Table 2.
Data source types in pediatric critical care research.
| Type of data | Clinical detail | Example data sources | Represented population | Accessibility/cost | Notes |
|---|---|---|---|---|---|
| Public use datasets from funded studies and networks | High | BioLINCC, CPCCRN, PECARN | Study-specific or network-specific | High/free | Often available online (e.g., www.pecarn.org and www.cpccrn.org), requires data use agreement and IRB approval. May have limited data support |
| Pediatric critical care registries | High | Virtual PICU systems (VPS) | 119 participating hospitals (fee) with ~600,000 PICU cases | Moderate/free to VPS members | Use by non-members requires partnership with a member hospital investigator. Requires review by the Research Committee, which is primarily intended to ensure that multiple investigators are not attempting to answer the same question |
| High | ANZPIC registry | 24 PICUs in Australia and New Zealand | Moderate/free | ||
| High | PICANet registry | 32 PICUs in the United Kingdom and Ireland | High/free | May be merged with continental European PICU registries in the future | |
| Therapy-specific registries | High | Society of thoracic surgeons – STS congenital database | Children who have undergone cardiac surgery at participating centers | Moderate/low | Cardiac anesthesia-specific data was introduced in 2010. Linked to PHIS |
| Moderate | ECMO registry of the Extracorporeal Life Support Organization | 230 voluntarily contributing centers. Internationally representative sample of ECMO utilization | Moderate/free | Available to member centers, special requests may be made to the ECMO Registry steering committee. Minimal ability to risk adjust, but plans in place to improve this in 2013. Also contains neonatal and adult ECMO runs. Interpretation of outcome and complication data should be done with care | |
| Disease-specific registries | High | Department of Defense Trauma Registry | U.S. military, coalition soldiers, and civilian trauma patients | Moderate/free | Department of Defense only. Use requires partnership with a military investigator |
| Moderate | American Heart Association Get With the Guidelines – Resuscitation | >400 voluntarily contributing hospitals (hospitals pay a fee to participate as a quality improvement initiative) | Moderate/free | Use requires approval of research request by AHA-GWTG-R Research Task Force | |
| Moderate | National Trauma Data Bank | >700 voluntarily contributing hospitals and >100,000 pediatric trauma admissions each year | High/$300 per year | Obtained from the American College of Surgeons. A nationally representative sample of adults treated at Levels I and II facilities is also available for purchase | |
| Population-based registries | High | Rochester Epidemiology Project (REP) | All residents of Olmsted County, MN from January 1, 1966 to the present, with ~500,000 individuals and ~1.2 million records | Moderate/free | Unique population-based resource. Use requires permission from the REP obtained through online application |
| Quality improvement or benchmarking databases | Moderate | Pediatric Health Information Systems (PHIS) | 44 free-standing children’s hospitals, >7 million inpatient cases and 20 million Emergency Department encounters | High/free to member hospitals | No physiologic variables. Resource utilization and charge data are detailed, but results of tests and studies are not currently widely available. Several linkages completed or planned |
| Claims databases | Moderate | State medicaid files | Data available from 1999 to present for all 50 states and D.C. | High/~$1,000–1,500 per year, per state | Limited use to date in PCCM research |
| Government administrative databases | Low | Healthcare Cost and Utilization Project (HCUP) databases (details below) | Prices for HCUP products frequently discounted for students | ||
| Kids inpatient database (KID) | Every 3 years: 1997, 2000, 2003, 2006, 2009 | High/$200–350 per year | Allows national-level estimates of pediatric conditions. Sample weighting requires analytic adjustment. Two to three million hospital discharges in each file | ||
| National Inpatient Sample (NIS) | Annual ~20% stratified sample of hospital discharges, ~1,000 hospitals per year | High/$160–350 per year | Sample weighting requires analytic adjustment | ||
| National emergency department sample (NEDS) | Annual ~20% stratified sample of ED visits in 28 states, 2006–2010 | High/$500 per year | Linked to state inpatient databases to determine ED outcomes. Sample weighting requires analytic adjustment | ||
| State inpatient databases (SID) | All inpatient discharge abstracts in participating states | High/~$35–3,000 per year, per state | Component files of the NIS and KID | ||
| State emergency department databases (SEDD) files | All ED visits that do not result in admission, for each participating state | High/~$35–3,200 per year, per state | Component files of the NEDS. Information about patients seen in an ED and admitted is found in the corresponding SID |
Level of clinical detail: high = includes many clinical variables such as vital signs, laboratory results, or severity of illness scores. Moderate = includes some clinical variables or utilization data (medications, imaging studies, etc.) or charge/cost information. Low = data limited to standard administrative content such as demographics, length of stay, disposition, and diagnosis and procedure codes.