Limitations
The results of this study are limited in more than one fashion. First, this study does not include any direct measures of the system users. For any given encounter, the system examined in this study only records use. While we can identify encounters without usage, we have no way of attributing this non-use to an individual user. In other words, while we can identify the users who employed the system, we cannot identify the users who did not use the system. Therefore, we cannot address any potential confounding due to user characteristics like job type, computer skills, or perceptions of system usability. Likewise, we did not systematically explore the potential the role of workplace characteristics. For example, computer terminals may not have been equally accessible to users in each ED or the speed of internet connection may have varied. While our random-intercept models statistically adjust for these differences, we do not explicitly model their effects. Third, our measure of usage does not include the specific data sought, search strategies, encounter workflow, or if the search was successful. Previous research demonstrates end users need a wide variety of patient information types,[51] engage in the system in very diverse ways,[52] and tighter integration into workflow improves usage rates [20,53]. Despite these limitations our measure of usage is more informative than previous research [54]. The logical next step in research is to examine these suggested factors qualitatively with users and the context workflow. Fourth, due to the user location and date restrictions in the linking process, we have excluded users accessing the system after a patient encounter to identify patients for disease management programs, social services, or public health. The factors associated with usage could reasonably be expected to differ among these types of users. Finally, caution is needed in generalizing this study to other types of healthcare encounters or other HIE efforts since this particular exchange only includes medically indigent patients and one particular HIE system. Other HIE systems may differ on key characteristics such as breadth of information types and sources, the display of information, or overall system usability.
Conclusions
These findings help fill a gap in knowledge about what type of HIE use occurs for children and under what conditions. Healthcare professionals, advocates and the government believe HIE has the potential to transform healthcare and children may benefit greatly from the improved information sharing. However, the simple existence of these systems is not likely to be sufficient and assuming use will automatically follow existence is unfounded. HIE systems must be applied to the delivery of care and the improvement of patient health. The improved use of HIE to avoid duplication and improve coordination of care for pediatric patients will be increasingly important as health reform moves to extend coverage to nearly all children.