The Association Between Nursing Factors and Patient Morality in the Veterans Health Administration

Association between In-Hospital Mortality and Nurse Staffing in VA Anne Sales, MSN PhD RN Faculty of Nursing University of Alberta Edmonton, Alberta June 5, 2007 Three year study   Funded by VA Health Services Research and Development Service, IIR 01-160 Project team included       Yu-Fang Li, PhD RN Nancy Sharp, PhD Gwen Greiner, MSW MPH Elliott Lowy, PhD Julie Sochalski, PhD RN Pamela Mitchell, PhD RN Background  Grounded in questions related to the effect of nursing on patient outcomes   Considerable number of prior studies in this area Issues related to quality and scope of evidence regarding the effect of nursing practice Cf. Lang et al. 2004 “A systematic review of the effects of nurse staffing on patient, nurse employee, and hospital outcomes” Journal of Nursing Administration 34 (7/8): 326-337 Key prior literature  Aiken et al.    Blegen et al.  JAMA 2002: Effects of RN staffing on surgical patient mortality and staff burnout and satisfaction at the hospital level JAMA 2003: Effects of BSN proportion on surgical patient mortality at the hospital level Nursing Economics and Nursing Research 1998: multisite unit-level study of staffing and patient occurrences  Estabrooks et al.   Mark et al.  Nursing Research 2005: Multi-level modeling of the relationship between hospitallevel nursing characteristics and 30 day mortality Health Services Research 2004: Longitudinal analysis of relationship between nurse staffing and patient outcomes over six years at the hospital level NEJM 2002: Association of nurse staffing with mortality, length of stay and development of complications at the hospital level  Needleman, Buerhaus et al.  Goals of the NSPO in VA study   To assess the relationship between nursing factors and patient outcomes at the nursing unit level Nursing factors include  Staffing, skill mix, perceptions of practice environment, job satisfaction, burnout Mortality, failure to rescue/conditional mortality, length of stay, satisfaction, complications  Patient outcomes include  Today’s talk    Focus on relationship between staffing, skill mix, and mortality Present hospital and unit level data Reflect on major questions this stream of research addresses Design  Retrospective, observational study    Large national data sets from VA Survey of all nursing personnel working in VA hospitals with acute inpatient care Followed design and approach of both the International Hospital Outcomes Consortium study and the Needleman and Buerhaus study in the US Context: the Veterans Health Administration  Single largest vertically and horizontally integrated health care system in the US (2006)       155 hospitals in each of the 50 states + District of Columbia and Puerto Rico Over 800 outpatient clinics 135 nursing homes 46 residential rehabilitation treatment centers Over 200 readjustment counseling centers 5 million unique users and 54 million outpatient visits   Annual budget of $35 billion in 2007 Divided into 21 regional networks First study ever conducted in VHA at the nursing unit level  Included 125 hospitals with acute inpatient care   Over 1900 units total with over 170,000 patients ~1200 inpatient units  Includes psychiatry and non-acute units such as rehab ~115 ICUs and 230 Med/Surg (acute non-intensive) units  430 acute inpatient units   126,382 patients with unique admissions between February and June 2003 Nursing personnel survey    Used instrument based on IHOC study (Aiken PI) Distributed through hospital nursing services to over 44,000 nursing personnel Overall response rate 26.4%  RN response rate ~30% ~7,000 RN responses  Over 11,000 responses  Staffing and skill mix data  Initially planned to obtain from survey responses  Low response rate  Used newly developed VA national staffing/cost database with unit level data   Able to match approximately 430 inpatient acute care units Provided hours worked for RNs, LPNs, unlicensed providers, and contract hours Key variables in our analysis  Hours per patient day   Aggregated all patient stays over the five month period February-June 2003 