HCUP NRD - Healthcare Cost and Utilization Project, Nationwide Readmissions Database

General description

1. Database primary purpose - Designed to support various types of analyses of national readmissions for all patients regardless of the expected payer for the hospital stay. This database addresses a significant data deficiency in healthcare: the absence of nationally representative data on hospital readmissions for all ages.
2. Overall data type - Health outcomes
3. Dataset type - Cross-sectional
4. Data source - Claims
5. Data level - Patient level, but includes a Hospital-level file
6. Geographic location of the data collection sites - United States (22 HCUP partner states: Arkansas, California, Florida, Georgia, Hawaii, Iowa, Louisiana, Maryland, Massachusetts, Missouri, Nebraska, New Mexico, Nevada, New York, South Carolina, South Dakota, Tennessee, Utah, Virginia, Vermont, Washington, and Wisconsin).
7. Sponsor, manager, or home institution - Healthcare cost and utilization project (HCUP), Agency for Healthcare Research and Quality's (AHRQ)
8. Date range - 2010 - 2014
9. Geolocation data - If the patient is a resident of the State in which he/she received hospital care
10. Hospital identifiers - NRD hospital ID (not linkable to any other HCUP or external databases)
11. Longitudinal tracking - Track patients within and across hospitals within a State (up to one year); track hospitals (up to one year)
12. Financial variables - Total hospital cost must be added to the NRD using the HCUP supplemental Cost-to-Charge Ratio files
13. Clinical areas of interest - all
14. Number of records - Unweighted, the NRD contains data from approximately 15 million discharges each year. Weighted, it estimates roughly 35 million discharges in the United States. The above-mentioned partner states are geographically dispersed and account for 51.2 percent of the total U.S. resident population and 49.3 percent of all U.S. hospitalizations (1)
15. Variables that are uniquely present in this dataset - The NRD is a unique and powerful database designed to support various types of analyses of national readmissions for all patients regardless of the expected payer for the hospital stay. The NRD includes discharges for patients with and without repeat hospital visits in a year and those who have died in the hospital. Also, it includes a "Severity File", containing additional data elements to aid in identifying the severity of the condition for a specific discharge. The NRD includes community hospitals, excluding rehabilitation or long-term acute care hospitals, and all discharges from the SID except discharges (1) from patients with an age of 0 from 10 of the 22 SID in 2014 because patient linkage numbers are inconsistently reported for this age; (2) with missing patient linkage numbers; (3) with questionable patient linkage numbers, defined as 20 or more discharges in a year, hospitalized after discharged dead, and overlapping inpatient stays; and (4) from hospitals with more than 50 percent of their total discharges excluded for any of the above causes, because patients treated at these hospitals may not be reliably tracked over time.
16. Database caveats and limitations - (1) Repeat stays may or may not be related. (2) Excludes all patients younger than 1 year from more than half of the SID in 2010–2013 NRD and 10 of the 22 SID in 2014 NRD. (3) The weights for pediatric discharges were often higher than for adult discharges. (4) 30- or 60-day readmissions for patients admitted in the latter part of the year may not be captured if the subsequent admission crossed into the next year. In addition, one year of discharge data is probably an insufficient length of time for examining readmissions that are more than 90 days apart. (5) Cannot be combined across data years to create a multi-year database. (6) Patients who were hospitalized in one State and readmitted or transferred to a hospital in another State cannot be tracked.

Applicable methods

1. Assocition methods, such as generalized linear mixed effects models (2), logistic regression (3, 4, 5), Cox proportional hazards regression (6), Kruskal-Wallis tests (7), Poisson regression (8)
2. Descriptive analyses (9)
3. Difference-in-differences (10, 11)
4. Dose-response (12)
5. Interrupted time-series (13)
6. Propensity scores (14, 15, 16)

High-impact designs

• Assess the length of stay (LOS) and costs for 30-day readmissions in children (2)

• Examine risk factors and outcomes of hospital readmission following surgery (17)

• Compare characteristics and outcomes of patients with different myocardial infarction types (18)

• Characterize the readmission rate associated with hospitalization for treatment (19)

• Investigate the association between the probability of a procedure at the patient level and profit per procedure (20)

• Evaluate and compare readmission burden, causes and timing (6, 21)

• Compare strategies for treatment (22)

• Evaluate the associations between demographics, social needs, hospital characteristics, and comorbidities on readmission (12)

• Compare emergency general surgery (EGS) patient outcomes following index and non-index hospital readmissions (23)

• Assess whether a program implementation was associated with a readmission decline for patients across all insurance types (10)

Data dictionary

To access the HCUP NRD data dictionary, click here

Variable categories

1. Patient demographics (e.g., age, sex, primary payer, NCHS urban-rural classification, median household income quartiles for patient's ZIP Code)
2. Discharge records (e.g., died, discharge quarter and month, elective admission, ED service, transfer, length of stay)
3. Diagnosis (e.g., diagnosis-related group, principal and secondary ICD-CM codes, number of chronic conditions, number of diagnosis and procedures coded)
4. Comorbidity measures (e.g., AIDS, alcohol, anemia, congestive heart failure, depression, diabetes, drug, obesity)
5. Hospital information (e.g., control/ownership, number of beds, teaching status, urban-rural location)