A Comparative Study of Factors Inﬂuencing the Survival Time of Invention Patents in Chinese Universities

. Based on the fact that the granted amount of patents in Chinese universities have increased signiﬁcantly, we took the invention patents applied by universities in Sichuan Province from 2001 to 2015 as a sample and adopted Kaplan-Meier survival analysis and Cox proportional hazards model. We found signiﬁcant di ﬀ erences in the survival time of invention patents in the three types of universities Based on this, we conducted an empirical study and comparative analysis on the factors inﬂuencing the survival time of the three types of universities. It was found that the number of citations and the number of patent families have di ﬀ erent e ﬀ ects on the survival time of invention patents in the three types of universities; the examination time and the number of inventors have negative e ﬀ ects on the survival time of invention patents in the three types of universities; the number of claims, the number of IPCs and the number of forward citations have positive e ﬀ ects on the survival time of invention patents in the three types of universities. This paper can provide some reference for universities to cultivate high-value patents


Introduction
Scientific and technological achievements of colleges and universities are an important driving force for industrial progress and regional economic development [1]. Invention patents are an important carrier of universities' scientific and technological innovation achievements, as well as a bridge link connecting innovation and the market, and promoting the transformation of innovation achievements into real productive forces. The Law of the People's Republic of China on Promoting the Transformation of Scientific and Technological Achievements, as amended in 2015, encourages universities and enterprises to jointly implement the transformation of scientific and technological achievements, and the National Plan for the Protection and Utilization of Intellectual Property Rights promulgated in 2021 points out the need to enhance the effectiveness of the transfer and transformation of intellectual property rights. These new requirements for deepening reforms in the field of intellectual property highlight the important role of intangible assets such as patents in serving social and economic development.
Since the formal implementation of the Patent Law of the People's Republic of China in 2001, the total number of patents and authorizations of universities in China has increased significantly. According to the statistics of science and technology of colleges and universities in 2021 [2], Chinese colleges and universities applied for 193474 invention patents and got 116043 invention patents granted in 2020. O'donoghu [3] found that the granting of a patent only means that the applicant can start to use the patent right, but the effective maintenance of the patent right is the beginning of its function. Therefore, the number of invention patents cannot fully and objectively reflect the scientific and technological innovation ability of universities. At present, there is a phenomenon of "weight but not quality" in university invention patents, and these low-value patents are difficult to transfer and transform. Therefore, colleges and universities have a large number of "dormant" patents that have not been put into production and application. Many theoretical studies have shown that patent maintenance time is one of the key indicators to evaluate the performance of the patent system. Since the concept of patent quality is not yet clear, scholars have considered patent maintenance time as an important factor to measure the technical and economic value of patents, and thus indirectly studied the issues related to "patent quality". Pakes et al. [4] first started to use patent duration as an indicator of patent value from the perspective of production activities. Schankerman researched French patents and found that the patent value was related to the maintenance time [5]. In the annual report on effective patents in China, the State Intellectual Property Office of China pointed out that the longer the maintenance time of a patent, the longer the time it takes to create economic benefits and the higher the market value [6]. The patent maintenance time can reflect both the operational performance of the patent system and the ability of the innovation subject to manage the patent. Therefore, how to prolong the survival time of university invention patents, promote the high-quality development of university invention patents, and better play the role of universities in serving economic and social development has become an important problem to be solved. This paper selects the invention patents of universities in Sichuan Province as the research object, refers to the list of "DoubleFirst − Class" university project, conducts Kaplan-Meier survival analysis on the invention patents of universities in Sichuan Province applied from 2001 to 2015, uses Cox proportional hazards model to empirically study the impact of legal and technical factors on the survival time, and deeply compares and analyzes the impact factors of the survival time of invention patents of different types of universities in Sichuan Province, and put forward targeted suggestions to promote the survival time of invention patents from the perspective of examination time. This study scientifically identifies the factors influencing the survival time of invention patents in universities, provides a reference for the cultivation of high-value patents in universities in China, and makes certain contributions to building China into a strong country in intellectual property strategy.

Literature review
Many experts have investigated the maintenance time of patents from the perspective of patent quality, and the core issue involved in the study is the factors influencing patent maintenance time. Tong and Frame [7] analyzed the US patents granted from 1970 to 1990 and found that the more the number of claims, the better the maintenance of the patent, but they did not conduct a detailed study on the relationship between the number of claims and other indicators. Scholars incorporated the IPC classification into patent studies. Qiao [8] found that there are differences in the maintenance time of patents in different technical fields; Brown [9] proved that the patent maintenance rate in the chemical and electrical fields is higher. Hall et al. [10] found by studying US patents and citations from 1963 to 1995 that the number of citations was a decisive factor in whether a patent was maintained, and the more citations a patent had, the higher its value. Harhoff et al's results [11] showed that the number of citations in the patent literature and the number of patent forward citations were positively correlated with its value. Many subsequent studies have used the number of citations and the number of forward citations as an indicator of patent value. Griliches [12] and Liu et al. [13] found that the larger the scope of patent families, the stronger the patentee's willingness to maintain the patent. Xiao et al. [14] used patent maintenance time as a proxy variable for patent quality and found that patent examination time had a negative impact on patent maintenance time, but their rationality of patent examination as an evaluation indicator was not explored more deeply. In addition, Liu et al. [15] used multiple linear regression models to find that the number of patent families, the number of inventors, innovation subjects, and patent examination time had an impact on patent maintenance time.
For the current situation of patent maintenance in Chinese universities, scholars have used various methods to carry out research. In terms of evaluation methods, Shen [16] defined three dimensions of patent indicators and got an evaluation system with 12 indicators, and used entropy value method and characteristic root method for analysis and comparison; Sun et al. [17] constructed a patent quality measurement index system from three stages and two dimensions and conducted an empirical study on the example of C9League. The realization of patent value in universities mainly depends on marketization approaches such as patent transactions. Li et al. [18] constructed a patent maintenance survival analysis model to study the influence of technology market characteristics on patent maintenance time in universities. Hu et al. [19] used the regression analysis method to select several indicators including the number of inventors to analyze the factors influencing the patent maintenance time in Jiangsu universities.
The existing literature focuses on the maintenance time of patents in a certain region or field, but there are few papers investigate the factors influencing the maintenance time of patents in Chinese universities; the literature that has studied the maintenance time of patents in Chinese universities mainly adopts linear regression methods, and the research perspective is relatively single. Taking the universities in Sichuan Province as an example, this study adopts Cox proportional hazards model to investigate the factors that affect the survival time of invention patents in different types of universities, and analyzes and compares the results, which is conducive to discovering the differences of factors influencing the survival time of invention patents in different types of universities, and on this basis, suggesting suggestions for the long-term development of invention patents in different types of universities.

Data
We use the IncoPat (i.e., an important business partner of Clarivate in China, and its address is https://www.incopat.com/ )for data collection. On April 10, 2022, we selected lapsed granted invention patents in the IncoPat database with application dates from January 1, 2001, to December 31, 2015, where the applicants were 134 universities in Sichuan Province, and obtained a total of 10,084 invention patents. According to the type of applicants, universities are divided into four categories: world − class university construction universities, f irst − class discipline construction universities, ordinary undergraduate universities and colleges, and the number of invention patents is 6148, 2773, 1131, and 32 in order. Since the number of invention patents whose applicants are colleges is too few, only 32, they are not included in the scope of the study.
Because of the long period of research and the complexity of invention patents, the data collection process is described as follows. Firstly, during the research period, for universities that only have name changes, the names of universities in different periods are used for searching, and then the search results were combined; for universities with mergers, the names of component colleges and universities were searched first, and then the search results were combined. Secondly, The reasons for choosing the research period from 2001 to 2015 are as follows: In 2001, the Patent Law of the People's Republic of China was officially implemented, and the revised law not only helps to mobilize the enthusiasm of university researchers and improve the efficiency of scientific research but also helps to promote the transformation of university scientific research results into real productivity. Considering the lag of the expiration date and citation of invention patents, as well as the development trend of the survival time of invention patents taking a long time, 2015 is finally chosen as the data cut-off time.

