Readmission to the PICU is associated with worse outcomes, but factors associated with PICU readmission within the same hospitalization remain unclear. We sought to describe the prevalence of, and identify factors associated with, early PICU readmission.
We performed a retrospective analysis of PICU admissions for patients aged 0 to 26 years in 48 tertiary care children’s hospitals between January 1, 2016 and December 31, 2019 in the Pediatric Health Information System. We defined early readmission as return to the PICU within 2 calendar days of floor transfer during the same hospitalization. Generalized linear mixed models were used to analyze associations between patient and clinical variables, including complex chronic conditions (CCC) and early PICU readmission.
The results included 389 219 PICU admissions; early PICU readmission rate was 2.5%. Factors with highest odds of early PICU readmission were CCC, with ≥4 CCCs (reference: no CCC[s]) as highest odds of readmission (adjusted odds ratio [95% confidence interval]: 4.2 [3.8–4.5]), parenteral nutrition (2.3 [2.1–2.4]), and ventriculoperitoneal shunt (1.9 [1.7–2.2]). Factors with decreased odds of PICU readmission included extracorporeal membrane oxygenation (0.4 [0.3–0.6]) and cardiopulmonary resuscitation (0.8 [0.7–0.9]). Patients with early PICU readmissions had longer overall length of stay (geometric mean [geometric SD]: 18.2 [0.9] vs 5.0 [1.1] days, P < .001) and increased odds of mortality (1.7 [1.5–1.9]).
Although early PICU readmissions within the same hospitalization are uncommon, they are associated with significantly worse clinical outcomes. Patients with medical complexity and technology dependence are especially vulnerable.
Early readmission to the ICU within 48 hours of transfer, otherwise known as a “bounce back,” has become an important quality metric in both the adult and pediatric populations.1–4 In adults, readmission to the ICU has been associated with increased risks of in-hospital morbidity, increased healthcare utilization, and 10 times higher odds of mortality.5–9 Similar trends have been seen in pediatrics, but existing data surrounding ICU readmission in pediatrics is derived from single-center studies,10–15 the pediatric cardiac ICU,16–18 or includes patients readmitted after hospital discharge.19 Recent studies suggest that readmission within 2 calendar days is the optimal definition for assessment of ICU readmissions20,21 and the Society of Critical Care Medicine delineates readmission within 48 hours as a “top indicator of ICU quality.”1,3,22 However, little is known about how admission characteristics, PICU treatment exposures, and medical complexity influence odds of early PICU readmission.
As medical technology has advanced, the proportion of children and adolescents living with complex medical conditions has increased substantially. Children with medical complexity (CMC) account for a disproportionate number of pediatric hospitalizations23–25 and healthcare costs.26 Additionally, CMC are more likely to experience recurrent critical illness requiring PICU readmission following hospital discharge.27–29 These children may also be at increased risk of bounce back to the PICU during the same admission.
Identifying patients at risk for early ICU readmission may inform decisions both in the ICU and after transfer to the acute care floor. Further, quantifying the frequency of early PICU readmissions, identifying risk factors, and comparing outcomes of patients requiring readmission may result in more judicious healthcare utilization and improved patient-level outcomes. Thus, we sought to evaluate a large, diverse population of pediatric patients to measure the rate of readmission to the PICU within 2 calendar days of transfer during the same hospitalization. Our secondary objectives were to evaluate baseline risk factors for PICU readmission and compare outcomes of readmitted versus nonreadmitted patients, hypothesizing that CMC would be at an increased odds of requiring PICU readmission within 2 calendar days of transfer.
Methods
Study Design and Setting
We conducted a retrospective cohort study of patients admitted to tertiary children’s hospitals participating in the Pediatric Health Information System (PHIS; Children’s Hospital Association, Lenexa, KS). The PHIS database is a comparative administrative database that includes demographic and daily resource utilization data for inpatient, ambulatory surgery, emergency department, and observation patient encounters for 48 US children’s hospitals. Up to 41 diagnoses and 41 procedures are recorded for each encounter using the International Classification of Diseases version 10. The local Institutional Review Board determined this study was not human subjects’ research because it was an analysis of a pre-existing, deidentified dataset with no patient contact.
