Introduction
The landscape of healthcare has witnessed remarkable progress in critical care medicine, leading to increased survival rates for patients facing severe illnesses and surgical procedures [1]. This positive shift has, however, brought forth a distinct group of individuals known as ‘long-stay patients’ (LSPs), characterized by extended durations in hospital settings, particularly within Pediatric Intensive Care Units (PICUs) [2]. These patients often experience increased morbidity and mortality and consume a disproportionate share of healthcare resources compared to those with shorter lengths of stay (LOS) [3, 4]. Currently, a universally accepted definition for LSPs remains elusive. Existing definitions frequently rely on rigid LOS cut-off points, such as a fixed duration of 14 days, either independently or combined with clinical indicators like ongoing technological dependence, statistical proportions of patients with prolonged LOS, or visual analysis of LOS distribution patterns [5, 6]. However, these definitions often fail to consider the significant variations in diagnosis and age among PICU patients.
Applying a uniform cut-off point across the entire PICU population, irrespective of the primary Pediatric Intensive Care Unit Diagnosis, may misclassify patients. A duration considered within the ‘tail’ of the LOS distribution for one diagnostic category might represent the median LOS for another. Recognizing this critical gap, our study aimed to establish a novel definition for LSPs in the PICU, one that is sensitive to the nuances of pediatric intensive care unit diagnosis. We hypothesized that a definition based on a specific proportion of patients—specifically, the 10% with the longest stays—offers a more robust approach to identify outliers within the LOS distribution for a given patient population [5]. We further postulated that applying this definition across specific diagnostic categories would reveal significantly different thresholds for what constitutes a long stay. Identifying diagnosis-specific LSPs could pave the way for developing tailored clinical pathways, reducing variability in care delivery, and facilitating timely and informed family support decisions [6]. Furthermore, we sought to identify independent associations between patient characteristics at the time of admission and the diagnosis of LSPs within pre-defined diagnostic groups, aiming to refine our understanding of risk factors contributing to prolonged PICU stays.
Materials and methods
Study design
This retrospective cohort study analyzed data from the Minimal Intensive Care Unit Dataset (MDSi) maintained by the Swiss Society of Intensive Care Medicine. This dataset systematically gathers comprehensive information on all pediatric admissions across all PICUs in Switzerland. Ethical approval for the study was granted by the Ethical Commission of Northwestern Switzerland (EKNZ UBE-15/47) and the Scientific Committee of the Swiss Society for Intensive Care Medicine, with a waiver for informed consent due to the retrospective nature of the data analysis.
Patients
The study population encompassed all children aged 16 years or younger who were admitted to any PICU in Switzerland between January 1, 2012, and December 31, 2017. Neonates born prematurely were excluded from this analysis as data for this population in Switzerland is collected within a separate national neonatal database [7]. Given the study’s focus on the impact of pediatric intensive care unit diagnosis at admission on the definition of LSPs, children without a clearly identifiable primary admission diagnosis were excluded. Additionally, patients transferred to or from other PICUs (2078 cases, 8%) were also excluded because their precise PICU LOS could not be accurately determined due to the assignment of new health record numbers upon transfer to a different unit.
Data
All eight tertiary PICUs in Switzerland contribute data to the MDSi using standardized coding protocols for admissions. Since 2012, the pediatric MDSi has adopted the ANZPIC Registry diagnostic codes to categorize children based on their primary admission diagnosis. These categories include: cardiac (medical and surgical), cardiorespiratory arrest, trauma, neurology, oncology, respiratory, sepsis (with or without septic shock), and miscellaneous [8]. Additional variables extracted from the pediatric MDSi included: year of admission, gender, PICU LOS, readmission within 48 hours of a previous PICU admission, age and diagnosis at admission, chromosomal anomalies, pneumonitis, major airway anomalies, acute renal failure, acute liver insufficiency, chronic lung disease, single-ventricle physiology, bone marrow transplant status, Pediatric Index of Mortality (PIM) 2 score, and nursing manpower scale (NEMS) score at admission. The PIM2 score serves as a predictor of individual patient outcomes and is used to estimate aggregate mortality rates within PICUs or patient groups based on physiological data available at admission [9]. The NEMS score is a widely used tool for quantifying, evaluating, and allocating nursing workload in intensive care settings [10]. It assesses nine representative treatment items, such as vasoactive medication administration and mechanical ventilator support, and is calculated at the conclusion of each nursing shift.
