ince we cannot measure the expected benefits and costs of
HIT adoption directly, we estimate a reduced-form model of
HIT adoption as a function of hospital characteristics that
influence costs and expected benefits, including hospital
cost inefficiency. However, HIT implementation is a
process rather than an event and it may take from one to
two years from the moment of signing a contract with a
vendor to having an operational system [17, 40]. In
addition, the implementation process can temporarily cause
a short-term loss of efficiency due to the learning and
adjustment processes [37]. Figure 2demonstrates a hypothetical example, in which Hospitals A and B make a
decision about HIT adoption in 2006. Hospital A may decide not to introduce HIT since the potential benefits can
be smaller compared to the costs, given that Hospital A is
already on the path to be more efficient without additional
interventions. Hospital B, the more inefficient hospital, may
decide to introduce HIT as a means of reducing inefficiencies in its operations. The HIT adoption process takes about
2 years and leads to a temporary loss of efficiency by 2008.
At the end, the investment decision was successful for
Hospital B and led to a reduction of inefficiencies
comparable to the level of the more efficient Hospital A.
Based on the empirical strategy presented in Fig.2, we used
a 2-year lag between an outcome variable and covariates
that allows modeling the HIT adoption decision as a
function of hospital characteristics at the point in time
when the decision is made and, most importantly, before
hospital performance has been affected by the HIT
adoption. Particularly, we are interested in the probability
of HIT adoption in a periodt+2 conditional on not having
HIT in a periodt as a function of hospital cost inefficiency
in period t: