Potentially Preventable 30-Day Hospital Readmissions at a Children’s Hospital
BACKGROUND AND OBJECTIVES: Hospital readmission rates are increasingly used to assess quality. Little is known, however, about potential preventability of readmissions among children. Our objective was to evaluate potential preventability of 30-day readmissions using medical record review and interviews.
METHODS: A cross-sectional study in 305 children (<18 years old) readmitted within 30 days to a freestanding children’s hospital between December 2012 and February 2013. Interviews (N = 1459) were conducted with parents/guardians, patients (if ≥13 years old), inpatient clinicians, and primary care providers. Reviewers evaluated medical records, interview summaries, and transcripts, and then rated potential preventability. Multivariate regression analysis was used to identify factors associated with potentially preventable readmission. Adjusted event curves were generated to model days to readmission.
RESULTS: Of readmissions, 29.5% were potentially preventable. Potentially preventable readmissions occurred sooner after discharge than non–potentially preventable readmissions (5 vs 9 median days; P < .001). The odds of a readmission being potentially preventable were greatest when the index admission and readmission were causally related (adjusted odds ratio [AOR]: 2.6; 95% confidence interval [CI]: 1.0–6.8) and when hospital (AOR: 16.3; 95% CI: 5.9–44.8) or patient (AOR: 7.1; 95% CI: 2.5–20.5) factors were identified. Interviews provided new information about the readmission in 31.2% of cases.
CONCLUSIONS: Nearly 30% of 30-day readmissions to a children’s hospital may be potentially preventable. Hospital and patient factors are associated with potential preventability and may provide targets for quality improvement efforts. Interviews contribute important information and should be considered when evaluating readmissions.
- AOR —
- adjusted odds ratio
- CCI —
- Chronic Condition Indicator
- CI —
- confidence interval
- CMS —
- Centers for Medicare and Medicaid Services
- PCP —
- primary care provider
What’s Known on This Subject:
Hospital readmission rates are increasingly used as measures of hospital quality. Limited information is available on whether pediatric readmissions are potentially preventable; however, no studies in children have incorporated medical record review and interviews in assessing potential preventability of readmissions.
What This Study Adds:
Nearly 30% of 30-day readmissions to a children’s hospital were potentially preventable. Hospital and patient factors were often identified for these readmissions. Parent and provider interviews, in combination with medical record reviews, provided important information about potential preventability of readmissions.
Hospital readmissions are disruptive to patients and families and costly to our health care system. Given that studies in adults suggest that preventable readmissions can result from the care received during the index (ie, initial) admission, inadequate discharge planning or postdischarge follow-up, or poor coordination among providers, adult readmission rates are widely used as a quality measure for public reporting and pay-for-performance.1,2 Increasingly, pediatric hospital readmission rates are also being used as a quality measure.3,4 The Centers for Medicare and Medicaid Services (CMS) and many state Medicaid agencies have levied financial penalties for hospitals with high 30-day readmission rates. Although the CMS does not apply financial penalties for pediatric readmissions, a growing number of states are applying penalties for pediatric readmissions.5–11
A common criticism of using readmission rates as a health care quality measure is that many readmissions are not causally related to the hospital care provided during the index admission; rather, they may represent worsening of underlying chronic disease.12,13 In addition, some argue that hospitals should not be held accountable for readmissions because factors that contribute to readmissions often are not hospital factors (eg, postoperative complication) but instead are outpatient (eg, nurse unable to visit home) or patient (eg, not taking prescribed medications) factors and thus beyond the hospital’s direct control.14,15
Currently, 3 main strategies for ascertaining the potential preventability of hospital readmissions include the following: claims-based algorithms, medical record review, and structured interviews. Claims-based algorithms are perceived to be limited in identifying reasons for readmission.16,17 Medical record review and structured interviews may offer more information. Although interviews provide opportunities to obtain even more detailed information, it is unclear whether interviews enhance information obtained from medical records or are duplicative. Because interviews are time intensive and costly, it is important to understand their value.
