 1 Median (interquartile range). 2 Admission Type – Mental Health admissions were excluded from this analysis. 3 Other: Emergency, Hyperbaric, Gynaecology Data Quality Assessment Evaluation of EMR data quality revealed substantial limitations in the completeness and standardisation of critical data fields. While admission note completion rates were high within 72 hours, structured fields were often bypassed in favour of free-text documentation, limiting downstream extractability. High variability in documentation practises was noted across clinical teams and disciplines, contributing to inconsistent data capture and reducing the reliability of certain structured fields for operational analysis. Mandatory fields were frequently populated with inaccurate or generic responses, for example, patients admitted from residential aged care were commonly recorded as arriving from private residences. Structured problem lists, intended to support longitudinal care continuity, were populated in less than 75% of cases. High-acuity events such as the massive transfusion pack (MTP) activation were significantly underreported in the EMR compared to external validation sources (e.g., 144 EMR entries in total cohort versus 355 blood bank activations in a single year). Consult orders were inconsistently used across clinical teams, particularly in urgent settings where informal referrals and verbal communication were preferred over electronic pathways. In some cases, consults were entered by the admitting team for their own service as part of the admission process, which distorted flow stream allocation. Due to inconsistencies in naming conventions between consult orders and admitting team identifiers, manual coding was necessary to accurately match and exclude these entries for valid classification (see Appendix 2 for variable-level missing data). No linkage failures were identified in the final dataset. Manual validation of a subset of patient records confirmed concordance between administrative admission identifiers and consolidated care episodes, supporting the integrity of the linkage process. Flow Stream Assignment All eligible admissions were classified into one of four resource-based flow streams using the allocation framework. The results are summarised in Table 4 , comparing the initial projected distribution of patients across streams with actual classifications at 24 hours, 72 hours and discharge, allowing assessment of classification stability and operational alignment with early admission profiles. As the model reflects resource utilisation, patient allocation may shift during admission in response to changes in care intensity or clinical deterioration, a characteristic designed to support real-time operational insight, as visualised in Fig. 3 . Table 4 Flow Stream Classification of Admissions at 24 Hours, 72 Hours, and Discharge (N = 332,711) Flow Stream 24 hours, n (%) 72 hours, n (%) Discharge, n (%) FS1 149,682 (45.0) 149,682 (45.0) 149,682 (45.0) FS2 111,339 (33.5) 108,991 (32.8) 108,840 (32.7) FS3 70,708 (21.3) 64,435 (19.4) 61,936 (18.6) FS4 982 (0.3) 9,603 (2.9) 12,253 (3.7) Note : Percentages are based on total admissions (N = 332,711) at each time point. Values are rounded to one decimal place for clarity. Fewer than 5% of patients changed flow stream assignment during admission. These changes predominantly involved transitions from FS3 to FS4 as clinical complexity emerged, and from FS2 to FS4 in cases requiring escalation to higher levels of care. This pattern is evident in the transition flow visualised in Fig. 3 , where movement from FS2 and FS3 to FS4 accounts for most reclassifications. For example, a patient initially classified as FS2 due to a stable embolic stroke presentation was later reclassified as FS4 following rapid clinical deterioration, requiring ICU admission and neurosurgical intervention. This low reallocation rate suggests strong initial model stability and highlights stream transitions are not frequent, instead representing meaningful changes in patient condition or care requirements. Establishing baseline ratios of the flow stream is critical to evaluating responses to external and internal system characteristics over time, to allow changes to resource reallocation or other strategic interventions. Operational Utility of Flow Stream Classification Initial application of the classification model revealed clear differentiation in resource-use profiles across streams. FS1 patients followed short-stay, low-resource pathways, with minimal investigations or consults, a median LOS of 0.2 days, rare ICU admission (3.1%) and polypharmacy in 0.6% of cases. FS2 admissions remained within specialty teams with relatively short lengths of stay (median 2.0 days) and focused care delivery, with low Allied Health input (median 0 (0,1)). FS3 captured patients of intermediate complexity shown by median LOS of 8.4 days, 21% requiring multidisciplinary input from medical and allied health. Polypharmacy was high (93%) and ICU use was greater than FS1 and FS2. FS4 admissions represented high acuity care: 96% were admitted to ICU, 36% had multiple specialise consultations and the median length of stay was 11.2 days. DISCUSSION This retrospective observational study evaluated the operational utility of a novel, resource-based patient classification model using routinely collected electronic medical record (EMR) data. The study aimed to determine whether structured EMR variables could reliably stratify patients into flow streams reflective of care complexity and resource use to inform hospital-wide resource allocation. The analysis included inpatient admissions (n = 332,711) to a quaternary hospital in South Australia between January 2020 and January 2025. While the model was developed in an Australian hospital, the operational challenges addressed, such as managing resource-intensive cohorts, identifying system bottlenecks, and supporting real-time planning, are shared across many hospital systems globally. The framework is designed to be adaptable to local infrastructure, EMR capabilities, and policy environments. A comprehensive dataset was constructed through deterministic linkage and rigorous data cleaning, incorporating demographic, clinical, diagnostic, and operational variables. Patients were stratified into one of four flow streams based on resource utilisation rather than diagnostic grouping. Despite limitations in structured EMR documentation, particularly variability in consult orders and data entry, the model demonstrated early utility. Flow streams captured distinct care complexity and service utilisation patterns, which remained stable over the inpatient stay.[ 7 ] Fewer than 5% of patients changed stream assignment during admission, suggesting allocation stability and that transitions likely reflect meaningful clinical changes. By grounding classification in observed resource consumption, this model enables dynamic, real-time operational planning. In line with Theory of Constraints (TOC) principles, it supports early identification of operational bottlenecks and enhances system responsiveness for staffing, bed management, and escalation planning, before broader strain emerges. While initial results demonstrated differentiation between streams, the broader value lies in its potential to inform continuous operational insight and pre-emptive decision making across the hospital, while avoiding the unintended consequence of directing additional workload toward already strained services. Many hospital performance constraints are internally generated, arising from how care delivery is structured, prioritised, and resourced. High resource patients in FS4, for instance, exert disproportionate demand on critical care infrastructure, which can delay care delivery and limit access for lower-acuity patients in FS1, FS2 and FS3. This cascading effect highlights the need to monitor high-acuity cohorts to mitigate system-wide impacts. TOC asserts that inefficiencies stem from unidentified or unmanaged constraints, which lead to cascading delays, access blocks, and disproportionate demands on high-acuity services. These constraints are not static and can be mitigated through targeted redesign and strategic resource realignment. The flow stream model offers a mechanism to identify operational pressure points by stratifying patients according to resource demand, supporting both direct benefits, such as early identification of capacity limits, and indirect benefits, including improved patient flow for other cohorts. These dynamics are especially relevant during periods of high occupancy, where marginal increases in volume can lead to disproportionate delays, reduced flexibility and compromised overall system efficiency. By anticipating pressure points and supporting proactive resource alignment, the model may help sustain performance under demand stress. Without such visibility, health services risk reinforcing inefficiencies by directing demand to constrained areas, inadvertently exacerbating stress where there is least capacity. Embedding this model into real-time operational planning may improve system resilience and performance at scale. A critical next step is evaluating how the distributions shift over time in response to internal or external pressures. For example, a rise in FS4 admissions may signal emerging strain. These patients often bypass routine scheduling, displacing lower-acuity patients and contributing to delays, blocked beds, and inefficiencies across the hospital. Monitoring these shifts in near real-time could support decision-makers in enacting timely interventions, reallocating resources, or adjusting discharge pathways before system-wide dysfunction occurs. Although EMRs are structured to standardise data capture and support visibility, the reality of frontline practice often diverges. Critical workflows remain undocumented due to time pressures, perceived irrelevance of fields, or the convenience of informal workarounds. Clinicians bypass digital systems in favour of direct communication (e.g., phone calls) or offline tools (e.g., spreadsheets), which, while efficient in context, obscure visibility at the organisational level. This gap between system design and care delivery impairs the ability to learn from frontline adaptations,[ 18 ] which carries significant operational risks. Planning, funding, and workforce forecasting frameworks increasingly rely on EMR-derived data, yet these decisions are compromised by inconsistencies in data entry, underutilisation of structured fields, and underreporting of resource-intensive activity.[ 19 ] While technically available, structured fields are often bypassed due to poor integration with clinical workflows and a lack of demonstrated value for accurate documentation, to provide system-level insight and responsiveness.[ 19 ] The absence of reliable visibility at scale impairs the organisation’s ability to anticipate demand, respond to emergent complexity, measure care delivery and allocate resources intelligently. To support adoption, decision tools must be embedded into clinicians’ natural workflow rather than additional external overlays. Operationally relevant nudges, and prioritisation prompts, could be integrated directly into the EMR to drive meaningful use.[ 20 ] These challenges also intersect with important policy and ethical considerations. As the study progresses toward predictive analytics and machine learning applications, the ethical integration of such models into clinical and operational practice must be prioritised. Attention to bias mitigation, model transparency[ 21 ] and stakeholder engagement (including clinicians, executives and consumers), will be essential in guiding methodological refinement and governance. This framework offers a structured approach to real-time operational stratification. By monitoring complexity and care intensity, the model supports an adaptive and data-driven approach to hospital-wide resource management. The flow stream model developed will serve as the foundation for machine learning-based predictive analytics in the next phase, aimed at real-time decision making and surge response. Future phases could identify delay points, such as prolonged ED stays or diagnostic bottlenecks, and support earlier escalation, targeted interventions, and more responsive resource alignment. LIMITATIONS This study has several limitations. First, it is based on retrospective EMR data, which is inherently subject to constraints in completeness, accuracy, and consistency of historical documentation. Where structured data were missing, it was assumed that the information had not been recorded during the admission. No imputation was used to preserve record integrity and reflect real-world availability. While aligned with operational practice, this may have underestimated the frequency of some resource utilisation events, particularly those captured only in free-text notes or omitted entirely due to documentation variability. These limitations are well-recognised in real-world data, where incomplete and heterogeneous data collection can limit interpretability, to reliably analyse, contextualise and apply to operational decision making.[ 22 ] Second, the analysis was conducted at a single quaternary hospital and may not be generalisable to settings with different infrastructure, case mix, staffing or EMR capability. Third, some components of care remain inconsistently documented in structured EMR fields, particularly in high-acuity or multidisciplinary environments. Verbal referrals, Medical Emergency Team (MET) calls, and non-standardised documentation of specialist consult input were frequently observed in unstructured notes, leading to an underestimation of resource use and complexity. Additionally, a structural limitation of the EMR system relates to demographic fields that are stored at the client level rather than at the individual admission level. Variables such as usual accommodation and postcode, are retrospectively overwritten across prior admissions if updated during a later encounter. For example, if a patient was admitted from a private residence early in the study period but subsequently moved to a residential aged care facility, all historical admissions may be updated to reflect the latter. While this has limited impact on real-time model development, it introduces a source of temporal inaccuracy for retrospective review and flow stream assignment. In the absence of audit trail access for these fields, this limitation reflects a broader challenge in the underlying EMR data architecture. Finally, this study presents descriptive findings only. While the model demonstrates early operational value, it does not yet incorporate real-time analytics or predictive capabilities. These will be developed and evaluated in subsequent phases of the study. CONCLUSION This study introduces a novel, resource-based flow stream framework to better classify patients by resource use rather than diagnosis. The model has operational relevance for hospital planning and performance monitoring but is constrained by EMR documentation gaps that limit visibility of care delivery. System inefficiencies are compounded by informal workflows and inconsistent structured data use, creating hidden constraints. From a TOC perspective, these bottlenecks restrict system throughput and remain unaddressed without real-time data insight. Future work must focus on strengthening real-time analytics, embedding decision support into EMR workflows, and developing predictive tools grounded in clinical and operational practice. This study forms the foundation for a multi-year evaluation of EMR adaptability, with the next phase focused on near real-time identification of bottlenecks and AI-driven forecasting to support smarter resource allocation. Aligning data with action is essential to building a learning health system that enables agile, proactive hospital operations. Abbreviations 4AT 4 'A's Test (Delirium Screening Tool) AHP Allied Health Professional AI Artificial Intelligence CALHN Central Adelaide Local Health Network CT Computed Tomography DRG Diagnosis-Related Group ED Emergency Department EMR Electronic Medical Record FS Flow Stream HREC Human Research Ethics Committee ICD-10 International Classification of Diseases, 10th Revision ICU Intensive Care Unit LOS Length of Stay MDC Major Diagnostic Category MET Medical Emergency Team MR Magnetic Resonance MRN Medical Record Number MTP Massive Transfusion Pack MUST Malnutrition Universal Screening Tool PET Positron Emission Tomography RAH Royal Adelaide Hospital RECORD REporting of studies Conducted using Observational Routinely collected Data TOC Theory of Constraints Declarations Ethics approval and consent to participate This study was conducted as part of a health service quality improvement initiative and was reviewed by the Central Adelaide Local Health Network (CALHN) Human Research Ethics Committee (HREC). The project was approved under the National Statement on Ethical Conduct in Human Research (reference number: 2024/HRE00225), which is consistent with the principles in the Declaration of Helsinki (2013). Consent to participate was waived due to the retrospective nature of the data and the use of de-identified records. Clinical trial number: not applicable. Consent for publication Not applicable. Availability of data and materials The datasets generated and/or analysed during the current study are not publicly available due to institutional and legal data access restrictions. De-identified data may be made available upon reasonable request and subject to appropriate ethics and governance approvals from SA Health and the Central Adelaide Local Health Network. Competing interests The authors declare that they have no competing interests . Funding This study was supported by grant funding from The Hospital Research Foundation, to facilitate data extraction, analysis, and project coordination. The funder had no role in the study design, data interpretation, or manuscript preparation. Authors' contributions GO and DL led the study conceptualisation, classification framework and design of the study. MW acquired and cleaned the EMR data. MB led the refinement of the classification framework, drafted the manuscript, and contributed to data interpretation and operational application. AC conducted all statistical analyses and supported interpretation of results. TB, BS and TL contributed to overall study design, review and interpretation of the findings and manuscript development and review. JK and SB provided guidance on model development. GO provided overall project supervision and manuscript review. All authors read and approved the final manuscript. Acknowledgements The authors wish to acknowledge Dr Ian Sturgess for his mentorship and guidance in hospital system improvement. His vision and dedication to advancing patient flow and operational performance were instrumental in shaping the conceptual foundations of this study. We gratefully dedicate this manuscript to him. References Australian Medical Association. Public hospitals in logjam as funding pressures grow . 2023. https://ama.com.au (accessed 19 Aug 2025). Duckett S, Breadon P. Controlling costs and improving care: reforming the hospital funding system . Grattan Institute 2014. Jackson T, Michel JL, Roberts RF, et al. A classification of hospital admissions. BMC Health Serv Res 2014;14:346. Almeida MA, Marinho MMO. Theory of constraints in healthcare: a systematic literature review. Int J Qual Reliab Manag 2022;39(3):716–37. Yadav S, Kumar R, Tran T, et al. Assessing the predictive and analytics capability of electronic medical records for operational planning. J Healthc Inform Res 2023;7(2):123–35. Zozus MN, Stansbury DW, Raskin S, et al. Assessing data quality for healthcare systems. eGEMs 2019;7(1):20. Reimer AP, Milinovich A, Madigan EA. Data quality assessment framework to assess electronic medical record data for use in research. Int J Med Inform 2016;90:40–7. McGinn CA, Gagnon MP, Shaw N, et al. Users’ perspectives of key factors to implementing electronic health records in Canada. JMIR Med Inform 2011;13(3):e73. Ajami S, Bagheri-Tadi T. Barriers for adopting electronic health records (EHRs) by physicians. Acta Inform Med 2013;21(2):129–34. Benchimol EL, Smeeth L, Guttmann A, et al. The reporting of studies conducted using observational routinely-collected health data (RECORD) statement. PLoS Med 2015;12:e1001885. Braden B, Bergstrom N. A conceptual schema for the study of the etiology of pressure sores. Rehabil Nurs 1987;12(1):8–12. Bellelli G, Morandi A, Davis DHJ, et al. Validation of the 4AT, a new instrument for rapid delirium screening: a study in 234 hospitalised older people. Age Ageing 2014;43(4):496–502. Elia M. The 'MUST' report: nutritional screening of adults: a multidisciplinary responsibility. Redditch, UK: BAPEN 2003. SA Health. Fall injury and prevention clinical guideline: screening, assessment, care planning and discharge planning . Government of South Australia 2018. https://www.sahealth.sa.gov.au/... (accessed 19 Aug 2025). Independent Health and Aged Care Pricing Authority. AR-DRG Version 11.0. https://www.ihacpa.gov.au/resources/ar-drg-version-110 (accessed 27 Aug 2025). World Health Organization. International Classification of Diseases, 10th revision (ICD-10) . https://www.who.int/classifications/icd/en/. Sjoberg D, Whiting K, Curry M, et al. Reproducible summary tables with the gtsummary package. R J 2021;13:570–80. https://doi.org/10.32614/RJ-2021-053. Verhagen MJ, de Vos MS, Sujan M, Hamming JF. The problem with making Safety-II work in healthcare . BMJ Qual Saf. 2022;31(5):402–408. doi:10.1136/bmjqs-2021-014396 de Groot K, de Bruijne M, Paans W, et al. Effective and feasible interventions to improve structured EHR data quality: a systematic review. Int J Med Inform 2023;174:105050. Alexiuk M, Ashcroft R, Pijl-Zieber E, et al. Clinical decision support tools in the electronic medical record. Kidney Int Rep 2023;8(9):929–38. Sendak MP, D’Arcy J, Kashyap S, et al. Clinical implementation of predictive models embedded within electronic health records. J Am Med Inform Assoc 2020;27(4):639–46. Weiskopf NG, Weng C. Methods and dimensions of electronic health record data quality assessment: enabling reuse for clinical research. J Am Med Inform Assoc 2013;20(1):144–51. Additional Declarations No competing interests reported. Supplementary Files AdditionalFileAppendix1.docx AdditionalFileAppendix2.docx Cite Share Download PDF Status: Published Journal Publication published 27 Mar, 2026 Read the published version in BMC Health Services Research → Version 1 posted Editorial decision: Revision requested 28 Jan, 2026 Reviews received at journal 05 Jan, 2026 Reviews received at journal 24 Nov, 2025 Reviewers agreed at journal 24 Nov, 2025 Reviewers agreed at journal 23 Nov, 2025 Reviewers agreed at journal 23 Nov, 2025 Reviewers invited by journal 21 Sep, 2025 Editor invited by journal 17 Sep, 2025 Editor assigned by journal 16 Sep, 2025 Submission checks completed at journal 16 Sep, 2025 First submitted to journal 09 Sep, 2025 You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-7577977","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":521223299,"identity":"3b97e66f-91cf-4986-82c3-d1a74d04f35d","order_by":0,"name":"Madison Bills","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA/0lEQVRIiWNgGAWjYBACgwMgBAT2zMyNBxgqiNBiCdNi2MzYcIDhDBFa7IFaoNYBtTC2EaHF7PjhjQd/VNxhMGxnbDj4c16dPAP/4mMSDDWHcWs5k1ZwmOfMM6BfGBsO8247bNgg8SxNguEYHi0HcgwOM7YdBvvlMOM2oPMkzhgbMLDh1mJw/o3BwZ9ALUCNQIfNqbOHaPmHR8uNHIMDvFAtB3gbmBMb+HsMH4Dsxa3lGcgvh3nADuM5dji5TYIt8UFiXzoehyVv/vij4rCcPf/hgw9/1NTZ9vMfPnDgwzdrnFpggAfOYpNIYGBIIKgBBfAfIE39KBgFo2AUDHsAACojYg4Dwu7jAAAAAElFTkSuQmCC","orcid":"","institution":"CALHN EpiCentre, Central Adelaide Local Health Network, SA Health","correspondingAuthor":true,"prefix":"","firstName":"Madison","middleName":"","lastName":"Bills","suffix":""},{"id":521223300,"identity":"20ed1efa-c0a3-4ae7-b149-1aee6c02722f","order_by":1,"name":"Taryn Bessen","email":"","orcid":"","institution":"Central Adelaide Local Health Network, SA Health","correspondingAuthor":false,"prefix":"","firstName":"Taryn","middleName":"","lastName":"Bessen","suffix":""},{"id":521223301,"identity":"ab30d67e-563b-4828-861d-0e5f03500b00","order_by":2,"name":"Matthew Williams","email":"","orcid":"","institution":"EMR Optimisation, Central Adelaide Local Health Network, SA Health","correspondingAuthor":false,"prefix":"","firstName":"Matthew","middleName":"","lastName":"Williams","suffix":""},{"id":521223302,"identity":"6ee6a18e-4215-4876-ad2a-71ace10d9d52","order_by":3,"name":"Annie Conway","email":"","orcid":"","institution":"CALHN EpiCentre, Central Adelaide Local Health Network, SA Health","correspondingAuthor":false,"prefix":"","firstName":"Annie","middleName":"","lastName":"Conway","suffix":""},{"id":521223303,"identity":"be07f590-d09c-4e20-9a87-f110dd822295","order_by":4,"name":"Brandon Stretton","email":"","orcid":"","institution":"Central Adelaide Local Health Network, SA Health","correspondingAuthor":false,"prefix":"","firstName":"Brandon","middleName":"","lastName":"Stretton","suffix":""},{"id":521223304,"identity":"2cc28848-82dd-4d5a-957a-1f540200510e","order_by":5,"name":"Tony Loan","email":"","orcid":"","institution":"Central Adelaide Local Health Network, SA Health","correspondingAuthor":false,"prefix":"","firstName":"Tony","middleName":"","lastName":"Loan","suffix":""},{"id":521223305,"identity":"9a2d5fc1-18ce-4133-9991-1d079f55c5d8","order_by":6,"name":"Joshua Kovoor","email":"","orcid":"","institution":"Southern Adelaide Local Health Network, SA Health","correspondingAuthor":false,"prefix":"","firstName":"Joshua","middleName":"","lastName":"Kovoor","suffix":""},{"id":521223306,"identity":"61e5c0df-9d92-4218-83fe-0b71cac9b040","order_by":7,"name":"Danny Liew","email":"","orcid":"","institution":"The University of Adelaide 