There is growing emphasis on improving hospital discharge planning and transitional care for children.1 Many of the potential elements, such as medication reconciliation, patient education, scheduling of outpatient follow-up visits, and postdischarge telephone or home nursing outreach, can be added or improved through process change and new resource allocation.
In adult and elderly patients, poor quality-of-care transitions have been linked to higher readmission rates, and bundled transitional care interventions have decreased readmissions in these patients.2 There is less published evidence on the impact of such interventions in hospitalized children.3 The value of readmissions as a metric of pediatric hospital care quality and the potential impact of interventions aimed at improving hospital discharge remains in question.4–6
Despite this uncertainty, pediatric readmission metrics are being incorporated into quality and payment reform, which drives efforts to further understand the etiologies of pediatric readmissions.7 As part of a broader quality improvement initiative to reduce readmissions, we notified inpatient providers about their patients’ readmissions and sought their perceptions about the preventability and causes of readmissions. In this study, we analyzed their responses to describe (1) the preventability of 30-day readmissions, and (2) causes of potentially avoidable readmissions as judged by inpatient clinicians.
Population and Setting
This work occurred at Children’s Hospital Colorado, a free-standing tertiary care center with more than 20 000 hospital discharges per year. The project was performed by our readmissions team within the context of a national quality improvement collaborative.8 We developed a process of e-mail notification for providers of children with unplanned 30-day readmissions between June and December 2013. E-mails were sent to the discharging provider from the index admission and the readmitting provider. Children discharged from all hospital services and units were included. The project was approved as quality improvement by the Children’s Hospital Organizational Research Risk and Quality Improvement Review Panel which exempts the project from institutional review board review.
Weekly 30-day readmission reports were generated by our hospital’s information technology department. Planned readmissions were excluded via an electronic flag indicating that the admission encounter was entered into the electronic medical record on a date before the admission date, and therefore was planned ahead of time.9 The report included patient name, discharge and readmitting dates, and discharge and readmitting attending. A research assistant sent weekly e-mail batches to both the discharging and readmitting attending provider of each readmitted child based on these reports. The e-mail included an introduction of the quality improvement project and goals, the patient’s name and medical record number, and a brief survey within the body of the e-mail. E-mails were sent simultaneously to both providers so that each could see and respond to the other’s replies. Responses were e-mailed back to the research assistant and transposed to the readmissions report in Microsoft Excel (Microsoft Corp, Redmond, WA) for subsequent analysis. If 2 providers responded differently to a specific question for a given patient, a “yes” or “likely” response was preferentially entered into the spreadsheet over a neutral or negative response. As such, in analysis, readmissions were categorized as potentially avoidable if one of the providers answered “yes” or “likely” and the other provider answered “no” or “likely not” to the question of whether the readmission could have been avoidable.
Our quality improvement team designed the survey to include queries about factors that are hypothesized to contribute to readmissions.10 We asked if the readmission was planned to further distinguish planned readmissions erroneously included in the automated report. We also queried if the readmission could have been avoided. For ease of use, we requested categorical (Yes/No) responses, although sometimes providers responded with less certainty (eg, maybe or not sure). We also included an open-ended question at the end of the survey (ie, what other factors contributed to this readmission?) (Fig 1).
Although e-mails were sent for all medical and surgical readmissions, we excluded oncology and bone marrow transplant readmissions from analysis, as the reviews were conducted by the medical director of oncology (by request) and not by discharging and readmitting providers. We analyzed readmissions of all remaining surgical, medical, and medical subspecialty patients.
We described the frequency of responses by discharging provider only, readmitting provider only, both providers, or same provider (ie, same person on service for discharge and readmission). We described the percentages of readmissions that were considered potentially avoidable (response: yes or likely), uncertain (response: possibly, maybe, unsure), and unavoidable (response: no or likely not).
