This retrospective study used the Virtual Pediatric Systems (VPS) web-based database (Los Angeles, CA; http://www.myvps.org). The data were entered into the VPS by a trained medical records officer and reviewed after retrieval by researchers (pediatric intensive care physicians). Institutional review board (IRB) approval was obtained from the local IRB committee at King Fahad Medical City (IRB log number: 21-355).

The KFMC is a tertiary hospital with a multidisciplinary PICU that provides specialized care and serves as a referral center for other regional healthcare institutes. The patients were admitted to the KFMC PICU with complex medical and surgical diseases.

All children younger than 14 years of age admitted to the KFMC PICU between January 1, 2014, and December 31, 2019, were included in this study. Patients older than 14 years and neonates with a corrected gestational age of <44 weeks were admitted to different units and were not included in the VPS database. Patients with incomplete data and those still admitted to the PICU at the end of the study period were excluded from the study because LOS could not be determined.

The data collected included baseline characteristics, LOS in the PICU, patient origin, history of previous PICU admission, diagnosis, comorbidities at admission, and Pediatric Risk of Mortality (PRISM) III, which calculated the probability of death. PRISM III is a score developed in 1996 and is calculated based on 21 physiological variables to estimate the severity of illness and probability of death (2). Other admission-related variables were collected, including procedures such as continuous renal replacement therapy; central venous catheter (CVC) insertion; exchange transfusion; Foley catheter insertion; intracranial pressure (ICP) monitor/extra-ventricular drain (EVD) insertion; peripherally inserted central venous catheter (PICC); tracheostomy and arterial line insertion; respiratory support needed; cardiac arrest events and complications such as ventilator-associated pneumonia (VAP), central line-associated bloodstream infection (CLABSI), catheter-associated urinary tract infection (CAUTI), and hospital-acquired pneumonia (HAP) acquired during admission. Information on do not resuscitate (DNR) and mortality status was also collected.

The LOS was calculated by subtracting the date and time of PICU admission from the date and time of PICU medical discharge. The PICU physical discharge time was excluded to avoid the influence of logistic causes that were not documented as a reason for the delay in patient charts. As previous studies have concluded, prolonged LOS was defined based on the 95th percentile for patients included in this study (6). The 95th percentile was calculated to be 29.5 days in this study population.

After analyzing the distribution of LOS, the patients were stratified into normal and prolonged stay groups. The baseline characteristics were compared between the two groups. The median and interquartile range (IQR) were calculated for continuous numerical data, whereas frequencies and percentages were used for nominal data. Mann-Whitney U, Kruskal-Wallis, and Chi-square tests were used to compare the characteristics of patients who stayed for 30 days or less and those with prolonged LOS. Statistical significance was set at P < 0.05.

Multivariate analysis was performed using logistic regression analysis. Variables were found to be significantly different (p < 0.05) between the two groups included in the multivariate analysis. The odds ratio (OR) was estimated with a 95% confidence interval (CI) for each included variable presumed to contribute to increasing length of stay.

Data were collected and analyzed using IBM SPSS Statistics (version 25.0; IBM Corp., Armonk, NY, USA). The statistical analyses performed in this study were thoroughly reviewed by a trained biostatistician.

LOS was found to have a rightward skewed distribution with a median of 2.8 (IQR, 1.08–7.04) and a mean of 7.43 ± 14.34 days. The 95th percentile was calculated to be 29.5 days. Thus, LSPs were defined as patients who stayed in the PICU for > 30 days.

The duration of admission was measured at 0–30 days for 3,233 (95.2%) patients and > 30 days for 163 (4.8%) patients. 44.8% of LSPs were less than 12 months of age (Table 1). LSPs had significantly lower growth parameters than patients with shorter stays (height of 73 cm and 87 cm, respectively, and weight of 8 kgs and 11.2 kgs, respectively) with a p-value of < 0.001. The sex of the admitted patients did not differ significantly between the groups (See Table 1 below).

Patient origin differed significantly between the two groups (p < 0.001). Patients admitted from the ward comprised the majority (50.92%) of LSPs, followed by patients admitted from the emergency room (34.97%; Table 1).

