Study Data And Methods
In an earlier study we analyzed data from
799 nonfederal acute care general hospitals in
eleven states. Discharge abstracts and nurse
staffing data were obtained from the states;
data on hospital size, location, teaching status,
from the American Hospital Association
(AHA) annual survey; and cost-to-charge ratios,
from Medicare cost reports.
In regression analyses we found an association
of nurse staffing and (1) lengths-of-stay,
urinary tract infections, upper gastrointestinal
bleeding, hospital-acquired pneumonia,
shock, or cardiac arrest among medical patients
and (2) “failure to rescue,” defined as the
death of a patient with one of five life-threatening
complications—pneumonia, shock or
cardiac arrest, upper gastrointestinal bleeding,
sepsis, or deep vein thrombosis—among surgical
patients. Details of that study are described
elsewhere.3 Exhibit 1 presents rates of these
outcomes and descriptive statistics for the
799-hospital sample.
In this study we simulated the effect of
three options to increase nurse staffing: raise
the proportion of hours provided by registered
nurses (RNs) to the seventy-fifth percentile for
hospitals below this level; raise the number of
licensed (that is, RNs and licensed practical/
vocational nurses, or LPNs) nursing hours per
day to the seventy-fifth percentile; and raise
staffing to each of these levels in hospitals
where each is below the seventy-fifth percentile.
This percentile was chosen based on our
judgment that attaining this level of staffing is
feasible for most hospitals (Exhibit 2).
The required number of additional nurse
hours to meet the seventy-fifth-percentile levels
was estimated from the original sample. Estimates
of avoided adverse outcomes and days
of care were simulated from the regression
models from the earlier study, and estimates of
avoided costs and deaths were made with additional
regression modeling in the original
data. Costs of avoided adverse outcomes were
estimated from patient-level regressions of
costs per case on patient diagnosis and other
characteristics and variables for each adverse
outcome. Costs of avoided days were estimated
by multiplying average costs per day by
regression-based estimates of reduced days net
of the days associated with adverse outcomes.
Because many hospital costs are fixed in the
short run, hospitals might not fully recover the
average costs of avoided days or avoided complications.
Based on a review of studies of hospital
fixed and variable costs, we estimated
variable costs of hospitals to be 40 percent of
average costs, and we multiplied calculated
costs by this amount to estimate the shortterm
cost impact of reduced hospital patient
days and avoided adverse outcomes.4 Over
time, hospitals should be able to adjust their
fixed costs to reflect the change in volume. We
present estimates of cost savings assuming
short-term savings of 40 percent of average
costs and with full recovery of fixed costs.
We projected the results from the sample to
all nonfederal U.S. acute care hospitals and updated
the estimates of needed staffing, avoided
adverse outcomes and days, and costs to reflect
hospital costs, admissions, and lengthsof-stay
in 2002. Specifically, our sample had 26
percent of the discharges from U.S. nonfederal
acute care hospitals in 1997. We constructed
national estimates of adverse outcomes, nursing
full-time equivalents (FTEs), and costs by
multiplying estimates from the sample by 100
divided by 26. We used data on RN wages
from the 1997 and 2002 Current Population
Surveys (CPS) and the change in admissions,
lengths-of-stay, spending per admission, and
spending per day between 2002 and 1997 from
the AHA annual survey to update the estimates
of avoided adverse outcomes, avoided
days, deaths, and costs. In aggregate, between
1997 and 2002, licensed hours per day and the
proportion of licensed hours provided by RNs
reported to the AHA, and average case-mix,
measured by the Medicare case-mix index, did
not change substantially; thus, no adjustments
were made to the staffing variables.5
Because neither our prior work nor other
studies capture all of the effects of nurse staffing
on patient care, and because we do not
have direct measures of patient-reported quality,
we do not attempt a cost-effectiveness
analysis of the impact of raising nurse staffing.
We do present estimates of the cost per
avoided death.