The initial goal was to collect profession-specific workforce information from each expert panel for each type of implementation: physician office EHR, institutional EHR, and community health information infrastructure. It was also important for the experts to validate and contribute to the overall study approach, assumptions, and limitations.
With respect to the overall approach of the study, there was general consensus that collecting data on specific activities needed to implement the NHIN and then building a tool to use that data for an overall workforce estimation made sense. There was also universal agreement that there is a workforce problem – i.e., at this time, we do not have the trained professionals we need to fully implement the NHIN. The experts also were unanimous in welcoming the effort to begin to quantify this problem with the current research.
However, there were a number of difficult issues. First, there is not consensus on the types or definitions of health IT professions involved in NHIN work. We were able to elucidate fifteen different types of personnel and develop definitions that were generally acceptable (Table I). However, each expert panel wanted to add additional subtypes and variants.
Another difficult problem was the categorization of different types of EHR implementations, both in terms of the type of organization and the architecture of the installation. It is generally agreed that an EHR installation in a solo practitioner’s office is quite different from one in a large multispecialty group practice, but creating meaningful and consistent categories is problematic. Initially, it was suggested that the number of physicians should be the key parameter for categorization. However, other experts suggested that it is the decision-making process in a practice that differentiates small and large installations. When one person is making decisions, the process is faster and more efficient than when several people or even several committees must be involved in each implementation step. The final categorizations we attempted to use were a hybrid of these approaches (Table II).
Finally, it was hoped that the experts would be able to provide profession-specific estimates of the personnel types and time required for EHR installation in various settings. However, the tenfold (or greater) variance observed in these estimates (even after discussion among the experts) clearly indicated that the results were quite speculative and largely the result of guesses.
The expert panelists themselves suggested that the best sources of data would be the EHR vendors, since they actually deploy staff for EHR installations on a regular basis. Accordingly, the members of the EHR Vendors Association were asked to submit data anonymously to assist in compiling information for this research. However, only a very small number of vendors had responded to this request as of this writing. Possible reasons for the lack of response include insufficient time, proprietary concerns, and low priority for support of such research activities.
There were even more difficult problems for collection of data regarding the personnel needed to build community health information infrastructures. First, only a handful of communities have operational systems and, therefore, could be considered to have completed the implementation process. Second, each community’s system is different and, therefore, not necessarily comparable to any of the others. This also reflects the fact that a consensus has yet to emerge regarding the best architecture for creating such systems. Finally, because of the long-term nature of the projects, the varied funding, and the typically extensive in-kind contributions of time, communities may not have accurate records of personnel types and time that have been used.
The experts were very helpful in clarifying and adding to the initial list of assumptions for the current research. In particular, the focus of the present study on the implementation phase of EHR systems needs to be emphasized. While it is clear that additional personnel time is necessary to plan for EHR implementation and that personnel are needed for maintenance after the installation, the present study attempted to focus exclusively on the implementation period only. Furthermore, additional burdens placed on current personnel were excluded, as the goal was to determine what additional workers are needed. A complete list of the assumptions is shown in Table III.
Similarly, the expert panels were quite helpful in expanding the scope of limitations for this study. For example, the short duration of this project and the planned methodology depended exclusively on capturing existing data as opposed to making new, independent observations of the workforce used for various EHR installations. The complete list of identified limitations is shown in Table IV.
In that context, the limitation with respect to attrition of the workforce (# 18) needs to be specifically highlighted, particularly as it has a substantial impact on the interpretation of the personnel estimates for physician office EHR installation. Given the short estimated installation time (23.7 FTE days -- or about 10 installation cycles per year), even a small attrition rate of personnel for each installation cycle could result in a substantially increased workforce need. For example, a 10% attrition rate over 10 cycles per year would mean that the entire workforce would need to be replaced each year. In other words, to maintain the needed workforce of 7,600 FTEs, an additional 7,600 workers would need to be added every year. Therefore, while the number of FTEs performing installations would not change over time, the number of actual personnel that would be needed is 7,600 each year (or a total of 38,000 in five years). Since there is substantial anecdotal evidence that some personnel performing EHR installations in physician offices are commonly retained by the practice for ongoing support and maintenance, this is an important factor to consider. Note that this has much less impact in the areas of hospital EHRs and community health information infrastructure systems because of the longer installation times.