You are on page 1of 7

Confidential. Do not distribute. Pre-embargo material.


Hospital Nurse Staffing and

Patient Mortality, Nurse Burnout,
and Job Dissatisfaction
Linda H. Aiken, PhD, RN Context The worsening hospital nurse shortage and recent California legislation
Sean P. Clarke, PhD, RN mandating minimum hospital patient-to-nurse ratios demand an understanding of
how nurse staffing levels affect patient outcomes and nurse retention in hospital
Douglas M. Sloane, PhD
Julie Sochalski, PhD, RN Objective To determine the association between the patient-to-nurse ratio and pa-
Jeffrey H. Silber, MD, PhD tient mortality, failure-to-rescue (deaths following complications) among surgical pa-
tients, and factors related to nurse retention.

bulent time for US hospitals and Design, Setting, and Participants Cross-sectional analyses of linked data from
10 184 staff nurses surveyed, 232 342 general, orthopedic, and vascular surgery
practicing nurses. News media
patients discharged from the hospital between April 1, 1998, and November 30,
have trumpeted urgent con- 1999, and administrative data from 168 nonfederal adult general hospitals in Penn-
cerns about hospital understaffing and sylvania.
a growing hospital nurse shortage.1-3
Main Outcome Measures Risk-adjusted patient mortality and failure-to-rescue
Nurses nationwide consistently report within 30 days of admission, and nurse-reported job dissatisfaction and job-related
that hospital nurse staffing levels are in- burnout.
adequate to provide safe and effective
Results After adjusting for patient and hospital characteristics (size, teaching status,
care.4-6 Physicians agree, citing inad- and technology), each additional patient per nurse was associated with a 7% (odds
equate nurse staffing as a major impedi- ratio [OR], 1.07; 95% confidence interval [CI], 1.03-1.12) increase in the likelihood
ment to the provision of high-quality of dying within 30 days of admission and a 7% (OR, 1.07; 95% CI, 1.02-1.11) in-
hospital care.7 The shortage of hospital crease in the odds of failure-to-rescue. After adjusting for nurse and hospital charac-
nurses may be linked to unrealistic nurse teristics, each additional patient per nurse was associated with a 23% (OR, 1.23; 95%
workloads.8 Forty percent of hospital CI, 1.13-1.34) increase in the odds of burnout and a 15% (OR, 1.15; 95% CI, 1.07-
nurses have burnout levels that exceed 1.25) increase in the odds of job dissatisfaction.
the norms for health care workers.4 Job Conclusions In hospitals with high patient-to-nurse ratios, surgical patients expe-
dissatisfaction among hospital nurses is rience higher risk-adjusted 30-day mortality and failure-to-rescue rates, and nurses
4 times greater than the average for all are more likely to experience burnout and job dissatisfaction.
US workers, and 1 in 5 hospital nurses JAMA. 2002;288:1987-1993
report that they intend to leave their cur-
rent jobs within a year.4
densome workloads and high levels of Author Affiliations: Center for Health Outcomes and
In 1999, California passed legisla- Policy Research, School of Nursing (Drs Aiken, Clarke,
job-related burnout and job dissatis-
tion mandating patient-to-nurse ra- Sloane, and Sochalski), Leonard Davis Institute of Health
faction. Stakeholder groups advo- Economics (Drs Aiken, Clarke, Sochalski, and Silber), De-
tios for its hospitals, which goes into partment of Sociology (Dr Aiken), Population Studies
cated widely divergent minimum ra-
effect in July 2003. The California leg- Center (Drs Aiken, Sloane, and Sochalski), and Depart-
tios. On medical and surgical units, ments of Pediatrics and Anesthesia, School of Medi-
islation was motivated by an increas-
recommended ratios ranged from 3 cine (Dr Silber), University of Pennsylvania, Philadel-
ing hospital nursing shortage and the phia; and Center for Outcomes Research, Childrens
to 10 patients for each nurse.9-11 In
perception that lower nurse retention Hospital of Philadelphia, Philadelphia, Pa (Dr Silber).
early 2002, Californias governor an- Corresponding Author and Reprints: Linda H. Aiken,
in hospital practice was related to bur-
nounced that hospitals must have at PhD, RN, Center for Health Outcomes and Policy Re-
search, University of Pennsylvania, 420 Guardian Dr,
least 1 licensed nurse for every 6 medi- Philadelphia, PA 19104-6096 (e-mail: laiken@nursing
For editorial comment see p 2040.
cal and surgical patients by July 2003,

2002 American Medical Association. All rights reserved. (Reprinted) JAMA, October 23/30, 2002Vol 288, No. 16 1987
Confidential. Do not distribute. Pre-embargo material.