4","correspondingAuthor":false,"prefix":"","firstName":"Danny","middleName":"","lastName":"Liew","suffix":""},{"id":521223307,"identity":"e2122162-b5e5-4178-9b27-b2566d67a81b","order_by":8,"name":"Stephen Bacchi","email":"","orcid":"","institution":"Lyell McEwin Hospital, Northern Adelaide Local Health Network, The University of Adelaide","correspondingAuthor":false,"prefix":"","firstName":"Stephen","middleName":"","lastName":"Bacchi","suffix":""},{"id":521223308,"identity":"a8a0dc6e-08ca-4e5d-8fda-32e3c7a1dbeb","order_by":9,"name":"Gerry O’Callaghan","email":"","orcid":"","institution":"Central Adelaide Local Health Network, SA Health","correspondingAuthor":false,"prefix":"","firstName":"Gerry","middleName":"","lastName":"O’Callaghan","suffix":""}],"badges":[],"createdAt":"2025-09-10 02:38:18","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-7577977/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-7577977/v1","draftVersion":[],"editorialEvents":[{"content":"https://doi.org/10.1186/s12913-026-14448-8","type":"published","date":"2026-03-27T16:11:07+00:00"}],"editorialNote":"","failedWorkflow":false,"files":[{"id":92682902,"identity":"43636b65-c083-4355-8f4f-07723a0dcb2b","added_by":"auto","created_at":"2025-10-03 01:23:41","extension":"docx","order_by":0,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":259599,"visible":true,"origin":"","legend":"","description":"","filename":"MBillsFullTextHospitalOperationsEMRDataHealthServicesResearchFinal1.docx","url":"https://assets-eu.researchsquare.com/files/rs-7577977/v1/b4198e658d125499b3d300c4.docx"},{"id":92680967,"identity":"a3223659-d046-4344-8b21-35e863224111","added_by":"auto","created_at":"2025-10-03 01:07:41","extension":"png","order_by":1,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":194958,"visible":true,"origin":"","legend":"","description":"","filename":"Figure1Distributionofinpatientadmissions.png","url":"https://assets-eu.researchsquare.com/files/rs-7577977/v1/98013c0f53bb684ee79f38c5.png"},{"id":92680972,"identity":"229a32fd-74c0-4e9c-b317-b75b804c3b58","added_by":"auto","created_at":"2025-10-03 01:07:41","extension":"png","order_by":2,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":19078,"visible":true,"origin":"","legend":"","description":"","filename":"Figure2LengthofStayDistributionstratifiedbystream.png","url":"https://assets-eu.researchsquare.com/files/rs-7577977/v1/3a2f70d7d6b34c286abe63f9.png"},{"id":92680986,"identity":"aa8e1ec0-c626-465e-a3c7-698e4dcf2f12","added_by":"auto","created_at":"2025-10-03 01:07:42","extension":"png","order_by":3,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":159124,"visible":true,"origin":"","legend":"","description":"","filename":"Figure3FlowStreamClassificationtransitions.png","url":"https://assets-eu.researchsquare.com/files/rs-7577977/v1/033706eb1c1b8c3887a22c60.png"},{"id":92682903,"identity":"579599e6-ba3f-4f00-ab2f-115d8160c9a7","added_by":"auto","created_at":"2025-10-03 01:23:41","extension":"json","order_by":4,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":11972,"visible":true,"origin":"","legend":"","description":"","filename":"4396fb2332cd4227888c3f8cc9452a8c.json","url":"https://assets-eu.researchsquare.com/files/rs-7577977/v1/65f535496d51bc00dcb0d8b0.json"},{"id":92680976,"identity":"8ae7b532-f62c-4902-8c37-1e01b736af59","added_by":"auto","created_at":"2025-10-03 01:07:41","extension":"docx","order_by":5,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":41917,"visible":true,"origin":"","legend":"","description":"","filename":"AdditionalFileAppendix1.docx","url":"https://assets-eu.researchsquare.com/files/rs-7577977/v1/aaf13459f0da524dbccc313d.docx"},{"id":92680987,"identity":"d0d257b3-c57d-4273-ae16-c60b03985de1","added_by":"auto","created_at":"2025-10-03 01:07:42","extension":"docx","order_by":6,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":54315,"visible":true,"origin":"","legend":"","description":"","filename":"AdditionalFileAppendix2.docx","url":"https://assets-eu.researchsquare.com/files/rs-7577977/v1/d17979190c39247f747af3c9.docx"},{"id":92682905,"identity":"6013a3da-8da9-4f5a-87b1-5745e246c13e","added_by":"auto","created_at":"2025-10-03 01:23:42","extension":"xml","order_by":7,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":96443,"visible":true,"origin":"","legend":"","description":"","filename":"4396fb2332cd4227888c3f8cc9452a8c1enriched.xml","url":"https://assets-eu.researchsquare.com/files/rs-7577977/v1/6931ccc940fa41b09d4259f0.xml"},{"id":92680992,"identity":"d7f852b6-4812-409e-8954-ca3aae41c008","added_by":"auto","created_at":"2025-10-03 01:07:42","extension":"png","order_by":8,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":194958,"visible":true,"origin":"","legend":"","description":"","filename":"Figure1Distributionofinpatientadmissions.png","url":"https://assets-eu.researchsquare.com/files/rs-7577977/v1/e09be8e6981811e3a14406c0.png"},{"id":92680980,"identity":"6658c4c6-47e5-4063-89f8-aa6f1193a344","added_by":"auto","created_at":"2025-10-03 01:07:42","extension":"png","order_by":9,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":19078,"visible":true,"origin":"","legend":"","description":"","filename":"Figure2LengthofStayDistributionstratifiedbystream.png","url":"https://assets-eu.researchsquare.com/files/rs-7577977/v1/df2fce3f55de1d0e27f24163.png"},{"id":92680977,"identity":"c0cae434-ca4b-490a-90a6-02ebd7773c00","added_by":"auto","created_at":"2025-10-03 01:07:42","extension":"png","order_by":10,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":159124,"visible":true,"origin":"","legend":"","description":"","filename":"Figure3FlowStreamClassificationtransitions.png","url":"https://assets-eu.researchsquare.com/files/rs-7577977/v1/d21f622832f2ad14e5b00d9b.png"},{"id":92682451,"identity":"690017bd-2746-48cc-ae1d-660d3712fb3c","added_by":"auto","created_at":"2025-10-03 01:15:42","extension":"png","order_by":11,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":88223,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage1.png","url":"https://assets-eu.researchsquare.com/files/rs-7577977/v1/a87e52de1eeb8fafe5e6bd0c.png"},{"id":92680984,"identity":"3fc1f38a-6f7c-46b8-adc5-372ee8a9a2a6","added_by":"auto","created_at":"2025-10-03 01:07:42","extension":"png","order_by":12,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":19078,"visible":true,"origin":"","legend":"","description":"","filename":"Figure2LengthofStayDistributionstratifiedbystream.png","url":"https://assets-eu.researchsquare.com/files/rs-7577977/v1/ff89bf84ee7b3958bbb0d52c.png"},{"id":92682455,"identity":"ca3db2db-84e4-4458-b72d-05baa5b067c0","added_by":"auto","created_at":"2025-10-03 01:15:42","extension":"png","order_by":13,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":61554,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage3.png","url":"https://assets-eu.researchsquare.com/files/rs-7577977/v1/05ba53badcad103123365487.png"},{"id":92680991,"identity":"078f8155-6cf0-4428-b2c4-a73f4ca14ef5","added_by":"auto","created_at":"2025-10-03 01:07:42","extension":"png","order_by":14,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":33408,"visible":true,"origin":"","legend":"","description":"","filename":"OnlineFigure1Distributionofinpatientadmissions.png","url":"https://assets-eu.researchsquare.com/files/rs-7577977/v1/b034405d83a254ee76a09b23.png"},{"id":92680983,"identity":"3a9364a0-b33f-4f1c-b985-3cee2e113b8f","added_by":"auto","created_at":"2025-10-03 01:07:42","extension":"png","order_by":15,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":9606,"visible":true,"origin":"","legend":"","description":"","filename":"OnlineFigure2LengthofStayDistributionstratifiedbystream.png","url":"https://assets-eu.researchsquare.com/files/rs-7577977/v1/ec4ff29d5c17998e509769fd.png"},{"id":92682450,"identity":"d76e6a2d-3868-451f-9d19-c7763694b918","added_by":"auto","created_at":"2025-10-03 01:15:42","extension":"png","order_by":16,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":18678,"visible":true,"origin":"","legend":"","description":"","filename":"OnlineFigure3FlowStreamClassificationtransitions.png","url":"https://assets-eu.researchsquare.com/files/rs-7577977/v1/d236c65b2ee5fba809b260f1.png"},{"id":92680988,"identity":"8961566f-434c-4098-810f-64234b9a340e","added_by":"auto","created_at":"2025-10-03 01:07:42","extension":"png","order_by":17,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":22983,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage1.png","url":"https://assets-eu.researchsquare.com/files/rs-7577977/v1/71f35be896f24227a65d927c.png"},{"id":92680978,"identity":"1b8cbe54-cd98-4fcb-8533-586ca2e41e3a","added_by":"auto","created_at":"2025-10-03 01:07:42","extension":"png","order_by":18,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":9606,"visible":true,"origin":"","legend":"","description":"","filename":"OnlineFigure2LengthofStayDistributionstratifiedbystream.png","url":"https://assets-eu.researchsquare.com/files/rs-7577977/v1/1b6943530c111b3e32dd4e16.png"},{"id":92682454,"identity":"74c0177b-e042-4310-a8e1-9e601ba3e6f2","added_by":"auto","created_at":"2025-10-03 01:15:42","extension":"png","order_by":19,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":13474,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage3.png","url":"https://assets-eu.researchsquare.com/files/rs-7577977/v1/56072b54982c281bd9b22429.png"},{"id":92682453,"identity":"1ea66562-b992-492d-9bb3-3833ab668729","added_by":"auto","created_at":"2025-10-03 01:15:42","extension":"xml","order_by":20,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":94261,"visible":true,"origin":"","legend":"","description":"","filename":"4396fb2332cd4227888c3f8cc9452a8c1structuring.xml","url":"https://assets-eu.researchsquare.com/files/rs-7577977/v1/f1ad2f67e6119c591172386b.xml"},{"id":92680989,"identity":"fa9d9892-1156-4681-a0c4-457370987193","added_by":"auto","created_at":"2025-10-03 01:07:42","extension":"html","order_by":21,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":108145,"visible":true,"origin":"","legend":"","description":"","filename":"earlyproof.html","url":"https://assets-eu.researchsquare.com/files/rs-7577977/v1/a076b12a7f15c004e243f873.html"},{"id":92680966,"identity":"b6cd136c-da41-4488-9fb0-15e021699415","added_by":"auto","created_at":"2025-10-03 01:07:41","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":73880,"visible":true,"origin":"","legend":"\u003cp\u003eDistribution of inpatient admissions by source and final Flow Stream classification.\u003c/p\u003e","description":"","filename":"1.png","url":"https://assets-eu.researchsquare.com/files/rs-7577977/v1/9f37b199bcd7d8353f5f0307.png"},{"id":92680968,"identity":"9997955f-f7a0-4fa3-a892-0fd216102aee","added_by":"auto","created_at":"2025-10-03 01:07:41","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":51532,"visible":true,"origin":"","legend":"\u003cp\u003eLength of Stay Distribution Stratified by Flow Stream*\u003cstrong\u003e \u003c/strong\u003e\u003cem\u003e*All FS1 LOS = 1 day or less\u003c/em\u003e\u003c/p\u003e","description":"","filename":"2.png","url":"https://assets-eu.researchsquare.com/files/rs-7577977/v1/d4f14023aefe0c401b9fb37c.png"},{"id":92680971,"identity":"de97b2e7-1c18-42a2-b37d-28a9d31a867c","added_by":"auto","created_at":"2025-10-03 01:07:41","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":41115,"visible":true,"origin":"","legend":"\u003cp\u003eFlow Stream classification transitions from 24 hours to discharge (N=332,711).\u003c/p\u003e","description":"","filename":"3.png","url":"https://assets-eu.researchsquare.com/files/rs-7577977/v1/300a9479a11f5a7fb2c1b9bc.png"},{"id":105756019,"identity":"bc38a049-c73c-4911-8c73-c72834f580a2","added_by":"auto","created_at":"2026-03-30 16:34:24","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":1107725,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-7577977/v1/616d3746-2677-4c0d-8f5b-c10f5871e032.pdf"},{"id":92682447,"identity":"60519f6c-0540-400c-9e5a-9b3e5e33865c","added_by":"auto","created_at":"2025-10-03 01:15:41","extension":"docx","order_by":1,"title":"","display":"","copyAsset":false,"role":"supplement","size":41917,"visible":true,"origin":"","legend":"","description":"","filename":"AdditionalFileAppendix1.docx","url":"https://assets-eu.researchsquare.com/files/rs-7577977/v1/f6bb05af695829601700d29d.docx"},{"id":92683075,"identity":"c07ecc0c-3aa7-44c0-bc73-adef5f954973","added_by":"auto","created_at":"2025-10-03 01:31:41","extension":"docx","order_by":2,"title":"","display":"","copyAsset":false,"role":"supplement","size":54315,"visible":true,"origin":"","legend":"","description":"","filename":"AdditionalFileAppendix2.docx","url":"https://assets-eu.researchsquare.com/files/rs-7577977/v1/1c25d1b2673e3fe89b39d427.docx"}],"financialInterests":"No competing interests reported.","formattedTitle":"\u003cp\u003eCan routinely collected electronic medical record (EMR) data support hospital resource allocation? A retrospective analysis of 332,711 presentations to a public quaternary teaching hospital in South Australia (2020-2025)\u003c/p\u003e","fulltext":[{"header":"BACKGROUND","content":"\u003cp\u003eThe performance of Australian public hospitals is increasingly challenged by finite funding, rising patient demand, and growing clinical complexity. These pressures contribute to persistent underperformance and capacity constraints across the hospital system, leading to delays in access to care and reduced operational efficiency. [\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e, \u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e]\u003c/p\u003e\u003cp\u003eAdmitted patients are typically assigned to general or subspecialty teams, with all aspects of care documented in the electronic medical record (EMR). Diagnosis-Related Groups (DRGs) are retrospectively assigned post-discharge to support hospital reimbursement. While DRGs are well established for funding, they offer limited operational utility in real-time and lack prospective insight into resource consumption or patient complexity during admission.