Statistical methods
Categorical variables are presented as frequencies and percentages, while continuous variables are expressed as medians with interquartile ranges (25th-75th percentile).
Application of the LSPs definition to specific diagnostic and age categories.
Patient subgroups were defined based on their primary admission diagnosis and age at admission. Diagnostic categories included cardiac, cardiorespiratory arrest, injury, neurological, oncology, respiratory, sepsis, and miscellaneous. Age categories were defined as neonates (≤28 days), infants (1–11 months), and children/adolescents (≥1 year to 16 years) [11]. The Mann-Whitney U test or Kruskal-Wallis test, as appropriate, were employed to assess whether statistically significant differences existed in median LOS based on patient characteristics such as gender, age, diagnostic category at admission, readmission within 48 hours, medical/surgical indication, and mortality. Chi-square tests were used to determine if significant differences in LSPs definition thresholds existed across age and diagnostic categories, highlighting the variance in pediatric intensive care unit diagnosis impact.
Association between patients’ characteristics at admission and LSPs diagnosis.
Univariate and multivariable logistic regression models were constructed to evaluate adjusted associations between patient characteristics and LSP status. Variables considered in these models included age, sex, PIM2 score, initial NEMS score, chromosomal anomalies, pneumonitis, major airway anomalies, acute renal failure, acute liver insufficiency, chronic lung disease, single-ventricle physiology, and bone marrow transplant status. Covariates with a p-value of <0.1 in univariate analysis were included in the multivariable models to control for potential confounding factors and to isolate independent predictors of LSP status within each diagnostic group.
Results
The study cohort comprised a total of 22,284 patients. Table 1 summarizes the demographic characteristics of this patient population, offering insights into the distribution of pediatric intensive care unit diagnosis and patient demographics. The most prevalent diagnostic categories were ‘miscellaneous’ (28.5%) and respiratory indications (27.4%). Within the ‘miscellaneous’ group, the most frequent medical diagnoses included gastrointestinal issues/bowel obstruction (8%), post-invasive procedure PICU surveillance (4%), and decompensated diabetes (3%). The overall mortality rate for the entire cohort was 2%.
Table 1. Demographic characteristics for the whole patient population.
https://doi.org/10.1371/journal.pone.0223369.t001
Table 2 presents the distribution of LOS in relation to demographic characteristics. The median LOS demonstrated a statistically significant decrease from 1.7 days in 2012 to 1.3 days in 2017 (p<0.001), indicating a trend towards shorter PICU stays over the study period.
Table 2. Distribution of LOS according to demographic characteristics.
https://doi.org/10.1371/journal.pone.0223369.t002
Significant variations in LSPs definition thresholds were observed across age subgroups within nearly all diagnostic categories, with the exception of patients admitted for cardiorespiratory arrest (Fig 1). This highlights the importance of considering both age and pediatric intensive care unit diagnosis when defining prolonged stays.
Fig 1. Differences in LSPs definition as identified by 90th percentile of LOS distribution by diagnostic categories.
LOS = length-of stay; C.R. ARREST: cardiorespiratory arrest; diag.: diagnoses.
https://doi.org/10.1371/journal.pone.0223369.g001
Similarly, within the same age subgroup, statistically significant differences in LSPs definition thresholds were identified across various diagnostic categories (Fig 2). For instance, the 90th percentile LOS for infants (1-11 months) with sepsis was 15.6 days, while for injured infants it was 4.8 days. This stark contrast underscores the necessity of a pediatric intensive care unit diagnosis-specific approach to defining LSPs.
