Article Text
PDF
Abstract
Discharge from hospitals to postacute care settings is a vulnerable time for many older adults, when they may be at increased risk for errors occurring in their care. We developed the Extension for Community Healthcare Outcomes-Care Transitions (ECHO-CT) programme in an effort to mitigate these risks through a mulitdisciplinary, educational, case-based teleconference between hospital and skilled nursing facility providers. The programme was implemented in both academic and community hospitals. Through weekly sessions, patients discharged from the hospital were discussed, clinical concerns addressed, errors in care identified and plans were made for remediation. A total of 1432 discussions occurred for 1326 patients. The aim of this study was to identify errors occurring in the postdischarge period and factors that predict an increased risk of experiencing an error. In 435 discussions, an issue was identified that required further discussion (known as a transition of care event), and the majority of these were related to medications. In 14.7% of all discussions, a medical error, defined as ‘any preventable event that may cause or lead to inappropriate medical care or patient harm’, was identified. We found that errors were more likely to occur for patients discharged from surgical services or the emergency department (as compared with medical services) and were less likely to occur for patients who were discharged in the morning. This study shows that a number of errors may be detected in the postdischarge period, and the ECHO-CT programme provides a mechanism for identifying and mitigating these events. Furthermore, it suggests that discharging service and time of day may be associated with risk of error in the discharge period, thereby suggesting potential areas of focus for future interventions.
- Transitions in care
- Human error
- Communication
Data availability statement
No data are available.
This is an open access article distributed in accordance with the Creative Commons Attribution Non Commercial (CC BY-NC 4.0) license, which permits others to distribute, remix, adapt, build upon this work non-commercially, and license their derivative works on different terms, provided the original work is properly cited, appropriate credit is given, any changes made indicated, and the use is non-commercial. See: http://creativecommons.org/licenses/by-nc/4.0/.
Statistics from Altmetric.com
WHAT IS ALREADY KNOWN ON THIS TOPIC
The transition from hospital to postacute care is a vulnerable time for patients, especially older adults, and can result in errors in care.
WHAT THIS STUDY ADDS
The Extension for Community Healthcare Outcomes-Care Transitions (ECHO-CT) conference was implemented to improve the transition for patients from hospital to postacute care. Through this programme, we identified a number of potentially adverse clinical and operational events occurring in the transitional period. This article describes the events identified and factors that predict an increased risk of experiencing an error in the transition period.
HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY
This study shows that patients often have ongoing care needs in the postdischarge period that require discussion between the hospital and postacute care team. 14.7% of patients had an error identified in the postdischarge period. Patients discharged from the emergency room were more likely to have an error and a number of these errors were medication related, suggesting that further attention to and intervention in these areas may decrease errors.
Introduction
‘Transitions of care’ refers to the movement of patients between healthcare providers or settings as their care needs change. Unfortunately, these transitions can be challenging as miscommunication and changes in providers and treatment plans can lead to adverse events and increased healthcare costs.1 These transitions disproportionately affect older adults, and a significant percentage of Medicare patients are discharged from the hospital to a skilled nursing facility (SNF) or nursing home. Hospital readmissions from SNFs are high and associated with a higher mortality rate.2 During these transitions, nearly 4 out of 10 patients discharged back to a long-term care facility developed an adverse event, most of which were preventable or ameliorable. These adverse events were similar in urban and rural patients.3 4 Poor communication during the handoff is a common reason, and a number of barriers exist including the physical and social setting, time, language barriers, health literacy and inadequate education and engagement of patients and family caregivers.5–8
Given that poor provider communication is one of the most modifiable causes of poor care transitions, in an attempt to identify and mitigate this problem in 2013, we adapted the Extension for Community Healthcare Outcomes (ECHO) model to develop a programme called ECHO–Care Transitions (ECHO-CT) that aimed to improve the care of patients discharged from the hospital to postacute settings.9 10 The process and outcomes of the ECHO-CT programme have been previously published.11 12
Within the case discussions that occurred during ECHO-CT conferences, we were able to investigate and better characterise potentially adverse clinical and operational events during care transitions, and also assist and influence posthospitalisation clinical decision-making. We characterised each of these discussions as transition-of-care events (TCEs). In a prior study, out of 675 total patients discussed, 139 TCEs were identified; 41.7% involved discharge communication or coordination errors and 37.4% were classified as medication issues.13 Similarly, another study implementing the ECHO-CT model found 327 errors identified in 391 patients discussed; the majority of errors were related to medications and communication.14 In this study, we sought to further classify TCEs, using a system for reporting TCEs developed in our prior study. Furthermore, we subcategorised TCEs in which an error occurred and sought to identify factors that predicted an increased risk of experiencing an error in the transition period in both an academic medical centre and a community hospital.
