In the original submission, we had planned to compare the effectiveness of the OLE and IPE for increasing the delivery of NIV in appropriate patients hospitalized with COPD exacerbation. Due to the changes in the study to a stepped-wedge design, we have revised this aim to compare the effectiveness of IPE to standard care before the implementation of the IPE strategy. Our primary outcome remain the hospital-level risk-standardized (RS) NIV proportion however we will assess the change in rate from the period prior to training (not compared with the active control, OLE). Similar changes will apply to the secondary outcome measures of RS-hospital rates of NIV failure (invasive mechanical ventilation (IMV) after a trial of NIV), mortality, length of stay, and 30-day readmission among all patients with COPD. We did not change the time periods for comparison. We have planned three 18-month periods of analysis. (1) Baseline—18 months prior to the start; (2) immediate/short-term impact—18 months after start; and (3) sustainability—18 months after period 2. We have revised the original analysis by developing four risk-standardization models for the baseline 18 months, then moving forward 6 months and modeling an 18-month period. We will remove the four months after the IPE training for each step (cohort) from the analysis to allow for the completion of the educational sessions. This is a change from the original protocol based on the learning from the first cohort; it took on average 4 months for the champions to get the other providers enrolled in the educational activity. In a sensitivity analysis, we will repeat these analyses, using baseline data from the period prior to March 2020. We will compare RS-NIV rates from the pre-COVID-19 period to the “baseline” prior to implementation of IPE to gain some understanding of the changes in ventilation practice with COPD patients related to the COVID era.
Patient and hospital information revisions
We will add the 7-day average of hospitalized patients with COVID-19 and the 7-day average of bed occupancy (adult inpatient and intensive care units) alongside the staffing: number of RTs, hospitalists, and emergency room physicians and nurses, to gain some understanding of the impact of COVID-19 pandemic. These factors will be used to describe participant hospitals.
Statistical analysis Aim 1 revisions
With the new stepped-wedge study design, we will generate descriptive statistics overall, by hospital, and pre-and post-implementation, including counts and percentages for categorical data and means, standard deviations, and percentile distributions for continuous data. We will compare characteristics of hospitals started at each step, including size, ownership, teaching status, location, baseline NIV proportion, staffing of RTs, nurses, hospitalists, intensivists, and emergency room physicians via chi-square tests, and analysis of variance or Kruskal-Wallis tests.
Characteristics of eligible COPD patients derived from de-identified administrative and billing data of the enrolled hospitals will initially be compared via GEE models accounting for the pre- and post-implementation periods. Then, for each hospital, for each period, we will calculate the percentage of patients treated according to each of the primary ventilatory strategies: no assisted ventilation, NIV, and IMV. We will then calculate the percentage of patients initially treated with NIV among those who received assisted ventilation.
Power and sample size for Aim 1 revisions
Originally, we had calculated that a total sample of 20 hospitals, 10 in each arm, will give 80% power to detect a difference of 15% in change (e.g., 5% increase among OLE hospitals, vs. 20% increase among IPE hospitals). For the stepped-wedge design, using a type I error rate of 0.05 and standard deviation of change in rates over time derived from our prior work with the Premier database, a total sample of 20 hospitals will give 80% power to detect a difference of 15% in change from baseline RS-NIV rates in hospitals before to the implementation to RS-NIV rates after the IPE education. However, to account for potential loss in recruited hospital sites, we aim to recruit up to 30 hospitals to achieve 80% power.
Aim 2 analysis revision
As in our original submission, we plan to only examine the effect of the IPE education on RT autonomy and team functionality as potential mediators of NIV uptake.
Statistical analysis revision
We will develop a series of models evaluating associations; however, instead of this being among the intervention (OLE and IPE), we will develop this for pre- and post-implementation and the mediators and outcome, including a structural equation model (SEM) to estimate the role of mediators as well as the direct effect of IPE education on the outcome.
Power and sample size for Aim 2 revisions
The original study design power analysis for this aim accounted for the clustering of the 20 hospitals into their respective IPE and OLE arms. The revised design will have 20 hospitals in the IPE education. We calculate that to achieve 80% power and using a type I error rate of 0.05, a sample of 10 RT’s per hospital will allow us to detect a moderate (Cohen’s d=.4) difference at 1-year post-intervention. We will use an intraclass correlation (ICC) in the range of .10–.20.
Aim 3 revisions
We plan to evaluate the strategies used and the barriers the COPD-NIV Champion teams faced to implement the NIV-IPE courses, to refine the implementation strategies further.
We will perform semi-structured interviews with the COPD-NIV Champions to assess the strategies they used and the barriers they faced when implementing the NIV-IPE courses, including the impact of the COVID-19 pandemic. We expect to enroll 2–3 providers from each hospital for a total of 14–21 Champions, enabling us to reach thematic saturation.
This qualitative aim will allow us to refine the intervention and the Champion’s role for each profession for future implementation strategies using an interprofessional team approach.
Current status of the study
We have completed the training for the first cohort of seven hospitals and thirty champions. We are interviewing the champions to understand barriers and facilitators for engaging clinicians in the educational intervention. We have continued recruiting hospitals, and we have a pool of at least 23 hospitals interested to join the learning collaborative in 2022.