Big Data in Healthcare: New Methods of Analysis

Abstract

With the ubiquitous availability of health-related data such as insurance claims, discharge abstracts, electronic health records, personal fitness devices or mobile phone applications, the amount of health data is increasing in size, but also in speed and in complexity. “Big data” provides new opportunities for nurse clinicians and researchers to improve patient health, health services and patient safety. Following this unprecedented amount and complexity of information available from different types of data sources, the processing and the analysis of big data challenges traditional analytical methods. For these reasons, a range of analytical approaches such as text mining and machine learning often developed in bioinformatics or engineering fields become of highest relevance to nurses wanting to work with big data. This chapter provides a brief overview of the main definitions and the analytical approaches of big data. The chapter gives two nursing research examples in the context of patient experience in cancer care and older people with dementia in nursing homes. In both cases the analytical approach (text mining and machine learning) is highly integrated into traditional research designs (a cross-sectional survey and a retrospective observational study), which highlights how traditional research designs become increasingly influenced by analytical strategies from big data or data science.

Case Study 5.1: Value-Based Nursing Care Model Development

• John M. Welton PhD, RN, FAAN &
• Ellen Harper DNP, RN-BC, MBA, FAAN 

1. Health Systems Research, University of Colorado College of Nursing, Denver, CO, USA

   John M. Welton PhD, RN, FAAN

2. School of Nursing, University of Minnesota, Minneapolis, MN, USA

   Ellen Harper DNP, RN-BC, MBA, FAAN

Abstract

This case study will describe the development of a national consensus model to measure patient-level nursing intensity and costs-per-patient in multiple care settings that support the continuum of care and produce objective measures of nursing value. Specifically, this case study will describe the creation of a common data dictionary that describes patient, nurse, and system-level data elements extracted from existing data sets to populate a conceptual model to measure nursing value.

Keywords

Data dictionary • Data models • Value-based nursing care • Business intelligence • Nursing care quality and costs • Room and board • Diagnosis related group • Florence Nightingale

Jennifer hurried to finish her charting at the end of a busy night shift in the cardiac stepdown unit (CSU). The last item on her list was to complete her bill for each of the 4 patients she was assigned. This is a new change as her hospital recently implemented a value-based nursing care model. Billing for nursing care is a way to link individual nurses with each patient to identify the unique resources expended for each patient and use these data internally to allocate nursing time and costs. At 8:00 a.m. Jennifer attended her monthly practice council. Selected nurses from the CSU reviewed the overall patient care, adverse events, patient satisfaction and patient level nursing costs. In the past six months, the CSU was able to reduce length of stay and nursing care costs for congestive heart failure patients, their top diagnosis, by assigning a more experienced nurse on admission who often had a reduced assignment. While the first-day costs were higher than average, the nursing business intelligence analytics clearly demonstrated better outcomes by improving nursing care in the vulnerable first 24 hours after admission. Lastly, the nurses reviewed the real-time quality metrics assigned to each nurse. The CSU nursing performance metrics included medication administration delays, pain management, and glycemic control. Each nurse was rated by using time and event stamped data from the electronic health records: for example when a medication was due and the time difference for when it was administered. Jennifer smiled when she saw all her scores had improved from last month mostly from being better organized and “keeping her head in the game.”

This is a fictional story. However emerging new techniques for extracting data from the electronic health record (EHR) can identify the added value of nursing care as well as the individual contribution of each nurse. Nursing care value, in its simplest form, is the relationship between quality and costs, or quality and outcomes of care (Pappas 2013; Simpson 2013). When nursing care is appropriate and optimum, adverse events such as injuries, pressure ulcers, infections, and medication errors are reduced thereby decreasing the added costs associated with morbidity and mortality (Spetz et al. 2013; Staggs and Dunton 2014; Yakusheva et al. 2014).

