Abstract
Objectives
To determine the rate of unplanned PICU readmissions, examine the characteristics of index admissions associated with readmission, and compare outcomes of readmissions versus index admissions.
Design
Retrospective cohort analysis.
Setting
Ninety North American PICUs that participated in the Virtual Pediatric Intensive Care Unit Systems.
Patients
One hundred five thousand four hundred thirty-seven admissions between July 2009 and March 2011.
Interventions
None.
Measurements and Main Results
Unplanned PICU readmission within 48 hours of index discharge was the primary outcome. Summary statistics, bivariate analyses, and mixed-effects logistic regression model with random effects for each hospital were performed. There were 1,161 readmissions (1.2%). The readmission rate varied among PICUs (0–3.3%), and acute respiratory (56%), infectious (35%), neurological (28%), and cardiovascular (20%) diagnoses were often present on readmission. Readmission risk increased in patients with two or more complex chronic conditions (adjusted odds ratio, 1.72; p < 0.001), unscheduled index admission (adjusted odds ratio, 1.37; p < 0.001), and transfer to an intermediate unit (adjusted odds ratio, 1.29; p = 0.004, compared with ward). Trauma patients had a decreased risk of readmission (adjusted odds ratio, 0.67; p = 0.003). Gender, race, insurance, age more than 6 months, perioperative status, and nighttime transfer were not associated with readmission. Compared with index admissions, readmissions had longer median PICU length of stay (3.1 vs 1.7 d, p < 0.001) and higher mortality (4% vs 2.5%, p = 0.002).
Conclusions
Unplanned PICU readmissions were relatively uncommon, but were associated with worse outcomes. Several patient and admission characteristics were associated with readmission. These data help identify high-risk patient groups and inform risk-adjustment for standardized readmission rates.
Keywords: child, hospital readmission, intensive care units, pediatric
Unplanned readmissions have increasingly become a focus of healthcare quality measurement, given that they may be preventable and may reflect threats to patient safety or hospital efficiency (1–3). Similarly, readmissions to ICUs during the same hospitalization have been suggested as an indicator of quality of critical care (4–7). However, determining the appropriate timing for ICU discharge can be a challenge for patients at risk for complications of their resolving critical illness and/or new sequelae of their chronic conditions.
Among adults, multi-institutional studies have found ICU readmission rates of 2–6% (8, 9), depending on the timeframe considered, and that readmissions are associated with greater risk of in-hospital mortality (10, 11). In pediatrics, however, there have only been a few single-institutional studies (12–16). Thus, little generalizable information about the frequency, risk factors, or outcomes of unplanned readmissions is available.
We sought to estimate the rate of early unplanned readmissions to PICUs, using a multi-institutional cohort, and to examine the factors associated with readmission. We also sought to determine whether index admissions and readmissions have different outcomes.
MATERIALS AND METHODS
We performed a retrospective cohort study of consecutive patients admitted between July 2009 and March 2011 to 90 North American PICUs that participated in the Virtual Pediatric Intensive Care Unit Performance Systems (VPS) (VPS, LLC, Los Angeles, CA) (17). No endorsement or editorial restriction of the interpretation of these data or opinions of the authors have been implied or stated. Because all data were deidentified, this study qualified for exemption from review by the Columbia University Institutional Review Board.
Primary Outcome
Our primary outcome was unplanned PICU readmission within 48 hours of discharge from the PICU, during the same and different hospitalizations. This timeframe has been suggested by the Society of Critical Care Medicine (4). However, no consensus exists in the literature about what constitutes “early” ICU readmission; both 48 (9, 14, 16, 18) and 72 (15, 19–21) hours have been used. VPS unique patient identifiers permitted detection of readmission to the same PICU.
Admission Characteristics
We described characteristics of admissions, reporting them as proportions and 95% CIs, means and sds, or medians and interquartile ranges (IQRs). Clinical characteristics included gender, age, race, insurance, presence of complex chronic conditions (CCCs), and baseline and discharge Pediatric Overall Performance Category (POPC) and Pediatric Cerebral Performance Category (PCPC) (22). Because age had a nonlinear relationship with readmission, it was divided into ordinal categories of 2–6 months. A logistic regression was fitted with just this ordinal age variable to determine where the direction and magnitude of associations clustered together. These categories were then collapsed according to this clustering to create a less arduous ordinal age variable (0–2 mo, 2–6 mo, 6–12 mo, 1–18 yr, > 18 yr). CCCs were defined using Feudtner’s definition (23) and identified among diagnoses using a list of VPS codes developed elsewhere (24). An ordinal variable for CCC was created (no CCC, 1 CCC, ≥ 2 CCC). POPC and PCPC scores range from 1 (normal function) to 6 (brain death). Scores of 2, 3, and 4 indicate mild, moderate, and severe disability, respectively. A score of 5 indicates coma or vegetative state. Race, insurance, POPC, and PCPC were optional data elements and were not available for all admissions.
Admission characteristics included planned (yes/no), perioperative (yes/no), trauma (yes/no), patient origin and disposition, total duration of positive pressure ventilation (PPV), discharge season and time, risk of mortality score, and PICU length of stay (LOS). PPV was an ordinal variable, grouped by the cohort’s mean total time on PPV (no time on PPV, < 1.5 d, ≥ 1.5 d). PPV included conventional mechanical ventilation, high-frequency ventilation, and bi-level positive airway pressure via tracheal, mask, or nasal interface. Discharge season was included to control for seasonal illness variation (25–27) and was categorized as September–November, December–February, March–May, and June–August. Two binary discharge time variables were included to explore the relevance of evening (14, 28, 29) and weekend transfers (30). As a proxy for severity of illness at admission, a predicted probability of mortality was estimated using Paediatric Index of Mortality (PIM) 2 (31). Institutional variables included number of licensed PICU beds (≤ 17, 18–24, ≥ 25 beds), presence of a separate intermediate unit, and presence of a pediatric critical care fellowship program.
