Publicly Available Large Data Sets for Health Outcomes Research: - - PowerPoint PPT Presentation

Publicly Available Large Data Sets for Health Outcomes Research: Pearls, Pitfalls, Prices & More

October, 2018

LAKSH IKA TEN N AKOON - MD , MS C, MP H IL, D TM&H R ESEAR C H SC IEN TIST TR AU MA, AC U TE C AR E AN D C R ITIC AL C AR E SU R GERY STAN FO R D U N IVER SITY

Aims

• To encourage use of public data for

Research

• To characterize existing large clinical

databases

Databases Dates Nationwide Inpatient Sample (NIS) 1988- 2016 Nationwide Emergency Department Sample (NEDS) 2006-2016 Nationwide Readmissions Database (NRD) 2010-2016 KID Inpatient Data (KID) 1997,2000, 2003,2006, 2009, 2012, 2016 National Trauma Databank (NTDB) 2002-2016 National Surgical Quality Improvement program (NSQIP) 2005-2016 National Ambulatory Medical Care Survey (NAMCS) 1993-2015

Best Currently Available Databases

Source

HCUP

HCUP HCUP HCUP ACS ACS CDC

Databases Dates National Health and Nutrition Examination Survey (NHANES) 1999-2015 National Hospital Ambulatory Medical Care Survey (NHAMCS) 1992-2015 Medicare/SEER 1991-2015 MarketScan 2002-2011 Hospital based Registry data

Databases……….

Source CDC CDC Government Private Hospital Based

• The largest publicly available all-payer inpatient care

database in the United States

• Samples include all discharges from 20% stratified

sample of US hospitals

• NIS data can be weighted

to generate national estimates

• Years available: 1988 to 2016
• Has 8 million hospital stays a year
• NIS_2015_CORE data file has: 7,153,989 Records

Nationwide Inpatient Sample (NIS)

Cost & Data Load Software

• Cost of 2016 NIS : $625
• Original Data comes as CSV or ASCII files
• Load programs are available in:

STATA SAS SPSS

• Data storage : Large databases need a server or

BOX

NIS-Requirements

Citing HCUP Databases

• Citing HCUP Databases in Abstract and Manuscript:

As specified in the HCUP DUAs, include the database name, HCUP, and AHRQ as demonstrated below for each HCUP database:

• HCUP Nationwide Inpatient Sample (NIS). Healthcare Cost and

Utilization Project (HCUP). 2007-2009. Agency for Healthcare Research and Quality, Rockville, MD. www.hcup- us.ahrq.gov/nisoverview.jsp

Data Files ▪ Core Data ▪ Hospital Data ▪ Illness Severity Data ▪ Cost to charge ratio Data ▪ Diagnosis & Procedure Groups Data

▪ https://www.hcup-us.ahrq.gov/db/nation/nis/nisdde.jsp

What Data Elements Are in the NIS?

• Age at admission
• Gender of patient
• Race of Patient
• Location of patient’s residence
• Median household income for patient's ZIP code
• ICD-9-CM

diagnoses: primary and secondary diagnoses, number of diagnoses, diagnosis coding system

• External causes of injury and poisoning: ECODE 1-4, number
• f external cause of injury
• ICD-9-CM Procedures: primary and secondary procedures,

number of procedures, procedure systems, duration of primary and secondary procedures

• Total charges
• Disposition
• Length of stay

Core Data File

• Hospital bed size
• Type
• f

Hospital: government

• r

private; government, nonfederal, public; private, non- profit; private, investor-own

• Hospital Location: rural or urban,
• Location/teaching

status

• f

Hospital: rural, urban non-teaching, urban teaching

• Region of Hospital: Northeast, Midwest, South,

West

• Hospital Weights: weight to hospitals in AHA

universe, weight to hospitals in the State

Hospital Data File

Severity of Illness Data

• Severity of Illness Subclass
• Risk of Mortality Subclass
• 29 Comorbid conditions: Alcohol Abuse, Depression, Drug

Abuse, Liver Disease, Renal Failure, Obesity…………..

• Defined by Elixhauser Comorbid Scale
• https://www.hcup-

us.ahrq.gov/toolssoftware/comorbidityicd10/comorbidity_icd10.jsp

Cost-to-Charge- Ratio Data

• Year
• Hospital Unique Identifier
• Wage Index
• CCR_NIS (an Identifier, linking NIS 2012 to current )
• Calculate “Total Cost” based on above data and “Total Charges”

(TOTCHG) variable which is available in NIS core data file

• Formula : gen Total_COSTS= TOTCHG*CCR_NIS

• NEDS is the largest all-payer ED database in the

United States

• Samples include stratified samples of 20% of US

hospital-based Emergency Departments

• Years available: 2006-2016
• Number of ED visits: Between 25 and 30 million

(unweighted) records for ED visits from 950 hospitals

• Cost of NEDS 2016 $1000

Nationwide Emergency Department Sample (NEDS)

Four Data Files per year ▪ Core data ▪ Emergency department data ▪ Inpatient data ▪ Hospital Weights data

What Data Elements Are in the NEDS?