In the current environment, quality and outcomes of care are measured at the individual patient level and aggregated across many patients within an identifiable entity such as a hospital inpatient unit or skilled nursing facility. The actual or true costs of nursing care are not directly measured for each patient and typically averaged across many nurses and many patients (Sanford 2010). Nursing care is rolled up to daily room and board charges (Thompson and Diers 1991), which hides the added value nurses bring to the bedside. Without a direct way to measure the actual or “true” cost and resources expended by nurses for each patient, it will be impossible to measure nursing care value (Welton 2010).

1.1 5.1.1 Value-Based Nursing Care and Big Data

In attempting to arrive at the truth, I have applied everywhere for information, but in scarcely an instance have I been able to obtain hospital records fit for any purposes of comparison. If they could be obtained, they would enable us to decide many other questions besides the one alluded to. They would show subscribers how their money was being spent, what amount of good was really being done with it, or whether the money was not doing mischief rather than good; they would tell us the exact sanitary state of every hospital and of every ward in it, where to seek for causes of insalubrity and their nature; and, if wisely used, these improved statistics would tell us more of the relative value of particular operations and modes of treatment than we have any means of ascertaining at present. They would enable us, besides, to ascertain the influence of the hospital with its numerous diseased inmates, its overcrowded and possibly ill-ventilated wards, its bad site, bad drainage, impure water, and want of cleanliness—or the reverse of all these—upon the general course of operations and diseases passing through its wards; and the truth thus ascertained would enable us to save life and suffering, and to improve the treatment and management of the sick and maimed poor.

Florence (Nightingale 1863, p. 176)

If we only had the data … Nightingale’s lament may be coming close to realization. A group of nurses and other professionals began meeting in June 2013 at the University Of Minnesota School Of Nursing to address the problem of growing amounts of healthcare and nursing data (Clancy et al. 2014; Westra et al. 2015a). An action plan committed nearly 100 attendees to address a wide range of issues related to data science, informatics, and how to leverage the burgeoning amounts of information contained with the EHRs to develop new approaches, methods, and analytics that ultimately will improve patient care outcomes and decrease costs (Westra et al. 2015b).

One expert workgroup was formed to address the issue of how to measure nursing value and develop new techniques that will provide real-time metrics to monitor quality, costs, performance, effectiveness, and efficiency of nursing care (Welton 2015). During an initial one-year interaction, members of the nursing value expert workgroup identified core issues needed to explicate nursing value (Pappas and Welton 2015; Welton and Harper 2015):

• 1. Identify individual nurses as providers of care

• 2. Define nursing care as the relationship between an individual nurse and patient, family, or community

• 3. Link nurses directly to patients within the EHRs and measure value at each unique nurse encounter

• 4. Identify nurse and nursing care performance at the patient, unit and hospital or business entity unit of analysis

• 5. Develop patient level nursing costs based on the direct care time and other measures of resources or services provided to patients (or families, communities)

The value-based nursing care model focuses on the individual nurse rather than nursing care as the basic unit of analysis. The model is software agnostic and setting neutral. The primary analytic approach is to use events and time-stamps to link nurses and patients. For example, in the vignette, Jennifer organized her care delivered to her four patients as distinct services, interventions, assessments, etc. Each patient has unique needs and these vary across the trajectory of an illness or episode of care such as a hospitalization. The ability to discern differences in time spent with individual patients as well as the associated dollars expended across a patient population provides much greater detail and more timely and actionable information about nursing care and added value that can be used for clinical and operational decision making.

Because the value-based nursing care model is focused on individual nurses, performance can be measured for each nurse using EHR data. In the vignette, the CSU is focusing on pain management. Each nurse conducts assessments, identifies problems (e.g., acute pain), provides interventions such as administering PRN opioid medication, and reassesses a patient’s response to interventions. Unit practice guidelines can be used to develop useful and objective information. For example, if the standard of care in the CSU is assessing for pain every 4 hours and follow up within 30 minutes after an intervention, extraction of nursing assessment documentation time data and pain acuity scores as well as the time and dose of PRN medications can be used to identify practice guideline adherence, patterns in using PRN opioids, and overall response to a nurse’s care for patients in pain. These data can be posted or used in value-based nursing performance metrics shared within the unit.

