FormalPara Take-home message

This qualitative study, with direct clinical observations and 87 staff interviews in eight hospitals in the United States, found four hospital-level factors that influenced intensive care unit (ICU) admission practices: (1) availability of ICU beds or alternative care locations; (2) availability, experience, comfort, and rapport of staff; (3) the hospital’s place in the hierarchy of interhospital transfer networks; and (4) hospital policies related to ICU admission. Health care systems should evaluate their use of ICU care to ensure that ICU admission decisions are centered around the needs of their critically ill patients while also considering the resources and constraints of their hospitals

Introduction

Intensive care is recognized as lifesaving for some patients but non-beneficial, costly, and even harmful for others [1,2,3,4]. Guidelines for intensive care unit (ICU) admission recommend patients receive care based solely on their ability to benefit [5]. “ICU benefit” has traditionally been defined at the patient-level, as a function of a patient’s severity of illness, likelihood of survival, and treatment preferences [6].

Yet, the degree of variability in ICU use between hospitals substantially exceeds that which can be explained by observable variation in severity of illness or treatment preferences. Some hospitals in the United States (US) use intensive care 20 times more than others, suggesting there are additional, unrecognized drivers of ICU admission [7].

There is a need for a more complete understanding of how ICU admission decisions are made. We adopted a strategy of open-ended inquiry, working on the assumption that clinicians in practice have identified other hospital-level factors important to decision-making—factors that may be overlooked in guidelines and quantitative work. We sought to identify these factors to build upon the current paradigm focusing on severity of illness and preferences, support the development of next-generation guidelines, and refine the use of intensive care.

Methods

To understand hospital-level factors that influence ICU admissions, we conducted clinical observations and one-on-one interviews with health care staff involved with ICU triage at eight hospitals. We focused on how ICU admission decisions might be made for patients who could reasonably receive ICU or general care (i.e., those without obvious indications for ICU care). ICU use for these patients is highly variable and most likely to be influenced by hospital factors [1, 2, 4, 5, 8]. In contrast, patients with clear ICU indications (e.g., those receiving invasive mechanical ventilation or vasopressors) nearly always receive ICU care in the US [1, 6, 9, 10].

Hospital selection

Eight unaffiliated US hospitals in Michigan were selected for variation across four dimensions: ICU admission rates, number of ICU beds, teaching status, and rurality. Additional details about hospital selection, interviews, and data analysis are in Appendix A in supplementary material.

Site visits

A primary informant was identified at each hospital to guide site visits and recruitment. Prior to the site visit, the primary informant (typically an ICU nurse supervisor) completed a survey of hospital characteristics (Appendix B in supplementary material) and a semi-structured interview to understand hospital context (Appendix C supplementary material). A preliminary diagram was then created to describe the ICU admission process (Appendix D supplementary material). It was used to inform site visits and as a discussion prompt in interviews.

Site visits were conducted by TSV, an ICU physician and health services researcher, and JM and LM, research assistants with training in sociology. Site visits included guided hospital tours and observations of routine care. Field notes were collected during site visits (Appendix E supplementary material).

Interviews

Participants were purposively sampled based on their roles within hospitals: physicians, nurses, and other key informants (e.g., bed managers, quality officers). Between eight and 12 participants were recruited at each site [11].

Separate interview guides were developed specific to ICU physicians, emergency department (ED) or hospitalist physicians, and other key informants (Appendix C supplementary material). Interview guides were pilot tested with ten physicians and nurses from an academic medical center that did not participate. These data were used to refine the interview guides and develop a draft codebook.

One-on-one, semi-structured interviews were conducted by JM or LM. Interviews lasted 30–75 min and were transcribed verbatim by a medical transcriptionist. All data were collected from May 2019 to February 2020. We used the Journal Article Reporting Standards for Qualitative Research to report this study [12].

Data analysis

Analysis of field notes and interviews occurred iteratively with data collection. Interpretive description, an inductive approach to context-sensitive qualitative inquiry in clinical settings, framed the analysis [13, 14]. Four study team members (TSV, AS, LM, KL) iteratively developed a codebook and identified themes through discussion (Appendix F supplementary material).

