Measuring Direct Nursing Cost Per Patient in the Acute Care Setting

Transcription

1 JONA Volume 44, Number 5, pp Copyright B 2014 Wolters Kluwer Health Lippincott Williams & Wilkins THE JOURNAL OF NURSING ADMINISTRATION Measuring Direct Nursing Cost Per Patient in the Acute Care Setting Peggy Jenkins, PhD, RN John Welton, PhD, RN OBJECTIVE: The objective of the study was to measure the variability of direct nursing cost for similar patients and to examine the characteristics of nurses assigned to different types of patients. BACKGROUND: There is no standard method for measuring direct nursing cost by patient. METHODS: Deidentified data were collected from 3 databases for patients admitted from January 2010 through December 2012 on 1 medical/surgical unit in a large Magnet A hospital. Direct nursing care time and costs were calculated from the nurse-patient assignment. RESULTS: Variability in nursing intensity ( hours) and cost per patient day ($132-$1,455) was significant for similar patients. Higher cost nurses were not assigned sicker patients (F 3, 3029 = 87.09, P G.001, R 2 = 0.124). Mean (SD) nursing direct cost per day was $96.48 ($55.73). CONCLUSIONS: Standard measurement of nursing cost per patient could be benchmarked across hospitals and inform nursing administration care delivery decisions. How much does nursing cost in the hospital? The simple answer is we do not know. The longstanding method of averaging all nursing labor hours and costs among all patients within a cost center, commonly termed nursing hours or costs per patient day (NHPPD/NCPPD), has several weaknesses. First and foremost, the individual Author Affiliations: Senior Director (Dr Jenkins), VHA, Inc, Irving, Texas; Professor, Director of PhD Program and Senior Scientist (Dr Welton), University of Colorado College of Nursing, Aurora. The authors declare no conflicts of interest. Correspondence: Dr Jenkins, 9506 E 4th Ave, Denver, CO (peggy.jenkins@ucdenver.edu). Supplemental digital content is available for this article. Direct URL citations appear in the printed text and are provided in the HTML and PDF versions of this article on the journal s Web site ( DOI: /NNA variation of nursing time (intensity) and cost per patient is unknown. Because hospitals are paid by the diagnosis or procedure, there is no way to identify the nursing cost structure relative to reimbursement for care. The purpose of the study was to examine the variability of direct nursing care time and costs per patient in a hospital setting and to investigate the relationships among patient characteristics, nurse characteristics, nursing intensity, and nursing cost. The study used the nurse-patient assignment captured in the electronic health record (EHR). This is a rapidly emerging capability at US hospitals and poses some interesting possibilities of measuring nursing care at the individual nurse-patient level rather than aggregate or average time and cost measures. Background No standard method to measure patient-level nursing time or cost is used today despite evidence spanning decades, suggesting the importance of such measurement. 1-6 Contemporary nurse staffing software provides a repository for data collected at the individual patient and nurse level. Patient assignment data offer a new source for nursing cost research in which actual nurse assignments can be used to derive actual care hours and costs for each patient rather than NHPPD or NCPPD. These new data can provide additional evidence about optimum staffing levels and patient care assignments. A standardized methodology for comparing nursing cost is necessary to benchmark costs across units and hospitals. Patient assignment data are available at many hospitals today stored in the EHR. 7,8 Direct nursing care time, also commonly referred to as nursing intensity, is a more precise measure how nursing care can be allocated to each patient compared to NHPPD. 7 Nursing intensity can be multiplied by either a standard JONA Vol. 44, No. 5 May

