Henry Thomas Stelfox, David W. Bates, Donald A. Redelmeier. Safety of Patients Isolated for Infection Control. JAMA. 2003;290(14):1899–1905. doi:10.1001/jama.290.14.1899
Author Affiliations: Department of Medicine (Drs Stelfox and Redelmeier) and Department of Health Policy, Management and Evaluation (Dr Redelmeier), University of Toronto, and Clinical Epidemiology Unit, Sunnybrook and Women's College Health Sciences Centre (Dr Redelmeier), Toronto, Ontario; Graduate School of Arts and Sciences, Harvard University, Cambridge, Mass (Dr Stelfox); and Division of General Medicine and Department of Medicine, Brigham and Women's Hospital and Harvard Medical School (Drs Stelfox and Bates), and Partners HealthCare Systems (Dr Bates), Boston, Mass.
Context Hospital infection control policies that use patient isolation prevent
nosocomial transmission of infectious diseases, but may inadvertently lead
to patient neglect and errors.
Objective To examine the quality of medical care received by patients isolated
for infection control.
Design, Setting, and Patients We identified consecutive adults who were isolated for methicillin-resistant Staphylococcus aureus colonization or infection at 2 large
North American teaching hospitals: a general cohort (patients admitted with
all diagnoses between January 1, 1999, and January 1, 2000; n = 78); and a
disease-specific cohort (patients admitted with a diagnosis of congestive
heart failure between January 1, 1999, and July 1, 2002; n = 72). Two matched
controls were selected for each isolated patient (n = 156 general cohort controls
and n = 144 disease-specific cohort controls).
Main Outcome Measures Quality-of-care measures encompassing processes, outcomes, and satisfaction.
Adjustments for study cohort and patient demographic, hospital, and clinical
characteristics were conducted using multivariable regression.
Results Isolated and control patients generally had similar baseline characteristics;
however, isolated patients were twice as likely as control patients to experience
adverse events during their hospitalization (31 vs 15 adverse events per 1000
days; P<.001). This difference in adverse events
reflected preventable events (20 vs 3 adverse events per 1000 days; P<.001) as opposed to nonpreventable events (11 vs 12
adverse events per 1000 days; P = .98). Isolated
patients were also more likely to formally complain to the hospital about
their care than control patients (8% vs 1%; P<.001),
to have their vital signs not recorded as ordered (51% vs 31%; P<.001), and more likely to have days with no physician progress
note (26% vs 13%; P<.001). No differences in hospital
mortality were observed for the 2 groups (17% vs 10%; P = .16).
Conclusion Compared with controls, patients isolated for infection control precautions
experience more preventable adverse events, express greater dissatisfaction
with their treatment, and have less documented care.
Patient safety has emerged as an important health care issue because
of the consequences of iatrogenic injuries.1 Preventable
injuries result largely from system failures, not from individual inadequacy.2- 5 Human
factors research in nonmedical settings (eg, aviation) suggests that people
tend to take the path of least effort; hence, demanding greater vigilance
from providers of medical care may not result in meaningful safety improvement.2,3 Instead, redesigning faulty systems
appears to be a more promising way to reduce human error.1
The infection control technique of patient isolation may be a system
change that predisposes patients to errors and adverse events.6- 8 Such
strategies, sometimes referred to as transmission-based precautions, are intended
to prevent the spread of pathogens by airborne, droplet, or contact transmission.
The recommended precautions depend on the infectious agent but typically involve
placing the patient in a private room, requiring visitors to wear protective
apparel (eg, gloves, gowns, and masks), and restricting the movement of the
patient outside of the room.9 Infection control
authorities view patient isolation as an important tool for management of
selected established (eg, vancomycin-resistant enterococcus) and emerging
(eg, severe acute respiratory syndrome) infectious diseases.9- 14
Critics of isolation policies have raised questions about quality of
care and whether isolated patients receive less attention.6- 8,15,16 In
1 medical intensive care unit, isolated patients had half as many hourly contacts
with clinicians as control patients.15 In another
study, isolated patients were less likely than other patients to be examined
by physicians during rounds.16 Inevitably,
isolation policies place physical barriers between clinicians and patients,
are time consuming, and impede visitors. However, less contact is not an indication
of inadequate care, and data are lacking regarding the quality of care received
by isolated patients. To examine the safety of isolating patients for infection
control, we conducted a study at 2 teaching hospitals, 1 in Canada and 1 in
the United States.
