From 2015 to 2022, a total of 1,938 deaths were collected in the hospital, including 287 hospitalized cases and 1,651 non-hospitalized cases, of which the number of deaths in 2022 was the highest, accounting for 262 cases (13.52%). The overall number of deaths showed a slight upward trend, which may be related to the improvement of medical insurance reimbursement ratio and medical treatment level. The number of non-hospital deaths in the hospital is five times that of in-hospital deaths. Therefore, attaching importance to pre-hospital and in-hospital first aid is the key work of the hospital, strengthening the discipline construction of the emergency department, and improving the ability of multi-disciplinary comprehensive cooperation in treatment.
Male patients died more than female patients in this hospital, and the ratio of male to female was 2.22:1, which was consistent with the research results of Salaj D and Campbell JE et al. [20–21]
The reason why men tend to have unhealthy lifestyles such as smoking, drinking and staying up late, and engage in more high-risk industries and heavy physical activities than women may also be related to the different physiological and anatomical characteristics of men and women. There were differences in age, marital status, education level and place of residence among different genders (P < 0.05). The death cases were distributed in all ages, among which 6.66% were from 0 to 19 years old, mainly including congenital heart disease, traffic accidents and accidental injuries. 20 to 59 years old accounted for 42.10%, mainly including sudden death, acute myocardial infarction and traffic accidents; ≥60 years old accounted for 51.24%, mainly including sudden cardiac death, cerebral infarction, lung cancer, liver cancer and lung infection. Hence, society and hospitals need to pay attention to the screening work of newborns in the perinatal period, pay attention to the physical and mental health of children and adolescents, strengthen the learning of road traffic safety knowledge of residents, improve the emergency treatment capacity of hospitals, and establish a sound gerontic disease prevention and treatment system. Among different marital status, married people accounted for the highest proportion (75.13%), followed by unmarried people (13.67%) and widows (8.00%). Among different educational backgrounds, junior high school education accounted for the highest proportion (75.15%), followed by high school (11.15%) and junior college (4.80%). In the distribution of residential areas, the proportion of the dead in the city was the highest (53.56%), followed by the outside of the province (38.44%).
The spectrum of diseases in China has changed in the past 20 years, from infectious diseases to the current non-communicable diseases such as cardiovascular, malignant tumors and accidental injuries [22–23]. According to the statistical analysis results of ICD-10 categories, it is found that the first four causes of death are circulatory system diseases, injury-poisoning, tumor and respiratory system diseases, which is roughly the same as the results of some domestic studies, and is in line with the development trend of diseases in China [24–25]. Cardiovascular disease is the leading cause of death worldwide [26]. This study found that the main causes of death from circulatory diseases were sudden cardiac death, acute myocardial infarction and cerebral hemorrhage. Huang S et al. [27] reported in the literature that the annual incidence of sudden cardiac death in China is about 41.84 per 100,000, which is the country with the highest incidence of sudden cardiac death in the world. injury-poisoning contribute to a significant disease burden in society, with Mulima G et al. [28] reporting that more than five million people die from injuries each year, accounting for 9% of all deaths worldwide. This study found that traffic accidents accounted for about 40%, falling from heights accounted for about 30%, and drowning accounted for about 10%. It is related to the sharp increase in the number of motor vehicles, the rapid development of the construction industry, illegal driving and the weak traffic safety awareness of the public. The global burden of malignant tumors is huge and growing [29]. Lung cancer accounted for a high proportion (22.03%) of the deaths from malignant tumors in our hospital, which was consistent with the findings of Zhang JY [30] and Ha L [31] et al. The occurrence of liver cancer, pancreatic cancer, colorectal cancer and other digestive system tumors may be related to unbalanced diet, excessive drinking, overeating and lack of exercise. Respiratory diseases in our hospital included pulmonary infection (39.9%), chronic obstructive pulmonary disease (9.85%) and hypostatic pneumonia (8.37%), which were consistent with the findings of Halpin DMG et al [32]. Zhang JW et al. [33] reported that the intensification of urban air pollution and the increase of population density could easily lead to the spread of viruses and bacteria and increase the incidence of respiratory diseases.
SARIMA model not only considers seasonal periodicity, but also extracts non-seasonal components according to the sequence changes within seasonal cycle, which has the characteristics of strong practicality and high accuracy[34–35]. Yang WJ et al.[36] used SARIMA(2,2,2)(0,1,1)12 model to predict the number of inpatients in a top three hospital in Zhejiang Province, and the prediction effect was good. Liu JC et al. [37] modeled and predicted the number of inpatients with acute mountain disease (AMS) through ARIMA seasonal product model, and finally determined that ARIMA (1,1,1) (1,0,1)12 model was the optimal model. In this study, the preliminary model was determined based on ACF and PACF diagrams, and the seasonal and non-seasonal factors were comprehensively considered. After several selections and attempts, the ARIMA(2,1,1)(1,1,1)12 model was finally determined as the optimal model. It is found that the relative error between the predicted value and the actual value in 2022 is 9.54%, indicating that the model has good forecasting ability. The SARIMA(2,1,1)(1,1,1)12 model predicted 241 deaths in 2023, with a higher number of 22 deaths in March, May and November. Compared with previous years, the fluctuation is not large, and the hospital can take relevant policies or improvement measures in advance to reduce the number of hospital deaths as much as possible.
Based on the development trend of new deaths, we can make the following recommendations for hospital management. First, the hospital should strengthen the treatment of cardiovascular diseases, on-site personnel should race against time to participate in cardiopulmonary resuscitation and obtain automatic external defibrillator (AED), regularly carry out first aid drills, strengthen emergency rescue ability, and further improve the diagnosis and treatment mode of chest pain center and stroke center. Second, elevating injury-poisoning rescue capabilities, establish trauma green channels, and enhance multidisciplinary collaboration. Third, attach great importance to the construction of tumor and respiratory system disciplines, improving the rate of specialized treatment for the elderly and their own professional capabilities. Finally, it is suggested to add a SARIMA model into the hospital information system to update and predict the changing trend of the number of deaths from various diseases and reserve the urgently needed medical resources in advance.
Some limitations of this study need to be acknowledged. First, the number of in-hospital deaths collected is missing because some patients and family members give up continuing treatment in the hospital, resulting in information bias. Second, we only studied death data for the hospital from 2015–2022, which decreases in accuracy as the SARIMA model's forecast time horizon expands. Finally, the data came from a tertiary general hospital in Hangzhou area, which can only reflect the death situation of the area where the hospital belongs. The research and analysis of the death case data from the perspective of the hospital is not enough to reflect the death situation of the whole Hangzhou area, and the representativeness is limited.
In the future, more death data will be collected to make the study more specific and in-depth, and further research will consider using SARIMA model to build a prediction model of specific disease incidence. We can also try to compare SARIMA model with other models to select a model with higher prediction efficiency. Such as seasonal autoregressive piecewise comprehensive moving average (SARFIMA)[38], SARIMA-ETS-SVR hybrid model [39] and Holt-Winters model [40] .