The Association Between Hospital Capacity Strain and Inpatient Outcomes in Highly Developed Countries: A Systematic Review

Abstract

Background

Increases in patient needs can strain hospital resources, which may worsen care quality and outcomes. This systematic literature review sought to understand whether hospital capacity strain is associated with worse health outcomes for hospitalized patients and to evaluate benefits and harms of health system interventions to improve care quality during times of hospital capacity strain.

Methods

Parallel searches were conducted in MEDLINE, CINAHL, the Cochrane Library, and reference lists from 1999-2015. Two reviewers assessed study eligibility. We included English-language studies describing the association between capacity strain (high census, acuity, turnover, or an indirect measure of strain such as delayed admission) and health outcomes or intermediate outcomes for children and adults hospitalized in highly developed countries. We also included studies of health system interventions to improve care during times of capacity strain. Two reviewers extracted data and assessed risk of bias using the Newcastle-Ottawa Score for observational studies and the Cochrane Collaboration Risk of Bias Assessment Tool for experimental studies.

Results

Of 5,702 potentially relevant studies, we included 44 observational and 8 experimental studies. There was marked heterogeneity in the metrics used to define capacity strain, hospital settings, and overall study quality. Mortality increased during times of capacity strain in 18 of 30 studies and in 9 of 12 studies in intensive care unit settings. No experimental studies were randomized, and none demonstrated an improvement in health outcomes after implementing the intervention. The pediatric literature is very limited; only six observational studies included children. There was insufficient study homogeneity to perform meta-analyses.

Discussion

In highly developed countries, hospital capacity strain is associated with increased mortality and worsened health outcomes. Evidence-based solutions to improve outcomes during times of capacity strain are needed.

Electronic supplementary material

The online version of this article (doi:10.1007/s11606-016-3936-3) contains supplementary material, which is available to authorized users.

INTRODUCTION

As hospitals strive to improve efficiency, increases in patient volume, acuity, and complexity can strain hospital resources. Strain can be defined as an “excessive demand on the strength, resources, or abilities”¹ of a hospital, and any resource the hospital uses to provide care (e.g., beds, nurses, physicians, equipment) can experience strain. Resource strain resulting from a mismatch between supply and demand exists on a continuum from mild strain due to routine fluctuations in patient needs to severe strain resulting from patient surges during public health emergencies. Resource strain has been well studied in the emergency department (ED), often focusing on overcrowding due to high patient volume, and ED overcrowding is associated with delayed care and increased mortality.^{2 – 4} However, less is known about the relationship between resource strain in hospital inpatient units and patient health outcomes.

Capacity strain is a subset of resource strain originally described in the intensive care unit (ICU). There is no universally accepted definition of capacity strain, but it has been defined as increased patient census, acuity, and/or turnover affecting an ICU’s ability to provide high-quality care.⁵ This concept can also be applied to non-ICU settings. Expert groups recommend strategies to improve the ability of hospitals to provide care during times of strain, particularly in response to public health emergencies.⁶ The effect of these strategies on patient outcomes is not clear. To decide whether to institute and promote these strategies, hospital leaders and health policymakers need to understand the effects of inpatient capacity strain on outcomes and the effectiveness of interventions to address these effects.

We conducted this systematic review to (1) review the association between capacity strain and health outcomes for patients receiving inpatient care and (2) evaluate the benefits and harms of health system interventions to improve quality of inpatient care during times of capacity strain. As the US hospital system has significantly less capacity for children than for adults and thus may be at particularly high risk for severe pediatric capacity strain,⁷ our goal was to focus on hospitalized children; however, we expected to find limited pediatric literature and thus expanded our scope to include adult and pediatric patients.

METHODS

Study Protocol

The study protocol was created a priori based on PRISMA-P guidance⁸ and registered with the International Prospective Register of Systematic Reviews (available at http://www.crd.york.ac.uk/PROSPERO/display_record.asp?ID=CRD42015024758). This study was deemed not human subjects research by our Institutional Review Board.