We identified consecutive adults admitted to Sunnybrook and Women's
College Health Sciences Centre (Toronto, Ontario) between January 1, 1999,
and January 1, 2000, who were isolated for at least 2 days because of methicillin-resistant Staphylococcus aureus (MRSA) colonization or infection.
Control patients were selected by identifying the 2 patients who occupied
each isolated patient's hospital bed immediately before and after his/her
admission. Isolated patients are typically treated in regular single-bed rooms
at this hospital. Matching by hospital bed ensured that isolated and control
patients were treated by similar clinicians, located within similar proximity
to the nursing station, and cared for at similar times of the year. This cohort
of patients was selected to evaluate isolation policies in a general teaching
We also selected a second, disease-specific cohort of patients by identifying
consecutive adults admitted to Brigham and Women's Hospital (Boston, Mass)
between January 1, 1999, and July 1, 2002, who had a principal admitting diagnosis
of congestive heart failure, a previously recorded isolate of MRSA, and an
admission order for isolation. Control patients were selected by identifying
the 2 patients who were admitted with a diagnosis of congestive heart failure
immediately before and after the isolated patient. Matching by primary diagnosis
ensured that isolated and control patients had similar indications for admission,
were treated by similar clinicians, and were cared for at similar times of
the year. This cohort of patients was selected to evaluate isolation policies
in patients admitted with a common cardiac disorder for which established
standards of care are available.
Isolation precaution policies for MRSA (contact precautions) at both
study institutions are based on the most recent recommendations from the Centers
for Disease Control and Prevention.9 The precautions
are designed to prevent transmission through both direct (person-to-person)
contact and indirect (via environmental surface) contact. Patients are cared
for in private rooms, visitors are required to wear gloves and gowns, patient
movement from the room is limited to essential purposes, and dedicated equipment
(eg, stethoscope and blood pressure cuff) is used for each patient.
Each patient's chart was reviewed for demographic, hospital, and clinical
data. Median household income was imputed on the basis of principal residence
using geocoding (postal codes and ZIP codes) and census data (1996 Canadian
Census and 2000 US Census).17 Patient comorbidities
were summarized using the Charlson comorbidity index18 and
illness severity was measured by Acute Physiology and Chronic Health Evaluation
(APACHE) II scores.19 Treatment preferences
were summarized by recording do-not-resuscitate orders issued within the first
2 days of admission. The most recent ejection fraction was recorded for patients
admitted with congestive heart failure.
Documentation of patient vital signs and clinicians' narrative notes
were recorded as general process-of-care measures and markers for thoroughness
of care.20- 22 These
measures were selected because they are appropriate for patients with a broad
range of diagnoses. The following disease-specific inpatient process-of-care
measures were recorded for patients with congestive heart failure: inpatient
evaluation of left ventricular function, inpatient ischemia evaluation (eg,
stress test or coronary angiogram), documentation of daily weight, efforts
toward heart failure education, timely (within 4 weeks) follow-up appointment
scheduled at discharge, and admission and discharge cardiovascular medications.23,24
Adverse events were defined as injuries caused by medical management.
We included injuries that prolonged the hospital stay or produced disability
(the definition used for the Harvard Medical Practice Study),25 as
well as injuries that resulted in transient disability or abnormal laboratory
value measurements (which would not have met those criteria). We applied this
more inclusive definition to capture events that would be clinically significant
yet might not have met the standard for malpractice.
A trained medical record analyst abstracted each patient's hospitalization
into a 1-page summary (excluding mention of MRSA or isolation procedures).
Two independent physicians, blinded to isolation status, reviewed each summary
for adverse events.25 Clinical reviewers were
also asked to grade the severity of any injury (on a 6-point scale ranging
from a single day of symptoms to death) and rate their confidence in the preventability
of the adverse event (on a 6-point scale ranging from no evidence to virtual
certain evidence of preventability). A third reviewer resolved discrepancies.
