A Reassessment of Sodium Correction Rates and Hospital Length of Stay Accounting for Admission Diagnosis

Background Recent studies have challenged assumptions about slow correction of severe hyponatremia and have shown that rapid correction is associated with shorter hospital length of stay. However, the confounding effect of admission diagnosis has not been fully explored. The

of this study was to determine whether rapid correction is still associated with shorter length of stay when controlling for admission diagnosis.

This retrospective cohort study is based on the Medical Information Mart for Intensive Care, including data from both MIMIC-III (2001-2012) and MIMIC-IV (2008-2019). Patients were identified who presented to the hospital with initial sodium <120 mEq/L and were categorized according to total sodium correction achieved in the first day (10 mEq/L). Linear regression was used to assess for an association between correction rate and hospital length of stay, and to determine if this association was significant when controlling for admission diagnosis classifications based on diagnosis related groups (DRGs).

There were 636 patients included in this study. Median [IQR] hospital length of stay was 7 [4, 11] days. Patients had a median [IQR] initial sodium value of 117 [114, 118] mEq/L and final sodium value of 124 [119, 128] mEq/L. In a univariate linear regression, the highest rate of correction (>10 mEq/L) was associated with a shorter length of stay than a moderate rate of All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprintthis version posted March 11, 2024. ; https://doi.org/10.1101/2024.03.08.24303993doi: medRxiv preprint 3 correction (coef. -2.363, 95% CI [-4.710, -0.017], p=0.048), but the association was not significant when controlling for admission diagnosis group (coef. -1.685, 95% CI [-3.836, 0.467], p=0.125).

Faster sodium correction was not associated with shorter length of stay when controlling for admission diagnosis categories, suggesting that the disease state confounds this association. While some patients may be discharged earlier if sodium is corrected more rapidly, others may not benefit or may be harmed by this strategy.

Hyponatremia, Sodium Correction, ODS All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprintthis version posted March 11, 2024. ; https://doi.org/10.1101/2024.03.08.24303993doi: medRxiv preprint 4 Abbreviations: CPM - Central Pontine Myelinolysis DRG - Diagnosis Related Group ICD - International Classification of Diseases ICU - Intensive Care Unit MIMIC - Medical Information Mart for Intensive Care ODS - Osmotic Demyelination Syndrome All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprintthis version posted March 11, 2024. ; https://doi.org/10.1101/2024.03.08.24303993doi: medRxiv preprint 5

Hyponatremia is a common primary or secondary hospital admission diagnosis. For patients with chronic hyponatremia, rapid correction of sodium levels is commonly thought to increase the risk of osmotic demyelination syndrome (ODS), previously known as central pontine myelinolysis (CPM), a potentially catastrophic neurological event. 1 As a result, longstanding recommendations have stressed slow correction of chronic hyponatremia, although exact guidelines have evolved over time.2 However, the risks and benefits of slow correction have recently been questioned, given the possibility of adverse effects of delaying normalization of sodium levels balanced against the rarity of ODS and debate about the connection between sodium correction rates and ODS.3 For example, one recent study showed that faster correction was associated with shorter hospital length of stay and lower mortality,4 and another showed that faster correction was associated with lower mortality and more hospital-free days and ICU-free days.5 While these findings could be practice-changing, it is important to recognize that the studies did not directly account for admission diagnosis, leaving open the potential for significant confounding. Other factors such as the competing risk of mortality, which also varies by admission diagnosis, need to be considered as well. The Medical Information Mart for Intensive Care (MIMIC), a deidentified critical care research database developed and maintained through a longstanding collaboration between the Beth Israel Deaconess Medical Center (BIDMC) in Boston, MA and the Massachusetts Institute of Technology, can be used to assess the impact of admission diagnosis by providing information on Diagnosis Related Groups (DRGs). DRGs classify admissions according to the primary All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprintthis version posted March 11, 2024. ; https://doi.org/10.1101/2024.03.08.24303993doi: medRxiv preprint 6 problem addressed and may be tied to hospital reimbursement, but may or may not overlap with chronic comorbidities that are more frequently used in retrospective analyses.6,7 In this study, we use these data to determine how admission diagnosis modifies the association between sodium correction rate and hospital length of stay in patients with severe hyponatremia, defined as serum sodium <120 mEq/L. 3

