Research On The Competence Of Clinical Medical Data Mining Talents Based On Demand Analysis

Objective:At present,the analysis models obtained by data mining research on clinical medical big data often cannot meet the actual clinical demands,and there is still a big gap between solving practical problems.The reason is that most studies are based on available data rather than actual clinical demands.Therefore,how to develop data mining for clinical practice demands is an urgent task.In the face of this urgent matter,the questions we need to consider include: What are the actual demands of clinical medical data mining analysis? Based on this demand,what kind of knowledge and skills should the personnel conducting clinical data mining research have? To solve the above problems,this study uses text mining and Delphi method to investigate the demands of clinical medical big data mining and talent competencies,hoping to provide a theoretical basis for future clinical medical big data development strategy research and talent training.Methods:(1)Demand analysis of clinical medical data mining:(a)Text mining analysis: Retrieve clinical medical big data mining related papers from international authoritative medical literature databases,and extract Me SH terms related to clinical medical data mining applications,purposes and intentions in the papers.Then,apply biclustering algorithm according to the Me SH terms co-occurrence matrix automatic to obtain the actual demands of clinical medical data mining,that is,the demands expressed in the form of literature.The result provides a basis for the preparation of follow-up clinical medical big data mining questions and demands questionnaire survey.(b)Delphi survey: According to the clues provided by text mining,under the framework of the anonalous state of knowledge theory,combined with cross-industry standard process for data mining model(CRISP-DM),design and construct the clinical medical big data mining demand questionnaire,and use the Delphi method to analysis.Distribute the clinical medical data mining demand questionnaire to 16 experts engaged in clinical data mining research and management in the fields of computer,clinical medicine,medical informatics and epidemiology,after 2 rounds of questionnaires,the opinions of experts gradually converged,and finally sorted out clinical medical big data mining problems,knowledge status and demands.(2)Construction of a competency model for clinical medical data mining talents: By referring to the concept of talent competency model,this paper aims to timely discover the clinical diagnosis and treatment demands and clinical management demands in clinical medical big data mining by improving the personal ability of clinical medical data mining personnel,and solve many problems caused by the mismatch of knowledge and ability in data mining process.According to the iceberg model,the competency of clinical medical big data mining talents is divided into the part of inner quality below the iceberg and the part of basic knowledge and skills above the iceberg.Based on the results of demand analysis and the principles of the competency iceberg model,the first-level and second-level indicators of the competence of clinical medical big data mining talents are constructed.The clinical medical data mining talent competency questionnaire was distributed to 16 experts in the fields of computer,clinical medicine,medical informatics and epidemiology.After two rounds of questionnaire distribution,the indicators at all levels were selected and optimized according to the mean and coefficient of variation.Preliminarily establish a competency evaluation model for clinical medical data mining talents.(3)Model empirical research: With 46 researchers engaged in clinical medical data mining as the main survey subjects,a questionnaire survey was conducted.Reliability,validity test and variance analysis were conducted to ensure the reliability and objectivity of the evaluation model.At the same time,discover the actual situation and influencing factors of the competence of the current data mining researchers.Results:(1)The biclustering algorithm and Delphi method discover the demands of clinical medical big data mining: The biclustering algorithm is used to extract and anchor part of the clinical medical data mining actual demands expressed in the form of literature,which mainly covers clinical diagnosis and treatment related demands,such as precision medicine,disease risk factor analysis,drug adverse event monitoring,disease prognosis,disease diagnosis,disease treatment and diagnostic imaging.The Delphi survey results supplement the actual demands of clinical medical data mining.The Delphi survey results show that in addition to the demand for clinical diagnosis and treatment,clinical medical data mining also has certain demands for clinical management,such as rehospitalization prediction,adverse event monitoring,clinical pathway optimization,drug prescription sequence,and nosocomial infection prediction,hospitalization time prediction,etc.Under the framework of anonalous state of knowledge theory,combined with the results of two rounds of Delphi surveys,it is found that there is a "gap" between the knowledge of clinical data mining practitioners and their data mining purposes and intentions.This "gap" is specifically manifested as: data understanding problems caused by the mismatch between mining knowledge and data analysis knowledge and purpose intent,especially the problem of insufficient understanding of the specific value of data;data preparation problems caused by the mismatch between data analysis knowledge and purpose intent,especially difficulties in data preprocessing and checking data consistency;model evaluation problems caused by the mismatch between interdisciplinary knowledge and data analysis knowledge and purpose intent;data acquisition problems caused by the mismatch between objective conditions of clinical medical big data mining and purpose intent,especially the low degree of data standardization and the difficulty of data acquisition due to privacy and security requirements.The existence of the "gap" reflects the potential demands of clinical medical data mining,that is,the demands of the knowledge and capabilities should have and the demands of objective conditions of clinical medical data mining.