Article Text
PDF
Abstract
Background Readmissions of chronic obstructive pulmonary disease (COPD) have devastating effects on patient quality-of-life, disease progression and healthcare cost. Effective interventions to reduce COPD readmissions are needed.
Objectives Reduce 30-day all-cause readmissions by (1) creating a COPD care bundle that addresses care delivery failures, (2) using improvement science to achieve 90% bundle adherence.
Setting An 800-bed academic hospital in Ohio, USA. The COPD 30-day all-cause readmission rate was 22.7% from August 2013 to September 2015.
Method We performed a cross-sectional study of COPD 30-day readmissions from October 2014 to March 2015 to identify care delivery failures. We interviewed readmitted patients with COPD to identify their needs after discharge. A multidisciplinary team created a care bundle designed to mitigate system failures. Using a quasi-experimental study and ‘Model for Improvement’, we redesigned care delivery to improve bundle adherence. We used statistical process control charts to analyse bundle adherence and all-cause 30-day readmissions.
Results Cross-sectional review of the index (first-time) admissions revealed COPD was the most common readmission diagnosis and identified 42 system-level failures. The most prevalent failures were deficient inhaler regimen at discharge, late or non-existent follow-up appointments, and suboptimal discharge instructions. Patient interviews revealed confusing discharge instructions, especially regarding inhaler use. The COPD care-bundle components were: (1) appropriate inhaler regimen, (2) 30-day inhaler supply, (3) inhaler education on the device available postdischarge, (4) follow-up within 15 days (5) standardised patient-centred discharge instructions. The adherence to completing bundle components reached 90% in 5.5 months and was sustained. The COPD 30-day readmission rate decreased from 22.7% to 14.7%. Patients receiving all bundle components had a readmission rate of 10.9%. As a balancing measure for the targeted reduction in readmission rate, we assessed length of stay, which did not change (4.8 days before vs 4.6 days after; p=0.45).
Conclusion System-level failures and unmet patient needs are modifiable risks for readmissions. Development and reliable implementation of a COPD care bundle that mitigates these failures reduced COPD readmissions.
- Quality improvement
- Chronic disease management
- Teamwork
Statistics from Altmetric.com
Introduction
Chronic obstructive pulmonary disease (COPD) is a highly prevalent, chronic disorder with significant effects on morbidity, mortality and healthcare costs. COPD exacerbations are the leading cause of healthcare utilisation and cost in COPD care.1 2 In the USA, there are approximately 7 00 000 hospitalisations yearly for COPD exacerbations. One in five patients require rehospitalisation within 30 days of discharge,3–5 which adds considerably to the cost of COPD care. Reducing these early readmissions is therefore a high priority and is a standard performance measure used to judge quality of healthcare delivery.6 7
In 2014, the US Centers for Medicare and Medicaid Services (CMS) included COPD in the Hospital Readmission Reduction Program, incentivising hospitals to reduce 30-day all-cause readmissions after an initial hospitalisation for COPD.8 Despite these incentives and numerous controlled -trials for COPD therapeutics, the optimal approach to reduce COPD readmissions remains elusive.9–11 Many recommendations to reduce readmissions are based on controlled clinical trials in highly defined and selected populations which do not closely match real -life case definitions of COPD.12 There is a dearth of evidence for effective methodologies that bridge the gap between controlled trials and day-to-day care practices that reduce COPD readmissions.13
Disparities in COPD readmissions among hospitals may be reflective of varied adherence to known best practices.14 Among general medicine patients, more than a quarter of 30 -day readmissions can be attributed to a preventable cause due to failure of optimal care delivery.15 The prevalence of such preventable readmissions can vary from 9% to 48%.16 Care bundles are one way to improve adherence to best practices. Although many different COPD discharge bundles have been developed, they have been variably effective.17 Translation of successful bundles into practice is often limited by poor reporting of bundle component selection process and method of implementation. In addition to improving system-level care -delivery factors, interventions that increase patient empowerment, such as education in disease management and inhaler device training, also reduce readmissions.18 19
In this study, we investigated system-level failures and unmet patient needs to identify gaps in the care delivered to patients who were readmitted within 30 days after a COPD exacerbation admission. We integrated these findings with best -practices in COPD management to create a practical five-element COPD care bundle designed specifically to mitigate system-level failures and fulfil patient needs. In our quasi-experimental study, we tested our theory that reliable adherence to our COPD care bundle would reduce COPD readmissions.