Data Sources

We searched Medline, the Cochrane Library, CINAHL, and ClinicalTrials.gov for relevant English-language studies from 1999 (the year the Institute of Medicine report To Err Is Human was published) until August 2015; we also manually identified studies from reference lists. Due to the lack of standard terms to define capacity strain, a wide variety of search terms was used (e.g., capacity strain, occupancy, surge capacity, hospital crowding). The complete search strategy can be found in the Online Appendix.

Study Selection

We included randomized and non-randomized trials, prospective and retrospective cohort studies, and case-control studies. Included studies described health outcomes or intermediate outcomes (i.e., outcomes such as length of stay or delay to emergent surgery, which are plausibly associated with health outcomes) for children and adults receiving inpatient care in acute non-psychiatric hospitals during times of inpatient capacity strain. The Online Appendix contains complete inclusion and exclusion criteria. We defined capacity strain as high patient census, acuity, or turnover⁵; we also accepted indirect measures reflecting changes in care due to inpatient capacity strain (admitted patients boarding in the ED, or patients refused ICU admission or admitted to alternate units due to lack of beds). Studies that primarily evaluated the effects of ED crowding or did not focus on outcomes of admitted patients were excluded. Included experimental studies described health system interventions to improve care for patients in either the hospital or the ED after the decision to admit to the hospital. Two reviewers assessed study eligibility (COE, RCS); disagreements were resolved through discussion and third-party review (JMG). We assessed risk of bias for experimental studies using the Cochrane Collaboration Risk of Bias Assessment Tool,⁹ summarizing risk of bias for each outcome as low, unclear, or high. For observational studies, we assessed risk of bias using the Newcastle-Ottawa Scale (NOS),¹⁰ which awards studies up to 9 total stars for participant selection (4 stars), comparability of participant groups (2 stars), and ascertainment of outcome or exposure (3 stars).

Data Abstraction and Synthesis

Two reviewers (COE, RCS) independently abstracted data on study design, patient and hospital characteristics, metrics to describe strain, outcomes, and interventions. Due to heterogeneity in measures of capacity strain, we qualitatively describe measures used to describe strain, the association between strain and health outcomes or intermediate outcomes, and the health effects of interventions to improve care during times of strain. There was insufficient study homogeneity to perform meta-analyses.

RESULTS

Of 5,702 potentially relevant studies, 52 were included for review (Fig. 1). Of 44 observational studies, 21 were performed in the US,^{11 – 31} 16 in Europe,^{32 – 47} 5 in Canada,^{48 – 52} and 2 in Australia.^{53 , 54} Thirty studies analyzed the association between capacity strain and mortality.^{11 – 25 , 32 – 43 , 48 , 49 , 53} All observational studies were cohort studies; all but four were retrospective.^{32 , 42 , 43 , 46} Eight studies described interventions to improve care during times of capacity strain,^{55 – 62} none using randomization to assign treatment category. Outcomes for children were separately analyzed in only six of the observational studies^{23 , 26 , 32 , 45 , 46 , 51} and none of the experimental studies. Characteristics of included studies are described in the Online Appendix.

Figure 1.

PRISMA flow diagram.

Measures and thresholds used to describe capacity strain varied widely (Online Appendix). One study used a composite measure that included components of census, acuity, and/or turnover¹³; a larger group used measures that included one of these concepts (most often census), often indexed to usual conditions. Other, often small studies used indirect measures of strain, most often ED boarding after the decision to admit to the hospital. Even among studies using similar concepts, there was great variation in specific measures to define strain. For example, some studies using census-based strain measures treated census as a continuous variable without a specific cutoff to define strain; other studies used cutoffs based on occupancy (e.g., >80% occupancy) or census percentiles (e.g., highest quartile of daily census). Timing of strain also varied: While most defined strain based on conditions on the day of admission, others used averages from the first 3 days or throughout the hospitalization, and one study evaluating the effect of strain on ICU readmission defined strain on the day of ICU discharge.¹⁴

NOS scores for included observational studies ranged from 3 to 8 of 9 possible stars (Online Appendix). The NOS does not include thresholds for distinguishing high- or low-quality studies. All experimental studies were assessed as having high risk of bias by the Cochrane Risk of Bias Tool because of high or unclear risk of bias for at least four of the tool’s eight criteria (Online Appendix).

