Children requiring long-term mechanical ventilation are at high risk of mortality. Setting ventilator alarms may improve safety, but best practices for setting ventilator alarms have not been established. Our objective was to increase the mean proportion of critical ventilator alarms set for those children requiring chronic mechanical ventilation followed in our pulmonary clinic from 63% to >90%.
Using the Institute for Healthcare Improvement Model for Improvement, we developed, tested, and implemented a series of interventions using Plan-Do-Study-Act cycles. We followed our progress using statistical process control methods. Our primary interventions were: (1) standardization of the clinic workflow, (2) development of an algorithm to guide physicians in selecting and setting ventilator alarms, (3) updating that algorithm based on review of failures and inpatient testing, and (4) enhancing staff engagement to change the culture surrounding ventilator alarms.
We collected baseline data from May 1 to July 13, 2017 on 130 consecutive patients seen in the pulmonary medicine clinic. We found that 63% of critical ventilator alarms were set. Observation of the process, standardization of workflow, and adaptation of an alarm algorithm led to an increase to 85.7% of critical alarms set. Through revising our algorithm to include an apnea alarm, and maximizing provider engagement, more than 95% of critical ventilator alarms were set, exceeding our goal. We sustained this improvement through January 2021.
Our stepwise approach, including process standardization, staff engagement, and integration of an alarm algorithm, improved the use of ventilator alarms in chronically ventilated pediatric patients.
Children requiring long-term, invasive mechanical ventilation in the home setting represent a growing population with variable underlying pathologies.1–4 Their complexity, reliance on technology, and fragility place them at a high risk of mortality.5 Many deaths in invasively ventilated patients are due to preventable causes, such as tracheostomy plugging or accidental decannulation.5–7 The American Thoracic Society8 produced guidelines recommending the use of pulse oximetry, in addition to internal ventilator alarms or other monitoring devices, to identify emergencies. However, evidence supporting this recommendation was limited.9
Home mechanical ventilators contain multiple internal alarms. Some alarms need to be manually set by a provider to sound when a given respiratory parameter falls outside of the set range.10 Minimal evidence exists regarding the effectiveness of ventilator alarms and how they should best be used. Two small studies suggesting concerns about ventilator alarms prompted an American Thoracic Society recommendation for pulse oximetry in ventilated patients.11,12 Ventilator manufacturers recommend using multiple manual alarms to detect serious events but do not specify how alarms should be set; differences in patient characteristics, tracheostomy tube size, and ventilation goals all impact the optimal alarm settings.13 No existing literature describes methods to determine ventilator alarm settings, or how alarms are set in a clinical setting.
Following a series of critical events at home, a multidisciplinary team was convened with the global aim of improving the safety of our chronic ventilator-dependent population. Although inappropriate use or malfunction of ventilator alarms was not implicated in any events, the use of appropriate alarms emerged as a clear opportunity to improve safety. Noting inconsistencies in our approach to setting home ventilator alarms, we developed this improvement project to increase the rate of ventilator alarms set during pulmonary clinic visits.
Methods
Context
We performed this improvement project at Cincinnati Children’s Hospital Medical Center in the pulmonary medicine clinic. Nine pediatric pulmonologists and dedicated nursing staff manage our >300 patients requiring chronic invasive ventilation. Most patients in our practice use a Philips Trilogy 100 ventilator (Philips Respironics, Murrysville, PA), with a smaller proportion using the LTV 1150 ventilator (Carefusion, Yorba Linda, CA). Nearly all our patients use a pressure mode of ventilation, most commonly, pressure control, synchronized intermittent mandatory ventilation. Dedicated respiratory therapists (RTs) evaluate ventilator-dependent patients at each clinic visit, take detailed histories, and obtain vital signs and ventilator data. Specifically, they obtain a Philips NM3 (Philips Respironics, Murrysville, PA) respiratory profile, including tidal volumes, and end-tidal carbon dioxide readings. Patients using the Trilogy ventilator also have an internal data card that may be accessed. Prior improvement work generated reliable processes for documentation of ventilator settings, including alarms, in both inpatient and outpatient settings.
During baseline data collection, between May and July 2017, we documented the presence or absence of alarm settings for 130 consecutive patients in our outpatient practice. We found that 63% of critical ventilator alarms had settings established.