2 nursing wage per hour or the actual wage for each nurse to provide a nursing cost metric for each patient separate from the NCPPD. This study used data from the EHR in combination with the nurse-patient assignment data to examine differences in nursing intensity and costs among different groups of patients in a diagnosis-related group (DRG). We also examined assignment patterns in relationship to patient characteristics and nursing acuity to assess how nurse characteristics such as academic preparation and experience are matched to patient needs and the associated costs of that care. The research questions were the following: 1. What is the variation in nursing cost per acute care episode (NCACE) for patients with the same DRGs with and without complications? 2. What are the relationships among nursing characteristics (years of service, educational degree) on NCACE? 3. What are the characteristics of nurses assigned to patients with complications and major complications? 4. What is the average nursing cost per day (NCDAY) measured at the patient level? Methods Study Design, Setting, and Sample The study consisted of all patients admitted to a single inpatient unit at an academic medical center from January 2010 to December 2012 comprising 3111 patients and 150 nursing personnel. There were 1051 patients excluded from the final analysis because they were transferred to other nursing units such as intensive care unit (ICU). Also, 427 patients with missing nursing intensity data were excluded from analysis. Data Acquisition and Management The primary source of data was the Clairvia Care Value Management Patient Assignment software plus medical management and human resources databases. 9 All protected health identifiers as well as time and date data were eliminated before submission to the investigators. Relevant patient and nurse identifiers used for linking files were anonymized. Procedure for Handling Missing Data Data were systematically analyzed for patterns of missing data. Microeconomic studies typically use only the original and not imputed data. 10 Cases with missing nursing intensity observations when a nurse should have been available were dropped. For instance, an RN should be assigned each shift and if no RN was assigned during a shift, the case was dropped from analysis because cost per acute care episode would be minimized in cases with missing data. Imputing averages would bias the sample and would have been burdensome given the multiple variables tied to each identifier. Variables A complete list of variables can be located in Table, Supplemental Digital Content 1, JONA/A303. Forty-five variables were collected and cleansed, and new variables were constructed for a total of 49 final study variables. Predictor variables included patient and nurse characteristicssuchasdrg, complication code, age, gender, years of service, float nurse indicator and education level. For each patient included in the analysis, NCDAY and episode of care variables were constructed by summing the product of nursing intensity per shift multiplied by actual nursing wage. Nursing intensity was calculated in the patient assignment software based on the following methodology. The acuity score was a 1-to-12-point scale derived from the Nursing Outcomes Classification based on the Pesut and Herman conceptual framework. 11 Nurses rated patients every shift or when condition changed on several outcome measures using a 1-to-5-point scale that contributed to an algorithm producing the nursing intensity score. A monthly audit was completed to ensure interrater reliability of acuity measurement, and the result was 86% agreement. Human Subject Compliance Expedited approval and continuation was obtained from Colorado multiple institutional review board (IRB) (COMIRB) as well as the study site IRB. Results Nurse Personnel Characteristics Table 1 displays the descriptive statistics in the study population by RN and non-rn. Eight nurses did not identify education level, so the assumption was made that they were not bachelor s of science in nursing (BSN) prepared. The range of actual wage is not shown because of the sensitivity of sharing nursing wage data. Patient Characteristics There were 3,111 patients in the study. Women comprised 54% of the sample (n = 1,688), and men, 46% (n = 1,423). The mean age of patients was 55 years (range, years). Total days in the hospital ranged from 0 to 25 days (mean, 3.64 days). Length of stay (LOS) ranged from 4.43 to hours (mean [SD], [69.13] hours). Patients stayed on the study unit an average of hours (range, hours). There were 247 unique DRGs, with the most common DRGs 392 esophagitis, gastroenterology, and 258 JONA Vol. 44, No. 5 May 2014

