Mapping the process of ICU care delivery to improve treatment decisions in acute respiratory failure

Abstract

Evidence suggests system-level norms and care processes influence individual patients’ medical decisions, including end-of-life decisions for patients with critical illnesses like acute respiratory failure. Yet, little is known about how these processes unfold over the course of a patient’s critical illness in the intensive care unit (ICU). Our objective was to map current-state ICU care delivery processes for patients with acute respiratory failure and to identify opportunities to improve the process. We conducted a process mapping study at two academic medical centers, using focus groups and semi-structured interviews. The 70 participants represented 17 distinct roles in ICU care, including interprofessional medical ICU and palliative care clinicians, surrogate decision makers, and patient survivors. Participants refined and endorsed a process map of current-state care delivery for all patients admitted to the ICU with acute respiratory failure requiring mechanical ventilation. The process contains four critical periods for active deliberation about the use of life-sustaining treatments. However, active deliberation steps are inconsistently performed and frequently disrupted, leading to prolongation of life-sustaining treatment by default, without consideration of patients’ individual goals and priorities. Interventions to standardize active deliberation in the ICU may improve treatment decisions for ICU patients with acute respiratory failure.

Keywords:

• Process mapping
• communication
• critical care
• end-of-life care
• shared decision making
• qualitative research

1. Introduction

Over 700,000 adults in the United States receive mechanical ventilation each year for acute respiratory failure. Unfortunately, about 35% die in the hospital (Mehta, Syeda, Wiener, et al., 2015), while another 10% survive but remain dependent on prolonged mechanical ventilation (Mehta, Syeda, Bajpayee, et al., 2015). These patients undergo lengthy hospitalization and post-acute care and approximately 60% die within a year (Carson et al., 1999; Damuth et al., 2015; Unroe et al., 2010). Yet, stopping mechanical ventilation for patients who are not recovering usually leads to end of life. These consequential decisions about prolonged life support or end-of-life care can be burdensome for patients’ surrogate decision makers and families (Wendler & Rid, 2011) and contribute to burnout in intensive care unit (ICU) clinicians (Moss et al., 2016).

Efforts to improve decision making about mechanical ventilation have focused on clinician communication during discrete events, such as formal family-clinician meetings (Bibas et al., 2019; Scheunemann et al., 2011). However, the process that culminates in a decision to stop or prolong mechanical ventilation encompasses a series of interactions throughout an ICU stay among patients, families, and a large interprofessional care team (Bruce et al., 2013; Kruser et al., 2019; Kruser et al., 2017; Needle et al., 2021). Evidence suggests these consequential decisions are strongly influenced by norms of care delivery and embedded care processes (Barnato et al., 2012; Drought & Koenig, 2002; Dzeng et al., 2018; Lynn et al., 2000). Yet, little is known about these processes and how they unfold over time to influence treatment decisions, within the complex ICU system and interprofessional team (Michalsen et al., 2019). Thus, our objective was to map the longitudinal, current-state ICU care delivery process for patients with acute respiratory failure and identify opportunities to improve this process and its outcomes.

2. Materials and methods

2.1. Study design

We used the systems engineering method of flowchart process mapping combined with data collection through focus groups and semi-structured interviews and directed qualitative content analysis. Process mapping is a tool from industrial engineering that illustrates the steps of a process, their temporal sequence, and the pathways that lead to outcomes (Jun et al., 2009; Trebble et al., 2010). We used this tool to build a detailed, visual model of how care is currently delivered to patients in the ICU. This method has been recently translated to healthcare environments to investigate and improve care delivery processes and their outcomes (Antonacci et al., 2021; NHS England and NHS Improvement, n.d.). This study adhered to published process mapping quality criteria, including scoping the process, educating participants about the method, ensuring stakeholder representation, gathering data from multiple sources, iteratively analyzing and refining the model, using charting software, and confirming model accuracy with stakeholders (Antonacci et al., 2021). The Northwestern University Institutional Review Board (IRB) reviewed and approved the study. Participants provided written informed consent and received a gift card incentive. Study procedures followed institutional IRB requirements and were in accordance with the Helsinki Declaration of 1975. The online supplement includes additional methodological details including research team, participant sampling and recruitment, data collection, and process map validation; this report also adheres to COREQ qualitative reporting guidelines (Tong et al., 2007).

