Discharge Communication Practices for Children With Medical Complexity: A Retrospective Chart Review

We conducted a retrospective cohort study through rigorous chart review of the electronic medical record (EMR). We evaluated the presence and content of documented hospitalist-to-PCP communication at the time of discharge in CMC who were discharged from the hospitalist service at 1 large, urban tertiary care hospital. We also analyzed the correlation between 30-day unplanned readmission and communication practices.

Our population consisted of pediatric patients with medical complexity discharged from a pediatric hospitalist service at a large tertiary care children’s hospital over a 5-month period from September 1, 2016, through January 1, 2017. We limited the population to those patients discharged only from the hospitalist service because we postulated that these patients were most likely to rely on a PCP, rather than a subspecialty provider, for care coordination. Our system does have a complex care consult team to provide continuity during recurrent admissions; however, the primary hospitalist is still responsible for communication at the time of discharge. The involvement of the consultant team was not included as a factor that might impact communication practices.

We defined CMC as any patient whose medical record reflected 1 or more complex chronic condition (CCC) on the basis of International Classification of Diseases, 10th Revision coding.¹¹  We identified our population by first using the Pediatric Health Information System database to query patients who were discharged from our general pediatric services that had ≥1 CCC.¹²  This system was chosen both for ease of patient identification because CCC fields were already in place in the original data set and so that patients could be further classified by type of CCC to clarify functional limitations. This system encompasses the 4 domains of CMC, including chronic conditions, functional limitations (eg, technology dependence), health care needs, and health care use.¹³  Patients who had a scheduled readmission within 30 days, who were discharged from a subspecialty service, or who died while hospitalized were excluded from this study.

Once our cohort of patients was identified, we conducted an institutional chart review for each encounter. Demographic data were collected for each patient that could impact either the likelihood of hospitalist communication at the time of discharge or propensity for readmission. Demographics included number and type of CCC, discharging service (attending-only team versus teaching service), insurance type (public, private, or no insurance listed), length of hospital stay in days, sex, patient age, if the patient had >5 prescription medications at the time of discharge, and if the patient spent time in the ICU while hospitalized.¹⁴ 

The presence or absence of documented physician to physician communication at time of discharge was recorded. Communication was defined as 1 or more of the following: (1) a faxed discharge notification form completed by the attending physician on the day of discharge, which usually included a short description of the hospital course and pertinent information for follow-up care, (2) a note in the chart indicating a phone call or e-mail was made to the PCP by the attending or hospitalist fellow on the day of discharge, or (3) a notation in the attending addendum to the discharge summary stating that the patient’s case was discussed with the PCP. We also collected information indicating that a completed resident-authored discharge summary was faxed to the PCP through an automated system and the components present in each discharge summary as a proxy of communication quality.¹⁵  These quality metrics were based on a previously studied hospital-to-home transition bundle, which highlighted factors needed for completion of handoff to a patient’s PCP and included a complete medication list, physical examination with vitals, follow-up appointments needed, the status of any pending laboratory studies, vaccines given during hospitalization, discharge diagnosis, hospital course, and admission and discharge dates.¹⁵  Communication practices were analyzed for patients with technology dependence and >2 CCCs to describe communication for the most complex patients.

Unplanned readmission within 30-days of discharge to any service in our hospital was also collected to assess if there was a difference in readmission in patients with PCP communication at discharge compared with patients without PCP communication. Study data were collected and managed by using Research Electronic Data Capture tools hosted at our children’s hospital.¹⁶  This study was approved by the Children’s National Hospital Institutional Review Board.

Patient characteristics and communication practices were analyzed by using χ² testing, t testing, Wilcoxon rank testing, and odds ratios (ORs). Normality was evaluated by using Shapiro-Wilk testing and quantile-quantile plots. χ² and Fisher’s exact tests were used to identify statistical differences in communication practices and discharge summary quality. P values were considered significant at <.05. We conducted power analysis to identify the ability to detect differences in readmissions between patients with PCP communication and those without using Power Analysis and Sample Size version 15.0.¹⁷  The sample size provided the power to identify a 9% change in readmission rate. OR and logistic regression modeling was used to analyze correlations between communication practices and readmissions. Statistical analysis was completed in R version 3.4.3.¹⁸ 

There were 377 patient encounters reviewed who were discharged from the hospitalist team from September 1, 2016, to January 1, 2017, in which the patient had 1 or more CCC. Nine of these patients were excluded from our analysis: 6 patients had a planned readmission such as a scheduled surgery, 2 were transferred to and eventually discharged from a subspecialty service, and 1 patient died while hospitalized. This resulted in 368 eligible patient encounters. Of these 368 eligible patients, 217 patients (59%) had at least 1 type of communication with their PCP documented on the day of discharge (Fig 1).

Bivariate analysis revealed patients who had documented physician communication at the time of discharge were less likely to be technology dependent (OR 0.5; confidence interval [CI] 0.33 to 0.84; P value .01), more likely to be younger (mean difference in age of −0.8 years; CI −2.0 to 0.001; P = .02), and more likely to be discharged from the attending-only service than a teaching service (OR 2.55; CI 1.03 to 7; P = .04) (Table 1). Other demographic factors that indicated illness severity, including length of stay and a PICU stay at any point during their admission, did not impact the likelihood of hospitalist-to-PCP communication at discharge.