Variables
(1)Dependent variable The dependent variable of this study is the survival time (T imesurvival), which is the time in days from the date of grant announcement to the expiration date. To maintain the validity of a patent, the patentee will first measure the possible economic value of the patent and then decide whether to pay the annual fee, which will increase with the increase of the patent survival time. Therefore, the act of patent maintenance has certain strategic and tactical significance. Since the order of magnitude of a patent lifetime is far greater than the explanatory variable, the logarithm method is adopted in the subsequent analysis.
(2)Independent variables Combining the existing literature and actual data, this study considers the impact of the following independent variables on the survival time of university patents in Sichuan Province.
a. the number of claims (Num_claims) The number of claims is the number of items and the number of patent rights requested for protection. It reflects the breadth of protection of a patent and correlates well with the value of a patent. Generally speaking, the more the number of claims, the stronger the technical complexity contained in that patent and the higher the value of the patent. The unit is pcs.
b. the number of IPCs (Num_IPC) The International Patent Classification (hereinafter referred to as IPC) table is an internationally used tool for classifying and searching patent documents. An invention patent may involve many different types of technology, and there will be multiple IPC numbers. In this paper, the data is pre-processed and the number of IPC groups is used to indicate the number of IPCs. The unit is pcs.
c. the number of inventors (Num_inventors) An inventor is a person who makes creative contributions to the substantive features of invention and creation. Generally speaking, the more inventors, the more scientific knowledge can be gathered, reflecting the application of different inventors' knowledge levels, and the higher the innovation of patents. The unit is pcs.
d. the number of citations (Num_citations) The number of patent citations is the number of times that the patent mentions other patents in the background technology. The number of patent citations reflects information about the economic value of the patent, such as the underlying innovative technology, the economic performance of the company, the patentee's performance in the technical field, etc [20]. The unit is times.
e. the number of forward citations (Num_ f orwardcitations) The number of forward citations is the number of times a patent has been mentioned in the background technology by other patents. The citation analysis of patents can find important patents. The higher the number of forward citations, the higher the influence and value of patents on subsequent patents. The unit is times.
f. the number of patent families (Num_ f amilies) A patent family is a group of patent documents that have at least the same priority and have been applied for, published, or approved many times in different countries or international patent organizations with the same or basically the same content. It represents the market coverage of a patent, and to some extent is proportional to the importance of the patent. The unit is pcs.
g. the examination time (T ime_examination) The patent examination time in this paper is the time from the application date to the grant date of the invention patent, and the sum of it and the patent survival time is the patent maintenance time, in days. The patent examination time is obtained by pre-processing the data. Because the magnitude difference between the data and other explanatory variables is too large, the method of taking its logarithm is used in the subsequent analysis.

