Covid-19 virus (severe acute respiratory syndrome coronavirus 2, SARS-CoV-2) threatens the health of human beings worldwide, imposing a concern for the world and prompting governments to control the contagion. Although vaccination is a proper tool to control the transmission, the efficient allocation of limited health-care resources to massive patients can improve the effectiveness of medical services. Relying on the Artificial Neural Network (ANN), the aim of this research is to enhance the future efficiency of Covid-19 treatment centers by forecasting their efficiency and providing benchmarks. To do this, we use the congestion approach of data envelopment analysis (DEA) based on the theory of economies of scale principles. In the traditional input-oriented DEA, inefficient decision-making units (DMUs) can become efficient merely by reducing the inputs. However, this may not always be true in real-world applications such as improving the efficiency of COVID-19 treatment centers (DMUs). Meaning that the treatment centers with less congested inputs (e.g., ventilators, test equipment, pulmonologists, and nurses, etc.) normally have higher mortality rates. For this reason, in this study, we take the congested inputs approach into account to provide proper benchmarks for the inefficient treatment centers. According to the congestion approach of DEA, an optimum increase in congested inputs can lead to a greater than a proportional increment in outputs. In other words, if more respiratory equipment, pulmonologists, patient rooms, nurses and beds, etc. are allocated to Covid-19 treatment centers, not only the number of deaths (undesirable outputs) are decreased, but also the number of recoveries (desirable outputs) are increased. Such an optimal rise in the congested inputs is determined in pairwise comparisons derived from the model. Accordingly, in this study, first, considering the congestion approach of DEA and historical data of five periods, we identify the initial efficiency of Iranian Covid-19 treatment centers. Then, by running ANN, we forecast the future inputs and outputs, the overall efficiency, and rank of the treatment centers. By doing this, the prospective efficient and inefficient DMUs are identified, and appropriate benchmarks are determined.