Statistical Analysis
To examine the characteristics that were associated with increased risk for readmission, index admissions were sorted by whether or not they resulted in a readmission. Bivariate comparisons of admission characteristics were performed using Pearson’s chi-square test, two-tailed t test, or Kruskal-Wallis rank test. Each initial PICU admission during a new hospitalization was treated as a new index admission. Index admissions were excluded if they resulted in death, as there was no possibility of readmission. Subsequent readmissions after the first and during the same hospitalization were excluded. Admissions missing diagnoses were excluded, because we considered chronic conditions particularly relevant to readmission risk.
To estimate the adjusted association between characteristics and unplanned readmission, mixed-effects logistic regression models were fitted. The models included random effects for each hospital to account for correlation of subjects within hospitals. Predictors for the final regression models were included if their p value was less than 0.2 in multivariate analysis. Because of nonlinearity, PIM2 and LOS were transformed into cubic splines. Because PIM2 is derived from categorical and physiologic data at admission to the PICU and is likely attenuated by LOS, an interaction term between PIM2 and LOS was included. C-statistics and Hosmer-Lemeshow goodness-of-fit tests were used to assess discrimination and calibration, respectively.
Three subanalyses were performed. In order to focus on functionality as a predictor, a regression model was fitted on index admissions that supplied discharge POPC and PCPC. Bivariate comparisons of index admissions that did and did not supply functionality scores were performed to examine differences between these subgroups. Also, we refitted the CCC model using 24 and 72 hours as the outcomes, because these times are endorsed by the National Quality Forum (6) and have been used in other studies (15, 18, 19, 21), respectively. To explore the reasons for readmissions, acute and chronic diagnoses that were present at admission were compared using chi-square tests. To compare the outcomes of all index admissions and readmissions, we performed bivariate analyses of admission outcomes (discharge functionality scores, disposition, LOS, duration of PPV, and PICU mortality). For this analysis only, the index admission group included those that resulted in death. We also performed a matched comparison of the readmissions and those index admissions that preceded them, using McNemar’s chi-square test, paired t test, or Wilcoxon signed rank tests.
Statistical significance was determined using a p value of less than 0.05 and 95% CI. Stata 12 (StataCorp, College Station, TX) was used for statistical analyses.
RESULTS
There were 96,189 admissions from 87 PICUs included in this study (Supplemental Fig. 1, Supplemental Digital Content 1, http://links.lww.com/CCM/A714); of these, 1,161 (1.2%) were early unplanned readmissions. The prevalence rate was 12.3 readmissions per 1,000 PICU discharges (95% CI, 11.6–13). The prevalence of readmission varied among PICUs, ranging 0–3.3% (median, 1%; IQR, 0.6–1.5%). The patient, clinical, and institutional characteristics of the index admissions with and without readmissions are presented in Table 1. Index admissions that preceded a readmission involved younger patients (median age, 38 vs 56 mo; p < 0.001) and those with CCC (50% vs 42%, p < 0.001), compared with index admissions that did not precede a readmission. Patients who were readmitted were more commonly discharged to a general ward (78% vs 65%, p < 0.001) or intermediate unit (19% vs 11%, p < 0.001).
Table 1.
Characteristics of PICU patients and their index admissions by readmission status
| Characteristic, % (95% CI) | Not readmitted n=93,867 (98.8%) |
Readmitted <48 hrs n=1,161 (1.2%) |
|---|---|---|
| Male sex | 55.8 (55.5–56.2) | 56 (53.1–58.8) |
| Age | ||
| 0–2 months† | 8.3 (8.1–8.5) | 10.2 (8.5–11.9) |
| 2–6 months* | 8.7 (8.5–8.8) | 14.7 (12.7–16.8) |
| 6–12 months | 7.6 (7.4–7.8) | 8.7 (7.1–10.4) |
| 1–18 years* | 71.1 (70.8–71.4) | 60.8 (58–63.6) |
| >18 years | 4.3 (4.2–4.5) | 5.5 (4.2–6.8) |
| Racea | ||
| Caucasian | 52.1 (51.8–52.5) | 53.7 (50.6–56.9) |
| African American | 18.3 (18–18.6) | 18.6 (16.1–21) |
| Hispanic | 18.1 (17.8–18.4) | 17.4 (15–19.7) |
| Asian/Indian/Pacific Islander | 2.6 (2.5–2.7) | 2.5 (1.5–3.4) |
| American Indian | 0.7 (0.7–0.8) | 0.8 (0.3–1.4) |
| Other/mixed/unspecified | 8.1 (7.8–8.3) | 7.1 (4.8–9.4) |
| Insuranceb | ||
| Medicaid/Medicare | 49.7 (49.2–50.2) | 53.5 (49.2–57.8) |