Inclusion and Exclusion Criteria
Patients less than 26 years of age with a PICU admission between January 1, 2016 and December 31, 2019 were eligible for inclusion. We included patients up to age 26 as per the age range delineated in the American Academy of Pediatrics policy on health benefits.15 PICU admissions were identified by a billed room charge for a PICU bed at any time during the hospital stay. We excluded patients who did not survive their first PICU admission and patients who were either discharged from the hospital or transferred to another facility directly from the PICU. We used Clinical Classification Software codes30 to exclude patients requiring cardiac intensive care and/or cardiac surgeries as these represent a unique subset of critically ill patients because of the complexity of their procedural interventions and postoperative care. An estimated 10% to 20% of patients undergoing congenital heart surgery require unplanned readmission, with postoperative complications representing one of the most common indications for readmission.31 Patient outcome after cardiac surgery is subject to substantial institutional variation influenced by center case volume, surgical technique, and provider experience.32,33 Since we were unable to control for this institutional variation, we elected to exclude this patient population from our analysis.30
Variables of Interest
The cohort was separated into patients requiring PICU readmission within 2 days of transfer out of the PICU (early PICU readmission group) and those that did not require early PICU readmission. We abstracted the following data from each index PICU encounter: demographic data, dates of admission, diagnoses, procedures, and mortality. Procedure codes and Centers for Medicare and Medicaid Services operative flags were used to identify patients who underwent surgical procedures.34 We defined children with medical complexity (CMC) as the presence of 1 or more complex chronic conditions (CCCs), as defined by Feudtner et al.35 We identified 3 categories of children assisted with medical technology: ventriculoperitoneal (VP) shunt, tracheostomy, and “other” (including patients with gastrostomy tubes (GT), gastrostomy-jejunal (GJ) tubes, cardiac pacemakers, ileostomies, and/or dialysis). Procedure codes and billing data were used to identify aspects of the critical care course, such as use of extracorporeal membrane oxygenation (ECMO), continuous renal replacement therapy, invasive mechanical ventilation, and medications administered, among others (Table 1). We used the 25 Major Diagnostic Categories (MDC) to classify reasons for admission.34
Baseline Demographics and Index PICU Course Characteristics by Early Readmission Status
Characteristic . | No Early PICU Readmission, N (%) 379 356 (97.5) . | Early PICU Readmission, N (%) 9863 (2.5) . | P . |
---|---|---|---|
Patient characteristics | |||
Sex | .008 | ||
Female | 170 041 (44.8) | 4295 (43.6) | |
Male | 209 102 (55.2) | 5566 (56.4) | |
Age in years | <.001 | ||
Median (IQR) | 4 (12–1) | 4 (12–0) | |
0–4 | 197 606 (52.1) | 5017 (50.9) | |
5–9 | 57 694 (15.2) | 1452 (14.7) | |
10–14 | 62 120 (16.4) | 1669 (16.9) | |
15–18 | 49 554 (13.1) | 1266 (12.8) | |
19–26 | 12 382 (3.3) | 459 (4.7) | |
Race | <.001 | ||
Non-Hispanic white | 177 104 (46.7) | 4707 (47.7) | |
Non-Hispanic black | 77 101 (20.3) | 1807 (18.3) | |
Hispanic | 73 056 (19.3) | 2047 (20.8) | |
Asian, Indian, or Pacific Islander | 12 301 (3.2) | 349 (3.5) | |
Other or unspecified | 39 794 (10.5) | 953 (9.7) | |
Patient residence | <.001 | ||
Rural | 47 962 (13.0) | 1351 (14.1) | |
Urban | 321 297 (87.0) | 8224 (85.9) | |
Payer | <.001 | ||
Government | 217 014 (57.2) | 6021 (61.0) | |
Private | 139 891 (36.9) | 3263 (33.1) | |
Other | 22 451 (5.9) | 579 (5.9) | |
Complex chronic conditions (CCC) | <.001 | ||
No CCC | 142 630 (37.6) | 1472 (14.9) | |
1 CCC | 103 262 (27.2) | 2251 (22.8) | |
2 CCC | 62 909 (16.6) | 2442 (24.8) | |
3 CCC | 39 786 (10.5) | 1884 (19.1) | |
≥4 CCC | 30 769 (8.11) | 1814 (18.4) | |
Technology dependence | |||
VP shunt | 6346 (1.7) | 390 (4.0) | <.001 |
Tracheostomy | 6289 (1.7) | 437 (4.4) | <.001 |
Othera | 95 486 (25.2) | 3890 (39.4) | <.001 |
Admission characteristics | |||
Seasonality at hospital admission | .8655 | ||
Fall: September – November | 95 334 (25.1) | 2440 (24.7) | |
Winter: December – February | 104 016 (27.4) | 2711 (27.5) | |
Spring: March – May | 94 560 (24.9) | 2464 (25.0) | |
Summer: June – August | 85 446 (22.5) | 2248 (22.8) | |
Transferred from outside facility | 99 869 (26.3) | 2615 (26.5) | .296 |
Admission flow | <.001 | ||
Admit to ICU from ED | 147 363 (38.8) | 3411 (34.6) | |