Association Between Capacity Strain and Mortality

Of 30 studies examining mortality as an outcome, 12 were performed in ICU settings (Table 1). Thirteen were single-institution studies, while six included data from more than 100 hospitals; overall, mortality was analyzed for over 4 million hospitalizations. To define capacity strain, all but six of the multi-institution studies used hospital or unit census,^{12 , 13 , 16 , 19 , 20 , 36} often indexed to usual conditions (Online Appendix). Five multi-institution studies used measures of patient turnover (e.g., number of admissions) or acuity to define strain.^{12 – 15 , 19} Meanwhile, all but three of the single-institution studies used indirect measures of strain,^{24 , 41 , 49} such as time spent boarding in the ED after the decision to admit to the hospital. Most studies examined hospital mortality as the outcome of interest, while others used ICU^{14 , 22 , 40} or time-specific mortality.^{34 – 36 , 42 , 53}

Table 1.

Studies Describing the Association Between Hospital Capacity Strain and Mortality

Metrics to describe strain
Author, year	Number of hospitals	Number of admissions	Census	Admissions	Acuity	Indirect measure*	Mortality	ICU study	Children included	Key finding(s)
Rubinson, 2013	661	NR	●				↑			Increased hospital mortality for patients with stroke (15% increase) and acute myocardial infarction (20%) at hospitals experiencing high surge during the influenza pandemic compared with no-surge hospitals
Evans, 2006	∼400	28,561+		●			↔			No association between a high number of admissions on Friday and Saturday following a Thursday admission and hospital mortality
Tucker, 2002	186	13,334	●				↑	●	●	For each 10% increase in percentage of maximum occupancy on day of admission, odds of hospital mortality increased 9%; infants admitted at 50% maximum occupancy had about 50% lower odds of mortality compared to maximum occupancy
Jenkins, 2015	156	230,621		●	●		↑			Two-fold increase in hospital mortality for patients admitted during times of high trauma surge (composite of admissions and acuity), 7-fold increase in mortality for patients with firearm injuries
Gabler, 2013	107	264,401	●	●	●		↑	●		2% increase in odds of in-hospital death for each standard deviation increase in ICU census. 2% decrease in odds of in-hospital death for each 10% increase in number of admissions. No effect of ICU acuity on mortality
Wagner, 2013	107	200,730	●	●	●		↔	●		Capacity strain on day of ICU discharge not associated with increased odds of subsequent in-hospital death; discharge on days with increased admissions associated with 3% lower odds of in-hospital death
Chalfin, 2007	90	50,332				●	↑	●		29% lower odds of hospital survival among patients with delayed admission to ICU
Iapichino, 2004	89	12,615	●				↑	●		Admission to an ICU with average occupancy >80% was associated with 32–35% increase in odds of hospital mortality compared to ICUs with average occupancy ≤80%
Schwierz, 2011	72	NR	●				↔			No increase in 24-h mortality among patients admitted on days with unexpectedly high census (after adjusting for changes in unobserved risk characteristics on high-census days)
Madsen, 2014	72	2,651,021	●				↑			1.2% increase in relative risk of inpatient and 30-day mortality per 10% increase in median bed occupancy rate. Occupancy rates ≥110% associated with a 9% increase in mortality compared to occupancy rate <80%
Iwashyna, 2009	48	200,499	●				↔	●		No difference in hospital mortality with increasing census on the day of admission
Schilling, 2010	39	166,920	●				↑			6% increase in hospital mortality among patients admitted on high occupancy days. 12% increase in mortality among patients admitted during widespread or regional influenza activity
Marcin, 2004	39	102,008		●			↔			No association between high annual, quarterly, or monthly number of trauma admissions and hospital mortality
Derose, 2014	13	136,740				●	↔			No association between increased ED boarding time and inpatient mortality
Robert, 2012	10	1,332				●	↑	●		1.8-fold increased adjusted odds of 60-day mortality in patients admitted to ICU after subsequent referral compared to patients admitted immediately; no significant effect on 28-day mortality
