Objective Factors Associated with Physicians’ and Nurses’ Perceptions of Intensive Care Unit Capacity Strain

Abstract

Rationale: Time-varying demand for critical care may strain the capacities of intensive care units (ICUs) to provide optimal care. Intensivists and ICU nurses may be the best judges of the strain on their ICU. Yet, it is not clear what ICU and hospital factors contribute to this perceived sense of strain among ICU providers.

Objectives: To identify measureable ICU and hospital factors associated with perceived strain by intensivists and ICU nurses.

Methods: During a 6-month prospective cohort study, we surveyed nurses and physicians responsible for bed management regarding the ability of a 24-bed medical ICU (MICU) to provide optimal critical care. We simultaneously assessed time-varying ICU-level factors, including patient census, number of admissions, average patient acuity, number of interhospital transfer requests, and censuses of other hospital units. To identify factors associated with strain, we used an algorithm for covariate selection in regression models that selects variables that contribute sufficiently to model prediction to justify their inclusion.

Measurements and Main Results: Of 254 surveys, 226 (89%) were completed by 18 charge nurses and 17 physicians. On a scale of 1 to 10 (where a higher score indicated more strain), the median perceived strain score among nurses was 6 (interquartile range, 3–7) and among physicians was 5 (interquartile range, 3–7), with moderate correlation within days (interclass correlation coefficient, 0.45; 95% confidence interval: 0.30, 0.60). Average patient acuity, MICU census, number of MICU admissions, and general ward census were included in the most efficient model of strain perceived by nurses. Only MICU census was strongly associated with strain perceived by physicians.

Conclusions: A model containing commonly available metrics of ICU census, average patient acuity, and the proportion of new admissions has validity as a model of ICU nurses’ perceived ICU capacity strain. However, only ICU census was associated with increased perceived capacity strain by physicians, highlighting the need for involvement of multiple stakeholder groups to improve our understanding of ICU capacity strain.

Studies have identified several intensive care unit (ICU)–level factors that influence patient outcomes, such as the use of clinical protocols, the volume of patients cared for, physician staffing models, and the presence of multidisciplinary teams on rounds (1–6). Beyond these “fixed” ICU characteristics, it is also possible that time-varying ICU-level factors play a role in patient care and ICU outcomes. We have defined “ICU capacity strain” as a set of temporally varying influences on the ability of an ICU to provide high-quality care for everyone who is or could become a patient in that ICU on that day (7). Several time-varying ICU factors are associated with patient outcomes, including ICU census, the proportion of new admissions, the mean acuity of other ICU patients, and bed occupancy (8–14). However, not all studies have documented such associations (15), and the impact of these and other variables on other potential consequences of capacity strain is unclear.

As front-line providers, ICU physicians and nurses are well positioned to determine when the capacity of an ICU is strained and when optimal care may not be delivered. However, the factors that contribute to clinicians’ perceptions of strain are unknown. Understanding the relationship between time-varying ICU factors and clinicians’ perceptions would further inform and expand our understanding of ICU capacity strain. Therefore, in this study, we sought to identify measureable ICU and hospital factors that are associated with ICU physicians’ and nurses’ perceptions of capacity strain. We measured many potential measures of capacity strain that would be available in most institutions with electronic medical records or patient-flow repositories, and assessed which were related to front-line clinicians’ concurrent perceptions of capacity strain in an ICU. Given the unique characteristics of each institution, our goal was not to develop a single capacity strain “score,” but to identify the key data elements that would need to be measured to most efficiently capture the construct of perceived capacity strain in many ICUs.

Methods

Overview of Study Design

We performed a prospective cohort study of patients admitted to the medical ICU (MICU) of the Hospital of the University of Pennsylvania from June 15 to December 14, 2010. This tertiary-care hospital serves as a regional referral center for patients in all specialties, including medical critical care. The MICU is a 24-bed closed unit staffed by critical care attending physicians, fellows, and medical residents that admits adults with nonsurgical diagnoses from the emergency department, other hospital units, and as interhospital transfers. Patients who survive their ICU admission are most commonly discharged to one of the general medical wards (which have a total of 205 beds) before hospital discharge, but they are also sometimes discharged to another ICU, a nonmedical (e.g., surgical) ward, to another institution, or directly to home. The MICU is a high-acuity unit, with a median Acute Physiology and Chronic Health Evaluation III (APACHE III) (16) score of 71 (interquartile range [IQR], 58–81) during the study period.

