Effective Clinical Practice
Effective Clinical Practice, September/October 2001
Peter J. Pronovost, MD, PhD, Eborah Dang, MSN, Todd Dorman, MD, Pamela A. Lipsett, MD, Elizabeth Garrett, PhD, Mollie Jenckes, MSc, Eric B. Bass, MD, MPH
For author affiliations, current addresses, and contributions, see end of text.
See related editorial.
Context. We previously found that length of stay in the intensive care unit (ICU) after abdominal aortic surgery increased when fewer ICU nurses were available per patient. We hypothesized that having fewer nurses increases the risk for medical complications.
Objective. To evaluate the association between nurse-to-patient ratio in the ICU and risk for medical and surgical complications after abdominal aortic surgery.
Design. Observational study.
Setting. All nonfederal acute care hospitals in Maryland.
Data Sources. Information about patients came from hospital discharge data on all patients in Maryland with a principal procedure code for abdominal aortic surgery from 1994 through 1996 (n = 2606). The organizational characteristics of ICUs were obtained by surveying ICU medical and nursing directors in 1996 at the 46 Maryland hospitals that performed abdominal aortic surgery. Thirty-nine of the ICU directors (85%) completed the survey.
Exposure. Surgery in hospitals with fewer ICU nurses (in which each nurse cared for three or four patients) compared with hospitals with more ICU nurses (in which each nurse cared for one or two patients).
Outcome. Proportion of patients who developed postoperative complications.
Results. Seven hospitals with 478 patients had fewer ICU nurses, and 31 hospitals with 2128 patients had more ICU nurses. Patients in hospitals with fewer nurses were more likely than patients in hospitals with more nurses to have complications: 47% vs. 34% had any complication, 43% vs. 28% had any medical complication, 24% vs. 9% had pulmonary insufficiency after a procedure, and 21% vs.13% were reintubated (P < 0.001 for all comparisons). After adjustment for patient, hospital, and surgeon characteristics, having fewer versus more ICU nurses was associated with an increased risk for any complication (relative risk, 1.7 [95% CI, 1.3 to 2.4]), any medical complication (relative risk, 2.1 [CI, 1.5 to 2.9]), pulmonary insufficiency after procedure (relative risk, 4.5 [CI, 2.9 to 6.9]) and reintubation (relative risk, 1.6 [CI, 1.1 to 2.5]).
Conclusion. Having fewer ICU nurses per patient is associated with increased risk for respiratory-related complications after abdominal aortic surgery.
Patients in intensive care units (ICUs) require substantial nursing care. Nevertheless, nurse staffing varies widely among ICUs, and the optimal nurse-to-patient ratio remains unclear. (1) Efforts to reduce costs of hospital care have resulted in decreased nurse-to-patient ratios; however, the effects of this reduction on quality of care have not been well evaluated, causing concern among patients, providers, and policymakers. (2-5) Previous studies that measured nurse staffing at the hospital level found an association between nurse staffing and specific complications, including postoperative infections and pulmonary complications. (6-8) However, nurse staffing varies within hospitals. By evaluating the impact of nurse staffing at the unit level rather than the hospital level, we may be more likely to identify an association between nurse staffing and patient outcomes. Because critically ill patients in an ICU require close nursing surveillance to ensure early detection of potential adverse events and prompt intervention to prevent morbidity, a reduction in ICU nurse staffing may negatively affect quality and outcomes of care.
Previously, we reported that a decreased ICU nurse-to-patient ratio was associated with a significant increase in length of stay in the ICU for patients who had abdominal aortic surgery in Maryland. (9) We hypothesized that the increased length of stay may be due to a greater risk for medical complications. We therefore sought to evaluate the association between the ICU nurse-to-patient ratio and risk for medical and surgical complications.
After obtaining approval from our institutional review board, we gathered nonconfidential patient data from the Uniform Health Discharge Data Set. This data set, which is maintained by the Maryland Health Services Cost Review Commission (HSCRC), contains information on all patients discharged from the 52 nonfederal, short-term-stay hospitals in Maryland. We obtained information on all patients 30 years of age or older who were discharged from a Maryland hospital between January 1994 and December 1996 with a principal procedure code for abdominal aortic surgery (International Classification of Diseases, Ninth Revision, Clinical Modification [ICD-9-CM] code 3844 for resection of abdominal aorta with replacement and ICD-9-CM code 3925 for aortoiliac-femoral bypass). We excluded 9 patients who were younger than 30 years of age, all of whom had had an injury to a blood vessel (ICD-9-CM code 902).