Admit to ward from ED, transfer to ICU | 44 678 (11.8) | 1544 (15.7) | |
Direct admit to ICU | 160 180 (42.4) | 3579 (36.3) | |
Direct admit to ward, transfer to ICU | 27 135 (7.2) | 1329 (13.5) | |
Day of PICU to ward transfer | <.001 | ||
Weekday | 286 982 (75.6) | 7301 (74.0) | |
Weekend | 92 374 (24.4) | 2562 (26.0) | |
Trauma patient | 25 189 (6.6) | 630 (6.4) | .467 |
Critical care course characteristicsb | |||
Operative procedure | 105 268 (27.7) | 3157 (32.0) | <.001 |
Sedative medication | 185 288 (48.8) | 5695 (57.7) | <.001 |
Paralytic medication | 108 314 (28.6) | 3528 (35.8) | <.001 |
Vasoactive medication | 59 720 (15.7) | 2037 (20.7) | <.001 |
Blood products | 44 967 (11.9) | 1904 (19.3) | <.001 |
Nutrition delivery | |||
GJ, GT, or NG tube | 41 610 (11.0) | 2393 (24.3) | <.001 |
TPN | 34 235 (9.0) | 2478 (25.1) | <.001 |
Treatment of infection | |||
Antibiotic | 232 135 (61.2) | 7119 (72.2) | <.001 |
Antiviral | 16 319 (4.3) | 636 (6.4) | <.001 |
Antifungal | 35 212 (9.3) | 1736 (17.6) | <.001 |
Invasive access | 31 968 (8.4) | 1367 (13.9) | <.001 |
CRRT | 4589 (1.2) | 226 (2.3) | <.001 |
Intubation | 107 842 (28.4) | 3492 (35.4) | <.001 |
ECMO | 1787 (0.5) | 43 (0.4) | .664 |
Cardiopulmonary resuscitation | 7603 (2.0) | 222 (2.3) | .054 |
Characteristic . | No Early PICU Readmission, N (%) 379 356 (97.5) . | Early PICU Readmission, N (%) 9863 (2.5) . | P . |
---|---|---|---|
Patient characteristics | |||
Sex | .008 | ||
Female | 170 041 (44.8) | 4295 (43.6) | |
Male | 209 102 (55.2) | 5566 (56.4) | |
Age in years | <.001 | ||
Median (IQR) | 4 (12–1) | 4 (12–0) | |
0–4 | 197 606 (52.1) | 5017 (50.9) | |
5–9 | 57 694 (15.2) | 1452 (14.7) | |
10–14 | 62 120 (16.4) | 1669 (16.9) | |
15–18 | 49 554 (13.1) | 1266 (12.8) | |
19–26 | 12 382 (3.3) | 459 (4.7) | |
Race | <.001 | ||
Non-Hispanic white | 177 104 (46.7) | 4707 (47.7) | |
Non-Hispanic black | 77 101 (20.3) | 1807 (18.3) | |
Hispanic | 73 056 (19.3) | 2047 (20.8) | |
Asian, Indian, or Pacific Islander | 12 301 (3.2) | 349 (3.5) | |
Other or unspecified | 39 794 (10.5) | 953 (9.7) | |
Patient residence | <.001 | ||
Rural | 47 962 (13.0) | 1351 (14.1) | |
Urban | 321 297 (87.0) | 8224 (85.9) | |
Payer | <.001 | ||
Government | 217 014 (57.2) | 6021 (61.0) | |
Private | 139 891 (36.9) | 3263 (33.1) | |
Other | 22 451 (5.9) | 579 (5.9) | |
Complex chronic conditions (CCC) | <.001 | ||
No CCC | 142 630 (37.6) | 1472 (14.9) | |
1 CCC | 103 262 (27.2) | 2251 (22.8) | |
2 CCC | 62 909 (16.6) | 2442 (24.8) | |
3 CCC | 39 786 (10.5) | 1884 (19.1) | |
≥4 CCC | 30 769 (8.11) | 1814 (18.4) | |
Technology dependence | |||
VP shunt | 6346 (1.7) | 390 (4.0) | <.001 |
Tracheostomy | 6289 (1.7) | 437 (4.4) | <.001 |
Othera | 95 486 (25.2) | 3890 (39.4) | <.001 |
Admission characteristics | |||
Seasonality at hospital admission | .8655 | ||
Fall: September – November | 95 334 (25.1) | 2440 (24.7) | |
Winter: December – February | 104 016 (27.4) | 2711 (27.5) | |
Spring: March – May | 94 560 (24.9) | 2464 (25.0) | |
Summer: June – August | 85 446 (22.5) | 2248 (22.8) | |
Transferred from outside facility | 99 869 (26.3) | 2615 (26.5) | .296 |
Admission flow | <.001 | ||
Admit to ICU from ED | 147 363 (38.8) | 3411 (34.6) | |
Admit to ward from ED, transfer to ICU | 44 678 (11.8) | 1544 (15.7) | |
Direct admit to ICU | 160 180 (42.4) | 3579 (36.3) | |
Direct admit to ward, transfer to ICU | 27 135 (7.2) | 1329 (13.5) | |
Day of PICU to ward transfer | <.001 | ||
Weekday | 286 982 (75.6) | 7301 (74.0) | |
Weekend | 92 374 (24.4) | 2562 (26.0) | |
Trauma patient | 25 189 (6.6) | 630 (6.4) | .467 |
Critical care course characteristicsb | |||
Operative procedure | 105 268 (27.7) | 3157 (32.0) | <.001 |
Sedative medication | 185 288 (48.8) | 5695 (57.7) | <.001 |
Paralytic medication | 108 314 (28.6) | 3528 (35.8) | <.001 |
Vasoactive medication | 59 720 (15.7) | 2037 (20.7) | <.001 |
Blood products | 44 967 (11.9) | 1904 (19.3) | <.001 |
Nutrition delivery | |||
GJ, GT, or NG tube | 41 610 (11.0) | 2393 (24.3) | <.001 |
TPN | 34 235 (9.0) | 2478 (25.1) | <.001 |
Treatment of infection | |||
Antibiotic | 232 135 (61.2) | 7119 (72.2) | <.001 |
Antiviral | 16 319 (4.3) | 636 (6.4) | <.001 |
Antifungal | 35 212 (9.3) | 1736 (17.6) | <.001 |
Invasive access | 31 968 (8.4) | 1367 (13.9) | <.001 |
CRRT | 4589 (1.2) | 226 (2.3) | <.001 |
Intubation | 107 842 (28.4) | 3492 (35.4) | <.001 |
ECMO | 1787 (0.5) | 43 (0.4) | .664 |
Cardiopulmonary resuscitation | 7603 (2.0) | 222 (2.3) | .054 |
CRRT, continuous renal replacement therapy; ECMO, extracorporeal membrane oxygenation; ED, emergency department; GJ, gastrojejunostomy tube; GT, gastrostomy tube; IQR, interquartile range; VP shunt, ventriculoperitoneal shunt.