We administered a questionnaire about perceptions of capacity strain on a daily basis to charge nurses and physician arbitrators during the study period, as detailed subsequently. The results of this questionnaire were merged, by calendar day, to an analytic database containing 11 MICU-level and hospital-level variables that we identified a priori as potential contributors to ICU capacity strain (Table 1). The rationale for studying these particular parameters was that we believed their daily variability could place uneven demand on constant resources and staffing and, therefore, strain the ICU’s capacity. We then assessed relationships between these candidate measures of strain and the “perceived strain score,” defined as the perception of strain identified by the charge nurse and physician arbitrator. Charge nurses oversee the nurse staffing and bed management issues for the entire unit. One of the two ICU attending physicians on service functions as the arbitrator, with authority to accept or decline outside hospital transfer requests and to prioritize beds for patients already within the hospital (including those in the emergency room).

Table 1.

Hospital and intensive care unit characteristics

Variable	Definition	Survey Days (n = 117)	Nonsurvey Days (n = 60)	P Value*
ICU characteristics
APACHE III score	Mean of admission APACHE III scores of all MICU patients as a measure of daily overall acuity	70.9 (20.5)	70.2 (17.7)	0.50
MICU census	Number of patients spending at least 2 hours on the MICU service during the calendar day	28 (26–30)	27 (25–29)	0.02
MICU admissions	Number of new admissions to the MICU	5 (4–7)	5 (3–6)	0.12
Days with nonmedical discharges, n (%)	Number of days with patients discharged to nonmedical general wards	10 (8)	5 (8)	0.94
Transfer requests	Number of accepted patient transfer requests from an outside facility	0 (0–1)	0 (0–1)	0.83
Transfer summary, n (%)	Categorical variable for ability to accept scheduled transfers
	No scheduled transfers	59 (50)	32 (53)
	All scheduled transfers occurred	22 (19)	9 (15)
	One or more scheduled transfers did not occur	36 (31)	19 (32)
Hospital characteristics
Medical ward census†	Number of patients on all general medical wards	234 (227–240)	213 (203–222)	<0.001
Nonmedical ward census	Number of medical patients on all nonmedical general wards	6 (3–9)	6 (3–9)	0.91
Emergency department census	Number of patients registered in emergency department	49 (43–53)	43.5 (38–48.5)	<0.001
Transition unit census	Number of patients admitted to transition unit (observation unit)	32 (29–35)	26.5 (21.5–30)	<0.001
Other ICU census	Number of patients admitted to any other hospital ICU	82 (78–85)	70 (66.5–75)	<0.001

Definition of abbreviations: APACHE = Acute Physiology and Chronic Health Evaluation; ICU = intensive care unit; MICU = medical ICU.

All results are presented as median (IQR) except where noted. All census variables were defined as the number of patients admitted to the respective unit for at least 2 hours during the calendar day.

^*P values reflect comparison of survey and nonsurvey days, using the Wilcoxon rank-sum test.

^†The medical wards have a total of 205 beds.

Candidate Variables

All candidate capacity strain variables (Table 1) were calculated over the calendar day. For example, the number of interhospital transfer requests for new admissions was counted from 12:01 a.m. until midnight. The MICU census included all patients spending at least 2 hours on the MICU service during the calendar day, including patients who overflowed to other units. All other census measures were defined as the number of patients occupying a bed in the respective unit for at least 2 hours during the calendar day. Because of daily discharges and admissions, a unit’s census could be greater than the total number of beds in that unit.

Questionnaire Development and Administration

We developed a questionnaire about perceptions of ICU capacity strain with the input of a focus group of critical care specialists and refined it during a 2-week pilot administration. The final questionnaire consisted of five questions, four of which inquired about the perceived adequacy of the supply of specific ICU resources that day (ICU beds, nurses, respiratory therapists, and residents and/or mid-level practitioners). Respondents indicated how many more or fewer units of each of these resources than were available would be necessary “to provide optimal care to all patients today.” The questionnaire specified that respondents should consider all patients already admitted to the MICU as well as those for whom a request for a MICU bed had been made. The final question asked respondents to rate their overall perception of ICU capacity strain on that day, using a scale from 1 to 10. We specifically asked about strain on the capacity of the ICU as a whole, not on an individual practitioner.