[\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e] EMRs in their current form are largely optimised for documentation, funding reconciliation and regulatory reporting, rather than as dynamic tools that support real-time operational decision making. This retrospective nature hinders timely adjustments to care pathways or resource allocation, as decisions are based on historical data rather than current patient needs. Without active visibility of resource consumption and patient complexity, healthcare providers face significant challenges in delivering timely, efficient and resource-appropriate care.\u003c/p\u003e\u003cp\u003eThe Theory of Constraints (TOC) provides a structured framework for identifying and managing bottlenecks within complex adaptive systems, such as healthcare.[\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e] By focusing on performance limiting constraints, TOC enables targeted interventions to improve throughput and efficiency contributing to workflow redesign by redirecting patient pathways away from identified bottlenecks resulting from workforce shortages, increased demand for specific services or rigid legacy rule-based models of care. Applied to hospital operations, TOC can facilitate real-time decision-making and resource optimisation, addressing the limitations posed by retrospective data systems like DRGs.\u003c/p\u003e\u003cp\u003eDespite claims that EMR systems can enhance clinical care and support proactive resource planning,[\u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e], these benefits have yet to be fully realised in practice.[\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e, \u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e] Outcomes remain highly dependent on the operational characteristics and workflow alignment within each healthcare organisation. A disconnect remains between EMR system design and clinical workflows, resulting in limited visibility of system-wide operations and actual resource use.[\u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e, \u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e]\u003c/p\u003e\u003cp\u003eWe hypothesise that stratifying patients into flow streams based on care demands early in their admission, using routinely collected EMR data, could unlock potential to understand service demand in real-time and provide actionable intelligence to guide hospital-wide resource management and improve system efficiency.\u003c/p\u003e\u003cp\u003eThe aim of this study was to evaluate whether routinely collected data in the EMR could feasibly be used to classify admitted patients into cohorts based on resource use patterns during early admission.\u003c/p\u003e"},{"header":"METHODOLOGY","content":"\u003cp\u003eThe study is a retrospective analysis of EMR data of all patients admitted to a quaternary hospital in South Australia between January 2020 and January 2025. The study consists of three key components: data extraction, creation of new patient flow streams, and descriptive analysis. This study is reported in accordance with the RECORD (Reporting of studies Conducted using Observational Routinely-collected health Data) guidelines.[10]\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eData\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cem\u003eEthics\u003c/em\u003e\u003c/p\u003e\n\u003cp\u003eApproval was granted by the Central Adelaide Local Health Network (CALHN) Human Research Ethics Review Committee (2024/HRE00225), with a waiver of individual consent due to the retrospective nature and scale of the cohort. Access to EMR data was authorised by the data custodian, and data handling complied with relevant governance and privacy legislation.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cem\u003eData source, extraction and internal linkage to capture encounters\u003c/em\u003e\u003c/p\u003e\n\u003cp\u003eData were extracted from the hospital\u0026rsquo;s fully integrated Altera Digital Health Sunrise EMR system, using the SA Health Data Analytics Platform, an integrated data infrastructure capable of large-scale interrogation and extraction from the EMR. To accurately represent complete admission episodes, internal deterministic linkage was performed using medical record numbers (MRNs) and time-stamped metadata to consolidate care events (e.g., emergency department visits and inpatient stays) into a unified patient admission journey for each presentation.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eA time-based threshold associated related care events. MRNs were subsequently mapped to a study-specific unique patient identifier in a separate, secure mapping file maintained outside the analytical dataset. This approach enabled the dataset to remain fully de-identified during analysis, while preserving the capacity for controlled re-identification. Manual review was conducted for a small number of cases with documentation inconsistencies. Linkage accuracy was verified through sample checks and cross-validation with official admission counts from the health network\u0026rsquo;s Business Intelligence Unit.\u003c/p\u003e\n\u003cp\u003e\u003cem\u003eData variables \u0026nbsp;\u003c/em\u003e\u003c/p\u003e\n\u003cp\u003eNo sampling or exclusion criteria were applied to the admission cohort in this study, to capture the complex operational reality of diverse, dynamic, and evolving nature of hospital resource demands. This comprehensive approach ensures that infrequent but high-impact events, such as seasonal demand surges, unplanned service disruptions, or escalation responses are represented in the dataset. Such inclusivity enables a more accurate understanding of system pressures across routine and exceptional conditions. For example, instances where a \u0026ldquo;Code White\u0026rdquo; is declared, signalling severe and sustained access block under the CALHN Demand Escalation Framework (unpublished internal document, Central Adelaide Local Health Network, 2025), can now be examined in relation to real-time admission volumes, resource consumption patterns, and flow stream distribution.\u003c/p\u003e\n\u003cp\u003eA comprehensive range of variables were extracted from the EMR including demographic details, clinical information such as comorbidities, diagnostics (pathology and imaging) and treatments (medications, procedures, operations, and therapies). Risk assessment data were captured using validated tools for skin integrity,[11] cognition,[12] malnutrition,[13] and mobility.[14] In addition, resource utilisation metrics were collected, including allied health and pharmacy input, specialty consultations, hospital length of stay, one to one nursing care, intra-hospital care transfers, ward locations, clinical coding data (Table 1). To explore the relationship between evolving care needs and classification stability, a convenience sample of cases with clearly observed changes in flow stream allocation were reviewed to assess corresponding shifts in patients care requirements and resource utilisation.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eTable 1. Summary of Variables Extracted from the Electronic Medical Record (EMR)\u003c/strong\u003e\u003c/p\u003e\n\u003ctable border=\"1\" cellspacing=\"0\" cellpadding=\"0\"\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 208px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eVariable\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 350px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eDescription\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 208px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eDemographic Details\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 350px;\"\u003e\n \u003cp\u003eAge\u003c/p\u003e\n \u003cp\u003eGender\u003c/p\u003e\n \u003cp\u003ePostcode\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 208px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eClinical Information\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 350px;\"\u003e\n \u003cp\u003eComorbidities\u003c/p\u003e\n \u003cp\u003eTreatment received \u0026nbsp;\u003c/p\u003e\n \u003cp\u003eClinical course\u003c/p\u003e\n \u003cp\u003eDischarge disposition\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 208px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eRisk Assessments\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 350px;\"\u003e\n \u003cp\u003eValidated tools for skin integrity (Braden Score[11]), cognition (4AT[12]), malnutrition (MUST[13]), and mobility[14]).\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 208px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eDiagnostic Information\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 350px;\"\u003e\n \u003cp\u003eBlood Tests \u0026nbsp;\u003c/p\u003e\n \u003cp\u003eCT and MR orders\u003c/p\u003e\n \u003cp\u003eInterventional Radiology orders\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 208px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eResource Utilisation Metrics\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 350px;\"\u003e\n \u003cp\u003eAllied health* and pharmacy consultation\u003c/p\u003e\n \u003cp\u003eSub-specialty consultations\u003c/p\u003e\n \u003cp\u003eHospital length of stay\u003c/p\u003e\n \u003cp\u003eOne to one nursing care\u003c/p\u003e\n \u003cp\u003eIntra-hospital transfers\u003c/p\u003e\n \u003cp\u003eWard type (home vs. outlier)\u003c/p\u003e\n \u003cp\u003eICU admission\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 208px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eCoding Data\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 350px;\"\u003e\n \u003cp\u003eInternational Classification of Diseases (ICD)-10 codes**\u003c/p\u003e\n \u003cp\u003eDRG Procedure codes [15]\u003c/p\u003e\n \u003cp\u003eDRG Severity codes [15]\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n\u003c/table\u003e\n\u003cp\u003e\u003cem\u003e*Allied Health Professional (AHP) considered as Physiotherapy, Occupational Therapy, Speech Pathology, Dietetics and Social Work.\u003cbr\u003e\u0026nbsp;** ICD-10 is the 10th revision of the International Classification of Diseases, a medical classification list by the World Health Organization.[16]\u003c/em\u003e\u003c/p\u003e\n\u003cp\u003e\u003cem\u003eData de-identification and cleaning\u003c/em\u003e\u003c/p\u003e\n\u003cp\u003eAll patient data were de-identified using a unique study identifier. The dataset was cleaned to ensure consistency, including removal of duplicate entries, standardising formats (e.g., date/time fields), correction of errors, and exclusion of clinically implausible outliers (e.g., negative timestamped data). Missing data were addressed pragmatically: if a variable was not recorded in structured EMR fields, it was assumed not collected. This assumption reflects routine clinical documentation practices and the study\u0026rsquo;s focus on operational visibility from real-time data. However, this may underestimate some clinical characteristics or care events captured only in free text or omitted due to documentation variability. \u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cem\u003eData Quality Assessment\u0026nbsp;\u003c/em\u003e\u003c/p\u003e\n\u003cp\u003eData quality was assessed to evaluate the completeness, consistency, and operational reliability of key variables prior to analysis. This included verifying demographic information, inconsistencies in service use documentation, and limitations in patient journey mapping across care transitions. Particular attention was given to data elements central to modelling patient flow and resource utilisation such as time-stamped transitions of care, moves to critical service delivery areas for urgent interventions and diagnostics. Front-end clinical documentation was compared with analytics extracts to identify discrepancies and assess the extent to which informal workflows (often recorded in unstructured free-text fields) were missing from the structured EMR data that is used for analysis and planning. These insights informed downstream modelling decisions and guided interpretation of care delivery patterns.