Methods
The ECHO-CT videoconferences were conducted at a large (720-bed) academic medical centre and smaller (73 bed) affiliated community hospital in the Boston, Massachusetts area from April 2019 to February 2021 (with a disruption between March 2020 and June 2020 due to the COVID-19 pandemic). The programme has been previously described in detail.10 In brief, the programme is a weekly, multidisciplinary videoconference between a hospital-based team and postacute care providers to discuss patients discharged from inpatient services to postacute care sites during the preceding week. The hospital team includes a hospitalist, pharmacist, case manager and programme manager. The team at the postacute care sites includes physicians, nurse practitioners, physical therapists, social workers and case managers. During the videoconference, each patient’s hospital course and discharge documentation are reviewed by a hospitalist and issues related to that patient are discussed with the postacute care staff. This includes a summary of the patient’s hospital course, an update from the postacute care team on the patient’s status, and an opportunity to discuss any concerns or questions raised by the postacute care or inpatient care teams. Additionally, a pharmacist performs a medication reconciliation of each patient’s admission, discharge and postacute care medication lists. Between 10 and 20 patients are discussed in a 2-hour session each week, with each postacute care site getting 10–15 min to discuss their patients.
1432 discussions occurred during our study period. Patients were discussed during the ECHO-CT conference and any medical care discrepancies or concerns from the care teams were recorded as a TCE. This was a retrospective, observational study; therefore, two physicians retrospectively and independently reviewed each TCE to determine if a medical error had occurred. A medical error was defined as ‘any preventable event that may cause or lead to inappropriate medical care or patient harm’.15 Clinical judgement, review of laboratory results or admission documentation was used to make each determination. If there was disagreement, the event was discussed with other team members in an attempt to reach consensus. If consensus could not be reached, the potential TCE was excluded from the analysis (24 TCEs). If a medication was inappropriately changed with no clear documentation of reasoning for doing so, this was considered an error.
Patient demographic and hospitalisation characteristics were obtained from the medical record and included: age, sex (male vs female), race (non-white vs white), college education (graduated college vs other), primary language (non-English vs English), marital status (married vs other), length of stay (days), zip code (to impute median household income), inpatient service (surgical vs medical), location of care (emergency vs inpatient), number of medications on discharge, discharge time (morning vs after noon), discharge day (weekday working hours vs other), hospital type (community hospital vs academic hospital) and Elixhauser score (calculated with RStudio Statistical Software, V.1.4.1717, ‘Comorbidity’ package V.1.0.2), which uses comorbidities to predict mortality risk and is a proxy for severity of disease.16 17 These variables were chosen as we believed they provided insight into each patient’s support system and medical complexity and potential systemic communication barriers at discharge. Missing data were imputed using the MICE package in R, with five iterations.18 Of 1432 discharges, 185 were missing at least 1 variable. Pooled bivariate logistic regression, with generalised estimating equations to account for multiple TCEs (during separate hospitalisations) per patient, was performed to predict TCEs (bivariate results not shown). Variables with p <0.25 in bivariate models were included in a subsequent multivariate logistic regression model, and age and sex were included in the multivariate model regardless of bivariate association. Bivariate and multivariate models were conducted separately for each site, and then combined in a third model. All analyses were conducted using RStudio Statistical Software. An alpha level of 0.05 was used to determine statistical significance in the multivariate model. All data were kept in secure, password-protected files, and were deidentified for the analysis of transitional events.