| Commercial | 23.1 (22.7–23.5) | 22.8 (19.1–26.4) |
| Managed Care | 17.6 (17.2–18) | 15.4 (12.2–18.5) |
| Government | 3.7 (3.6–3.9) | 2.9 (1.5–4.4) |
| Self-Pay/other | 5.8 (5.5–6.1) | 5.4 (2.6–8.2) |
| Complex chronic condition* | ||
| No CCC* | 58 (57.7–58.4) | 50 (47.1–52.8) |
| 1 CCC | 24.5 (24.2–24.7) | 25.5 (23–28) |
| ≥2 CCC* | 17.5 (17.3–17.7) | 24.5 (22–27) |
| Baseline POPC Scorec | ||
| No disability* | 48.6 (48–49.1) | 37.1 (32–42.2) |
| Mild disability | 24.5 (24–24.9) | 27 (22.3–31.7) |
| Moderate disability | 17.9 (17.5–18.4) | 23 (18.5–27.4) |
| Severe disability | 8.8 (8.5–9.1) | 12.6 (9.1–16.2) |
| Coma/vegetative state | 0.2 (0.1– 0.2) | 0.3 (0–0.9) |
| Baseline PCPC Scorec | ||
| No disability† | 73.1 (72.6–73.6) | 66.7 (61.9–71.6) |
| Mild disability | 11.5 (11.1–11.8) | 14.9 (11.2–18.7) |
| Moderate disability | 8.6 (8.2–8.9) | 8.6 (5.7–11.6) |
| Severe disability | 6.7 (6.4–6.9) | 9.5 (6.4–12.6) |
| Coma/vegetative state | 0.2 (0.1–0.2) | 0.3 (0–0.9) |
| Unplanned* | 72.1 (71.8–72.4) | 78 (75.6–80.3) |
| Pre-/post-operative | 33.5 (33.2–33.8) | 30.9 (28.3–33.6) |
| Origin | ||
| Emergency Department† | 48.7 (48.4–49) | 43.8 (40.9–46.6) |
| General ward* | 11.5 (11.3–11.7) | 19 (16.8–21.3) |
| Another ICU† | 3.3 (3.2–3.5) | 4.9 (3.7–6.2) |
| Intermediate unit | 1.4 (1.3–1.5) | 2 (1.2–2.8) |
| OR/PACU/Procedure suite† | 31.1 (30.8–31.4) | 27.3 (24.8–30) |
| Chronic/Rehabilitation facility | 0.2 (0.2–0.3) | 0.3 (0–0.6) |
| Outpatient/Home | 3.3 (3.2–3.4) | 2.3 (1.5–3.2) |
| Other | 0.5 (0.4–0.5) | 0.2 (0–0.6) |
| Trauma* | 8.1 (8–8.3) | 5.3 (4–6.6) |
| Duration of PPV, days, mean (SD)† | 1.5 (6.9) | 2 (5.4) |
| Disposition | ||
| General ward* | 64.9 (64.6–65.2) | 77.9 (75.4–80.3) |
| Another ICU† | 2.3 (2.3–2.4) | 1.5 (0.8–2.2) |
| Intermediate unit* | 11.1 (10.9–11.3) | 19.4 (17.1–21.7) |
| OR | 0.5 (0.4–0.5) | 0.5 (0.1–0.9) |
| Chronic/Rehabilitation facility | 0.9 (0.9–1) | 0.5 (0.1–0.9) |
| Home* | 19.4 (19.1–19.6) | 0.1 (0–0.3) |
| Other† | 1 (0.9–1) | 0.2 (0–0.4) |
| Discharge season | ||
| September – November | 30.7 (30.3–30.9) | 28.9 (26.3–31.6) |
| December – February | 26 (25.8–26.3) | 27.7 (25.2–30.3) |
| March – May† | 17.9 (17.6–18.1) | 20.2 (17.8–22.5) |
| June – August | 25.5 (25.2–25.8) | 23.2 (20.7–25.6) |
| Discharged between 5 PM & 8 AM† | 29.5 (29.2–29.8) | 33.6 (30.9–36.3) |
| Weekend discharge† | 24.7 (24.5–25) | 22.2 (19.8–24.5) |
| Admitted to PICU with | ||
| ICU beds | ||
| ≤17 | 47.9 (47.6–48.2) | 46.3 (43.4–49,2) |
| 18-24* | 18.3 (18.1–18.5) | 12.5 (10.6–14.4) |
| ≥25* | 33.8 (33.5–34.1) | 41.2 (38.3–44) |
| Separate intermediate care unit† | 40.1 (39.7–40.4) | 44.8 (41.9–47.7) |
| Affiliated fellowship program* | 58.6 (58.3–59) | 69.6 (66.9–72.2) |
| PIM2 risk of mortality, %, mean (SD)* | 1.8 (5.3) | 2.7 (6.8) |
| LOS, days, median (IQR)* | 1.6 (0.9–3.6) | 2 (1–5.1) |
CCC, complex chronic condition; CI, confidence interval; PACU, post-anesthesia care unit; ICU, intensive care unit; IQR, interquartile range; LOS, length of stay; OR, operating room; PIM, Paediatric Index of Mortality; PCPC, Pediatric Cerebral Performance Categories; PICU, pediatric intensive care unit; POPC, Pediatric Overall Performance Categories; PPV, positive pressure ventilation; SD, standard deviation
P < 0.05;
P < 0.001 by chi2 test, unpaired two-tailed t-test, or Kruskal-Wallis rank test
Based on 77,954 (80%) admissions from 68 sites that supplied race
Based on 42,825 (44%) admissions from 28 sites that supplied insurance
Based on 31,434 (33%) admissions from 27 sites that supplied Pediatric Overall and Cerebral Performance Categories
The mixed-effects CCC model fitted on the entire cohort (Table 2) had reasonable discrimination and calibration—C-statistic 0.72 (95% CI, 0.71–0.74), Hosmer-Lemeshow p value 0.2. Significant predictors included age 2–6 months (compared with age 0–2 mo), having one or more CCC, and having an unscheduled index admission. Similarly, patients discharged to an intermediate unit were more likely to return, compared with those sent to general wards. In contrast, being admitted for trauma and being cared for in a moderate-sized PICU (18–24 beds) were associated with decreased odds of readmission. Patients discharged to chronic care or rehabilitation facilities and home also had significantly lower odds of returning than those discharged to general wards. Weekend discharge was negatively associated with readmission. Age more than 6 months was not associated with readmission. Gender, race, insurance, perioperative status, patient origin, nighttime transfer, and presence of a separate intermediate unit were left out of the final model (p > 0.2).
Table 2.