Gastrostomy tube, dependence on dialysis, cardiac pacemaker, ileostomy.
Refers to index PICU stay.
Statistical Analysis
We presented categorical variables as counts and percentages, age as median with interquartile range, and length of stay (LOS) as geometric mean and geometric SD. We used a generalized linear mixed model with appropriate link function and random intercept for hospitals for univariate and multivariable analysis to identify variables that were associated with early PICU readmission, LOS, and mortality. We used a multilevel approach to control for clustering of patients within hospitals. Variables with a P value < .2 in the univariate analysis were included in the multivariable analysis model. Akaike information criterion was used for model selection. We risk-adjusted LOS and mortality using Hospitalization Resource Intensity Scores for Kids (H-RISK); a resource intensity weight derived from the 2019 Healthcare Cost and Utilization Project Kids’ Inpatient Database.36 A P value < .05 was considered statistically significant. Data analyses were performed using Stata version 15 (StataCorp).
Results
Characteristics of the Study Population
There were 389 219 admissions that met inclusion criteria (Fig 1). The median patient age was 4 years (interquartile range: 1–12) and 55.2% of patients (n = 214 668) were male. Table 1 depicts differences in demographic, admission characteristics, and features of the index critical care course that differ between groups.
Cohort isolation and categorization of subjects by timing of PICU readmission.
Early PICU Readmission Rate
The number of encounters that had a PICU readmission at some point during the hospitalization was 19 137 (4.9%), of which 9863 (51.5%) were early PICU readmissions. The rate of early PICU readmission in the centers contributing to our study varied from 1.1% to 8.8%.
Bivariable Associations of Patient, Admission, and Critical Care Course Characteristics and Likelihood of Early PICU Readmission
There was a significantly greater proportion of patients with 1or more CCCs in the early PICU readmission group (85.1% vs 62.1%, P < .001). Additionally, more patients requiring early readmission had a VP shunt and/or tracheostomy. Elements of the critical care course also varied between groups: more patients with early PICU readmissions required nutrition delivery via GT, GJ, or nasogastric (NG) tube or via total parenteral nutrition (TPN); received sedative, paralytic, and/or vasoactive medications; and were treated for infection with antibiotics, antiviral agents, and/or antifungal medications compared with those that did not require early readmission (Table 1).
Multivariable Associations of Patient, Admission, and Critical Care Course Characteristics and Likelihood of Early PICU Readmission
Patient Characteristics
On multivariable analysis, children with 1 or more CCC had the highest adjusted odds of requiring early PICU readmission (Table 2). The adjusted odds ratio (aOR) of early PICU readmission increased sequentially with each additional CCC (aOR [95% confidence interval, (CI)]: 2.0 [1.89–2.17] for 1 CCC, 3.2 [3.02–3.49] for 2 CCCs, 3.7 [3.40–3.98] for 3 CCCs, 4.2 [3.83–4.53] for ≥4 CCCs; P < .001 for all). Children assisted with medical technology were also at increased risk of early readmissions: patients with VP shunts and tracheostomies had 1.9 (95% CI 1.71–2.17) and 1.4 (95% CI 1.23–1.55) times higher adjusted odds of early PICU readmission, respectively.