a ratio that will move to 1 to 5 when agnosis Related Groups (DRGs) dur- or smaller trainee:bed ratios (minor
the mandates are fully implemented.12 ing the study period, as well AHA data, teaching hospitals) and those with ra-
This study reports on findings from a and survey data from 10 or more staff tios that were higher than 1:4 (major
comprehensive study of 168 hospitals nurses. Six of the excluded hospitals teaching hospitals). Finally, hospitals
and clarifies the impact of nurse staff- were Veterans Affairs hospitals, which with facilities for open heart surgery
ing levels on patient outcomes and fac- do not report discharge data to the state. and/or major transplants were classi-
tors that influence nurse retention.13 Spe- Twenty-six hospitals were excluded be- fied as high-technology hospitals and
cifically, we examined whether risk- cause their administrative or patient out- contrasted with other hospitals.19
adjusted surgical mortality and rates of comes data could not be matched to our Nurses and Nurse Outcomes. Sur-
failure-to-rescue (deaths in surgical pa- surveys because of missing variables, pri- veys were mailed in the spring of 1999
tients who develop serious complica- marily because they reported their char- to a 50% random sample of registered
tions) are lower in hospitals where nurses acteristics or patient data as aggregate nurses who were on the Pennsylvania
carry smaller patient loads. In addition, multihospital entities. In 10 additional Board of Nursing rolls and resided in the
we ascertained the extent to which more small hospitals, the majority of which state. The response rate was 52%, which
favorable patient-to-nurse ratios are as- had fewer than 50 beds, fewer than 10 compares favorably with rates seen in
sociated with lower burnout and higher nurses responded to the survey. other voluntary surveys of health pro-
job satisfaction among registered nurses. A nurse staffing measure was calcu- fessionals.20 Roughly one third of the
We also estimated excess surgical deaths lated as the mean patient load across all nurses who responded worked in hos-
associated with the different nurse staff- staff registered nurses who reported pitals and included the sample of 10184
ing ratios vigorously debated in Califor- having responsibility for at least 1 but nurses described here. No special re-
nia. Finally, we estimated the impact of fewer than 20 patients on the last shift cruiting methods or inducements were
nurse staffing levels proposed in Cali- they worked, regardless of the spe- used. Demographic characteristics of the
fornia on nurse burnout and dissatisfac- cialty or shift (day, evening, night) respondents matched the profile for
tion, 2 precursors of turnover.13 Our find- worked. This measure of staffing is Pennsylvania nurses in the National
ings offer insights into how more superior to those derived from admin- Sample Survey of Registered Nurses.21
generous registered nurse staffing might istrative databases, which generally Nurses employed in hospitals were asked
affect patient outcomes and inform cur- include registered nurse positions that to use a list to identify the hospital in
rent debates in many states regarding the do not involve inpatient acute care at which they worked, and then were que-
merits of legislative actions to influence the bedside. Staffing was measured ried about their demographic character-
staffing levels. across entire hospitals because there is istics, work history, workload, job sat-
no evidence that specialty-specific staff- isfaction, and feelings of job-related
METHODS ing offers advantages in the study of burnout. Questionnaires were returned
Patients, Data Sources, patient outcome17 and to reflect the fact by nurses employed at each of the 210
and Variables that patients often receive nursing care Pennsylvania hospitals providing adult
Our study combines information about in multiple specialty areas of a hospi- acute care. To obtain reliable hospital-
hospital staffing and organization ob- tal. Direct measurement also avoided level estimates of nurse staffing (the ra-
tained from nurse surveys with patient problems with missing data common tio of patients to nurses in each hospi-
outcomes derived from hospital dis- to the AHAs Annual Survey of hospi- tal), attention was restricted to registered
charge abstracts and hospital character- tals, which imputed staffing data in 1999 nurses holding staff nurse positions in-
istics drawn from administrative data- for 20% of Pennsylvania hospitals. volving direct patient care and to hos-
bases.14 The study protocol for linking Three hospital characteristics were pitals from which at least 10 such nurses
anonymized nurse data and handling de- used as control variables: size, teaching returned questionnaires. In 80% of the
nominalized patient data was ap- status, and technology. Hospitals were 168 hospitals in the final sample, 20 or
proved by the institutional review board grouped into 3 size categories: small more nurses provided responses to our
of the University of Pennsylvania. (100 hospital beds), medium (101- questionnaire. There were more than 50
Hospitals. Data were collected on all 250 hospital beds), and large (251 hos- nurse respondents from half of the hos-
210 adult general hospitals in Pennsyl- pital beds). Teaching status was mea- pitals. We examined 2 nurse job out-
vania. Information about hospital char- sured by the ratio of resident physicians comes in relation to staffing: job satis-
acteristics was derived from the 1999 and fellows to hospital beds, which has faction (rated on a 4-point scale from very
American Hospital Association (AHA) been suggested as superior to univer- dissatisfied to very satisfied) and burn-
Annual Survey and the 1999 Pennsyl- sity affiliations and association member- out (measured with the Emotional Ex-
vania Department of Health Hospital ships as an indicator of the intensity of haustion scale of the Maslach Burnout
Survey.15,16 Ultimately, 168 of the 210 teaching activity.18 Hospitals with no Inventory, a standardized tool).22,23
acute care hospitals had discharge data postgraduate trainees (nonteaching) Patients and Patient Outcomes. Dis-
for surgical patients in the targeted Di- were contrasted with those that had 1:4 charge abstracts representing all admis-
1988 JAMA, October 23/30, 2002Vol 288, No. 16 (Reprinted) 2002 American Medical Association. All rights reserved.
Confidential. Do not distribute. Pre-embargo material.