The adequacy of reviewer blinding was not tested.
Evidence of dissatisfaction with hospital care was ascertained from
2 sources. First, a structured implicit review of each medical record was
performed using the method of Nettleman and Nelson.26 Documented
evidence of dissatisfaction included patients leaving against medical advice,
recorded complaints about medical care, attempted suicide, and altercations.
Second, files from the public relations offices at both institutions were
reviewed for unsolicited complaints. Narratives were coded for specific complaints
using a standardized set of codes comprising 6 categories: perceptions of
treatment, access to staff, communication, humaneness of staff, environmental
cleanliness, and billing.27- 29
The primary analysis tested associations between patients' isolation
status and all of the quality-of-care measures. Baseline patient characteristics
were compared using t tests and χ2 tests
and the Fisher exact test for outcomes with rare events. A team of 6 medical
record reviewers tested the reliability of the measurement instruments. All
medical records were reviewed and abstracted by a trained medical record analyst
using a standard data abstraction tool. One other reviewer independently evaluated
a 10% random sample of all medical records. Four physician-reviewers provided
judgments on adverse events. Agreement was assessed with Cohen κ reliability
Linear, logistic, and Poisson regression analyses were used to test
for differences between isolated and control patients after adjustment for
study cohort and patient demographic characteristics (age, sex, race/ethnicity,
primary language, household income), hospital characteristics (mode of hospital
arrival, admitting service, ward, admission day), and clinical characteristics
(health habits, institutional status, individual comorbidities, Charlson comorbidity
index score, admitting diagnosis, APACHE II score, do-not-resuscitate status).
The data for our general process-of-care measures were longitudinal in nature,
with repeat daily hospital observations clustered within each patient. In
analyzing these data, we adjusted for day-level variables (day of hospital
stay, hospital ward, day of week) in addition to patient-level variables.
To account for the interdependence of these observations, we used robust estimates
of variance (generalized estimating equation).31 Cardiovascular
medications were examined using paired analyses that compared changes in medications
from admission to discharge for isolated and control patients. Statistical
analyses were performed using Stata (version 8.0; Stata Corp, College Station,
Tex), with 2-tailed significance levels of .05.
The ethics committee of Sunnybrook and Women's College Health Sciences
Centre and the institutional review board of Brigham and Women's Hospital
approved the study.
We identified 78 isolated patients and 156 control patients in our general
cohort. In addition, we identified 72 isolated patients and 144 control patients
in our congestive heart failure cohort. The matching process was straightforward
and complete. Four patients had missing nursing notes (1 isolated patient
and 3 control patients) while 1 isolated patient had missing physician notes.
Medical records were available for the remaining 445 patients (99%). The interrater
reliabilities for our measurement instruments ranged from 0.78 to 0.97, and
the κ coefficients for the presence and preventability of adverse events
were 0.72 and 0.40, respectively.
The baseline characteristics of the study groups were similar, with
a few notable exceptions (Table 1).
In the general cohort, there were fewer noninstitutionalized patients in the
isolated group than the control group (P<.001).
In the congestive heart failure cohort there were more cases of diabetes (P = .048), more arrivals by ambulance (P = .03), and generally higher baseline ejection fractions (P = .009) for isolated patients than for controls. All of the isolated
patients in both cohorts were isolated for MRSA during their hospital stay
(144 [96%] colonizations and 6 [4%] infections), while 4 control patients
were temporarily isolated for other infectious disorders (2 vancomycin-resistant
enterococcus infections, 1 Acinobacter infection,
and 1 for infectious diarrhea).
Isolated and control patients had similar numbers of daily vital sign
recordings (Table 2). However,
isolated patients were more likely to have their vital signs incompletely
recorded (14% vs 9%; P<.001) and to have days
with no vital sign recordings (5% vs 1%; P = .02)
at all. A third of the respiratory rates were recorded as exactly 20/min for
both groups. Isolated patients were almost twice as likely to have their vital
signs not recorded as ordered (51% vs 31%; P<.001),
and they were also more likely to have days with no nursing narrative notes
(14% vs 10%; P<.001) or physician progress notes
(26% vs 13%; P<.001) recorded.