Setting
This project was performed at an 800-bed, academic health centre that serves as a regional referral hospital in USA. There are nine general internal medicine teams and four medicine subspecialty teams (cardiology, gastroenterology, nephrology and haematology-oncology). Patients admitted with COPD exacerbations are assigned to any of nine medicine teams with occasional assignment to subspecialty teams. Medicine patients are distributed on six hospital units (non-geographical) including one medical step-down unit.
Each medicine team is composed of an attending physician, house-staff residents (one to three) and medical students (zero to two). Each team is assigned a care coordinator to assist in discharge planning. Each care coordinator is assigned to two to four teams. There are three inpatient pharmacists who assist all general medicine teams. On weekends, there is one pharmacist for all teams. There can be up to 12 patients on each general medicine team and 16 patients on each subspecialty team. One respiratory therapist (RT) is assigned to one or two floor units, and typically cares for approximately 30 active patients. The electronic medical record (EMR) system is Epic (General Electronics, Wisconsin, USA).
COPD readmission definition and local problem
COPD readmission was defined by CMS guidelines as any unplanned readmission within 30 days of discharge after an index admission for COPD exacerbation. An index admission is one with a primary diagnosis of COPD exacerbation or a primary diagnosis of respiratory failure with COPD as a secondary diagnosis8 (see online supplementary eTable 1).
Among all patients, the baseline COPD 30-day all-cause readmission rate before starting this project was 22.7% from August 2013 to September 2015 (figure 1A). Among the Medicare and Medicaid patients the observed to expected COPD readmission ratio was 1 from June 2011 to July 2014,20 suggesting that our performance was ‘as expected’ for our patient population and case-mix index.
Baseline system performance and bundle development: (A) Control chart (p-chart) of baseline 30-day all-cause COPD readmissions. Only common cause (random, chance) variation is observed. No special cause variation occurred during the 2-year period. The 6-month period sampled and studied for system failures analysis is identified by a bold box from October 2014 to March 2015. (B) Pareto chart for system-level failures at the time of index discharges during the baseline sample period. Failure categories are aligned on the x axis in descending order from left to right. The cumulative percentage of failures from left to right are shown with an overriding curved line. *Late appointment is defined as >30 days after discharge. **Likely under-represented on retrospective chart review. This failure was selected if there was a mention of lack of medicine supply in a patient’s chart at readmission. +Influenza and pneumococcal vaccines. COPD, chronic obstructive pulmonary disease; f/u, follow-up; meds, medications.
Methods
Cross-sectional analysis to identify system failures
As a first step, we performed a cross-sectional analysis to understand failures in care delivery through retrospective chart review and patient interviews. A convenient sample of 2 years was performed to measure baseline COPD 30-day all-cause readmissions rate. Since the variations in COPD readmission rate during the baseline period showed only common cause (random, chance) variation, we postulated that a 6-month representative sample period for detailed analysis would be sufficient to identify system failures. We performed a chart review of all readmitted patients during the 6-month period from October 2014 to March 2015 (figure 1A). Patient variables included age, gender, forced expiratory volume in 1 s (FEV1), Charleston Comorbidity Index Score,21 smoking status and psychological disorders. Care delivery variables included length of stay (LOS), discharge disposition, vaccination status at discharge, COPD medications prescribed at discharge, follow-up appointments, number of readmissions during the preceding 6 months, time to readmission and reason for readmission. Pareto analysis was performed for identifiable failures in care delivery in order to rank the failures by their frequency and understand which interventions will be needed to mitigate the majority of the failure (figure 1B).22
Over a 2-week period, we interviewed all COPD readmission patients, during their rehospitalisation to obtain accurate and relevant information about their recent discharge. Patients were asked three structured, open-ended questions: (1) How was your experience at home after your last hospital discharge? (2) What in your opinion led to this readmission? (3) What could we have done to help prevent this readmission? Qualitative data from these interviews were used to understand unmet patient needs after discharge.