It remains unclear how many readmissions in children may be preventable and whether factors contributing to preventable readmissions can be identified. Furthermore, no previous study in children has compared medical record review and structured interviews in assessing readmissions. Thus, our study aimed to determine the potential preventability of unplanned 30-day hospital readmissions, to identify factors that contribute to readmissions, and to assess whether structured interviews add substantial valuable information beyond medical record review.
Study Design and Participants
We conducted a cross-sectional study of unplanned 30-day hospital readmissions at a freestanding children’s hospital. Patients <18 years old who were readmitted between December 10, 2012, and February 22, 2013 (ie, discharged from the same hospital in the 30 days before the readmission), were potentially eligible for the study. Exclusion criteria included the following: planned readmissions (eg, scheduled surgeries or chemotherapy), psychiatric readmissions, and readmissions for end-of-life care. Patients were only included once during the study period. For 2 common conditions (fever and neutropenia in oncology patients, bleeding after tonsillectomy/adenoidectomy), only a random subsample (1 of every 3 readmissions) was included. Of the 928 patients readmitted during the study period, 378 were potentially eligible on the basis of study exclusion criteria. Thirty-nine patients with 1 of the 2 frequently occurring conditions were randomized out. Parents declined participation or could not be contacted for 9 patients. An additional 25 patients were excluded for other reasons (eg, possible child abuse). In total, the study included 305 patients.
Research nurses and research assistants conducted 1192 semistructured interviews with parents or caregivers, henceforth referred to as “parents” (n = 305), adolescent patients (n = 42), inpatient nurses (n = 297) and attending physicians at time of readmission (n = 304), social workers (n = 91) and case managers (n = 104) who were involved in the index admission, and subspecialists who comanage children with medical complexity (n = 49). Parent and adolescent interviews were conducted by using a semistructured interview script, recorded, and transcribed. The inpatient provider interviews were conducted by using a structured interview form. We conducted training sessions and a 9-day pilot before the collection of actual study data to refine the study logistics and survey instrument. Study data were collected and managed by using REDCap, a secure, Web-based application.18 A pediatrician conducted telephone interviews with primary care providers (PCPs; n = 267) with the use of a structured interview form. Thirteen PCPs were unreachable or declined participation. PCP interviews were excluded for international patients (n = 13), patients without a designated PCP (n = 8), and incomplete PCP interviews (n = 4).
Four pediatric physician reviewers (3 inpatient/outpatient providers and 1 outpatient provider) who did not conduct interviews independently conducted structured medical record reviews for all readmissions. Reviewers received standardized training in which they were presented with guidelines for assessing preventability and related factors by using Institute for Healthcare Improvement background frameworks.19 After completing medical record review, the physician reviewers assessed the readmission, examined transcripts and interview summaries, and determined whether the interviews changed their assessment of the readmission. Subsequently, a consensus review process was conducted that included the physician reviewer(s) and an academic general pediatrician with >1 decade of postgraduate experience. Because of the complexity of assessing preventability, we used a consensus review process similar to that used in other readmission-preventability studies.13,20,21 Each case had 1 primary reviewer, with no fewer than 3 pediatricians participating in the consensus review. In circumstances in which the readmission involved significant subspecialty care (eg, congenital heart disease postoperative management), relevant subspecialists were consulted.
The primary outcome measure was potential preventability of the readmission using a 4-point Likert scale (very likely to have been preventable, somewhat likely to have been preventable, somewhat unlikely to have been preventable, and very unlikely to have been preventable). Additional dichotomous outcomes included the following: causal relationship between index admission and readmission; index admission and readmission related by chronic illness; and presence of hospital, PCP, or patient factors contributing to the readmission. We characterized the contributing factors by using the overarching categories from the Institute for Healthcare Improvement’s STAAR (STate Action on Avoidable Rehospitalizations) Readmissions Diagnostic Worksheet: “typical failures associated with patient assessment,” “typical failures found in patient and family caregiver education,” “typical failures in handover communication,” and “typical failures following discharge from the hospital.”22 We created 3 additional categories: postoperative complication/hospital-acquired condition, contributing social factors/parental anxiety, and other. Secondary outcome measures covered whether the addition of the interviews to the medical record review changed the reviewers’ assessment of preventability (eg, from somewhat likely to have been preventable to somewhat unlikely to have been preventable) or enabled identification of new contributing factors for readmission (eg, new hospital contributing factor identified).