Exposure: ICU Nurse Staffing
In December 1996, we mailed a survey to the ICU directors at the 46 Maryland hospitals recorded in the HSCRC database as having performed abdominal aortic surgery in our target period. These directors were asked to have the ICU nurse managers complete the information about ICU nurse staffing. The details of our survey methods are published elsewhere. (9)
The survey included questions about the average ICU nurse-to-patient ratio during the day and evening; possible responses were 1:1, 1:2, 1:3, or 1:4. On the basis of the few responses with ratios of 1:1 and 1:4 and data from exploratory analyses, we created a dichotomous variable: We considered nurse-to-patient ratios of 1:1 or 1:2 as "more ICU nurses" and ratios of 1:3 or 1:4 as "fewer ICU nurses." Because our previous study demonstrated that ICU nurse staffing during the day rather than the evening was associated with ICU length of stay, we limited the current analysis to ICU nurse staffing during the day. (9)
We obtained information about complications and comorbid conditions by using one of the 14 secondary diagnoses and procedure codes included in the HSCRC database. The list of complications was created before the start of our study by a panel of four critical care physicians who used a Delphi process to determine the most important complications. In our previous analysis, several of these complications were found to be associated with in-hospital mortality or length of stay, which supports construct validity. (9)
In this paper, we focused on medical and surgical complications that were independently associated with in-hospital mortality. These complications are defined in Table 1. Although the discharge diagnosis codes we used do not distinguish between complications and comorbid conditions, the medical diagnoses listed here are for acute problems and are therefore more likely to represent complications than comorbid conditions.
We assessed the number of hospital beds and the volume of aortic surgery performed during the study period by each hospital and each surgeon in the database.
We included variables for patient age (in years), sex, and race (white or nonwhite).
We used the Romano-Charlson comorbidity index to identify potentially important comorbid conditions. (10-13) Instead of using a single comorbidity index in our analysis, each disease in the index was included as a separate variable.
Severity of Illness
To account for severity of illness, we classified patients as having a ruptured (ICD-9-CM code 441.3) or nonruptured aneurysm and used the nature of admission field, coded at the time of admission, to identify each case as elective, urgent, or emergent (Dardik A. Personal communication. 1998). (14)
We performed a descriptive analysis of patient characteristics, in-hospital mortality, hospital length of stay, days in the ICU, and rates of medical and surgical complications for patients with more versus fewer ICU nurses. Using the patient as the unit of analysis, we performed bivariate analysis with simple logistic regression to evaluate the association between the ICU nurse staffing (independent variable) and each of the complications (dependent variables).
We then performed a multivariate logistic regression to evaluate the independent association between ICU nurse staffing and each of the complications. In each of these models, we adjusted for patient age (in years), sex, race (white or nonwhite), nature of admission (emergent vs. elective and urgent vs. elective), type of aneurysm (ruptured or nonruptured), and comorbid conditions (yes or no for each of the 10 diseases in the Romano-Charlson index). Because routinely extubating patients in the operating room was independently associated with increased length of stay in our original study, (9) we included this variable in our multivariate model. We also included terms for surgeon volume and hospital volume.
We modeled hospital and surgeon volume as dichotomous variables. Using Lowess smoothing curves, (15) we defined low volume as fewer than 36 cases per year for hospitals and fewer than 8 cases per year for surgeons. (A Lowess smoothing curve is a method used to identify cut-points in data by obtaining a locally weighted regression. The output of this analysis is a graph with volume on the x axis and probability of death on the y axis. The cut-points for high and low volume were chosen from these graphs.)
Patients in our study were naturally clustered within hospitals, which artificially decreased our variance estimates. We used hierarchical models that took such clustering into account. (16) For the bivariate and multivariate analyses, we converted the odds ratios obtained from logistic regression to relative risks by using the method described by Zhang. (17) All reported P values are two-tailed. P values were considered significant if they were less than 0.05. We used STATA 6.0 software (Houston, Texas) to perform all calculations.