Aggregated nursing hours by type of nurse over the same period  RN, LPN, UAP (primarily nurse aides in the inpatient setting)  Divided nursing hours by patient days   Skill mix is the proportion of RN hours to total hours over the five month period Both constructed at both the hospital and unit level Definitional issues  Measure of staffing is not uniform across studies   Partially driven by the available data Partially driven by analytic preference 24 hours per patient day of care = 1:1 nurse: patient ratio  12 HPPD = 1:2  8 HPPD = 1:3  2 HPPD = 1:12  Analysis strategy  Dependent variables  In-hospital mortality All analyses are at the patient level For facility level analysis, standard errors are corrected for hospital level auto-correlation for hospital analyses For unit level analysis, standard errors are corrected for hospital and unit level auto-correlation for unit analyses  Multi-level modeling appropriate to analysis     Mortality models are estimated using general linear latent and mixed model (gllamm) estimation In-hospital mortality  Facility level in-hospital mortality ranges from 0 – 11%   Mean 2.9%, s.d. 1.1% Median 2.7%, IQR 2.1 – 3.5  Unit level in-hospital mortality ranges from 0 – 29%   Mean 2.9%, s.d. 3.2% Median 1.8%, IQR 0.8 – 3.6 Facility and unit level results are very different Facility level with RN hours and skill mix Unit level with RN hours and skill mix Odds ratio 1.20 p-value <0.01 Lower limit 95% CI 1.19 Upper limit 95% CI 1.21 Odds ratio 1.20 p-value <0.01 Lower limit 95% CI 1.19 Upper limit 95% CI 1.21 Probability that patient experienced a major complication Patient had medical DRG Facility Acute Care Case-Mix for prior year RN hours per patient day RN skill mix (RN hours/total nursing hours) 1.92 1.18 0.91 <0.01 0.44 <0.01 1.75 0.78 0.88 2.10 1.78 0.95 1.75 0.88 1.01 <0.01 0.63 0.22 1.58 0.51 0.99 1.92 1.50 1.04 0.99 0.02 0.98 Reference group 1.00 0.66 0.27 0.32 Reference group 1.37 Admitted to facility with Level 1 ICU 0.99 0.94 Admitted to facility with Level 2 ICU Admitted to facility with Level 3 ICU 0.85 1.16 1.29 0.04 1.01 1.64 0.76 0.02 0.60 0.96 0.90 0.48 0.67 1.21 Admitted to facility with Level 4 ICU 0.76 0.87 0.07 0.44 0.57 0.60 1.02 1.24 0.83 1.00 0.30 0.99 0.59 0.61 1.17 1.63 Admitted to facility with no ICU Admitted to facility where >50% RNs have BSN or higher Patient had ICU stay 0.96 1.55 0.56 <0.01 0.84 1.39 1.10 1.71 1.03 1.51 0.76 0.03 0.86 1.05 1.22 2.17 Effect of nursing care depends on patient risk group Unit level with RN hours and skill mix: Patients with ICU stay Unit level with RN hours and skill mix: Patients with no ICU stay Odds ratio 1.16 p-value <0.01 Lower limit 95% CI 1.15 Upper limit 95% CI 1.17 Odds ratio 1.49 p-value <0.01 Lower limit 95% CI 1.45 Upper limit 95% CI 1.53 Probability that patient experienced a major complication Patient had medical DRG Facility Acute Care Case-Mix for prior year RN hours per patient day RN skill mix (RN hours/total nursing hours) 1.69 0.82 1.02 1.54 <0.01 0.54 0.13 0.40 1.51 0.43 0.99 0.56 Reference group 1.89 1.56 1.04 4.25 2.04 0.84 0.89 1.03 <0.01 0.57 0.01 0.96 1.66 0.45 0.82 0.35 Reference group 2.52 1.55 0.97 3.02 Admitted to facility with Level 1 ICU 1.22 0.14 Admitted to facility with Level 2 ICU 0.94 1.58 1.28 0.09 0.96 1.69 Admitted to facility with Level 3 ICU 0.65 0.01 0.46 0.91 1.39 0.08 0.96 2.01 Admitted to facility with Level 4 ICU 0.56 0.79 1.03 <0.01 0.39 0.74 N/A 0.38 0.47 0.86 0.82 1.35 1.24 1.17 1.05 0.97 0.45 0.87 0.78 N/A 0.77 0.62 0.79 1.77 1.77 1.20 Admitted to facility with no ICU Admitted to facility where >50% RNs have BSN or higher Patient had ICU stay Results: The bottom line    There are considerable differences in findings between hospital and unit level analysis A primary driver of those differences is the difference among patients in their risk of complications Mixing patients who had any ICU stay with those without an ICU stay creates tremendous heterogeneity Policy implications  If our goal is to improve in-hospital mortality, where should we add RN hours?  From this analysis, adding RN hours in acute non-ICU units will probably yield more benefit than in ICU