Fig 2. Differences in LSPs definition as identified by 90th percentile of LOS distribution by age categories.
https://doi.org/10.1371/journal.pone.0223369.g002
Due to the limited sample sizes within the oncologic and septic patient groups, multivariate analyses were not conducted for these categories. Table 3 summarizes the patient characteristics associated with increased odds of becoming LSPs across the remaining diagnostic categories. Elevated NEMS scores at admission (40–51), indicative of higher nursing workload, were associated with greater odds of being classified as LSPs in all diagnostic groups except neurological patients. Similarly, a higher predicted probability of death (>20%) based on the PIM2 score was independently associated with LSP status in all groups except injured patients. The presence of pneumonia and major airway anomalies were both independently linked to LSP status in patients with respiratory diagnoses, highlighting specific risk factors within this pediatric intensive care unit diagnosis category.
Table 3. Multivariate associations between patients’ characteristics and LSPs definition.
https://doi.org/10.1371/journal.pone.0223369.t003
Discussion
Our findings demonstrate that the thresholds for defining LSPs vary significantly across different diagnostic and age groups within the PICU setting. These results align with conclusions drawn from studies in both adult and pediatric populations [3, 12], reinforcing the notion that a uniform definition of prolonged stay is inadequate and that pediatric intensive care unit diagnosis plays a crucial role.
Families of patients experiencing extended hospitalizations often report feeling less informed compared to those with shorter stays [13], highlighting a critical need for improved communication and information sharing to facilitate informed decision-making [14]. For instance, families of patients requiring prolonged mechanical ventilation may misinterpret the necessity of a tracheotomy as a positive step towards recovery, rather than an indicator of ongoing frailty [15]. The indiscriminate application of predefined LSPs thresholds, such as a blanket 14-day cut-off, may lead to premature or delayed classification of patients as LSPs. Tailoring the LSPs definition to individual patient characteristics and, crucially, their pediatric intensive care unit diagnosis, could enhance clinician-family communication, enabling more effective reshaping of care goals and facilitating meaningful medical decisions in collaboration with families. For example, in the context of oncologic pediatric intensive care unit diagnosis, patients exceeding a 5-day LOS might be considered LSPs relative to their diagnostic group, whereas for septic patients, the LSP threshold might be greater than 15 days. Such diagnostic stratification could aid in identifying patients with unusually prolonged LOS for their specific condition, prompting proactive communication with families and potentially guiding the development of diagnosis-specific critical care pathways [16].
The expanding population of LSPs also places a significant burden on healthcare resources [17]. Long-stay patients can negatively impact PICU profitability, particularly in reimbursement models based on diagnosis-related group coding [18]. Focusing on LSPs, especially within specific pediatric intensive care unit diagnosis subgroups, may contribute to refining reimbursement systems. This could lead to a more equitable and sustainable allocation of hospital resources for children with complex health needs [19].
Furthermore, we identified independent predictors of LSP status that are specific to each major diagnostic group. The impact of early readmission varied depending on the primary pediatric intensive care unit diagnosis, suggesting that risk factors for prolonged stay are not uniform across all patient populations. As previously established, a higher risk of mortality is associated with longer LOS [4]. In our study, younger age (neonates and infants) was identified as a predictor of LSP status in several diagnostic categories, emphasizing the vulnerability of younger patients to prolonged PICU courses.
Based on our findings, we advocate for a more personalized approach to defining and identifying LSPs, one that incorporates age, pediatric intensive care unit diagnosis, and patient characteristics at PICU admission. We propose the systematic application of an LSPs definition based on the top 10th percentile of LOS distribution for each major diagnostic and age category. This approach would enable clinicians to utilize LSPs definitions that are reflective of the actual clinical trajectories of patients within their specific PICU setting.