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eCreation of new patient flow streams \u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eA novel classification framework was developed to stratify patients into four distinct flow streams based on resource use. The initial framework was established through a modified Delphi Technique, involving structured consensus from a multidisciplinary panel of expert clinicians with operational and frontline insight. This iterative, expert informed approach enabled the development of clinically meaningful flow streams grounded in real world care delivery. Stream definitions were subsequently refined throughout the analysis to better reflect observed variations in patient care profiles resulting in four discrete resource-utilisation categories as shown in Table 2.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eTable 2. Flow Stream Definitions and Classification Criteria\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003ctable border=\"0\" cellspacing=\"0\" cellpadding=\"0\" width=\"109%\"\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 8px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eFlow Stream\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 15px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eDefinition\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 29px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eCriteria\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 22px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eOperational Considerations\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 23px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eOpportunities for Future Improvement\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 8px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e1\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n \u003cp\u003e\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 15px;\"\u003e\n \u003cp\u003e\u003cem\u003eShort-stay or ambulatory care patients\u003c/em\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 29px;\"\u003e\n \u003cul\u003e\n \u003cli\u003eDay Procedure or Treatment\u003c/li\u003e\n \u003cli\u003eLOS \u0026lt;24 hours\u003c/li\u003e\n \u003c/ul\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 22px;\"\u003e\n \u003cp\u003eOperational interest when expected discharge is delayed or procedural outcomes are not achieved.\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 23px;\"\u003e\n \u003cp\u003eRefine patient selection protocols to maximise efficiency and minimise avoidable overnight stays.\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 8px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e2\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n \u003cp\u003e\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 15px;\"\u003e\n \u003cp\u003e\u003cem\u003eAdmissions for a primary condition under a single clinical specialty\u003c/em\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 29px;\"\u003e\n \u003cul\u003e\n \u003cli\u003e\u0026le;1 medical consult\u003c/li\u003e\n \u003cli\u003eLength of stay \u0026lt; 5 days\u003c/li\u003e\n \u003c/ul\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 22px;\"\u003e\n \u003cp\u003eWell-suited to care standardisation through clinical pathways.\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 23px;\"\u003e\n \u003cp\u003eOptimise clinical workflow and care timeliness through pathway adherence and streamlined delivery.\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 8px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e3\u003c/strong\u003e\u003c/p\u003e\n \u003cp\u003e\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 15px;\"\u003e\n \u003cp\u003e\u003cem\u003ePatients of intermediate complexity requiring multidisciplinary input\u003c/em\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 29px;\"\u003e\n \u003cul\u003e\n \u003cli\u003eDo not meet criteria for FS1, FS2, or FS4\u003c/li\u003e\n \u003c/ul\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 22px;\"\u003e\n \u003cp\u003eBenefit from early senior clinical decision-making to determine care pathways.\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 23px;\"\u003e\n \u003cp\u003eImplement targeted risk stratification and escalation triggers to reduce care delays and prevent deterioration.\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 8px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e4\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n \u003cp\u003e\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 15px;\"\u003e\n \u003cp\u003e\u003cem\u003eHypercomplex, high-acuity admissions requiring intensive resource mobilisation\u003c/em\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 29px;\"\u003e\n \u003cul\u003e\n \u003cli\u003eICU admission \u0026ge; 24 hours\u003c/li\u003e\n \u003c/ul\u003e\n \u003cp\u003e\u003cstrong\u003e\u003cu\u003eOR\u003c/u\u003e\u003c/strong\u003e\u003c/p\u003e\n \u003cul\u003e\n \u003cli\u003eCritical Location \u0026ndash; ED Resus, Theatre, ICU, Spinal\u003c/li\u003e\n \u003cli\u003eEmergency surgery or theatre listed\u003c/li\u003e\n \u003cli\u003eTreatment Intensity (CT or MR) and \u0026ge; 5 bloods\u003c/li\u003e\n \u003cli\u003eGroup + Hold ordered\u003c/li\u003e\n \u003cli\u003eWithin 24hours:\u003c/li\u003e\n \u003cli\u003e\u0026ge;3 specialist consults\u003c/li\u003e\n \u003cli\u003e\u0026ge;4 intra-hospital transfers\u003c/li\u003e\n \u003c/ul\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 22px;\"\u003e\n \u003cp\u003eBenefit from proactive care coordination to manage complexity and mitigate system strain.\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 23px;\"\u003e\n \u003cp\u003eDesign robust models of care with embedded coordination, workforce agility, and real-time monitoring to manage surge and mitigate systemic disruption.\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n\u003c/table\u003e\n\u003cp\u003e\u003cstrong\u003eNote:\u003c/strong\u003e Some criteria (e.g. length of stay \u0026lt; 5 days are assessed dynamically during admission. Patients may be reclassified as their care evolves. For instance, a patient initially meeting Stream 2 criteria may transition to Stream 3 or 4, if their admission exceeds five days or requires additional resources. This dynamic assignment supports real-time planning and operational responsiveness.\u003c/p\u003e\n\u003cp\u003eTo minimise misclassification and reduce potential confounding, a structured hierarchical allocation was used to classify patients into flow streams based on observed care delivery patterns. Each admission was retrospectively assessed for eligibility beginning with Flow Stream 1 (FS1), which reflects short-stay, low-resource admissions. If criteria were not met, patients were sequentially evaluated for assignment to Flow Stream 4 (FS4), 2 (FS2), and finally 3 (FS3), based on observed resource intensity and care characteristics. This approach ensured assignment to the highest appropriate resource-intensity stream and prioritised early identification of high-demand cases for operational planning. While this study focused on retrospective classification using complete admission data, the same logic may be applied at key early points in the patient journey (e.g. 24 or 72 hours) to predict likely stream assignment based on partial data. Patients who did not meet defined criteria for Streams 1, 2, or 4 were allocated to FS3, a heterogeneous group retained for future refinement through machine learning-based classification models.\u003c/p\u003e\n\u003cp\u003e\u003cem\u003eClinical Validation of Flow Stream Allocation\u003c/em\u003e\u003c/p\u003e\n\u003cp\u003eClinical validation was undertaken on a stratified sample across all flow streams, using full EMR records accessed through the clinical interface. Classification accuracy was assessed by comparing assigned stream categories against actual care pathways and resource use. Discrepant cases were reviewed in multidisciplinary consensus discussions, and findings were used to refine stream definitions and allocation logic. Full methodology and examples are provided in Appendix 1.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eStatistical analysis\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eDescriptive statistics examined hospital resource utilisation trends at three time points: 24 hours, 72 hours, and at discharge. This enabled assessment of transitions between flow streams or shifts in resource-use profiles over the course of an admission.\u003c/p\u003e\n\u003cp\u003eSummary measures included medians and interquartile ranges of key resource variables, as these are more appropriate than means for data with skewed distributions and operational outliers, such as length of stay and consult counts. Comparative analysis across flow streams was undertaken to evaluate the alignment between anticipated and actual resource intensity, informing iterative refinement of stream definitions and identification of potentially misclassified or transitional cases. \u0026nbsp;For example, the original FS2 criteria limited patients to a single sub-specialty consult; however, analysis identified a subset of short-stay patients who received a second consult from their usual treating team to support continuity of care. In response, the classification rule was modified to allow up to two consults in FS2.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eStatistical analyses were performed using R, version 4.4.3 and STATA, version 18.5. The summary statistics in Table 3 were generated with the R package, gtsummary.[17]\u003c/p\u003e"},{"header":"RESULTS","content":"\u003cp\u003eBetween January 2020 and January 2025, a total of 332,711 inpatient admissions were recorded. The distribution of these admissions by source and final flow stream classification is shown in Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e. Descriptive characteristics, including age, gender, medication burden, and list of investigations, were analysed and stratified across flow streams to identify differences in patient complexity and care patterns. A high-level summary is presented in Table\u0026nbsp;\u003cspan refid=\"Tab3\" class=\"InternalRef\"\u003e3\u003c/span\u003e, with the full set of descriptive statistics provided in Appendix 2. Patterns of hospital length of stay by flow stream, also shown in Table\u0026nbsp;\u003cspan refid=\"Tab3\" class=\"InternalRef\"\u003e3\u003c/span\u003e, are visualised in Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003e to highlight distributional differences. These variables were used to inform flow stream classification and support early assessment of hospital resource distribution.\u003c/p\u003e\u003cp\u003eAll findings presented in this section are descriptive and unadjusted, consistent with the exploratory objectives of this study\u0026rsquo;s initial phase. No statistical modelling or hypothesis testing was performed, as the focus was on classification accuracy and resource profiling.