Managing Unnecessary Variability in Patient Demand to Reduce Nursing Stress and Improve Patient Safety

Abstract

Background:
Increases in adverse clinical outcomes have been documented when hospital nurse staffing is inadequate. Since most hospitals limit nurse staffing to levels for average rather than peak patient census, substantial census increases create serious potential stresses for both patients and nurses. By reducing unnecessary variability, hospitals can reduce many of these stresses and thereby improve patient safety and quality of care.

The source and nature of variability in demand:
The variability in the daily patient census is a combination of the natural (uncontrollable) variability contributed by the emergency department and the artificial (potentially controllable) peaks and valleys of patient flow into the hospital fromelective admissions. Once artificial variability in demand is significantly reduced, a substantial portion of the peaks and valleys in census disappears; the remaining censsus variability is largely patient and disease driven. When artificial variability has been minimized, a hospital must have sufficient resources for the remaining patient-driven peaks in demand, over which it has no control, if it is to deliver an optimal level of care.

Discussion:
Study of operational issues in health care delivery, and acting on what is learned, is critical. Al forms of artificial variation in the demand and supply of health care services should be identified, and pilot programs to test operational changes should be conducted.

Full text

330 June 2005 Volume 31 Number 6
Managing Unnecessary Variability
in Patient Demand to Reduce
Nursing Stress and
Improve Patient Safety
Timeliness and Efficiency
Eugene Litvak, Ph.D.
Peter I. Buerhaus, Ph.D, R.N.
Frank Davidoff, M.D.
Michael C. Long, M.D.
Michael L. McManus, M.D., M.P.H.H.
Donald M. Berwick M.D., M.P.P.
T
hree major stresses are intrinsic to all health care
delivery systems.1The first is flow stress, repre-
sented as variability in the appearance rate of
patients for hospital care. This variability routinely pro-
duces peaks and valleys, sometimes extreme, in the
demand for hospital resources. The second is clinical
stress, which is expressed in the variability in type and
severity of disease, and also produces peaks and valleys
in resource demand. Finally, stress can originate from
variability in the professional abilities and competing
responsibilities of health care providers, such as teach-
ing medical and nursing students while at the same time
caring for patients.
An analogy is illustrative. If automobile design is inad-
equate to compensate for the stresses of travel, then
more accidents will occur, and the damage resulting
from any accident is likely to be more severe. In health
care, compelling evidence demonstrates that if
resources are insufficient to tolerate the three stresses
described above, then more medical errors will result.
System stress introduced by demands for nurses to care
for more or sicker patients has been shown to be a lead-
ing cause of adverse patient outcomes.2–5 Preventable
medical errors and excessive mortality rates will contin-
ue to occur, therefore, as long as the health care system
in the United States fails to mitigate the stresses that
contribute to operational dysfunction.
To date, the primary solution has been to spend
enough to provide the resources needed to handle peaks
Background: Increases in adverse clinical outcomes
have been documented when hospital nurse staffing is
inadequate. Since most hospitals limit nurse staffing to
levels for average rather than peak patient census, sub-
stantial census increases create serious potential stress-
es for both patients and nurses. By reducing unnecessary
variability, hospitals can reduce many of these stresses
and thereby improve patient safety and quality of care.
The Source and Nature of Variability in Demand: The
variability in the daily patient census is a combination of
the natural (uncontrollable) variability contributed by
the emergency department and the artificial (potentially
controllable) peaks and valleys of patient flow into the
hospital from elective admissions. Once artificial vari-
ability in demand is significantly reduced, a substantial
portion of the peaks and valleys in census disappears; the
remaining census variability is largely patient and disease
driven. When artificial variability has been minimized, a
hospital must have sufficient resources for the remaining
patient-driven peaks in demand, over which it has no
control, if it is to deliver an optimal level of care.
Discussion: Study of operational issues in health care
delivery, and acting on what is learned, is critical. All
forms of artificial variation in the demand and supply of
health care services should be identified, and pilot pro-
grams to test operational changes should be conducted.
Article-at-a-Glance

331
June 2005 Volume 31 Number 6
in resource demand and ignore the waste that results
during the valleys. This solution is based on the simple
reality that one cannot store the resources accumulated
during the demand valley to use during the next peak.
