Improving hospital efficiency is a critical concern for health care managers and policy makers. Hospital technical efficiency is measured as the ratio of what quantity and quality of care is produced to what could be produced given the level of resources available to the hospital (its budget). What a hospital should produce given the resources at its disposal is called the “production frontier”. In order to improve hospital performance, health policy makers need knowledge and information about how well the hospitals they fund are utilizing the resources they receive. Data Envelopment Analysis, a non-parametric technique, is applied to administrative data on hospitals in Canada to produce the “technical frontier” and get insight into the variation of technical efficiency of acute hospitals at the Pan-Canadian level (except for the province of Quebec, which does not report its data on hospitals in a way that would make them comparable to the rest of Canada). DEA is preferred to the alternative method of stochastic frontier for the following reasons: DEA does not require to impose a specification on the production function of hospitals (for which theory is clearly lacking), and it allows the analyst to estimate a multi-output frontier (a stochastic frontier would have to weight arbitrarily the value of quantity versus that of quality of care in hospitals, whereas the DEA approach generates these weights from the data). Efficiency scores are serially de-correlated using a bootstrap technique and then entered as the dependent variable in regressions to identify the main factors of efficiency or inefficiency. Specifically, this thesis aims to: 1) estimate the level of technical efficiency of acute inpatient care in 35 teaching hospitals, 54 large hospitals and 90 medium-size hospitals respectively in Canada and identify the potential factors that have influence on technical efficiency; 2) uncover and measure the existence of possible spatial spillovers of hospital efficiency in Canada and examine its potential determinants while taking into account the interaction between hospitals by means of spatial regression; and 3) examine the technical and scale efficiency of the 229 small and rural hospitals across Canada (outside Quebec), as well as estimate the impact of institutional and contextual variables on hospital technical and scale efficiency respectively. The major findings are: 1) hospital output (combination of number and quality of stays; quality being measured as the inverse of in-hospital mortality) in Canada could be increased by 24 percent with the same resources by eliminating inefficiency. Highly efficient teaching hospitals benefit from producing care under favourable environments. Higher efficiency could be achieved by increasing cooperation within the health system and making more post- acute care beds available to both large and medium hospitals; 2) There is a substantial and significantly positive spatial spillover effect on the efficiency of acute inpatient care (elasticity of 0.3): Canadian hospitals are clearly complements to each other and work in networks much more than in competition. The hospital size (the number of beds), the percent of transfers between acute hospitals, and the percent of patient transfers to home care are the main drivers of efficiency among acute hospitals in Canada while controlling of the dependence between hospitals; and 3) Among small hospitals, the average output orientation technical efficiency on all types of services is 54% at the current input-output mix. To improve their technical efficiency, small hospitals should provide with more home care facilities to discharge their patients to (so-called Alternative Level of Care patients) and strengthen their cooperation with larger, urban hospitals. Small hospitals are scale inefficient, specifically, rural hospitals could reduce their size by 34% on average (around 6 acute beds) to achieve the optimal size. The study also found that the spending on diagnosis tests and the nursing as the percentages of total hospital spending (cost shares) are positively and significantly related to the scale efficiency. / Thesis / Doctor of Philosophy (PhD) / A hospital is technically efficient if it uses its resources (its budget) to get the most in terms of quantity (number of stays) and quality of care it can. A hospital can be more or less technically efficient for reasons independent of its control (typically, because of the environment in which the hospital operates) and efficiency is a value-neutral measure. This thesis aims to: 1) estimate the technical efficiency of acute inpatient care in Canada and identify the potential factors that influence the level of efficiency achieved by a given hospital; 2) uncover the existence of possible geographic clusters of efficiency (hospitals that are close geographically are also close in the efficiency scale, something called spatial spillovers in the literature) in Canada.; and 3) examine the role that size plays in the variation of technical efficiency among small and rural hospitals across Canada. The major findings are: 1) hospital output could be increased by 24 percent with the same resources by eliminating technical inefficiency; 2) There is a substantial and significantly positive spatial spillover effect on the efficiency of acute inpatient care: being close to an efficient hospital increases the efficiency score of a hospital, everything else being the same; and 3) The level of technical efficiency of small and rural hospitals across Canada is low overall and, perhaps surprisingly, larger rural hospitals are among the least efficient: among small hospitals, scale does not yield economies of resources.