Factors Associated With Increased Risk of Early PICU Readmission on Multivariable Analysis
Characteristic . | aOR (95% CI) . | P . |
---|---|---|
Patient characteristics | ||
Sex | ||
Male | Ref | |
Female | 0.93 (0.89–0.97) | .001 |
Age in years | .039 | |
0–4 | Ref | |
5–9 | 0.94 (0.88–1.01) | .075 |
10–14 | 1.06 (1.00–1.13) | .050 |
15–18 | 1.01 (0.94–1.08) | .806 |
19–26 | 1.05 (0.94–1.17) | .394 |
Race | .007 | |
Non-Hispanic white | Ref | |
Non-Hispanic black | 0.91 (0.85–0.97) | .002 |
Hispanic | 0.99 (0.94–1.06) | .866 |
Asian, Indian, or Pacific Islander | 1.05 (0.93–1.18) | .456 |
Other or unspecified | 0.92 (0.85–0.99) | .032 |
Payer | <.001 | |
Government | Ref | |
Private | 0.87 (0.83–0.91) | <.001 |
Other | 0.90 (0.82–0.99) | .023 |
Complex chronic conditions | <.001 | |
No CCC | Ref | |
1 CCC | 2.02 (1.89–2.17) | <.001 |
2 CCC | 3.24 (3.02–3.49) | <.001 |
3 CCC | 3.68 (3.40–3.98) | <.001 |
≥4 CCCs | 4.17 (3.83–4.53) | <.001 |
Technology dependence | ||
VP shunt | 1.93 (1.71–2.17) | <.001 |
Tracheostomy | 1.38 (1.23–1.55) | <.001 |
Admission characteristics | ||
Transferred from outside facility | 1.14 (1.08–1.20) | <.001 |
Admission flow | <.001 | |
Admit to ICU from ED | Ref | |
Admit to ward from ED, transfer to ICU | 1.22 (1.15–1.31) | <.001 |
Direct admit to ICU | 0.86 (0.81–0.91) | <.001 |
Direct admit to ward, transfer to ICU | 1.22 (1.13–1.31) | <.001 |
Day of PICU to ward transfer | <.001 | |
Weekday | Ref | |
Weekend | 1.09 (1.04–1.15) | <.001 |
Trauma patient | 1.13 (1.03–1.24) | .006 |
Critical care course characteristicsa | ||
Operative procedure | 0.75 (0.71–0.80) | <.001 |
Sedative medication | 1.24 (1.17–1.31) | <.001 |
Vasoactive medication | 0.87 (0.82–0.93) | <.001 |
Blood products | 1.15 (1.08–1.23) | <.001 |
Nutrition delivery | ||
GJ, GT, or NG tube | 1.54 (1.46–1.64) | <.001 |
TPN | 2.27 (2.14–2.41) | <.001 |
Intubation | 0.83 (0.79–0.88) | <.001 |
ECMO | 0.41 (0.29–0.56) | <.001 |
Cardiopulmonary resuscitation | 0.79 (0.68–0.92) | .002 |
Characteristic . | aOR (95% CI) . | P . |
---|---|---|
Patient characteristics | ||
Sex | ||
Male | Ref | |
Female | 0.93 (0.89–0.97) | .001 |
Age in years | .039 | |
0–4 | Ref | |
5–9 | 0.94 (0.88–1.01) | .075 |
10–14 | 1.06 (1.00–1.13) | .050 |
15–18 | 1.01 (0.94–1.08) | .806 |
19–26 | 1.05 (0.94–1.17) | .394 |
Race | .007 | |
Non-Hispanic white | Ref | |
Non-Hispanic black | 0.91 (0.85–0.97) | .002 |
Hispanic | 0.99 (0.94–1.06) | .866 |
Asian, Indian, or Pacific Islander | 1.05 (0.93–1.18) | .456 |
Other or unspecified | 0.92 (0.85–0.99) | .032 |
Payer | <.001 | |
Government | Ref | |
Private | 0.87 (0.83–0.91) | <.001 |
Other | 0.90 (0.82–0.99) | .023 |
Complex chronic conditions | <.001 | |
No CCC | Ref | |
1 CCC | 2.02 (1.89–2.17) | <.001 |
2 CCC | 3.24 (3.02–3.49) | <.001 |
3 CCC | 3.68 (3.40–3.98) | <.001 |
≥4 CCCs | 4.17 (3.83–4.53) | <.001 |
Technology dependence | ||
VP shunt | 1.93 (1.71–2.17) | <.001 |
Tracheostomy | 1.38 (1.23–1.55) | <.001 |
Admission characteristics | ||
Transferred from outside facility | 1.14 (1.08–1.20) | <.001 |
Admission flow | <.001 | |
Admit to ICU from ED | Ref | |
Admit to ward from ED, transfer to ICU | 1.22 (1.15–1.31) | <.001 |
Direct admit to ICU | 0.86 (0.81–0.91) | <.001 |
Direct admit to ward, transfer to ICU | 1.22 (1.13–1.31) | <.001 |
Day of PICU to ward transfer | <.001 | |
Weekday | Ref | |
Weekend | 1.09 (1.04–1.15) | <.001 |
Trauma patient | 1.13 (1.03–1.24) | .006 |
Critical care course characteristicsa | ||
Operative procedure | 0.75 (0.71–0.80) | <.001 |
Sedative medication | 1.24 (1.17–1.31) | <.001 |
Vasoactive medication | 0.87 (0.82–0.93) | <.001 |
Blood products | 1.15 (1.08–1.23) | <.001 |
Nutrition delivery | ||
GJ, GT, or NG tube | 1.54 (1.46–1.64) | <.001 |
TPN | 2.27 (2.14–2.41) | <.001 |
Intubation | 0.83 (0.79–0.88) | <.001 |
ECMO | 0.41 (0.29–0.56) | <.001 |
Cardiopulmonary resuscitation | 0.79 (0.68–0.92) | .002 |
ED, emergency department.
Refers to index PICU stay.