Mismatch between costly hospital resources and peaks
in demand is thus a major source of reduced quality of
care, nursing dissatisfaction, and reduced access to
care.2,6–9
In this article we examine variability in hospital
patient demand and its relationship to nurse staffing and
patient safety. By smoothing—reducing peaks and val-
leys in—patient demand, hospitals can reduce many of
the stresses needlessly placed on its nursing (and other
staff), and by so doing improve patient safety and quali-
ty of care. We examine the rationale behind long-stand-
ing nurse staffing strategies, explain how they have led
to staffing shortfalls or excesses, and describe the con-
cept “staffing to demand.” We conclude by offering prac-
tical ideas to help clinicians and hospital managers to
exert greater control over patient scheduling and moni-
tor the impact of changes in scheduling on nurse
staffing.
Inefficiency in Nurse Staffing
Undisciplined hospital spending in the 1970s and 1980s,
largely attributable to cost-based reimbursement, affect-
ed many aspects of hospital operations, including nurse
staffing. Many hospitals staffed at levels that ensured
that enough nurses were always on hand, even during
peak patient demand. For example, hospitals frequently
staffed nursing units according to the number of beds,
regardless of whether all were occupied, a strategy that
was feasible and affordable10,11 at a time when an abun-
dance of registered nurses (RNs) was available. During
periods of average or below-average patient demand,
however, nurse staffing in many hospitals was excessive,
although few administrators seemed to know or care
what proportion of nursing resources was wasted or
inappropriately used.
The advent of managed care brought with it efforts to
reduce costly excess capacity and the annual double-
digit cost inflation of the postMedicare era while empha-
sizing efforts to improve the quality of care. By the mid
1990s, the rate of growth in health expenditures slowed
to levels not seen in the previous three decades,12 and the
number of hospitals and beds decreased significantly.
Because a large proportion of hospitals’ operating cost is
labor related, and nursing represents the largest segment
of that cost, hospitals acted to decrease expenditures on
nursing. Consequently, many hospitals began to staff
their nursing units closer to their average demand.
Although these staffing reductions helped to reduce
labor-related costs, an undesirable side effect was that
nursing units were increasingly understaffed during peri-
ods of peak demand. At the same time, however, allow-
ing a deterioration in the quality of care related to
inadequate nurse staffing would only have made matters
worse. These realities continue to the present day, and
nurse executives and senior hospital managers should
pause and consider the following questions:
■
Are all the peaks in the demand pattern for hospital
care inevitable?
■
Are they all patient driven or are they at least partly
the result of uninformed or less than optimal manage-
ment?
■
Does today’s economic reality mean that we cannot
achieve both—adequate nurse staffing and affordable
cost?
The Nursing Shortage
The restrictions in nurse staffing resulting in recent
years from fiscal constraints have been substantially
aggravated by the current and projected shortage of
RNs. Since the 1960s, hospitals in the United States
have reported six periods of RN shortages, but until
the most recent of these, each shortage lasted an aver-
age of one to two years. The current RN shortage
began in 1998 and, despite an increase in hospital
employment of nearly 185,000 RNs from 2001 to 2003,13
continues—in fact, the shortage has now advanced
into its eighth year and continues to take its toll on
nurses, patients, and hospitals. By the late 1990s,
national and state level surveys began reporting that a
majority of RNs were dissatisfied with the hospital
workplace climate; they felt overworked and believed
that inadequate staffing was harming the quality of
care.14 Today, many RNs believe that state or federal
government regulations are the only mechanism that
can raise nurse staffing levels to a point where they are
adequate to meet patient needs.15

332 June 2005 Volume 31 Number 6
A more recent (2002) national survey of 4,000 RNs2
found that roughly two-thirds believed the shortage was
negatively affecting their ability to provide patient
care—by limiting their time to collaborate with the care
team, carry out physician orders, and detect patient
complications early enough to prevent their escalation.
Hospital-employed RNs’ perceptions14,16 that they are tak-
ing care of too many patients and are working double
shifts and overtime more frequently are complemented
by a growing number of studies demonstrating a rela-
tionship between low hospital RN staffing and an
increased risk of adverse patient outcomes.3,4,17–19 For
example, Needleman et al. estimate that the length of
stay for patients treated in hospitals with higher relative
RN staffing is 3%–5% shorter and that complication rates
are 2%–9% lower than in hospitals with lower RN
staffing, controlling for other factors.3
Looking ahead, long-range projections of the RN
workforce suggest that the shortage will only get worse,
as large numbers of baby-boom-generation RNs begin
retiring.21,22 Although many hospitals, policy makers, and
private organizations are working to address problems
in the nursing workforce, current initiatives are unlikely
to avert the large shortage that lies ahead. Thus, health
care organizations and policy makers must anticipate
that future RN shortages will exert a severe and negative
impact on access to care, quality of care, and patient
safety.