Research Method
Cox proportional hazards model is a semi-parametric regression model proposed by British statistician D.R. Cox in 1972. By establishing a regression model, we can evaluate the impact of covariates on the survival function and then predict the risk rate. Cox proportional hazards model has two advantages. First, it has no requirements for data distribution and can utilize all data, which facilitates a more realistic and objective analysis of the data; second, it can handle truncated data and is suitable for analyzing the effect of covariates on survival time. We use Cox proportional hazards model to analyze patents. Firstly, there will always be invalid and valid patents at any time when we collect data; secondly, patents will go through multiple status changes from the application, grant to invalid, which is typical survival data.
Expression for the Cox proportional hazards model is h(t, X) = h 0 (t)exp(β 1 x 1 )exp(β 2 x 2 ) · · · exp(β n x n ). Where X = (x 1 , x 2 · · · x n ) is a covariate; h(t, X) e risk function related to the covariate at time t; h 0 (t) is the benchmark risk rate, which is the risk rate at moment t without the influence of covariates; β is the bias regression coefficient of the independent variable; β 1 , β 2 · · · β n can be obtained iteratively by maximum likelihood method, β i < 0 indicates that x i is a protective factor and β i > 0 indicates that x i is a risk factor. Table 1 shows the descriptive statistical results of each variable. Comparing the standard deviation, we can see that the distribution of other variables is relatively centralized except for the survival time and examination time. Table 2 shows the results of Kaplan-Meier survival analysis on the sample data, and the test analysis results are all significant at the 0.001 level from the perspective of applicant categories, indicating that the survival time of invention patents in different categories of universities has significant differences. Comparing the survival time of 25%, 50%, and 75% of the observed individuals in table 2, it can be found that the universities that world − class   Figure 1 shows the survival curve graph obtained by studying the survival time of the sample invention patents. Observing the figure, in terms of applicant categories, the survival of invention patents is, in order from good to bad: the world-class university construction universities, the first-class discipline construction universities, and ordinary undergraduate universities.

Regression result analysis 4.2.1 Correlation test
The covariates of the three types of universities were first analyzed for correlation, and the correlation coefficients between each covariate were less than 0.2, with low correlation and no co-linearity, enabling the inclusion of all covariates in the model.

Proportional risk hypothesis test
As can be seen from Table 3, the invention patents of world-class university construction universities do not pass the proportional risk assumption of continuous variables in the number of citations, so the Cox regression model should not be directly used, adding the role term of the number of citations and time, and using Cox regression model with time-dependent variables for analysis. The invention patents of first-class discipline construction universities and ordinary undergraduate universities meet the proportional risk hypothesis test and can be university construction universities perform best, followed by the f irst − class discipline construction universities and ordinary undergraduate universities. Figure 1 shows the survival curve graph obtained by studying the survival time of the sample invention patents. Observing the figure, in terms of applicant categories, the survival of invention patents is, in order from good to bad: the world − class university construction universities, the f irst − class discipline construction universities, and ordinary undergraduate universities.

Regression Result Analysis
(1)Correlation test The covariates of the three types of universities were first analyzed for correlation, and the correlation coefficients between each covariate were less than 0.2, with low correlation and no co-linearity, enabling the inclusion of all covariates in the model.
(2)Proportional risk hypothesis test As can be seen from table 3, the invention patents of world − class university construction universities do not pass the proportional risk assumption of continuous variables in the number of citations, so the Cox regression model should not be directly used, adding the role term of the number of citations and time, and using Cox regression model with time-dependent variables for analysis. The invention patents of f irst − class discipline construction universities and ordinary undergraduate universities meet the proportional risk hypothesis test and can be directly analyzed by regression with the Cox model. (

3)Result analysis
The three Cox regression models obtained by the applicant category all passed the significance test and are analyzed in the following order. As shown in table 4, the number of claims, the number of IPCs, the number of citations, and the number of patent families can reduce the survival risk of invention patents, and their relative risk rates decrease by 3.9%, 4.9%, 6.2% and 7.5% for each unit increase, respectively. In addition, the number of inventors, number of citations, and examination time are risk factors, and each unit increase in them increases the relative risk rate by 2.8%, 5.2%, and 18.4%, respectively.
It can be seen from table 5 that the number of citations jeopardizes the survival time of invention patents in the first-class discipline construction universities, but the effect is not significant because it does not pass the hypothesis test. The number of claims, the number of IPCs, the number of citations, and the number of patent families can reduce the survival risk of invention patents, and their relative risk rate decreases by 3%, 5.1%, 7%, and 7.7% for each unit increase. In addition, the number of inventors and examination time are risk factors, and each unit increase in them increases the relative risk rate by 3.3% and 13.9%, respectively.
As can be seen from table 6, the number of patent families has a positive effect on the survival time of invention patents in ordinary undergraduate universities, but it does not pass the hypothesis test, so the effect is not significant. The number of claims, the number of IPCs, the number of citations, and the number of patent families can reduce the survival risk, and their relative risk rate decreases by 3.4%, 2.4%, 2.2%, and 16.4% for each unit increase. In addition, the number of inventors and examination time jeopardize the survival time of   invention patents, and each unit increase in them increases the relative risk rate by 2% and 20.3%.