Research Ethics This retrospective cohort study is based on the Medical Information Mart for Intensive Care (MIMIC), including data from both the third (MIMIC-III, years 2001-2012)8 and fourth (MIMIC-IV, data from years 2008-2019)9 iterations. MIMIC was approved for research by the institutional review boards of BIDMC (2001-P-001699/14) and MIT (0403000206) without a requirement for individual patient informed consent because data are deidentified and publicly available. Cohort Selection Data were obtained via the Google BigQuery (Alphabet Inc.) cloud platform using RStudio Version 2023.6.1.524 (Posit Software, PBVC) with the R 4.3.2 programming language (R Foundation for Statistical Computing). Hospital admissions were identified where the first sodium value was less than 120 mEq/L. Patients were excluded if they were less than 18 years of age and/or were admitted for less than 24 hours. For each patient, only the last hospital admission was included to maximally All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprintthis version posted March 11, 2024. ; https://doi.org/10.1101/2024.03.08.24303993doi: medRxiv preprint 7 assess the competing risk of mortality. This study followed the STROBE guidelines for observational studies.10 Sodium Value Determination The initial serum sodium value used for each patient was the first charted value for the admission. The final sodium value was the value from closest to 24 hours after the initial value, but was limited to values between 20 and 28 hours. This method was used rather than approximating a 24 hour value to better reflect real-world management. The total sodium correction during this period was calculated as the difference between the final value and the initial value and was classified for analysis as less than 6 mEq/L, 6-10 mEq/L, or greater than 10 mEq/L, as was done previously. 4 Disease and Organ System Classifications DRG codes for admissions were classified by disease state and/or organ system, based on manual review of the data set to identify common categories, as well as disease states commonly associated with hyponatremia. The final categories chosen were: cardiopulmonary, digestive, hematology/oncology, infection, liver, toxic/metabolic, neurologic/psychiatric, orthopedic, renal/urologic, and other diagnosis. If a patient fit both a disease state and an organ system (e.g. infection and liver), the patient was classified into the non-organ-based disease state group (infection). Patients categorized primarily by the electrolyte disorder were included in the toxic/metabolic category. The final classifications for all DRG codes used in this study are provided in e-Table 1. All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprintthis version posted March 11, 2024. ; https://doi.org/10.1101/2024.03.08.24303993doi: medRxiv preprint 8 Determination of Elixhauser Scores International classification of diseases (ICD) revisions 9 and 10 codes were extracted to evaluate acute diagnoses and chronic comorbidities collectively. For each hospital admission, the Elixhauser comorbidity score was calculated11 using the R ‘comorbidity’12 package. This score aggregates ICD codes into a set of 31 binary diagnostic categories.13,14 For each DRG category specified in this study, the proportion of patients who carried ICD codes consistent with the DRG was also determined, as shown in e-Table 2. Statistical Analysis Univariate linear regression analysis was used to assess for an association between correction rate category in the first day (10 mEq/L) and hospital length of stay, and multivariable regression analysis with the addition of a multi-level variable for diagnosis category was used to determine if the univariate association persisted when controlling for diagnosis group. The reference correction rate was 6-10 mEq/L and the reference diagnosis in the multivariable models was toxic/metabolic. This analysis was repeated with the exclusion of patients for whom that hospitalization ended in death, to assess the impact of the competing risk of death on the association, if any, between sodium correction rate and hospital length of stay. For all analyses, a p-value of <0.05 was considered statistically significant.