We established an improvement team, including RTs, nurses, medical assistants, and physicians, that care for our chronic invasive ventilator-dependent population. Though not formally represented on our quality improvement team, families provided continuous insight into the process and concerns related to alarms during clinic visits and inpatient stays. The team reviewed ventilator alarms used in inpatient and outpatient settings and the process for collecting and evaluating alarm information in the pulmonary clinic. Barriers to maximizing our use of ventilator alarms included (1) lack of a standardized process for collecting alarm data in the clinic and for discussing it with the primary provider, (2) lack of standardized guidelines for setting alarms, (3) propensity for nuisance alarms in certain patient groups, and (4) low priority placed on setting alarms during clinic visits. This assessment provided the basis for our theory for improvement. We theorized that iterative interventions directed at standardizing clinic workflow process, developing and implementing an algorithm for setting alarms, and engaging team members in discussions about alarms would facilitate improvement (Fig 1). Our Key Driver Diagram includes our specific, measurable, actionable, realistic, time-bound (SMART) Aim14 ; we sought to improve the rate of critical alarms set at departure from the pulmonary clinic for long-term, mechanically ventilated patients from 63% to >90% within 6 months.
Interventions
Standardizing Clinic Workflow
We prioritized standardizing clinic workflow to ensure that RTs could efficiently obtain ventilator data; data availability was key for conversations with physicians about alarm settings. We explored optimal data extraction methods for each ventilator type. For patients requiring Trilogy ventilators, we compared the values obtained from the respiratory profile to the data downloaded from the ventilator data card. After multiple Plan-Do-Study-Act (PDSA) cycles, we routinely downloaded ventilator data upon arrival to clinic, rather than later in the visit, to ensure data availability for RT and physician discussions. We also provided the alarm algorithm, described below, in the clinic for reference when setting alarm parameters. RTs documented any reasons why alarms were not set, along with the final ventilator alarm settings in the electronic health record.
Alarm Algorithm
First Iteration
We developed an algorithm for setting ventilator alarms that we tested in the outpatient setting and in our inpatient unit. As there was no evidence to guide the creation of this document, we based the algorithm on consensus recommendations from our physicians that manage patients requiring chronic mechanical ventilation and RTs on our improvement team. The algorithm offered guidance for alarm parameters based on data obtained during the clinic visit. It was intended to function as a starting point from which providers could customize alarms to a specific child’s disease and ventilatory goals. We attempted to set alarms as tightly as possible but also recognized that an alarm set at even the least sensitive level was more likely to detect a serious event than a disabled alarm.
We defined critical alarms based on ad hoc simulations on the inpatient unit; simulations were performed by ventilating a test lung through various size tracheostomy tubes, plugging or decannulating the tracheostomy tube, and measuring the time to alarm. Our inpatient trials demonstrated that in the LTV ventilator, the activation of the low minute ventilation alarm, in addition to default alarms that cannot be disabled, were sensitive to the following: accidental disconnect events, obstructive mucous plugs, or accidental decannulation events. The Trilogy ventilator was reasonably sensitive to similar events, without generating an untenable number of nuisance alarms, when using a combination of alarms: the circuit disconnect alarm, the low minute ventilation alarm, and the low tidal volume alarm. After implementation of the algorithm in the clinic, we performed follow-up phone calls with families for 1 week to assure there was no substantial increase in nuisance alarms. Following this test, discussion regarding alarm changes and mitigation in the event of increased nuisance alarms was left to individual provider discretion. We encouraged families to call if they noted more nuisance alarms.
Second iteration
Learnings from our initial efforts to improve ventilator alarm use informed the second iteration of our alarm algorithm. When alarms were not set, we reviewed charts to determine why. The most common reason for failure was a high rate of nuisance alarms when using the low exhaled tidal volume (Vte) alarm in children with low average tidal volumes. (The Trilogy ventilator allows the low Vte alarm to be set at a minimum level of 40 mL; for infants and small children, this leads to frequent nuisance alarms.) The second most common reason for failure was large leaks occurring in children with uncuffed tracheostomy tubes; in these cases, variable tidal volumes due to large leaks led to frequent nuisance alarms. These 2 reasons accounted for 72.9% of failures during implementation of the first algorithm. Low tidal volumes were generally problematic in infants, whereas large leaks generally occurred in older and more stable patients.
We began trialing an additional Trilogy alarm, the apnea alarm, in our inpatient unit to address the most common reason for failure (low average tidal volumes, and inability to tolerate the low tidal volume alarm on the Trilogy ventilator without excessive nuisance alarms). The apnea alarm is designed to sound if a patient’s spontaneous respiratory rate drops below a set threshold. Children that require high mandatory breath rates, and those that have impaired respiratory drive or inadequate respiratory muscle strength to trigger the ventilator, often generate excessive nuisance apnea alarms. However, in other populations, especially the majority of infants with primarily parenchymal lung disease, the apnea alarm was a reasonable substitute to the low Vte alarm based on our testing, with similar sensitivity to potentially life-threatening events. The use of the apnea alarm was therefore added to our alarm algorithm in the second iteration (Fig 2).
