Evaluation of the Influenza-Associated Intensive Care Unit Hospitalization Surveillance System in Illinois

BACKGROUND: Since 2010, Illinois has mandated case-based reporting of laboratory-confirmed influenza-associated intensive care unit (ICU) hospitalizations to the Illinois National Electronic Disease Surveillance System (I-NEDSS) in order to define and characterize the total burden of severe illness and monitor for changes between influenza seasons. Reporting ICU hospitalizations versus all hospitalizations provides information on influenza severity without overwhelming users.  The surveillance system was evaluated to assess its usefulness and provide recommendations for improvement.

METHODS: The evaluation was structured using CDC’s Updated Guidelines for Evaluating Public Health Surveillance Systems. Local health department staff and hospital infection preventionists were interviewed to evaluate system simplicity, flexibility, and acceptability. 2010-2013 I-NEDSS data were used to assess data quality and reporting timeliness. I-NEDSS and the Illinois Hospital Discharge Database (HDD) data were compared to estimate sensitivity and representativeness. Sensitivity was calculated by dividing the number of I-NEDSS cases by the number of HDD ICU hospitalizations with any influenza-specific ICD-9 code. To assess representativeness, the number of I-NEDSS cases was compared to all influenza hospitalizations in the HDD.   

RESULTS: Stakeholders found the system’s information useful when disseminating public health messages and making hospital policy decisions during influenza season. The case definition is simple; however, case identification is time consuming in hospitals where determining ICU admission requires manually tracking patients. Users can manage the current volume of reported cases, but most believe resources would be stressed if all influenza-related hospitalizations were made reportable. From October 2010 to May 2013, 1330 cases were reported in I-NEDSS. 35% were reported within 24-hours. The proportion of missing values ranged from 0.1-24% for demographic and 7-36% for clinical fields. The sensitivity of the system was 37.4%, which remained consistent for each influenza season evaluated. I-NEDSS cases represented 9.8% of all influenza-related hospitalizations in the HDD. The distribution of demographic data was comparable between cases reported in I-NEDSS and the HDD. 

CONCLUSIONS: Educating users on reporting requirements and refining the case definition to include a defined time interval between a positive influenza test and ICU admission could improve the system’s timeliness and simplicity. The apparent low sensitivity of the system merits further investigation as it suggests that the total burden of illness is not being accurately captured.  Limiting reporting of severe influenza illnesses to ICU hospitalizations allows Illinois to efficiently obtain useful information regarding the severity of each influenza season.