Adjusted odds ratios for predictor variables in models of early unplanned PICU readmissiona
| CCC model |
Functionality modelb |
|||
|---|---|---|---|---|
| Patient or index admission variable |
OR (95% CI) | P value | OR (95% CI) | P value |
| Age | ||||
| 0–2 months (reference) | ||||
| 2–6 months | 1.47 (1.16–1.87) | 0.002 | 1.27 (0.82–1.96) | 0.29 |
| 6–12 months | 1.02 (0.78–1.34) | 0.86 | 0.81 (0.49–1.34) | 0.41 |
| 2–6 months | 0.83 (0.68–1.02) | 0.08 | 0.69 (0.47–1.01) | 0.06 |
| >18 years | 1.06 (0.77–1.45) | 0.73 | 0.75 (0.4–1.39) | 0.36 |
| Complex chronic condition | ||||
| No CCC (reference) | ||||
| 1 CCC | 1.22 (1.05–1.41) | 0.009 | – | |
| ≥2 CCC | 1.72 (1.46–2.01) | <0.001 | – | |
| Discharge POPC Score | ||||
| No disability (reference) | ||||
| Mild disability | – | 1.98 (1.45–2.7) | <0.001 | |
| Moderate disability | – | 2.44 (1.65–3.63) | <0.001 | |
| Severe disability | – | 3.24 (1.61–6.5) | 0.001 | |
| Coma/vegetative state | – | 2.05 (0.26–16.3) | 0.59 | |
| Discharge PCPC Scorec | ||||
| No disability (reference) | ||||
| Mild disability | – | 1.12 (0.81–1.55) | 0.49 | |
| Moderate disability | – | 0.8 (0.52–1.25) | 0.34 | |
| Severe disability | – | 0.76 (0.37–1.59) | 0.47 | |
| Unplanned index admission | 1.37 (1.17–1.61) | <0.001 | 1.75 (1.3–2.37) | <0.001 |
| Trauma | 0.67 (0.51–0.87) | 0.003 | 0.79 (0.5–1.24) | 0.31 |
| Duration of PPV, days | ||||
| Not on PPV (reference) | ||||
| <1.5 | 0.77 (0.63–0.95) | 0.016 | 1.0 (0.69–1.44) | 0.99 |
| ≥1.5 | 1.12 (0.9–1.4) | 0.29 | 0.94 (0.63–1.42) | 0.78 |
| Disposition | ||||
| General ward (reference) | ||||
| Another ICU | 0.37 (0.22–0.6) | <0.001 | 0.56 (0.27–1.16) | 0.11 |
| Intermediate unit | 1.29 (1.08–1.53) | 0.004 | 1.59 (1.09–2.32) | 0.017 |
| Operating room | 0.66 (0.29–1.5) | 0.32 | 0.56 (0.14–2.31) | 0.43 |
| Chronic/Rehabilitation facility | 0.37 (0.16–0.85) | 0.018 | (0–∞) | 0.99 |
| Home | 0.004 (0.001–0.03) | <0.001 | 0.01 (0.002–0.08) | <0.001 |
| Other | 0.16 (0.41–0.66) | 0.011 | (0–∞) | 0.99 |
| Discharge season | ||||
| September – November (reference) | ||||
| December – February | 1.09 (0.93–1.27) | 0.29 | 0.86 (0.65–1.15) | 0.32 |
| March – May | 1.19 (1.00–1.41) | 0.044 | 1.18 (0.87–1.58) | 0.29 |
| June – August | 0.99 (0.84–1.16) | 0.87 | 0.83 (0.62–1.11) | 0.21 |
| Weekend discharge | 0.84 (0.73–0.97) | 0.018 | 0.74 (0.57–0.97) | 0.029 |
| Admitted to PICU with | ||||
| ICU beds | ||||
| ≤17 (reference) | ||||
| 18–24 | 0.64 (0.5–0.82) | <0.001 | 0.48 (0.31–0.74) | 0.001 |
| ≥25 | 1.02 (0.83–1.27) | 0.83 | 1.3 (0.93–1.82) | 0.13 |
| Affiliated fellowship program | 1.25 (1.03–1.53) | 0.027 | 0.65 (0.44–0.95) | 0.026 |
CCC, complex chronic condition; CI, confidence interval; PACU, post-anesthesia care unit; ICU, intensive care unit; IQR, interquartile range; OR, odds ratio; PICU, pediatric intensive care unit; PPV, positive pressure ventilation
Models also adjusted for length of stay and PIM2 score as cubic splines and for an interaction between them
Based on 31,434 (33%) index admissions from 27 sites that supplied POPC and PCPC
PCPC level 5 (coma/vegetative state) was perfectly collinear with POPC level 5 (also coma/vegetative state), which is a testament to the accuracy of the submitted data. Because of this collinearity, one or the other had to be eliminated; we chose PCPC level 5 to omit.
In the subgroup analysis of 31,434 admissions (33%) with discharge POPC and PCPC (Supplemental Table 1, Supplemental Digital Content 1, http://links.lww.com/CCM/A714; Table 2), progressively worse overall disability was associated with increasing odds of readmission, with the exception of coma/vegetative states (POPC 5). All levels of cerebral disability were not associated with readmission, after adjusting for overall disability scores and other covariates. Age 2–6 months and trauma were not significant predictors. Having an unscheduled index admission and being discharged to an intermediate care unit remained predictors. Being in a PICU with a fellowship became a negative predictor. This model also discriminated and calibrated well—C-statistic 0.77 (95% CI, 0.74–0.79); Hosmer-Lemeshow p value 0.53.
The 24- and 72-hour models (Supplemental Table 2, Supplemental Digital Content 1, http://links.lww.com/CCM/A714) were similar to the 48-hour model with a few exceptions. In the 24-hour model, having one CCC, trauma, and discharge to a chronic facility were not associated with readmissions. Weekend discharge was not associated with 24- or 72-hour readmission. Being on PPV for less than 1.5 days was negatively associated with 48- and 72-hour readmission, but being on PPV for greater than or equal to 1.5 days was positively associated with 24-hour readmission.
The PICU outcomes of unplanned readmissions and the matched and unmatched index admissions are presented in Table 3. Readmissions had longer median LOS (3.1 d) compared with both all index admissions (1.7 d) and their preceding admission (2 d) (p < 0.001). Readmissions had lower mean PIM2 risk of mortality compared with their matched preceding index admissions (2.2% vs 2.7%, p = 0.04). However, readmissions had a higher mortality rate (4%) compared with index admissions (2.5%, p = 0.002).
Table 3.