2.2. Setting and study participants

The study was conducted at two academic medical centers in the Midwestern United States, focusing on the adult medical ICUs within each hospital. All members (clinicians) of the interprofessional medical ICU and palliative care teams (advance practice providers, chaplains, nurses, occupational, physical, and respiratory therapists, physicians, and social workers) from the study sites were eligible and invited to participate. All clinicians who responded to the invitation were included, except those unable to attend due to scheduling conflicts.

Patients and surrogate participants were recruited from an ongoing research study of mechanically ventilated patients at one of the study sites (details in Supplement). Surrogates of patients who received mechanical ventilation during an ICU stay of at least 3 days and were discharged between 1 and 12 months prior to study recruitment were eligible, including surrogates of patients who died. Eligible patient survivors were also invited to participate. We purposively sampled patients and surrogates to achieve variation in patient outcome (survival versus death), race/ethnicity, and surrogate-patient relationship. We excluded non-English speaking participants.

2.3. Data collection

We generated a preliminary process map based on our previous work using electronic health records (Kruser et al., 2019) and direct observations in one study ICU by a healthcare systems engineer and a physician investigator. We bound the process from onset of acute respiratory failure requiring mechanical ventilation in the ICU to one of three endpoints: (1) extubation (i.e. discontinuation of mechanical ventilation due to recovery); (2) prolonged mechanical ventilation (i.e. placement of tracheostomy or ≥ 14 days of mechanical ventilation), or (3) end of life. Qualitative data collection then followed three stages: site 1 clinician focus groups, site 1 surrogate/patient interviews, and site 2 clinician focus groups. Within each stage, data collection and analysis took place concurrently and ceased at thematic saturation (Saunders et al., 2018), defined as the point at which no additional modifications to the process map were necessary to represent participant experiences. The process map was iteratively revised during each stage and an updated map was generated before moving to the next stage.

We conducted 10 focus groups (site 1, n = 6; site 2, n = 4) with 59 clinicians representing 9 different healthcare professions. To reduce response bias due to hierarchy, focus groups were separated by professional role. During focus groups, clinicians were shown the process map and were asked to describe how the map represents their experiences and to offer modifications that better represent current care delivery. To preserve participant confidentiality, professional roles with 5 or fewer study participants are grouped as “ICU Team Member” to label direct quotes in the manuscript. To validate and refine the map using individual patient narratives, we conducted 8 semi-structured interviews with 11 participants (9 surrogates and 2 patient survivors). To reduce recall bias, we did not use the map diagram during surrogate/patient interviews, but instead followed an open-ended guide that elicited the patient’s unique longitudinal care delivery narrative. All focus groups and interviews were audio recorded and transcribed verbatim (guides in Supplement).

2.4. Analysis

We used MAXQDA 2020 (Berlin, VERBI Software) to store and code transcripts and Visio 2016 (Redmond, Washington, Microsoft Corporation) to build the map. Five investigators participated in directed qualitative content analysis (Hsieh & Shannon, 2005). An initial coding taxonomy based on the preliminary map was iteratively refined as new codes and themes emerged. Four investigators independently coded every transcript. Weekly meetings of at least three investigators were held to reconcile independent coding, reach consensus on applied codes, and revise the taxonomy. At the end of each data collection stage, all five investigators met to establish themes and relationships between codes, compare data with the map, and iteratively revise the map. We validated the map by using surrogate/patient individual narratives, iteratively revising the diagram until it reflected the care delivery and experiences of every surrogate/patient participant. We further validated the process map using clinician member checking and site-level comparative analysis (see Supplement).

3. Results

3.1. An identifiable, current-state process of ICU care delivery

The 70 study participants represented 17 distinct roles in the delivery and receipt of intensive care (Table 1). Clinician participants refined and endorsed the process map as a representation of ICU care delivery that universally applies to any medical ICU patient with acute respiratory failure requiring mechanical ventilation:

Table 1. Demographics and characteristics of study participants.

