Methods
We conducted a prospective cohort study including all adult patients (age ≥18) transferred from any acute care hospital to the general medicine (GMS), cardiology, oncology or intensive care unit (ICU) services at an 792-bed tertiary care hospital in Boston, Massachusetts. This study was approved by the hospital’s Institutional Review Board.
Due to lack of direct interaction with patients as part of this study, patients or the public were not involved in the design, or conduct, or reporting, or dissemination plans of our research.
Study population
Patients were included if they were transferred to any of the included services during our baseline data collection period between August 2020 and June 2021 and during our intervention data collection period between August 2021 and June 2022. We excluded patients transferred from somewhere other than an acute care hospital (eg, emergency room transfers, direct hospital admissions). We also intentionally excluded patients who were transferred during July 2021 to allow for a sufficient wash-in period for our intervention. Notably, we were unable to include ICU service patient transfers between August 2020 and November 2020 (4 months of baseline time period) due to a delay in the decision to incorporate ICU service transfers in our cohort, or any patient transfers during September 2021 (1 month of the intervention time period) due to change in research support personnel.
Baseline practices
At baseline (ie, prior to this intervention), documentation of clinical information regarding patient transfers was variable. Nurses that staff the hospital’s ‘Access Centre’, responsible for coordinating all patient transfers to our hospital and documentation of clinical information at the time of IHT patient acceptance, would document some clinical information within the ‘Transfer Module’ in the EHR. Documentation of a more formal Accept Note at time of patient acceptance was variable and expectations for documentation depended on the service of admission. There was no expectation for documentation among patient transfers to the ICU or oncology services, though some accepting clinicians chose to document via a note within the EHR. Among patient transfers to GMS or cardiology, there was an expectation to document a note within the EHR, though this responsibility was diffused across all attending physicians responsible for accepting patients for transfer, and there were not standard guidelines for documentation, leading to unreliable and varied documentation practices.
Design of the standardised accept note
Over the course of 9 months (September 2020 through June 2021), we engaged a group of key stakeholders within the hospital involved in IHT, including: (1) medicine residents and physician assistants responsible for admitting IHT patients at the time of arrival; (2) attending medicine physicians responsible for accepting IHT patients for transfer and (3) frontline nurses and leadership within our hospital’s ‘Access Centre responsible for coordinating all patient transfers to our hospital and documentation of clinical information at the time of IHT patient acceptance.
The working group met bi-weekly and used foundational quality improvement tools to develop a standardised accept note for use among all IHT patients to included services, including:
Generation of a process map of collection and documentation of clinical information at time of patient acceptance for transfer, used to identify key barriers to access and documentation.
Identification of most essential clinical information to include within a standardised accept note (eg, from the perspective of admitting and accepting clinicians).
Feasibility of documentation strategies (eg, from the perspective of Access centre nurses) using a priority/pay-off matrix.
Based on multidisciplinary working group input, a draft standardised accept note was created using Illness severity, Patient summary, Action items, Situational awareness, Synthesis by receiver (IPASS) format,23 given clinician familiarity and use of IPASS among other care transitions within our hospital.25 26 The IPASS standardised accept note was iteratively refined with group input and additional stakeholder input outside of the working group and converted into a ‘dot-phrase’ (eg, templated note within Epic EHR) (figure 1).
Figure 1Standardised accept note template. The standardised accept note was implemented as a ‘dot-phrase’ within the electronic health record (Epic). The note was developed in IPASS format (Illness Severity, Patient Summary, Action Items, Situational Awareness, Synthesis)23 with rearrangement of the data elements to better match the workflow of the Access centre nurses completing the documentation. Fields marked as ‘***’ indicate areas that require documentation. Fields highlighted in yellow indicate drop-down options within the note template. Blue text within the box includes instructional text to help guide accurate selection of patient illness severity. Black test within the box at the end of the note directs note readers to additional clinical information within the patient electronic medical record. All data fields were developed with input from key stakeholders over the course of 9 months. HCP, Healthcare Proxy; MAR, Medication Administration Record; HR, Heart Rate; BP, Blood Pressure; RR, Respiratory Rate.
Of note, in addition to the working group’s primary goal of developing a standardised accept note, the generated process map identified that lack of advanced notification of the impending patient transfer was an additional key barrier to the IHT process. Prior work had targeted improving advanced notification of GMS patients for admitting clinicians via implementation of a new workflow for Access Centre nurses to page the admitting clinician at the time that the incoming IHT patient received a bed assignment in the receiving hospital (in advance of arrival).27 Thus, in addition to implementation of the developed standardised accept note, the decision was made to expand the advanced notification initiative to page the admitting clinician for IHT patients across all services, including cardiology, oncology, and the medical ICU.
Implementation of the standardised accept note
The IPASS standardised accept note was implemented for use within our EHR in July 2021. At the time of implementation, an instructional tutorial was created to guide users on accessing and documenting the standardised note within the patient’s medical record. All access centre nurses (ie, those responsible for coordinating all hospital transfers, conducting a three-way conference call with the transferring and accepting clinicians and for documentation of the accept note at time of patient acceptance for transfer) were trained using this tutorial with oversight from access centre leadership and directed feedback when necessary. Use of the standardised accept note went live for use on GMS and cardiology in July 2021. Use among oncology patients was delayed until November 2021 due to limited staffing within the Access centre to conduct the documentation. Use among ICU patients was delayed until after completion of the study, also due to limited staffing.
Implementation of the expanded advanced notification paging initiative went live across all services (ie, beyond GMS for which it was already in use) in November 2021.