Among the patients admitted with congestive heart failure, isolated
and control patients received similar care in the emergency department. They
were equally likely to have an electrocardiogram (85% vs 94%; P = .22), chest radiograph (93% vs 88%; P =
.49), general blood work (100% vs 99%; P = .72),
and cardiac enzyme measurement (69% vs 66%; P = .50)
on arrival. Once admitted to the ward, however, isolated patients were far
less likely to have a stress test or angiogram if they had angina (8/59 [14%]
vs 42/93 [45%]; P<.001), to have their weight
recorded on at least half of the days of the hospitalization (58% vs 87%; P = .01), or to have an evaluation of left ventricular
function while in the hospital (57% vs 69%; P = .049).
Overall, 199 patients with a diagnosis of congestive heart failure were
discharged from the hospital alive. Among these, isolated patients were less
likely to have documentation of congestive heart failure education (18/63
[29%] vs 69/136 [51%]; P = .004) and timely (within
4 weeks) follow-up appointments scheduled (15/63 [24%] vs 63/136 [46%]; P = .001). When cardiovascular medications on admission
(mean number of medications, 4.4 vs 4.2; P = .38)
and discharge (mean number of medications, 4.6 vs 5.0; P = .09) were compared, isolated patients were found to have a smaller
increase in the mean number of medications prescribed than control patients
(+0.2 vs +0.8; P = .02). Specifically, admission-to-discharge
changes in the proportion of patients prescribed angiotensin-converting enzyme
inhibitors (−9% vs +8%; P = .009), digoxin
(−6% vs +7%; P = .045), and diuretics (+3%
vs +11%; P = .03) favored control patients. No significant
differences in medication changes were noted for angiotensin II receptor antagonists,
hydralazine, β-blockers, spironolactone, antiplatelet agents, anticoagulants,
lipid-lowering medications, nitrates, amiodarone, or calcium channel blockers.
Isolated patients had longer hospitalizations and higher rates of adverse
events compared with control patients (Table 3). Overall, there were 161 independent adverse events experienced
by 121 patients; 88 patients with a single adverse event (45 isolated patients
vs 43 control patients), 27 patients with 2 adverse events (22 isolated patients
vs 5 control patients), and 6 patients with 3 or more adverse events (6 isolated
patients vs 0 control patients) (P = .002 for test
of proportions). Isolated patients were twice as likely as control patients
to experience adverse events (31 vs 15 adverse events per 1000 days; P<.001) during their hospital stay. This difference
reflected preventable (20 vs 3 adverse events per 1000 days; P<.001) as opposed to nonpreventable (11 vs 12 adverse events per
1000 days; P = .98) adverse events. Specifically,
isolated patients were 8 times more likely than control patients to experience
supportive care failures such as falls, pressure ulcers, and fluid or electrolyte
disorders; in contrast, no significant differences in diagnostic, operative,
anesthesia, medical procedure, or adverse drug events were noted. The overall
severity of patient injuries from adverse events was similar for both groups
(Table 3), and no differences
in total hospital mortality were observed (26 isolated patients [17%] vs 30
control patients [10%]; odds ratio, 1.69; 95% confidence interval, 0.49-3.21; P = .16).
Isolated patients expressed greater dissatisfaction with their treatment
than control patients (Table 4).
These differences were reflected by both informal and formal complaints. Twelve
isolated patients (8%) submitted unsolicited complaints to the hospital compared
with only 3 control patients (1%). Complaints were associated with negative
perceptions of treatment (5 isolated patients vs 0 control patients), access
to staff (3 isolated patients vs 0 control patients), communication (3 isolated
patients vs 0 control patients), humaneness of staff (0 isolated patients
vs 1 control patient), cleanliness of the environment (1 isolated patient
vs 1 control patient), and billing and payment (0 isolated patients vs 1 control
patient). Only 2 unsolicited compliments were identified, both from control
Our study examined the quality of medical care received by patients
isolated for infection control during hospitalization. The results demonstrate
a strong association between patient isolation and shortfalls of processes,
outcomes, and satisfaction. Isolated patients were less likely than control
patients to have their vital signs accurately recorded, to have daily physician
progress notes documented, and to achieve selected disease-specific standards
of care for heart failure management. Isolated patients also were more likely
to experience a preventable adverse event and to express dissatisfaction with
their care. Hospital mortality rates were similar for both groups.