It is increasingly clear, therefore, that hospitals will
need to take decisive actions to ensure that RNs are
spending their time wisely and productively, and, above
all, that this time is not wasted. Staffing to demand offers
managers a way to make the best use of their RN nursing
workforce.
Why Staff to Demand?
Although the evidence is still accumulating on the rela-
tionship between inpatient nurse staffing and mortality,
a recent study estimated that for each additional surgical
patient assigned to an RN above the level of adequate
staffing (1:4) the mortality rate increases by 7% for all
patients cared for by that RN.4While recognizing that the
magnitude of the association of mortality and nurse
staffing is not known with precision, we base the calcu-
lations reported here on these results. (Even if this
mortality figure is overestimated by 50%, the increase in
patient mortality rate associated with insufficient nurse
staffing is still significant, at ~3.5%.) Results are reported
in Sidebar 1 above) to illustrate significant effect of vari-
ability in patient demand. (The results may not be gener-
alizable to other individual hospital units or to entire
hospital populations.)
Sidebar 1. Effect of Variability in
Patient Demand
Consider a 200-bed surgical unit with an average cen-
sus of 160. Nurse staffing in this unit is adequate at
40 (1 RN per 4 patients) when demand is average.
Now assume that patients are distributed evenly
among the RNs—4 patients per RN. How will the mor-
tality rate change in this unit with an overall 15%
increase (24 patients) in demand?
If these additional 24 patients are distributed even-
ly among 24 RNs (one additional patient, for a total of
4 + 1 = 5 patients per RN), these 24 nurses will now
be responsible for a total of 120 patients. Because the
mortality rate increases by 7% on average for every
patient added above the optimal staffing ratio of 1:4,*
a 15% increase in unit census can result in an
increase in the mortality rate for 120/184 = 65% of
all patients cared for in this unit (that is, those
patients being cared for by the now-overburdened
RNs). If these 24 additional patients are distributed
evenly among only 12 RNs, then each RN will take
care of 4 + 2 = 6 patients. These 12 RNs will take care
of 72 patients (39% of all patients), whose mortality
rate increases by 14% over the rate in adequately
staffed units.
More generally, for every 5% increase in census
over the adequate staffing level, an additional 20% of
patients will be unnecessarily exposed to a 7% risk of
increased mortality (for details of these calculations,
see Appendix [page 000]). Census increases up to 25%
above an adequate staffing level subject all patients in
the nursing unit in question to the 7% increase in risk;
census increases over 25% result in the addition of
new patients with a 14% increase in mortality rate;
and so on.
* Aiken L.H., et al.: Hospital nurse staffing and patient mortality, nurse
burnout, and job dissatisfaction. JAMA 288:1987–1993, Oct. 23, 2002.

333
June 2005 Volume 31 Number 6
The linkages between nurse understaffing and sen-
tinel events are also well-established; the data indicate
that inadequate numbers of nursing staff contribute to
24% of all sentinel events in hospitals. Inadequate orien-
tation and in-service education of nursing staff are addi-
tional contributing factors in > 70% of sentinel events.9
Taken together, these data suggest that the many
other patient safety measures (for example, computer-
ized data entry) currently being implemented, however
important in their own right, cannot substitute for ade-
quate nurse staffing.
To summarize, a fundamental cause of error and
injury in health care is stress, which appears in turn to
reflect variation in workload. The pursuit of safety and
quality must therefore include understanding and man-
aging variation in workload. Yet it is a mistake to assume
that the best (or the only) way to manage that variation
is with increased resources. In our view, that “quality at
any cost” approach was neither very effective nor effi-
cient in times of plenty; it is bankrupt in times of severe
constraint.
The Source and Nature of Variability
in Demand
Recognizing that variability in demand is intrinsic to
health care delivery, it is useful to focus in greater detail
on the nature of the peaks in resource demand. Data
from many hospitals indicate a high variability in week-
day bed occupancy, with a typical daily deviation of 15%
or more from the mean (and occasionally up to 40%) in
the midnight census, which results in a day-to-day vari-
ability of 30%, even when holidays and weekends (with
their highly predictable changes in census) are not
included. How should the units in these hospitals be
staffed with RNs to provide quality care in the face of
such unpredictable demand?