There were 636 patients included in this study. Overall cohort characteristics are shown in Table 1. Patients were 55% female and had a median [interquartile range, IQR] age of 65 [55, 78] years. There was an overall mortality rate of 13% and a median [IQR] hospital length of stay All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprintthis version posted March 11, 2024. ; https://doi.org/10.1101/2024.03.08.24303993doi: medRxiv preprint 9 of 7 [4, 11] days, but both outcomes varied widely by organ system/disease state (Table 2, Figure 1). Overall, 75% of patients had an Elixhauser diagnosis that matched their DRG category, but also with substantial variation by DRG category (e-Table 2). Patients had a median [IQR] initial sodium value of 117 [114, 118] mEq/L and final sodium value of 124 [119, 128] mEq/L. Proportions of patients with corrections of 10 mEq/L were 36%, 36%, and 28%, respectively. As shown in Table 2, these proportions differed widely by organ system/disease state. In a univariate linear regression, the highest rate of correction (>10 mEq/L) was associated with a shorter length of stay than a moderate rate of correction (6-10 mEq/L) (coef. - 2.363, 95% CI [-4.710, -0.017], p=0.048). This association persisted among the majority subset of patients who survived the hospitalization (coef. -2.552, 95% CI [-5.080, -0.024], p=0.048). There was no difference in length of stay between slow (<6 mEq/L) and moderate correction patients either in the full cohort (coef. 1.420, 95% CI [-0.779, 3.619], p=0.205) or in survivors (coef. 1.948, 95% CI [-0.468, 4.364], p=0.114) in a univariate analysis. When controlling for organ system/disease class, the association between rapid correction and length of stay was not significant in either the full cohort (coef. -1.685, 95% CI [-3.836, 0.467], p=0.125) or in survivors (coef. -1.657, 95% CI [-3.871, 0.556], p=0.142).

In this study, we leverage high-resolution critical care and administrative data to assess how controlling for hospital diagnosis categories modifies associations between early sodium correction rate and hospital length of stay in patients with severe hyponatremia. While consistent with another recent study showing that a correction of >10 mEq/L in the first 24 hours of All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprintthis version posted March 11, 2024. ; https://doi.org/10.1101/2024.03.08.24303993doi: medRxiv preprint 10 hospitalization was associated with shorter length of stay,4 when controlling for admission diagnosis categories, the association is clearly not significant. This suggests that the reported association between rapid correction and shorter length of stay is likely confounded by diagnosis, and faster correction does not consistently reduce length of stay. Management of severe hyponatremia is a controversial subject in nephrology and critical care, and it can have a significant impact on the hospital course. It can also have medicolegal implications, as adverse outcomes may be attributed to sodium correction practices. 2 Slow correction is recommended to reduce the risk of the exceedingly rare but potentially catastrophic potential outcome of ODS, but data showing improved outcomes with faster correction could shift this balance.15 However, the rarity of this outcome and confounding factors also make it difficult to study prospectively.16 Confounding by diagnosis has presented a critical shortcoming in prior analyses. Conceptually, it can be understood by considering that patients with certain diagnoses, such as decompensated liver disease, may present in a critically ill state that portends a prolonged hospitalization and it may not be safe or physiologically possible to rapidly correct their sodium values. 17 Meanwhile patients presenting with more readily reversible conditions that do not portend a prolonged admission may also be clinically better candidates for rapid sodium correction. However, it is not clear that correcting sodium more rapidly in a patient with decompensated liver disease, if possible, would reduce length of stay, and doing so could lead to other complications that could even increase length of stay. The use of diagnosis related groups (DRG), a standard classification of hospital admission diagnoses which we categorized according to organ systems or general disease states, allows us to reassess previous findings while addressing confounding by diagnosis. For this All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprintthis version posted March 11, 2024. ; https://doi.org/10.1101/2024.03.08.24303993doi: medRxiv preprint 11 analysis, DRGs are preferable to other more commonly used sources of information about diagnosis, such as ICD codes, which may not always distinguish between acute presentations and underlying chronic conditions.18 As we show, using common comorbidity burden indices such as the Elixhauser score may also capture some admission diagnoses better than others, making them poor substitutes for direct classification of admission diagnosis. It is important to reiterate that while our study does not support an overall shorter length of stay with faster correction, it does not show that rapid correction is detrimental either. However, it does support the need for more nuanced consideration of hospital diagnosis and comorbidities in future studies and at the bedside. It also stresses that it may be inadvisable to manage sodium correction according to one-size-fits-all guidelines. Strengths of this study include the use of DRG codes to precisely identify the reason for admission. We also provide all applicable DRGs and our associated classifications in e-Table 1, allowing for scrutiny and reproducibility. Finally, this study was based on high-resolution data, and we assessed correction rates in a way that mimicked real-world conditions rather than artificial averages.