Comparison of outcomes of index admissions and early unplanned readmissions
| Characteristic, % (95% CI) | All index admissions n=97,527a |
Index admissions that preceded unplanned readmission n=1,161 |
Unplanned readmissions n=1,161 |
|---|---|---|---|
| Discharge POPC Scoreb | |||
| No disability‡* | 39.9 (39.4–40.5) | 23.6 (19.1–28.1) | 19.8 (15.6–24) |
| Mild disability | 29.3 (28.8–29.8) | 38.6 (33.5–43.8) | 34.5 (29.5–39.5) |
| Moderate disability‡* | 18.8 (18.3–19.2) | 25.1 (24–32.9) | 29.6 (24.8–34.4) |
| Severe disability | 9 (8.7–9.4) | 12.4 (9.9–15.9) | 11.8 (8.4–15.2) |
| Coma/vegetative state | 0.3 (0.2–0.4) | 0.3 (0–0.9) | 0.3 (0–0.9) |
| Brain death | 2.6 (2.5–2.8) | – | 4 (1.9–6.1) |
| Discharge PCPC Scoreb | |||
| No disability‡O | 68.9 (68.3–69.4) | 63.1 (58–68.2) | 55.1 (49.9–60.4) |
| Mild disability‡ | 12.5 (12.1–12.9) | 17.3 (13.3–21.3) | 21 (16.7–25.3) |
| Moderate disability | 8.9 (8.6–9.2) | 10.1 (6.9–13.3) | 12.4 (8.9–15.8) |
| Severe disability* | 6.8 (6.5–7.1) | 9.2 (6.2–12.3) | 7.2 (4.5–9.9) |
| Coma/vegetative state | 0.3 (0.2–0.4) | 0.3 (0–0.9) | 0.3 (0–0.9) |
| Brain death | 2.6 (2.5–2.8) | – | 4 (1.9–6.1) |
| Disposition | |||
| General ward†O | 63.4 (63.1–63.7) | 77.9 (75.4–80.3) | 66.8 (64.1–69.6) |
| Another ICU | 2.3 (2.2–2.4) | 1.5 (0.8–2.2) | 1.9 (1.1–2.7) |
| Intermediate unit‡ | 10.9 (10.7–11.1) | 19.4 (17.1–21.7) | 18.3 (16–20.5) |
| OR† | 0.4 (0.4–0.5) | 0.5 (0.1–0.9) | 0.9 (0.3–1.4) |
| Chronic/Rehabilitation facility | 0.9 (0.8–0.9) | 0.5 (0.1–0.9) | 1 (0.4–1.6) |
| Home‡O | 18.6 (18.4–18.9) | 0.1 (0–0.3) | 7.1 (5.6–8.5) |
| Morgue/Medical examiner† | 2.5 (2.4–2.6) | – | 4 (2.8–5.1) |
| Other | 1 (0.9–1.1) | 0.2 (0–0.4) | 0.1 (0–0.4) |
| LOS, days, median (IQR)‡O | 1.7 (0.9–3.8) | 2 (1–5.1) | 3.1 (1.8–6.2) |
| Duration of PPV, days, mean (SD) | 1.5 (8) | 2 (5.4) | 2 (6.4) |
| PIM2 risk of mortality, %, mean (SD)* | 2.5 (8.9) | 2.7 (6.8) | 2.2 (6.9) |
| PICU mortality† | 2.5 (2.4–2.6) | – | 4 (2.8–5.1) |
CI, confidence interval; PACU, post-anesthesia care unit; ICU, intensive care unit; IQR, interquartile range; LOS, length of stay; OR, operating room; PIM, Paediatric Index of Mortality; PCPC, Pediatric Cerebral Performance Categories; POPC, Pediatric Overall Performance Categories; PPV, positive pressure ventilation
P < 0.05;
P < 0.001 comparing all index admissions and readmissions by chi2 test, unpaired two-tailed t-test, or Kruskal-Wallis rank test
P < 0.05;
P < 0.001 comparing index admissions that led to readmissions and their matched readmission by McNemar’s chi2 test, paired t-test, or Wilcoxon signed-rank test
Includes 95,028 index admissions in other analyses and 2,486 index admissions that resulted in death
Based on 32,228 (33%) index admissions of patients (31,434 in other analyses and 854 that resulted in death) and 348 (30%) readmissions from 27 sites that supplied POPC and PCPC
Acute diagnoses that were present at admission are presented in Table 4. Respiratory and neurological diagnoses were 10% and 7%, respectively, more common in readmissions compared with matched index admissions (p < 0.001). Adverse events or complications were present in 6% of both index admissions and readmissions. Chronic diagnoses of our cohort sorted by readmission status are found in Supplemental Table 3 (Supplemental Digital Content 1, http://links.lww.com/CCM/A714). Several chronic diagnoses were found in greater proportions of readmitted patients, although diabetes and asthma were found less frequently in readmitted patients.
Table 4.