sions to nonfederal hospitals in Penn-

sylvania from 1998 to 1999 were obtained Box. Surgical Patient Diagnosis Related Groups Included
from the Pennsylvania Health Care Cost in the Analyses of Mortality and Failure-to-Rescue
Containment Council. These discharge General Surgery
abstracts were merged with Pennsylva-
146-155, 157-162, 164-167, 170, 171, 191-201, 257-268, 285-293, 493, and 494
nia vital statistics records to identify
patients who died within 30 days of hos- Orthopedic Surgery
pital admission to control for timing of 209-211, 213, 216-219, 223-234, 471, 491, and 496-503
discharge as a possible source of varia- Vascular Surgery
tion in hospital outcomes. We exam- 110-114, 119, and 120
ined outcomes for 232 342 patients
between the ages of 20 and 85 years who
underwent general surgical, orthope- ables indicating the presence of chronic of the hospital admission, comorbidi-
dic, or vascular procedures in the 168 preexisting health conditions reflected ties, and relevant interaction terms. For
hospitals from April 1, 1998, to Novem- in the ICD-9-CM codes in the dis- analyses of both patient and nurse out-
ber 30, 1999. Surgical discharges were charge abstracts (eg, diabetes melli- comes, we adjusted for hospital size,
selected for study because of the avail- tus), as well as a series of interaction teaching status, and technology.
ability of well-validated risk adjustment terms. The final set of control variables All logistic regression models were es-
models.24-29 The number of patients dis- was determined by a selection process timated by using Huber-White (ro-
charged from the study hospitals ranged that paralleled an approach used and re- bust) procedures to account for the clus-
from 75 to 7746. Only the first hospital ported previously.27-29 The C statistic tering of patients within hospitals and
admission for any of the DRGs listed in (area under the receiver operating char- adjust the SEs of the parameter esti-
the BOX for any patient during the study acteristic curve) for the mortality risk ad- mates appropriately.31,32 Model calibra-
period was included in the analyses. justment model was 0.89.30 tion was assessed with the Hosmer-
In addition to 30-day mortality, we Lemeshow statistic.33 We used direct
examined failure-to-rescue (deaths Data Analysis standardization to illustrate the magni-
within 30 days of admission among Descriptive data show how patients and tude of the effect of staffing by estimat-
patients who experienced complica- nurses in our sample were distributed ing the difference in the numbers of
tions).24-29 Complications were identi- across the various categories of hospi- deaths and episodes of failure-to-
fied by scanning discharge abstracts for tals defined by staffing levels and other rescue under different staffing sce-
International Classification of Diseases, characteristics. Logistic regression mod- narios. Using all patients in the study and
Ninth Revision, Clinical Modification els were used to estimate the effects of using the final fully-adjusted model, we
(ICD-9-CM) codes in the secondary di- staffing on the nurse outcomes (job dis- estimated the probability of death and
agnosis and procedure fields that were satisfaction and burnout) and 2 patient failure-to-rescue for each patient un-
suggestive of 39 different clinical events. outcomes (mortality and failure-to- der various patient-to-nurse ratios (ie,
Distinguishing complications from pre- rescue). We computed the odds of 4, 6, and 8 patients per nurse) with all
viously existing comorbidities in- nurses being moderately or very dissat- other patient characteristics un-
volved the use of rules developed by ex- isfied with their current positions and changed. We then calculated the differ-
pert consensus and previous empirical reporting a level of emotional exhaus- ences in total deaths under the differ-
work, as well as examination of dis- tion (burnout) above published norms ent scenarios.34 Confidence intervals
charge records for each patients hospi- for medical workers and of patients ex- (CIs) for these direct standardization es-
talizations 90 days before the surgery of periencing mortality and failure-to- timates were derived with the method
interest for overlap in secondary diag- rescue under different levels of regis- described by Agresti.35 All analyses were
nosis codes.27-29 Examples of complica- tered nurse staffing, before and after performed using STATA version 7.0
tions included aspiration pneumonia control for individual characteristics and (STATA Corp, College Station, Tex), and
and hypotension/shock. Patients who hospital variables. For nurse out- P<.05 was considered statistically sig-
died postoperatively were assumed to comes, we adjusted for sex, years of ex- nificant in all analyses.
have developed a complication even if perience in nursing, education (bacca-
no complication codes were identified laureate degree or above vs diploma or RESULTS
in their discharge abstracts. associate degree as highest credential in Characteristics of Hospitals,
Risk adjustment of mortality and fail- nursing), and nursing specialty. For Nurses, and Patients
ure-to-rescue for patient characteris- analyses of patient outcomes, we con- Distributions of hospitals with various
tics and comorbidities was accom- trolled for the variables in our risk ad- characteristics, distributions of nurses
plished by using 133 variables, including justment model, specifically, demo- surveyed, and patients whose out-
age, sex, surgery types, and dummy vari- graphic characteristics of patients, nature comes were studied are shown in
2002 American Medical Association. All rights reserved. (Reprinted) JAMA, October 23/30, 2002Vol 288, No. 16 1989
Confidential. Do not distribute. Pre-embargo material.