1.1.1 5.1.1.1 Extracting Nursing Data from the EHR

One of the vexing problems in building new patient and nurse level analytic models is the difficulty in finding and extracting key data from the many tables in a modern EHR and developing ways to do this across multiple software platforms. Substantial resources are needed for even simple data inquiries and reports. If hospitals or other health care settings wish to compare results and information across multiple settings, a common method and model for extracting similar data is needed.

Part of the efforts of the value-based nursing care expert workgroup is to develop a common data model that can allow multiple healthcare settings to extract similar nursing related data. The model is a roadmap for information technology and business analyst professionals to develop extraction code and pull data into a common repository for planned or ad hoc analysis. A preliminary model has been proposed by the value-based nursing care expert workgroup (Welton & Harper 2016). The common data model allows extraction and collection of complex nursing related data across many different software platforms and settings (Fig. 5.1.1). Ultimately, this common data model provides a framework for using and analyzing data about nursing care in many different settings and across many different nurses.

1.1.2 5.1.1.2 Nursing Business Intelligence and Analytics (NBIA)

The data model provides a template as well as the key data that can be used in complex analysis and business intelligence efforts. For example, future systems will be able to monitor performance of nurses administering medications by deriving the time between when a medication was due and when administered using bar code technology (Welton 2013). Pattern detection algorithms will be able to detect when medication administration is becoming increasingly delayed due to high workload, which may be a precursor to late medication doses. Medication administration times can be used to analyze the relationship between unit churn, patient acuity and medication administration performance. Individual nurse performance can be monitored and specific questions addressed that may indicate difficulty in meeting clinical needs due to high complexity and the amount of drugs administered (Kalisch et al. 2011, 2014; Ausserhofer et al. 2014). Focused examination of high-risk drugs such as aminoglycoside antibiotics can be used to link operational aspects of nursing care such as staffing and assignment, with short term clinical goals such as avoidance of nephrotoxicity.

1.2 5.1.2 The Cost of Nursing Care

In the value-based nursing care model, the actual services delivered as well as the associated time are allocated to each patient. This overcomes a longstanding problem of using average time to identify costs of nursing care. What is an “average” patient? The ability to link nurses to patients and apply different nursing care resources accurately to each patient provides a means to detect differences in nursing intensity and costs across an episode of care such as a hospitalization as well as compare a similar patient within a specific diagnosis located in a Diagnosis Related Group (DRG). Having the actual cost of nursing care and overhead costs such as management, benefits, and so forth, provides a way to estimate actual dollars expended for each patient and links to the billing and reimbursement system. In Fig. 5.1.1, components of the data model link patients, nurses, and charges.

This nursing value common data model provides a way to extract similar data across different EHRs and link nurses directly to patients. For example, if we were interested in examining the effects of young nurses (new graduates with less than 1 year experience), the DateRN field in the Nurse table would be used as the date the nurse was first licensed and then calculate the difference in time in months for the current patient that nurse was caring for and then identify the cost of care (Wage) multiplied by the hours of care given to a specific patient. Data on nursing costs and experience could be summed for an individual patient across a hospitalization, and compared to other patients based on similar or different outcomes such as length of stay, total nursing time and costs, etc.

Patient-level costing for nursing care achieves the goal of better understanding cost drivers within the health care system (Kaplan and Porter 2011). Using nursing business intelligence and analytic tools previously described, nursing and healthcare finance leaders could identify nursing cost drivers and possible nursing intensity outlier patients. Using this new information, nursing care could be adjusted to achieve better matching of nurses to patients, and optimize assignments to achieve best clinical outcomes at the lowest costs of nursing care.

1.3 5.1.3 Summary

Value-based nursing care represents a new approach to using data to identify the added value nurses bring to patient care. It is a way to realize Nightingale’s vision. Collecting and analyzing data at the individual nurse-patient encounter provides greater granularity for identifying clinical and operational intelligence that can inform providers in near real-time to a range of important clinical and operational information needs.