Examples of NEDS-Based Research

Nationwide Readmissions Database (NRD)

• Calculate national readmission rates for all payers and the uninsured
• Available nationally representative information on hospital

readmissions for all ages

• Unweighted NRD data from approximately 12 million discharges each

year

• Has Core data, Hospital data, Illness severity, Cost to Charge Ratio

data

• Available years 2010- 2016
• Cost of NRD 2016 data $1000

KID (Kids’ Inpatient Database )

• Only all-payer pediatric inpatient care database in the USA
• Contains 2-3 million hospital stays
• Helps to develop national & regional estimates on diseases
• Data available for Demographics, Injury characteristics,

Diagnosis, Hospital characteristics, Outcomes and Healthcare Cost

• Need to sign a DUA
• Cost of KID 2016 data $500

• The largest registry of trauma patients admitted to

trauma centers in the United States

• Data is not weighted
• No DUA (data user agreement)
• Samples are obtained from trauma center
• registries

▪ In 2011, 747 trauma centers were included

• Years available: 2002 -2016
• Data files are in CSV format
• Cost of 2016 NTDB data $300

National Trauma Data Bank (NTDB)

NTDB Data

• Demographic data
• Injury severity data
• Emergency department data
• Mechanisms of Injury data
• ICD9 and ICD10 Procedure data
• ICD9 and ICD10 Diagnosis data
• Discharge disposition data
• Facility data
• Vital signs data
• Protective devices & transportation data
• Comorbid and complications data

National Surgical Quality Improvement Program (NSQIP)

• A nationally validated, risk-adjusted, and outcomes-

based program

• NSQIP has prospective and outcomes data
• Years available: 2005 - 2011
• NSQIP will measure and improve the quality of surgical

care across surgical specialties

• 680 hospitals are participating NSQIP in 2017

What Data Elements Are in the NSQIP?

• Preoperative risk factors
• Intraoperative variables
• 30-day postoperative mortality and morbidity
• utcomes
• Demographic data
• Current Procedural Terminology (CPT) data
• Health and behavior data
• Physical examination data

• Free data for NSQIP participating hospitals
• Data Request Process

www.facs.org/quality-programs/acs-nsqip

• Need to sign a DUA (Data User Agreement)
• Download the data

www.facs.org/quality-programs/acs-nsqip

• Data files available in 3 different formats: Text, SPSS,

SAS

• Private database
• MarketScan is broadly representative of the commercially

insured population of United States

• High quality, longitudinal, and patient level data
• Low percentage of missing data
• Years available: 2002 – 2011
• Need to sign a DUA (Data User Agreement)

Cost around $50,000/year

MarketScan Data

• Patient socio-demographic data
• Admission date and type
• Diagnosis code (principal and secondary)
• Discharge status
• Procedure code (principal and secondary)
• Length of stay
• Place of service
• Provider ID
• Data on drugs/medications

What Data Elements Are in the MarketScan?

SEER-Medicare Data

• SEER-Medicare Linked Database
• Medicare beneficiaries with cancer
• Data derived from Surveillance, Epidemiology and End Results
• Diagnosis & Procedure codes: ICD9, ICD10, CPT,
• HCPCS (Healthcare Common Procedure Classification System)
• Patient Demographic and Socioeconomic Characteristics
• Comorbidity
• Breast, Colorectal, and Prostate Cancer Screening
• Radiation Therapy (includes codes to identify radiation therapy)
• Chemotherapy Use (includes codes to identify chemotherapy)
• Complications of Cancer Treatment
• Surveillance After Cancer Treatment
• Data sets available from 1991-2015
• Need to sign a DUA (Data User Agreement)

• Physician Characteristics
• Hospital Characteristics
• Health Care Costs Related to Cancer Treatment

National Health and Nutrition Examination Survey (NHANES)

• Cross-sectional and high quality survey data of adults and children

in United States

• Data available on nationally representative sample of about 5,000

persons/each year

• Years available:

1971-75—NHANES I 1976-80—NHANES II 1982-84—Hispanic Health and Nutrition Examination Survey (HHANES) 1988-94—NHANES III 1999-present--National Health and Nutrition Examination Survey (Continuous NHANES)

• Free to download the data from CDC website

• Socio-Demographic data
• Dietary data
• Clinical examination data (medical, dental, and

physiological measurements)

• Laboratory data
• Questionnaire data
• Genetic data
• Mortality data
• NHANES Medicare Utilization and Expenditure Linked

Files (Restricted data)

• NHANES Linked Mortality files
• NHANES Linked Social Security Administration Files

(Restricted Data)

What Data Elements Are in the NHANES?

• National survey has ambulatory medical care

services data in the United States

• Data will represent a sample of visits to non-

federal employed, office-based physicians who are primarily engaged in direct patient care

• NAMCS has high quality cross-sectional data
• Years available: 1973-Current
• Free to download the data from CDC

website

National Ambulatory Medical Care Survey (NAMCS)

• Socio-demographic data
• Source of payment and number of past visits
• Patient’s Primary Care Physician Information

Diagnosis

• Chronic disease checklist and disease management

programs

• Screening and diagnostic services
• Treatments and drugs prescribed
• Physician specialty
• EMR use and practice parameters
• sources of revenue
• Providers seen and duration of care under those

providers

What Data Elements Are in the NAMCS?

Evidence Based Research-NAMCS

National Hospital Ambulatory Medical Care Survey (NHAMCS)

• National probability sample survey of visits to Emergency and

Outpatient departments in nonfederal, general, and short-stay hospitals in United States

• Records-based survey data, producing annual estimates of the

number and attributes of visits to hospital emergency departments (EDs) in the U.S

• Survey is a visit based and cannot calculate prevalence and incidence

rates

• Years available: 1992-Current
• Free to download the data from CDC website

Thank you ! Lakshika email: lakshika@stanford.edu