Admission Characteristics
Patients who were initially admitted to the PICU as a transfer from the acute care floor were at increased odds of requiring early PICU readmission compared with patients admitted to the PICU directly from the emergency department (Table 2). Other factors associated with increased odds of early PICU readmission included transfer from outside facilities (aOR [95% CI]: 1.1 [1.08–1.20]) or admission caused by trauma (1.1 [1.03–1.24]). Patients transferred out of the PICU on a weekend were also at increased odds of requiring early PICU readmission (1.1 [1.04–1.15]).
The most common admission diagnoses were disorders of the respiratory system (N = 130 514; 33.5% of cohort), nervous system (N = 60 992; 16.1% of cohort), and circulatory system (N = 24 978; 6.4% of cohort). Encounters categorized in the respiratory MDC were significantly less likely to require early PICU readmission (odds ratio 0.8, 95% CI 0.72–0.79), whereas encounters categorized as nervous system and circulatory system MDCs were significantly more likely to require early PICU readmission (odds ratio 1.2, 95% CI 1.17–1.30 and OR 1.2, 95% CI 1.14–1.33, respectively) (Supplemental Table 4).
Critical Care Course
Patients who received TPN during their index PICU stay had significantly higher odds of readmission (aOR [95% CI]: 2.3 [2.14–2.41]). Other factors associated with increased odds of early PICU readmission include receiving nutrition via GJ, GT, or NG tube (1.5 [1.46–1.64]), sedative medications (1.2 [1.17–1.31]), or blood products (1.2 [1.08–1.23]) during the index PICU course. Several features of the critical care course were associated with decreased odds of early PICU readmission: operative procedures (0.8 [0.71–0.81]), vasoactive medications (0.9 [0.82–0.93]), intubation (0.8 [0.79–0.88]), ECMO (0.4 [0.29–0.56]), and experiencing cardiac arrest (0.8 [0.68–0.92]) during the index PICU stay (Table 2).
Multivariable Associations of Early PICU Readmission With Patient Outcomes
Early PICU readmission was associated with 1.7 increased odds of mortality (95% CI 1.54–1.92) and 3 times longer LOS (adjusted incident risk ratio [aIRR] 3.0, 95% CI 2.96–3.08) (Table 3). Length of stay was significantly longer in patients requiring early PICU readmission (geometric mean [geometric SD]: 18.2 [0.9] vs 5.0 [1.1] days, P < .001) even after adjusting for H-RISK (aIRR 2.1, 95% CI 2.03–2.10). Early PICU readmission was associated with increased LOS across all number of CCCs (Table 3). Patients with 4 or more CCCs who required early PICU readmission had an average LOS of 33.3 days, significantly longer than those patients with 4 or more CCCs that did not require early PICU readmission (aIRR 1.7, 95% CI 1.62–1.78).
Patient Outcomes by Early PICU Readmission Status and Number of Complex Chronic Conditions
n (%) . | Length of Stay . | Mortality . | |||||
---|---|---|---|---|---|---|---|
LOS, Days (GM, GSD) . | IRR (95% CI) . | aIRRb (95% CI) . | Mortality, n (%) . | OR (95% CI) . | aORb (95% CI) . | ||
Entire cohort | |||||||
No early PICU readmission | 379 356 (97.5) | 5.0, 1.1 | Ref | Ref | 9188 (2.4) | Ref | Ref |
Early PICU readmission | 9863 (2.5) | 18.2, 0.9 | 3.02 (2.96–3.08) | 2.06 (2.03–2.10) | 395 (4.0) | 1.72 (1.54–1.92) | 0.99 (0.87–1.12) |
No CCC | |||||||
No early PICU readmission | 142 630 (99.0) | 3.4, 0.8 | Ref | Ref | 512 (0.4) | Ref | Ref |
Early PICU readmission | 1472 (1.0) | 10.50, 0.64 | 2.81 (2.70–2.92) | 2.19 (2.12–2.26) | 5 (0.3) | a | a |
1 CCC | |||||||
No early PICU readmission | 103 262 (97.9) | 4.38, 0.96 | Ref | Ref | 1920 (1.9) | Ref | Ref |
Early PICU readmission | 2251 (2.1) | 13.26, 0.72 | 2.69 (2.59–2.79) | 2.14 (2.07–.21) | 36 (1.6) | 0.89 (0.61–1.31) | 0.52 (0.34–0.81) |
2 CCC | |||||||
No early PICU readmission | 62 909 (96.3) | 6.68, 1.06 | Ref | Ref | 2620 (4.2) | Ref | Ref |
Early PICU readmission | 2442 (3.7) | 17.61, 0.75 | 2.30 (2.21–2.39) | 1.95 (1.88–2.02) | 74 (3.0) | 0.67 (0.49–0.91) | 0.47 (0.34–0.65) |
3 CCC | |||||||
No early PICU readmission | 39 786 (95.5) | 8.55, 1.16 | Ref | Ref | 1924 (4.8) | Ref | Ref |
Early PICU readmission | 1884 (4.5) | 23.86, 0.86 | 2.37 (2.25–2.49) | 1.90 (1.82–1.98) | 122 (6.5) | 0.84 (0.41–1.74) | 0.89 (0.46–1.73) |
≥4 CCC | |||||||
No early PICU readmission | 30 769 (94.4) | 12.46, 1.34 | Ref | Ref | 2212 (7.2) | Ref | Ref |