(2)Construction of a competency model for clinical medical data mining talents: This research has initially constructed a three dimensions of clinical medical data mining talent competency evaluation model,including motivation(interest,achievement motivation),personality and characteristics(foresight,insight,data sensitivity,planning and organization ability,expression and presentation ability,continuous learning ability),innovation ability and interpersonal communication ability)and knowledge technology ability(medical knowledge,data mining knowledge,interdisciplinary knowledge,with data literacy,data mining algorithm,model analysis and evaluation,scientific research ability),17 secondary indicators.In terms of content,it includes not only obvious elements such as comprehensive knowledge and data mining skills,but also implicit elements such as achievement motivation and personal qualities.(3)Status of self-evaluation of competence of clinical medical data mining talents: The model constructed in this study was used to evaluate 46 clinical medical data miners.The result found that the overall job competency self-evaluation score of the surveyed miners was 4.094 ±0.762.According to the scale level(1-5points),the surveyed miners' overall competence of the subject is at a relatively high level.Among the three dimensions of motivation(4.488 ± 0.763),personality and traits(4.138 ± 0.930),and knowledge and technical ability(3.914 ± 1.047),the self-rating of knowledge and technical ability is low,which is worthwhile researchers are concerned.Among the self-evaluation indicators,interest,continuous learning ability,and data literacy have higher self-ratings in each dimension.The higher interest and data literacy scores indicate that the surveyed clinical medical data miners are interested in their own data mining and data.In terms of literacy,they are more confident,which is more consistent with the scoring of experts.The continuous learning ability is inconsistent with the element(insight)that experts scored higher in the dimensions of personality and traits,indicating that the clinical medical data miners under investigation still need to strengthen their ability to insight into clinical problems.The competence of clinical medical big data mining talents will be affected by personal sociological characteristics.The results of single factor analysis and multiple linear regression analysis show that gender and age factors have a significant influence on the total self-assessment of clinical medical data mining talents competence score.(p<0.05).In terms of motivation,factors such as gender,age,working years,job title,education background and other factors have no statistically significant influence on the competence of clinical medical data mining talents(p>0.05);in terms of personality and traits,gender and age factors have an impact on clinical Personality and trait self-ratings of medical data mining talents will have a significant influence(p<0.05);in terms of knowledge and technology capabilities,gender factors will have a significant impact on the clinical medical big data mining talents self-ratings(p<0.05).Conclusion:(1)From the perspective of demands,this research focuses on clinical issues,under the framework of the anonalous state of knowledge theory combined with CRISP-DM model,the biclustering algorithm and the Delphi method are used to finally summarize the demands of clinical medical big data mining research as three categories:(a)The actual demands of clinical diagnosis and treatment and clinical management;(b)The demands of clinical medical data mining knowledge and ability should have;(c)The demands of clinical medical data mining objective conditions.(2)This study has constructed a clinical medical data mining talents competency evaluation model with 3 core dimensions and 17 secondary indicators.It has good reliability and validity and can objectively reflect the clinical medical data mining talents' competency requirements.(3)The overall competency of the surveyed clinical medical data mining miners is at a relatively high level.Among the three dimensions of motivation,personality and traits,and knowledge and technical ability,the self-rating of knowledge and technical ability is low.Therefore,the key to the competence of data mining personnel lies in the knowledge and technical ability training,which stimulates intrinsic motivation and deeply digs out personal characteristics and qualities.Generally speaking,age and gender have a greater influence on the competence of clinical medical data mining talents,which may determine the career plan of clinical medical data miners,that is,male clinical medical data miners may be willing to engage in clinical practice for a long time,in order to obtain higher competence and social reputation.