The remaining survey categories pertained to causal factors of the readmissions (eg, disease progression, problems with discharge instructions). Responses about causal factors were similarly categorized as likely yes, uncertain, or likely no. We compared the frequency of “likely or yes” responses for each factor between potentially avoidable and unavoidable readmissions using the χ2 or Fisher’s exact tests.
The primary investigator (MB) reviewed qualitative comments to identify additional perceptions about causes of potentially avoidable readmissions. Comments were mapped to categories of disease progression, complication from previous admission, and new disease process to gain more insight into these factors. Microsoft Excel and SAS v. 9.4 (SAS Institute, Inc, Cary, NC) were used for data storage and analysis.
We received provider responses for 351of 493 unplanned readmissions. Ninety-two oncology and bone marrow transplant readmissions were excluded, and we analyzed 259 (66%) of 401 readmissions. Forty-five (17%) readmissions were considered potentially avoidable, 29 (11%) uncertain, and 185 (71%) unavoidable.
Table 1 shows the breakdown of responses and preventability ratings by type of provider. We received the most responses from readmitting providers only (60%), followed by discharging providers only (24%), both providers (10%), and same provider on service for both hospitalizations (6%). Readmissions were more often judged to be potentially avoidable when the same provider was on service for both hospitalizations, and when both providers responded (Table 1).
Providers attributed potentially avoidable (n = 45) versus unavoidable (n = 185) readmissions to the following factors: disease progression (40.0% vs 45.9%; P = .5), previous admission complication (33.3% vs 15.1%; P = .006), new disease process (26.7% vs 40.0%; P = .07), problems/adhering to discharge instructions (24.4% vs 3.2%; P < .0001), insufficient family/community supports (17.8% vs 5.4%; P = .006), inability to obtain/correctly administer medications (8.9% vs 2.2%; P = .05), home health problem (11.1% vs 2.7%; P = .03), and suboptimal outpatient follow-up (8.9% vs 0.5%; P = .006).
Table 2 shows examples of provider comments for 3 categories of potentially avoidable readmissions: disease progression, complication from previous admission, and new disease process. The responses suggested problems with care coordination, community resources, clinical practice, and communication across settings in preventable readmissions. A few readmissions were thought unnecessary (ie, prompted by undue outpatient provider or parent concerns). Preventable complications from the previous admission were most often linked to surgery, including postoperative infection. One response revealed how readmission notification might spur additional review/action: “This is one that I think we all need to discuss. A care conference is in the works again, but there is a breakdown in the system in my opinion for this patient.”
To our knowledge, this is the first pediatric study to describe provider perceptions of readmission preventability and etiologies via e-mailed survey responses. Seventeen percent of unplanned 30-day readmissions were deemed potentially avoidable and 11% were considered uncertain. Rates of potentially preventable readmission have varied from 20% to 50% in previous studies using medical record review, application of preventability software to administrative data, and qualitative methods.6,11–13
We identified varied and complex reasons for potentially avoidable readmissions, some of which would be addressed by broad improvements to health care systems and hospital discharge. Factors relating to the quality of hospital discharge and care transitions were more prominently identified for potentially avoidable readmissions than unavoidable ones. The most frequently identified factors for avoidable readmissions were disease progression, previous admission complication (often related to surgery), new disease process, and discharge instructions/adherence. Respondents attributed avoidable disease progression or new disease process to clinical management (ie, diagnosis and treatment), care coordination (communication across settings, outpatient services), and discharge preparation/instructions. Providers infrequently attributed readmission to premature hospital discharge. We noted in a previous study that parents more often attributed readmission to premature hospital discharge than providers.14 We did not gather perceptions of other stakeholders in this study (parents, outpatient providers, allied health workers), which would yield more comprehensive and balanced perspectives.
Structured feedback to providers may reduce readmissions by allowing clinicians to reflect on their own clinical and transitional care practices, although there is not yet published evidence to support this premise in pediatrics. We have limited data tying our surveys to readmission rates and we are currently evaluating the effect of our surveys as part of a multifaceted intervention on readmission rates in children with special health care needs.
