Primary Care Physicians' Antibiotic Prescribing Patterns And Their Influencing Mechanism Based On Capability-Opportunity-Motivation Behavior Model

[Purpose] Antibiotic resistance poses a serious threat to global health and socio-economic development,and the irrational use of antibiotics has further exacerbated the resistance situation,which is particularly severe in primary care.In order to respond to relevant threats,promoting rational antibiotic prescribing behaviors of physicians is essential to alleviate the hazards of drug resistance and maintain public health.However,the current research on antibiotic prescribing patterns is insufficient,and the influencing mechanism is still unclear.This study mainly focuses on primary care physicians,and proposes the following research purposes:(1)identify potential antibiotic prescribing patterns and identify physician groups who need key intervention;(2)develop a measurement tool for factors influencing antibiotic prescribing behavior;(3)establish the influencing mechanism of antibiotic prescribing behavior to identify significant elements and key links of prescribing behavior,as well as supplement theoretical value and practical significance,so as to provide evidence support for policy practice transformation.[Methods](1)Based on outpatient prescription data in primary care institutions,literature research,focus group discussion,descriptive analysis,and univariate analysis were used to summarize the overall level of antibiotic use in primary care and the frequency as well as characteristics of different types of antibiotics.From a comprehensive perspective of the resistance risk,patient health,price,and policy regulation effects,latent profile analysis was performed to explore potential antibiotic prescribing patterns,and to identify key intervention physicians.(2)Literature research and focus group discussion were used to explain the definition of each element of Capability-Opportunity-Motivation Behavior model(COM-B)and put forward research hypotheses,and construct the theoretical model of antibiotic prescribing behavior of primary care physicians.(3)Referring to Hinkin's questionnaire development steps,the measurement tool for factors influencing antibiotic prescribing behavior of primary care physicians was developed following item generation,cognitive interview,questionnaire survey,factor analysis,and internal consistency test.(4)Based on the measurement tool above,face-to-face questionnaire survey was conducted in 49 primary care institutions.Multinomial logistic regression was performed to identify the key influencing factors of antibiotic prescribing patterns.Meanwhile,confirmatory factor analysis,internal consistency,and structural equation model were used to identify key links that affect antibiotic prescribing behavior.Multiple Indicators,Multiple Causes model was established to explore the impact of covariates on the basic model.(5)Based on the identified key physician groups,elements and links that need key intervention during the above empirical analysis,focus group discussion in conjunction with Behavior Change Wheel(BCW)was adopted to understand target behaviors,identify intervention options,determine intervention content and implementation options,so as to develop intervention strategies.[Results](1)Identification of antibiotic prescribing patterns: Based on average number of medicines issued per prescription,average number of antibiotics issued per prescription,percentage of prescriptions involving antibiotics,percentage of prescriptions involving combination of antibiotics,percentage of antibiotic prescriptions involving restricted antibiotics,percentage of antibiotic prescriptions involving antibiotics in Watch and Reserve List,percentage of antibiotic prescriptions involving broad-spectrum antibiotics,percentage of antibiotic prescriptions involving parenteral administered antibiotics,three different levels of potential antibiotic prescribing patterns were systematically identified:high,medium,and low.Among them,22.1% of physicians showed high antibiotic prescribing patterns,belonging to the key intervention group,who were more common among female,40-50 years old,surgery and other departments,as well as physicians who had worked for 20-30 years.(2)Construction of theoretical model: The characteristics of antibiotic prescribing behavior were discussed based on motivation theory,regulation theory,and principal-agent theory,emphasizing the significance of prescribing motivation,incentive regulation,and the principal-agent relationship in deeply understanding antibiotic prescribing behavior.The COM-B theoretical model for antibiotic prescribing behavior of primary care physicians was constructed,and hypothetical relationship between model elements were put forward: capability and opportunity have an interaction with each other,capability and opportunity can directly affect antibiotic prescribing behavior or indirectly affect antibiotic prescribing behavior through motivation,and motivation has a direct impact on antibiotic prescribing behavior.Capability included knowledge,skills,behavior regulation,decision processes;Opportunity included environmental context and resources,and social influence;Motivation included reinforcement,emotions,goals,professional role and identity,beliefs about consequences,beliefs about capabilities,optimism,and behavior intentions.