Staff Engagement & Sustainability
We were able to ensure awareness of this project among clinic staff through repeated interventions, ongoing discussions, and huddles with key team members. We used our weekly previsit planning meeting to discuss alarm settings for each patient. To increase reliability, we incorporated alarm settings into an automated electronic template with growth parameters, recent visits and admissions, and other relevant information discussed at previsit planning meetings. We also included alarm notes in weekly previsit e-mails to providers. Additional detail about individual PDSA cycles for all interventions are in Supplemental Table 1.
Study of the Interventions
We measured the number of critical alarms ordered following each clinic visit as a percentage of the total possible number of critical alarms. If a patient used a Trilogy ventilator, we focused on 3 critical alarms to set: circuit disconnect, low minute ventilation, and low Vte. If the patient used an LTV ventilator, we focused on 1 critical alarm to set: low minute ventilation. We calculated the set number of critical alarms (numerator) and total possible number of critical alarms (denominator) for groups of 10 consecutive patients seen in clinic. The variation in denominator between groups was due to the number of patients seen on either a Trilogy or LTV ventilator. We recorded data on an annotated p-chart, a statistical process control chart showing a proportion of failures over time.15 Of note, after incorporation of the apnea alarm on the second iteration of our algorithm, we counted the apnea alarm as a substitute for the low Vte alarm only in the scenario where the patient’s low Vte alarm could not reasonably be set due to frequent nuisance alarms. The use of the apnea alarm in our algorithm did not affect the denominator of our measure.
We documented all alarm settings in the electronic health records (EHR) as part of our standard clinic workflow using existing processes. We extracted data from the EHR and manually reviewed charts for a Pareto analysis of failures.
Analysis
We used a quantitative, time-series study design and statistical process control methods for analysis. We used accepted industry criteria to designate change as random variation (common cause variation) or attributable to a specific cause (special cause variation).16 We updated the annotated p-chart monthly during the period of frequent interventions and then approximately quarterly once our goal was sustained. We calculated the mean and upper and lower control limits displayed on the p-chart.
Human Subjects Protection
The current study fell within the Cincinnati Children’s Hospital Medical Center Institutional Review Board’s guidance for quality improvement projects that did not constitute human subjects research.
Results
The study period began on May 1, 2017, with interventions trialed through April 1, 2018, when we achieved our SMART Aim goal. Data collection continued through January 2021 to demonstrate sustainability. There were 1210 clinic visits during the study period.
Baseline data were collected from May 1, 2017 to July 13, 2017 on 130 consecutive visits. Critical alarms were set, on average, 63% of the time (Fig 3A). After assembling our improvement team and beginning observation of the clinic workflow on July 19, 2017, we saw an increase in the mean percentage of critical alarms set to 80.8% based on p-chart rules for special cause (Fig 3B). After changes to clinic workflow occurred through August and September of 2017, and adoption of our alarm algorithm on August 17, 2017, we noted further improvement to 85.7% of potential alarms set (Fig 3C). We also observed decreased variation in the proportion of alarms set when comparing new data to the baseline study period. The consistent failure to set alarms on children with small tidal volumes and large leaks became more evident as the processes became better standardized. After adoption of the second iteration of the alarm algorithm (which included the apnea alarm) on November 30, 2017, and continued efforts to engage care providers through weekly previsit planning meetings, we noted an increase in mean percentage of alarms set to 95.1%, exceeding our initial goal of 90% (Fig 3D). During the time when we continued high-reliability interventions, including the alarm algorithm and alarm discussions in weekly meetings, we maintained our goal of >90% of alarms set through January 2021. We had one instance of special cause where the rate of alarms set fell outside of the lower control limits. This occurred in March 2020, and we suspect it was secondary to temporary interruption in normal clinic practice due to coronavirus disease 2019. Specifically, we rapidly adopted telehealth visits, so RTs were unable to evaluate ventilator settings. In some instances, ventilator settings may have been documented during a telehealth encounter without visual verification, though this is difficult to confirm using EHR data.
Discussion
Among our population of pediatric patients with chronic ventilator dependence, we improved the use of critical ventilator alarms in the home setting through modifying our outpatient practice and instituting a systematic approach to setting alarms in our clinic. We engaged our core clinic staff members, which allowed for rapid adoption of interventions. Additionally, we modified clinic workflow to streamline the data collection needed to inform decisions about alarm settings. We developed the first algorithm we are aware of to guide the use of ventilator alarms in children, and modified this algorithm based on new learnings to maximize safety for our patients at home. Collectively, our improvement efforts led to an increase in the proportion of critical ventilator alarms set to more than 90%. We have maintained this improvement for over 3 years at the time of this report.
