Clinical Informatics John Welton, PhD, RN, FAAN CU College of Nursing BIOS 6660 University of Colorado College of Nursing November 3, 2015.

Clinical InformaticsJohn Welton, PhD, RN, FAANCU College of Nursing

BIOS 6660

University of Colorado College of Nursing

November 3, 2015

Big Data Concepts

Data Explosion

https://www.intelethernet-dell.com/wp-content/uploads/2011/09/Screen-shot-2011-10-05-at-1.50.14-PM1.png

Challenges

▪ Storage, processing, computational limitations

▪ Security, confidentiality, privacy

▪ Obsolescence of current technology

▪ Accessing data across multiple settings

http://blog.codinghorror.com/content/images/uploads/2006/01/6a0120a85dcdae970b0128776fd5cc970c-pi.png

http://oldcomputers.net/pics/osborne1.jpg

Big Data Concepts

▪ Volume

▪ Velocity

▪ Variety▪ Diverse representations of data▪ Complexity and multiple/mixed

media, e.g. video, sound, pictures, texting, Twitter, Facebook, etc.

▪ Autonomous data sources with distributed and decentralized controls

Wu, X., et al. (2014) Data mining with big data. Knowledge and Data Engineering, IEEE Transactions on 26, 97-107

http://d1mpb3f4gq7nrb.cloudfront.net/img/toons/cartoon6517.png

Items and Issues

▪ Data accuracy and missing data

▪ Extraction and common data models

▪ Archiving and persistence of data

▪ Data consistency across time and settings

▪ Version control, obsolescence

▪ Structured vs. unstructured data

▪ Lack of common data model

▪ Lack of IT support (resources)

▪ Lack of expertise in working with large data

▪ Resources needed to manage “the machine”

Healthcare Data?

▪ Assessments, Physical Exam

▪ Order entry

▪ Results reporting: Labs, xrays, pharmacy (prescription)

▪ Flow sheet data, vital signs, point of care testing

▪ Problem list, treatment plan

▪ Diagnosis, billing, reimbursement

▪ Staffing/assignment (workforce)

▪ Medication administration (bar code)

▪ RFID

Some Interesting Data

▪ RFID (time and position)▪ Tracking patients and

nurses/personnel▪ Finding resources

▪ Call light and response

▪ Continuous data streams from devices, e.g. monitors, beds, etc.

▪ Medication administration (BCMA/eMAR)

http://www.rfidc.com/docs/indoor_rfid_tracking.htm

What are the “Big” Healthcare Questions

Clinical/Patient Focus

▪ Improve health/nursing care

▪ Optimize outcomes

▪ Population management

▪ Better patient experience

Operational/Organizational Focus▪ Healthcare workforce

▪ Resource utilization

▪ Costs, quality, value

▪ Performance, efficiency and effectiveness

Other/Healthcare System Focus

▪ Payment

▪ Policy, etc

Research Perspectives

▪ Continuous data streams

▪ Large volumes of clinical / operational data

▪ Complete data on entire population

▪ Span multiple clinical settings

▪ Examine all provider “touch points”

▪ Multiple/simultaneous natural experiments

Rethinking Healthcare Research

▪ Very large and complex data systems (volume)▪ Statistical significance of large data▪ Time referenced data (e.g. stock

market)

▪ Sipping from a fire hose (velocity)▪ Continuous data streams▪ Natural experiments

▪ Large data sets Complex data sets (variety)▪ Span multiple settings▪ Complex questions and answers

Rethinking Healthcare Systems

Clinical

▪ Real-time clinical decision making

▪ AI potential for pattern recognition

▪ Mapping trajectories of care

▪ Acuity trending (patient, unit, hospital/agency)

Operational

▪ Real-time operational decision making

▪ Quality = acting on poor quality before it occurs

▪ Cost monitoring = higher efficiency and effectiveness

▪ Performance metrics at individual nurse-patient encounter

Data Quality

▪ Structured data▪ Data entry/recopy errors▪ Programming errors▪ Work arounds (BCMA)▪ Event time vs. document time

▪ Unstructured data▪ Narrative hard to quantify▪ Natural Language Processing

(Siri?)▪ Pattern recognition (xray)▪ Expert systems

Real-Time Clinical and Operational Performance

Performance vs Outcomes

Missed Care Potential Quality/Safety Issues

Pain Management Pt Satisfaction; Increased LOS*

Administer meds on timePt Satisfaction; Increased LOS*; Clinical deterioration, e.g. renal effects from improper aminoglycoside admin

Prepare Pt/Family for discharge Readmission < 30d*

Adequate pt surveillance Infections; Clinical deterioration; Increased LOS*;

Oral hygiene Infections; Increased LOS*; Ventilator acquired pneumonia

Educating pts/families Readmission < 30d*

Comfort/talk w patients Pt satisfaction

Change patient position Pressure ulcers*

* Potential for increased cost of care

Quality Performance Metrics for Nursing

Unit/Hospital

▪ Infection rates

▪ Falls & injuries

▪ Pressure ulcers

▪ Patient level nursing costs and intensity

▪ Staffing and assignment

▪ Staff turnover, vacancy rates

Individual Nurse(s)

▪ Medication administration delays and omissions

▪ Pain assessment and management

▪ Other symptom management, e.g. hyper or hypoglycemia

▪ Patient progression (achieving nursing outcomes)▪ Mobility, activity▪ Nutrition▪ Respiratory/cardiac▪ Pain management

Clinical Performance Indicators

▪ Medication Administration

▪ Time delays and omissions

▪ 1 and 2 hour windows

▪ Critical medications, e.g. aminoglycoside antibiotics

▪ PRN medications

▪ Time

▪ Med Admin – Med Due

▪ Med Admin – Med pickup (Pyxis)

▪ Patterns

▪ PRN dose time and amount

▪ Delays and omissions

Medication Administration

Clinical Issues

▪ High risk drugs▪ Insulin, heparin▪ Aminoglycoside Antibiotics

▪ High volume drugs

▪ Pain control (PRN med usage)

▪ Delayed/omitted doses and hospital outcomes

▪ Medication administration volume and complexity

Operational Issues

▪ Patterns of delays & omissions▪ Relationship to workload ▪ Staffing vs. med admin complexity▪ Patterns and trends

▪ PRN practice patterns▪ Day/night shift▪ PRN opioid distribution▪ Relationship with patient satisfaction

▪ Performance▪ Unit level▪ Nurse level▪ Patient level

Hospital Medication Administration

Prescription• MD: Physician Order• CPOE

Dispensing• PharmD: 1. Drug Scheduling 2. Dispensing

• eMAR/Pyxis (or equivalent)

Administration• RN: Medication Administration • BCMA

Process vs. Performance in Med Admin

Prescription• Delayed Rx• Contraindicated• Drug-drug interaction• Polypharmacy• Allergy• Off label• Not standard of care• Inexperienced MD (resident)

Dispensing• Delayed dispensing• Scheduling conflicts• Wrong dose, route, time +• Label errors (cannot scan)• Wrong patient• Lack of drug (shortages, supply issues, surge use, etc.)• Inexperienced PharmD

Administration

• Delayed administration or omission• Multiple patients• Med admin complexity (stool softener vs intropic agent)• ↓PRN med admin (e.g. narcotic analgesics)• Practice variation• Equipment failure (BCMA eMAR)• Float/traveler nurse• Inexperience RN (new grad, float nurse)

Real-Time Medication Administration Analysis

▪ Due vs. admin time

▪ Delays and omissions

▪ Pyxis to BCMA – interruptions?

▪ PRN med patterns (pain management)

▪ Dose to dose variation (antibiotics)

▪ High alert drugs: insulin, anticoagulants, etc.

▪ Nurse – patient – unit analysis

Clinical Performance Indicators

Some Research Questions

▪ Do late/early doses of aminoglycoside antibiotics have direct clinical effects that influence outcomes of care?

▪ Are their practice differences among nurses in administering opioids for pain control?

▪ Is there a relationship between medication administration complexity and nurse workload?

▪ Are delays in administering medications related to high workload, high acuity shifts?

▪ Do long time between drug pickup (Pyxis) and administration identify potential interruptions in nurse workflow?

Future Directions

▪ Real-time information systems

▪ Comparison across different settings

▪ Follow “patient” across all encounters

▪ Link all providers to each patient, family, community

▪ Performance based analysis

▪ Share/compare data

▪ Value-driven healthcare

Common Data Model

▪ Patient focused

▪ Setting neutral

▪ Identifies nurse as provider of care

▪ Direct care hours and costs based on nurse-patient encounter

▪ Ability to directly bill for nursing care

▪ Problem/intervention/outcomes

Nurse_Patient_Encounter

PK ID_Encounter

FK3 ID_EpisodeFK2 ID_Nurse DayTime_Start DayTime_End Shift Type

Patient

PK ID_Patient

Age Race Sex OtherDemographics

Nurse

PK ID_Nurse

FK1 ID_Unit DOB Race Sex JobClass Date RN Date Hire Wage HighestDegree AssignedUnit FTE Agency NPI

