Chapter 39. Nurse Staffing, Models of Care Delivery, and Interventions (continued)

You Are Here: AHRQ Home > Clinical Information > EPC Evidence Reports > Archived Reports > Making Health Care Safer: A Critical Analysis of Patient Safety Practices

Chapter 39. Nurse Staffing, Models of Care Delivery, and Interventions

(continued)

Table 39.3 Structural measures: availability of nurses and patient outcomes (First 11 studies showed positive associations; final 5 studies detected no significant effect)

Study Setting	Study Design, Outcomes	Availability of Nurses	Effect Size (coefficient, mean differences, OR)
1. Data were collected from 1,205 consecutively admitted patients in 40 units in 20 acute care hospitals and on 820 nurses in the US¹¹⁵	Level 3, Level 1&3	0.8 mean nurse/ patient day with a range of 0.5-1.5 nurses/patient day	This measure was significantly associated with 30-day mortality (OR .46, 95% CI: 0.22-0.98). An additional nurse per patient day reduces the odds of dying by one-half.
2. All patients who developed a central venous catheter bloodstream infection during an infection outbreak period (January 1992 through September 1993) and randomly selected controls. Cohort study: all SICU patients during the study period (January 1991 through September 1993)¹²⁶	Level 3, Level 1	1.2 patient/nurse and 20 nursing hours per patient day (HPPD) 1.5 patient/nurse and 16 nursing HPPD 2 patient/nurse and 12 nursing HPPD	There was a significant relationship between nurse to patient ratios and nursing hours and central venous catheter bloodstream infection in the SICU. For 1.2 patients/nurse and 20 HPPD the adjusted odds ratio was 3.95 (95% CI: 1.07-14.54), 1.5 patients/nurse and 16 nursing HPPD, 15.6 (95% CI: 1.15-211.4), and for 2 patients/nurse and 12 HPPD, 61.5 (95% CI:1.23-3074).
3. 39 nursing units in 11 hospitals for 10 quarters of data between July, 1993 and December, 1995 in the US⁸⁴	Level 3, Level 1&2	Proportion of direct care RN hours; total direct care hours; up to 87.5% RN skill mix	With patient acuity controlled, direct care RN proportion of hours was inversely associated with medication errors (-0.525 p<0. 05), decubiti (-0.485 p<0.05), and complaints (-0.312, p<0.10). Total direct care hours was positively associated with decubiti (0.571, p<0.10), complaints (0.471, p<0.10), and mortality (0.491, p<0.05). A curvilinear relationship was found so that as RN proportion increased, rates of all adverse events decreased up to a proportion of 88% RNs. Above that level, as RN proportion increased, the adverse outcomes increased.
4. 42 inpatient units in one 880-bed hospital in the US⁸³	Level 3, Level 1&2	8.63 mean total hours of care; 69% RN skill mix; up to 85% skill mix	With patient acuity controlled, direct care RN proportion of hours was inversely associated with medication errors/doses (-0.576, p<0.05) and falls (-0.456, p<0.05). Total direct care hours was positively associated with medication errors/doses (0.497, p<0.05). A curvilinear relationship was found so that as RN proportion increased, medication error rates decreased up to a proportion of 85% RNs. Above that level, as RN proportion increased, the medication error increased
5. Data from hospital cost disclosure reports and patient discharge abstracts from acute care hospitals in California and New York for fiscal years 1992 and 1994¹²⁵	Level 3, Level 1&2	7.56-8.43 mean total hours of care/nursing intensity weight (NIW); 67.7% to 70.5% RN skill mix	Total hours/NIW was inversely associated with pressure ulcer rates (-15.59, p<0.01). RN hours in California, but not New York, was inversely associated with pneumonia (-0.39, p<0.01) Nonsignificant association with postoperative infection rates.
6. Data from hospital cost disclosure reports, patient discharge abstracts and Medicare data from acute care hospitals in Arizona, California, Florida, Massachusetts, New York, and Virginia for 1996¹²³	Level 3, Level 1&2	5.76 mean licensed hours of care/ 83.3% RN skill mix	Skill mix was inversely associated with pneumonia (-0.20, p<0.01), postoperative infection (-0.38, p<0.01), pressure ulcers (-0.47, p<0.01), and urinary tract infections (-0.61, p<0.01).