plications and were included in our

Table 1. Study Hospitals, Surgical Patients Studied, and Nurse Respondents in Hospitals*
analyses of failure-to-rescue were simi-
No. (%)
lar to the broader group of patients in our
Hospitals Patients Nurses mortality analyses with respect to their
Characteristic (N = 168) (N = 232 342) (N = 10 184)
comorbidities, but orthopedic surgery
Staffing, patients per nurse
4 20 (11.9) 41 414 (17.8) 1741 (17.1) patients were less prominently repre-
5 64 (38.1) 111 752 (48.1) 4818 (47.3) sented among patients with complica-
6 41 (24.4) 48 120 (20.7) 2114 (20.8) tions than in the overall sample.
7 29 (17.3) 21 360 (9.2) 1106 (10.9)
8 14 (8.3) 9696 (4.2) 405 (4.0) Staffing and Job Satisfaction
Size, No. of beds and Burnout
100 41 (24.4) 16 123 (6.9) 842 (8.3)
Higher emotional exhaustion and greater
101-250 95 (56.6) 110 510 (47.6) 4927 (48.4)
job dissatisfaction in nurses were strongly
251 32 (19.1) 105 709 (45.5) 4415 (43.4)
and significantly associated with patient-
Not high 121 (72.0) 103 824 (44.7) 4706 (46.2) to-nurse ratios. TABLE 4 shows odds
High 47 (28.0) 128 518 (55.3) 5478 (53.8) ratios (ORs) indicating how much more
Teaching status likely nurses in hospitals with higher
None 107 (63.7) 98 937 (42.6) 4553 (44.7) patient-to-nurse ratios were to exhibit
Minor 44 (26.2) 80 127 (34.5) 3435 (33.7)
burnout scores above published norms
Major 17 (10.1) 53 278 (22.9) 2196 (21.6)
and to be dissatisfied with their jobs. Con-
*Percentages may not add up to 100 because of rounding.
trolling for nurse and hospital charac-
teristics resulted in a slight increase in
veyed were drawn from the 61 hospi- these ratios, which in both cases indi-
Table 2. Characteristics of Nurses
(N = 10 184) in the Study Hospitals* tals (36.3%) that reported postgradu- cated a pronounced effect of staffing. The
Characteristic No. (%)
ate medical trainees in 1999. final adjusted ORs indicated that an
Women 9425 (94.1)
As shown in TABLE 2, 94.1% of the increase of 1 patient per nurse to a hos-
BSN degree or higher 3980 (39.6) nurses were women and 39.6% held a pitals staffing level increased burnout and
Years worked as a nurse, mean (SD) 13.8 (9.8) baccalaureate degree or higher. The job dissatisfaction by factors of 1.23 (95%
Clinical specialty
Medical and surgical 3158 (31.0) mean (SD) work experience in nurs- CI, 1.13-1.34) and 1.15 (95% CI, 1.07-
Intensive care 1992 (19.6) ing was 13.8 years (9.8). Thirty-one per- 1.25), respectively, or by 23% and 15%.
Operating/recovery room 998 (9.8)
Other 4026 (39.6) cent of the nurses in the sample worked This implies that nurses in hospitals with
High emotional exhaustion 3926 (43.2) on medical and surgical general units, 8:1 patient-to-nurse ratios would be 2.29
Dissatisfied with current job 4162 (41.5)
while 19.6% and 9.8% worked in in- times as likely as nurses with 4:1 patient-
*Sample size for individual characteristics varied be-
cause of missing data. BSN indicates bachelor of sci- tensive care and perioperative set- to-nurse ratios to show high emotional
ence in nursing. High emotional exhaustion refers to lev-
els of emotional exhaustion above the published high
tings, respectively. Forty-three per- exhaustion (ie, 1.23 to the 4th power for
norm for medical workers.20 Dissatisfied with current job cent of the nurses had high burnout 4 additional patients per nurse=2.29) and
combines nurses who reported being either very dis-
satisfied or a little dissatisfied. scores and a similar proportion were 1.75 times as likely to be dissatisfied with
dissatisfied with their current jobs. their jobs (ie, 1.15 to the 4th power for
Of the 232342 patients studied, 53813 4 additional patients per nurse=1.75).
TABLE 1. Fifty percent of the hospitals (23.2%) experienced a major complica- Our data further indicate that, although
had patient-to-nurse ratios that were 5:1 tion not present on admission and 4535 43% of nurses who report high burnout
or lower, and those hospitals dis- (2.0%) died within 30 days of admis- and are dissatisfied with their jobs intend
charged 65.9% of the patients in the sion. The death rate among patients with to leave their current job within the next
study and employed 64.4% of the nurses complications was 8.4%. The surgical 12 months, only 11% of the nurses who
we surveyed. Hospitals with more than case types and clinical characteristics of are not burned out and who remain sat-
250 beds accounted for a disproportion- the patient cohort are shown in TABLE 3. isfied with their jobs intend to leave.
ate share of both patients and nurses Slightly more than half of patients
(45.5% and 43.4%, respectively). Al- (51.2%) were classified in an orthope- Staffing and Patient Mortality
though high-technology hospitals ac- dic surgery DRG, with the next largest and Failure-to-Rescue
counted for only 28.0% of the institu- group of patients (36.4%) undergoing di- Among the surgical patients studied,
tions studied, more than half (55.3%) gestive tract and hepatobiliary surger- there was a pronounced effect of nurse
of the patients discharged and 53.8% ies. Chronic medical conditions, with the staffing on both mortality and mortal-
of nurses surveyed were from high- exception of hypertension, were rela- ity following complications. Table 4 also
technology hospitals. A majority of tively uncommon among these pa- shows the relationship between nurse
the patients studied and nurses sur- tients. Patients who experienced com- staffing and patient mortality and failure-
1990 JAMA, October 23/30, 2002Vol 288, No. 16 (Reprinted) 2002 American Medical Association. All rights reserved.
Confidential. Do not distribute. Pre-embargo material.