PtLocation

PK ID_PtLocation

FK1 ID_EpisodeFK2 ID_Unit Unit_ID DayTime_Start DayTime_End Admit (y/n) Discharge (y/n)

Episode

PK ID_Episode

FK2 ID_Patient EpisodeType DateAdmit DateDischarged AdmissionSource DischargeDispition DRG APRDRG Payer ProcedureCode(1-15) Primary DX Secondary DX (2-15) Readm<30d

Outcomes

PK ID_Outcome

FK1 ID_EpisodeFK2 ID_FlowSheetData OutcomeDayTime OutcomeItem OutcomeScore

Intervention

PK ID_Intervention

FK1 ID_Episode InterventionDayTime InterventionCode InterventionClass

Nurse_Certifications

PK ID_Certification

FK1 ID_Nurse Certification Type DateStart DateExpire

Nurse_Credential

PK ID_Credential

FK1 ID_Nurse Credential_Type DateAwarded DateExpire

Unit

PK ID_Unit

UnitName UnitType NDNQI class Beds

Charges

PK,FK1 ID_Episode

FK2 ChargeID ChargeItem Units Charge

UnitBudget

PK ID_UnitBudget

FK1 ID_Unit BudgetPeriod RN_salaries RN_hours NurAide_hours NurAide_salaries Other_hours Other_salaries RN_FThires RN_FTterminate RN_BudgetedFTE NurAide_BudgetFTE TotalPatientDays

FlowSheetData

PK ID_FlowSheetData

FlowSheetDateTime ItemLabel ItemValue

Nursing Value Generic rev18 Nursing Common Data Model

ChargeMaster

PK ChargeID

Charge Description Charge

PtProblem

PK ID_PtProblem

FK1 ID_NurseFK2 ID_EpisodeFK3 ID_FlowSheetData ProblemIdentDateTime ProblemItem ProblemDesc ProbResolutionDate

CostItem

PK ID_CostItem

FK1 ID_UnitBudgetFK2 ID_EncounterCost TotalHours TotalCosts SumDirecCareCosts IndirectCareHours IndirectCareCosts IndirectCareCostAverage Benefit Costs

EncounterCost

PK ID_EncounterCost

FK1 ID_Encounter DirectCareHours DirectCareCost NurseWage ShiftDifferential OtherShiftCosts

ChargeCost

PK ChargeCost_ID

FK1 ChargeIDFK2 ID_CostItemFK3 ID_EncounterCost BudgetPeriod IndirectCareCostAverage PatientNursingCost

Green = costs; Blue = patient; Purple = nurse/provider; Red = facility/business entity

Nursing Management Minimum Data Set

Business Intelligence and Analytics

▪ Real-time clinical data▪ Sepsis algorithms▪ Care trajectory▪ Pain management

▪ Healthcare Business and Intelligence▪ Optimizing care delivery systems▪ Trending, forecasting, volatility

analysis, pattern recognition, etc.

▪ Outlier analysis▪ Adjust clinical care▪ Optimize outcomes

http://www.equest.com/wp-content/uploads/2013/08/dashboard-snockered-624x418.png

Quality Framework

▪ Traditional View

▪ Monitor/surveillance

▪ Root cause

▪ React to poor quality

▪ Nursing time and costs allocated as a department mean per patient day

▪ Future View

▪ Predictive models

▪ Multiple/interactive cause

▪ Predict and prevent

▪ Nursing time and costs allocated directly to each patient in real-time

Welton, J. M. (2008). Implications of Medicare reimbursement changes related to inpatient nursing care quality. Journal of Nursing Administration, 38, 325-330.

Nursing Value Work Group (7)

▪ Key consensus items▪ Nursing as a practice discipline ▪ Nurses as providers of care▪ Nursing measured at individual nurse-patient encounter▪ Need for common data model to extract relevant costs and

quality data▪ Patient level nursing costing model

PhD Student Core Competencies

Big Data Core Competencies

▪ Database theory and extraction methods

▪ Business intelligence and analytics

▪ Applying statistical techniques to real world problems

▪ Real-time data

Informatics Competencies

▪ Information systems, data storage, processing retrieval

▪ Performance using large data sets, e.g. genomics

▪ Developing common data models

▪ Nursing terminologies, representation of nursing and health care

▪ Natural language processing

▪ Data mining tool kit

What do you take away today?