Acute diagnoses present on index admission and on unplanned readmissions among patients with an unplanned readmission
| Diagnosis, % (95% CI)a | Index admissions that preceded readmission n=1,161 |
Unplanned readmissions n=1,161 |
|---|---|---|
| Adverse event or complication | 6.1 (4.7–7.5) | 6.4 (5–7.8) |
| Arrest, cardiac or respiratory | 4.8 (3.6–6.1) | 4.2 (3.1–5.4) |
| Cardiovascular, any | 17.1 (15–19.3) | 19.8 (17.5–22.1) |
| Arrhythmia | 3.1 (2.1–4.1) | 4.7 (3.4–5.9) |
| Heart failure | 2.8 (1.9–3.8) | 3.5 (2.5–4.6) |
| Hypertension, systemic† | 2.2 (1.3–3) | 3.8 (2.7–4.9) |
| Hypotension without shock | 2 (1.2–2.9) | 2.3 (1.4–3.2) |
| Shock | 2.9 (2–3.9) | 2.7 (1.7–3.9) |
| Fluid/electrolyte abnormality | 9.6 (7.9–11.3) | 9.2 (7.5–10.9) |
| Gastrointestinal, any | 13.4 (11.5–15.4) | 14.3 (12.3–16.3) |
| Liver failure | 0.6 (0.2–1) | 0.4 (0.1–0.8) |
| Infection, any | 34.2 (31.5–36.9) | 35.8 (33.1–38.6) |
| CNS | 3.2 (2.2–4.2) | 3.4 (2.4–4.5) |
| Gastrointestinal | 1.1 (0.5–1.7) | 1 (0.4–1.5) |
| Respiratory | 23.6 (21.2–26) | 24.1 (21.7–26.6) |
| Sepsis, bacteremia, viremia, fungemia | 6.2 (4.8–7.6) | 6 (4.7–7.4) |
| Neurological, any* | 20.7 (18.3–23) | 27.7 (25.2–30.3) |
| Cerebrovascular event | 3.2 (2.2–4.2) | 4.7 (3.5–6) |
| Encephalopathy† | 6.3 (4.9–7.7) | 9.3 (7.6–11) |
| Increased intracranial pressure/cerebral edema | 2.5 (1.6–3.4) | 2.2 (1.3–3) |
| Seizure* | 13.2 (11.2–15.1) | 19.6 (17.3–21.8) |
| Renal, any | 5.2 (3.9–6.4) | 6.5 (5–7.9) |
| Renal failure | 2.4 (1.5–3.3) | 2.6 (1.7–3.5) |
| Respiratory, any* | 45.5 (42.6–48.3) | 55.6 (52.8–58.5) |
| ARDS/ALI | 18 (15.8–20.2) | 15.8 (13.7–18) |
| Bronchospasm | 4.1 (3–5.3) | 3.6 (2.5–4.7) |
| Pneumothorax/pleural effusion† | 4.4 (3.2–5.6) | 7.2 (5.7–8.7) |
| Upper airway obstruction† | 3.4 (2.4–4.5) | 5.5 (4.2–6.8) |
ALI, acute lung injury; ARDS, acute respiratory distress syndrome; CNS, central nervous system; CI, confidence interval
P < 0.05;
P < 0.001 by chi2 test
Patients may have had more than one acute diagnosis
DISCUSSION
To our knowledge, this is the first study to estimate the frequency of early unplanned readmissions to PICUs, explore the factors associated with higher odds of readmission, and compare outcomes using a multi-institutional cohort. We demonstrated that readmissions were relatively uncommon but had significantly longer PICU LOS and higher mortality compared with index admissions. Furthermore, we identified several characteristics that were independently associated with readmission.
Earlier studies of PICU readmission reported data from single institutions. Two such studies reported that 7–8% of patients had planned or unplanned readmissions within any timeframe during the hospitalization (12, 13). Bernard and Czaja (16) found that 1.8% of their patients had unplanned readmissions within 48 hours; Bastero-Miñón et al (15) found that 2.4% of pediatric cardiac ICU patients had unplanned readmissions within 72 hours.
Our regression analyses went beyond those earlier studies by adding institutional covariates, incorporating more clinical variables (e.g., CCC, functionality scores, disposition location, and PICU characteristics), using splines to better model risk of mortality scores and LOS, and including interactions. Several of these variables were found to be significant predictors of readmission. Patients with greater than or equal to two CCC were especially likely to return. Those discharged to intermediate units were also at increased risk; intermediate units likely cared for patients whose critical illness had not fully resolved and were at risk for deterioration. In contrast, patients discharged directly home were very unlikely to return in 48 hours. Although we could not assess if these patients required care again at an emergency department or were readmitted to a general ward or to another PICU, this may suggest that ICU teams were good at determining when patients’ clinical condition did not warrant further inpatient care. Waiting for this level of clinical stability may have resulted in longer PICU LOS for these patients. Patients who were on PPV for shorter periods were less likely to be readmitted than those who did not receive PPV, and being on PPV for longer periods was not associated with readmission. This is possibly because patients previously on PPV were observed for longer periods after weaning and before discharge. In all models, patients cared for in PICUs with 18–24 beds were significantly less likely to be readmitted compared with smaller units. However, being in a larger PICU (≥25 beds) did not confer any decreased risk. Having a fellowship program was a positive predictor of readmission in all models incorporating CCC.
In our model incorporating functionality at PICU discharge, increasing overall disability was associated with increasing odds of readmission, with the exception of coma or vegetative state. This may reflect the impact of disabilities on patients’ need for highly specialized care and vulnerability even after discharge, which has been noted in studies of readmissions to children’s hospitals (32–34). Patients with CCC or substantial disability are likely at risk for new complications of their underlying states and take longer to fully recover from critical illness. After adjusting for functionality, there were notable changes in magnitude and direction of association of some of our predictors compared with the CCC model. For instance, having a fellowship program was a negative predictor. These variations in the associations are possibly related to POPC and PCPC being a more granular assessment of chronic condition status, PICU management idiosyncrasies, and/or case-mix differences that are not fully explained in the model but manifest when analyzing the POPC/PCPC subgroup (Supplemental Table 3, Supplemental Digital Content 1, http://links.lww.com/CCM/A714). Differences in some predictors’ direction of association with 24-, 48-, and 72-hour readmission may reflect variation in the types of patients who are readmitted in those timeframes and/or variation in discharge criteria for different patients.
Examining acute diagnoses present on readmission, respiratory, infectious, and neurological conditions were especially common (all > 25%), suggesting that these conditions were the reasons for most readmissions. VPS coders may designate a diagnosis as the primary diagnosis for admission. However, we chose not to focus on this designation because some patients had more than one primary diagnosis, and we believe it is often difficult to discern a single reason for ICU admission (e.g., should one code the respiratory distress or the pneumonia as primary?). This is especially true in PICU patients who commonly have multiple, interrelated reasons for admission. Interestingly, no acute adverse event or complication was listed as the primary diagnosis present on readmission, and the proportion of index and readmissions with these diagnoses was similar. Unless these diagnoses were undercoded, this may suggest that adverse events were not large drivers of readmission. Likely some readmissions could be anticipated, but prospective research on the reasons for unplanned readmissions is needed to assess their preventability. Higher mortality rates and longer LOS among unplanned PICU readmissions also give impetus to better understand why they occur.