Results

The study team spent 40 h observing ICU operations in the eight hospitals. Of 120 individuals identified for recruitment, 33 did not participate in the study. Most non-participants were ED or general care physicians from larger hospitals, who were unable to be contacted or did not respond to interview requests. Eighty-seven participants were interviewed: 20 ICU physicians (with five ICU directors); 15 ED physicians; 17 hospitalists (with four who worked in both general and ICU care); 16 ICU nurses; eight ED, general care, or rapid response nurses; and 11 other staff (including supervisors, advanced practice professionals, transfer staff, and bed managers) (Table 1).

Table 1 Hospital and participant characteristics
Full size table

ICU admission rates ranged from 3 to 21% across hospitals. Three of eight hospitals were in rural areas, and four were teaching hospitals. ICU bed capacity ranged from eight to 156. Half of hospitals had closed admission models (i.e., patients were admitted under the sole care of intensivists). In varying combinations, most hospitals employed intensivists, utilized advanced practice professionals in ICUs, or had intermediate care units (Table 1). ICU admission processes varied substantially across hospitals (Appendix D supplementary material).

We identified four hospital-level factors influencing whether a patient might receive ICU care: (1) availability of ICU beds or alternative care locations; (2) availability, experience, comfort, and rapport of staff; (3) the hospital’s place in the hierarchy of interhospital transfer networks; and (4) hospital policies (actual and perceived) related to ICU admission (Fig. 1).

Fig. 1
figure 1

Hospital-level factors affecting ICU admission for critically ill patients

Full size image

Availability of ICU beds or alternative care locations

Participants spoke about ICUs routinely operating near capacity (Table 2, quotations 1–3 (Q1-3)), which triggered difficult conversations about patient placement. To accommodate new patients, existing ICU patients were often reevaluated and rapidly transferred out of ICUs to free up space (Q3-4). However, this became complicated when no existing ICU patients were ready for transfer to general care or when general care units in the hospital were also at capacity.

Table 2 Illustrative quotations
Full size table

During these situations, clinicians were forced to identify alternatives to ICU care. Boarding critically ill patients in EDs when ICU beds were unavailable was common across hospitals. For example, Hospital E had a special ED location with ICU-trained ED staff for patients waiting for an ICU bed. Since they were able to provide ICU-level care, ED providers reported their patients were often at lowest priority for ICU beds compared to general care patients and incoming interhospital transfers. This resulted in further delay in ICU admission for these ED patients and conflict about which patients should be prioritized for open ICU beds (Q5).

Intermediate care units provided an alternative care location for patients deemed too sick for general care but not ill enough for ICU care (Q6-7). Participants at hospitals without intermediate care described decision-making as more “binary” than hospitals with intermediate care (Q8-9).

Availability, experience, comfort, and rapport of staff

Participants described ICU admission as dependent on an ICU bed but also, critically, on the availability of an ICU nurse (Q10–11). Hospital C, for example, had nine ICU beds, and the nurse-to-patient ratio was 1:2. Typically, this ICU staffed four nurses per shift, who could care for eight patients. However, if a ninth ICU patient were to be admitted, two options were available. A fifth nurse could be called in to work; however, it often took hours for an on-call nurse to arrive. Alternatively, a nurse with ICU experience could be shifted from another hospital unit. To prevent shortages, nurses described keeping patients in the ICU longer than necessary to maintain nurse staffing (Q12–13).

There were times when patients were admitted to ICUs because general care staff were uncomfortable caring for higher-acuity patients, even if ICU staff felt the patient did not require ICU-level care (Q14). Participants spoke about recognizing different staff members’ levels of comfort and experience: “It's important to maintain collegial relations with everyone in the hospital, so if someone genuinely wants someone in the ICU, I don't put up a big fight” (ICU physician/director, Hospital F).

Experience also varied by hospital size. Patients in smaller hospitals were more likely to be admitted to ICUs (Q15) or transferred out of the hospital for higher levels of care: “We have very limited resources, we're in the middle of nowhere, there aren't a lot of specialists. If there are, they're at home sleeping. If a patient decompensates, it's…us against the world” (ED physician, Hospital C). This reliance on ICUs to care for patients who might receive general care at other hospitals ran the risk of perpetuating a cycle in which general care staff had fewer opportunities to gain experience caring for sicker patients, resulting in increasing dependence on ICU care over time (Q16). Over-reliance on ICUs was also a common source of dissatisfaction among ICU nurses, who disliked when patients were placed in ICU care for reasons such as increased nursing attention (Q17–18).

