MOdifiable facTors to Improve VentilAtion ThErapy in ICU

Introduction:

In low and middle-income countries (LMICs), up to two-thirds of ICU admissions necessitate invasive mechanical ventilation (IMV), compared to half of ICU admissions globally1-3. Though IMV is a life saving organ support technique, it's also linked to complications like ventilator associated lung injury and requires complex processes of care4. Despite being younger, ventilated patients in LMICs have 2-4 times higher mortality than high-income countries (HICs)5-8. Patients with acute respiratory distress syndrome (ARDS) in LMICs face an even greater risk, 70% higher than in HICs9.

While HICs have ample data on modifiable factors and improvement strategies that curtail ICU complications, LMICs may face preventable complications associated with IMV1,10-13. Complications like ventilator-induced lung injury and related pneumonia lead not only to excess deaths but also longer ICU stays and escalating costs14. Unfortunately, data on determinants of poor outcome among ventilated patients in LMICs remains scarce 12,15-17. Directly applying HIC evidence based management strategies to LMICs isn't always effective, as seen in several African trials 18,19. LMICs grapple with challenges like understaffing, limited training, insufficient infrastructure, and different patient characteristics and comorbidities, all potentially compromising the quality of ICU care15,16,20,21.

Recently, the use of critical care registries for data collection, including patient care, indicators, and outcomes, has expanded in both HICs and LMICs22-24. This streamlines research and offers cost benefits. Uganda has launched the Intensive Care Registry of Uganda (ICRU) for quality improvement and research infrastructure. This registry-embedded study aims to identify modifiable factors impacting outcomes for mechanically ventilated patients in LMICs leveraging ICRU's data pipeline. We hypothesize that specific patient-level and organizational factors can be identified that contribute to the ICU mortality of ventilated patients in Uganda.

Methods and Analysis:

Study design:

The 'Potentially MOdifiable factors To ImproVe outcomes of mechanically Ventilated patients in ICUs in a Low-income Country' (MOTIVATE-ICU) study is a prospective multicentre registry-embedded observational study with two main components. The first will be a prospective observational multicenter registry-embedded study aimed at assessing patient-related and process-related factors. The second component will be a cross-sectional survey to assess the organizational structure of the ICUs included. This study protocol was registered at Clinicaltrials.gov (registration number: NCTXXXXXXX). The results of the study will be reported according to the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) Statement25.

Study setting:

ICUs will be defined as geographically designated area/unit within a hospital that routinely provides invasive mechanical ventilation therapy with continuous vital sign monitoring [electrocardiographic (ECG) monitoring, heart/pulse rate, non-invasive blood pressure (NIBP), peripheral oxygen saturation (SpO2)] and designated nursing care for each bed in the unit i.e at least three patients per week for at least 24 hours. A descriptive study conducted in Uganda assessing the ICU capacity before the Corona virus induced disease 2019 (COVID-19) revealed that at the time, there were 14 ICUs with 12 being functional20. However, post-COVID-19, anecdotal evidence shows nearly a doubling of the ICU capacity with nearly 25 functional ICUs to-date. All ICUs currently functional in Uganda will be eligible to participate (eTable 1).

Participants:

Patients aged ≥ 15 years admitted to study ICUs during the study period and receiving invasive mechanical ventilation will be eligible for recruitment. Exclusion criteria include successful extubations within 24 hours of intubation, admissions for end-of-life care and/or ICU palliative support as well as patients transferred from ICUs not participating in the study 24 hours after initiation of MV.

Outcomes:

Primary outcomes The primary outcomes will be ICU mortality, ICU length of stay (LOS) and duration of mechanical ventilation. Outcome definitions are detailed in Table 1.

Secondary outcomes These will include duration of non-pulmonary organ support, ICU-free days, ventilator-free days at day 28, ventilator-associated pneumonia (VAP), tracheobronchitis, non-infectious pulmonary complication (clinician suspicion or radiological diagnosis of pleural effusion, atelectasis or pneumothorax), readmission, unplanned extubations. Tracheostomy related outcomes will be explored as part of a preplanned substudy, including tracheostomy timing, decannulation rate, decannulation failure and complications such as stoma infection, major bleeding, tube dislodgement and malfunction.

Associated factors (Table 2):

Factors potentially associated with outcomes among mechanically ventilated patients will be analyzed as 'non-modifiable' factors and 'potentially modifiable' factors. Non-modifiable factors include patient-related factors such as age, illness severity at initiation of mechanical ventilation, comorbidities, medical vs surgical admission, and indication for IMV (Table 2). Potentially modifiable factors will subcategorized into care process-related factors such as Richmond Agitation Sedation Scale (RASS) targets, initial ventilator settings on day 0, use of daily spontaneous awakening and spontaneous breathing trials, use of stress ulcer and deep vein thrombosis (DVT) prophylaxis. The second subcategory will be ICU organizational structure such ICU staffing, staff-patient ratios, empowerment of non-physician staff, multidisciplinary team rounds and use of protocols and checklists (Details on associated factors in supplementary material Table 2).