In the vignette, Jennifer was an active participant in interpreting complex data derived from the clinical assessments, interventions, and outcomes identified in the nursing documentation of the EHR. In this fictional setting, nurses are individually and collectively accountable for their care. The goal is to achieve high value outcomes by optimizing nursing care at all the “touchpoints” where nurses interact with patients.

New analytic techniques will inform clinical and operational decision making in the here and now rather than waiting weeks or months. These new data tools will decrease the time between information and action. A value driven data environment will create new information to measure the nurse-patient encounter and share and compare data across the broad spectrum of healthcare that ultimately will achieve Nightingale’s vision to seek and realize excellence in everything we do.

References

• Ausserhofer D, Zander B, Busse R, Schubert M, De Geest S, Rafferty AM, et al. Prevalence, patterns and predictors of nursing care left undone in European hospitals: results from the multicountry cross-sectional RN4CAST study. BMJ Qual Saf. 2014;23(2):126–35.

• Clancy TR, Bowles KH, Gelinas L, Androwich I, Delaney C, Matney S, et al. A call to action: engage in big data science. Nurs Outlook. 2014;62(1):64–5.

• Kalisch BJ, Tschannen D, Lee H, Friese CR. Hospital variation in missed nursing care. Am J Med Qual. 2011;26(4):291–9.

• Kalisch BJ, Xie B, Dabney BW. Patient-reported missed nursing care correlated with adverse events. Am J Med Qual. 2014;29(5):415–22.

• Kaplan RS, Porter ME. How to solve the cost crisis in health care. Harv Bus Rev. 2011;89(9):3–18.

• Nightingale F. Notes on hospitals. Third ed. London: Longman, Green, Longman, Roberts & Green; 1863.

• Pappas SH. Value, a nursing outcome. Nurs Adm Q. 2013;37(2):122–8.

• Pappas SH, Welton JM. Nursing: essential to health care value. Nurse Leader. 2015;13(3):26–9. 38

• Sanford KD. Nurse staffing. Finding the right number and mix. Healthc Financ Manage. 2010;64(9):38–9.

• Simpson RL. Guest editorial. What is the value of nursing care? Nurs Adm Q. 2013;37(2):92–4.

• Spetz J, Brown DS, Aydin C, Donaldson N. The value of reducing hospital-acquired pressure ulcer prevalence: an illustrative analysis. J Nurs Adm. 2013;43(4):235–41.

• Staggs VS, Dunton N. Associations between rates of unassisted inpatient falls and levels of registered and non-registered nurse staffing. Int J Qual Health Care. 2014;26(1):87–92.

• Thompson JD, Diers D. Nursing resources. In: Fetter RB, Brand DF, Gamache D, editors. DRGs their design and development. Ann Arbor: Health Administration Press; 1991. p. 121–83.

• Welton JM. Value-based nursing care. J Nurs Adm. 2010;40(10):399–401.

• Welton JM. Nursing and the value proposition: how information can help transform the healthcare system. Minneapolis, MN, August 12–13: University of Minnesota, School of Nursing, Center for Nursing Informatics; 2013. Accessed 1 May 2016

• Welton JM. Conference report: big data in nursing 2015. Voice of Nursing Leadership. 2015;13(5):10–1.

• Welton JM, Harper EM. Nursing care value-based financial models. Nurs Econ. 2015;33(1):14–9.

• Welton JM, Harper EM. Measuring nursing care value. Nurs Econ. 2016;34(1):7–14.

• Westra BL, Clancy TR, Sensmeier J, Warren JJ, Weaver C, Delaney CW. Nursing knowledge: big data science-implications for nurse leaders. Nurs Adm Q. 2015a;39(4):304–10.

• Westra BL, Latimer GE, Matney SA, Park JI, Sensmeier J, Simpson RL, et al. A national action plan for sharable and comparable nursing data to support practice and translational research for transforming health care. J Am Med Inform Assoc. 2015b;22(3):600–7.

• Yakusheva O, Lindrooth R, Weiss M. Nurse value-added and patient outcomes in acute care. Health Serv Res. 2014;49(6):1767–86.