Three possible scenarios are available: (1) staff con-
tinuously to peak load; (2) staff to average demand and
add additional nurses selectively from dynamic pool of
nurses (for example, on-call nurses) as needed when
census rises above specified levels; or (3) staff constant-
ly to average load. Although all these scenarios have
their strengths and limitations, only the first appears to
both ensure safety and quality. But that scenario is both
unaffordable and wasteful. Are there other alternatives
to these scenarios? To answer that question, consider
the sources of census peaks.
Patients who contribute to the midnight census of a
typical hospital enter in two major ways: (1) through the
emergency room or department (ED)—generally the
source of about 45%–60% of all admissions; and
(2) through scheduling for elective procedures in the
operating room (OR)—generally the source of about
30%–35% of all admissions. The remainder enter through
outpatient referrals and transfers from other hospitals.
Because of its intrinsically variable nature, random
patient arrival to the ED would appear to be the
major source of daily census variability. Conversely,
because elective procedures are scheduled, they
would be expected to create a smooth demand pattern
and be a minor, perhaps nonexistent, source of census
variability.
In fact, these two major sources of demand for hos-
pital beds vary over time by about equal amounts.1,7
Stated more concretely, it is about equally possible
to predict when a patient will come to the ED as it is
to predict when elective surgery will be performed in
the OR. The daily census variability is, therefore, a
combination of both the natural (uncontrollable)
variability contributed by the ED and the artificial
(potentially controllable, hence unnecessary) peaks
and valleys of patient flow into the hospital from
elective scheduled surgery in the OR.1This variation in
the OR elective schedule is one of many examples
of artificial variability that routinely produce severe
operational dysfunction—both waste and stress—in all
health care delivery organizations.
Consider, for example, a hospital that admits 40
scheduled patients one day and another day admits 60
scheduled patients, which is not an unusual degree of
variation. Throughout this time, the hospital budgets
for a constant average of 50 scheduled daily admis-
sions, and routinely staffs for that average number.
Sixty admissions create a significant demand above the
hospital’s staffing capacity; these artificial peaks in
patient demand then become the single major source of
several important “downstream” problems. First, com-
petition for scarce hospital resources between sched-
uled admissions and those that arrive through the ED
leads to ED overcrowding, and, at times, ambulance

334 June 2005 Volume 31 Number 6
diversion and boarding of patients awaiting admission.
Second, these additional admissions lead to nursing
overload and understaffing, increased risk of medical
errors, and a stressful work environment. Finally, the
resultant bottleneck in patient flow will likely reduce
hospital revenue.
Would a “good” staffing ratio eliminate the effect of
artificial variability on nurse workload and quality of
care? Caught between the current fiscal realities that
make staffing to peak loads infeasible and the clinical
realities that staffing below peak census leads to
adverse clinical outcomes and worsening staff morale,
the problem of appropriate nurse staffing in hospitals
might seem to be insoluble. Fortunately, that is not
the case.
The obvious alternative to staffing to peak demand is
eliminating, or at least minimizing, artificial (that is, con-
trollable, hence unnecessary) variability in demand.
Once artificial variability in demand is significantly
reduced, a substantial portion of the peaks and valleys in
census disappears; the remaining census variability
(which is referred to as natural)1is then largely patient
and disease driven. Operations management techniques,
such as queuing theory, can help manage the type of nat-
ural variability seen in ED admissions more effectively
(see, for example, McManus M.C., et al.23). In addition,
addressing variability in demand puts hospitals in a bet-
ter position to ensure the necessary capacity to meet
increases in demand, routinely and cost-effectively,
when they do occur. It is important to mention that arti-
ficial variability is also present in scheduling hospital
discharges, which, combined with artificial variability in
admissions, creates a substantial additional burden on
nurses.
When artificial variability has been minimized, two
scenarios are then possible for any given hospital: (1) the
available resources may turn out to be sufficient to staff
to the remaining, and smaller, natural peaks in demand;
or (2) additional resources may be needed to staff to the
remaining natural peaks in demand. In either case, a hos-
pital must have sufficient resources to staff to the
remaining patient-driven peaks in demand, over which it
has no control, if it is to deliver an optimal level of care.
Any attempt to staff below the patient driven (uncon-
trollable) peaks in demand would lead to nurse under-
staffing, with all its attendant negative consequences. To
ascertain the amount of resources needed for patient-
driven demand, however, it is first necessary to eliminate
(or significantly reduce) the artificial peaks in demand
created by a broken, irrational scheduling system.
Reducing these peaks would reduce artificial demand
for nursing resources.
