Advantagesof the article
Aikenetal.(2002) present a useful analysis of the effects of nurse staffinglevels on patient outcomes. Advantages of this study include 1) largesample size, 2) specific descriptions of variables, 3) obtained riskwas adjusted using 133 potentially confounding variables, 4)descriptive statistics regarding the hospitals, patients, and nursesin the study were thoroughly presented, 5) logistic regression wasused to account for patient and nurse clustering in hospitals(because variables might not be independent), and 6) confidenceintervals and p levels were given for each odds ratio.
First,samples were taken from Pennsylvania hospitals (sample size, n=168),staff nurses (n=10,184), and general, orthopedic, and vascularsurgery patients who were discharged between April 1, 1998, andNovember 30, 1999 (n=232, 342). The large samples increased thelikelihood of the samples representing the populations and supportedthe calculated odds ratios. Second, the variables were clearlyspecified. For example, ICD-9 codes for diagnoses, and nurse staffingas the average patient load across all nurses in a given hospital.Third, the risk-adjustment process was described as using age, sex,surgery type, and comorbid conditions to account for differences inrisk not related to nurse staffing. Fourth, descriptive statisticsconcerning hospitals included staffing ratios, size, level oftechnology, and teaching status, and nurses were described by sex,degree, experience, specialty, emotional exhaustion, and jobdissatisfaction. Fifth, logistic regression was used since theindependent variables were of different types — continuous,categorical, dichotomous, and this type of model can be used topredict or estimate the outcome category for individual cases.Finally, each estimate (effect of additional patient on mortality,failure-to-rescue, and nurse burnout / job dissatisfaction) wasaccompanied by 95% confidence intervals and calculated p-levels.
Limitationsof the article
Thereare three possible limitations that might reduce the usefulness ofthe article in estimating the impact of the level of nurse staffingon patient outcome. First, the findings reported in the article werepossibly affected by response bias. Although the researcher used anoptimum number of the study subjects, only 52 % of them responded.The effect of response bias reduces the generalizability of thefindings on the effect of nurse staffing on patient outcome.
Secondly,the study sample used in the study resembled the participants in theNational Survey of Registered Nurses in terms of the demographiccharacteristics. Hence, the study lacks the longitudinal data set,which creates a notion that the low level of nurse staffing is thenot the cause, but the consequence of poor patient as well as nurseoutcome.
Third,the article reports a cross sectional study, which has no capacity toestablish causality. This is because the cross-sectional study is anon-randomized design that makes the unexposed as well as the exposedgroups differs depending on other factors that may themselves havebeen the causes of the reported outcome. For example, nurses’dissatisfaction may have been caused by poor interpersonalrelationship, which has nothing to do with the level of nursestaffing. These factors may be referred to as the confounders.Regardless of those limitations, Aiken etal.(2002) is an excellent source for learning more about nurse staffingimpacts.
Aiken,H., Clarke, P., Sloane, M., Sochalski, J. & Silber, H. (2002).Hospital nurse staffing and patient mortality, nurse burnout, and jobdissatisfaction. Journalof the American Medical Association,299 (16), 1987-1993.