Demographic/characteristic	Participants (N = 70)
Age, years, n (%)
20–30	15 (21.4)
31–40	27 (38.6)
41–50	14 (20.0)
51–60	6 (8.6)
61–70	3 (4.3)
> 70	5 (7.1)
Female, n (%)	46 (65.7)
Race, n (%)
White or Caucasian	57 (81.4)
Black or African American	6 (8.6)
Asian	3 (4.3)
Hispanic/Latinx	1 (1.4)
Other^a	3 (4.3)
Ethnicity, n (%)
Not Hispanic/Latinx	64 (91.4)
Hispanic/Latinx	5 (7.1)
Other^a	1 (1.4)
Process Stakeholder Role, n (%)
Attending Physician^b	17 (24.3)
Clinical Nurse or Unit Nursing Leadership^b	14 (20.0)
Surrogate^c	9 (12.9)
Critical Care Fellow	7 (10.0)
Chaplain^b	5 (7.1)
Occupational Therapist	4 (5.7)
Advanced Practice Provider^b^,d	3 (4.3)
Physical Therapist	3 (4.3)
Respiratory Therapist	3 (4.3)
Social Worker^b	3 (4.3)
Patient Survivor	2 (2.9)
Palliative Care Team Member, n (%)^e	13 (22)
Site 1 Team Member, n (%)^e	41 (69)
Years since completion of professional training, median (IQR); n = 52^e^,f	6.0 (4.0-12.0)
Years of experience in study site ICU, median (IQR); n = 57^e^,g	4.0 (2.0-6.0)
Total years of experience in an ICU setting, median (IQR); n = 56^e^,g	5.0 (3.5-9.0)

^aOther includes “other race” or “other ethnicity”, missing, or declined to answer.

^bBoth ICU and palliative care team members participated from this professional background.

^cThe relationship of surrogates to the patient included spouses (3), cousins (3), adult children (2), and parent (1).

^dNurse Practitioner or Physician Assistant.

^eCharacteristic applies only to clinician participants; n = 59.

^fA total of 7 participants had not completed training.

^gTotal respondents less than 59 due to incomplete responses.

Abbreviations: ICU = intensive care unit; IQR = interquartile range.

[This map is] accurately representing all the different steps that the entire care team goes through with any patient that comes to the [medical] ICU. (ICU nurse)

The individual experiences described by surrogate and patient participants were reflected in the final map, further validating its accuracy. Figure 1 provides a summary map of the full process, illustrating the three major care phases and the four primary deliberation periods. Figure 2 demonstrates the detailed process within each care phase: initiating critical care (Phase 1, Figure 2a); developing an initial plan of care (Phase 2, Figure 2b); and reevaluating the results of care (Phase 3, Figure 2c). The combined map is available in the Supplement.

Figure 1. A summary process map of intensive care delivery for adults with acute respiratory failure who require mechanical ventilation. The current-state process has four major deliberation periods with decision points, yet these opportunities are inconsistently acted on (inconsistent steps represented with dashed lines). The blue pathways represent default processes that unfold if/when deliberation periods are bypassed.

Figure 2. Detailed process maps of intensive care delivery for adults with acute respiratory failure who require mechanical ventilation. Figures 2a–c depict the detailed processes of care within each phase of care delivery. Squares represent key process steps; squares containing a circle represent a step during which existing information is verified (i.e. a verification step); diamonds represent decision steps; and upside-down triangles indicate system-level resources available during a process step. Dashed lines indicate that a step is performed inconsistently at the patient level. “YES” demonstrates the direction of the process if the step occurs and “NO” demonstrates the direction of the process if the step does not occur. Features of the process that may differ between intensive care units (ICU) are represented in blue. Major process events are numbered to demonstrate sequence and facilitate reference; sub-processes that arise from major process events are indicated with decimals. The term “conditional yes” refers to an agreement that a treatment is acceptable to use with specific limitations or conditions for continuation (e.g. a time-limited trial of mechanical ventilation to evaluate response to therapies or pending diagnostic/prognostic information).