to-rescue (mortality following compli- tients and 8.7 (95% CI, 3.9-13.5) addi- than 6 patients per nurse, we would ex-
cations) when other factors were ig- tional deaths per 1000 patients with pect 2.6 (95% CI, 1.2-4.0) additional
nored, after patient characteristics were complications. If the staffing ratio in all deaths per 1000 patients and 9.5 (95%
controlled, and after patient character- hospitals was 8 patients per nurse rather CI, 3.8-15.2) additional deaths per 1000
istics and other hospital characteristics
(size, teaching status, and technology)
were controlled. Although the ORs re- Table 3. Characteristics of the Surgical Patients Included in Analyses of Mortality and
flecting the nurse staffing effect were
No. (%)
somewhat diminished by controlling for
patient and hospital characteristics, they Patients With
All Patients Complications
remained sizable and significant for both Characteristic (N = 232 342) (n = 53 813)
mortality and failure-to-rescue (1.07; Men 101 624 (43.7) 25 619 (47.6)
95% CI, 1.03-1.12 and 1.07; 95% CI, Age, mean (SD) 59.3 (16.9) 64.2 (15.7)
1.02-1.11, respectively). An OR of 1.07 Emergency admissions 63 355 (27.3) 21 541 (40.0)
implies that the odds of patient mortal- Deaths within 30 days of admission 4535 (2.0) 4535 (8.4)
ity increased by 7% for every additional Major Diagnostic Categories (MDCs)
patient in the average nurses workload General surgery
in the hospital and that the difference Diseases and disorders of the 54 919 (23.6) 19 002 (35.3)
digestive system (MDC 6)
from 4 to 6 and from 4 to 8 patients per
Diseases and disorders of the hepatobiliary 29 660 (12.8) 6804 (12.6)
nurse would be accompanied by 14% and system (MDC 7)
31% increases in mortality, respectively Diseases and disorders of the skin, 12 771 (5.5) 3010 (5.6)
(ie, 1.07 to the 2nd power=1.14 and 1.07 subcutaneous tissue, and breast (MDC 9)
to the 4th power=1.31). Endocrine, nutritional, metabolic diseases, 4853 (2.1) 1535 (2.9)
and disorders (MDC 10)
These effects imply that, all else being
Orthopedic surgery
equal, substantial decreases in mortal-
Diseases and disorders of the musculoskeletal 118 945 (51.2) 17 403 (32.3)
ity rates could result from increasing reg- system (MDC 8)
istered nurse staffing, especially for pa- Vascular surgery
tients who develop complications. Direct Diseases and disorders of the circulatory 11 194 (4.8) 6059 (11.3)
standardization techniques were used to system (MDC 5)
predict excess deaths in all patients and Medical history (comorbidities)
Congestive heart failure 11 795 (5.1) 5735 (10.7)
in patients with complications that would
Arrhythmia 3965 (1.7) 1765 (3.3)
be expected if the patient-to-nurse ratio
Aortic stenosis 2248 (1.0) 848 (1.6)
for all patients in the study were at vari-
Hypertension 79 827 (34.4) 20 648 (38.4)
ous levels that figure prominently in the
Cancer 28 558 (12.3) 9074 (16.9)
California staffing mandate debates. If the
Chronic obstructive pulmonary disease 19 819 (8.5) 7612 (14.2)
staffing ratio in all hospitals was 6 pa-
Diabetes mellitus (insulin and noninsulin dependent) 31 385 (13.5) 9597 (17.8)
tients per nurse rather than 4 patients per Insulin-dependent diabetes mellitus 3607 (1.6) 1755 (3.3)
nurse, we would expect 2.3 (95% CI, 1.1- *Patients who died postoperatively were assumed to have developed a complication even if no complication codes
3.5) additional deaths per 1000 pa- were identified in their discharge abstracts.

Table 4. Patient-to-Nurse Ratios With High Emotional Exhaustion and Job Dissatisfaction Among Staff Nurses and With Patient Mortality and
Odds Ratio (95% Confidence Interval)

Adjusted for Adjusted for

P Nurse or Patient P Nurse or Patient and P
Unadjusted Value Characteristics Value Hospital Characteristics Value
Nurse outcomes
High emotional exhaustion 1.17 (1.10-1.26) .001 1.17 (1.10-1.26) .001 1.23 (1.13-1.34) .001
Job dissatisfaction 1.11 (1.03-1.19) .004 1.12 (1.04-1.19) .001 1.15 (1.07-1.25) .001
Patient outcomes
Mortality 1.14 (1.08-1.19) .001 1.09 (1.04-1.13) .001 1.07 (1.03-1.12) .001
Failure-to-rescue 1.11 (1.06-1.17) .004 1.09 (1.04-1.13) .001 1.07 (1.02-1.11) .001
*Odds ratios, indicating the risk associated with an increase of 1 patient per nurse, and confidence intervals were derived from robust logistic regression models that accounted for
the clustering (and lack of independence) of observations within hospitals. Nurse characteristics were adjusted for sex, experience (years worked as a nurse), type of degree, and
type of unit. Patient characteristics were adjusted for the patients Diagnosis Related Groups, comorbidities, and significant interactions between them. Hospital characteristics
were adjusted for high technology, teaching status, and size (number of beds).

