VHA employs thousands of physicians to provide care at over 800 hospitals and clinics. It bears a significant cost for hiring new physicians and training them on VHA systems and practices. There is a widespread belief that the VA research program improves physician recruitment and retention. The relation of research support to physician recruitment and retention was investigated in a recent HSR&D-funded project (Hysong, Best, and Bollinger 2005). The Hysong et al. (2005) study succeeded at its stated mission to develop and test a qualitative model. It confirmed that VA physicians believe what most others do, viz., that research support affects physician recruitment and retention. As the authors note, however, it has several limitations. There is significant need for a rigorous quantitative model of physician recruitment and retention. Such a model would enable policymakers to estimate the relative impacts on retention of different levels of research funding. It could also be used to develop estimates of the savings in direct (recruitment) and indirect (productivity) costs caused by varying the VA research budget. Such figures would inform Congressional testimony by VA and other supporters of the VA research program, as well as providing VISN and station managers with additional incentive to pursue VA research dollars.
The aim of this project is to develop and test quantitative models of VHA physician retention. The key independent variable will be VA research funding at the physician's station. There will be two outcome variables. The first will be physician job tenure, the number of years employed at the same VA station. The second will be retention, modeled as the probability that an individual physician remains at his/her VA station during a fiscal year. The final objective is to estimate the potential savings in recruitment costs if VA research funds were increased.
[original text: ] The study will combine individual-level data from the VA PAID payroll data system with facility-level administrative data. We will develop descriptive statistics and carry out regression analyses. There will be two outcomes of interest: physician job tenure at VA and physician retention probability (the probability that an individual physician remains at his/her VA station during a fiscal year). Results of these models, combined with published estimates of the cost of physician recruitment, will be used to estimate the potential savings in recruitment costs if VA research funds were increased.
[May 28, 2009 update:] We have found additional VA research funding data on the Allocation Resource Center (ARC) web site and have used it to determine two streams of VA funding: VA funds (which count dollar-for-dollar in determining the allocation of '101' funds), and other funds that VA managers (which are discounted in allocated '101' funds). We also have NIH total funds for affiliated medical schools where they exist. Thus for each station-year combination we have three funding figures.
[May 28, 2009 update:] When we look across sites within individual years FY2003 - FY2005, we find no relation of research funding to the retention rate of physicians. This is true even when we study the retention rate of PI-eligible physicians only (those scheduled 5/8ths time or more), when we limit the sample to stations having any VA funding, and when we limit the sample to stations having at least 5, at least 10, or at least 19 (the median) PI-eligible physicians.
What remains is to develop a pseudo-panel dataset that stacks data by station and year. Thus it will have one observation for each station-year combination over a period of 4-5 years. This will enable us to estimate a fixed-effects model that controls for year and station effects.
Reducing turnover of physician staff members may improve quality of care. Understanding the relation of retention to research funding may assist VACO in targeting research funding opportunities.
None at this time.