3 Table 1. Study Unit Nurse Personnel Characteristics (n = 150) Characteristics RN (n = 101) Non-RN (n = 49) Age 41.1 (11.2) (13.1) Years service organization 8.1 (4.6) (4.3) 2-13 Years service unit 6.1 (4.2) (4.2) Wage, mean (SD) $31.21 ($7.16) $14.19 ($4.81) Female 92 (91) 41 (84) BSN 41 (41) NA Clinical ladder 2 (2) NA Certification 10 (10) NA Float a 62 (41) 27 (18) Data are presented as mean (SD) range or n (%), unless otherwise indicated. a Only 4% of total shifts were staffed by float personnel. miscellaneous digestive disorder w/o major complications (MCC) (n = 165, 5%), 313 chest pain (n = 129, 4%), 191 chronic obstructive pulmonary disease (COPD) w/ complications (CC) (n = 92, 3%), 192 COPD w/o CC/MCC (n = 90, 3%). Female patients had fewer discharge DRGs without complications (n = 752, 58%). Men had higher frequency of discharge DRGs with major complications (n = 284, 59%). Nursing Intensity Variable Summary statistics for nursing intensity by complication code were calculated. As noted in Table 2, mean nursing intensity increased as the DRG complication code increased. Nursing Cost Length of stay, NCACE, nursing intensity per day (DAYNINT), and NCDAY for patients with similar DRGs are presented in Table 2. The log of NCACE was regressed on the log of total aggregated years of nurse experience. Logarithm is a common transformation for data that are positively skewed, minimizing variance in the high end of the distribution, hence normalizing the data. 12 In ordinary least squares regression, the arithmetic mean is used to predict y (response variable), and in log transformation, the geometricmeanisusedtopredicty. 13 The model was significant (F 1, 3090 = 18,242, P G.001; R 2 =0.86),or by patient episode, years of RN experience explains 86% of the variance in NCACE. Two additional models added BSN and float variables with minimal effect (Table 3). Relationship Among Nursing Characteristics and Patient Acuity Average nursing acuity for each patient episode was regressed on nurse characteristics. Two patient characteristics, LOS and age, were added as control variables. The model was significant (F 3, 3029 = 87.09; P G.001; R 2 = 0.124). Registered nurse years of experience was not significantly associated with patient acuity in the model. Number of float nurses and BSN were significant (P =.001). With each additional BSN nurse assigned during an acute care episode, the predicted acuity increased by 0.002%. Each additional float nurse assigned was associated with a 7% increase in acuity (Table 4). Discussion The study demonstrated statistically significant variability in direct nursing cost by day and acute care Table 2. Variability of Nursing Cost by DRG DRG LOS, d NCACE, $ DAYNINT NCDAY, $ 190 COPD w/ MCC 4.43 (1.99) 462 (316) 3.33 (1.75) (56.03) 191 COPD w/ CC 4.04 (2.98) 408 (427) 3.20 (1.71) (51.59) 192 COPD w/o CC/MCC 3.57 (2.11) 325 (242) 2.87 (1.56) (45.76) 291 Heart failure and shock w/ MCC 5.83 (3.54) 676 (507) 3.70 (1.79) (56.54) 292 Heart failure and shock w/ CC 4.44 (2.62) 465 (352) 3.31 (1.60) (48.31) 293 Heart failure and shock w/o CC/MCC 3.28 (1.6) 299 (196) 2.76 (1.62) (45.50) 682 Renal failure w/ MCC 6.06 (4.92) 748 (773) 3.94 (1.95) (58.97) 683 Renal failure w/ CC 4.19 (2.47) 464 (451) 3.40 (1.90) (60.62) 684 Renal failure w/o CC/MCC 2.99 (1.18) 338 (187) 3.39 (1.82) (49.53) Data are presented as mean (SD). JONA Vol. 44, No. 5 May

4 Table 3. Summary of Regression Models of Nursing Cost for Patient Acute Episode Variable Model 3 Model 2 Model 1 RN years org a 1.01 a a % BSN a a Float a _constant a 1.09 a 1.26 a r2_a rmse Model 1 regresses nursing cost on RN years of experience, model 2 adds percentage BSN, and model 3 adds float nurses. a P G.001. episode for patients with similar DRGs. Nursing intensity and costs increased as severity of patients increased. More experienced or higher educated nurses do not raise nursing cost per patient episode substantially. Nurses with greater experience and education were not consistently assigned sicker patients. Associations Among Nurse Characteristics and Cost The elasticity of RN years of experience on nursing cost was measured. Elasticity is an economic measurement of a change in 1 variable in response to a change in another variable. In the double log regression model used, the coefficients are elasticities of the log of NCACE. 