Data collection: primary outcome
The primary outcome was clinician-reported medical errors, obtained via a survey sent to admitting clinicians within 72 hours after IHT patient admission during baseline and intervention time periods (excluding the 1-month wash in period following implementation of the intervention). Admitting clinicians were identified via an administratively generated daily transfer report that listed that day’s eligible patients (ie, patients transferred to included services), and the author of the admission History+Physical (ie, H+P) note, indicating the admitting clinician. The transfer report was generated daily Monday through Friday, with Monday’s report, including all patient transfers and H+P note authors from the weekend prior. A trained research assistant reviewed the daily report to ensure accuracy of eligible patients and admitting clinicians, and then emailed a survey for each patient transfer to the corresponding admitting clinician via REDCap (Vanderbilt University, Nashville, TN), a free, secure, HIPAA compliant web-based application hosted by MGB. Survey questions were developed based on similar studies23 28 29 and asked about medical errors experienced by the patient following the transfer, including: unnecessary tests, procedures, medications, fluids or other therapies; delays in tests, procedures, medications, fluids or other therapies; orders written that were erroneous but intercepted before reaching the patient (near misses) or any other medical errors (online supplemental appendix A). The survey was sent daily up to three times via REDCap. After 3 days, any non-responding clinicians were sent a final email by the study principal investigator (SM). Survey respondents were given $5 USD gift cards for each completed survey. The presence of a clinician-reported medical error was defined as an ‘yes’ response to any of these questions. To obtain our primary outcome of clinician-reported medical error rates, we divided the total number of clinician-reported medical errors by 100 to obtain rates/100 patient transfers.
Secondary outcomes
Secondary outcomes included Clinician-reported failures in communication gathered via the data collection survey (above) and defined as the respondent answering ‘yes’ to any of the following: receipt of inaccurate patient information; missing important information about the patient; feeling uncertain about management decisions due to lack of clinical information; needing to spend extra time learning about the patient due to lack of clinical information; unable to provide accurate or complete information to the patient, family member or other member of the care team and the patient was more unstable than anticipated based on information received.
Additional secondary outcomes were collected administratively and included: presence of an accept note, defined as the presence of a documented accept note within the patient’s medical chart within 3 days of patient admission; ‘timeliness’ of the accept note, defined as the number of hours between accept note documentation and patient admission; patient rapid response or transfer to the ICU within 24-hours of patient admission, defined as a composite outcome for patients who were initially transferred to non-ICU services at the accepting hospital; patient LOS following transfer and patient in-hospital mortality (ie, death during hospitalisation following transfer).
Covariates
We additionally collected the following variables administratively for use in our adjusted analyses: patient demographics (age, gender, race, ethnicity); patient insurance categorised as Private, Medicare, Medicaid, other; income category by zip code, categorised into quartiles; diagnosis category, using standard ICD-10 grouper algorithms on the principal problem of the Hospital Problem List on admission; patient comorbidity as measured by the Elixhauser score30; illness severity as measured by the eCART score on admission, which uses vital sign and laboratory data to predict cardiac arrest, ICU transfer or death among hospitalised patients31; clinical service at time of admission; time of year of transfer categorised into quarters to adjust for training effects on residents and seasonal case mix; admitting team census on date of patient admission,32 to account for clinician work-compression on day of patient transfer; Census of the admitting service on date of patient admission, categorised into ‘normal’ vs ‘extreme’ census based on previously defined thresholds by service within the hospital, to account for variations in the hospital’s threshold of acceptance for transfer; and hospital COVID-19 census on date of patient admission, defined as the total number of hospitalised patients with COVID-19, to account for pandemic-related spikes in hospitalisation that occurred during data collection periods.
Analysis
Survey response rate was generated throughout and on completion of data collection. We first compared patient and transfer process characteristics between baseline and intervention patients and additionally between patients admitted by survey responders and patients admitted by survey non-responders.
For primary and secondary outcomes, we conducted a series of univariable then multivariable intention-to-treat analyses to evaluate the association between implementation of the standardised accept note and each outcome, adjusting for all covariates (SAS Statistical Software, Cary, North Carolina), notably using different cohorts depending on the outcome due to specific cohorts included in each outcome (online supplemental appendix B). For our primary outcome (clinician-reported medical error rates), we conducted Poisson regression analyses to evaluate the adjusted association between the standardised accept note implementation and clinician-reported medical error rates. Similar Poisson regression analysis was conducted to evaluate impact on clinician-reported failures in communication. For the outcomes of presence of accept note and in-hospital mortality, we conducted multivariate logistic regression models, including all patients in the cohort; for timeliness of the accept note, we conducted multivariate linear regression; for patient LOS following transfer, we conducted adjusted linear regression with gamma distribution (ie, using log LOS) to account for non-normal distribution (right-skew) of this outcome; and for the composite outcome of rapid-response or ICU transfer within 24 hours of patient admission, we conducted multivariate logistic regression including all patients in the cohort except for those that were initially transferred to ICU services.
Despite notable variation in the implementation of accept note documentation by service (as described in Methods section, implementation among oncology patients was delayed, and implementation among ICU patients did not begin during the study period), we conducted the aforementioned multivariable analyses of primary and secondary outcomes among all services combined on an intention-to-treat basis. We felt that an intention-to-treat analysis was important both due to the expansion of the pre-existing advanced notification initiative that was implemented as a result of the stakeholder work generated by this study as well as due to clinician cross-over between included services (eg, medical residents that rotate across all included services). However, to address the notable variation in implementation, we additionally conducted: (1) an unadjusted stratified analysis of documentation of accept note documentation by service (GMS, cardiology, oncology and ICU) and (2) an intention-to-treat multivariable analyses of primary and secondary outcomes among all services except for ICU service. Two-sided p values <0.05 were considered significant in all analyses.