Creating the COPD care bundle
A multidisciplinary team was formed to develop the COPD care bundle and was composed of four hospitalists, two pharmacists, two RTs, two care coordinators, two nurses and a pulmonologist. They reviewed published literature on best practices and COPD guidelines to identify gaps in care.23 The number of bundle components was limited to five to ensure feasibility and maintain the focus on high yield items. Items that could apply to all patients were given higher priority. Consensus was defined as simple majority. Different ideas for bundle components were discussed before finalising the bundle in two team meetings.
Measures
We created a COPD bundle with five components. We defined an opportunity each time a bundle component should have been performed as part of care. The process measures were (1) Adherence to COPD care bundle components (% of bundle components completed out of all available opportunities) and (2) Adherence to each individual bundle component. The primary outcome measure was rate (%) of 30 -day all-cause COPD readmissions. Secondary outcomes were readmission rates based on number of bundle components received out of all available opportunities and time to readmission. Hospital LOS was used as a balancing measure.
System redesign and PDSA cycles
We used the Model for Improvement24 for system transformation. The specific aim of the project was to increase adherence with COPD bundle components to 90%. The existing process was mapped and failure modes were highlighted by direct observation of processes and input from front-line workers. Newly proposed processes were explored through iterative improvements by rapid tests of change, plan-do-study-act (PDSA) cycles. Multiple sequential PDSA cycles were conducted to optimise new processes, interactions and care delivery.
Through these tests, new knowledge of failure modes and new ideas for interventions were developed. Real time feedback was obtained from patients and front-line care providers for each new intervention that was tested, and interventions were either abandoned, adapted or adopted based on tested outcomes and feedbacks. Successful interventions were implemented. Further modifications were made in EMR to increase reliability of key processes involved in COPD bundle completion.25
Data collection
Data for COPD index admissions and readmissions were obtained retrospectively each month from the EMR based on billing codes using ICD-9 and ICD-10 (see online supplementary eTable 1). Patients with a primary diagnosis of COPD or a primary diagnosis of respiratory failure with COPD as secondary diagnosis were identified as index cases using ICD codes according to CMS guidelines.8 Hospitalised patients with COPD were identified in real time by custom-built EMR reports that identified all patients with COPD in the active problem list. COPD exacerbation cases were identified though daily manual chart review of patients on this list. Adherence to the COPD bundle was recorded by chart review (process measure). For outcome measures (COPD readmission rate), we used the billing data to avoid measurement bias, since the baseline readmission rate was also calculated based on billing data. At the end of each month, data from both sources, that is, retrospective billing codes and real time COPD exacerbation cases from EMR report, were compared to identify any missing cases to measure accurately the bundle adherence and prevent fall-out of any case of readmission at the extreme ends of the month (end effects).
Analysis
Interrupted time-series analysis was performed using statistical process control (SPC) methods.26 27 Specifically, annotated control charts (p-chart) were developed and updated monthly for outcome measures and weekly for process measures prospectively using software package QI-charts V.2.0.23 (Process Improvement Products, USA) for Microsoft Excel 2013 (Microsoft, USA). The initial control limits and centre line (mean) were calculated using baseline data for outcome measures and by using first eight observations for process measures. Standard industrial criteria for special cause and system shift were used to determine if observed changes in measures were due to a specific assignable cause.26 27 A system shift was identified if eight consecutive measurements were persistently above or below the mean. New control limits and centre line were calculated if a system shift was observed. Secondary outcome measures were compared using unpaired Student’s t-test and χ2 test using SPSS V.21 (IBM, USA). A p value less than 0.05 was considered statistically significant.
Ethical consideration
This quality improvement initiative was approved by the Institutional Review Board at our institute and was designated as not human subject research as the aim was to improve outcomes and reliably deliver best practice care.
