Him ORIGINAL CONTRIBUTION Educational Levels of Hospital Nurses and Surgical Patient Mortality Linda H. Aiken, PAD, RN URSE UNDERSTAFFING IS ranked by the public and physicians as one of the greatest threats to patient safety in US hospitals. Last year we re- ported the results of a study of 168 Pennsylvania hospitals showing that ‘each additional patient added to the av- erage workload of stall registered nurses (RNs) increased the risk of death fol- lowing common surgical procedures by 796, and that the risk of death was more than 30% higher in hospitals where nurses’ mean workloads were 8 pa- lients or more each shift than in hos- pitals where nurses cared for 4 or fewer patients? These findings are daunting given the widespread shortage of nurses, increasing concern about re- cruiting an adequate supply of new nurses to replace those expected to re- lire over the next 15 years,’ and con- strained hospital budgets. These find- ings also raise questions about whether characteristics ofthe hospital RN work- force other than ratios of nurses to pa lients are important in achieving ex cellent patient outcomes. Nurses constitute the surveillance system for early detection of compli- cations and problems in care, and they are in the best position to initiate ac- ons that minimize negative out ccomes for patients." That the exercise of clinical judgment by nurses, as well asstaffing adequacy, is key to effective surveillance may explain the link be- Context. Growing evidence suggests that nurse tafing alec the qulty of aren Fora, but ile known about whether te educational composton of epseed rues (RNS) in hospital related Yo patent outcomes. Objective To examine whether the proportion of hospital RNs educated atthe bac- Gatrents level or higher = asrocted win skeadjurted morsty and falure tore cue (deaths in surgi! patients with serious complications). Design, Setting, and Population Cros-zectonal analyses of outcomes dats for Pa2342 general, orthopece and varclr surgery patents Gecharged from 168 non- federal adult general Pennyivania hospitals Setween Apel 1.1998, and November 30, 1999, inked to administrative and survey data providing information on educa- ona compost, staffing, and other chrscterstes Main Outcome Measures. icacjusted patient mortal and fluretorexcue within 40 days of admission assocated with nurse educatona lve Results. The proportion of hospital RNs holding a bachelor’s degree orhigheranged from 02 to 7/% aro the horas, After adjusting for patient characterises and hosp structural characterises ale, teaching sats level of technology), as wel as fornurse staffing nurse experence and whetherthe patient surgeon vite board ce tied a 10% increase in te proporbon of ruses Holding a bachelor’s degree was serocited with 45% detent n both the lkelinood of patients dying within 90 days Of admission and the od of falar o rescue (odds ratio, 095; 99% confidence n= terval, 091-0.9 in bath cases). Conclusion In hospitals wth higher proportions of nurses educated atthe bacea- ISureste level or higher, surgical patent experienced lower mortal and fure-to- rescue rates, Iau 2osao0 617-1623 ven jamacom ‘oween higher nursing skill mix (ie, a about the benefits ifany, othe substan higher proportion of RNs among the tal growth in the numbers of nurses with nursing personnel of a hospital) and bachelor’s degrees. Indeed the conver patient outcomes." tional wisdom is that nurses’ experi- Registered nurses in the United States ence is more important than their edu- generally receive their basic education in cational levels 1 of 3 types of programs: 3-year di- ploma programs in hospitals, associate Aythor Affiliations: Center for Health Outcomes and gree ntsing pogrine in commas alvisnat cise er Sa Bree murs programs 18 Om a Oa mi colleen an Yacelsurcte nur eas eecDsstaa Ing programs colleges and wives, senescent et In 1950, 92% fnew RNsgradted Or ee trom hospital diploma programs," reece rpm tl tes bo, et oer ca Wherry 2001" ony 9 graded sete eS from hospital diploma programs, 61% Corresponding Author and Reprints: Linda H. Aiken, cern ob ance deg progenee cemamenueanttis Men and 36% were baccalaureate program — pfilageiphia, PA 19104-6096 (e-mal sken@nursing grates supranglyilegicown ts (©2003 American Medical Association, All rights reserved. (Reprinted) JAMA, eptember 24, 2093 Nol 290, No, 12. 1647 Downloaded From: on 01/07/2018NURSE EDUCATION AND PATIENT MORTALITY Despite the diversity of educational programs preparing RNs, and a logical (but unconfirmed) connection be- tween education and clinical judgment, lide ifanything is known about the im pctof nurses education on paticat out- comes." Results of some studies have suggested that baccalaureate-prepared nurses are more likely to demonstrate professional behaviors important to pa- tient safety suchas problem solving, per formance of complex functions, and el- fecive communication." However few siudies have examined the elect of nurse ‘education on patint outcomes, and thelr findings have been inconclusive.” The 168 Pennsylvania hospitals in- cluded in our previous tudy*ofpatient- to-nurse stalling and patient mortality varied substantially in the proportion of stall nurses holding baccalaureate oF higher degeces. This variability pro- vides an opportunity to conduct a simi lar study examining the association be- tween the educational composition of a hospital's RN stalf and patient out- comes. Specifically, we ested whether hospitals with higher proportions of di rectecare RNs educated at the bacea- laureate level orabove have lower risk- adjusted mortality rates and lower rates of failure to reseue (deaths in patients with serious complications). We also examined whether the