2002 American Medical Association. All rights reserved. (Reprinted) JAMA, October 23/30, 2002Vol 288, No. 16 1991
Confidential. Do not distribute. Pre-embargo material.

patients with complications. Staffing hos- ments of 5 to 6 patients per nurse. Our dissatisfaction predict nurses inten-
pitals uniformly at 8 vs 4 patients per results do not directly indicate how tions to leave their current jobs within
nurse would be expected to entail 5.0 many nurses are needed to care for pa- a year. Although we do not know how
(95% CI, 2.4-7.6) excess deaths per 1000 tients or whether there is some maxi- many of the nurses who indicated inten-
patients and 18.2 (95% CI, 7.7-28.7) ex- mum ratio of patients per nurse above tions to leave their jobs actually did so,
cess deaths per 1000 complicated pa- which hospitals should not venture. it seems reasonable to assume that the
tients. We were unable to estimate ex- Our major point is that there are de- 4-fold difference in intentions across
cess deaths or failures associated with a tectable differences in risk-adjusted these 2 groups translated to at least a
ratio of 10 patients per nurse (one of the mortality and failure-to-rescue rates similar difference in nurse resigna-
levels proposed in California) because across hospitals with different regis- tions. If recently published estimates of
there were so few hospitals in our sample tered nurse staffing ratios. the costs of replacing a hospital medi-
staffed at that level. In our sample of 168 Pennsylvania cal and surgical general unit and a spe-
hospitals in which the mean patient-to- cialty nurse of $42 000 and $64 000,
COMMENT nurse ratio ranged from 4:1 to 8:1, 4535 respectively, are correct, improving staff-
Registered nurses constitute an around- of the 232342 surgical patients with the ing may not only save patient lives and
the-clock surveillance system in hospi- clinical characteristics we selected died decrease nurse turnover but also reduce
tals for early detection and prompt inter- within 30 days of being admitted. Our hospital costs.44
vention when patients conditions results imply that had the patient-to- Additional analyses indicate that our
deteriorate. The effectiveness of nurse nurse ratio across all Pennsylvania hos- conclusions about the effects of staff-
surveillance is influenced by the num- pitals been 4:1, possibly 4000 of these pa- ing and the size of these effects are simi-
ber of registered nurses available to assess tients may have died, and had it been 8:1, lar under a variety of specifications. We
patients on an ongoing basis. Thus, it is more than 5000 of them may have died. allowed the effect of nurse staffing to
not surprising that we found nurse staff- While this difference of 1000 deaths in be nonlinear (using a quadratic term)
ing ratios to be important in explaining Pennsylvania hospitals across the 2 staff- and vary in size across staffing levels (us-
variation in hospital mortality. Numer- ing scenarios is approximate, it repre- ing dummy variables and interaction
ous studies have reported an associa- sents a conservative estimate of prevent- terms) and found no evidence in this
tion between more registered nurses and able deaths attributable to nurse staffing sample of hospitals that additional reg-
lower hospital mortality, but often as a in the state. Our sample of patients rep- istered nurse staffing has different effects
by-product of analyses focusing directly resents only about half of all surgical at differing staffing levels. Limiting our
on some other aspect of hospital cases in these hospitals, and other pa- analyses to general and orthopedic sur-
resources such as ownership, teaching tients admitted to these hospitals are at gery patients and eliminating vascular
status, or anesthesiologist direc- risk of dying and similarly subject to the surgery patients (who have higher mor-
tion.19,27,36-42 Therefore, a simple search effects of staffing. Moreover, in Califor- tality and complication rates) did not
for literature dealing with the relation- nia, which has nearly twice as many acute affect our conclusions and effect-size
ship between nurse staffing and patient care hospitals and discharges and an estimates. Also, our findings were not
outcomes yields only a fraction of the overall inpatient mortality rate higher changed by restricting attention to inpa-
studies that have relevant findings. The than in our sample in Pennsylvania (2.3% tient deaths vs deaths within 30 days
relative inaccessibility of this evidence vs 2.0%), it would be reasonable to ex- of admission. Results were unaffected
base might account for the influential pect that the difference of 4 fewer pa- by restricting analyses to patients who
Audit Commission in England conclud- tients per nurse might result in 2000 or were discharged after our staffing mea-
ing recently that there is no evidence that more preventable deaths throughout a sures were obtained, rather than to the
more favorable patient-to-nurse ratios similar period. patients who were discharged from 9
result in better patient outcomes.43 Our results further indicate that nurses months before to 9 months following
Our results suggest that the Califor- in hospitals with the highest patient-to- the nurse surveys that produced our
nia hospital nurse staffing legislation nurse ratios are more than twice as likely staffing measures. They were also
represents a credible approach to re- to experience job-related burnout and unchanged by restricting the sample of
ducing mortality and increasing nurse almost twice as likely to be dissatisfied nurses from which we derived our staff-
retention in hospital practice, if it can with their jobs compared with nurses in ing measures to medical and surgical
be successfully implemented. More- the hospitals with the lowest ratios. This nurses, as opposed to all staff nurses.
over, our findings suggest that Califor- effect of staffing on job satisfaction and Finally, they were neither altered by
nia officials were wise to reject ratios burnout suggests that improvements in adjusting for patient-to-licensed prac-
favored by hospital stakeholder groups nurse staffing in California hospitals tical nurse ratios and patient-to-
of 10 patients to each nurse on medi- resulting from the new legislation could unlicensed assistive personnel ratios
cal and surgical general units in favor be accompanied by declines in nurse (neither of which were related to patient
of more generous staffing require- turnover. We found that burnout and outcomes) nor affected by excluding the
1992 JAMA, October 23/30, 2002Vol 288, No. 16 (Reprinted) 2002 American Medical Association. All rights reserved.
Confidential. Do not distribute. Pre-embargo material.