Data collection:

The following procedures will be performed to ensure the quality of the data collection Patient-level data will be entered directly via laptop or tablet through the cloud-based ICU registry platform (PROTECT cloud-based platform) of the ICRU currently in use (Figure 1).

Patient level variables of interest including age, sex, diagnostic categories, functional capacity before hospital admission (Clinical Frailty Score), the Charlson Comorbidity Index, the use of IMV on Day 1, modified Sequential Organ Function Assessment (mSOFA) and Tropical Intensive Care Score (e-TropICS) will be scored. ICU level variables of interest included ICU type (medical-surgical vs. specialty), presence of training programs in critical care, ICU admission volumes in the preceding year, staffing patterns, organizational and process characteristics; a full list of potential covariates is detailed in Table 2.

We shall extract prospectively collected data from the ICRU platform, which has both core data variables collected at admission and daily variables collected during the ICU stay.

A complete list of variables is detailed in Supplementary materials. The platform uses a common data model and provides real-time data on casemix, management features and outcomes. MOTIVATE-ICU study variables were incorporated into the PROTECT platform in order to design a registry embedded study (Figure 1). The following site-level procedures will be followed:

To ensure standardization of the study procedures, each site shall have a site lead and dedicated data entrant(s) responsible for data collection during use of the ICU registry platform. Prior to the beginning of recruitment, each site lead and data entrant(s) will receive training.

Each site will have a site lead responsible for weekly validation of patient episodes in terms of completion and data quality.

Each site will have IT support as well as participate in monthly meetings as a means of dispelling doubts and solving potential challenges related to data entry.

For each site, data on hospital and ICU organizational and care process characteristics will be collected through an interviewer-administered questionnaire from the ICU director and/or the in-charge nurse.

Follow-up:

Given the potential burden of data collection involved, the patient recruitment period in each participating ICU will be a minimum of three months up to a maximum of 6 months, with a 28-day maximum follow-up period for each patient recruited. The follow up period for individual patients will end at ICU discharge.

Procedures to ensure data quality

Sample size:

We aim to include all consecutive patients admitted to participating ICUs in order to reduce potential selection bias and minimise temporal variation of care processes. Based on aggregated data (published and unpublished) from LMICs including Uganda's ICU cohorts2,10,12,26-31 the lowest measured mortality among patients receiving invasive mechanical ventilation was approximately 40%. A total of 250 events (deaths) would allow us to evaluate at least 25 associated variables in multivariable models. We will therefore enroll at least 625 patients receiving invasive mechanical ventilation that fulfill the eligibility criteria.

Handling of missing data:

During the study period, all data pertinent to the study objectives will be mandated for all study ICUs. The study platform will flag missing data to the data collector in case of incompleteness. Each site lead will ensure validation of patients entered and completeness of the data weekly to ensure no missing data. This is being a prospective registry-embedded study, we expect a low percentage of missing data pertinent to the primary objectives of the study.

Statistical analysis:

We will describe ICU and patient characteristics using standard descriptive statistics and reported continuous variables as mean ± standard deviation or median (25%-75% interquartile range, IQR), as appropriate. We will test the univariate association between eligible variables and outcomes using analysis of variance, Kruskal-Wallis, Student t test, Mann-Whitney test, chi-square test, and Fisher exact test, as appropriate.

Using multilevel multivariable logistic regression models, we shall investigate the association between organizational factors and the primary outcome adjusting for patients' characteristics. A two-level model will be fit with patient-level fixed effects at the first level and ICU-level fixed effects at the second level, as well as an ICU-specific random effect. We will pre-specify several models to represent the patients and the unit organization dimensions. We shall then insert variables into the models if they are associated with outcome with a P value of <0.20 on univariate analysis. We shall not consider variables used for e-TropICS score computation. We shall apply several variables into the final model regardless of their statistical significance due to their clinical significance, including the presence of 24/7 intensivist, regular multidisciplinary rounds and nurse/bed ratios in the final model. In addition, we shall perform subgroup analyses stratifying patients according to the type of admission (medical vs. surgical) and the e-TropICS tertiles. We shall use the Akaike information criterion, the likelihood ratios and the model residuals to choose among the alternative models. Two-tailed P-values <0.05 were considered statistically significant. All statistical analyses will be conducted in R (http://www.r-project.org) and SPSS 21 (IBM Corp., Armonk, NY). A detailed statistical analysis plan will be made available online before completion of the study and locking of the database.