For all question responses, we provided descriptive anchors: a score of 1 represented the least strain the clinician had ever experienced and 10 the most. The full questionnaire is available in the online supplement. Research assistants administered questionnaires in person to the charge nurse and physician arbitrator in the Hospital of the University of Pennsylvania MICU between 4 and 6 p.m. during the study period, excluding weekends and holidays.

Statistical Analysis

We summarized all variables using standard descriptive statistics and present the raw values for all subjective responses (i.e., resources perceived to be needed and perceived strain scores). We assessed the unadjusted relationship between nursing and physician perceived strain scores on each day, using the intraclass correlation with date as the clustering variable. Because several candidate capacity strain variables had nonparametric distributions, we compared their values on days for which we did and did not have completed surveys, using Wilcoxon rank-sum tests.

We performed multivariable analyses to determine which resources were deemed necessary by survey respondents in times of higher perceived capacity strain. To do so, we used linear regression models, in which the perceived strain score was the dependent variable and the independent variables included each resource about which we inquired (numbers of ICU beds, nurses, respiratory therapists, and residents and/or mid-level practitioners). We selected these variables for inclusion a priori, and all were included in the final models. We built separate models for charge nurse and physician arbitrator responses. Because providers may differ in their general perceptions, we standardized perceived strain scores by individual for all multivariable models. Standardized values were calculated as the difference between the raw number and the mean of that provider’s responses across days, divided by the standard deviation of all responses given by the same provider. Standardized perceived strain scores account for the heterogeneous distributions of responses by the same individual.

We performed univariate analyses of candidate strain variables with physicians’ and nurses’ perceived strain scores, using linear regression models with standardized scores as described above as the dependent variable. We then performed a multivariable analysis to identify those strain variables that were independently associated with physicians’ and nurses’ perceived strain scores, again using standardized scores. To do so we used least-angle regression, a model-building algorithm that considers parsimony as well as predictive accuracy (17). This algorithm sequentially fits models in which predictors are added to or dropped from the active set, and the coefficients of each included variable are updated. The algorithm is similar to forward stepwise regression in that it begins with each candidate variable’s coefficient equal to zero and then finds the predictor most highly correlated with the outcome. Successive variables are then entered, individually and in tandem, to assess improvement in model fit while penalizing the model for each extra variable added. Thus, only covariates that contribute more to prediction than would be expected for any given covariate are retained. This process is repeated until the value of the likelihood ratio test decreases from the prior model. This final model is considered to be the optimal set of linearly independent predictors.

An α value less than 0.05 was considered statistically significant. All analyses were performed with Stata 12 (StataCorp, College Station, TX). The study was approved by the Institutional Review Board of the University of Pennsylvania.

Results

Perceptions of Capacity Strain

The study period included a total of 183 days, with 127 days eligible for surveys (i.e., nonweekend and nonholiday d). Of 254 possible surveys, 226 (89%) were completed by 18 charge nurses and 17 physician arbitrators on 122 of the 127 (96%) eligible days. Among responders, 14 charge nurses and 11 physicians provided voluntary demographic data. Among nurses, 2 (14%) were male; all were white; and 10 (71%) had no more than 5 years of experience as a charge nurse. Among physician arbitrators, 7 (63%) were male; 6 (55%) were white, 2 (18%) were black, and 3 (27%) were Asian. Five (45%) physicians had more than 5 years of experience as MICU attending physician.

Of the 122 days for which we obtained completed surveys, 117 (96%) had complete data about candidate strain variables. Hospital and ICU characteristics on these days are summarized and compared with nonsurvey days in Table 1. In general, candidate strain variables were lower on nonsurvey days, which included weekends and holidays.

Among all clinician ratings on these days, the median perceived strain score was 5 (IQR, 3–7). When stratified by type of respondent (charge nurse or physician arbitrator), the median and IQR were the same. There was moderate correlation between the scores of physicians and nurses on the same day (intraclass correlation, 0.45; 95% confidence interval, 0.30–0.60). Figure 1 illustrates the correlation as a scatter plot.

Figure 1.

Scatter plot with Lowess line of correlation between charge nurse and arbitrator perceived strain scores. Lowess: locally weighted least squares regression.