Results
Cross-sectional review and identification of system-level failure
During the 26 months baseline period (August 2013 to September 2015), the COPD readmission rate was 22.7% with common cause variation only (figure 1A). During the 6-month sample period (October 2014 to March 2015), there were 207 index admissions and 44 readmissions among 27 patients with mean (±SD) age 62.3 (±9.2) years, FEV1 % predicted 41 (±19.8)%, BMI 29.8 (±12.5) kg/m2 and Charleston Comorbidity Index Score of 5.1 (±2.4). The cause of readmission was COPD exacerbation in 55.5%, congestive heart failure in 13.3% and pneumonia in 7%. Other infrequent causes were cancer, fall, influenza, psychosis, cirrhosis and cardiac arrest. Median and mean (±SD) time to readmission were 7 days and 10.3 (±7.8) days and time to follow-up appointment were 10 days and 20 (±20.8) days. There were 42 system-level failures at discharge of index cases and their distribution frequencies are shown in the Pareto chart (figure 1B).
Three patient interviews were conducted. Two patients had poor understanding of inhalers with confusion between different inhaler types, colours and usage technique. Use of different respiratory medications during hospitalisation and at home, frequent change of inhaler brands at refills due to insurance formulary preferences, and lack of formal training in inhaler use contributed to their difficulties with inhaler use. All three patients expressed difficulty with understanding discharge instructions and none were provided an action plan in case of symptom worsening.
COPD care bundle
Items that were considered during team consensus for COPD bundle are presented in table 1.18 23 28 The final five components selected for the bundle were: (1) appropriate inhaler regimen, (2) 30-day inhaler supply, (3) inhaler education on the device available postdischarge, (4) follow-up appointment within 15 days and (5) standardised patient-centred discharge instructions. The bundle development process and the bundle components are summarised in online supplementary eFigure 1.
Items reviewed for COPD care bundle inclusion during multidisciplinary team consensus meetings
All components of the bundle were applicable to patients who were discharged to home (five opportunities). For patients who were discharged to skilled nursing or assisted living facilities, some components such as 30-day medication supply or follow-up within 15 days may not be applicable given available resources at the facility (three to four opportunities).
System redesign and PDSA cycles
The final key driver diagram which served as our theory to guide testing, is presented in online supplementary eFigure 2. The team designed a new process that served as the initial template for testing and modifying the existing process (figure 2A,B). A total of 36 PDSA cycles were performed to test different interventions within each team and their interactions with each other. A summary of PDSA cycles performed along with their associated challenges, learnings and ultimate status of each intervention is presented in online supplementary eTable 2. The chronological order of PDSA cycles is highlighted in the annotations on the control chart for bundle adherence displayed in figure 3A. Failure modes that were encountered at each step of the process and interventions used to mitigate them are shown in failure mode effects analysis in online supplementary efigure 3. The final process map after system redesign is shown in figure 2C.
Process map and roles of team members in COPD care: (A) Previous (baseline) process. Roles and tasks were not well described. Respiratory therapists were expected to educate patients about inhaler use. However, education was provided inconsistently and used hospital formulary brands that might differ from the discharge brands. Only patients labelled as high risk for readmission were seen by care coordinators. There were no standardised expectations or processes for scheduling follow-up appointments, or providing inhaler education and discharge instructions. Discharge medication prescriptions were ordered on the day of discharge and the discharge pharmacy did not have enough lead time to coordinate co-pays and provide 30 days of medications. Frequently, patients left with prescriptions to be filled from other pharmacies or were prescribed medications that were not covered by their insurance. (B) Initial system redesign plan proposed by the multidisciplinary team prior to plan-do-study-act (PDSA) cycles. Many processes and roles were ambiguous and were refined and clarified based on knowledge gained from PDSA cycles. (C) The final process map after system redesign outlining the roles and tasks for all members of the team. There is standardisation of role assignments, distribution of work among team members, and ease of communication. EMR (Epic) tools were developed to increase reliability of key processes and new educational tools were created for patients. *Level of reliability – 1 (90% success). **Level of reliability – 2 (95%–99% success). *^Level of reliability – 1, with added redundancy. COPD, chronic obstructive pulmonary disease; RT, respiratory therapist; Pt, patient.