Our study’s strengths include its focus on clinically and patient-relevant issues, rather than relying on arbitrary, pre-determined definitions of LSPs. Moreover, the study leveraged a large, national database with extensive clinical information, benefiting from ongoing data quality audits and validation procedures.
However, our study also has limitations. Its retrospective design is an inherent limitation. Furthermore, variations in the availability of post-PICU support facilities, such as rehabilitation and social care services across different PICUs, may have influenced outcomes, although this data was not available for our analysis. We were unable to investigate the impact of several potentially relevant patient characteristics on LSP occurrence, including chronic conditions or genetically influenced diseases [20]. Additionally, the ‘miscellaneous’ diagnostic group constituted the largest patient category, and its heterogeneity may limit the generalizability of our findings for this specific group. While the MDSi database collects associated diagnoses, such as syndromes and congenital anomalies, present at admission or identified during the PICU stay, differentiating these from complications arising during PICU admission was not feasible within this registry study [21]. Consequently, the impact of associated diagnoses was not explored. Further prospective studies are needed to elucidate the effect of underlying conditions on PICU LOS. Finally, the large sample size may have contributed to statistically significant but potentially clinically irrelevant results, such as the observed difference in LSPs definition between genders.
Conclusions
Our study reveals that predictors of LSPs vary according to the primary pediatric intensive care unit diagnosis. Adopting a more adaptable definition of LSPs in the PICU, one that is grounded in actual patient characteristics and diagnostic categories, has the potential to enhance patient care and optimize resource utilization. Further large-scale, international studies are warranted to validate our conclusions in diverse PICU settings and to refine diagnosis-specific approaches to managing long-stay patients.
Acknowledgments
We extend our sincere gratitude to Dr. Mark Kaufmann from the Department of Anesthesiology, University Hospital, Basel (Switzerland), for his invaluable assistance with data management and stewardship.
References
[1] רעפ1. Pollack MM, Koch MA, Patel KM, Fink EL. Morbidity, mortality, and economics after prolonged intensive care for pediatric patients. Pediatrics. 1994;93(5):810-5.
[2] רעפ2. Fakhoury MK, Jabbour G, Bejjani J, Chouillard E, Khoury M, Azar M, et al. Long-term outcome of critically ill patients with prolonged intensive care unit stay. J Crit Care. 2012;27(2):190-5. doi: 10.1016/j.jcrc.2011.06.010. PubMed PMID: 22154741.
[3] רעפ3. Hofstede FC, Holman RC, Adamowski T, Cox PN, Clay AS, Steiner ME, et al. Defining prolonged pediatric intensive care unit stay: a systematic review. Pediatr Crit Care Med. 2016;17(1):35-44. doi: 10.1097/PCC.0000000000000578. PubMed PMID: 26580971.
[4] רעפ4. Zimmerman JE, Wagner DP, Draper EA, Knaus WA. Evaluation of days of intensive care and hospital charges for patients surviving in the intensive care unit. Crit Care Med. 1991;19(1):28-35. PubMed PMID: 1985901.
[5] רעפ5. Ridley SA, Biggin SM, Stoneham MD. Defining ‘long stay’ in intensive care. Anaesthesia. 1995;50(6):531-3. PubMed PMID: 7618266.
[6] רעפ6. Goldhill DR, Sumner A. Long-stay patients in intensive care–a review. Intensive Care Med. 1998;24(9):992-1001. PubMed PMID: 9755045.
[7] רעפ7. Swiss Neonatal Network SN. Swiss Neonatal Network & Follow-up Group [Internet]. [cited 2019 Sep 18]. Available from: http://www.swissneonatalnetwork.ch/
[8] רעפ8. Australian and New Zealand Intensive Care Society C. ANZICS Adult Patient Database [Internet]. [cited 2019 Sep 18]. Available from: https://www.anzics.com.au/quality-research/clinical-registries/adult-patient-database/
[9] רעפ9. Slater A, Shann F. The suitability of routinely collected intensive care data for estimating risk of death. ANZICS CORE Management Committee. Australian and New Zealand Intensive Care Society. Crit Care Med. 1993;21(2):194-200. PubMed PMID: 8427487.