This study has several limitations. First, because it is retrospective and VPS lacks non-PICU data, we were unable to evaluate the extent to which readmissions were preventable. Second, some potential confounders could not be analyzed (e.g., discharge severity of illness and unit census at the time of discharge). Additionally, we were unable to study differences among PICUs in their tendencies to transfer or reaccept patients and how these differences impact readmissions. This confounding is somewhat mitigated using a mixed-effects model that incorporates within-unit correlations. Third, adverse event diagnoses could have been coded for care other than that immediately preceding readmission. So besides describing how common adverse events are, we do not know how they contributed to readmissions. Fourth, patients may have been readmitted to a different PICU or have unexpectedly died outside the PICU (though this number would presumably be small). Lastly, while our models’ accuracy was not exceptional, the C-statistics were higher than those reported for most adult multi-institutional ICU readmission studies (9, 17, 21).
CONCLUSIONS
The decision of when to discharge patients from the PICU can be a challenging balance of benefits (e.g., longer ICU observation) and burdens (e.g., unnecessary costs and greater risk of ICU-related infections) for patients who may still have complex needs. Intensivists at each PICU are charged with striking the proper balance for their patients, but pressure to avoid unplanned readmissions is rising. Knowing the frequency of early unplanned readmissions permits physicians to better understand how their unit compares to a larger group of PICUs. Furthermore, identifying high-risk patient groups (e.g., those with CCC or functional limitations at discharge) and admission characteristics (e.g., unscheduled admission and intermediate unit disposition) can guide risk-adjustment and future predictive scoring systems of PICU readmission (21, 35, 36). Finally, this knowledge could direct interventions to decrease unplanned readmissions.
Supplementary Material
ACKNOWLEDGMENTS
We thank the Virtual Pediatric Intensive Care Unit Performance Systems for providing the data for this study. We thank Haomiao Jia, PhD, for reviewing our statistical methods and analysis.
Supported, in part, by Pediatric Critical Care Scientist Development Program. Dr. Edwards was supported by a National Institutes of Health (NIH) Pediatric Critical Care Scientist Development Program grant (K12 HD047349) during the time of this study. Dr. Stone’s institution received grant support from NIH. Dr. Stone is employed by Columbia University.
Footnotes
Supplemental digital content is available for this article. Direct URL citations appear in the printed text and are provided in the HTML and PDF versions of this article on the journal’s website (http://journals.lww.com/ccmjournal).
The remaining authors have disclosed that they do not have any potential conflicts of interest.
REFERENCES
- 1. [Accessed July 13, 2012];National Quality Forum: NQF Endorses All-Cause Unplanned Readmissions Measures. Available at: http://www.qualityforum.org/News_And_Resources/Press_Releases/2012/NQF_Endorses_All-Cause_Unplanned_Readmissions_Measures.aspx.
- 2. [Accessed July 11, 2012];Partnership for Patients: Better Care, Lower Costs. 2011 Available at: http://www.healthcare.gov/news/factsheets/2011/04/partnership04122011a.html.
- 3. [Accessed July 11, 2012];CHIPRA Measures by CHIPRA Categories: Initial Core Set and PQMP CoE Measure Assignments. 2012 Available at: http://www.ahrq.gov/chipra/pqmpmeasures.html.
- 4.SCCM Quality Indicators Committee: Candidate Critical Care Quality Indicators. Anaheim, CA: Society of Critical Care Medicine; 1995. [Google Scholar]
- 5.Lacey S, Smith JB, Cox KS. Pediatric safety and quality. In: Hughes RG, editor. Patient Safety and Quality: An Evidence-Based Handbook for Nurses. Rockville, MD: AHRQ Publication; 2008. pp. 1–30. [Google Scholar]
- 6. [Accessed August 31, 2012];National Quality Forum: NQF Endorses Pulmonary and Critical Care Measures. 2012 Available at: http://www.qualityforum.org/News_And_Resources/Press_Releases/2012/NQF_Endorses_Pulmonary_and_Critical_Care_Measures.aspx.
- 7.Berenholtz SM, Dorman T, Ngo K, et al. Qualitative review of intensive care unit quality indicators. J Crit Care. 2002;17:1–12. doi: 10.1053/jcrc.2002.33035. [DOI] [PubMed] [Google Scholar]
- 8.Kramer AA, Higgins TL, Zimmerman JE. Intensive care unit readmissions in U.S. hospitals: Patient characteristics, risk factors, and outcomes. Crit Care Med. 2012;40:3–10. doi: 10.1097/CCM.0b013e31822d751e. [DOI] [PubMed] [Google Scholar]
- 9.Brown SE, Ratcliffe SJ, Kahn JM, et al. The epidemiology of intensive care unit readmissions in the United States. Am J Respir Crit Care Med. 2012;185:955–964. doi: 10.1164/rccm.201109-1720OC. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Rosenberg AL, Watts C. Patients readmitted to ICUs: A systematic review of risk factors and outcomes. Chest. 2000;118:492–502. doi: 10.1378/chest.118.2.492. [DOI] [PubMed] [Google Scholar]
- 11.Chen LM, Martin CM, Keenan SP, et al. Patients readmitted to the intensive care unit during the same hospitalization: Clinical features and outcomes. Crit Care Med. 1998;26:1834–1841. doi: 10.1097/00003246-199811000-00025. [DOI] [PubMed] [Google Scholar]
- 12.Odetola FO, Clark SJ, Dechert RE, et al. Going back for more: An evaluation of clinical outcomes and characteristics of readmissions to a pediatric intensive care unit. Pediatr Crit Care Med. 2007;8:343–347. doi: 10.1097/01.PCC.0000269400.67463.AC. [DOI] [PubMed] [Google Scholar]