(A) Control chart (p-chart) for adherence to COPD care bundle components. Annotations describe the state of interventions being performed. Initial control limits were formed using first eight observations. The control limits were recalculated each time a ‘system shift’ was observed with eight consecutive data points above the mean attributable to the interventions performed (annotations). Three system shifts are observed over time. Bundle adherence reached nearly 90% on 19 March 2016 and was sustained. All temporary communications and alerts were stopped on 16 April 2016 and the process has been sustained by design and ongoing monitroing. (B) Control chart (p-chart) for 30-day, all-cause COPD readmissions. A system shift is observed since initiation of COPD care bundle testing. A new stable system with mean readmission rate of 14.7% has been established. COPD, chronic obstructive pulmonary disease; RT, respiratory therapist; EMR, electronic medical record.
Interventions of higher reliability (level 2 reliability with 95%–99% success rate) were applied to key processes of bundle execution.25 These included an EMR automated best practice alert for physicians to use the COPD care bundle for appropriate patients, an EMR COPD care bundle order set that activates the bundle and communicates roles of all team members, and standardised EMR note templates with prepopulated text for pharmacy (to identify insurance-compatible inhaler brands), RT (inhaler education note) and standardised discharge instructions.
New patient education kits were developed that included handouts for inhaler instructions for each inhaler brand with illustrations and easy readability, practice inhalers for each brand type and a ‘my COPD care’ worksheet. This patient-centred worksheet was designed to be reviewed with the patient by RT to educate them about what to expect during their hospital stay, how to manage worsening symptoms at home and how to prepare for follow-up appointments (see online supplementary eFigure 4). All inhaler handouts and the ‘my COPD care’ worksheet were designed with grade 6 readability.29
Outcomes
Process measure: After initiation of PDSA cycles, we observed three system shifts in bundle adherence. Adherence increased from 50% at the start of the project in October 2015 to 89.1% in March 2016 and was sustained at this level (figure 3A and online supplementary eFigure 5).
Outcome measure: COPD all-cause 30-day readmissions declined from 22.7% to 14.7% with a system shift in SPC (figure 3B) associated with COPD care bundle utilisation during the period October 2015 to July 2016 (table 2).
Comparison of COPD Index and readmission rates and patterns of readmission before and after COPD care bundle implementation
Among the patients who were discharged home, those who received four or five components had readmission rates of 12.5% and 10.9%, respectively, which was lower than those who missed two or more bundle components, calculated at 26.5% (p<0.05). The effect of the level of adherence to COPD bundle on readmission patterns is shown in online supplementary eTable 3.
Discussion
In this study we describe an approach to reduce 30-day all-cause COPD readmissions using improvement science methodology. Through comprehensive understanding of system-level failures and unmet patient needs at our hospital, we developed an institution-specific COPD care bundle through multidisciplinary team consensus and COPD literature review. Then, using Model for Improvement and sequential tests of change (PDSA) cycles, we redesigned the care system for reliable delivery of the COPD care bundle to all patients with COPD exacerbations. Consistent, systematic implementation of the COPD care bundle led to a system shift producing a sustained decrease in all-cause, 30-day COPD readmissions from 22.7% to 14.7%. Patients who received all five components of the bundle had a readmission rate of 10.9%.