7. Data from hospital cost disclosure reports, patient discharge abstracts from acute care hospitals in California, Massachusetts, and New York for 1992 and 1994¹²²	Level 3, Level 1&2	7.67-8.43 mean total hours of care; 67.7-70.5% skill mix	RN hours were inversely associated with pneumonia (-0.39, p<0.01), pressure ulcer rates (-1.23, p<0.01), and postoperative infection (-0.47, p<0.01) but not significant for urinary tract infections.
8. Data from HCFA Medicare Hospital Mortality Information 1986 and the American Hospital Association 1986 annual survey of hospitals¹¹⁶	Level 3, Level 1	0.9 mean RN/ADC (average daily census); 60% skill mix	Controlling for hospital characteristics, number of RNs/ADC was not significantly related to adjusted 30-day mortality rate but proportion of RNs/all nursing staff was significantly related to adjusted 30-day mortality rate (adjusted difference between lower and upper fourth of hospitals -2.5, 95% CI: -4.0 to -0.9)
9. Data from the American Hospital Association 1986 annual survey of hospitals and medical record reviews from July 1987 to June 1988 in 6 large PPOs¹²⁸	Level 3, Level 3	52.2 (Texas)-67.6% (California) skill mix	Controlling for hospital characteristics, number of RNs/ADC was not significantly related to problem rate but proportion of RNs/all nursing staff was significantly related to lower problem rates (California lower rates 3.58, upper rates 2.30 p<0.0001)
10. Data from the American Hospital Association Annual Survey of Hospitals for 1993 and the Nationwide Inpatient Sample from the Agency for Health Care Policy and Research for 1993 (HCUP-3)¹²⁴	Level 3, Level 1	67.8% mean skill mix	Proportion of RN FTEs/all nursing FTEs was inversely related to thrombosis after major surgery (beta -33.22, 95% CI: -57.76 to -8.687), urinary tract infection after surgery (beta -636.96, 95% CI: -852.78 to -421.15), pneumonia after major surgery (beta -159.41, 95% CI: -252.67 to -66.16), and pulmonary compromise after major surgery (beta -59.69, 95% CI: -117.62 to 1.76).
11. Data were collected form March 1 to June 7, 1986 and included 497 patients¹²⁷	Level 3, Level 2	Adequate staffing	The adequately staffed unit had fewer complications than the inadequately staffed unit.
12. 390 patients admitted within 1 week after stroke onset in 9 acute care hospitals in The Netherlands. Surviving patients were interviewed 6 months post-stroke and asked about falls. Fall and other patient data were collected from medical records. Ward characteristics were provided by senior nurses. There is complete data on 349 patients⁸⁹	Level 3, Level 2	0.04 mean difference in nurse to patient ratios	There was no statistical difference in falls between case and control groups in number of nurses or nurse ratios on any shift. Days (mean difference -0.06, CI: -0.51 to 0.39); Evening (mean difference -0.24, 95% CI: -0.97 to 0.50); Nights (mean difference 1.24, 95% CI: 0.28 to 2.20); All shifts (mean difference 0.04, 95% CI, -0.33 to 0.40).
13. 17,440 patients across 42 ICUs in the US³⁰	Level 3, Level 1-3	Mean .66 patient/nurse with a range of 0.31-1.31	Neither nurse to patient ratio nor caregiver interaction was found to be significantly associated with risk-adjusted mortality.
14. Data were collected from April, 1994-March, 1995 from 23 trusts (groups of hospitals) in Scotland¹¹⁷	Level 3, Level 1	Mean RN FTE was 1.21 per patient	There was no association between RN FTE per occupied hospital bed and mortality
15. Data were collected form the American Hospital Association Annual Survey of Hospitals in 1989-1991, the observed and predicted 30-day post-admission mortality for patients with a primary diagnosis of COPD from the HCFA Hospital Information Reports from 1989-1991 and the Medicare Case Mix Index¹¹⁸	Level 3, Level 1	RN FTE/100 adjusted admissions	There was no association between RN FTE/100 adjusted admissions and 30-day post-admission mortality for patients with a primary diagnosis of COPD
16. Data from staffing and accounting records of 60 community hospitals across the US in 1985, hospital and nursing unit surveys, 1981 case mix indexes from the Federal Register, and the Health Area Resources File¹²⁹	Level 3, Level 3	52% RN skill mix; 33% LPN mean nursing HPPD was 4.93	None of the staffing variables of interest were associated with medication errors, patient injuries, IV administration errors, or treatment errors.