Abbreviations: ICU = intensive care unit; MV = mechanical ventilation

3.2. Critical periods for active deliberation and their triggers

The map contains four critical periods during an ICU stay where active deliberation about mechanical ventilation or additional life-sustaining treatments can occur: (1) intubation, (2) early deliberation after intubation, (3) physiologic deterioration, and (4) prolonged mechanical ventilation. The primary focus of each period is deliberation about appropriateness and acceptability of life-sustaining treatment. The first critical period (intubation) arises during initiation of critical care (Phase 1). Active deliberation during this period is triggered by either a “do not resuscitate” code status (Figure 2a, step 3) or a concern raised by a clinician or surrogate about the reversibility of respiratory failure (Figure 2a, step 4):

Even if someone is a full code, sometimes I stop here and I still have a decision, I have a conversation with the family about what I think is the reversibility of the process that’s going on. (ICU Physician)

Urgency and severity of acute respiratory failure often limit time for deliberation during this period. However, clinicians described how advancements in ICU technology like high-flow nasal cannula can “buy time,” alleviating time constraints.

The second period (early deliberation) arises in Phase 2 after the ICU team has started mechanical ventilation and initiated a comprehensive evaluation and plan of care. Active deliberation during this period may be triggered by new clinical information (e.g. imaging results) and/or addition of clinicians (e.g. specialists) to the care team (Figure 2b, Step 9). For example:

Once they’re intubated, you get scans, and their cancer spread. We’re not going to get the chemo anymore so we should focus on being comfortable. (ICU physician)

For patients with improving physiology who recover to extubation, no additional periods of deliberation occur. For patients with worsening physiology, a third period of deliberation (deterioration) arises in Phase 3 (Figure 2c, Step 11.1):

If they’re worsening, I always say that that’s a really good window of opportunity to make a difference. […] It makes me more confident when I’m initiating the discussion. (ICU physician)

Active deliberation during deterioration may be triggered by a need for additional life-sustaining treatments (e.g. renal replacement therapy). A fourth period of deliberation (prolonged mechanical ventilation) arises in Phase 3 (Figure 2c, step 13.1) if and when a patient’s physiology stabilizes (i.e. neither deteriorating nor improving enough to be extubated). The trigger for active deliberation during this period is typically a time threshold:

We may not want to proceed yet to the next steps, including discussion of the trach because it would be too early to do that, depending on how the patient’s physiology evolves; versus, stable physiology for 7 to 14 days, where you would definitely want to go down that pathway. (ICU physician)

These active deliberation periods share a parallel sub-process that begins with a concern raised about the appropriateness of life-sustaining treatment (Figure 2, steps 9, 11.1, 13.1), which may be followed by discussion among the patient (if able), surrogate, family, and clinicians about its acceptability (Figure 2, steps 9.1, 11.2, 13.2). Participants described a “gatekeeping” role of ICU attending physicians who decide whether to move from raising to discussing a concern,

Not necessarily because they are the most important person, but because they can put together input from everyone involved from their own team. (ICU physician)

This sub-process may encompass a series of conversations and may proceed to a decision about whether the treatment under consideration is acceptable to the patient, surrogate, family, and care team. Participants reported that ICU physicians and surrogates are typically the key decision-makers, but patients, other care team members (e.g. oncologists), or other family members can also actively contribute.

3.3. Decisions at the end of active deliberation

Participants described four potential responses to the decision of whether mechanical ventilation or another life-sustaining treatment is acceptable: (1) yes, (2) conditional yes, (3) no consensus, or (4) no (Figure 2, steps 3.1, 9.2, 11.3, 13.3). The first three lead to continuation or addition of life-sustaining treatment. If consensus is reached that the treatment is not acceptable (“no”), the patient takes an alternate pathway toward end-of-life care. “Conditional yes” describes an agreement to pursue a life-sustaining treatment with certain conditions, such as a limited period of time to evaluate clinical response or to obtain test results. Some participants referred to a specific conditional strategy known as a “time-limited trial,” and described using trials during all four deliberation periods. Clinicians noted that patients sometimes request a time-limited trial, and a surrogate participant described:

They did say if she didn’t get better within a certain amount of time, they would take her off of the machine and see what happened. (Surrogate)

Clinicians voiced differing uses of and perspectives on time-limited trials and the “conditional yes” pathway. Some physicians use trials to increase prognostic certainty:

I don’t know if you’re going to get better; how about if we […] give it 48 hours, and we should be able to tell. (ICU physician)

Other physicians use trials to help families prepare for and build consensus when end-of-life is felt to be imminent. Some clinicians expressed concerns about time-limited trials, including the concern that it is difficult to decide how long a trial should last. Several clinicians also described witnessing incomplete trials that were initiated but not followed through, whereby patients received prolonged mechanical ventilation despite having articulated preferences to avoid prolonged life-sustaining treatment.

Decision makers may also fail to reach consensus (“no consensus”) about acceptability of treatment. In Phase 1, “no consensus” leads directly to intubation, owing to the urgent nature of the phase. In Phases 2 and 3, however, “no consensus” leads to an overt disagreement step followed by a default path to continuation or addition of life-sustaining treatments. Clinicians identified these overt disagreement steps as typical opportunities to consult specialist palliative care and ethics for conflict resolution.

3.4. Inconsistency in active deliberation

The four deliberation periods arise naturally and passively as part of the current-state process, but the active response to these periods is highly inconsistent for individual patients (Table 2). Sometimes, active deliberation is bypassed entirely because the opportunity for deliberation is not recognized or acted upon. Both clinicians and surrogates described having an internal or informal discourse in which, “we question it in our minds” or informally discuss with others but do not raise a formal concern. Without someone raising a concern, active deliberation is bypassed, and life-sustaining treatments are continued or added. Individual clinicians also vary in their response to deliberation periods. For example, participants reported that experienced nurses and respiratory therapists will raise their concerns during a deliberation period while less-experienced clinicians may not. Clinicians can also be deterred from raising concerns if a physician is perceived as lacking skill or inclination to deliberate about appropriateness or acceptability of life-sustaining treatment:

Table 2. Key inconsistencies in the process of deliberation about life-sustaining treatments for patients with acute respiratory failure.

Central Theme	Exemplary Participant Quotes
Critical periods to deliberate about life-sustaining treatment are bypassed because the opportunity is not recognized or acted on	Surrogate: I didn’t want it to sound like I was giving up on him, or that I didn’t trust their ability to cure him. I just didn’t feel like me bringing it up—me bringing it up was not the right thing. ICU Team Member^a: we can also be better at raising these concerns. […] we are talking on the back end of, like, ‘Are-are you serious? We’re going to intubate this person again?’ […] we talk about it, but we don’t bring these concerns up.
Clinician-level variation leads to inconsistent actions during critical periods to deliberate	ICU Team Member: I feel like if you have a strong nurse, that’s like “no. You’re going to go talk to the family about this, because this is getting out of hand.” I think if you have a strong fellow as well […]. ICU Team Member: there are people that I will not raise concerns to. And there people that I [pause] feel comfortable […]. It’s hard to, kind of, um, voice those concerns and those frustrations, sometimes.
Regular rotation of ICU clinicians leads to avoidance or repetition of deliberation steps	ICU Nurse: we will drop the ball in a way and say, “Well, I’m only on for the weekend, so I don’t want to get involved. I’m only here for one shift. I don’t want to get involved.” ICU Team Member: one day, a physician might have a discussion with the family that ends with the family thinking that no additional interventions are acceptable, and that starts this process of consulting palliative and maybe moving toward end-of-life. And then while all that’s happening, the next day, the next care team has the same exact conversation with the family, but decides with the family that additional interventions are acceptable, and then that leads us back into that cycle.

^aTo preserve participant confidentiality, professional roles with 5 or fewer study participants were grouped as “ICU Team Member” to label direct quotes in table and manuscript text.

Abbreviations: ICU = intensive care unit.

If the attending that’s on or the fellow that’s on, they don’t really do a good job connecting or empathizing with patients and their families, I might not address it with them. (ICU nurse)

Regular rotation of ICU clinicians contributes to inconsistent deliberation through avoidance or repetition of steps.