14 A 10% increase or 9.3 total nurse years of experience in the organization for the patient episode of care was associated with a 9.9% or, on average, $34.92 increase in direct cost of nursing care per episode for patients on the study unit holding all other variables constant. Experienced nurses add only a small increase in cost for a patient episode of care. Use of BSN nurses was not associated with an increased cost per patient. Assignment of Nurses The study did not find that nurses with greater experience are assigned sicker patients. On the study unit, BSN nurses were not assigned patients with more complications; however, increasing the number of BSN nurses would not increase the direct nursing cost per patient. This raises a number of questions. Should assignment of nurse to patients be based on patient needs? For example, should more experienced nurses or nurses with BSN degrees be assigned more complex or higher acuity patients? Our study did not address the relationship between nurse characteristics on outcomes of care. However, our findings indicate that there would be minimal additional costs per case when more experienced or BSN-prepared nurses care for sicker patients. Float nurses were associated with increase in patient acuity. Are float nurses used because acuity is higher or are float nurses assigned patients with higher acuity? Future research can help answer such questions. Implications for Nurse Leaders The method used in this study provides a clearer picture of variable nursing intensity and cost per patient, which is not demonstrated when measuring nursing costs averaged across patients on a unit. The NHPPD measurement is dated and was built before sophisticated patient assignment software. 15 Patient-level nursing intensity/cost data, rather than average NHPPD/ NCPPD, could lead to better decision making and provide a basis to build a business case for nursing care delivery systems. The use of the nurse-patient assignment could provide an alternative costing metric for nursing care. 16,17 The current study found differences in direct nursing cost by shift, day, and acute care Table 4. Regress Average Patient Acuity on Nursing Characteristics avacu Coefficient SE t P 9 t 95% Confidence Interval ptlos to ptage to avyrorg j to pbsn to aflt to _constant to Abbreviations: aflt, total float nurses per episode; avacu, average patient acuity; avyrorg, average nurse years in organization; ptage, patient age; ptlos, patient LOS; pbsn, percentage BSN. 260 JONA Vol. 44, No. 5 May 2014

5 episode level of analysis. This demonstrates the potential of these data to provide a range of business analytics that have not been available. Using patient-level data opens up the possibility to describe, explore, and test a broad range of research questions that cannot be studied using aggregate unitor hospital-level data. The findings from this study revealed that the patient assignment might not be based on experience or educational level of the nurse. Do hospitals assign nurses to patients considering the impact on quality and outcomes of care? The findings show nursing years of experience was positively associated with direct nursing cost in the study unit. However, nurses with more experience add little additional cost per case. Previous nursing research has provided evidence supporting the experience level of the nurse contributed to better outcomes, so the value of nursing might be maximized through use of more experienced nurses Could the use of experienced nurses lead to decreased LOS and, thus, lower the overall cost of care? What is the best use of nurses with more experience and education? Limitations The study was completed using data from a single unit in 1 organization; hence, the results of the study are not generalizable beyond the study unit. Deidentified data from a secondary source were used and cases with missing data were excluded in this study. Overtime and differential wage data were not obtained because of the extraction burden; therefore, the cost metric used in this study measured only actual wage and intensity per patient. Additional costs not measured in this study include unit indirect costs such as vacation time and sick leave, full-time benefits, nondirect care time such as unit meetings and in-service education, or other hospital nursing indirect costs such as nurse manager and executive salaries. The sample included only patients whose entire LOS was on the study unit. Patients who were in ICU or other nursing units were excluded. Only 1 unit used the assignment software during the study period, and as more units adopt patient assignment modules, additional studies can be designed to measure nursing cost for an episode of care including all nursing units the patient is on during hospitalization. In addition, transfers off the unit for procedures were not