Subgroup analysis was done to explore types of communication documented for patients with increasing numbers of CCCs and technology dependence, as a proxy for increasing medical complexity. Patients with >2 CCCs were less likely to have a faxed discharge notification sent to their PCPs office (OR 0.6; CI 0.4 to 1; P = .04). Of note, there was no difference in the rate of trainee-authored discharge summaries as complexity increased. Children who were technology dependent were less likely to have an attending-authored fax (OR 0.6; CI 0.4 to 0.9; P = .02) or a notation in the discharge summary that the attending spoke with their PCP (OR 0.3; CI 0.1 to 0.9 P = .04) (Table 2). Bidirectional communication including an e-mail or phone call was uncommon in all groups and did not differ in patients with technology dependence or >2 CCCs (Table 2).

Discharge summary quality was assessed by identifying the presence or absence of components found in a high-quality discharge summary.¹⁵  There was no significant difference in the quality of discharge summaries between patients who were technology dependent and those who were not (Fig 2). Patients who were technology dependent had the same likelihood of having a discharge summary include all quality metrics as children without technology dependence (OR 0.57; CI 0.31 to 1.01).

Our cohort had a 30-day readmission rate of 19.6%. There was no difference in the OR of 30-day readmission between patients with discharge communication documented and those without (19.8% vs 19.2%; OR 1.04 [95% CI: 0.62 to 1.77]). Subgroup analysis of technology-dependent patients did not show any impact of documented PCP communication at discharge on readmission, with an OR of 1.29 (CI 0.57 to 2.99).

This study revealed that at a single urban tertiary care pediatric hospital, CMC had variability in documented communication with their PCP at the time of discharge. Children with increased complexity, measured by increasing numbers of CCCs and analysis of technology-dependent patients alone, were less likely to have attending-authored communication with their PCP at discharge.

There are several possible explanations for this finding. First, CMC with increasing complexity are generally admitted to the teaching service because the attending-only service is usually reserved for patients with a single, clear admitting diagnosis. Our study is consistent with previously published literature in that teaching services are less likely to complete PCP communication. This could be due to unclear roles and expectations with multiple learners per team.⁸  It is possible that the most medically complex, technology-dependent patients are not using their PCP as their main locus of care. A subspecialty doctor may be assuming care coordination for the children with more medical complexity. Additionally, some patients with multiple CCCs at the hospital are managed by a complex care coordination team, which liaises between the inpatient and outpatient settings. If patients are managed by this team, conversations about care coordination are continuous and, generally, face to face, and it is possible that these communications are not documented in the medical record. Finally, because of the complex nature of discharges for CMC, hospitalists may simply not prioritize PCP communication.

The majority of communication in this study took the form of a faxed discharge notification letter. Previous literature has not revealed consensus of PCP preferences regarding modality of communication, although there is agreement that discharge communication should be structured, timely, and direct.^4,19,20  Bidirectional communication such as telephone calls or e-mails were uncommon. Interestingly, attending-generated communication was less likely to occur with increasing patient complexity, but trainee-generated discharge summaries were of similar quality and faxed at similar frequency. This possibly reflects the increased workload on the attending physician in case management tasks at the time of discharge or increased reliance on consulting services to complete inpatient-to-outpatient handoff.

Discharge summary quality was the same when comparing technology-dependent children with those who were not technology dependent. This is possibly due to the templated nature of the discharge summary that prompts the author to enter many components of a high-quality summary. This template is the same for every patient being discharged leading to highly equitable discharge summaries.

We had hoped to link discharge communication with health care use in the form of 30-day unplanned readmissions. We were unable to demonstrate any impact of the presence of discharge communication on readmission. Our conclusions are limited by a sample size that provided a power to detect only a 9% difference in readmissions.

There are multiple limitations in this study. Communication was quantified in a retrospective way on the basis of documentation in the patient’s EMR. Communication may have occurred without documentation. Most communication consisted of a faxed letter summarizing the hospital stay written by the attending hospitalist. The timing and physical receipt of this fax was not confirmed or quantified. We did not collect information about what was contained in these faxed letters. The names and offices of PCPs are obtained and documented in our electronic health record through verification with a family member during the admission process. If the documented PCP is incorrect, the discharge notification letter could be sent to the wrong office.

The analysis was conducted at a single study center in a large tertiary care urban medical center, which limits generalizability. PCPs affiliated within the health system may have access to the inpatient medical record, whereas other community pediatricians may not. We did not delineate groups on the basis of PCP association with the medical system. As described above, the system uses a consultative service for the most medically complex patients. The consulting team coordinates with an outpatient clinic and provides care continuity. It is possible that communication was occurring through this consultative team without being documented in the EMR.

In this study, we highlight the need for improved and standardized communication for CMCs who are being discharged from the hospital. Varied care models for CMCs including multiple physicians and complicated home care needs may create barriers to timely and effective discharge communication between hospitalists and PCPs. Further study used to identify reasons for the communication gap identified in the most complex patients could include a survey of hospitalists, families, and PCPs used to identify communication barriers. Multicenter prospective studies would be useful to characterize communication practices across multiple care models. Finally, researchers of a prospective randomized controlled trial of discharge communication could further evaluate the relationship between communication and health care use in this population.

Communication with a PCP at the time of discharge was less likely to occur for patients with technology dependence at a tertiary care medical center. Attending-authored documentation was less likely to occur for patients with increasing numbers of CCCs. Discharge summary completion and quality were similar among all patients with CCCs. Fragmented medical care systems put pediatric patients with multiple medical conditions at risk for overuse because of poor care transitions. Standardization in discharge communication practices for CMC may increase the frequency of PCP communication, which could lead to improved care coordination. Improvement in care coordination and communication during discharge from the hospital for CMC would improve quality of care. Identifying reasons for the gap in discharge communication for CMCs could increase patient and provider satisfaction by improving overall communication and workflow.

We acknowledge Dr Padma Pavuluri, DO, Dr Karen Fratantoni, MD, MPH, and Dr Catherine Forster, MD, PhD, for their contributions to this article.