\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003cp\u003e\u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab3\" border=\"1\"\u003e\u003ccaption language=\"En\"\u003e\u003cdiv class=\"CaptionNumber\"\u003eTable 3\u003c/div\u003e\u003cdiv class=\"CaptionContent\"\u003e\u003cp\u003eSummary Characteristics of the Total Study Population (N\u0026thinsp;=\u0026thinsp;332,711), Stratified by Flow Stream\u003c/p\u003e\u003c/div\u003e\u003c/caption\u003e\u003ccolgroup cols=\"5\"\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e\u003cthead\u003e\u003ctr\u003e\u003cth align=\"left\" colname=\"c1\"\u003e\u003cp\u003eVariable\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c2\"\u003e\u003cp\u003eStream 1 \u003c/p\u003e\u003cp\u003eN\u0026thinsp;=\u0026thinsp;149,682\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c3\"\u003e\u003cp\u003eStream 2 \u003c/p\u003e\u003cp\u003eN\u0026thinsp;=\u0026thinsp;108,840\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c4\"\u003e\u003cp\u003eStream 3 \u003c/p\u003e\u003cp\u003eN\u0026thinsp;=\u0026thinsp;61,936\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c5\"\u003e\u003cp\u003eStream 4 \u003c/p\u003e\u003cp\u003eN\u0026thinsp;=\u0026thinsp;12,253\u003c/p\u003e\u003c/th\u003e\u003c/tr\u003e\u003c/thead\u003e\u003ctbody\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e\u003cb\u003eAge, years\u003c/b\u003e\u003csup\u003e\u003cb\u003e1\u003c/b\u003e\u003c/sup\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e61 (48, 72)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e62 (44, 77)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e70 (56, 81)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e62 (49, 73)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e\u003cb\u003eFemale, n (%)\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e59,799 (40.0)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e47,750 (44.0)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e27,548 (44.0)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e4,636 (38.0)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e\u003cb\u003eLength of Stay, days\u003c/b\u003e\u003csup\u003e1\u003c/sup\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e0.2 (0.2, 0.3)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e2.0 (1.1, 3.1)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e8.4 (6.2, 13.3)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e11.2 (6.4, 20.3)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e\u003cb\u003eAdmission Type\u003c/b\u003e\u003csup\u003e\u003cb\u003e2\u003c/b\u003e\u003c/sup\u003e, \u003cb\u003en (%)\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eMedical\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e108,437 (72.4)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e53,338 (49.0)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e37,144 (60.0)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e5,367 (43.8)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eSurgical\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e28,027 (18.7)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e54,471 (50.0)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e24,607 (39.7)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e6,877 (56.2)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eOther\u003csup\u003e3\u003c/sup\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e13,218 (8.8)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e1,031 (0.9)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e185 (0.3)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e9 (\u0026lt;\u0026thinsp;0.1)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003c/tbody\u003e\u003c/colgroup\u003e\u003c/table\u003e\u003c/div\u003e\u003c/p\u003e\u003cp\u003e\u003csup\u003e1\u003c/sup\u003e Median (interquartile range).\u003c/p\u003e\u003cp\u003e\u003csup\u003e2\u003c/sup\u003eAdmission Type \u0026ndash; Mental Health admissions were excluded from this analysis.\u003c/p\u003e\u003cp\u003e\u003csup\u003e3\u003c/sup\u003eOther: Emergency, Hyperbaric, Gynaecology\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003cdiv id=\"Sec13\" class=\"Section2\"\u003e\u003ch2\u003eData Quality Assessment\u003c/h2\u003e\u003cp\u003eEvaluation of EMR data quality revealed substantial limitations in the completeness and standardisation of critical data fields. While admission note completion rates were high within 72 hours, structured fields were often bypassed in favour of free-text documentation, limiting downstream extractability. High variability in documentation practises was noted across clinical teams and disciplines, contributing to inconsistent data capture and reducing the reliability of certain structured fields for operational analysis. Mandatory fields were frequently populated with inaccurate or generic responses, for example, patients admitted from residential aged care were commonly recorded as arriving from private residences.\u003c/p\u003e\u003cp\u003eStructured problem lists, intended to support longitudinal care continuity, were populated in less than 75% of cases. High-acuity events such as the massive transfusion pack (MTP) activation were significantly underreported in the EMR compared to external validation sources (e.g., 144 EMR entries in total cohort versus 355 blood bank activations in a single year). Consult orders were inconsistently used across clinical teams, particularly in urgent settings where informal referrals and verbal communication were preferred over electronic pathways. In some cases, consults were entered by the admitting team for their own service as part of the admission process, which distorted flow stream allocation. Due to inconsistencies in naming conventions between consult orders and admitting team identifiers, manual coding was necessary to accurately match and exclude these entries for valid classification (see Appendix 2 for variable-level missing data).\u003c/p\u003e\u003cp\u003eNo linkage failures were identified in the final dataset. Manual validation of a subset of patient records confirmed concordance between administrative admission identifiers and consolidated care episodes, supporting the integrity of the linkage process.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec14\" class=\"Section2\"\u003e\u003ch2\u003eFlow Stream Assignment\u003c/h2\u003e\u003cp\u003eAll eligible admissions were classified into one of four resource-based flow streams using the allocation framework. The results are summarised in Table\u0026nbsp;\u003cspan refid=\"Tab4\" class=\"InternalRef\"\u003e4\u003c/span\u003e, comparing the initial projected distribution of patients across streams with actual classifications at 24 hours, 72 hours and discharge, allowing assessment of classification stability and operational alignment with early admission profiles. As the model reflects resource utilisation, patient allocation may shift during admission in response to changes in care intensity or clinical deterioration, a characteristic designed to support real-time operational insight, as visualised in Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003e.\u003c/p\u003e\u003cp\u003e\u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab4\" border=\"1\"\u003e\u003ccaption language=\"En\"\u003e\u003cdiv class=\"CaptionNumber\"\u003eTable 4\u003c/div\u003e\u003cdiv class=\"CaptionContent\"\u003e\u003cp\u003eFlow Stream Classification of Admissions at 24 Hours, 72 Hours, and Discharge (N\u0026thinsp;=\u0026thinsp;332,711)\u003c/p\u003e\u003c/div\u003e\u003c/caption\u003e\u003ccolgroup cols=\"4\"\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e\u003cthead\u003e\u003ctr\u003e\u003cth align=\"left\" colname=\"c1\"\u003e\u003cp\u003eFlow Stream\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c2\"\u003e\u003cp\u003e24 hours, n (%)\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c3\"\u003e\u003cp\u003e72 hours, n (%)\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c4\"\u003e\u003cp\u003eDischarge, n (%)\u003c/p\u003e\u003c/th\u003e\u003c/tr\u003e\u003c/thead\u003e\u003ctbody\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e\u003cb\u003eFS1\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e149,682 (45.0)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e149,682 (45.0)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e149,682 (45.0)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e\u003cb\u003eFS2\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e111,339 (33.5)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e108,991 (32.8)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e108,840 (32.7)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e\u003cb\u003eFS3\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e70,708 (21.3)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e64,435 (19.4)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e61,936 (18.6)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e\u003cb\u003eFS4\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e982 (0.3)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e9,603 (2.9)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e12,253 (3.7)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003c/tbody\u003e\u003c/colgroup\u003e\u003ctfoot\u003e\u003ctr\u003e\u003ctd colspan=\"4\"\u003e\u003cb\u003eNote\u003c/b\u003e: Percentages are based on total admissions (N\u0026thinsp;=\u0026thinsp;332,711) at each time point. Values are rounded to one decimal place for clarity.\u003c/td\u003e\u003c/tr\u003e\u003c/tfoot\u003e\u003c/table\u003e\u003c/div\u003e\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003cp\u003eFewer than 5% of patients changed flow stream assignment during admission. These changes predominantly involved transitions from FS3 to FS4 as clinical complexity emerged, and from FS2 to FS4 in cases requiring escalation to higher levels of care. This pattern is evident in the transition flow visualised in Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003e, where movement from FS2 and FS3 to FS4 accounts for most reclassifications. For example, a patient initially classified as FS2 due to a stable embolic stroke presentation was later reclassified as FS4 following rapid clinical deterioration, requiring ICU admission and neurosurgical intervention.\u003c/p\u003e\u003cp\u003eThis low reallocation rate suggests strong initial model stability and highlights stream transitions are not frequent, instead representing meaningful changes in patient condition or care requirements. Establishing baseline ratios of the flow stream is critical to evaluating responses to external and internal system characteristics over time, to allow changes to resource reallocation or other strategic interventions.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec15\" class=\"Section2\"\u003e\u003ch2\u003eOperational Utility of Flow Stream Classification\u003c/h2\u003e\u003cp\u003eInitial application of the classification model revealed clear differentiation in resource-use profiles across streams. FS1 patients followed short-stay, low-resource pathways, with minimal investigations or consults, a median LOS of 0.2 days, rare ICU admission (3.1%) and polypharmacy in 0.6% of cases. FS2 admissions remained within specialty teams with relatively short lengths of stay (median 2.0 days) and focused care delivery, with low Allied Health input (median 0 (0,1)). FS3 captured patients of intermediate complexity shown by median LOS of 8.4 days, 21% requiring multidisciplinary input from medical and allied health. Polypharmacy was high (93%) and ICU use was greater than FS1 and FS2. FS4 admissions represented high acuity care: 96% were admitted to ICU, 36% had multiple specialise consultations and the median length of stay was 11.2 days.\u003c/p\u003e\u003c/div\u003e"},{"header":"DISCUSSION","content":"\u003cp\u003eThis retrospective observational study evaluated the operational utility of a novel, resource-based patient classification model using routinely collected electronic medical record (EMR) data. The study aimed to determine whether structured EMR variables could reliably stratify patients into flow streams reflective of care complexity and resource use to inform hospital-wide resource allocation. The analysis included inpatient admissions (n\u0026thinsp;=\u0026thinsp;332,711) to a quaternary hospital in South Australia between January 2020 and January 2025. While the model was developed in an Australian hospital, the operational challenges addressed, such as managing resource-intensive cohorts, identifying system bottlenecks, and supporting real-time planning, are shared across many hospital systems globally. The framework is designed to be adaptable to local infrastructure, EMR capabilities, and policy environments.\u003c/p\u003e\u003cp\u003eA comprehensive dataset was constructed through deterministic linkage and rigorous data cleaning, incorporating demographic, clinical, diagnostic, and operational variables. Patients were stratified into one of four flow streams based on resource utilisation rather than diagnostic grouping. Despite limitations in structured EMR documentation, particularly variability in consult orders and data entry, the model demonstrated early utility. Flow streams captured distinct care complexity and service utilisation patterns, which remained stable over the inpatient stay.[\u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e] Fewer than 5% of patients changed stream assignment during admission, suggesting allocation stability and that transitions likely reflect meaningful clinical changes.\u003c/p\u003e\u003cp\u003eBy grounding classification in observed resource consumption, this model enables dynamic, real-time operational planning. In line with Theory of Constraints (TOC) principles, it supports early identification of operational bottlenecks and enhances system responsiveness for staffing, bed management, and escalation planning, before broader strain emerges. While initial results demonstrated differentiation between streams, the broader value lies in its potential to inform continuous operational insight and pre-emptive decision making across the hospital, while avoiding the unintended consequence of directing additional workload toward already strained services.