Is It Feasible to Eliminate or
Significantly Reduce Artificial
Variability?
In most hospitals, surgeons and/or surgical specialty
groups are assigned fixed time intervals (blocks) when
ORs “belong” to them. Underuse of blocks result in val-
leys in surgical case volume; when surgeons extend their
OR use beyond their blocks, the result is peaks in vol-
ume. Proper assignment and use of block times are the
keys to smoothing elective surgical case volume, which,
in turn is a major determinant of artificial census vari-
ability and, hence, nursing stress. Yet changing block
times is a very sensitive issue, because it raises practical
concerns and questions of status and control. It is par-
ticularly important, therefore, for decisions about block
time scheduling to be data driven rather than emotional-
ly determined. Surgeons have other responsibilities:
office hours, teaching, staff meetings, and the like.
Irresponsible redistribution of block times is likely to
interfere with surgeons’ professional lives, and can
result in substantial backlash. Given the complexity and
sensitivity of OR scheduling and uncertainty about the
factors that actually determine overall hospital patient
flow, the feasibility and efficacy of reducing artificial
hospital census variability by rearranging OR schedules
have, until recently, remained open to question. Some
encouraging information on these important points has
begun to emerge from at least one hospital system, how-
ever, as documented in the case study in Sidebar 2 (page
342–343).
Discussion
The need to protect health care organizations from the
stresses introduced by both natural (uncontrollable)
and artificial (potentially controllable) variability means
that we are in effect paying two types of dollars to
provide a “financial bumper” to protect our health care

335
June 2005 Volume 31 Number 6
Boston Medical Center is the city’s “safety net hospital”;
the 547-bed facility is New England’s largest trauma
center. Its emergency department (ED) treats nearly
120,000 patients per year, including many who have
been shot or seriously injured in auto accidents. The
elective surgery schedule in the Merino Pavilion, where
trauma cases are done, was plagued by a 20% cancella-
tion rate and 15 to 20 add-ons per day. Moreover, the
hospital’s patient flow manager, Janet Gorman, noted
that the surgical stepdown unit had a chronic problem
with a lack of beds, especially on Wednesdays and
Thursdays. As a result, patients coming to the unit from
the OR were competing with patients coming from the
surgical ICU.
Taking the First Steps
BMC’s surgical scheduling project started with vascular
surgery. Gorman noticed that the vascular service did
their elective cases in batches—for example, they might
do four cases on one day and none the next. John B.
Chessare, BMC’s chief medical officer and senior vice
president for medical affairs, worked out a plan with the
chief of vascular surgery, James Menzoian, to cap the
number of elective vascular surgery patients going to
the stepdown unit at two per day. In exchange, the vas-
cular surgeons were offered more OR time on Mondays
and Fridays and a guarantee that their cases would
never be bumped. The results have included a smoother
flow of cases and a decrease in the stepdown unit’s
nursing hours per patient day by about 0.5 hours. “The
reason is that during the peaks, the unit has to call in
extra staff and pay overtime. And during the valleys, the
staff has idle time,” according to Dr. Chessare. “So by
getting rid of the stress, you can reduce cost significant-
ly and get more cases on the schedule.”
Spreading the Changes
The next project was cardiac surgery, for which elective
cases peaked in the middle of the week. Although the
chief of cardiothoracic surgery thought that emergency
cases were the problem, data showed that there
appeared to be an equal probability that an emergent
case will come on any day of the week. It was the
scheduled cases that were clearly responsible for the
peaks; in fact, the number of elective cases scheduled
on Tuesdays, Wednesdays, and Thursdays was almost
twice the number on Mondays and Fridays.
To smooth the schedule, the team asked one of the
cardiac surgeons to change his clinic day from Friday to
Wednesday and to do his elective cases on Friday
instead of Wednesday. The two projects combined—
smoothing the vascular and cardiac surgery schedules—
reduced variability in the surgical stepdown unit by
55%; nursing costs in that unit fell by an annualized
amount of $130,000.
Building on success, the team then decided to tackle
the schedule in the Merino Pavilion, which has eight ORs
and an annual volume of about 6,600 procedures. The
high cancellation rate for elective procedures in these
ORs had become a difficult problem for all concerned,
particularly for the elective surgery patients. As Dr.
Chessare stated, someone can walk into the room on the
morning of a woman who has waited three weeks for
her elective gynecological surgery and tell her, “We’re
sorry, but we have to cancel your case because we’ve
had three bad car wrecks.”