Sprivulis, 2006	3	62,495	●				↑			Compared to patients admitted while hospital occupancy was <90%, patients admitted while occupancy was ≥100% had 30% increase in 7-day mortality hazard
Yergens, 2015	3	1,036	●				↑			Compared to patients admitted to the hospital when ICU occupancy was <80%, patients admitted when ICU occupancy was ≥90% had 72% increased odds of hospital mortality
Singer, 2011	1	41,256				●	↑			Compared to patients who did not board in the ED, patients who boarded ≥12 h had 23–43% increased hospital mortality
Plunkett, 2011	1	23,114				●	↑			Compared to patients who waited <1 h between admission team referral and ward bed placement, each category of increasing delay (1–2.5 h, 2.5–6 h, 6–14 h, >14 h) was associated with an additional 7% increase in 30-day hospital mortality
Gilligan, 2008	1	13,357				●	↔			No association between number of ED boarders at 9 a.m. and hospital mortality
Pascual, 2014	1	8,626				●	↔	●		No difference in ICU mortality between patients boarding in an overflow ICU and non-boarded patients
Serafini, 2015	1	3,828				●	↑			Among medical patients boarding in surgical wards, 1.8-fold increased adjusted odds of hospital mortality
Bekmezian, 2012	1	1,792				●	↔		●	No association between ED boarding time and hospital mortality
O’Callaghan, 2012	1	1,609				●	↔			Delay in admission to the ICU from the ED was not associated with ICU mortality
Clark, 2012	1	1,433	●				↑	●		9% increase in hospital mortality for every unit increase in ICU occupancy
Clark, 2007	1	1,200				●	↑	●		1.5% increase in hospital mortality for every 10% increase in time between decision to admit and ICU admission
Tarnow-Mordi, 2000	1	1,050	●		●		↑	●		Highest quartile of each workload measure (initial occupancy, initial workload, peak occupancy, average occupancy, average workload, average acuity) associated with 90-130% increased odds of hospital mortality
Ball, 2006	1	861		●			↔			Patients treated during mass casualty incidents did not have higher hospital mortality than trauma patients admitted during other times
Stowell, 2013	1	483				●	↔			No association between admission to outlying wards instead of usual specialty ward and unadjusted mortality at 24 h, 28 days, and 90 days
Intas, 2012	1	200				●	↑	●		Patients with ED boarding times >6 h had 5.7 times higher hospital mortality than those with shorter boarding time

*Examples of indirect measures include boarding in the emergency department after decision to admit, admission to alternate unit because of full usual unit, and refused ICU admission due to full ICU

ED: Emergency department. h: hours. ICU: Intensive care unit. NR: Not reported

There was a statistically significant increase in mortality during times of capacity strain in 18 of 30 studies and in 9 of 12 studies in ICU settings (Table 1). While two studies reported over five-fold mortality associated with capacity strain,^{13 , 43} several studies found more modest 50–150% increases in mortality.^{32 , 36 , 39 , 41 , 48} Only two studies included children: a multi-institution UK study that found a doubling in mortality odds for patients admitted at maximum compared to 50% occupancy³² and a single-institution US study that found no significant association between ED boarding time and subsequent hospital mortality.²³ Study quality did not appear to affect the likelihood of reporting a positive relationship between strain and mortality. While most studies measured strain daily or more often, one study measured monthly, quarterly, and annual variation in strain and did not find a significant association with hospital mortality¹⁹; as conditions in acute care hospitals change rapidly, it is possible that measuring strain monthly may be too infrequent to detect strain-associated changes in outcomes. The only study finding a statistically significant decrease in mortality during strained times reported decreased mortality for patients discharged on days with increased ICU admissions, but no change in mortality using other measures of strain; this was also the only study to measure strain at ICU discharge.¹⁵