A culture of direct communication and rapport between units and staff contributed to decreased ICU admissions. One hospitalist described an incident in which she decided to keep a sick but stable patient in general care with a bedside nurse whom she trusted, despite protests from the charge nurse (Q19).

Hierarchy of interhospital transfer networks

The role of interhospital transfers on ICU admissions differed depending on the hospital’s place within the hierarchy. The largest hospitals (Hospitals D and E) had resources to care for all patients. Mid-sized hospitals (Hospitals G and H) were able to avoid transferring most patients, unless they needed highly specialized treatments, like extracorporeal membrane oxygenation.

However, smaller hospitals (Hospitals A and C) were frequently forced to consider transfer at the time of ICU admission (Q20–21). To clinicians in these hospitals, triage was described as less of a decision between general or ICU care, but rather ICU care at their hospital or at a larger hospital. Participants in smaller hospitals described some patients being transferred out of the hospital on a Friday if staff anticipated the patient might worsen and require a service not offered over the weekend, like renal replacement therapy or cardiac catheterization. Additional factors, such as geographic isolation, inclement weather, or shortages of ambulance staff sometimes made transfers difficult and had to be considered at the time of ICU admission (Q22–23). Patients who might need a higher level of care at some point during their hospitalization were sometimes kept in EDs rather than admitted to ICUs because interhospital ED transfers were perceived as more expeditious than interhospital ICU transfers (Q24).

Hospital policies related to ICU admission

Most hospital policies dealt with conditions clearly requiring ICU care or increased nursing attention, like mechanical ventilation, titrated medications, or alcohol withdrawal. Sometimes, certain policies made it easier to admit patients without obvious ICU needs to an ICU (Q25). However, more often, policies were confusing, unknown, or nonexistent (Q26). Participants were often unsure whether certain norms were codified in hospital policy. One participant described their hospital’s practice of admitting all patients with non-ST-elevation myocardial infarction to an ICU: “But when I pulled the policy, that wasn't in the policy. But we believed that to be the truth” (ED physician, Hospital E).

In several cases, admission guidelines dictated stable patients be placed in ICUs, erring on the side of caution but inefficiently using ICU resources (Q27–28). Participants often discussed a willingness to disregard both formal and informal admission policies if they felt strongly about where a patient would be best cared for, making it difficult to establish or enforce standardized admission rules: “We can tell if a patient is sick as a dog and about to crump just by looking at them. I can know nothing about a patient, I can never have looked in their chart, and I can walk in the room and be like, ‘ICU’. So, I think by allowing us that freedom—and we do respect the ICU—I think we really utilize it effectively and appropriately without having to put people there that don't require it” (Hospitalist, Hospital G). However, some participants described feeling pressured to use ICUs more frequently, even when they thought patients did not need ICU-level care: “I do think there is a well-intentioned, although overly simplistic view, from hospital administration, that the patient will be safest and get the best care in the ICU. So when in doubt, put them there. And again, that is well-intentioned, although not always true” (Hospitalist, Hospital B).

Participants in hospitals that had recently undergone changes to their organizational structure described ICU admission as a more contested process. For example, Hospital B had recently transitioned from an open (hospitalists admitted and cared for ICU patients) to closed (only intensivists admitted and cared for ICU patients) model, resulting in resistance from some staff and a breakdown in shared understandings of ICU admission. Hospitalists were particularly resistant to the shift in patient care, which now involved being restricted from using ICU care at their discretion and “handing over” patients to intensivists (Q29–30).

Discussion

This study found that actual decisions to use ICU care—and by proxy, beliefs about ICU benefit for individual patients—are not entirely patient-centered and vary based on the hospital. We identified four hospital-level determinants influencing whether a patient receives ICU care: (1) availability of ICU beds or alternative care locations; (2) availability, experience, comfort, and rapport of staff; (3) the hospital’s place in the hierarchy of interhospital transfer networks; and (4) hospital policies related to ICU admission.