Covariates of Interest
Child demographic covariates included age in years (<1, 1–4, 5–8, 9–12, and 13–17 years), sex, race/ethnicity (Asian, black/non-Hispanic, Hispanic, white/non-Hispanic, and other), and preferred language (English or non-English). To identify children with chronic disease, we used the Agency for Healthcare Research and Quality (AHRQ) Chronic Condition Indicator (CCI), a classification system based on International Classification of Diseases, Ninth Revision, codes.23 In addition, we counted the number of CCIs present for each patient (0, 1, 2, or ≥3).
The 4-point scale for preventability was collapsed into a binary variable: potentially preventable (somewhat/very likely to have been preventable) and potentially nonpreventable (somewhat/very unlikely to have been preventable). First, we examined whether preventability was associated with patient characteristics. We then examined whether potentially preventable readmissions were associated with the following: (1) the readmission being causally related to the index admission, (2) the readmission being related to the index admission by underlying chronic disease, and (3) the presence of identifiable contributing factors. We examined the association between preventability rating and the presence of different chronic conditions (measured by CCI category). We also assessed whether patients with chronic conditions were more likely to have certain contributing factors in their readmissions. Bivariate associations were tested by using Fisher’s exact tests or Wilcoxon rank-sum tests as appropriate; multivariate associations were tested by using multivariate logistic regression models adjusting for additional covariates.
The overall crude 30-day readmission rate for the hospital was calculated for the study period by using the AHRQ-CMS Pediatric All-Condition Readmission Measure SAS program.4 We compared the mean days to readmission by using the Wilcoxon rank-sum test for the following conditions: (1) potentially preventable and not preventable, (2) causally related and not causally related, (3) due to underlying chronic disease and not due to underlying chronic disease, and (4) contributing factor and no contributing factor identified. As an illustration of time to readmission, we generated adjusted event curves on the basis of Cox proportional hazard models to adjust for age, sex, language, and CCI count; and the model coefficient Wald tests were used to compare the event curves between groups.
We assessed the change in perception of potential preventability and the addition of contributing factors with the inclusion of the semistructured interviews. Statistical analysis was conducted by using SAS version 9.3 (SAS Institute, Cary, NC) and Stata version 13 (StataCorp, College Station, TX). The Boston Children’s Hospital Institutional Review Board determined that this study was exempt.
The overall readmission rate for the hospital during the study period was 6.5%. Of the readmissions included in the study, 29.5% were determined to be potentially preventable: 11.8% very likely, 17.7% somewhat likely, 16.1% somewhat unlikely, and 54.4% very unlikely. In bivariate analysis, potentially preventable readmissions and non–potentially preventable readmissions did not vary significantly on the basis of patient age, sex, language, or race/ethnicity (Table 1). Although potentially preventable and non–potentially preventable readmissions did not vary significantly by number of chronic conditions, for 2 of the 18 chronic condition categories, neoplasms (2.2% vs 9.8%; P = .03) and diseases of blood/blood-forming organs (1.1% vs 8.8%; P = .01), readmissions were more likely to be considered non–potentially preventable. In contrast, readmissions for congenital anomalies were more likely to be considered potentially preventable (40.0% vs 22.3%; P = .002).
Of all readmissions, 48.9% were deemed to be causally related to the index admission and 63.8% were deemed to be due to a chronic disease. Potentially preventable readmissions were more often causally related to the index admission (74.1% vs 38.6%; P < .001); however, no association between chronic disease and preventability was observed (P = .59).
At least 1 contributing factor was identified more often for potentially preventable than for non–potentially preventable readmissions (86.7% vs 20.4%; P < .001). Potentially preventable readmissions more often had hospital (77.1% vs 16.7%; P < .001), PCP (14.5% vs 0.5%; P < .001), and patient (39.2% vs 9.2%; P < .001) contributing factors identified. Among all readmissions, 36.2% were associated with hospital factors, 4.6% with PCP factors, and 18.1% with patient factors. Hospital factors were most often related to “patient assessment” (37.4%) and postoperative complications/hospital-acquired conditions (28.0%); PCP factors were most often related to “patient assessment” (58.3%) and “handover communication” (25.0%); and patient factors were most often related to “following discharge from the hospital” (60.4%) (Table 2).