educational backgrounds of hospital RNsare a pre- dlictor of patient mortality beyond fac- tors stch as nurse stalfing and experi- cence, These findings oller insights into the potential benelits ofa more highly educated nurse workforce. METHODS: Data Sources, and Variables \Weanalyzed outcomes data derived from hospital discharge abstracts that wer merged with information on the char- acteristics of the treating hospitals, in- cluding unique data obtained from sur- vyeys of hospital nurses. The institutional review board ofthe University of Penn- sylvanta approved the study protocol Hospitals. The sample consisted of 168 (80%) of the 210 adult acute-care general hospitals operating in Pennsyl- ania in 1999 that (1) reported surgi- A618 JAMA September 24, 200% Vol 290, No. 12 (Reprinted) Downloaded From: on 01/07/2018 cal discharges to the Pennsylvania Health Care Cost Containment Coun- cil in the specific categories studied here, (2) had data on structural char- acteristics available from 2 external ad- ‘ministrative databases (American Hos- pital Association [AHA] annual survey"* and Pennsylvania Department of Health Hospital Questionnaire”), and (3) had at least 10 nurses responding to our questionnaire, which previous emp cal work demonstrated was sufficient to provide reliable estimates of survey- based organizational characteristics of the hospitals, Six of the excluded hos- pitals were Veterans Affairs hospitals, Which do not report discharge data to the state. Twenty-six hospitals were e cluded because of missing data, most often because their reporting to exter- nal administrative sources was done as aggregate multihospital entities, Ten small hospitals, most of which had 50 or fewer beds, had an insufficient num- ber of nurses responding to the ques Udonnaire to be included, ‘A 50% random sample of RNs resid ing in Pennsylvania and on the rolls of the Pennsylvania Board of Nursing te ceived questionnaires att the spring of 1909, Surveys were com- pleted by 10184 nurses, an average of ‘more than 60 nurses per hospital, and the 52% response rale compares favor- ably with other voluntary, anonymous surveys of health professionals."” We compared our data with information from the AHA annual survey and found that the numberof nurses from each hos- pital responding to our survey was di- reclly proportional tothe number of RN positions in each hospital. This sug- gests similar response rates across hos- pitals and no response bias at the hos- pital level. Moreover, demographic characteristics of the respondents par- alleled those of Pennsylvania hospital nurses in the National Sample Survey of Registered Nurses."! For example, the mean ages of Pennsylvania hospital nurses in our sample and in the Na- ional Sample Survey of Registered Nurses were 40 and 41 years, respec tively; the percentages of Pennsylvania hospital nurses working full-time were ir homes in (66% and 69%, respectively; and those hhaving earned bachelor of science in nnursing (BSN) degrees were 30% and 31%, respectively Hospital stall nurses were asked to indicate whether their highest crede lial in nursing was @ hospital school diploma, an associate degree, a bach- lors degree,amastersdegree, or another degree. The proportion of nursesineach hospital who held each type of ereden- lial was computed. Because the educa Lional preparation of the 4.3% of nurses who checked “other” wasunknown, thelr answers were not included in our hos- pital-level measures of educational qual Fications. It was later verified that this deciston did not bias the results. Because there was no evidence that the relative proportions of nurses holding diplomas and associate degrees allected the patient ‘outcomes studied, those 2 categories of nurses were collapsed into a single cat- cegory and the educational composition ‘of the hospital staff was characterized in terms of the percentage of nurses hold ing bachelor’s or master's degrees. Two further variables were derived from the nurse survey. Nursing work- load was computed as the mean num- berof patients assigned toall staff nurses who reported caring for at least 1 but Fewer than 20 patients on the last shift, they worked. Because nurse experience ‘was an important potential confound ing variable related to both clinical judg- ment and education, the mean number of years of experience working asan RN for nurses from each hospital was also caleulated and used in the analyses. Three hospital characteristies were used as control variables: size, teaching status, and technology. Hospital-level data were obtained from the 1999 AHA annual survey and the 1999 Pennsylva- nia Department of Health Hospital Sur vey. Three size categories (<100 beds, 101-250 beds, =251 beds) were used. Hospitals without any postgraduate medical residents or fellows (nonteach- ing) were distinguished from those with 1:4 orsmaller trainee-to-bed ratios (mi- nor teaching) and those with ratios ‘higher than 1:4 (major teaching). High- technology hospitals were those that had (©2003 American Medical Association, All rights reserved.facilities for ether open-heart surgery, major organ transplantations, o both, Patients and Patient Outcomes Di charge abstracts for the universe of 252542 patients aged 20 085 years who underwent general surgical, orthope- dic, or vascular procedures from Apel! 1, 1998, to Novernber 30, 1999, in the 168, nonfederal hospitals were obiained from the Pennsylvania Health Care Cost Con- tainment Counell, which checks thedata for completeness and quality. Ais of the dlagnosis related groups studied was pro- vided previously? We examined the association be- tween the educational attainments of nurses across hospitals and both deaths ‘within 30 days