• Better understanding how to use existing data (including cost data) to improve care

• Optimize clinical and operational environments of care

• Move towards a data-driven and value-based nursing practice model

• Provide the “best” nursing care at the highest quality and lowest cost (the value equation) = best outcome

• The value of nursing can only be described when the financial impact is included

Summary Points

Healthcare Costs

32

Grocery Store Problem

▪ How much do things cost?

▪ How much do you have to spend?

▪ What are bargains?

▪ What if there was no price?

▪ What if everything was the same price?

Cost of Care

▪ How much does care cost at your institution?

▪ What are costs, quality, and outcomes of INDIVIDUAL patients?

▪ How does YOUR hospital compare to other hospitals?

http://www.bcbsm.com/home/images/rising_cost/dollar_is_spent.gif

5.0 10.0 15.0

Routine Care Nursing Intensity

0

200

400

600

800

1,000

1,200

1,400F

req

uen

cy

Mean = 9.761Std. Dev. = 2.79N = 35,723

Variability of Nursing Time

Welton, J. M., Fischer, M., DeGrace, S., & Zone-Smith, L. (2006). Hospital nursing costs, billing, and reimbursement. Nursing Economics, 24, 239-245.

Welton, J. M., Unruh, L., & Halloran, E. J. (2006). Nurse staffing, nursing intensity, staff mix, and direct nursing care costs across Massachusetts hospitals. Journal of Nursing Administration, 36, 416-425.

Healthcare Utilization

▪ Supply vs. demand for healthcare services

▪ Roemer’s Law

▪ Over served vs. under served?

▪ Rural vs. urban

▪ Utilization of services by population

Gawande, A. (2009). The Cost Conundrum What a Texas town can teach us about health care. The New Yorker. http://www.newyorker.com/reporting/2009/06/01/090601fa_fact_gawande?printable=true

The Healthcare Price Problem

▪ Why do hospital charges vary so much?

▪ How much does it cost me to . . .

▪ Does competition increase costs to patients?

▪ Why is utilization higher in some parts of the country?

The Healthcare Price Problem

▪ Tylenol $1.50/pill (Amazon, $1.49/100 pills

▪ Gauze pads: $77 Walgreens “a few dollars”

▪ Troponin lab: $199.50 (Medicare $13.94)

▪ CBC lab: $157.61 (Medicare $11.02)

▪ Accu-Check diabetes test strips $18/each (Amazon $27/box of 50 = $0.55)

Medicare Provider Utilization and Payment Data

▪ Provider and claims based

▪ Fee for service

Homework/Project

Problem

▪ Limited access to primary care in rural CO

▪ Changing demographics (older population, providers moving from rural areas, etc.)

▪ Lack of specialty care

▪ No hospitals in the community

Analysis

▪ How many counties in CO do not have hospitals:http://www.unitedstateszipcodes.org/co/ (merge zip -> county data)?

▪ How many MD and APRN/PA providers are in each county ?

▪ What is the change in providers from 2012 to 2013 data?

▪ What are the top 10 procedures for each county?

▪ What are total billables for each county?

▪ What are total unique patients for each year 2012-2013?

Potential Solution

Community Paramedics

▪ EMS/Fire Department based

▪ Knowledge of community

▪ Mobile

▪ Technology capable, e.g. telehealth, point of care labs

More information?

▪ How many ambulance runs per county and per number of unique patients (see prior) 2012-2013

▪ How many ambulance runs in counties with no hospitals?

▪ Number of emergency visits and hospitalizations for each year by county (note some counties will not have hospitals)

▪ How many non emergency transports

Assignment

Services▪ ## hcpcs_description

▪ ## [1,] "Pathology examination of tissue using a microscope, intermediate complexity"

▪ ## [2,] "Ambulance service, basic life support, non-emergency transport, (bls)"

▪ ## [3,] "Emergency department visit, problem with significant threat to life or function"

▪ ## [4,] "Ambulance service, advanced life support, emergency transport, level 1 (als1-emergency)"

▪ ## [5,] "Subsequent hospital inpatient care, typically 35 minutes per day"

▪ ## [6,] "Initial hospital inpatient care, typically 70 minutes per day"

▪ ## [7,] "Removal of cataract with insertion of lens"

▪ ## [8,] "Subsequent hospital inpatient care, typically 25 minutes per day"

▪ ## [9,] "Established patient office or other outpatient visit, typically 15 minutes"

▪ ## [10,] "Established patient office or other outpatient, visit typically 25 minutes"

New variables

▪ Hospital inpatient care

▪ Ambulance Service

▪ Ambulance service, basic life support, non-emergency transport

▪ Number of non emergency transports in counties without hospitals?