In multivariate analyses adjusting for age, sex, language, and CCI count, potentially preventable readmissions, when compared with non–potentially preventable readmissions, had increased odds of being causally related to the index admission (adjusted odds ratio [AOR]: 2.6; 95% confidence interval [CI]: 1.02–6.75). Potentially preventable readmissions also had increased odds of identification of a hospital (AOR: 16.3; 95% CI: 5.94– 44.81) or patient (AOR: 7.1; 95% CI: 2.45–20.48) factor (Table 3).
The mean days to readmission (Fig 1) was shorter for potentially preventable readmissions than for non–potentially preventable readmissions (4 vs 9 median days; Wilcoxon test, P < .001). Similar findings were observed for readmissions that were causally related to the index admission (4 vs 14 median days; Wilcoxon test, P < .001) and for readmissions for which a contributing factor was identified in comparison with readmissions where one was not (5 vs 10 median days; Wilcoxon test, P < .001). The time to readmission was significantly longer for readmissions that were related to underlying chronic disease (9 vs 5 median days; Wald test, P < .001). These comparisons of the curves between preventable and nonpreventable and the other factors were also significant (Wald test, P < .01).
New information was learned from the interview regarding the readmission in 31.2% of all cases. For instance, a complicated patient was readmitted with respiratory distress 1 week after discharge. We learned from the physician that the family declined to have their child go to a rehabilitation center after the index admission. In another example, a patient readmitted with gastrostomy-tube cellulitis was initially discharged without a visiting nurse. However, we learned from the parent that they were frequently checking the gastrostomy-tube site and that the cellulitis developed rapidly and would not have been picked up earlier had a visiting nurse been coming to the home. In another case, we learned from the parent that because of delays in her child’s procedure, she missed her medications and pulmonary toileting, which ultimately led to increased respiratory distress the day after the procedure and resulted in a readmission. New information was learned about the potential preventability of the readmission in 20.3% of cases, the reasons for readmission in 12.8% of cases, and additional contributing factors in 15.1% of cases. In 11.8% of all cases, new information learned from the interviews changed the final assessment of potential preventability of the readmission. In cases in which the reviewers’ opinion regarding the potential preventability changed after the interview review, 47.2% of cases were revised from potentially nonpreventable to potentially preventable and 8.3% from potentially preventable to nonpreventable. For the remaining cases, the assessment shifted without changing the direction of the preventability rating (eg, a change from somewhat to very likely to have been preventable). New information learned from the interviews was predominantly gained from parents (21.6%), attending physicians (14.8%), PCPs (11.1%), and inpatient nurses (7.2%). The information learned was mainly about the hospital (19.0%), parents (15.1%), patients (7.9%), and PCPs (4.6%).
We found that 29.5% of all-cause, unplanned, 30-day readmissions to a tertiary care children’s hospital were potentially preventable. Potentially preventable readmissions were more likely to be causally related to the index admission and have an associated contributing factor identified than were readmissions that were non–potentially preventable. Hospital and patient factors were strongly associated with potentially preventable readmission. Although many children with underlying chronic disease are readmitted, preventability was not associated with the presence of an underlying chronic disease.