Results
There were 1432 hospitalisations discussed for 1326 patients (806 discussions were conducted for 666 patients at the community site and 626 discussions were conducted for 660 patients at the academic medical centre). A total of 435 TCEs were identified in 339 (26%) patients; 316 (73%) events were medication related, including discrepancies between home and discharge medication lists, incorrect medication dosages, frequency of medications and/or antibiotic stop dates (see Gonzalez (2021) for additional examples TCEs). Thirteen events (3%) were discharge communication related, such as a discharge summary not included in the transfer paperwork. The remainder were other types of events relating to either advanced directives (unclear goals of care discussion with patient and healthcare proxy) or medical events (changes to management of a clinical condition recommended by SNF physicians based on patient condition). For 350 (80%) of TCEs discussed, a change to the plan of care was recommended by the ECHO-CT multidisciplinary team.
Of the 1432 hospitalisations analysed, 211 (14.7%) had TCEs classified as medical errors (110 at the academic site and 101 at the community site). A greater number of errors was found in the community site compared with the academic site (16.1% and 13.6%, respectively), though this was not statistically significant (p=0.188). Discrepancies between home and discharge medication lists were some of the most common errors. Other examples of errors included unclear instructions for venous thromboembolism prophylaxis and weight-bearing status on discharge as well as ambiguities regarding medication dosage, frequency and stop date. The demographic and clinical characteristics of patients who experienced errors compared with patients without errors are shown in table 1. Notably, there were more men in the error group compared with the non-error group at the academic site; however, this was the opposite in the community site. The group of patients with errors at the academic site had a larger proportion of white race compared with the non-error group. The proportion of college-educated patients tended to be higher in the non-error group, but this was not statistically significant. For the academic medical centre site, race, college education, marital status, primary language, discharge time, location of care, inpatient service, discharge time and discharge day were marginally associated with transitional events (p <0.25) in bivariate logistic regression and therefore included in the multivariate analysis, along with age and sex. In the community site, sex, location of care and Elixhauser score were marginally associated with events (p <0.25) in the bivariate regression and were therefore included in the multivariate regression, along with age. Because race was coded differently between the hospitals and college education was not available in the community hospital, they were excluded in pooled analysis across both sites. Across both hospital sites, primary language, Elixhauser, inpatient service, location of care, discharge time and hospital type were marginally associated with events in bivariate regression at p <0.25 and therefore included in the multivariate regression, along with age and sex.
Patient demographics and clinical characteristics across participating hospitals
Most significantly, in multivariate regression (table 2) at the academic hospital site, patients who were on surgical services or discharged from the emergency department had a higher risk of a medical error during their transitions of care (1.61 OR, p=0.039; 4.85 OR, p=0.007, respectively). Black patients had a lower risk of medical error during their transitions of care (0.56 OR, p=0.047) compared with white patients.
Adjusted risk ratios and 95% CIs for the association between potential risk factors and TCEs
In multivariate regression at the community hospital site, patients discharged from the emergency department or who had lower Elixhauser scores had a higher risk of a medical error during their transitions of care (11.4 OR, p <0.001; 0.97 OR, p =0.008, respectively).
In multivariate regression across both hospital sites, patients who were on surgical services or discharged from the emergency department had a higher risk of a medical error during their transitions of care (1.71 OR, p =0.016; 9.02 OR, p <0.001, respectively). Patients discharged during the morning had a lower risk of a medical error during their transitions of care, compared with those discharged in the afternoon. (0.43 OR, p =0.049).
