NSQIP National Surgical Quality Improvement Program - onetomapanalytics/Meta_Data GitHub Wiki

NSQIP - National Surgical Quality Improvement Program

General description

  1. Database primary purpose - Provide robust, valid data to hospitals to analyze and take appropriate steps to fix problem areas.
  2. Overall data type - Health outcomes
  3. Dataset type - Cross-sectional
  4. Data source - Patient's medical charts
  5. Data level - Patient level
  6. Geographic location of the data collection sites - United States
  7. Sponsor, manager, or home institution - American College of Surgeons
  8. Date range - 2005 - 2019
  9. Dates - Year
  10. Clinical areas of interest - Surgery
  11. Number of records - More than 8.7 million cases from over 700 NSQIP-participating sites
  12. Variables that are uniquely present in this dataset - Risk-adjusted and case-mix-adjusted dataset developed by surgeons.
  13. Database caveats and limitations - (1) the variables are generic in nature, which may pose difficulties for researchers attempting in-depth research on specific conditions or operations. (2) Only patients over the age of 18 are available for assessment, and patients over the age of 90 are grouped into a 90+ category, so the age distribution is somewhat truncated. (3) There are no specific dates, and this precludes evaluation of time-of-day or day-of-week effects. Outcomes in the NSQIP data are limited to a 30-day follow-up period. (4) Preventative measures are not recorded, which can lead to an underestimation of the risk of certain conditions when such measures are routinely taken before surgery. (5) Data are submitted from hospitals participating in the ACS NSQIP, therefore, do not represent a statistically valid nationally representative sample. (6) It is not possible to assess variations among specific clinicians or account for the clustering of outcomes. (7) Cases are typically selected using random sampling to provide a hospital-level quality assessment, and they may not adequately capture rare outcomes or accurately portray outcomes for rare cases. (8) A select number of variables are missing or have undergone an evolution over time, making analysis difficult or impossible.

Applicable methods

  1. Associations analysis, such as hierarchical logistic regression (1), multivariable regression (2, 3)
  2. Machine learning (4, 5)
  3. Propensity score (6, 7)
  4. Interrupted time series (8)
  5. Univariate analysis (9)
  6. Sensitivity analyses (10)
  7. Dose-response analysis (11)

High-impact designs

  • Examine the association between risk factors and mortality (12, 13)

  • Assess the association between risk factors and complications (14, 15, 16)

  • Determine the accuracy of outcomes predictions (17, 18)

  • Evaluate safety practices (19)

  • Evaluate the effect of protocol adherence and outcomes (20)

  • Compare surgical techniques and outcomes (21

  • Characterize agreement between administrative and registry data in relation to patient-level comorbidities (22)

  • Evaluate the impact of an implemented intervention (policy, event, program, etc.) (23)

  • Compare variables between databases (24, 25, 26)

  • Analyze health care costs and savings associated with quality improvement (27)

  • Assess racial/ethnic disparities on surgical outcomes (28, 29

  • Enrichment of this bank through linkage to other datasets, for example, CMS (30), DCI (31)

  • Compare outcomes by gender (32

Data dictionary

To access the NSQIP data dictionary, click here

Variable categories

  1. Patient demographics (e.g., age, sex, race, ethnicity)
  2. Preoperative risk factors (e.g., height/weight, diabetes, smoker, drinking, dyspnea, previous conditions, and procedures)
  3. Intraoperative variables (e.g., primary procedure code, anesthesia technique, level of residency supervision, surgical specialty, emergency/elective surgery, operation time, complications, in/out-patient status)
  4. 30-day postoperative mortality and morbidity outcomes (e.g., post-op diagnosis, readmission, return to OR)