are that our study only included patients with severe hyponatremia, defined as initial sodium <120 mEq/L. Disease state classifications are subjective, and changing these classifications could theoretically affect the results. Our cohort was also smaller than prior work on this topic supporting faster sodium correction, but the change in significance with the addition of diagnosis categories is nonetheless compelling, 4 If there were a clear independent clinical benefit from rapid correction, we should be skeptical of the argument that a much larger sample is needed to show statistical significance. All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprintthis version posted March 11, 2024. ; https://doi.org/10.1101/2024.03.08.24303993doi: medRxiv preprint 12

When controlling for admission diagnosis categories, there was not a significant association between relatively rapid sodium correction rate and shorter hospital length of stay. Our findings do not support faster sodium correction as a strategy to reduce hospital length of stay in a diverse population of patients admitted to the hospital with severe hyponatremia. Further research is needed to determine subsets of patients likely to benefit from or be harmed by rapid correction. All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprintthis version posted March 11, 2024. ; https://doi.org/10.1101/2024.03.08.24303993doi: medRxiv preprint 13 Funding/Support: LAC was supported by the National Institutes of Health NIBIB R01 (EB017205). Conflict of Interest The authors have no conflicts of interest to report. Data Sharing Statement: MIMIC is publicly available with training in human subjects research and application. Statistical code is available upon request. All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprintthis version posted March 11, 2024. ; https://doi.org/10.1101/2024.03.08.24303993doi: medRxiv preprint 14

1. Pham CT, Kassab HS, Johnston JP. Evaluation of Serum Sodium Correction Rates for Management of Hyponatremia in Hospitalized Patients. Ann Pharmacother. 2021;56(2):131-138. doi:10.1177/10600280211019752 2. Sterns RH, Rondon-Berrios H, Adrogué HJ, et al. Treatment Guidelines for Hyponatremia. CJASN. 2023;19(1):129-135. doi:10.2215/cjn.0000000000000244 3. Hofmaenner DA, Singer M. Challenging management dogma where evidence is non-existent, weak or outdated. Intensive Care Med. 2022;48(5):548-558. doi:10.1007/s00134- 022-06659-4 4. Seethapathy H, Zhao S, Ouyang T, et al. Severe Hyponatremia Correction, Mortality, and Central Pontine Myelinolysis. NEJM Evidence. 2023;2(10). 5. Kinoshita T, Mlodzinski E, Xiao Q, Sherak R, Raines NH, Celi LA. Effects of correction rate for severe hyponatremia in the intensive care unit on patient outcomes. Journal of Critical Care. 2023;77:154325. doi:10.1016/j.jcrc.2023.154325 6. Doddi S, Tirumani SH. Hospital payment systems and physician reimbursement: A primer for radiology residents. Current Problems in Diagnostic Radiology. Published online October 2023. doi:10.1067/j.cpradiol.2023.10.019 7. Maniaci MJ, Cowdell JC, Maita K, et al. Diagnosis Related Groups of Patients Admitted from an Urban Academic Medical Center to a Virtual Hybrid Hospital-at-Home Program. RMHP. 2023;Volume 16:759-768. doi:10.2147/rmhp.s402355 All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprintthis version posted March 11, 2024. ; https://doi.org/10.1101/2024.03.08.24303993doi: medRxiv preprint 15 8. Johnson AEW, Pollard TJ, Shen L, et al. MIMIC-III, a freely accessible critical care database. Sci Data. 2016;3(1). doi:10.1038/sdata.2016.35 9. Johnson AEW, Bulgarelli L, Shen L, et al. MIMIC-IV, a freely accessible electronic health record dataset. Sci Data. 2023;10(1). doi:10.1038/s41597-022-01899-x 10. Vandenbroucke JP, von Elm E, Altman DG, et al. Strengthening the Reporting of Observational Studies in Epidemiology (STROBE): Explanation and Elaboration. PLoS Med. 2007;4(10):e297. doi:10.1371/journal.pmed.0040297 11. Elixhauser A, Steiner C, Harris DR, Coffey RM. Comorbidity Measures for Use with Administrative Data. Medical Care. 1998;36(1):8-27. doi:10.1097/00005650-199801000- 00004 12. Gasparini A. comorbidity: An R package for computing comorbidity scores. 2018;3:648. doi:10.21105/joss.00648 13. Quan H, Sundararajan V, Halfon P, et al. Coding Algorithms for Defining Comorbidities in ICD-9-CM and ICD-10 Administrative Data. Medical Care. 2005;43(11):1130- 1139. doi:10.1097/01.mlr.0000182534.19832.83 14. Menendez ME, Neuhaus V, van Dijk NC, Ring D. The Elixhauser Comorbidity