By separating urgent/emergent surgical flow from its
elective admissions, the OR schedule becomes much
more predictable.1 This separation is accomplished by
setting aside dedicated ORs each day for urgent/emer-
gent cases.
The first step for the planning team in this area was
to reach consensus on a definition of urgent and emer-
gent cases. Up to that point, an “urgent” case had
sometimes been more for a surgeon’s convenience than
for the patient’s medical condition. The team agreed on
the following definitions for when surgery must be
done:
■Emergent: within 30 minutes
■Urgent: 30 minutes to four hours
■Semi-urgent: 4 to 24 hours
■Non-urgent: over 24 hours
Cases in the first three categories would be done in the
dedicated urgent-emergent room.
After tracking data for several months, the team
found they had a choice of setting aside either one or
Sidebar 2. The Experience of Boston Medical Center*

336 June 2005 Volume 31 Number 6
“car” from damaging its precious passengers. Today, we
spend “necessary dollars” to accommodate inherent
natural variability, and “unnecessary dollars” to deal
with artificial variability. Without rigorous research, it is
difficult to know which expenditure is currently greater.
Until we can distinguish necessary from unnecessary
dollars, any broad-brush attempts to reduce overall
spending in health care delivery will almost certainly
reduce quality because necessary spending will be con-
strained along with unnecessary spending. Fortunately,
new methodologies are becoming available that make it
possible to measure and distinguish appropriate spend-
ing for these two types of variability, although these
techniques are just beginning to be applied to health
care operations.1
Technologically sophisticated support systems
such as electronic order entry and verification have
great potential to reduce medical errors. Once these are
widely implemented, our health care car will
have many sophisticated technical features that will help
prevent clinical damage. Because peaks in resource
demand are inevitable, however, preventable increases
in operational dysfunction and, hence, preventable
adverse patient outcomes will continue to occur unless
we learn how to free up necessary dollars by minimizing
the spending of unnecessary ones.
two urgent-emergent rooms daily. With one room so
assigned, they would occasionally have to bump an elec-
tive case; with two rooms assigned, they would never
have to cancel an elective case, but the second OR
would stand idle for a significant amount of time. They
decided to set aside one room.
“Blowing Up” the Schedule
When the BMC team presented the plan for dedicating
one OR solely to urgent-emergent cases to the Pavilion’s
chiefs of surgery and anesthesia, it was surprised when
the chiefs responded, “As long as we are going to have to
take a block away from someone, even though we know
they’ll be better off, why don’t we just ‘blow up’ block
scheduling?” The planning team and the chiefs promised
the surgeons that if the results were not as expected, the
schedule would return to the previous block system. The
changes in OR scheduling were made in April 2004. The
Pavilion now has one room for urgent-emergent cases,
five “open” rooms, and two rooms still blocked for ortho-
pedics (they are used at 100% capacity, and orthopedic
surgeons do their own bumping because of a lack of sur-
geons not a lack of ORs).
Although the volume of emergent surgical cases was
comparable for the April–September period for 2003
(157 cases) and 2004 (159 cases), delays and cancella-
tions were 99.5% lower in 2004, and only three elective
cases were cancelled (versus 334 cancellations in 2003).
“We’ve also saved the cost of human time, angst, over-
time, and the effort to reschedule all of those delayed
cases,” according to Dr. Chessare. Is the general message
that you’re better off without block scheduling? Dr.
Chessare responds: “You are better off with scientific
management. The problem with blocks is that when you
cut up the time into small segments, you lose flexibility.
When you leave the time open, you gain flexibility. If the
goal is to get more cases done, and to make it easier for
surgeons to get their cases done, blocks actually make it
somewhat harder. On the other hand, if blocks are fully
utilized, and cases aren’t being constantly bumped,
blocks may work fine.”
In practice, surgeons at BMC who maximize the use
of their blocks generally still have their cases scheduled
in the same time frames as before, but they don’t “own”
a block. “It’s not in the hospital’s interest either to have
a surgeon do a case on Monday morning and three cases
on Tuesday afternoon,” Dr. Chessare points out. But
because surgeons don’t own blocks, the hospital is free
to schedule into that time if it’s not used.
In parallel with the improved overall inflow through
the ORs that has resulted from this new approach to
operative scheduling, BMC shaved its average wait time
in the ED from 60 to 40 minutes and shorted its ED
throughput time by 45 minutes—effects that are pre-
sumably related at least in part to the changes in sched-
uling of elective surgical cases.