- 13.Kanter RK. Post-intensive care unit pediatric hospital stay and estimated costs. Crit Care Med. 2000;28:220–223. doi: 10.1097/00003246-200001000-00036. [DOI] [PubMed] [Google Scholar]
- 14.Linton S, Grant C, Pellegrini J, et al. The development of a clinical markers score to predict readmission to paediatric intensive care. Intensive Crit Care Nurs. 2009;25:283–293. doi: 10.1016/j.iccn.2009.07.003. [DOI] [PubMed] [Google Scholar]
- 15.Bastero-Miñón P, Russell JL, Humpl T. Frequency, characteristics, and outcomes of pediatric patients readmitted to the cardiac critical care unit. Intensive Care Med. 2012;38:1352–1357. doi: 10.1007/s00134-012-2592-2. [DOI] [PubMed] [Google Scholar]
- 16.Bernard AM, Czaja AS. Unplanned pediatric intensive care unit readmissions: A single-center experience. J Crit Care. 2013 Apr 16; doi: 10.1016/j.jcrc.2013.02.008. [Epub ahead of print] [DOI] [PubMed] [Google Scholar]
- 17.Wetzel RC, Sachedeva R, Rice TB. Are all ICUs the same? Paediatr Anaesth. 2011;21:787–793. doi: 10.1111/j.1460-9592.2011.03595.x. [DOI] [PubMed] [Google Scholar]
- 18.Campbell AJ, Cook JA, Adey G, et al. Predicting death and readmission after intensive care discharge. Br J Anaesth. 2008;100:656–662. doi: 10.1093/bja/aen069. [DOI] [PubMed] [Google Scholar]
- 19.Lee JY, Park SK, Kim HJ, et al. Outcome of early intensive care unit patients readmitted in the same hospitalization. J Crit Care. 2009;24:267–272. doi: 10.1016/j.jcrc.2007.12.019. [DOI] [PubMed] [Google Scholar]
- 20.Ho KM, Dobb GJ, Lee KY, et al. The effect of comorbidities on risk of intensive care readmission during the same hospitalization: A linked data cohort study. J Crit Care. 2009;24:101–107. doi: 10.1016/j.jcrc.2007.11.015. [DOI] [PubMed] [Google Scholar]
- 21.Frost SA, Tam V, Alexandrou E, et al. Readmission to intensive care : Development of a nomogram for individualising risk. Crit Care Resusc. 2010;12:83–89. [PubMed] [Google Scholar]
- 22.Fiser DH. Assessing the outcome of pediatric intensive care. J Pediatr. 1992;121:68–74. doi: 10.1016/s0022-3476(05)82544-2. [DOI] [PubMed] [Google Scholar]
- 23.Feudtner C, Christakis DA, Connell FA. Pediatric deaths attributable to complex chronic conditions: A population-based study of Washington State, 1980–1997. Pediatrics. 2000;106(1, Part 2):205–209. [PubMed] [Google Scholar]
- 24.Edwards JD, Houtrow AJ, Vasilevskis EE, et al. Chronic conditions among children admitted to U.S. pediatric intensive care units: Their prevalence and impact on risk for mortality and prolonged length of stay. Crit Care Med. 2012;40:2196–2203. doi: 10.1097/CCM.0b013e31824e68cf. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25.Pendergraft TB, Stanford RH, Beasley R, et al. Seasonal variation in asthma-related hospital and intensive care unit admissions. J Asthma. 2005;42:265–271. doi: 10.1081/jas-200057893. [DOI] [PubMed] [Google Scholar]
- 26.Dodek PM, Norena M, Keenan SP, et al. Intensive care unit admissions for community-acquired pneumonia are seasonal but are not associated with weather or reports of influenza-like illness in the community. J Crit Care. 2011;26:228–233. doi: 10.1016/j.jcrc.2010.08.011. [DOI] [PubMed] [Google Scholar]
- 27.Iwane MK, Edwards KM, Szilagyi PG, et al. New Vaccine Surveillance Network: Population-based surveillance for hospitalizations associated with respiratory syncytial virus, influenza virus, and parainfluenza viruses among young children. Pediatrics. 2004;113:1758–1764. doi: 10.1542/peds.113.6.1758. [DOI] [PubMed] [Google Scholar]
- 28.Pilcher DV, Duke GJ, George C, et al. After-hours discharge from intensive care increases the risk of readmission and death. Anaesth Intensive Care. 2007;35:477–485. doi: 10.1177/0310057X0703500403. [DOI] [PubMed] [Google Scholar]
- 29.Renton J, Pilcher DV, Santamaria JD, et al. Factors associated with increased risk of readmission to intensive care in Australia. Intensive Care Med. 2011;37:1800–1808. doi: 10.1007/s00134-011-2318-x. [DOI] [PubMed] [Google Scholar]
- 30.Obel N, Schierbeck J, Pedersen L, et al. Mortality after discharge from the intensive care unit during the early weekend period: A population based cohort study in Denmark. Acta Anaesthesiol Scand. 2007;51:1225–1230. doi: 10.1111/j.1399-6576.2007.01431.x. [DOI] [PubMed] [Google Scholar]
- 31.Slater A, Shann F, Pearson G Paediatric Index of Mortality (PIM) Study Group. PIM2: A revised version of the Paediatric Index of Mortality. Intensive Care Med. 2003;29:278–285. doi: 10.1007/s00134-002-1601-2. [DOI] [PubMed] [Google Scholar]
- 32.Feudtner C, Levin JE, Srivastava R, et al. How well can hospital readmission be predicted in a cohort of hospitalized children? A retrospective, multicenter study. Pediatrics. 2009;123:286–293. doi: 10.1542/peds.2007-3395. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 33.Berry JG, Hall DE, Kuo DZ, et al. Hospital utilization and characteristics of patients experiencing recurrent readmissions within children’s hospitals. JAMA. 2011;305:682–690. doi: 10.1001/jama.2011.122. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 34.Gay JC, Hain PD, Grantham JA, et al. Epidemiology of 15-day readmissions to a children’s hospital. Pediatrics. 2011;127:e1505–e1512. doi: 10.1542/peds.2010-1737. [DOI] [PubMed] [Google Scholar]
- 35.Gajic O, Malinchoc M, Comfere TB, et al. The Stability and Workload Index for Transfer score predicts unplanned intensive care unit patient readmission: Initial development and validation. Crit Care Med. 2008;36:676–682. doi: 10.1097/CCM.0B013E318164E3B0. [DOI] [PubMed] [Google Scholar]
- 36.DesHarnais SI, McMahon LF, Jr, Wroblewski RT, et al. Measuring hospital performance. The development and validation of risk-adjusted indexes of mortality, readmissions, and complications. Med Care. 1990;28:1127–1141. [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.