Patient and ICU nurse staffing data were available for 38 of the 46 hospitals in this study. Seven hospitals had fewer ICU nurses and 31 hospitals had more ICU nurses. Both groups included small and large hospitals and low and high surgical volumes (Table 2). The practice of routinely extubating patients in the operating room, which was associated with an increased risk for reintubation and postoperative pulmonary complications in our previous study, (9) was not related to nurse staffing.
Table 3 shows the characteristics of patients who had surgery in hospitals with fewer versus more ICU nurses. Demographic characteristics and severity of illness did not differ between patients in either nurse staffing model. Mild diabetes mellitus was the only comorbid disease that occurred significantly more often in patients in ICUs with fewer nurses. Mild diabetes mellitus was associated neither with ICU length of stay in our previous study (9) nor with risk for any complication or medical complication in this analysis. Patients in hospitals with fewer nurses were less likely to be treated in low-volume hospitals and to undergo surgery by a low-volume surgeon. The in-hospital mortality rate did not statistically differ between the two nurse staffing models.
Overall, 47% of patients treated in hospitals with fewer ICU nurses and 34% of patients treated in hospitals with more ICU nurses had at least one complication. Patients treated in hospitals with fewer ICU nurses were also more likely to experience any medical complication (43% vs. 28%) (Table 4). Pulmonary insufficiency (24% vs 9%) and reintubation (21% vs 13%) were more common among patients treated in hospitals with fewer ICU nurses. Nurse staffing did not affect the risk for a surgical complication.
The crude analysis showed that having fewer ICU nurses was associated with a significantly increased risk for having any complication, any medical complication, pulmonary insufficiency after a procedure, cardiac complications after a procedure, and reintubation (Table 4). In multivariate analysis adjusted for patient characteristics and hospital and surgeon volume, having fewer nurses was associated with an increased risk for any complication (relative risk, 1.7 [95% CI, 1.3 to 2.4]), any medical complication (relative risk, 2.1 [CI, 1.5 to 2.9]), pulmonary insufficiency after procedure (relative risk, 4.5 [CI, 2.9 to 6.9]), and reintubation (relative risk, 1.6 [CI, 1.1 to 2.5]).
In our previous analysis, daily rounds by an ICU physician was related to risk for complications. (9) After we added daily rounds by an ICU physician to the multivariate analysis used in this study, all regression coefficients remained unchanged, suggesting that ICU nurse staffing has an independent effect.
We found that having fewer versus more ICU nurses per patient was independently associated with an increased risk for medical complications, including pulmonary insufficiency after a procedure and reintubation. Although we cannot establish a causal relationship by using this type of analysis, our findings have face validity and support our hypothesis that ICU nurse staffing is associated with risk for pulmonary complications. Nurses who care for three or more patients in the ICU may have less time than nurses who care for one or two patients to devote to patient care—especially preventive measures, such as postoperative pulmonary care.
This study adds to our previous work in evaluating how the organization of ICU care is related to clinical and economic outcomes. We previously found that no daily rounds by an ICU physician is associated with increased risk for acute renal failure, septicemia, cardiac arrest, reintubation, and platelet transfusion. (9) In the current study, daily rounds by an ICU physician and ICU nurse staffing during the day were not significantly correlated; addition of daily rounds by an ICU physician to the multivariate analysis did not change the regression coefficients. These results suggest that ICU nurse staffing is associated with risk for specific complications independent of ICU physician staffing. Optimal ICU organization should probably address both physician and nurse staffing.