In the adult setting, introduction of an automated physician feedback system coincided with a significant decrease in readmissions.15 At another center, hospitalists were asked to review and identify factors contributing to their own patients’ readmissions. This prompted providers to initiate process change and led to a simultaneous and sustained reduction in readmissions.16 Although uncontrolled, these studies suggest that systematic provider feedback may be beneficial.
In the current study, readmitting providers more often responded to the e-mails than discharging providers, although their ratings of nonpreventability were very similar. Simplifying the process by sending e-mails only to readmitting providers may be a reasonable approach. However, if the goal is to raise awareness and encourage clinicians to reflect on their own clinical practice or systems-based issues, then multiprovider notifications may be more valuable. We did not assess whether providers perceived benefit from these notifications (eg, prompting clinicians to change clinical practice style or address care systems) or potential negative consequences (eg, too time-consuming, feelings of guilt), which would be an important avenue of future research.
Interestingly, when both providers responded, or when a provider who was on service for both hospitalizations responded, the perceptions more often favored avoidable readmission. The numbers, though, were low, and it is difficult to draw conclusions from these results. As one plausible explanation, the original goals of the quality improvement project and data entry process precluded assessment of interrater reliability. Specifically, we did not examine how often 2 providers disagreed. As positive responses trumped uncertain or negative responses in the data entry process, we may have overestimated preventability or the role of causal factors for readmissions rated by 2 respondents. Alternatively, the second provider may have received additional information from the first that influenced the latter’s perspective. Sending e-mails to providers/stakeholders independently (rather than on the same e-mail thread) may alter the number of respondents or the nature of responses (ie, by reducing censoring or response bias).
This study has other limitations beyond those mentioned previously. The survey instrument was designed to optimize the chance of feedback. Without more qualitative information some of the categorical responses were difficult to interpret. Given the providers’ willingness to provide qualitative comments (93% of readmissions), it seems useful to include open-ended questions as part of a readmission feedback tool. We did not assess gradations of opinion through a Likert scale, or undertake a formal qualitative analysis (ie, researcher triangulation) of provider comments. This may have resulted in an inaccurate measurement of preventability or magnitude of certain factors. Finally, these single-center findings cannot be generalized to other children’s hospitals or to nonacademic or community pediatric centers.
Despite these limitations, the method of notifying providers about readmissions and soliciting feedback can help identify areas for process improvement and may encourage reflective practice. The results of this project have prompted us to evaluate the content and processes around discharge instructions and education. Definitive research is needed to determine if readmission notification can decrease pediatric readmissions.
FINANCIAL DISCLOSURE: The authors have indicated they have no financial relationships relevant to this article to disclose.
FUNDING: Dr Brittan received salary support from an institutional Research Scholar Award and the project was supported by funding from an institutional Clinical and Operational Effectiveness and Patient Safety Grant.
POTENTIAL CONFLICT OF INTEREST: The authors have indicated they have no potential conflicts of interest to disclose.
- Bardach NS,
- Vittinghoff E,
- Asteria-Peñaloza R,
- et al
- Hain PD,
- Gay JC,
- Berutti TW,
- Whitney GM,
- Wang W,
- Saville BR
- Trudnak T,
- Kelley D,
- Zerzan J,
- Griffith K,
- Jiang HJ,
- Fairbrother GL
- Auger KA,
- Simon TD,
- Cooperberg D,
- et al
- Wallace SS,
- Keller SL,
- Falco CN,
- et al
- Amin D,
- Ford R,
- Ghazarian SR,
- Love B,
- Cheng TL
- Brittan M,
- Albright K,
- Cifuentes M,
- Jimenez-Zambrano A,
- Kempe A
- Kashiwagi DT,
- Burton MC,
- Hakim FA,
- et al
- Copyright © 2016 by the American Academy of Pediatrics