hospitals in our sample with smaller Drafting of the manuscript: Aiken, Clarke, Sloane, 20. Asch DA, Jedrziewski MK, Christakis NA. Re-
Silber. sponse rates to mail surveys published in medical jour-
numbers of patients or nurses. Critical revision of the manuscript for important intel- nals. J Clin Epidemiol. 1997;50:1129-1136.
One limitation of this study is the po- lectual content: Aiken, Clarke, Sloane, Sochalski, Silber. 21. The Registered Nurse Population. Rockville, Md:
Statistical expertise: Clarke, Sloane, Silber. US Dept of Health and Human Services; 1996.
tential for response bias, given a 52% re- Obtained funding: Aiken, Sloane, Sochalski. 22. Maslach C, Jackson SE. Burnout in health profes-
sponse rate. We find no evidence that the Administrative, technical, or material support: Aiken, sions: a social psychological analysis. In: Sanders GS, Suls
Clarke, Sochalski, Silber. J, eds. Social Psychology of Health and Illness. Hillsdale,
nurses in our sample were dispropor- Study supervision: Aiken, Clarke, Silber. NJ: Lawrence Erlbaum Associates; 1982:227-251.
tionately dissatisfied with their work rela- Funding/Support: This study was supported by grant 23. Maslach C, Jackson SE. Maslach Burnout Inven-
tive to Pennsylvania staff nurses from the R01 NR04513 from the National Institute of Nursing tory Manual. 2nd ed. Palo Alto, Calif: Consulting Psy-
Research, National Institutes of Health. chologists Press; 1986.
National Sample Survey of Registered Acknowledgment: We thank Paul Allison, PhD, from 24. Silber JH, Williams SV, Krakauer H, Schwartz JS.
Nurses (a national probability-based the University of Pennsylvania for statistical consul- Hospital and patient characteristics associated with
tation, and Xuemei Zhang, MS, Wei Chen, MS, and death after surgery: a study of adverse occurrence and
sample survey performed in 2000).21 Fur- Orit Even-Shoshan, MS, from the Center for Out- failure to rescue. Med Care. 1992;30:615-629.
thermore, with respect to demographic comes Research at the Childrens Hospital of Phila- 25. Silber JH, Rosenbaum PR, Ross RN. Comparing
delphia for their assistance. the contributions of groups of predictors: which out-
characteristics (sex, age, and educa- comes vary with hospital rather than patient charac-
tion) included in both surveys, our REFERENCES
teristics? J Am Stat Assoc. 1995;90:7-18.
sample of nurses also closely resembles 26. Silber JH, Rosenbaum PR, Schwartz JS, Ross RN,
1. Stolberg SG. Patient deaths tied to lack of nurses. Williams SV. Evaluation of the complication rate as a
those participating in the National New York Times. August 8, 2002:A18. measure of quality of care in coronary artery bypass
2. Parker-Pope T. How to lessen impact of nursing graft surgery. JAMA. 1995;274:317-323.
Sample Survey of Registered Nurses. We shortage on your hospital stay. Wall Street Journal. 27. Silber JH, Kennedy SK, Even-Shoshan O, et al. An-
are confident that these results are not March 2, 2001:B1. esthesiologist direction and patient outcomes. Anes-
3. Trafford A. Second opinion: less care for patients.
specific to this particular sample of Washington Post. August 20, 2002:HE01.
thesiology. 2000;93:152-163.
28. Silber JH, Rosenbaum PR, Trudeau ME, et al. Mul-
nurses. Ultimately, longitudinal data sets 4. Aiken LH, Clarke SP, Sloane DM, et al. Nurses re-
tivariate matching and bias reduction in the surgical
ports on hospital care in five countries. Health Aff
will be needed to exclude the possibil- (Millwood). 2001;20:43-53.
outcomes study. Med Care. 2001;39:1048-1064.
29. Silber JH, Kennedy SK, Even-Shoshan O, et al. An-
ity that low hospital nurse staffing is the 5. Henry J. Kaiser Family Foundation. Survey of phy-
esthesiologist board certification and patient out-
consequence, rather than the cause, of sicians and nurses. Available at:
comes. Anesthesiology. 2002;96:1044-1052.
/content/1999/1503. Accessed March 22, 2002.
poor patient and nurse outcomes. 30. Hanley JA, McNeil BJ. The meaning and use of
6. Shindul-Rothschild J, Berry D, Long-Middleton E.
the area under a receiver operating characteristic (ROC)
Our findings have important impli- Where have all the nurses gone? final results of our
Patient Care Survey. Am J Nurs. 1996;96:25-39. curve. Radiology. 1982;143:29-36.
cations for 2 pressing issues: patient 7. Commonwealth Fund. Doctors in five countries see 31. Huber PJ. The Behavior of Maximum Likeli-
decline in health care quality. Commonwealth Fund hood Estimates Under Non-standard Conditions: Pro-
safety and the hospital nurse shortage. ceedings of the Fifth Berkeley Symposium on Math-
Quarterly. 2000;6:1-4.
Our results document sizable and sig- 8. Joint Commission on Accreditation of Healthcare ematical Statistics and Probability. Berkeley: University