Our study has several limitations that may affect our conclusions. First, because the study is essentially a retrospective analysis, we cannot establish a causal relationship between ICU nurse staffing and complications. Nonetheless, our findings are consistent with clinical experience that nurse staffing can affect pulmonary complications. Second, coding of complications and comorbid diseases in the HSCRC database may not be as accurate as the coding of the principal procedure. (18) Because the percentage of patients with comorbid conditions (with the exception of diabetes) were similar, there does not appear to be systematic differences in coding of comorbid diseases at hospitals with more or fewer ICU nurses. Therefore, misclassification of comorbid diseases probably did not significantly bias our results. It is also unlikely that misclassification of complications biased our results. We reviewed a random sample of 25 medical records from one hospital and found that coding of complications associated with in-hospital mortality was 96% accurate. We helped establish construct validity for coding of complications by evaluating only complications that were independently associated with in-hospital mortality. (19) In addition, bias from coding of complications was probably minimal, since the association between nurse staffing and pulmonary complications is consistent with that in previous studies. (6) Despite the potential problems with coding, our analysis based on discharge data and survey data allows us to evaluate the association between health care organizational characteristics and outcomes. (20-23)
Third, we did not use a systematic scoring system, such as The Acute Physiology and Chronic Health Evaluation (APACHE) II that requires medical record review to adjust for severity of illness. (24) We classified patients as having ruptured or nonruptured aneurysms and elective, urgent, or emergent admissions, and the coding for these variables was found to be more than 98% accurate. (25) Patients who have abdominal aortic surgery are unique because the variables for ruptured versus nonruptured aneurysm and elective, urgent, or emergent admission seem to account for severity of illness relatively well, as evidenced by the odds ratios for in-hospital mortality in our original study. (9) Fourth, we did not adjust for differences in pre-ICU care, including surgical approach and type of anesthesia, and post-ICU care. However, we did adjust for hospital volume and used multilevel modeling that may account for some of these differences. Nevertheless, differences in pre- and post-ICU care, including nursing care on the regular surgical floor, could have affected some of our results.
The fifth limitation is the validity of our measure of nurse staffing. The ICU nurse-to-patient ratio is a relatively crude measure of nursing surveillance; it is a complex variable that may be affected by staff mix, experience, training, certification, fatigue, and nursing workload. Nonetheless, nurse-to-patient ratio is easy for managers to understand and act on, and our survey provided a unit-level measure of nurse staffing that might be more accurate than hospital-level measures. From the patients' and physicians' perspective, nurse-to-patient ratio may be a measure of the perception of the adequacy of nurse staffing. Random misclassification of either the exposure or the intervention biases a study toward the null. Nurse-to-patient ratio may be a relatively imprecise measure of nursing surveillance and as such, may result in random misclassification of exposure, biasing the study to the null. Had we used more detailed measures of nursing surveillance, our findings may have been even stronger. Our findings warrant further investigation to determine how these other measures of nursing surveillance may affect patient outcomes. Finally, we focused on one surgical procedure in one state; the applicability of these findings to other procedures and other states is therefore limited. However, our study is strengthened by inclusion of survey data from 38 of the 46 hospitals that performed abdominal aortic surgery in Maryland.
Because complications are influenced by a complex array of factors, many of which may be unrelated to nursing, it could be debated whether we should link nurse staffing and complications as a measure of quality. Although work in this area is just beginning, our study and recently published studies support the idea that the level of nursing surveillance distinguishes hospitals with lower complication rates from those with higher complication rates. (4, 6)
Our results are important because they identify specific and potentially modifiable ICU characteristics that could be altered to reduce ICU complications and length of stay in high-risk surgical patients. Additional research is needed to further define the optimal ICU nurse-to-patient ratio for an ICU with a given severity of illness. When hospital administrators consider having ICU nurses care for more patients to reduce costs, they should consider the impact of this change on quality of care. In patients undergoing abdominal aortic surgery, more ICU nurses per patient during the day is associated with reduced risk for medical, particularly pulmonary, complications.
|Take Home Points
1. Groeger JS, Strosberg MA, Halpern NA, et al. Descriptive analysis of critical care units in the United States. Crit Care Med. 1992;20:846-63.
2. Aiken LH, Sochalski J, Anderson GF. Downsizing the hospital nursing workforce. Health Aff (Millwood). 1996;15:88-92.
3. Brannon RL. Restructuring hospital nursing: reversing the trend toward a professional work force. Int J Health Serv. 1996;26:643-54.
4. Blegen MA, Goode CJ, Reed L. Nurse staffing and patient outcomes. Nurs Res. 1998;47:43-50.
5. Wunderlich GS, Sloan FA, Davis CK. Nursing Staff in Hospitals and Nursing Homes: Is It Adequate? Washington, DC: National Academy Press; 1996.