As estimated by PRISM III, LSPs had a significantly higher probability of death (1.08%%; p < 0.001). The actual mortality was as well significantly higher in LSPs (26.4%; p < 0.001), Most patients were admitted to the PICU because of respiratory diseases (29.77%), followed by infectious diseases (16.6%). Most LSPs were admitted with respiratory (51.53%) or infectious (22.09%) diseases (Table 2).

Comorbidities at admission for LSPs were significantly more frequent than that for shorter stay patients, with median values of 2 (IQR, 2–3) and 1 (IQR, 1–2), respectively. LSPs had more comorbidities related to infectious diseases (68.10%), central nervous diseases (54.60%), and respiratory diseases (50.90%) compared to shorter stay patients with p-value < 0.001 (See Table 2 below).

Procedure frequency, need for respiratory support, and complications during admission were predominant in LSPs. CVC, Foley catheter, nasogastric tube (NGT), PICC, arterial line, and ICP monitor/EVD insertions were significantly higher in patients who stayed for more than 30 days with 52.76%, 40.49%, 39.26%, 20.86%, 11.04% and 4.29%, respectively with p-value < 0.001. Furthermore, LSPs had more tracheostomies compared to other groups (27.61%) (p < 0.001; see Table 3 below).

LSPs also required more respiratory support during admission. High flow nasal cannula, non-invasive, conventional, and high-frequency ventilation were utilized in a significantly greater number of LSPs—38.04%, 65.03%, 77.91%, and 18.40%, respectively (p < 0.001; Table 3).

Regarding complications during PICU admission, out of the patients admitted for more than 30 days, 27 (16.56%) had CLABSI, while VAP, HAP, and CAUTI were reported for 4.29%, 4.29%, and 3.07%, respectively. Complications during PICU admission were significantly higher in patients with more than 30 days of hospital stay (p < 0.001; Table 3). During admission, LSPs were found to have significantly more arrest events (12.88%), and patient-signed DNR (17.20%; p < 0.001).

Multivariate logistic regression analysis was conducted for all variables that were significantly different between the groups. The remaining factors with significant correlations after the multivariate analysis are presented in Table 4. CAUTI, CLABSI, VAP, and procedures during admission, such as ICP monitoring, PICC insertion, and tracheostomy, had the highest OR compared to other variables included in the multivariate analysis.

Patients admitted as a referral from other hospitals' intensive care units had an OR of 3.03 (1.32–6.96, 95% CI; p = 0.009), while patients admitted postoperatively from the operating room had an OR of 0.12 (0.03–0.44, 95% CI; p = 0.001). Central nervous system (CNS) disease was the only comorbidity with significant relationships in the analysis, with an OR of 1.68 (1.07–2.63, 95% CI; p = 0.024; Table 4).

Procedures during PICU admission, including ICP monitor/EVD, PICC insertion, and tracheostomy, were found to significantly increase the LOS with ORs of 5.93 (1.72–20.42, 95% CI; p = 0.005), 5.67 (2.64–12.20, 95% CI; p < 0.001), and 7.72 (4.18–14.28, 95% CI; p < 0.001), respectively. Additionally, respiratory support required during admission also affected the LOS. Non-invasive ventilation had an OR of 4.92 (3.12–7.75, 95% CI; p < 0.001), conventional mechanical ventilation had an OR of 1.83 (1.09–3.06, 95% CI; p = 0.022), and high frequency oscillatory ventilation had an OR of 2.56 (1.31–5.02, 95% CI; p = 0.006; Table 4).

Complications during PICU admission had the highest impact on LOS. CLABSI had the highest OR among the variables included in the study (OR, 12.87; 95% CI 5.34–31.01; p < 0.001). While other complications such as CAUTI with an OR of 9.16 (1.62–51.97, 95% CI; p = 0.012), VAP with an OR of 8.66 (1.48–50.53, 95% CI; p = 0.016), and HAP with an OR of 4.88 (1.12–21.29, 95% CI; p = 0.035) also showed an association with increased LOS—although to a lesser degree than CLABSI. Patients with DNR status were also associated with a prolonged LOS with an OR of 5.55 (2.95–10.44, 95% CI); p < 0.001; see Table 4 below).

This study has several limitations. First, it was conducted at a single institution, limiting the generalizability of its findings. Moreover, this triggers questions of construct validity as well as concerns regarding the reliability and replicability of a study centered on a single institution. Second, the study was retrospective; thus, institutional practices and individualized physician care may have also influenced the study outcomes.