included in the study data and could be included in future research. Recommendation for Further Research Use of patient-level data opens up the study of nursing systems to a multitude of interesting research questions. As other inpatient settings use nurse assignment software, all hospitalized patients could be included in future studies. Comparing episode of care cost across units and hospitals could provide benchmarking to better understand the optimum use of nursing resources to affect quality patient outcomes. Future research could measure patient episode of care cost to include not only the acute care setting but also other nursing care delivery models. Linking cost data to quality data may provide better science on the value of nursing. Conclusion The study used data from electronically stored nursepatient assignments to calculate nursing intensity and nursing cost per patient shift, day, and episode of care based on the actual wage of the nurse caring for a particular patient. Large variation in nursing costs were found for patients with similar diagnoses/severity, and this finding is consistent with previous nursing research. The use of patient-level data can provide improved costing metrics over current use of NHPPD and NCPPD. As more hospitals use software to assist with nurse scheduling, staffing, and assignment to patients, it may be possible to consider national methods and standards to better measure and benchmark nursing intensity and costs and how nurse characteristics linked to individual patients can be used to identify optimum staffing and assignment patterns. Acknowledgments The author thanks Dr Amy Barton, Dr Joyce Verran, Dr Sharon Eck-Birmingham, Dr Jackie Jones for education, and Dr Esther Chipps, Valerie Moore, Louise Griffiths, Kraig McKinley for collaboration. References 1. Chiang B. Estimating nursing costsva methodological review. Int J Nurs Stud. 2009;46(5): Thompson JD, Diers D. DRGs and nursing intensity. Nurs HealthCare. 1985;6: Wilson L, Prescott PA, Aleksandrowicz L. Nursing: a major hospital cost component. Health Serv Res. 1988;22(6): Diers DJ, Bozzo J, and RIMS/Nursing Acuity Project Group. Nursing resource definition in DRGs. Nurs Econ. 1997;15(3): , Edwardson SR, Giovannetti PB. A review of cost-accounting methods for nursing services. Nurs Econ. 1987;3: Pappas SH. Describing costs related to nursing. J Nurs Adm. 2007;37(1): JONA Vol. 44, No. 5 May

6 7. Welton JM, Zone-Smith L, Bandyopadhyay D. Estimating nursing intensity and direct cost using the nurse-patient assignment. JNursAdm. 2009;39(6): Radwin LE, Cabral HJ, Chen L, Mowinski Jennings B. A protocol for capturing daily variability in nursing care. Nurs Econ. 2010;28(2): Caspers BA, Pickard B. Value-based resource management: a model for best value nursing care. Nurs Adm Q. 2013;37(2): Cameron AC, Trivedi PK. Microeconometrics Using Stata. College Station, TX: StataCorp LP; Eck Birmingham S, Nell K, Abe N. Determining staffing needs based on patient outcomes versus nursing interventions. In: Cowen PS, Moorhead S, eds. Current Issues in Nursing. Philadelphia, PA: Elsevier; 2010: Ferketich S, Verran J. An overview of data transformation. Res Nurs Health. 1994;17(5): Bruin J. Regression With Stata. Los Angeles, CA: UCLA, Statistical Consulting Group; stata/ado/analysis/. Accessed September 6, Baum CF. An Introduction to Modern Econometrics Using Stata. College Station, TX: StataCorp LP; Thompson JD, Diers D. Nursing resources. In: Gamache D, ed. DRGs: Their Design and Development.AnnArbor,MI:Health Administration Press; 1991: Pickard B, Eck Birmingham SA. Staffing and scheduling. In: Huber D, ed. Nursing Leadership. St Louis, MO: Elsevier; Eck Birmingham S. Evidence-based staffing: the next step. Nurs Lead. 2010;8: Aiken LH, Clarke SP, Cheung RB, Sloane DM, Silber JH. Educational levels of hospital nurses and surgical patient mortality. JAMA. 2003;290(12): Chang YK, Mark BA. Antecedents of severe and nonsevere medication errors. J Nurs Scholarsh. 2009;41: Doran DM, McGillis Hall L, Sidani S, et al. Nursing staff mix and patient outcome achievement: The mediating role of nurse communication. J Int Nurs Perspect. 2001;1(2-3): Tourangeau AE, Doran DM, McGillis Hall L, et al. Impact of hospital nursing care on 30-day mortality for acute medical patients. JAdvNurs. 2006;57(1): Blegen MA, Goode CJ, Park SH, Vaugh T, Spetz J. Baccalaureate education in nursing and patient outcomes. J Nurs Adm. 2013;43(2): JONA Vol. 44, No. 5 May 2014