Discussion
The comparative analysis of the three COX regression models mentioned above reveals that the same covariate may have different directions and degrees of influence on the survival time of the three types of university invention patents. It reflects the complexity and variability of studying survival when the amount of patent data is large. In addition, the existence of timedependent variables in world − class university construction universities indicates that the effect of covariates will change with time.
The number of citations has a significant negative effect on the invention patents of world − class university construction universities, has no significant negative effect on the invention patents of the f irst − class discipline construction universities, and has a significant positive effect on the survival time of invention patents of ordinary undergraduate universities. Sapsalis believes that there may be a negative correlation or no relationship between patent citation and patent value because any innovation subject can obtain public information such as scientific papers. The number of patent families is a protective factor for the survival risk of invention patents in world − class university construction universities and f irst − class discipline construction universities, but a risk factor for the survival risk of invention patents in ordinary undergraduate universities. Lanjouw et al. found that the number of patent families has a positive effect on the quality of patents by studying the American patent samples. It may be that the small amount of data on the number of patent families for invention patents in ordinary undergraduate universities causes patent families to be a risk factor.
The examination time and the number of inventors have a negative impact on the survival time of invention patents in the three types of universities. With the overall maintenance time of invention patents remaining relatively stable, longer examination time will limit the practical application of invention patents by patentees and is not conducive to the survival of invention patents. In addition, the examination time reflects a certain extent the examination efficiency of the examination authority on the uniqueness, novelty, and utility of the patent [21]. In the study of the number of inventors on the quality of patents, foreign scholar M.reitzing found that the number of inventors has a positive impact on the value of patents; while certain domestic scholars found that for various reasons, there are discrepancies between the actual inventors of patents in China and the inventors listed in the literature [22].
The number of claims, the number of IPCs, and the number of forward citations have positive effects on the survival time of three types of university invention patents. Liu, Barney et al. believe the higher the number of claims, the broader the scope of rights protection, and then the higher the level of quality of that patent in terms of the law. In addition, Chinese Patent Law stipulates that a patent application with more than 10 claims is subject to additional fees. Therefore, a patent with a larger number of claims has a higher chance of becoming a high-value patent. The more IPCs a patent has, the more technical fields it covers and the more it can be transposed and applied in multiple fields [23]. Lerner's research found that the technical field of a patent represents the importance of the technology and that the technical field of a patent is positively correlated with the quality of the patent. In domestic and foreign studies, the number of forward citations is positively correlated with patent maintenance time and patent survival time, and a similar conclusion is reached in this paper.

Summary
This paper takes the invention-granted patents applied by universities in Sichuan Province from 2001-2015 as the research object, and continuously tracks the grant date and expiration date of invention patents. The factors influencing the survival time of the three types of Based on the above findings, this paper concludes that shortening the examination time can help applicants obtain authorization faster, prolong the survival time of invention patents and promote the transfer and transformation of invention patents. Universities are the main body of knowledge and technological innovation and produce a large number of patents every year. We can establish diversified review channels and special review communication channels according to the characteristics of colleges and universities. In addition, for invention patents of universities in new technology fields, consider assigning high-level and high-quality experts and examiners.
This study draws some useful conclusions based on invention patent data, but there are still shortcomings. When analyzing the influence of the number of patent families on the survival time of invention patents in ordinary undergraduate universities, we only remove the data volume with the number of patents in the same family of 0 and then make a simple analysis, which cannot give a more reasonable explanation, and the specific reasons need to be further studied. This paper takes Sichuan province as an example to study the factors influencing the survival time of invention patents in different types of universities, but there are obvious differences in the scientific and technological innovation ability of universities in different provinces in China, and the results of this paper may not be universal.