Identifying In-Hospital Venous Thromboembolism (VTE)

A Comparison of Claims-Based Approaches with the Rochester Epidemiology Project VTE Cohort

Reports that medical errors account for substantial in-hospital mortality are difficult to substantiate due to the challenges of identifying hospital-acquired conditions using claims data. In most cases, discharge diagnosis codes are used to identify suspect conditions and exclusion rules are then applied to distinguish hospital-acquired conditions from those present when the patient was admitted. No estimates are available of hospital-acquired complications that are missed using claims-based approaches.

The researchers examined 37,845 inpatient encounters by Olmsted County, MN, residents at Mayo Clinic-affiliated hospitals 1995 through 1998. They used the Rochester Epidemiology Project (REP) to identify all cases of venous thromboembolism (VTE) among Olmsted County residents for which onset of symptoms occurred during a hospital stay. REP-identified cases were used as the gold-standard and were compared to cases identified with two different claims-based approaches, i.e., exclusion rules and present-on-admission (POA) indicators.

Key Findings:

  • Using the REP approach, 98 encounters (2.6 per 1,000 total encounters) met criteria for hospital-acquired VTE.
  • Using claims data, 252 encounters had one or more secondary diagnosis code for VTE; 208 (5.5 per 1,000 total encounters) met criteria for hospital-acquired VTE using the exclusion rules, and 1.4 per 1,000 total encounters met criteria for hospital-acquired VTE using POA indicators.
  • The proportion of encounters identified as hospital-acquired VTE using claims data that were confirmed as such using the REP approach (i.e., positive predictive value) was low (35%) for exclusion rules and high (74%) for POA indicators.
  • By contrast, of all REP-identified hospital-acquired VTE encounters, the proportion correctly identified as hospital-acquired VTE using claims data (i.e., sensitivity) was high (74%) for exclusion rules and low (38%) for POA indicators.

Further research is necessary to assess the utility and accuracy of claims data for identifying hospital-acquired VTE.