Our findings are largely consistent with the few studies that assessed the preventability of readmissions in the pediatric setting. One study on unplanned pediatric readmissions found that the most common reasons cited for them were disease progression and procedural complications.24 The rate of potentially preventable readmissions has varied across pediatric studies. For instance, 1 study that performed medical record reviews of 200 patients readmitted to a tertiary care children’s hospital within 15 days of discharge found that 20% of readmissions were considered likely to be preventable.13 In another study, 42.6% of readmissions or revisits to the emergency department within 90 days of discharge after an appendectomy admission were deemed potentially avoidable.25
In comparison with other methods for calculating potentially preventable readmissions in pediatric settings, our rate of potentially preventable readmissions is lower than previously reported rates that used claims-based software algorithms but higher than those that used medical record review.26,27 Discordance between automated and manual review of readmissions has been previously noted.28 Claims-based algorithms may overestimate the preventability by not fully accounting for patient characteristics (eg, severity of chronic condition) and planned readmissions.29,30 Medical record review, on the other hand, may provide added clinical detail but may be limited in ascertaining the contribution of sociodemographic characteristics that may not be documented in the medical record (eg, unmet social needs, medication adherence) or may occur outside of the hospital (eg, missed appointments, transportation barriers), thereby leading to underestimation of the likelihood of preventability.13,25,31–34 Interviews in combination with medical record review may provide a more global understanding of these circumstances leading up to, and after, a child’s hospitalization. Due to the complexity of assessing preventability, we chose to use a multistep approach that culminated in a consensus review.
In contrast to other studies that have shown an association between clinical characteristics and readmissions in other settings, our study did not observe an association between certain patient clinical characteristics and potentially preventable readmissions.13,24–26,31,32,34–38 This difference may be related to the high overall prevalence of chronic illness in our sample. Chronically ill children may be more vulnerable to potentially preventable readmissions due to medical fragility and increased care needs after discharge (eg, specialty appointments, multiple medications); however, their parents may be more comfortable navigating the health care system. Further research is needed to examine the role of chronic illness in potentially preventable readmissions.
Potentially preventable readmissions occurred sooner after hospital discharge than did non–potentially preventable readmissions. We observed a strong association between hospital factors and potential preventability and also found that readmissions with hospital contributing factors were more likely to occur sooner after hospital discharge than readmissions without. Our findings suggest that readmissions occurring closer to the index admission are more likely to be preventable.
A majority of potentially preventable readmissions had an identified hospital contributing factor. Previous studies have identified hospital and clinician factors as a large portion of potentially preventable readmissions.39,40 Hospital factors may serve as targets for quality improvement initiatives. In particular, we found that hospital factors were often related to patient assessment, such as failure to recognize worsening clinical status. Hospital factors were also often related to patient and family education and handover communication. Addressing these areas through improved discharge preparation and planning might help hospitals decrease preventable readmissions.
Our study has limitations. The retrospective interviews are subject to response bias. In addition, we conducted the study with children hospitalized at a tertiary care children’s hospital, so this may limit the generalizability to non–children’s hospitals. Because it was a single hospital, we were unable to examine characteristics of the treating hospital such as nurse staffing ratios, which have been shown to be associated with pediatric readmission rates.41
Our results suggest that nearly 30% of readmissions were likely to have been prevented with better care delivery and that interviews provided new information about readmissions in nearly one-third of all cases. Although they are not the majority of readmissions, potentially preventable readmissions are prevalent enough to warrant monitoring readmission rates and to provide targets for quality improvement. Interviews led to a change in the assessment of the potential preventability and provided important information regarding the reasons for and contributing factors to readmissions in a substantial proportion of cases. On the basis of this study, targets for reducing readmissions might include addressing hospital factors, such as the need for improved caregiver education, meeting social needs, and increasing rates of postdischarge follow-up. More broadly, parent and provider interviews offer important information regarding potential preventability and contributing factors for preventability.
We thank Dr Swee Hui Ang, MD, MPH, Dr Samantha Bannister, MD, MPH, Ms Annemarie Fayemi, MS, RN, CPNP-PC, Ms Diane McInerney, Ms Christine Finn, RN, Dr Kathleen Levin, MD, Ms Colleen Madden, RN, Mr Debanjan Pain, AB, and Ms Michelle Wang, AB, for their help with data collection and analysis.
- Accepted May 24, 2016.
- Address correspondence to Sara L. Toomey, MD, MPhil, MPH, MSc, Division of General Pediatrics, Department of Medicine, Boston Children’s Hospital, 300 Longwood Ave, Boston, MA 02115. E-mail:
FINANCIAL DISCLOSURE: The authors have indicated they have no financial relationships relevant to this article to disclose.