A common scenario in that is that I end up passing that on to another decision-maker. And so that patient kind of continues in the cycle, and they end up back with a different clinical team. (ICU physician)

Different attendings may have different values and then it is like a giant reset. So we have to go through this process again. We have to go through all these boxes again. (ICU team member)

A surrogate described this disruption in the care team as “discomforting” and “distressing.” Despite describing clinician rotation as problematic, some participants also recognized the potential benefit of a “fresh set of eyes.”

If a deliberation period is bypassed, incompletely acted upon, or repeated, the process defaults to continuation or addition of life-sustaining treatment. In Phase 1, for example, intubation is default if a surrogate cannot be located or “code status” cannot be verified, because the alternative is death. In Phase 3, when active deliberation is bypassed during deterioration, treatments and procedures quickly accumulate:

We need more fluid off because that’s why they can’t wean, and now we’re going to need a CT and then we’re going to put an EEG, and then we’re going to get a spinal tap. It’s like if we don’t cross everything off the list, then we can’t come to the family and say we’ve done everything. (ICU nurse)

4. Discussion

We mapped the process of ICU care delivery for patients with acute respiratory failure, demonstrating a default pathway that leads toward prolongation and accumulation of life-sustaining treatments. This process map is the first visual diagram of the longitudinal sequence of events and decisions that unfold for patients with respiratory failure over the course of an ICU stay. The diagram underscores the complexity of this process: 27 steps divided into 3 phases with 4 critical periods for deliberation about mechanical ventilation or other life-sustaining treatment. We found that patients’ individual pathways through the process are directed by physiology, available technology, clinicians’ practice patterns, and the relationships among and between clinicians, patients, and families.

System-level norms and care processes have an important impact on individual patient decisions, including near end-of-life (Barnato et al., 2012; Drought & Koenig, 2002; Dzeng et al., 2018; Hadler et al., 2023; Lynn et al., 2000). This study has generated the first structured, visual map of how these system-level processes unfold in the context of the intensive care unit for patients with respiratory failure. This detailed depiction of how and when individual patients with respiratory failure and their families are engaged in medical decisions has important implications for patients who are facing critical illness and their families. We found that this engagement is effortful, dependent on individual clinicians’ and surrogates’ actions and agency, and fragmented by the rotation of care team members. These vulnerabilities in the current-state process help explain why ICU care delivery often results in prolonged life-sustaining treatment, despite evidence suggesting this type of care is unwanted by many patients (Barnato et al., 2007; Kahn et al., 2015; Mehta et al., 2015; Rubin et al., 2016). For clinicians, our findings underscore the need to approach decision making in the ICU as a process rather than an event. Our map of the process highlights why this decision making is so challenging for clinicians and families, by revealing a complex process that requires sustained deliberation over time and across a rotating group of clinicians. By detailing this complex process and identifying its vulnerabilities, this study provides a foundation for future efforts to transform the ICU system to better meet the needs of patients, families, and clinicians.

Our findings are important because they can be used to identify, develop, and implement new strategies that improve the current-state process. For example, we identified four potential opportunities for deliberation about the use of life-sustaining treatments that arise during a typical ICU stay. Yet, acting on these opportunities currently depends on someone raising a concern about these treatments. Many clinicians, patients, and families lack agency or awareness to perform this step, resulting in the bypass of active deliberation. Our findings suggest that this vulnerability could be overcome by new interventions that prompt and formalize active deliberation at already existing moments in this longitudinal care process. One such intervention could be the use of the process map produced in this study as a bedside clinical tool. Our detailed research map could be adapted and refined into a concise clinical tool through a human-centered design approach (Krolikowski et al., 2022). A shared, visual process map has the potential to prevent unintentional bypass of deliberation periods, maintain deliberation across rotating clinicians, and overcome the lack of agency that discourages some individuals from engaging in the deliberation about life-sustaining treatment (Kruser et al., 2023). Similar applications of process maps as bedside tools to set expectations and guide decision making are being developed in surgery and chronic illness care (Ghaferi & Wells, 2021; Scherer et al., 2021). In addition, researchers, clinicians, and ICU leaders could use our process map to improve implementation of other, existing interventions designed to improve communication among ICU clinicians, patients, and families (Curtis et al., 2016; Scheunemann et al., 2011; White et al., 2018). The map may improve the effectiveness of communication interventions, by tailoring the timing and context of their use to align with the underlying processes of care.