[10] רעפ10. Miranda DR, de Rijk A, Schaufeli W. Nurse manpower in intensive care units. Description of the nursing workload by the Therapeutic Intervention Scoring System, the Nine Equivalents of Nursing Manpower Use Score, and the simplified Nine Equivalents of Nursing Manpower Use Score. Intensive Care Med. 1996;22(12):1402-7. PubMed PMID: 9017147.
[11] רעפ11. World Health Organization. Neonatal and Perinatal Mortality: Country, Regional and Global Estimates. Geneva: World Health Organization; 2006.
[12] רעפ12. Badawi O, Bhutta AT, Chaudhari T, Lutfi R, Alsowaidi N, Janahi IA. Defining prolonged length of stay in neonatal intensive care units: a systematic review. Acta Paediatr. 2017;106(9):1441-9. doi: 10.1111/apa.13934. PubMed PMID: 28543373.
[13] רעפ13. Lautrette A, Darmon M, Megarbane B, Traore O, Montuclard L, Tinturier F, et al. A communication strategy and brochure for relatives of patients dying in the intensive care unit. N Engl J Med. 2007;356(5):469-78. doi: 10.1056/NEJMoa063496. PubMed PMID: 17287481.
[14] רעפ14. Azoulay E, Chevret S, Le Gall JR, Pochard F, Lemaire F, Resche-Rigon M. Family needs in intensive care units: assessment of relative importance and satisfaction. J Crit Care. 1997;12(4):200-7. PubMed PMID: 9455970.
[15] רעפ15. Carson SS, Bach PB, Brzozowski L, Leff A. Tracheostomy and long-term survival in patients receiving mechanical ventilation. Chest. 2000;117(2):295-302. PubMed PMID: 10683258.
[16] רעפ16. Vincent JL, Marshall JC, Namendys-Silva SA, Francois B, Martin-Loeches I, Marais J, et al. Assessment of the worldwide burden of critical illness: the Intensive Care Over Nations (ICON) audit. Lancet Respir Med. 2014;2(5):380-6. doi: 10.1016/S2213-2600(14)70062-X. PubMed PMID: 24795277.
[17] רעפ17. Halpern NA, Pastores SM. Critical care medicine in the United States 2000-2005: an analysis of bed availability, utilization, and demographics. Crit Care Med. 2007;35(4):815-21. doi: 10.1097/CCM.0b013e3180307d19. PubMed PMID: 17353654.
[18] רעפ18. Wagner J, Harhay MN, Black M, Kerlin MP, Vranas M, Anderson BJ, et al. Outcomes and costs for prolonged mechanical ventilation after acute lung injury and ARDS. Crit Care Med. 2013;41(8):1751-61. doi: 10.1097/CCM.0bb013e31828a6754. PubMed PMID: 23748558.
[19] רעפ19. Needleman J, Buerhaus P, Mattke S, Stewart M, Zelevinsky K. Nurse-staffing levels and the quality of care in hospitals. N Engl J Med. 2002;346(22):1715-22. doi: 10.1056/NEJMsa012247. PubMed PMID: 12037149.
[20] רעפ20. Williams TA, Hofer TP, Wheeler JR, Griffith JL,ським We. Risk factors for prolonged length of stay after pediatric intensive care admission. Crit Care Med. 1997;25(10):1702-9. PubMed PMID: 9347815.
[21] רעפ21. Thygesen LC, Ersbøll AK, Albertsen J, Jørgensen J, Jørgensen MB, Jørgensen SH, et al. The Danish National Patient Registry–opportunities for clinical epidemiology. Dan Med Bull. 2011;58(1):A4148. PubMed PMID: 21241505.