The perceived need for additional resources during the study period is summarized in Table 2. For most resources, the median number of additional units was 0. Higher perceived strain scores reported by charge nurses were associated with their perceived need for additional ICU nurses (P < 0.001). Higher physician scores were associated with their perceived need for additional beds (P = 0.004) and residents and/or mid-level practitioners (P = 0.004).

Table 2.

Perceptions of additional resources needed with increasing intensive care unit strain

Perceptions	Median (IQR)
Charge nurse perceptions
ICU beds	2 (0, 3)
Nurses	1 (0, 2)
Respiratory therapists	0 (0, 1)
Residents and/or mid-level practitioners	0 (0, 1)
Physician perceptions
ICU beds	1 (0, 3)
Nurses	0 (0, 1)
Respiratory therapists	0 (0, 1)
Residents and/or mid-level practitioners	0 (0, 1)

Definition of abbreviations: ICU = intensive care unit; IQR = interquartile range.

Table 3 summarizes the unadjusted association of candidate strain variables with perceived strain scores. Table 4 summarizes the results of multivariable models that identify candidate strain variables that are independently associated with perceived strain scores. Among both nurses and physicians, higher ICU census was associated with higher perceived strain. Among nurses, higher mean APACHE III score and census on the general medical wards were also associated with higher perceived strain. A higher number of admissions to the ICU was associated with lower perceived strain. The proportions of variance explained by the most parsimonious models predicting the physician and nurse perceived strain scores (as measured by the model R²) were 0.067 and 0.130, respectively.

Table 3.

Univariate analyses of association between candidate strain variables and perceived strain scores^*

Variable	Charge Nurse			Arbitrator
	β Coefficient	95% CI	P Value	β Coefficient	95% CI	P Value
MICU census	0.10	0.03, 0.16	0.004	0.11	0.04, 0.17	0.001
Mean APACHE III score	0.01	−0.00, 0.02	0.094	0.00	−0.01, 0.01	0.396
MICU admissions	0.01	−0.08, 0.10	0.813	0.07	−0.02, 0.17	0.106
Nonmedical discharges	0.03	−0.46, 0.52	0.891	0.48	0.01, 0.95	0.046
Transfers accepted	−0.08	−0.31, 0.15	0.502	0.06	−0.17, 0.28	0.623
Transfers accepted and admitted	−0.21	−0.62, 0.20	0.308	−0.15	−0.52, 0.22	0.419
Transfers accepted but unable to admit	−0.02	−0.23, 0.19	0.845	0.08	−0.13, 0.29	0.439
Medical ward census	0.00	−0.02, 0.02	0.979	−0.01	−0.03, 0.01	0.343
Nonmedical ward census	0.01	−0.03, 0.04	0.746	−0.01	−0.04, 0.02	0.585
Emergency department census	−0.01	−0.04, 0.01	0.311	0.00	−0.03, 0.02	0.910
Transition unit census	0.00	−0.04, 0.04	0.893	0.00	−0.05, 0.04	0.833
Other ICU census	−0.01	−0.04, 0.03	0.614	0.02	−0.02, 0.06	0.255

Definition of abbreviations: APACHE = Acute Physiology and Chronic Health Evaluation; CI = confidence interval; ICU = intensive care unit; MICU = medical ICU.

^*Perceived strain scores were standardized by respondent to account for nonhomogeneous variability in scores given by individuals. Because the perceived strain score is subjective and standardized, the β coefficients and 95% confidence intervals cannot be easily interpreted quantitatively, but are provided to give some measure of the magnitude of the effect.

Table 4.

Predictors of clinicians’ perceived intensive care unit strain in final models

	β Coefficient*	95% CI	Cumulative R2
Charge nurse perceived strain score
MICU census	0.147	0.069, 0.225	0.052
Mean APACHE III score	0.007	−0.001, 0.015	0.072
No. of new MICU admissions	−0.126	−0.231, –0.021	0.110
Census on general medical wards	0.008	−0.010, 0.025	0.130
Physician perceived strain score
MICU census	0.104	0.040, 0.169	0.067

Definition of abbreviations: APACHE = Acute Physiology and Chronic Health Evaluation; CI = confidence interval; MICU = medical intensive care unit.

^*Because the perceived strain scores are subjective and standardized, the β coefficients and 95% confidence intervals have no quantitative interpretation but are provided to give some measure of the magnitude of the effect.