Sidebar 2. The Experience of Boston Medical Center,
continued
* Some of the material in this case study was used, with permission, from “Boston hospital sees big impact from smoothing elective schedule,” OR Manager
20:1–5, Dec. 2004.

337
June 2005 Volume 31 Number 6
In the present irrational health care environment, the
societal and managerial options are both clear and limit-
ed. Our options are as follows:
■
Start aggressive research to distinguish necessary
from unnecessary dollars in health care, minimize or
eliminate the spending of unnecessary dollars, and use
the resources released to improve protection against
natural variability.
■
Ignore waste and readopt a policy of, as we would call
it, “quality at any cost,” leading to another spiral of
health care inflation.
■
Ignore variability and continue to control overall cost
(the sum of necessary and unnecessary dollars). This
will almost certainly lead to more medical errors and
decreased quality of care.
The wise option is clearly the first: undertaking seri-
ous and sophisticated study of operational issues in
health care delivery, and acting on what we learn.
Essential components of such a program would include
the following:
■
An aggressive research agenda to identify all forms of
artificial (potentially controllable) variation in the
demand and supply of health care services
■
Pilot programs to test operational changes designed
to reduce system stress and improve flow and efficiency
(and thereby improve safety), based on the research
findings
■
Over the long term, the development of closer collab-
oration and two-way communication between opera-
tions researchers and the physicians in clinical care and
health care administration
The result will be an increase in the ability to make
optimal management decisions, that is, decisions that
simultaneously decrease overall cost and increase quality.
Attempts to improve the health care system by addressing
the highly interdependent issues of quality and operations
management separately are unlikely to be productive;
combining them would be a powerful strategy.
Research into the relationship between optimal
management and quality has been ongoing in many
industries and for many years yet is curiously deficient
in health care delivery. Providers or hospitals alone do
not have the necessary resources, and many would
rather return to the good old days of “quality at any
cost.” Managed care organizations generally absolve
themselves of responsibility or interest in addressing
these issues. Thus far, the U.S. federal government has
not invested adequately to meet the challenge.
Successful implementation of such a sophisticated
program of research and action is therefore likely to
require a combination of federal and private funding,
with collaboration among insurers, employers, inte-
grated health care systems, and providers. If we fail
to adopt such a research and action strategy, we are
likely to experience not only an unsustainable increase
in hospital-related health care spending but an
increase in preventable in-hospital medical errors.
The resultant further decline in the quality and safety
of the care we deliver to our patients would be all the
more lamentable because it almost certainly could be
prevented.
J
Eugene Litvak, Ph.D., is Director, Program for Management
of Variability in Health Care Delivery (MVP), and Professor
of Health Care and Operations Management, Boston
University, Boston. Peter I. Buerhaus, Ph.D, R.N., is Senior
Associate Dean for Research, Vanderbilt University School
of Nursing, Nashville, Tennessee. Frank Davidoff, M.D., is
Executive Editor, Institute for Healthcare Improvement
(IHI), Boston. Michael C. Long, M.D., and Michael L.
McManus, M.D., M.P.H.H., are Adjunct Associate
Professors of Operations Management, Boston University.
Donald M. Berwick, M.D., is President and Chief Executive
Officer, Institute for Healthcare Improvement, Cambridge,
Massachusetts. Please address requests for reprints to
Eugene Litvak, Ph.D., litvak@bu.edu.

338 June 2005 Volume 31 Number 6
1. Litvak E., Long M.C.: Cost and quality under managed care:
Irreconcilable differences? Am J Managed Care 6:305–312, Mar. 2000.
2. Nurse Week and American Organization of Nurse Executives,
National Survey of Registered Nurses, 2002: Highlights.
http://www.nurseweek.com/survey/summary.asp (last accessed Apr.
20, 2005).
3. Needleman J., et al.: Nurse-staffing levels and quality of care in hos-
pitals in the United States. N Engl J Med 346:1715–1722, May 30, 2002.
4. Aiken L.H., et al.: Hospital nurse staffing and patient mortality, nurse
burnout, and job dissatisfaction. JAMA 288:1987–1993, Oct. 23–30, 2002.
5. Berens M.J.: Nursing mistakes kill, injure thousands. Cost cutting
exacts toll on patients, hospital staffs. Chicago Tribune, Sep. 10, 2000,
p. 20.
6. Litvak E., et al.: Emergency room diversion: Causes and solutions.
Acad Emergency Med 8:1108–1110, Nov. 2001.