While almost all studies adjusted for risk of patient mortality in multivariable analyses (Online Appendix), two studies specifically analyzed mortality for patients who had diagnoses that were likely unrelated to the cause of strain.^{11 , 49} In a large study comparing US hospitals with an increased number of admissions during the 2009 Influenza H1N1 pandemic (“strained” hospitals) to hospitals with no increased admissions, Rubinson et al. reported an approximately 15–20% increase in the odds of hospital mortality for patients admitted to strained hospitals with stroke or acute myocardial infarction.¹¹

Association Between Capacity Strain and Other Outcomes

Eight studies examined the association between capacity strain and nonlethal adverse events,^{22 , 30 – 32 , 38 , 44 , 45 , 47} with five of eight identifying a statistically significant association between strain and aspiration pneumonia,²² methicillin-resistant Staphylococcus aureus infection,⁴⁴ Clostridium difficile infection,⁴⁷ or adverse events in general.^{31 , 45} Of two studies including children, one described an almost doubling in patient-related adverse events during times of high pediatric ICU occupancy,⁴⁵ while one found no significant association between neonatal ICU occupancy immediately before admission and development of nosocomial bacteremia.³²

Of 15 studies examining the relationship between capacity strain and hospital, ICU, or postoperative length of stay (LOS), 10 reported a significant association between strain and increased LOS^{16 , 20 , 21 , 23 , 26 , 27 , 34 , 42 , 49 , 54}; both pediatric studies reported an association between strain and increased LOS.^{23 , 26} The magnitude of increase in LOS ranged from 1 h²³ to more than 1 day²⁷; the greatest increases in LOS were reported in studies that examined strain during mass casualty incidents.^{27 , 49} The previously described study that measured capacity strain at ICU discharge reported that patients discharged from the ICU on days with high ICU census, acuity, and admissions had shorter ICU and post-ICU LOS.¹⁵

Seventeen studies examined the relationship between capacity strain and additional outcomes, such as ICU or hospital readmission,^{12 , 15 , 19 , 23 , 26 , 28 , 29 , 34 , 42} representation to the ED,⁵² ICU admission,^{21 , 46} delayed testing or treatment,^{30 , 54} low Apgar scores,⁴⁶ and composite measures including morbidity and mortality.^{32 , 50 , 51} All but six of these studies reported significant associations between capacity strain and outcomes.^{23 , 29 , 32 , 34 , 46 , 52}

Benefits and Harms of Interventions to Improve Care During Times of Capacity Strain

We did not find any randomized studies of interventions to improve care during times of capacity strain; all but one of the eight experimental studies utilized historical controls only⁶⁰ (Table 2). Types of interventions varied greatly and included interventions to increase bed availability,^{55 , 58 , 59 , 62} decrease inefficiency and improve patient flow in busy hospitals,^{57 , 61} coordinate care during mass casualty incidents,⁵⁶ and limit spread of an emerging infectious illness during an epidemic.⁶⁰ Interventions were not associated with improved health outcomes in any studies; seven studies described post-intervention improvements in time-based measures (e.g., hospital LOS, time to surgery), ambulance diversion, or use of non-trauma ICUs for trauma patients.^{55 – 59 , 61 , 62} None of the studies separately analyzed outcomes for children.

Table 2.