Despite a large body of evidence of effective interventions that reduce COPD exacerbations,23 their utilisation in day-to-day care is low and patient needs remain unmet. Major gaps in care transition from acute to community care persist30 and there are widespread regional variations in COPD readmission rates.31 Care bundles have been used to close this unbridged gap. Many different COPD discharge bundles using a wide variety of interventions have been reported in the literature with varying degrees of success.17 Poor reporting of bundle component selection and implementation methods limit the ability to reach conclusions and translate successful bundles into practice.17 Improvement bundles that are tested in controlled clinical trials are non-pragmatic due to exclusion of a significant portion of the COPD population who do not meet strict study selection criteria and are plagued by an inability to translate implementation methodology from controlled study settings that depend heavily on temporary research staff, care coordinators and financial resources.32–35
In our improvement study, we describe a successful strategy of mitigating system failures and improving outcomes by better defining the context of improvement, describing the methodology of bundle composition, identifying people and teams involved, providing a detailed process map of the redesigned system, and reporting both successful and unsuccessful testing. Due to a deliberate effort to develop a durable and reliable system, the effect of this intervention persists beyond the temporary testing period and resources. Our theory was based on data and systems learning and our implementation followed a scientific approach to develop a reliable system.24 All patients with a documented, clinical diagnosis of COPD exacerbation were included, replicating real life clinical management. The COPD bundle components used in this study had varying degrees of existing literature support and had a strong local need based on identified system failures (table 1). Redesigning the system to reliably deliver the bundle facilitated the execution of evidence based medicine (EBM) by front-line staff. In this manner, we were able to make EBM a system property rather than an individual responsibility.
For each health delivery system, an assessment of local failures and unmet patient needs will be needed to guide the selection of bundle components in order to achieve improved outcomes given large variations in care delivery.36–38 Our approach serves as a road map for improvement methodologies to develop and implement effective bundles to reduce readmissions for COPD and possibly other chronic diseases.
There are several limitations to our study. First, we did not have a control population for comparison and the influence of other factors such as patient demographics was not measured; however, since the system has been stable for several years with only common cause variation at baseline and reduced readmissions were associated with higher number of bundle components received (dose effect), we are confident that the change in outcome was attributable only to COPD care bundle adherence. There were no other improvement initiatives at our hospital nor any major changes in COPD management recommendations during this time that would influence the results of this project. Second, although we showed efficacy for the bundle as a whole and a dose response based on number of components received by individual patients, our study design did not allow us to assess the influence of each individual component of the COPD care bundle on COPD readmissions. Third, all bundle components are evidence based and selected based on local system failures which will vary in each healthcare setting. It would be naive to believe that this bundle will produce similar reductions in COPD readmissions in other facilities without analysis, identification and rectification of system failures for each unique healthcare system. Lastly, the long-term effects of the bundle were not evaluated.
Since improvement is an ongoing endeavour, our next steps will include improving the methodology for inhaler education and discharge instructions for patients to enhance retention, standardisation of COPD management during follow-up outpatient visits, and, additional personalised interventions for high-risk patient subgroups. Another challenge for all improvement work is sustainability of the processes that led to better outcomes. To enhance durable bundle utilisation and adherence, we have developed a four-member multidisciplinary team (a pulmonologist with improvement science expertise, a hospitalist, a pharmacist and an RT) that monitors monthly 30-day COPD readmission rates (provided by the data support team) and measures bundle component adherence (sample 10 patients each month). The team has triggers for action if special cause is observed in adherence to overall bundle or individual bundle components. The pulmonologist in the team is also responsible to suggest modifications to the COPD care bundle based on any new literature/guidelines.
Conclusion
System-level failures and unmet patient needs are modifiable risks for COPD readmissions. COPD care bundle design based on knowledge of local system-level failures followed by implementation and system redesign for reliable bundle delivery using improvement science reduces 30-day COPD readmissions.
Acknowledgments
We acknowledge the active participation of following team members in this improvement project: Rich Warman, RRT; Brave Nyugen, MD; Eric Warm, MD; Justin Held, MD; Arshia Ali, MD; Rajan Parkash, MD; Marshall Ashby, MBA; Brittany Woolf, PharmD, Eric Mueller, PharmD; Susie Foltz, RN; Greg Yosmali, MS; Nancy Gray, SW; Melissa Campell, RN; Myka Bowen, RN.
References
Footnotes
Twitter ACOPD care bundle designed to mitigate system-level failures & implemented through Model for improvement reduced COPD readmissions by 35%.
Competing interests The authors have no conflict of interests.
Ethics approval University of Cincinnati Institutional Review Board.
Provenance and peer review Not commissioned; externally peer reviewed.