Discussion

This study demonstrates that in a tertiary-care MICU, several objective factors are associated with clinicians’ perceptions of ICU capacity strain—that is, strain on the ability of an ICU to provide optimal care to patients that day. Specifically, increased ICU census was associated with increased perceived capacity strain among both physicians and nurses. Among charge nurses, higher average acuity of patients in the census and higher census on the general medical wards were also associated with increased perceived capacity strain, and higher proportion of the census comprising new admissions that day was associated with lower perceived capacity strain. This last finding is consistent with a previous study showing that the proportion of the census comprising new admissions on a given day is inversely associated with in-hospital mortality among patients admitted that day (12). These consistent, but perhaps unexpected, findings suggest that higher numbers of admissions occur at times when the threshold for admission is lower (i.e., admitted patients may have less severe disease), perhaps due to the availability of more beds.

This study also found that there was moderate correlation between nurse and physician overall perceptions of capacity strain. In addition, we found that the median number of most categories of additional resources needed was zero, with relatively narrow distributions, suggesting that on most days, physicians and nurses perceive adequate resources to meet the demands of the ICU. Physician arbitrators perceived a need for additional beds and clinicians in times of higher strain. Charge nurses identified a need for additional nursing resources with higher strain. The findings that the correlation of perceived strain overall was only moderate and that nurses and physicians differed in their perceived need for resources are not entirely surprising, as charge nurses and physician arbitrators have different roles and responsibilities in ICU bed management. Therefore, different information is likely to shape physicians’ and nurses’ perceptions of strain. Taken together, the differences in factors associated with perceived strain by nurses and physicians, the different perceived resource needs, and the moderate correlation of perceived strain scores highlight the importance of assessing multiple stakeholders in defining and understanding capacity strain.

This study is the first to assess the concept of ICU capacity strain using the perceptions of front-line care providers. The results complement prior studies that have sought to identify metrics of ICU capacity strain, using other approaches. In a retrospective cohort study of 264,401 patients admitted to 155 U.S. ICUs from 2001 to 2008, we found that the total ICU census on the day of a patient’s admission, standardized for the size of the ICU, was associated with a higher risk of mortality for that patient, and that higher average severity of illness among other admitted patients on the day of admission augmented the strength of that association (12). Another study found that three measures of capacity strain (ICU census, average acuity, and proportion of new admissions), when measured on the day of ICU discharge, were independently associated with patients’ preceding ICU length of stay and their odds of being readmitted to the ICU (13). Future studies should evaluate whether such measures of strain contribute to the high rates of reported symptoms of burnout among ICU clinicians (18–20).

This study also has several important limitations. First, it is a single-center study in an academic, MICU. It may have limited generalizability to other ICUs, such as ICUs in community hospitals, subspecialty ICUs, or ICUs with different staffing models. However, the findings are consistent with the results of our multicenter analyses of objective patient outcomes, which included 155 organizationally diverse ICUs, 70% of which were in community hospitals (12). Second, we performed questionnaires on nonholiday weekdays only and excluded weekends, which may differ from weekdays in terms of capacity strain and patient outcomes (21–24). Indeed, we found that the nonsurvey days had lower values of several candidate strain variables, which is consistent with the fact that occupancy and hospital activities tend to be less on weekends and holidays. If anything, these differences permit better characterization of capacity strain, as our sample of ICU days will be those most likely to be strained and perceived as such, and therefore most likely to reveal important associations with the candidate measures.

Third, we studied clinicians’ perceptions of capacity strain, which are subjective measures and potentially prone to bias. However, we believe that these perceptions among front-line clinicians provide an important barometer of capacity strain as a coherent construct given that such providers not only have expertise about when they feel strained, but may also be directly affected by strain in ways that can affect patient care.

In summary, daily measures of the ICU census, the average severity of illness in the ICU, the number of new admissions to the ICU, and the census in other units in the hospital each contribute to clinicians’ concurrent perceptions of strains on the ICU’s capacity to provide optimal critical care. These results are consistent with multicenter studies showing that these metrics are associated with patient flow and outcomes. Together, these findings suggest that a model of ICU capacity strain that accounts for patient census, acuity, and the proportion of new admissions may be used to reveal how ICU capacity strain influences a variety of care processes and outcomes in the ICU, such as communication with families, use of indicated prophylactic therapies, timeliness of resuscitation, and acquisition of nosocomial infections.