Discussion/conclusion
Our study demonstrated that 26% of patients discharged from the hospital to an SNF experienced a TCE in the postdischarge period, and the ECHO-CT programme effectively detected these potential adverse events in an attempt to mitigate their impact. In 14.7% (211) of all discharges, a medical error was detected (49% of TCEs). We found that patients discharged from the emergency and surgical departments had a higher likelihood of experiencing a medical error compared with the medical services and these results were similar for discharges from both hospital sites. This difference may be because many errors were medication related, and in our experience medication reconciliation is more commonly prioritised on the medical services compared with emergency and surgical services. This issue was also highlighted in a study of 36 patients who were discharged from the ED and 7 days following discharge were found to have a medication discrepancy rate of 42%.19
We also observed fewer errors occurring for patients discharged during the morning hours. This may be due to improved staffing levels at the postacute care facilities and acute hospital at the time that the patient was discharged. Additionally, the workflow on inpatient services (eg, more admissions later in the day) may allow staff to focus more time on discharges in the morning, and therefore commit fewer errors.
Based on the experience of us and others showing increased risk of error with increased number of diagnoses and more medication errors in the postdischarge period,19–22 we expected to see an increase in risk of error occurring with increasing medical complexity. However, this was not observed. In fact, higher Elixhauser scores were associated with fewer errors, possibly because these patients received more attention from staff involved in discharge planning. Additionally, if patient comorbidities were not accurately coded, this score may not be an accurate reflection of illness severity, and therefore may not have correlated with errors as we expected.
There was a decreased risk of errors for black patients at the academic site; however, no other demographic variables decreased the risk of errors in our study. While this result may suggest that clinicians pay closer attention to vulnerable patients and standardise care for such patients on discharge, it may also reflect lack of recognition of errors in this patient population. The type of discharging hospital (academic vs community) also did not affect the patient’s risk of an error occurring. Taken together, these findings suggest that future interventions could be targeted to specific populations for maximum benefit. For example, hospitals could use pharmacy technicians to help with medication reconciliation—especially for emergency and surgical departments, provide education for patients with low medical literacy or English language proficiency, incentivise morning discharges and implement other quality improvement programmes to mitigate errors in the discharge period, such as protocols to encourage adherence to established best practices.23–25
Our study has several notable limitations. We used a subjective determination of an error, but adjudicated this with two physician reviewers who needed to reach consensus to establish an event as an error. Given the retrospective observational design, misclassification bias is possible. We were unable to validate the errors by identifying adverse consequences associated with them and we also could not exclude the possibility that the errors were corrected; however, this was not documented in the record or specifically discussed in the conference and therefore not evident to us. Given that all physician reviewers were internists, their assessment of errors may have been biased in favour of medical discharges compared with other services; however, the discharging service was not always known by the reviewers. Additionally, there may have been errors that were not identified through ECHO-CT, resulting in an underestimation of errors that occurred in the transition period. Despite these limitations, this study suggests that there is a significant opportunity for ECHO-CT to identify potential errors and improve care in the postdischarge period.
Data availability statement
No data are available.
Ethics statements
Patient consent for publication
Ethics approval
The study was approved by the BIDMC and Brown University IRBs, which waived the requirement for subject consent.
References
Footnotes
ERB and AS are joint first authors.
Contributors EB: methodology, software, formal analysis, investigation, data curation, writing-original draft, review and editing. AS: investigation, data curation, writing-original draft, review and editing. LJ-M: methodology, investigation, data curation, writing-original draft, review and editing, project administration. KK: investigation, data curation, writing-review and editing. ABD: methodology, formal analysis, writing-review and editing. LL: conceptualisation, methodology, investigation, resources, writing-review and editing, supervision, funding acquisition. AM: conceptualisation, methodology, formal analysis, investigation, data curation, writing-original draft, review and editing, supervision, guarantor.
Funding Grant number R01HS025702 from the Agency for Healthcare Research and Quality.
Competing interests None declared.
Patient and public involvement Patients and/or the public were not involved in the design, or conduct, or reporting, or dissemination plans of this research.
Provenance and peer review Not commissioned; externally peer reviewed.