Early PICU readmission | 1814 (5.6) | 33.30, 0.95 | 2.14 (2.02–2.27) | 1.70 (1.62–1.78) | 158 (8.7) | 1.41 (0.87–2.28) | 1.02 (0.63–1.67) |
n (%) . | Length of Stay . | Mortality . | |||||
---|---|---|---|---|---|---|---|
LOS, Days (GM, GSD) . | IRR (95% CI) . | aIRRb (95% CI) . | Mortality, n (%) . | OR (95% CI) . | aORb (95% CI) . | ||
Entire cohort | |||||||
No early PICU readmission | 379 356 (97.5) | 5.0, 1.1 | Ref | Ref | 9188 (2.4) | Ref | Ref |
Early PICU readmission | 9863 (2.5) | 18.2, 0.9 | 3.02 (2.96–3.08) | 2.06 (2.03–2.10) | 395 (4.0) | 1.72 (1.54–1.92) | 0.99 (0.87–1.12) |
No CCC | |||||||
No early PICU readmission | 142 630 (99.0) | 3.4, 0.8 | Ref | Ref | 512 (0.4) | Ref | Ref |
Early PICU readmission | 1472 (1.0) | 10.50, 0.64 | 2.81 (2.70–2.92) | 2.19 (2.12–2.26) | 5 (0.3) | a | a |
1 CCC | |||||||
No early PICU readmission | 103 262 (97.9) | 4.38, 0.96 | Ref | Ref | 1920 (1.9) | Ref | Ref |
Early PICU readmission | 2251 (2.1) | 13.26, 0.72 | 2.69 (2.59–2.79) | 2.14 (2.07–.21) | 36 (1.6) | 0.89 (0.61–1.31) | 0.52 (0.34–0.81) |
2 CCC | |||||||
No early PICU readmission | 62 909 (96.3) | 6.68, 1.06 | Ref | Ref | 2620 (4.2) | Ref | Ref |
Early PICU readmission | 2442 (3.7) | 17.61, 0.75 | 2.30 (2.21–2.39) | 1.95 (1.88–2.02) | 74 (3.0) | 0.67 (0.49–0.91) | 0.47 (0.34–0.65) |
3 CCC | |||||||
No early PICU readmission | 39 786 (95.5) | 8.55, 1.16 | Ref | Ref | 1924 (4.8) | Ref | Ref |
Early PICU readmission | 1884 (4.5) | 23.86, 0.86 | 2.37 (2.25–2.49) | 1.90 (1.82–1.98) | 122 (6.5) | 0.84 (0.41–1.74) | 0.89 (0.46–1.73) |
≥4 CCC | |||||||
No early PICU readmission | 30 769 (94.4) | 12.46, 1.34 | Ref | Ref | 2212 (7.2) | Ref | Ref |
Early PICU readmission | 1814 (5.6) | 33.30, 0.95 | 2.14 (2.02–2.27) | 1.70 (1.62–1.78) | 158 (8.7) | 1.41 (0.87–2.28) | 1.02 (0.63–1.67) |
Compared with patients who did not experience an early PICU readmission, those patients who required early PICU readmission had 3 times longer hospitalizations and were 1.7 times more likely to die during their hospitalization. Patients with medical complexity experiencing early PICU readmission had significantly longer lengths of stay compared with those that were not readmitted across all number of CCCs. Almost 10% of patients with 4 or more CCCs that experienced early PICU readmission did not survive to hospital discharge. CD, calendar days; GM, geometric mean; GSD, geometric SD.
Model did not converge because of small n in readmission group.
Risk-adjusted length of stay (aIRR) and mortality (aOR) using H-RISK.
Discussion
In this retrospective study utilizing the PHIS database, 2.5% of all children admitted to tertiary care children’s hospital PICUs required readmission to the PICU within 2 calendar days of transfer. Patients with medical complexity, including those with VP shunts and/or CCCs, were especially vulnerable. Finally, LOS and mortality were significantly higher in patients requiring early PICU readmission compared with patients who did not require readmission within 2 calendar days.
A significantly higher proportion of CMC experienced early PICU readmission compared with those patients without medical complexity. Moreover, the odds of readmission increased significantly with each additional CCC. Our results demonstrating that CMC have a higher rate of early PICU readmission add to the growing body of literature on the unique considerations regarding the care of this vulnerable patient population.
Patients requiring early readmission to the PICU had longer hospital length of stay and higher mortality rates than those patients who did not require early PICU readmission. This finding is similar to others in the literature; one study found that adults who require ICU readmission at any time during the same hospitalization have almost 6 times the odds of mortality compared with those who do not require readmission, even after adjusting for disease severity.9 In pediatric patients, Kotaskis et al found that mortality was 1.8 times higher for patients with an unplanned readmission to the ICU within 48 hours of transfer.13 Our results support these findings while also exploring how certain admission characteristics and aspects of the critical care course may be associated with outcomes of interest. Although many of these variables are not modifiable, better understanding of their association with outcomes may influence transfer decisions, inform staffing models, and identify opportunities for quality improvement.