Studies Describing Interventions to Improve Care During Times of Capacity Strain

Author, year of publication, country	Study years	Number of patients. Hospital characteristics, Patient diagnoses	Metric to define capacity strain	Intervention	Key results
Bhakta, 2013, USA	2009-2011	529 adults Urban trauma center with 3 trauma ICUs Trauma	Trauma patients boarded in ED or admitted to a non-trauma ICU bed due to no available Trauma ICU bed. Analysis not restricted to times of capacity strain	Implementation of 24/7 open trauma bed protocol in designated trauma ICUs to facilitate rapid admission	No change in mortality, ICU and hospital LOS, and ICU readmissions. Mean ED LOS decreased from 4.2 to 3.1 h, percentage of ICU patients admitted to trauma ICU increased from 83 to 93%
Einav, 2009, Israel	2001-2006	531 adults Urban trauma center Victims of MCIs	MCI (terrorist attack sufficient in size to activate district emergency medical system) and ≥10 casualties or ≥4 severely injured casualties in a brief period of time	Assignment of a physician or nurse case manager to each MCI patient to guide workup, treatment, and transfer decisions	No change in mortality. Several time-based measures improved: Time to first chest x-ray decreased 24 min, time from admission to OR decreased >3 h, and hospital LOS for patients with severe injuries decreased by ≥50%
Howell, 2010, USA	2005-2007	33,721 adults Urban trauma center None specified	Hours on ambulance diversion due to “yellow alert” (ED experiencing temporary overwhelming overload) or “red alert” (no critical care beds available). Analysis not restricted to times of capacity strain	Implementation of active bed management: Hospitalist physician assigned as bed manager, makes triage decisions collaboratively with ED physicians, proactively assesses bed availability	No change in ICU mortality, ICU LOS, ICU admissions after original admission to non-ICU, or ICU readmissions. ED LOS decreased for patients admitted to ICUs. Yellow alert status hours decreased 6% and red alert status hours decreased 27%
Kastrup, 2012, Germany	2008-2011	9,286 adults & children Urban teaching hospital Post-surgical patients needing intensive or intermediate care	None. Analysis not restricted to times of capacity strain	Institution of 24-h staffing of PACU by ICU nurse and in-house critical care physician to allocate postoperative patients to ICU, intermediate care unit, or PACU	No change in mortality for ICU patients. Mean hospital LOS (includes additional post-surgical patients) decreased (8.3 to 7.7 days), mean time from admit to surgery for all operative patients decreased
Lo, 2014, Hong Kong	2009	1,834 adults Regional hospital None specified	None. Analysis not restricted to times of capacity strain	Institution of emergency medicine ward (similar to short-stay unit)	Decrease in mean hospital LOS from 5.2 to 4.1 days. No change in ED LOS
Stukel, 2008, Canada	2000-2003	Not reported 7 urban hospitals in 3 cities 7 common diagnoses	Toronto SARS epidemic (defined by dates). Compared pre-SARS to SARS for Toronto and 2 other cities	Restricted medical admissions during SARS outbreak to limit nosocomial spread	No change in mortality. No systematic change in hospital readmission
Toomath, 2014, New Zealand	2009-2013	49,319 adults Urban regional hospital None specified	None. Analysis not restricted to times of capacity strain	Redesign of medical inpatient service: Geographically based teams, daily patient redistribution, less uneven staffing	Mean hospital LOS decreased from 3.5 to 3.1 days. No change in 7-day readmission rates
Wertheimer, 2014, USA	2011-2013	5,812 adults 2 acute care units in urban academic hospital None specified	None. Analysis not restricted to times of capacity strain	Intervention to increase discharges before noon: Interdisciplinary rounds, pre-discharge checklists, communication to stakeholders, daily leadership meetings, real-time feedback	Decrease in standardized hospital LOS by 10%. No change in 30-day readmission rates

ED: Emergency department. ICU: Intensive care unit. LOS: Length of stay. MCI: Mass casualty incident. OR: Operating room. SARS: Severe acute respiratory syndrome

DISCUSSION

This systematic review found that hospital capacity strain in highly developed countries was associated with increased patient mortality in 9 of 12 studies in ICU settings and in 18 of 30 studies overall. Only 5 of 41 included observational studies did not find a statistically significant association between strain and worsened patient health outcomes or intermediate outcomes.^{15 , 17 , 29 , 38 , 40 , 52} The pediatric literature is very limited, with only four observational studies and no experimental studies separately analyzing outcomes for children.^{23 , 26 , 32 , 45} There was marked heterogeneity in study methods, including the metrics used to define capacity strain, hospital settings, patient populations, outcomes examined, and study quality. We found only eight reports of the health effects of interventions to improve care during times of capacity strain, none of which were randomized studies. Though seven of these eight studies described improvements in process measures, none reported an improvement in patient health outcomes after implementing the intervention.