7. McManus M.L., et al.: Variability in surgical caseload and access to
intensive care services. Anesthesiol 98:1491–1496, Jan. 2003.
8. United States General Accounting Office: Hospital Emergency
Departments: Crowded Conditions Vary Among Hospitals and
Communities. Washington, D.C.: United States General Accounting
Office, 2003.
9. Personal communication between the author [E.L.] and Dennis S.
O’Leary, president, Joint Commission on Accreditation of Healthcare
Organizations, Oakbrook Terrace, Illinois, Jul. 28, 2004.
10. Buerhaus P.: Economics of managed competition and conse-
quences to nurses: Part I. Nurs Econ 12:10–17, Jan.–Feb., 1994.
11. Buerhaus P.: Economics and reform: Forces affecting nurse
staffing. Nursing Policy Forum 1(2):8–14, 1995.
12. Levit K., et al.: Health spending in 1998: Signals of change. The
Health Accounts Team. Health Aff (Millwood)19:124–132, Jan.–Feb.,
2000.
13. Buerhaus P., Staiger D., Auerbach D.: Is the current shortage
of hospital nurses ending? Health Aff (Millwood) 22:191–198,
Nov.–Dec. 2003.
14. The Kaiser Family Foundation and Harvard School of
Public Health: Survey of Physicians and Nurses, 1999:
http://www.kff.org/kaiserpolls/1503-index.cfm (last accessed Apr. 20,
2004).
15. Seago J.A.: The California experiment: Alternatives for minimum
nurse-to-patient ratios. J Nurs Adm 32:48–58, Jan. 2002.
16. Buerhaus P., et al.: State of the oncology nursing workforce:
Problems and implications for strengthening the future. Nurs Econ
19:198–208, Sep.–Oct. 2001.
17. Kovner C., Gergen P.J.: Nurse staffing levels and adverse events fol-
lowing surgery in U.S. hospitals. Image J Nurs Sch 30(4):315–321, 1998.
18. Lichtig L., Knauf R., Milholland K.: Some impacts of nursing on
acute care hospital outcomes. J Nurs Adm 29:25–33, Feb. 1999.
19. Unruh L.: Licensed nurse staffing and adverse events in hospitals.
Med Care 41:142–152, Jan. 2003.
20. First Consulting Group, American Hospital Association: The
Healthcare Workforce Shortage and Its Implications for America’s
Hospitals, Fall 2001. http://www.hospitalconnect.com/aha/key_issues/
workforce/resources/Content/FcgWorkforceReport.pdf (last accessed
Apr. 20, 2005).
21. Buerhaus P., Staiger D., Auerbach D.: Implications of a rapidly aging
registered nurse workforce. JAMA 283:2948–2954, Jun. 14, 2000.
22. Health Resources and Services Administration (HRSA): Projected
Supply, Demand, and Shortages of Registered Nurses: 2000–2020.
Washington, D.C.: HRSA, Bureau of Health Professions, National
Center for Health Workforce Analysis, 2002.
23. McManus M.C., et al.: Queuing theory accurately models the need
for critical care resources. Anesthesiol 100:1271–1276, May 2004.
References
In general, if a surgical unit consists of Ppatients with
adequate staffing of 4 patients per nurse, the number
of nurses in this unit is fixed and equals 0.25•P.
Suppose that xis a proportion of patients that have
been added to P (xcould be 5%, 20%, etc.), then the
total number of patients in this unit will be P•(1 + x).
According to Aiken et al.,* in this case the number of
patients with 7% increase in mortality rate will be
5•x•P, as the additional x•Ppatients will be distributed
among x•Pnurses: one additional patient per nurse.
Then, given that the total number of patients has
become P•(1 + x), the proportion of patients in this unit
who are subjected to an 7% increase in mortality rate
will be 5•x•P/P•(1 + x) = 5•x/(1 + x) (in order for this
ratio to be < 1, x should not be > .25). This ratio means
that for every 5% increase in census over the adequate
staffing level, 20% more patients will be exposed to an
overall average increase in mortality of 7%. A census
increase in 25% over adequate staffing level would
result in all patients in that unit being subjected to a
7% increase in mortality rate. Census increases > 25%
would result in adding new patients with a 14%
increase in mortality rate, and so on.
* Aiken L.H., et al.: Hospital nurse staffing and patient mortality, nurse
burnout, and job dissatisfaction. JAMA 288:1987–1993, Oct. 23, 2002.
Appendix 1. Estimating the Effect of Increasing Census on Mortality Rate