\u003c/p\u003e\u003cp\u003eMany hospital performance constraints are internally generated, arising from how care delivery is structured, prioritised, and resourced. High resource patients in FS4, for instance, exert disproportionate demand on critical care infrastructure, which can delay care delivery and limit access for lower-acuity patients in FS1, FS2 and FS3. This cascading effect highlights the need to monitor high-acuity cohorts to mitigate system-wide impacts. TOC asserts that inefficiencies stem from unidentified or unmanaged constraints, which lead to cascading delays, access blocks, and disproportionate demands on high-acuity services. These constraints are not static and can be mitigated through targeted redesign and strategic resource realignment. The flow stream model offers a mechanism to identify operational pressure points by stratifying patients according to resource demand, supporting both direct benefits, such as early identification of capacity limits, and indirect benefits, including improved patient flow for other cohorts. These dynamics are especially relevant during periods of high occupancy, where marginal increases in volume can lead to disproportionate delays, reduced flexibility and compromised overall system efficiency. By anticipating pressure points and supporting proactive resource alignment, the model may help sustain performance under demand stress. Without such visibility, health services risk reinforcing inefficiencies by directing demand to constrained areas, inadvertently exacerbating stress where there is least capacity. Embedding this model into real-time operational planning may improve system resilience and performance at scale.\u003c/p\u003e\u003cp\u003eA critical next step is evaluating how the distributions shift over time in response to internal or external pressures. For example, a rise in FS4 admissions may signal emerging strain. These patients often bypass routine scheduling, displacing lower-acuity patients and contributing to delays, blocked beds, and inefficiencies across the hospital. Monitoring these shifts in near real-time could support decision-makers in enacting timely interventions, reallocating resources, or adjusting discharge pathways before system-wide dysfunction occurs.\u003c/p\u003e\u003cp\u003eAlthough EMRs are structured to standardise data capture and support visibility, the reality of frontline practice often diverges. Critical workflows remain undocumented due to time pressures, perceived irrelevance of fields, or the convenience of informal workarounds. Clinicians bypass digital systems in favour of direct communication (e.g., phone calls) or offline tools (e.g., spreadsheets), which, while efficient in context, obscure visibility at the organisational level. This gap between system design and care delivery impairs the ability to learn from frontline adaptations,[\u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e18\u003c/span\u003e] which carries significant operational risks. Planning, funding, and workforce forecasting frameworks increasingly rely on EMR-derived data, yet these decisions are compromised by inconsistencies in data entry, underutilisation of structured fields, and underreporting of resource-intensive activity.[\u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e19\u003c/span\u003e] While technically available, structured fields are often bypassed due to poor integration with clinical workflows and a lack of demonstrated value for accurate documentation, to provide system-level insight and responsiveness.[\u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e19\u003c/span\u003e] The absence of reliable visibility at scale impairs the organisation\u0026rsquo;s ability to anticipate demand, respond to emergent complexity, measure care delivery and allocate resources intelligently. To support adoption, decision tools must be embedded into clinicians\u0026rsquo; natural workflow rather than additional external overlays. Operationally relevant nudges, and prioritisation prompts, could be integrated directly into the EMR to drive meaningful use.[\u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e20\u003c/span\u003e]\u003c/p\u003e\u003cp\u003eThese challenges also intersect with important policy and ethical considerations. As the study progresses toward predictive analytics and machine learning applications, the ethical integration of such models into clinical and operational practice must be prioritised. Attention to bias mitigation, model transparency[\u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e21\u003c/span\u003e] and stakeholder engagement (including clinicians, executives and consumers), will be essential in guiding methodological refinement and governance.\u003c/p\u003e\u003cp\u003eThis framework offers a structured approach to real-time operational stratification. By monitoring complexity and care intensity, the model supports an adaptive and data-driven approach to hospital-wide resource management. The flow stream model developed will serve as the foundation for machine learning-based predictive analytics in the next phase, aimed at real-time decision making and surge response. Future phases could identify delay points, such as prolonged ED stays or diagnostic bottlenecks, and support earlier escalation, targeted interventions, and more responsive resource alignment.\u003c/p\u003e\u003cdiv id=\"Sec17\" class=\"Section2\"\u003e\u003ch2\u003eLIMITATIONS\u003c/h2\u003e\u003cp\u003eThis study has several limitations. First, it is based on retrospective EMR data, which is inherently subject to constraints in completeness, accuracy, and consistency of historical documentation. Where structured data were missing, it was assumed that the information had not been recorded during the admission. No imputation was used to preserve record integrity and reflect real-world availability. While aligned with operational practice, this may have underestimated the frequency of some resource utilisation events, particularly those captured only in free-text notes or omitted entirely due to documentation variability. These limitations are well-recognised in real-world data, where incomplete and heterogeneous data collection can limit interpretability, to reliably analyse, contextualise and apply to operational decision making.[\u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e22\u003c/span\u003e]\u003c/p\u003e\u003cp\u003eSecond, the analysis was conducted at a single quaternary hospital and may not be generalisable to settings with different infrastructure, case mix, staffing or EMR capability. Third, some components of care remain inconsistently documented in structured EMR fields, particularly in high-acuity or multidisciplinary environments. Verbal referrals, Medical Emergency Team (MET) calls, and non-standardised documentation of specialist consult input were frequently observed in unstructured notes, leading to an underestimation of resource use and complexity.\u003c/p\u003e\u003cp\u003eAdditionally, a structural limitation of the EMR system relates to demographic fields that are stored at the client level rather than at the individual admission level. Variables such as usual accommodation and postcode, are retrospectively overwritten across prior admissions if updated during a later encounter. For example, if a patient was admitted from a private residence early in the study period but subsequently moved to a residential aged care facility, all historical admissions may be updated to reflect the latter. While this has limited impact on real-time model development, it introduces a source of temporal inaccuracy for retrospective review and flow stream assignment. In the absence of audit trail access for these fields, this limitation reflects a broader challenge in the underlying EMR data architecture.\u003c/p\u003e\u003cp\u003eFinally, this study presents descriptive findings only. While the model demonstrates early operational value, it does not yet incorporate real-time analytics or predictive capabilities. These will be developed and evaluated in subsequent phases of the study.\u003c/p\u003e\u003c/div\u003e"},{"header":"CONCLUSION","content":"\u003cp\u003eThis study introduces a novel, resource-based flow stream framework to better classify patients by resource use rather than diagnosis. The model has operational relevance for hospital planning and performance monitoring but is constrained by EMR documentation gaps that limit visibility of care delivery. System inefficiencies are compounded by informal workflows and inconsistent structured data use, creating hidden constraints. From a TOC perspective, these bottlenecks restrict system throughput and remain unaddressed without real-time data insight.\u003c/p\u003e\u003cp\u003eFuture work must focus on strengthening real-time analytics, embedding decision support into EMR workflows, and developing predictive tools grounded in clinical and operational practice. This study forms the foundation for a multi-year evaluation of EMR adaptability, with the next phase focused on near real-time identification of bottlenecks and AI-driven forecasting to support smarter resource allocation. Aligning data with action is essential to building a learning health system that enables agile, proactive hospital operations.\u003c/p\u003e"},{"header":"Abbreviations","content":"\u003cp\u003e4AT\u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp;\u0026nbsp;4 'A's Test (Delirium Screening Tool)\u003c/p\u003e\n\u003cp\u003eAHP\u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp;Allied Health Professional\u003c/p\u003e\n\u003cp\u003eAI\u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp;Artificial Intelligence\u003c/p\u003e\n\u003cp\u003eCALHN\u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp;Central Adelaide Local Health Network\u003c/p\u003e\n\u003cp\u003eCT\u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp;\u0026nbsp;Computed Tomography\u003c/p\u003e\n\u003cp\u003eDRG\u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp;Diagnosis-Related Group\u003c/p\u003e\n\u003cp\u003eED\u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp;\u0026nbsp;Emergency Department\u003c/p\u003e\n\u003cp\u003eEMR\u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp;\u0026nbsp;Electronic Medical Record\u003c/p\u003e\n\u003cp\u003eFS\u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp;Flow Stream\u003c/p\u003e\n\u003cp\u003eHREC\u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp;Human Research Ethics Committee\u003c/p\u003e\n\u003cp\u003eICD-10\u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp;International Classification of Diseases, 10th Revision\u003c/p\u003e\n\u003cp\u003eICU\u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp;Intensive Care Unit\u003c/p\u003e\n\u003cp\u003eLOS\u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp;\u0026nbsp;Length of Stay\u003c/p\u003e\n\u003cp\u003eMDC\u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp;\u0026nbsp;Major Diagnostic Category\u003c/p\u003e\n\u003cp\u003eMET\u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp;Medical Emergency Team\u003c/p\u003e\n\u003cp\u003eMR\u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp;Magnetic Resonance\u003c/p\u003e\n\u003cp\u003eMRN\u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp;\u0026nbsp;Medical Record Number\u003c/p\u003e\n\u003cp\u003eMTP\u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp;Massive Transfusion Pack\u003c/p\u003e\n\u003cp\u003eMUST\u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp;\u0026nbsp;Malnutrition Universal Screening Tool\u003c/p\u003e\n\u003cp\u003ePET\u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp;\u0026nbsp;Positron Emission Tomography\u003c/p\u003e\n\u003cp\u003eRAH\u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp;Royal Adelaide Hospital\u003c/p\u003e\n\u003cp\u003eRECORD\u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp;\u0026nbsp;REporting of studies Conducted using Observational Routinely collected Data\u003c/p\u003e\n\u003cp\u003eTOC \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp;Theory of Constraints\u003c/p\u003e"},{"header":"Declarations","content":"\u003cp\u003e\u003cstrong\u003eEthics approval and consent to participate\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThis study was conducted as part of a health service quality improvement initiative and was reviewed by the Central Adelaide Local Health Network (CALHN) Human Research Ethics Committee (HREC). The project was approved under the National Statement on Ethical Conduct in Human Research (reference number: 2024/HRE00225), which is consistent with the principles in the Declaration of Helsinki (2013). Consent to participate was waived due to the retrospective nature of the data and the use of de-identified records. Clinical trial number: not applicable.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eConsent for publication\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eNot applicable.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAvailability of data and materials\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe datasets generated and/or analysed during the current study are not publicly available due to institutional and legal data access restrictions. De-identified data may be made available upon reasonable request and subject to appropriate ethics and governance approvals from SA Health and the Central Adelaide Local Health Network.