Table 39.4 Structural variables: nursing organization models and patient outcomes

Study Setting	Study Design, Outcomes	Organization of Care/Models	Effect Size (coefficient, mean differences, OR)
Data were collected from 39 "magnet" hospitals, which are hospitals designated as good places for nurses to work, and 195 nonmagnet matched hospitals²⁹	Level 3, Level 1	Magnet hospitals	Magnet hospitals had a 4.6% lower adjusted Medicare mortality rates (p=0.026, 95% CI: 0.9-9.4 fewer deaths per 1,000)
Data were collected form 1,205 consecutively admitted patients in 40 units in 20 acute care hospitals and on 820 nurses in the US¹¹⁵	Level 3, Level 1&2	Magnet hospitals (nurse control over practice variable)	Nurse control over practice was not significantly associated with any clinical outcomes, but was significantly associated with patient satisfaction (coefficient 0.56 (95% CI: 0.16-97)
17,440 patients across 42 ICUs in the US³⁰	Level 3, Level 1-3	Magnet hospitals (nurse unit culture captured in caregiver interaction variable)	Caregiver interaction was not significantly associated with clinical outcomes, but was significantly associated with lower risk-adjusted length of stay (-0.16, p<0.05) and lower nurse turnover (-0.21, p<0.05)
Data were collected at 3 points in time; 6 month before the intervention, 6 months, and 12 months after the introduction of the new model and included the time between October 1996 to December 1997⁷⁹	Level 3, Level 2	Patient Focused Care	There was a significant reduction in medication errors between the pre-model change (0.97%) and the post-model change (0.78%, p=0.016) and no difference in the other measures
Data were collected 6 months before and 6 months after the introduction of the new model and included the time between January-June, 1992 and January-June, 1993¹³⁰	Level 3, Level 2	RN-UAP Partnership similar to Patient Focused Care	There was a significant reduction in falls (4.7732, p< 0.05) and no difference in the other measures between the pre- and post-measures.
Review article: Pierce, 1997¹³¹	Level 3A, Level 1&2	Nursing Environment	There are mixed results in studies about whether the predictor variables related to nurses and nursing are related to the outcomes of interest or whether the conceptual models being used are incomplete.
Review article: MEDLINE® from 1966-1996, CINAHL from 1982-1996, Expanded Academic Index from 1989-1996, search by author for investigators known to be working in the field, manual searches of the bibliographies of review articles and monographs (Mitchell)¹¹¹	Level 3A, Level 1&2	Nursing Environment	Mixed results in studies about whether nursing surveillance, quality of working environment, and quality of interaction with other professionals predict hospitals with lower mortality. With more sophisticated risk adjustment, evidence suggests that mortality and complications are related more to patient variables and adverse events may be more closely related to organizational characteristics.

Table 39.5 Process measures: nurse intervention and patient outcomes

Study Setting	Study Design, Outcomes	Intervention	Effect Size (coefficient, mean differences, OR)
Data were collected from 60 hospitalized patients on 1 surgical service in a university hospital in Turkey between September 1996 and September 1997⁴⁴	Level 2, Level 2&3	Added education to intervention group	Positive colonization of catheter hub was 68.6% in the control group and 25% in the intervention group (chi square=5.75, p<0.05); mean positive nurse practice scores in control group was 45.7 and 66.5 after education (p<0.05)
2 surgical and 2 medical wards in one hospital in Sweden were randomly assigned to either a control or experimental group. 18 nurses on the experimental wards and 18 nurse on the control wards; 90 patients on the experimental wards and 39 patients on the control wards; 112 Peripheral IVs on the experimental wards and 60 PIVs on the control wards⁴⁷	Level 1, Level 2&3	Added education to intervention group	50% of the PIV lines in the control group had thrombophlebitis/complications compared with 21% in intervention (p<0.001); positive association observed for nurse practices related to care of PIV lines was 12% in the control group and 72% in the experimental group; there was complete nursing documentation in 10% of the control group and 66% of the experimental group
One hospital in Spain; all nosocomial infection data between March 1982 and December 1990⁵⁴	Level 3, Level 1	Added education to intervention group	Additional training was associated with a significant 3.63% decrease (p<0.01) in nosocomial infection rates
One university hospital in Washington, DC; all adult patients with bloodstream Infections between July 1984 and February 1994 (n=432)⁴⁵	Level 3, Level 2	Added education	No significant difference in total BSI rates or central line BSI rates before, during or after the program
One general hospital in Illinois; all omitted and wrong dose medication errors between October 1992 and March 1993⁴³	Level 3, Level 2	Added education	No difference in wrong dose IV medication errors for 12 months after training; there was a decrease in omitted dose IV mediation errors for 12 months after training (p<0.01).
All urinary catheter-patient-days between January 1995 and September 1996 in 1 VA hospital⁵⁵	Level 3, Level 2	Provided infection rate data to nurses	Pre-intervention there were 32/1000 catheter-patient days (95% CI: 22.9-43.7); for the 5 quarters post intervention, there was a significant decrease (p<0.01) in the average infection rate (17.4/1000 catheter-patient-days (95% CI: 14.6-20.6)) compared to pre-intervention
Stanford University Hospital; all pressure ulcers and nosocomial pressure ulcers during 1992 through 1996⁵⁷	Level 3, Level 2	Provided nosocomial pressure rate data to nurses plus added education	After Intervention #1, total pressure ulcer rate went from 20% to 21%; nosocomial pressure ulcer rates went from 19% to 21%. After Intervention #2 total pressure ulcer rates stayed at 21% but nosocomial pressure ulcer rates went from 21% to 13%. One-year later, total pressure ulcer rates were 10.9% and nosocomial pressure rates were 8.1%.
8. Stanford University Hospital 52 bed medical surgical unit; all falls between 1995 through 1996⁵⁶	Level 3, Level 2	Provided fall rate data to nurses and added education	Pre-intervention the fall rate ranged from 4.2 to 3.7 fall per thousand patient days (FPTPD); after Intervention #1 the fall rate was 5.2 FPTPD; after Intervention #2 the fall rate ranged from 5.1 to 3.7 FPTPD.

Return to Contents
Proceed to Next Section