Although the prevalence of PICU readmissions is low, these encounters represent resource-intensive hospitalizations that encompass a large proportion of healthcare costs.27 Additionally, the caregiver burden and psychological distress associated with repeat admission to the ICU cannot be overstated, as evidenced by high rates of post-traumatic stress disorder in family members of critically ill children,37,38 with unexpected PICU admissions particularly distressing.39 Given our findings that CMC are at high risk for early PICU readmission and the fact that caregivers of CMC are already highly stressed,40,41 interventions that reduce the frequency of PICU readmission in this population may have dramatic implications for caregiver wellbeing in addition to patient outcomes.
Identification of a population of children at high risk of PICU readmission may enable development of targeted interventions. Early readmissions are more likely to be because of potentially modifiable factors, such as ICU management or transfer decisions.3,8 However, a delicate balance exists between delaying PICU discharge to ensure clinical stability, which may lead to increased ICU length of stay and healthcare utilization, with premature PICU discharge that may result in unplanned PICU readmission after decompensation on the acute care floor. Thus, identification of high-risk patients creates opportunities to be proactive before, during, and after the transfer process. Before transfer, intensivists may be more conservative when it comes to decisions about patients identified as high risk. Additional systems-level modifications could include in-person sign out, transfer checklists, and/or limiting transfer of high-risk patients to daylight hours.42–47 In CMC, parent perception of their child’s health at hospital discharge has been associated with risk of subsequent unplanned hospital readmission.48 Although further studies are needed to determine whether this applies to unplanned PICU readmissions, partnering with families to address caregiver concerns before transfer to the floor may help mitigate readmission risk.
Identifying patients at high risk of PICU readmission can also focus preventative efforts after transfer to the floor. The hospital readmission literature suggests that interventions providing enhanced support after hospital discharge are most effective in reducing readmissions.49–51 Similar post-transfer surveillance, such as intermediate or “step-down” units,52 PICU provider outreach services such as nurse liaisons,53 and/or complex care teams for CMC,54 may improve outcomes in high-risk patients and future research should explore the effect of these measures.
A limitation to our work is the absence of a validated method to identify unplanned versus planned admission events in PHIS. To try to isolate unplanned readmissions, we limited our analysis to patients readmitted to the PICU within 2 days, as it is unlikely that a patient would be transferred from the PICU if it was known that they would need to return within such a short timeframe. Rates of same-hospitalization unplanned PICU readmission within 48 hours in single center studies range from 1.8% to 2.5%,12–14 consistent with our study’s rate of 2.5%. However, despite our strict cutoff, it remains a limitation of our study that we may have inadvertently included planned PICU readmissions in our cohort. Likewise, because of the nature of the PHIS database, we were unable to determine the exact number of hours between transfer to the floor and readmission to the PICU.
Several additional limitations to our study should be acknowledged. Although we examined 4 years’ worth of data, the relatively low incidence rate of our outcome may have limited the power to detect significant associations despite increased risk estimates. Additionally, the PHIS database is an administrative database and therefore clinical information is limited to that captured through billing and claims data. The use of administrative data are potentially subject to errors in coding information and may not contain all clinically relevant features that could influence readmission, such as whether medical technologies were present on transfer or acquired during the PICU course. We were unable to determine causation for any readmissions, and in fact some of the “index” PICU admissions at the beginning of our study period in 2016 may have themselves been readmissions. Further, we lack data on hospital-level resource availability and utilization, such as PICU census and bed availability, use of step-down or intermediate units, physician coverage models, nurse staffing, all of which are likely to influence the decision to transfer a patient from the PICU to the ward. The significant variation in early readmission rates across centers in our study suggests that institutional differences may impact outcomes. These results may serve as a benchmark of early PICU readmission rates and create opportunities for improvement work both within those centers with higher rates and across institutions for best practice collaboration. Relatedly, the institutions represented in our cohort are tertiary-care children’s hospitals in the United States and as a result, our findings may not be representative of sites with different healthcare delivery systems.
Conclusions
In this multicenter study of over 380 000 PICU discharges from tertiary care children’s hospitals, we found that patients with medical complexity have significantly increased risk of readmission to the PICU within 2 calendar days of transfer. Although readmissions are uncommon, only 2.5% of the overall cohort, they are associated with worse outcomes and increased healthcare utilization. These results may inform risk-stratification when deciding when to discharge a patient from the PICU.
FUNDING: This project was supported in part by the National Institutes of Health through Grant UL1-TR-001857. The content is solely the responsibility of the authors and does not necessarily represent the official views of the funding organizations. The funding organizations had no role in the design, preparation, review, or approval of this paper.
CONFLICT OF INTEREST DISCLOSURES: The authors have indicated they have no conflicts of interest relevant to this article to disclose.
Dr Sharp conceptualized and designed the study, led data collection, analysis, and interpretation, and drafted the initial manuscript; Ms Wang developed the data analysis plan and conducted analysis and interpretation of data; Drs Hall and Berry supervised the design of the study and supervised data collection, analysis, and interpretation; Dr Forster contributed to conceptualization and supervised the design of the study and supervised data collection, analysis, and interpretation; and all authors critically reviewed and revised the manuscript, approved the final manuscript as submitted, and agree to be accountable for all aspects of the work.
Comments