of hospital admission (dee rived by linking discharge abstract data and Pennsylvania vital statisti data and deaths within 30 days of admission among patients who experienced com- plications (Failure to rescue). Patient ‘complications were determined with i lemational Classification of Diseases, Ninth Revision, Clinical Modification (ICD- CM) codes in the secondary diagnosis and procedure fields of discharge ab- siracis indicative of 39 clinical events s- ing protacols drawing on expert con- sensits as well as empirical evidence to distinguish complications fom preex- isting comorbidities The 2 patient outcomes stadied were risk-adjusted by including 133 vari- ablesinour models, including ae, sex, whether the admission was a transler from another hospital, whethceit was an cmergency admission, aseries of 48 vari ables indicating surgery type, dummy variables indicating the presence of 28 chronic preexisting conditions as cls- sified by ICD-9-CM codes, and interac- lion terms chosen on the basis of their ability predict morality and failure to rescuin the present dataset. Construc- lion ofthe patient risk adjustinent miod- cls used an approach similar to that re- ported by silber and colleagues The statistic for the mortality sk adjuste ment model was 0.89 and for the fail- ure to reseue model, 081 We also estimated and controlled for the elfect of having a board-certified sur ‘geon on risk for mortality and failure to (©2003 American Medical Association, All rights reserved, Downloaded From: on 01/07/2018 ‘NURSE EDUCATION AND PATIENT MORTALITY rescue, Fr each patient, the license num ber of the operating physician of record was matched to a physician's name us- ing a public use file rom the Pennsyl- vania Bureau of Professional and Oceu- pational Affairs, and subsequently t0 records from the American Board of Medical Specialties directory of board- certified medical specialists” A dummy variable was constructed to indicate whether oF not the operating physician was board-certified in general surgery or another surgical specialty. A second dummy variable was used to identi pa tients (8% ofall patients) with operat- ing physicians whose license numbers could not be linked to names to dete ‘mine board-certification status. Use of these 2 variables in tandem produced a reasonable way of controlling for sur- geon qualifications in our models. Data Analysis Descriptive statistics (means, SDs, and percentages) and significance tests (x! and F tests) were computed to compare groups of hospitals that varied in terms oftheir educational composition on hos- pital characteristics, including nurse ex- perience and nurse staffing, and patient characteristics, Logistic regression mod- cls were used to estimate the effects of 10% increase inthe proportion of nurses who had a bachelor's oF master's degree fon patient mortality and failure to res- cue, and to estimate the effects of nurse stalling, nurse experience, and surgeon board certification. The associations of educational composition, stalling, expe- rience of nurses, and surgeon board cer- ification with patient outcomes were computed before and after controlling for patient characteristics (demographic characteristics, nature of the hospital d= mission, comorbidities, and relevant in- teraction terms) and hospital character- Isties (bed size, teaching status, and technology). To account for the clustering of p= tients within hospitals in oursample, all model estimates were computed using Huber-White (robust) procedures to ad just the SEs of the estimated para- meters, Direct standardization esti- mates derived from the final model are presented to indicate the size of the ef Feets of educational composition of nurs- ing staff independently of and jointly ‘with nurse staffing levels. With all pa- Lents and using the final fully adjusted models for predicting death and failure to rescue, the probabilities of poor out ccomes were caleulated for patients in hospitals assuming that 20%, 40%, and (60% ofthe hospital RNs held bachelor's lor mastersdegrees and under various pa- tient-to-nurse ratios (4, 6, and 8 pa tients per nurse), with all other patient and hospital characteristics un- changed. Allanalyses were conducted using STATA version 7.0 (STATA Corp, College Station, Tex), using P<.05 as the level of statistical significance. RESULTS Characteristics of Hospitals and Patients TABLE 1 provides information on char- acteristics of the 168 hospitals in our sample, About 19% of the hospitals had more than 250 beds, 36% were teach- ing hospitals, and 28% had high- technology facilities. Across all hos} tals, nurses had a mean (SD) of 14.2(2.7) years of experience and a mean (SD) ‘workload on their last shit of 5.7 (1.1) patients, The proportion of staf nurses ‘with bachelor’s degrees or higher grees ranged from 0% to 70% across the hospitals. In 20% of the hospitals (34/ 168) less than 20% of staff nurses had BSN or higher degrees, while in 11% of the hospitals (10/168) 50% or more of the nurses had BSN or higher degrees. Hospitals with higher percentages of nurses with BSN or master's degrees tended to be larger and have postgrad ate medical raining programs, as well as high-technology facilities. Hospitals with higher proportions of bacealaureate-and rmusters-prepared nurses also had slightly less experienced nurses on average and significantly lower mean workloads. The strong association between the educa- tional composition of hospitals and other hospital characteristics, including nurse ‘workloads, makes clear the need to con- trol for these later characteristis ines Limating the effects of nurse education fon patient mortality (Reprinted) JAMA, September 24, 2093 Nol 290, No, 12. 1619