A key challenge to understanding the health effects of hospital capacity strain is the lack of standard terminology and classification to define strain. Studies included in this review used very different terms to describe capacity strain, from general descriptions (e.g., “strained,”¹⁴ “busy,”¹⁵ or high “workload”^{31 , 32 , 41}) to descriptions based on the number of admitted patients (e.g., “crowding” or “overcrowding,”^{26 , 53} high “census” or “occupancy,”^{17 , 18 , 24 , 35 , 41 , 44 , 47 , 50}) to descriptions based on the number of new patients (e.g., “admission volume”¹⁹). A subset of studies used terms often found in trauma or public health emergency literature (e.g., “surge,”¹¹ “mass casualty incident,”²⁷ or “multiple casualty incident”⁴⁹); others used sequelae of high inpatient census to define strain (e.g., “ED boarding”^{16 , 20 , 21 , 23 , 30 , 38 , 43} or “time from receiving the order for a bed and leaving the ED,”²⁵ “lack of beds”⁴² or “refused admission due to full unit,”³⁶ “delayed admission,”⁴⁰ “boarding”²² or “outlying”⁴² or “bedspacing”⁵² in non-primary units). This lack of standard terminology contributed to differences in how strain was conceptualized by study authors, i.e., whether it was based on changing patient occupancy, acuity, or turnover or whether it was measured indirectly (Online Appendix). Even studies that used similar underlying terminology (e.g., occupancy) varied greatly in strain classification; while some studies examined the effect of occupancy as a continuous variable, others classified strain based on a threshold in percentage occupancy, and others chose occupancy thresholds indexed to a hospital’s or unit’s usual occupancy as reflected by deviation from mean or median occupancy. One study examined strain at the time of ICU discharge, and the finding that ICU and post-ICU length of stay decreased for patients discharged from the ICU during strained times raises the possibility that strain may have led to improved care efficiency, though at the cost of increased ICU readmissions.¹⁵ The resulting heterogeneity in strain classification precluded meta-analysis and limited more detailed qualitative assessments of the relationship between strain and health outcomes. To improve individual study quality and between-study consistency, we suggest that future studies (1) use direct measures of capacity strain based on patient census, acuity, and turnover rather than indirect measures, (2) index strain to usual conditions, such as median and interquartile range, (3) explore non-linear relationships between strain and outcomes, and (4) measure strain over a short enough time interval (minutes to days) to avoid contamination of “strained” vs. “non-strained” groups.

A second key challenge to understanding the relationship between hospital strain and outcomes is the lack of a consistent conceptual approach. Hospitals must consistently provide high-quality care in spite of relatively fixed resources and significant fluctuations in patient needs. An imbalance between patient needs and available resources creates resource strain, but such an imbalance may be due to either increased needs or decreased resources and may be acute or chronic. Chronic resource strain may result from lack of any essential resource and may compromise a hospital’s ability to provide safe care. Adequate staffing may be the single most critical resource to maintaining care quality, and shortfalls in nurse or physician staffing have been linked to increased patient mortality and decreased care efficiency.^{63 – 68} We focused on a subset of resource strain called “capacity strain,” applying a conceptual model that defines the cause of capacity strain as the temporal variation in patient needs as defined by census, acuity, and turnover.⁵ Thus, the purpose behind the concept of capacity strain, and the focus of this review, is not to assess a hospital’s inherent ability to provide high-quality care, but rather its ability to provide high-quality care when patient needs rise. Even within a hospital, individual units may face strain resulting from increased patient needs at different times; it is unclear to what extent strain and its relationship to care quality are localized to specific areas within a hospital and to what extent this localization varies. A coherent theoretical framework for hospital resource strain, which includes both hospital- and patient-driven factors and accounts for both temporal variation in supply and demand as well as the interdependence of different settings (e.g., EDs, inpatient units, nearby hospitals) would provide essential context when planning and interpreting studies of strain.