Ethics and dissemination:

Ethical approval and consent to participate This study will be conducted according to the principles of the Declaration of Helsinki (revision Fortaleza, Brazil, October 2013). Approval to carry out this study will be sought from the relevant ethical review and scientific review committees and from hospital administrations.

We shall obtain deferred consent within 72 hours of recruitment into the study from the legal representative since most ICU admissions are emergency admissions and informed consent during the initial 48 hours of admission may not be possible. In addition, without exception, patients admitted to ICU for ventilator support are unable to give informed consent. Persons who may take the role of legal representative in accordance with the Medical Treatment Agreement Act (WGBO) are: a predefined representative, husband or wife, registered partner or other life partner, a parent or child, brother or sister, and incidentally a curator appointed by a judge. For patients with no legal representative and are unable to give informed consent during in-hospital stay, we shall obtain a waiver of consent. The experience of ICU patients enrolled under deferred consent is mainly positive, as shown in the NICE-SUGAR trial, in which participants were included using deferred consent32. A majority of the patients were happy with the decision made by the representative (93%) and would have granted consent if asked (96%)33.

Oral consent from each of the ICU directors and/or in-charge nurse will be sought to conduct the interviewer-administered survey.

Dissemination:

The study results will be submitted for publication regardless of the results after completion and analysis. We shall make the study findings widely available including dissemination through conferences and peer-reviewed journals.

Data sharing:

The authors encourage interested parties to contact the corresponding author with data sharing requests, including for access to additional unpublished data. ICRU is part of a larger network which has appointed a Data Access Committee (DAC). Curated data may be shared with third parties for research purposes following written approval from the DAC.

Discussion and study status:

The MOTIVATE-ICU registry-embedded study aims to identify potentially modifiable factors associated with outcomes among invasively ventilated patients in a low income country.

According to the Donabedian model improving the quality of healthcare and improving outcomes of interest, rests on three components; the structure, care processes and outcomes.34 Various studies from HICs on ICU organization structure, care-bundles and protocols have shown significant impact in reduction of mortality, duration of ventilation and ICU length of stay; albeit with some inconsistencies in some aspects.6,11,35-37 Based on what we currently know from the ORCHESTRA study done in Brazil, an upper middle income country (UMIC), organizational factors including protocol implementation are the key targets to improve ICU outcomes and efficiency38,39. There has been considerable data describing the physical structural set-up and staffing of ICUs in LMICs20,26,40,41, but limited data is available on organizational features and care-processes, and their impact on outcomes in lower income settings like Uganda.

Direct implementation of evidence-based care processes from higher income settings may not necessarily yield similar results when implemented in resource-poor settings. For example, two African studies that implemented fluid therapy as per protocols used in HICs revealed more harm than benefit in both critically ill adults and paediatric populations18,19. For other processes, implementation in LMICs has improved consistently but outcomes remain lagging when compared to HICs, probably affected by a multitude of factors. For instance, use of protective mechanical ventilation in patients without ARDS was shown to be comparable between MICs and HICS during first days of ventilation, but with a large difference in ICU mortality14. The same was seen for ventilation in ARDS patients, albeit with a small difference in protective ventilation implementation1,6.

One of the key potential interventions likely to show impact in low-resource settings compared to HICs is tracheostomy placement and timing. Tracheostomy timing, while improving ICU length of stay and duration of mechanical ventilation has not had an impact on mortality in HICs42-44, but may have a significant impact in a setting with limited resources. A preplanned substudy of MOTIVATE-ICU will explore timing, complications and outcomes in patients undergoing tracheostomy in a low income country like Uganda, complementing data from LMICs from other African countries45,46.

The strengths of the current study are its prospective design, registry-embedded nature, and multicenter inclusion. There are several potential limitations of this study. For example, while efforts are made to ensure data quality and reduce missing data, the reliance on electronic data collection and the prospective nature of the study may still encounter challenges with data completeness and accuracy across multiple centers. The study's findings, especially those related to organizational structure and process-related factors, may be specific to the context of low-income countries and thus may not be generalizable to settings with different healthcare systems, resources, and ICU capabilities. The cross-sectional survey assessing the organizational structure of ICUs, rely on self-reported data, which can introduce biases and inaccuracies.

Acknowledgements:

We thank all the site leads and data entrants of each site at each of the participating hospitals as well as Critical Care Asia Africa network for the support and involvement in the success of this study.

Funding:

The study will be supported by the Wellcome award, Collaboration for Research, Implementation and Training in Critical Care in Asia and Africa (Grant number: 224048/Z/21/Z).