Our findings also identify opportunities to improve a recently proposed, promising care delivery model known as a time-limited trial of critical care (Aslakson, 2015; Chang et al., 2021; Quill & Holloway, 2011; Schenker et al., 2013). During time-limited trials, clinicians and patients or their surrogates agree to attempt life-sustaining treatment for a specified time period, with a plan to reassess continued use based on the patient’s response. Our study confirms that clinicians use and endorse time-limited trials (Aslakson, 2015; Chang et al., 2021; Schenker et al., 2013) and describes how these trials are initiated during deliberation periods. However, our qualitative data reveal that such trials often fail to be completed. These findings suggest that efforts to optimize time-limited trials should focus on supporting trials beyond initiation, including standardizing a new period for deliberation at the end of a time-limited trial.

Finally, this application of process mapping has broader implications for the science of healthcare delivery. Recognition that undesirable but avoidable healthcare outcomes result more from system and process failures than individual human errors has not been accompanied by adequate tools to investigate and modify the system (Kaplan et al., 2013). We demonstrate the utility of process mapping to characterize complex cognitive and behavioral processes that underpin ICU care delivery. This study further adds support for the observation that process mapping in healthcare is a valuable method to uncover, visualize, and explain norms of decision making that are difficult to observe directly and that arise over time (Antonacci et al., 2021). This study also demonstrates the value of combining process mapping with inductive qualitative analysis to generate rich insights into behaviors and relationships within a complex social network.

This study has limitations. Our map represents the current state of intensive care, not an ideal state of deliberation and decision making. Thus, we have not addressed whether the current process ought to be completely redesigned or whether outcomes can be improved by addressing identified vulnerabilities. We used qualitative data to enrich our description of the process and important vulnerabilities in the current state, but this study does not quantify the frequency nor the statistical association between individual process failures and patient outcomes. Future work could use our process model to prioritize risks and vulnerabilities using a method such as a failure mode, effects, and criticality analysis (McElroy et al., 2016). A key strength of our approach was including a broad range of stakeholders (17 distinct process roles, including 9 surrogates), but patient enrollment was limited because most eligible patients had died or were unable to participate due to ongoing illness. This study also took place at highly-resourced, academic medical centers and excluded non-English speakers. Thus, the map may not resonate with all stakeholders and in other contexts. Nevertheless, our version of the map can be revised and adapted to the context of any ICU. Finally, the map is detailed but not exhaustive. Future studies should focus on specifying additional sub-processes.

5. Conclusions

A visual map of ICU care delivery for patients with acute respiratory failure reveals a process that is predisposed toward prolongation and accumulation of life-sustaining treatment. Standardizing active deliberation among patients, families, and clinicians during critical opportunities that arise during an ICU stay may improve this consequential process and its outcomes.

Consent and approval statement

The Northwestern University Institutional Review Board (IRB) reviewed and approved the study. Written informed consent was obtained from each subject.

Prior presentations

A portion of the findings were previously presented in abstract form at the Society of Medical Decision Making virtual conference in October 2020

Role of the funder

NIH/NHLBI were not involved in the writing or the decision to submit the article for publication.

Acknowledgements

Contributors: We would like to acknowledge Drs. Margaret (Gretchen) Schwarze, Christopher Cox and Theodore (Jack) Iwashyna for comments on an earlier version of the submitted manuscript and Dr. Joy Moy for editorial assistance in manuscript preparation.

Disclosure statement

The first author’s spouse receives honoraria for lectures and speakers bureaus from Astra Zeneca. The remaining authors report no conflict of interest.

Additional information

Funding

This work was supported, in part, by NIH/NHLBI under Grant numbers K23HL146890 and K23HL157364.