A third challenge to understanding the relationship between strain and outcomes is that variation in settings is likely to influence the relationship. Some settings may have high baseline levels of strain and thus have less reserve to cope with increases in patient needs. While quality of care may degrade gradually and linearly as patient needs increase, we agree with prior authors that it is more likely that the system is resilient to changing patient needs up to a certain “tipping point,” after which care may degrade rapidly.^{5 , 69} Thus, chronically high hospital census in Denmark (almost 40% of patient days were spent in hospitals with over 100% bed occupancy) may partly explain the association between high occupancy at the time of hospital admission and a 9% increase in mortality.³⁵ It is possible that patient risk is dependent on the timing of capacity strain relative to an individual patient’s hospitalization or even on the total amount of time above a threshold level of strain; these concepts are not well addressed in the existing literature. In addition, more fundamental differences in care delivery systems may affect the relationship between strain and outcomes. For example, the finding by Intas et al. of an almost six-fold increase in hospital mortality among patients with >6 h of ED boarding prior to ICU admission may be due in part to the inability of Greek EDs to initiate ICU treatments at the time of the study,⁴³ and may not be generalizable to settings where ICU-level care is routinely provided in the ED. Lastly, some hospitals may be more resilient than others to the potentially negative effects of strain, which may have contributed to heterogeneity in findings among studies included in this review. While factors influencing such resilience were not reported in these studies, understanding and replicating practices employed by resilient hospitals is an important area of future research.

Limitations

The lack of consistent terminology, classification, and theoretical model in this field makes study selection very challenging. While we consistently applied inclusion and exclusion criteria based on a clear construct and clearly defined outcomes, we acknowledge that no inclusion strategy is perfect when the divisions between related fields are blurred. This may be particularly true for experimental studies; many types of interventions implemented for different reasons (e.g., reorganizing care processes to decrease inefficiencies or improving bed utilization data capture) could lead to improved care quality during times of capacity strain, regardless of their intent. Though almost all included observational studies analyzed the association between strain and outcomes adjusted for patient-level factors, unmeasured differences in patient risk characteristics may have been partly responsible for worsened outcomes during times of strain.

Quality assessment is subject to limitations of the assessment tools. The NOS was not specifically designed to assess risk of bias for hospital-based studies of mortality and thus may not have been an ideal tool; similarly, the Cochrane Risk of Bias Tool was not primarily designed to assess risk of bias in pre-post experimental studies. The significant heterogeneity among included studies, particularly in study settings and strain classification, precluded meta-analysis and limited the ability to draw more precise conclusions regarding the relationship between strain and health outcomes. Lastly, findings of studies performed in highly developed countries may not apply in resource-restricted settings; even in highly developed countries, there may be significant variation in the relationship between strain and outcomes in different hospital settings.

CONCLUSIONS

Hospital capacity strain is likely associated with increased patient mortality and worsening of other health outcomes in highly developed countries, indicating that care quality may degrade during times of strain. There are no interventions that have been shown to improve patient outcomes during times of capacity strain. Understanding the relationship between strain and outcomes is challenged by lack of consistent terminology, classification, and theoretical framework, and by variation in study settings. It is likely that some hospitals are more resilient than others during times of strain, and understanding and replicating practices employed by resilient hospitals are essential to improving care in busy hospitals.

Electronic supplementary material

Below is the link to the electronic supplementary material.

Compliance with Ethical Standards

Conflict of Interest

Carl Eriksson and Craig Newgard received funding from the US Agency for Healthcare Research and Quality (AHRQ) during the conduct of the study. All other authors declare no conflicts of interest.

Funding Source

US Agency for Healthcare Research and Quality, 1 K12 HS022981 01. The funding agency played no role in the design or conduct of this research.