Patient safety has become an increasingly prominent issue.1 Much
of the attention has focused on medications32 and
surgery,33 yet any medical intervention can
have adverse consequences. Isolation represents one such intervention. A large
body of observational studies supports the effectiveness of isolation policies
in preventing nosocomial infections.10,11 Furthermore,
prominent authorities endorse such procedures for selected patients.9,12,13 However, persistent
concerns remain about the safety of isolation policies because infection control
is only one component of patient safety.6- 8,14- 16
Our study provides some of the strongest data to date on how a systems
change can result in frequent, systematic, and predictable medical errors.
It is unlikely that the lapses in processes, outcomes, and satisfaction documented
in this study were deliberate; hence, our results underscore the importance
of examining any intervention (eg, isolation precautions) for unintended consequences.
In addition, multicomponent interventions (eg, barriers, restricted access,
reduced mobility) should have their individual parts examined to determine
whether all elements are essential. For example, it may be possible to disentangle
which isolation policy components are most important for infection control
and which may be most deleterious to the isolated patient. Finally, the need
for individualization is highlighted because the patients who experienced
the most negative effects from isolation strategies might not necessarily
be those who presented the highest risk of disease transmission. The interdependence
of individual patient characteristics, clinician factors, environmental constraints,
and organizational climate is likely to significantly influence safety.
Our study also underscores some of the challenges of improving patient
safety. First, adverse consequences are not always easily detected; furthermore,
rigorous clinical trials may minimize adverse events because special attention
by research assistants provides an extra layer of safeguard and error interception.
Second, difficult trade-offs may emerge, such as the tension between preventing
injury to other patients (such as reducing nosocomial transmission of pathogens)
and maximizing the well-being of an individual patient (by avoiding barrier
restrictions). Third, because most medical interventions are well intentioned,
faulty systems often arise when a focus on one priority detracts from other
unrelated clinical concerns. Interventions that simplify tasks (such as minimizing
barriers between patients and clincians) and minimize distractions are likely
to improve safety.2,5 Fourth,
while educational or regulatory policies that update clinicians with new information
on an intervention's risks and benefits (such as unintentional discrepancies
in the care of isolated patients) have intuitive appeal, they are unlikely
to have a significant effect on safety.34 Rather,
creative solutions that recognize the limitations of clinicians are badly
needed. Ultimately, the only foolproof way to eliminate medical errors is
by eliminating disease—for example, effective eradication techniques
might allow patients with multidrug-resistant pathogen colonization to avoid
Our study also highlights the limitations of science applied to medical
errors. Similar to most research in this field, our data are quasi-experimental
in nature and raise the question of whether the differences between the 2
study groups are simply a function of illness severity. This concern seems
less worrisome in our study because the patient characteristics of our groups
were remarkably similar. Nevertheless, unmeasured differences in illness severity
in our study may cause our results to underestimate quality-of-care differences
if isolated patients were inherently sicker than control patients because
the standard of care for isolated patients should therefore have been higher.
In addition, our data were obtained through retrospective chart review, as
with most articles on patient safety. Chart review is problematic because
it tends to miss subtle lapses in care.35,36 Finally,
our results are based on patients with MRSA at 2 major North American teaching
hospitals. Patients isolated because of other infectious agents or treated
in smaller hospitals may have different experiences.
In summary, hospital infection control policies may prevent the spread
of communicable infections but may also inadvertently lead to poorer quality
of care and adverse events. Our results illustrate the importance of balancing
the risks and benefits of an intervention while highlighting that mandatory
policies may not always be appropriate. In addition, our findings provide
greater incentives for the eradication of chronic disease states and promote
the use of human factors research to develop safer medical policies. The complexities
of health care are likely to increase in the future, making the detection
of unintended adverse consequences even more difficult. Well-designed, carefully
evaluated, and appropriately implemented interventions will be essential in
ensuring the safety of all patients.