Outperforms the Charlson Index in Predicting Inpatient Death After Orthopaedic Surgery. Clinical Orthopaedics & Related Research. 2014;472(9):2878-2886. doi:10.1007/s11999-014-3686-7 15. Sumi H, Imai N, Shibagaki Y. Incidence and risk factors of overcorrection in patients presenting with severe hyponatremia to the emergency department. Clin Exp Nephrol. 2022;26(11):1086-1091. doi:10.1007/s10157-022-02252-7 All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprintthis version posted March 11, 2024. ; https://doi.org/10.1101/2024.03.08.24303993doi: medRxiv preprint 16 16. Mustajoki S. Severe hyponatraemia (P-Na /i1< 116 mmol/l) in the emergency department: a series of 394 cases. Intern Emerg Med. 2023;18(3):781-789. doi:10.1007/s11739- 023-03221-y 17. Lovett GC, Ha P, Roberts AT, et al. Healthcare utilisation and costing for decompensated chronic liver disease hospitalisations at a Victorian network. Internal Medicine Journal. 2022;53(9):1581-1587. doi:10.1111/imj.15962 18. Arthur R, Mayberry RM, Odum S, Kempton LB. Can researchers trust ICD-10 coding of medical comorbidities in orthopaedic trauma patients? OTA International. 2024;7(1). doi:10.1097/oi9.0000000000000307 All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprintthis version posted March 11, 2024. ; https://doi.org/10.1101/2024.03.08.24303993doi: medRxiv preprint 17 Table 1: Characteristics of Patients by Correction Rate Variable Overall 10 mEq/L N 636 227 231 178 Age (years) 65.3 [54.8, 78.3] 68.0 [56.4, 80.0] 67.1 [55.5, 78.5] 61.4 [50.4, 74.6] Male (%) 287 (45.1) 106 (46.7) 114 (49.4) 67 (37.6) Race/Ethnicity (%) Asian 50 (7.9) 14 (6.2) 14 (6.1) 22 (12.4) Black 44 (6.9) 16 (7.0) 12 (5.2) 16 (9.0) Hispanic 26 (4.1) 12 (5.3) 7 (3.0) 7 (3.9) White 462 (72.6) 175 (77.1) 171 (74.0) 116 (65.2) Other/Unknown Ethnicity 54 (8.5) 10 (4.4) 27 (11.7) 17 (9.6) Diagnosis Toxic/Metabolic 260 (40.9) 93 (41.0) 94 (40.7) 73 (41.0) Renal/Urologic 44 (6.9) 8 (3.5) 20 (8.7) 16 (9.0) Cardiopulmonary 104 (16.4) 37 (16.3) 39 (16.9) 28 (15.7) Digestive 13 (2.0) 3 (1.3) 5 (2.2) 5 (2.8) Hematology/Oncology 21 (3.3) 12 (5.3) 5 (2.2) 4 (2.2) Infection 67 (10.5) 20 (8.8) 21 (9.1) 26 (14.6) Liver 60 (9.4) 29 (12.8) 23 (10.0) 8 (4.5) Neurologic/Psychiatric 25 (3.9) 7 (3.1) 10 (4.3) 8 (4.5) Orthopedic 15 (2.4) 6 (2.6) 6 (2.6) 3 (1.7) Other 27 (4.2) 12 (5.3) 8 (3.5) 7 (3.9) Hypertonic Saline (%) 123 (19.3) 42 (18.5) 58 (25.1) 23 (12.9) Initial Na 117.0 [114.0, 118.0] 117.0 [115.0, 118.0] 117.0 [114.0, 119.0] 116.0 [113.0, 118.0] Final Na 124.0 [119.0, 128.0] 119.0 [117.0, 120.5] 125.0 [122.0, 126.0] 131.0 [128.0, 133.8] Length of Stay (days) 6.6 [4.2, 11.1] 7.7 [5.0, 11.7] 6.8 [4.6, 12.6] 5.5 [3.3, 8.9] Deaths (%) 80 (12.6) 35 (15.4) 29 (12.6) 16 (9.0) All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprintthis version posted March 11, 2024. ; https://doi.org/10.1101/2024.03.08.24303993doi: medRxiv preprint 18 Table 2: Proportions of Patients with Each Diagnosis and Mortality, Length of Stay, and Correction Rate Categories by Diagnosis Organ/Disease Category Proportion of Patients with Admission Diagnosis (%) Mortality (%) Length of Stay (days)1 Proportion 10 mEq/L (%) Toxic/Metabolic 41 3 5 [3, 7] 36 36 28 Cardiopulmonary 16 14 10 [6, 17] 36 38 27 Infection 11 30 7 [5, 12] 30 31 39 Liver 9 33 16 [7, 35] 48 38 13 Renal/Urologic 7 5 6 [5, 9] 18 45 36 Other 4 11 9 [6, 18] 44 30 26 Neurologic/Psychiatric 4 20 6 [3, 8] 28 40 32 Hematology/Oncology 3 19 6 [4, 10] 57 24 19 Orthopedic 2 0 10 [5, 16] 40 40 20 Digestive 2 15 13 [10, 17] 23 38 38 1Median [IQR] All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprintthis version posted March 11, 2024. ; https://doi.org/10.1101/2024.03.08.24303993doi: medRxiv preprint 19 Table 3: Linear Regression Models of Associations Between Correction Rate and Length of Stay, Controlling for Organ System/Disease State All +Organ/Disease Survived +Organ/Disease (Intercept) 10.551 [9.003,12.099] 6.260 [4.512,8.008] 10.339 [8.652,12.025] 5.892 [4.121,7.663] p<0.001 p<0.001 p<0.001 p<0.001 Rate 10 -2.363 [-4.710,-0.017] -1.685 [-3.836,0.467] -2.552 [-5.080,-0.024] -1.657 [-3.871,0.556] p=0.048 p=0.125 p=0.048 p=0.142 Cardiopulmonary 7.482 [4.993,9.972] 7.914 [5.339,10.489] p<0.001 p<0.001 Digestive 8.282 [2.179,14.385] 5.697 [-0.740,12.133] p=0.008 p=0.083 Hematology/Oncology 3.733 [-1.149,8.616] 2.192 [-3.049,7.434] p=0.134 p=0.412 Infection 4.702 [1.755,7.649] 5.035 [1.713,8.357] p=0.002 p=0.003 Neurologic/Psychiatric 1.890 [-2.605,6.385] 2.288 [-2.561,7.137] p=0.409 p=0.354 Orthopedic 5.383 [-0.317,11.083] 5.498 [-0.052,11.048] p=0.064 p=0.052 Renal/Urologic 1.487 [-2.025,4.999] 1.756 [-1.740,5.253] p=0.406 p=0.324 Other 9.467 [5.126,13.809] 10.505 [6.044,14.966] p<0.001 p<0.001 N 636 636 556 556 All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprintthis version posted March 11, 2024. ; https://doi.org/10.1101/2024.03.08.24303993doi: medRxiv preprint 20 Figure Captions Figure 1: Hospital Length of Stay for all Patients Stratified by Organ/Disease Category All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprintthis version posted March 11, 2024. ; https://doi.org/10.1101/2024.03.08.24303993doi: medRxiv preprint All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprintthis version posted March 11, 2024. ; https://doi.org/10.1101/2024.03.08.24303993doi: medRxiv preprint