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eCompeting interests\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe authors declare that they have no competing interests\u003cstrong\u003e.\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eFunding\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThis study was supported by grant funding from The Hospital Research Foundation, to facilitate data extraction, analysis, and project coordination. The funder had no role in the study design, data interpretation, or manuscript preparation.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAuthors' contributions\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eGO and DL led the study conceptualisation, classification framework and design of the study. MW acquired and cleaned the EMR data. MB led the refinement of the classification framework, drafted the manuscript, and contributed to data interpretation and operational application. AC conducted all statistical analyses and supported interpretation of results. TB, BS and TL contributed to overall study design, review and interpretation of the findings and manuscript development and review. JK and SB provided guidance on model development. \u0026nbsp;GO provided overall project supervision and manuscript review. All authors read and approved the final manuscript.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAcknowledgements\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe authors wish to acknowledge Dr Ian Sturgess for his mentorship and guidance in hospital system improvement. His vision and dedication to advancing patient flow and operational performance were instrumental in shaping the conceptual foundations of this study. We gratefully dedicate this manuscript to him.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\n\u003cli\u003eAustralian Medical Association. \u003cem\u003ePublic hospitals in logjam as funding pressures grow\u003c/em\u003e. 2023. https://ama.com.au (accessed 19 Aug 2025).\u003c/li\u003e\n\u003cli\u003eDuckett S, Breadon P. \u003cem\u003eControlling costs and improving care: reforming the hospital funding system\u003c/em\u003e. Grattan Institute 2014.\u003c/li\u003e\n\u003cli\u003eJackson T, Michel JL, Roberts RF, et al. A classification of hospital admissions. \u003cem\u003eBMC Health Serv Res\u003c/em\u003e 2014;14:346.\u003c/li\u003e\n\u003cli\u003eAlmeida MA, Marinho MMO. Theory of constraints in healthcare: a systematic literature review. \u003cem\u003eInt J Qual Reliab Manag\u003c/em\u003e 2022;39(3):716\u0026ndash;37.\u003c/li\u003e\n\u003cli\u003eYadav S, Kumar R, Tran T, et al. Assessing the predictive and analytics capability of electronic medical records for operational planning. \u003cem\u003eJ Healthc Inform Res\u003c/em\u003e 2023;7(2):123\u0026ndash;35.\u003c/li\u003e\n\u003cli\u003eZozus MN, Stansbury DW, Raskin S, et al. Assessing data quality for healthcare systems. \u003cem\u003eeGEMs\u003c/em\u003e 2019;7(1):20.\u003c/li\u003e\n\u003cli\u003eReimer AP, Milinovich A, Madigan EA. Data quality assessment framework to assess electronic medical record data for use in research. \u003cem\u003eInt J Med Inform\u003c/em\u003e 2016;90:40\u0026ndash;7.\u003c/li\u003e\n\u003cli\u003eMcGinn CA, Gagnon MP, Shaw N, et al. Users\u0026rsquo; perspectives of key factors to implementing electronic health records in Canada. \u003cem\u003eJMIR Med Inform\u003c/em\u003e 2011;13(3):e73.\u003c/li\u003e\n\u003cli\u003eAjami S, Bagheri-Tadi T. Barriers for adopting electronic health records (EHRs) by physicians. \u003cem\u003eActa Inform Med\u003c/em\u003e 2013;21(2):129\u0026ndash;34.\u003c/li\u003e\n\u003cli\u003eBenchimol EL, Smeeth L, Guttmann A, et al. The reporting of studies conducted using observational routinely-collected health data (RECORD) statement. \u003cem\u003ePLoS Med\u003c/em\u003e 2015;12:e1001885.\u003c/li\u003e\n\u003cli\u003eBraden B, Bergstrom N. A conceptual schema for the study of the etiology of pressure sores. \u003cem\u003eRehabil Nurs\u003c/em\u003e 1987;12(1):8\u0026ndash;12.\u003c/li\u003e\n\u003cli\u003eBellelli G, Morandi A, Davis DHJ, et al. Validation of the 4AT, a new instrument for rapid delirium screening: a study in 234 hospitalised older people. \u003cem\u003eAge Ageing\u003c/em\u003e 2014;43(4):496\u0026ndash;502.\u003c/li\u003e\n\u003cli\u003eElia M. The \u0026apos;MUST\u0026apos; report: nutritional screening of adults: a multidisciplinary responsibility. Redditch, UK: BAPEN 2003.\u003c/li\u003e\n\u003cli\u003eSA Health. \u003cem\u003eFall injury and prevention clinical guideline: screening, assessment, care planning and discharge planning\u003c/em\u003e. Government of South Australia 2018. https://www.sahealth.sa.gov.au/... (accessed 19 Aug 2025).\u003c/li\u003e\n\u003cli\u003eIndependent Health and Aged Care Pricing Authority. AR-DRG Version 11.0. https://www.ihacpa.gov.au/resources/ar-drg-version-110 (accessed 27 Aug 2025).\u003c/li\u003e\n\u003cli\u003eWorld Health Organization. \u003cem\u003eInternational Classification of Diseases, 10th revision (ICD-10)\u003c/em\u003e. https://www.who.int/classifications/icd/en/.\u003c/li\u003e\n\u003cli\u003eSjoberg D, Whiting K, Curry M, et al. Reproducible summary tables with the gtsummary package. \u003cem\u003eR J\u003c/em\u003e 2021;13:570\u0026ndash;80. https://doi.org/10.32614/RJ-2021-053.\u003c/li\u003e\n\u003cli\u003eVerhagen MJ, de Vos MS, Sujan M, Hamming JF. The problem with making Safety-II work in healthcare\u003cem\u003e. BMJ Qual Saf.\u003c/em\u003e 2022;31(5):402\u0026ndash;408. doi:10.1136/bmjqs-2021-014396\u003c/li\u003e\n\u003cli\u003ede Groot K, de Bruijne M, Paans W, et al. Effective and feasible interventions to improve structured EHR data quality: a systematic review. \u003cem\u003eInt J Med Inform\u003c/em\u003e 2023;174:105050.\u003c/li\u003e\n\u003cli\u003eAlexiuk M, Ashcroft R, Pijl-Zieber E, et al. Clinical decision support tools in the electronic medical record. \u003cem\u003eKidney Int Rep\u003c/em\u003e 2023;8(9):929\u0026ndash;38.\u003c/li\u003e\n\u003cli\u003eSendak MP, D\u0026rsquo;Arcy J, Kashyap S, et al. Clinical implementation of predictive models embedded within electronic health records. \u003cem\u003eJ Am Med Inform Assoc\u003c/em\u003e 2020;27(4):639\u0026ndash;46.\u003c/li\u003e\n\u003cli\u003eWeiskopf NG, Weng C. Methods and dimensions of electronic health record data quality assessment: enabling reuse for clinical research. \u003cem\u003eJ Am Med Inform Assoc\u003c/em\u003e 2013;20(1):144\u0026ndash;51.\u003c/li\u003e\n\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":false,"highlight":"","institution":"","isAcceptedByJournal":true,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"bmc-health-services-research","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"bhsr","sideBox":"Learn more about [BMC Health Services Research](http://bmchealthservres.biomedcentral.com/)","snPcode":"","submissionUrl":"https://www.editorialmanager.com/BHSR/default.aspx","title":"BMC Health Services Research","twitterHandle":"BMC_series","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"em","reportingPortfolio":"BMC Series","inReviewEnabled":true,"inReviewRevisionsEnabled":true},"keywords":"Electronic Medical Records, Hospital Resource Management, Health Information Systems, Data Quality, Patient Flow","lastPublishedDoi":"10.21203/rs.3.rs-7577977/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-7577977/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003e\u003cb\u003eBackground\u003c/b\u003e\u003c/p\u003e\u003cp\u003eHospitals face increasing strain from rising clinical complexity and demand. Traditional resource allocation approaches often lack the granularity and timeliness needed for responsive planning. This study evaluates whether routinely collected electronic medical record (EMR) data can be used to classify hospital inpatients into resource-based groups to support real-time planning and hospital-wide operational management.\u003c/p\u003e\u003cp\u003e\u003cb\u003eMethods\u003c/b\u003e\u003c/p\u003e\u003cp\u003eA retrospective analysis was conducted on 332,711 inpatient admissions to a quaternary public hospital in South Australia between January 2020 and January 2025. Patients were classified into one of four flow streams within 72 hours of admission using a resource-based classification framework developed through a modified Delphi process and validated by clinical review. Summary statistics were used to assess differences in resource use across streams and to evaluate classification stability. Data quality limitations and documentation variability were also assessed.\u003c/p\u003e\u003cp\u003e\u003cb\u003eResults\u003c/b\u003e\u003c/p\u003e\u003cp\u003eFlow streams demonstrated distinct differences in length of stay, diagnostic testing, consultations, and allied health input. The model showed strong initial stability, with fewer than 5% of patients changing streams during admission. Key data quality issues included inconsistent consultation documentation, underuse of structured fields, and retrospective overwriting of demographic information, affecting visibility of resource use. Despite these limitations, flow stream classification effectively differentiated patients by resource intensity and care complexity, offering a practical framework to support real-time hospital operations, complementing diagnosis-based groupings.\u003c/p\u003e\u003cp\u003e\u003cb\u003eConclusion\u003c/b\u003e\u003c/p\u003e\u003cp\u003eA structured classification model using routinely collected EMR data can differentiate inpatient resource needs. Flow stream stratification offers a complementary approach to traditional coding-based systems and may help identify operational bottlenecks. With improved documentation and system integration, this approach could enhance hospital responsiveness, resource planning, and overall system performance.\u003c/p\u003e","manuscriptTitle":"Can routinely collected electronic medical record (EMR) data support hospital resource allocation? A retrospective analysis of 332,711 presentations to a public quaternary teaching hospital in South Australia (2020-2025)","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-10-03 01:07:36","doi":"10.21203/rs.3.rs-7577977/v1","editorialEvents":[{"type":"communityComments","content":0},{"type":"decision","content":"Revision requested","date":"2026-01-28T11:54:24+00:00","index":"","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2026-01-05T05:16:07+00:00","index":"hide","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2025-11-24T22:43:51+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"120210212359667412795960335993139812485","date":"2025-11-24T21:48:09+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"44347545153158729317805334883452446296","date":"2025-11-24T00:47:18+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"222517254718546759012626286950839574044","date":"2025-11-23T20:53:21+00:00","index":"hide","fulltext":""},{"type":"reviewersInvited","content":"","date":"2025-09-21T08:42:50+00:00","index":"","fulltext":""},{"type":"editorInvited","content":"","date":"2025-09-17T18:43:26+00:00","index":"","fulltext":""},{"type":"editorAssigned","content":"","date":"2025-09-16T04:56:06+00:00","index":"","fulltext":""},{"type":"checksComplete","content":"","date":"2025-09-16T04:54:16+00:00","index":"","fulltext":""},{"type":"submitted","content":"BMC Health Services Research","date":"2025-09-10T02:29:28+00:00","index":"","fulltext":""}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"bmc-health-services-research","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"bhsr","sideBox":"Learn more about [BMC Health Services Research](http://bmchealthservres.biomedcentral.com/)","snPcode":"","submissionUrl":"https://www.editorialmanager.com/BHSR/default.aspx","title":"BMC Health Services Research","twitterHandle":"BMC_series","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"em","reportingPortfolio":"BMC Series","inReviewEnabled":true,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"06539b8e-6387-4168-b811-31dccc6f8616","owner":[],"postedDate":"October 3rd, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"published-in-journal","subjectAreas":[],"tags":[],"updatedAt":"2026-03-30T16:29:46+00:00","versionOfRecord":{"articleIdentity":"rs-7577977","link":"https://doi.org/10.1186/s12913-026-14448-8","journal":{"identity":"bmc-health-services-research","isVorOnly":false,"title":"BMC Health Services Research"},"publishedOn":"2026-03-27 16:11:07","publishedOnDateReadable":"March 27th, 2026"},"versionCreatedAt":"2025-10-03 01:07:36","video":"","vorDoi":"10.1186/s12913-026-14448-8","vorDoiUrl":"https://doi.org/10.1186/s12913-026-14448-8","workflowStages":[]},"version":"v1","identity":"rs-7577977","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-7577977","identity":"rs-7577977","version":["v1"]},"buildId":"XKTyCvWXoU3ODBz1xrDgd","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}