6. Kovner C, Gergen PJ. Nurse staffing levels and adverse events following surgery in U.S. hospitals. Image J Nurs Sch. 1998; 30:315-21.
7. Aiken LH, Smith HL, Lake ET. Lower Medicare mortality among a set of hospitals known for good nursing care. Med Care. 1994;32:771-87.
8. American Nurses Association. Implementing Nurse's Report Care: A Study of RN Staffing, Length of Stay and Patient Outcomes. Washington, DC: American Nurses Publishing; 1997.
9. Pronovost PJ, Jenckes MW, Dorman T. Organizational characteristics of intensive care units related to outcomes of abdominal aortic surgery. JAMA. 1999;281:1310-7.
10. Romano PS, Roos LL, Jollis JG. Adapting a clinical comorbidity index for use with ICD-9-CM administrative data: differing perspectives. J Clin Epidemiol. 1993;46:1075-9.
11. Charlson ME, Pompei P, Ales KL, MacKenzie CR. A new method of classifying prognostic comorbidity in longitudinal studies: development and validation. J Chronic Dis. 1987;40: 373-83.
12. Roos LL, Fisher ES, Sharp SM, Newhouse JP, Anderson G, Bubolz TA. Postsurgical mortality in Manitoba and New England. JAMA. 1990;263:2453-8.
13. Wennberg DE, Lucas FL, Birkmeyer JD, Bredenberg CE, Fisher ES. Variation in carotid endarterectomy mortality in the Medicare population: trial hospitals, volume, and patient characteristics. JAMA. 1998;279:1278-81.
14. Dardik A, Burleyson GP, Bowman H, et al. Surgical repair of ruptured abdominal aortic aneurysms in the state of Maryland: factors influencing outcome among 527 recent cases. J Vasc Surg. 1998;28:413-20.
15. Cleveland WS. Robust locally weighted regression and smoothing scatterplots. Journal of the American Statistical Association. 1979;74:829-36.
16. Paterson L, Goldstein H. New statistical methods for analyzing social structures: An introduction to multilevel models. British Educational Research Journal. 1991;17:387-93.
17. Zhang J, Yu KF. What's the relative risk? A method of correcting the odds ratio in cohort studies of common outcomes. JAMA. 1998;280:1690-1.
18. The Johns Hopkins Hospital Report of Clinical Data Quality Review. Baltimore: Deloitte, Haskins, and Sells; 1996.
19. Romano PS. Can administrative data be used to compare the quality of health care? Med Care Rev. 1993;50:451-77.
20. Steiner CA, Bass EB, Talamini MA, Pitt HA, Steinberg EP. Surgical rates and and operative mortality for open and laparoscopic cholecystectomy in Maryland. N Engl J Med. 1994;330:403-8.
21. Iezzoni LI. Data sources and implications: administrative databases. In: Iezzoni LI, ed. Risk Adjustment for Measuring Healthcare Outcomes. Chicago: Health Administration Press; 1997.
22. Tunis SR, Bass EB, Steinberg EP. The use of angioplasty, bypass surgery, and amputation in the management of peripheral vascular disease. N Engl J Med. 1991;325:556-62.
23. Angus DC, Linde-Zwirble WT, Sirio CA, et al. The effect of managed care on ICU length of stay: implications for medicare. JAMA. 1996;276:1075-82.
24. Knaus WA, Draper EA, Wagner DP, Zimmerman JE. APACHE II: a severity of disease classification system. Crit Care Med. 1985;13:818-29.
25. Dardik A, Lin JW, Gordon TA, Williams GM, Perler BA. Results of elective abdominal aortic aneurysm repair in the 1990s: a population-based analysis of 2335 cases. J Vasc Surg. 1999;30:985-95.
Peter J. Pronovost, MD, PhD, Department of Anesthesiology/CCM, The Johns Hopkins University School of Medicine, 600 North Wolfe Street, Meyer 297A, Baltimore, MD 21205-7294; telephone: 410-955-9080; fax: 410-955-8978; e-mail: firstname.lastname@example.org.