of California Press; 1967:221-233.
nificant effects of registered nurse staff- Organizations. Health care at the crossroads: strate-
32. White H. Maximum likelihood estimation of mis-
gies for addressing the evolving nursing crisis, 2002.
ing on preventable deaths. The associa- Available at: + room specified models. Econometrica. 1982;50:1-25.
tion of nurse staffing levels with the /news + release + archives/health + care + at + the 33. Hosmer DW, Lemeshow S. Applied Logistic Re-
+crossroads.pdf. Accessed September 17, 2002. gression. New York, NY: John Wiley & Sons Inc; 1989.
rescue of patients with life-threatening 9. Spetz J. What should we expect from Californias 34. Bishop YM, Fienberg SE, Holland PW. Discrete
conditions suggests that nurses contrib- minimum nurse staffing legislation. J Nurs Adm. 2001; Multivariate Analysis: Theory and Practice. Cam-
31:132-140. bridge, Mass: MIT Press; 1975.
ute importantly to surveillance, early de- 10. Seago JA. The California experiment: alterna- 35. Agresti A. Categorical Data Analysis. New York,
tection, and timely interventions that tives for minimum nurse-to-patient ratios. J Nurs Adm. NY: John Wiley & Sons Inc; 1990.
2002;32:48-58. 36. Shortell SM, Hughes EFX. The effects of regula-
save lives. The benefits of improved reg- tion, competition, and ownership on mortality rates
11. Friedman L. Nurse ratios are still too high. San Fran-
istered nurse staffing also extend to the cisco Chronicle. January 30, 2002:A19. among hospital inpatients. N Engl J Med. 1988;318:
12. Governor Gray Davis announces proposed nurse- 1100-1107.
larger numbers of hospitalized patients 37. Institute of Medicine. Nursing Staff in Hospitals
to-patient ratios [press release]. Sacramento, Calif: Of-
who are not at high risk for mortality but fice of the Governor; January 22, 2002. and Nursing Homes: Is It Adequate? Washington, DC:
nevertheless are vulnerable to a wide 13. Lake ET. Advances in understanding and predict- National Academy Press; 1996.
ing nurse turnover. Res Sociol Health Care. 1998;15: 38. Aiken LH, Sloane DM, Lake ET, Sochalski J, We-
range of unfavorable outcomes. Improv- 147-171. ber AL. Organization and outcomes of inpatient AIDS
ing nurse staffing levels may reduce 14. Aiken LH, Clarke SP, Sloane DM. Hospital staffing, care. Med Care. 1999;37:760-772.
organizational support, and quality of care: cross-national 39. Mitchell PH, Shortell SM. Adverse outcomes and
alarming turnover rates in hospitals by findings. Int J Qual Health Care. 2002;14:5-13. variations in organization of care delivery. Med Care.
reducing burnout and job dissatisfac- 15. AHA Annual Survey Database. 1999 ed. Chi- 1997;35(suppl 11):NS19-NS32.
cago, Ill: American Hospital Association; 1999. 40. Moses LE, Mosteller F. Institutional differences in
tion, major precursors of job resigna- 16. Commonwealth of Pennsylvania. Hospital Ques- postoperative death rates. JAMA. 1968;203:492-494.
tion. When taken together, the im- tionnaire: Reporting Period July 1, 1998-June 30, 41. Pronovost PJ, Jenckes MW, Dorman T, et al. Or-
pacts of staffing on patient and nurse 1999. Harrisburg, Pa: Department of Health, Divi- ganizational characteristics of intensive care units re-
sion of Statistics. lated to outcomes of abdominal aortic surgery. JAMA.
outcomes suggest that by investing in 17. Needleman J, Buerhaus PI, Mattke S, Stewart M, 1999;281:1310-1317.
registered nurse staffing, hospitals may Zelevinsky K. Nurse staffing and patient outcomes in 42. Needleman J, Buerhaus P, Mattke S, Stewart M,
hospitals. Available at: Zelevinsky K. Nurse-staffing levels and the quality of
avert both preventable mortality and low /staffstudy.htm. Accessed August 6, 2002. care in hospitals. N Engl J Med. 2002;346:1715-
nurse retention in hospital practice. 18. Ayanian JZ, Weissman JS, Chasan-Taber S, Ep- 1722.
stein AM. Quality of care for two common illnesses 43. Audit Commission. Acute Hospital Portfolio: Re-
Author Contributions: Study concept and design: in teaching and nonteaching hospitals. Health Aff view of National Findings: Ward Staffing. London, En-
Aiken, Clarke, Sloane, Sochalski, Silber. (Millwood). 1998;17:194-205. gland: The Audit Commission; 2001:3.
Acquisition of data: Aiken, Clarke, Sochalski, Silber. 19. Hartz AJ, Krakauer H, Kuhn EM, et al. Hospital 44. Nursing Executive Committee. Reversing the Flight
Analysis and interpretation of data: Aiken, Clarke, characteristics and mortality rates. N Engl J Med. 1989; of Talent: Nursing Retention in an Era of Gathering
Sloane, Silber. 321:1720-1725. Shortage. Washington, DC: Advisory Board Co; 2000.

2002 American Medical Association. All rights reserved. (Reprinted) JAMA, October 23/30, 2002Vol 288, No. 16 1993