FUNDING: Support for this work was provided by the US Department of Health and Human Services Agency for Healthcare Research and Quality and Centers for Medicare and Medicaid Services, CHIPRA (Children’s Health Insurance Program Reauthorization Act of 2009) Pediatric Quality Measures Program Centers of Excellence under grant U18 HS 020513. The content is solely the responsibility of the authors and does not necessarily represent the official views of the Agency for Healthcare Research and Quality.
POTENTIAL CONFLICT OF INTEREST: The authors have indicated they have no potential conflicts of interest to disclose.
COMPANION PAPER: A companion to this article can be found online at www.pediatrics.org/cgi/doi/10.1542/peds.2016-1643.
- ↵Centers for Medicare and Medicaid Services. Readmissions reduction program. Published April 26, 2013. Available at: www.cms.gov/Medicare/Medicare-Fee-for-Service-Payment/AcuteInpatientPPS/Readmissions-Reduction-Program.html. Accessed May 20, 2013
- ↵Agency for Healthcare Research and Quality; Centers for Medicare and Medicaid Services. Pediatric All-Condition Readmission Measure. Available at: www.ahrq.gov/sites/default/files/wysiwyg/policymakers/chipra/factsheets/chipra_14-p008-1-ef.pdf. Accessed August 17, 2015
- ↵Patient Protection and Affordable Care Act. Vol Stat 119; 2010:318–319. Available at: www.gpo.gov/fdsys/pkg/PLAW-111publ148/pdf/PLAW-111publ148.pdf. Accessed September 8, 2015
- Medicare Payment Advisory Commission
- Illinois Department of Healthcare and Family Services. State fiscal year (SFY) 2014 Potentially Preventable Readmissions (PPR) Reports. Published October 4, 2014. Available at: www.hfs.illinois.gov/html/100413n1.html. Accessed January 13, 2016
- Potentially preventable readmissions in the Texas Medicaid population, state fiscal year 2012. Texas Health and Human Services Commission; 2013:11–25. Available at: www.hhsc.state.tx.us/reports/2013/PPR-Report.pdf. Accessed August 17, 2015
- NY State Potentially Preventable Readmissions. 2011. Available at: https://www.health.ny.gov/regulations/recently_adopted/docs/2011-02-23_potentially_preventable_readmissions.pdf. Accessed January 13, 2016
- Oklahoma Health Care Authority. Hospital Potentially Preventable Readmissions (PPR) Program. 2015. Available at: http://www.okhca.org/providers.aspx?id=16078&terms=Hospital%20Potentially%20Preventable%20Readmissions%20(PPR. Accessed January 13, 2016
- ↵Colorado Department of Health Care Policy and Financing. Creating a culture of change. 2014. Available at: https://www.colorado.gov/pacific/sites/default/files/Accountable%20Care%20Collaborative%202014%20Annual%20Report.pdf. Accessed January 13, 2016
- Institute for Healthcare Improvement
- Meisenberg BR,
- Hahn E,
- Binner M, et al
- ↵Institute for Healthcare Improvement. IHI STAAR Readmissions Diagnostic Worksheet. Available at: http://www.ihi.org/resources/pages/tools/readmissionsdiagnosticworksheet.aspx. Accessed June 20, 2016
- ↵Healthcare Cost and Utilization Project (HCUP). Chronic Condition Indicator (CCI) for ICD-9-CM. Available at: https://www.hcup-us.ahrq.gov/. Accessed August 17, 2015
- Williams EI,
- Fitton F
- Hallerbach M,
- Francoeur A,
- Pomerantz SC, et al
- Berry JG,
- Ziniel SI,
- Freeman L, et al
- Flores G,
- Abreu M,
- Tomany-Korman S,
- Meurer J
- Clarke A
- Grim RD,
- McElwain D,
- Hartmann R,
- Hudak M,
- Young S
- Zmistowski B,
- Restrepo C,
- Hess J,
- Adibi D,
- Cangoz S,
- Parvizi J
- Tubbs-Cooley HL,
- Cimiotti JP,
- Silber JH,
- Sloane DM,
- Aiken LH
- Copyright © 2016 by the American Academy of Pediatrics