International guidelines recommend ICU admission be offered to patients based solely on their ability to benefit from intensive care [5]. However, the use of ICU benefit to guide admission decision-making is problematic for several reasons. First, a patient’s likelihood to benefit from ICU care depends on several patient factors (e.g., severity of illness, chance of survival, treatment preferences) that are poorly measured and dynamic. Second, ICU benefit depends on a subjective hospital-specific counterfactual—the difference in a patient’s outcome (typically, survival) if a patient received ICU care compared to an alternative to ICU care (e.g., general or intermediate care). Therefore, ICU benefit may vary depending on both the patient and the hospital.

Most studies examining variation in ICU use have focused on ICU bed capacity and availability [15,16,17,18,19]. Bed availability plays an important role in ICU admissions. Our study, in addition to others [20,21,22], supports the belief that the likelihood of a patient receiving ICU care diminishes with fewer available ICU beds. As bed capacity becomes limited, clinicians were more likely to use alternatives to ICU care. This theme of ICU bed availability is particularly important when considering resource constraints across high-, middle-, and low-income health systems globally. For example, some health systems are shifting ICU patients towards intermediate care, with its use rising steadily over the past 2 decades [23, 24]. Yet, the concept of intermediate care and the types of patients who should receive intermediate care remain poorly defined.

Nurse staffing ratios have previously been associated with outcomes for critically ill patients [25]. However, our study also identified nurse staffing as crucial to true ICU bed availability. Because nurse staffing ratios in the US are typically fixed within intensive care, participants described strategies (e.g., delaying transfers of stabilized patients out of ICUs to keep nurses in ICUs, re-allocating general care nurses to ICUs, calling in nurses from home) to ensure there were enough nurses available to avoid closing off ICU beds. These strategies were particularly important in hospitals dealing with nursing shortages. However, nurse staffing is generally unmeasured in most research datasets used to evaluate ICU use, highlighting a critical gap [26].

Experienced physicians and nurses were often felt to be more comfortable caring for sick patients, whether in general or ICU care. At the same time, ICU nurses and physicians described dissatisfaction when forced to provide ICU care to patients without obvious critical care needs. These themes are particularly pertinent, given increased concerns about workforce turnover and burnout [27]. Critical care outreach teams could play a role in improving ICU admission decision-making, though evidence about their ability to improve outcomes is mixed [28]. Participating hospitals had nurse-led rapid response teams, but dominant themes did not emerge about their role in decision-making.

Factors such as bed availability and staff experience also influenced decisions to transfer patients from smaller or rural hospitals to larger hospitals for higher levels of care. These decisions were often time-sensitive, as interhospital transfers were felt to be most streamlined when they occurred in EDs, as compared to ICUs. Prior studies have demonstrated pressure to expedite transfers, combined with the cumbersome nature of the transfer process, often resulted in decisions to transfer patients to hospitals based on existing relationships and ease of transfer, rather than to hospitals offering the highest quality of care [29, 30].

Our study has several strengths. This is a large, multi-center qualitative study that examines determinants of ICU admission, with prior studies conducted within single centers [31, 32]. In addition, we compiled rich data, with contributions from clinical observations within a diverse selection of hospitals supplemented by perspectives from several types of hospital staff (e.g., ICU, ED, hospitalist physicians; ICU, ED, rapid response nurses; administrators).

This study also has certain limitations. Because this is an exploratory qualitative study, we cannot comment on causality between hospital-level factors and ICU admission. Furthermore, practices at these eight fee-for-service US hospitals may not generalize to all hospitals, particularly internationally. This study was also conducted prior to the coronavirus disease 2019 (COVID-19) pandemic, which may have altered ICU admission practices in some hospitals.

This study has important implications for patients, clinicians, and health systems. While focus has traditionally been placed on minimizing between-hospital variation in ICU use, an alternative goal may be to reduce unwarranted within-hospital variation. This strategy would recognize that ICU admission decisions may reasonably differ between hospitals, depending on their available resources. As such, individual hospitals would ideally develop institutional consensus about the broad patterns of patients who should receive ICU care in their hospital.

Conclusion

Guidelines recommend that ICU admission be based on a patient’s likelihood to benefit from ICU care. However, these guidelines have been nearly impossible to operationalize. ICU use should be tailored by individual hospitals to reflect the needs of their patients, often driven by severity of illness and treatment preferences, while also accounting for the resources and constraints specific to the hospital.