Staff awareness and engagement were crucial to the success of this work in a variety of ways. The RTs contributed to the practical aspects of this work by collecting information necessary to inform discussions with physicians on alarm settings. Staff engagement was key in modifying the culture around use of ventilator alarms in our practice. This change was noted even during our observation period (Fig 3 A and B). Without the engagement of our dedicated staff, our improvement efforts and culture shift would have been unlikely to succeed.
Designing our ventilator alarm algorithm was a challenging task given the lack of existing literature to guide the use of home ventilator alarms. We took a pragmatic approach and performed frequent bedside testing in our inpatient unit. Our learnings from these bedside trials informed our algorithm. Our focus on specific critical alarms does not discount the utility of other user-set or default alarms in certain scenarios. Supportive hospital leaders and unit-based staff were essential to our ongoing efforts. Without them, it would not have been possible to perform the bedside testing needed to ensure the safety of our approach in an expedited fashion.
Throughout the course of this work, our focus was improving patient safety. Outcomes, such as significant morbidity, or mortality, related to a failure in ventilator alarms are, thankfully, rare. To our knowledge, there have been no home deaths in our population related to inappropriate alarm function since the implementation of our interventions. There have been critical events resulting in significant morbidity in this time. However, interrogation of ventilator alarms has not revealed inappropriate alarm function or failures. We suspect that if adverse events within our patient population occurred, we would identify these, as ours is the only children’s hospital in the area. This series of process interventions was part of a larger initiative aimed at improving the safety of our patients as quickly and effectively as possible. We did not design a mechanism for testing which of these initiatives was most impactful in improving patient safety outcomes.
There are several limitations to this work. First, although our algorithm represents our best practice currently concerning ventilator alarms in the home setting, it is difficult to determine how well these alarms capture events that require immediate intervention. It is also difficult to quantify the frequency of nuisance alarms after implementing settings according to our algorithm. Anecdotally, families report that the alarm settings are not cumbersome with excessive, unactionable alarms. However, this does not exclude the possibility of increased nuisance alarms after our change in practice. Finally, although the processes and interventions we outlined were appropriate for our practice, they may not be ideal in all settings.
We care for a unique and fragile population at high risk of mortality that may be mitigated through improved use of internal ventilator alarms. Therefore, we will continually refine our approach to ventilator alarms and maximize safety in the hospital, at home, and in other community settings. Ongoing ventilator bench testing may better identify deficiencies in alarms and us to test alarm combinations through simulation. New knowledge and innovative strategies will be needed to solidify best practices in ventilator alarm use, test and incorporate new ventilators into our approach, and improve safety for all ventilator-dependent children. Finally, we wish to emphasize that setting ventilator alarms does not replace the need for additional forms of monitoring and the presence of an alert, well-trained caregiver. Children with invasive chronic ventilation are always at high risk and our hope is that a better process for ventilator alarm use can reduce that risk.
This report uses the Revised Standards for Quality Improvement Reporting Excellence (SQUIRE 2.0) guidelines for reporting quality improvement research in healthcare.17
ACKNOWLEDGEMENTS
We thank Alicia West, RT; Beth Koch, RT; Amanda Waits, RT; Shannon Johnson, RN; Denise Leonard, RN; Alyssa Mohr, RN; Janice Gramke, RN; Molly Stratman, RN; and Yuping Guo.
Dr Pajor conceptualized and designed the study, led the improvement team, conducted and interpreted the data analyses, drafted the initial manuscript, reviewed and revised the manuscript; Ms Kaiser and Ms Brinker were part of the improvement team, conceptualized, designed, and executed the study interventions, interpreted the data, and reviewed and revised the manuscript; Ms Mullen was part of the improvement team, conceptualized and designed study interventions and assisted with data analysis and data interpretation; Drs Hart, Britto, Torres-Silva, Hysinger, and Amin assisted with conceptualization of the study, provided feedback on intervention design, interpreted the data, assisted with implementation, and critically reviewed the manuscript; Dr Schuler helped interpret the data and critically reviewed and revised the manuscript; Dr Benscoter conceptualized and designed the study, provided guidance to the improvement team, oversaw data analyses, helped interpret the data and critically reviewed the manuscript; and all authors approved the final manuscript as submitted and agree to be accountable for all aspects of the work.
FUNDING: The pediatric study was supported, in part, by the National Center for Advancing Translational Sciences of the National Institutes of Health, under Award Number 2UL1TR001425-05A1. The content is solely the responsibility of the authors and does not necessarily represent the official views of the NIH.
CONFLICT OF INTEREST DISCLOSURES: Dr Britto is a member of the American Board of Pediatrics Foundation Board. The other authors have indicated they have no financial relationships relevant to this article to disclose.
Comments