(3)Development of measurement tool: Following Hinkin's questionnaire development steps,an initial item pool containing 68 items was generated in the item generation stage.An initial measurement tool containing 58 items was formed in the cognitive interview stage,and 421 valid questionnaires were retained in the questionnaire survey stage.In the factor analysis and internal consistency test stage,9 dimensions and 26 items were finally retained(Cronbach's alphas for single dimension were 0.754—0.892,overall Cronbach's alpha was 0.904),and a validated measurement tool for factors influencing antibiotic prescribing behavior of primary care physicians was eventually developed(RMSEA=0.061,CFI=0.981,TLI=0.977,SRMR=0.049).(4)Establishment of influencing mechanism:(1)Multinominal Logistic regression results showed that behavior regulation,social influence were the significant determinants of antibiotic prescribing patterns.Compared with physicians with poorer behavioral regulation capability,physicians with stronger behavioral regulation capability were more likely to show a medium antibiotic prescribing pattern rather than a high pattern(?=1.029,P=0.003).Compared with physicians with lower social influence scores,physicians with higher social influence scores were more likely to show a low antibiotic prescribing pattern rather than a high pattern(?=0.380,P=0.047).Gender,age,department,and working years also had effects on antibiotic prescribing patterns.Male were more likely to be medium antibiotic prescribers rather than high prescribers(?=1.660,P<0.001);Physicians who were40-50 years old were more likely to be high antibiotic prescribers compared with those over50 years old(?=-1.584,P=0.015);General practitioners were more likely to be low antibiotic prescribers rather than high prescribers(?=1.142,P=0.003);Internist physicians were more likely to be medium antibiotic prescribers rather than high prescribers(?=1.632,P<0.001);Physicians who had worked for less than 10 years were more likely to be low antibiotic prescribers rather than high prescribers,compared with those who had worked for more than 30 years(?=1.967,P=0.028).(2)COM-B model for antibiotic prescribing behavior of primary care physicians was established(RMSEA=0.069,CFI==0.970,TLI=0.966,SRMR=0.056).Structural equation model results showed that knowledge and skills,and behavior regulation,social influences had positive effects on motivation(?=0.435,P<0.001;?=0.509,P<0.001;?=0.087,P=0.021);Knowledge and skills,social influences were significantly related to behavior regulation(?=0.557,P<0.001;?=0.440,P<0.001).Knowledge and skills were significantly associated with social influence(?=0.160,P=0.001).When associated with different types of antibiotic prescribing behavior,motivation and social influence both had negative effects on self-reported antibiotic prescribing behavior(?=-0.377,P=0.009;?=-0.125,P=0.026);Behavior regulation was positively correlated with self-reported antibiotic prescribing behavior(?=0.261,P=0.007),but it can have a significant regulatory effect on behavior(?=-0.377,P=0.009)through the mediating effect of motivation(?=0.509,P<0.001);Knowledge and skills were directly associated with the average number of antibiotics issued per prescription and the percentage of prescriptions involving antibiotics(?=0.240,P=0.003;?=0.305,P=0.001);Behavior regulation had negative influence on the average number of antibiotics issued per prescription,and percentage of prescriptions involving combination of antibiotics(?=-0.255,P=0.042;?=-0.266,P=0.041);Social influence had negative effects on the percentage of antibiotic prescriptions involving broad-spectrum antibiotics and percentage of antibiotic prescriptions involving parenteral administered antibiotics(?=-0.129,P=0.031;?=-0.149,P=0.015).Multiple Indicators,Multiple Causes model results showed that age,annual family income both had negative effects on knowledge and skills(?=-0.112,P=0.047;?=-0.110,P=0.041),and department was a significant indicator of knowledge and skills,as well as motivation(?=-0.117,P=0.038;?=-0.165,P=0.004).(5)Development of intervention strategy: Based on the above empirical results,a targeted intervention strategy was developed following the main steps and principles of BCW.Five intervention functions including education,training,restriction,modeling,and enablement,and four policy categories including financial measures,public services,communication,and management regulations,as well as nine behavior change techniques including information about health consequences,feedback on behavior results,tips,behavioral norms,instructions on how to perform behaviors,social support(unspecified),social support(practical),goal setting(results),and problem solving were eventually determined.Finally,intervention strategies based on the Template for Intervention Description and Replication(TIDie R)were briefly described.[Conclusions] The antibiotic prescribing behaviors of primary care physicians can be divided into three different levels of antibiotic prescribing patterns: high,medium,and low.Among them,the populations who need key intervention were more common among female,40-50 years old,surgery and other departments,and physicians who had worked for20-30 years.The measurement tool and COM-B model had good reliability and validity,and can be used to identify significant influencing factors and key links of antibiotic prescribing behavior.The intervention strategy following the BCW steps and principles included 5 intervention functions,4 policy categories,and 9 behavior change techniques.It has a strong targeted intervention effect and can provide evidence-based support for policy practice transformation.In the future,it is necessary to improve the establishment of the information system and laboratory capacity,optimize the primary pharmaceutical service system and encourage medication consultation services,promote patient-centered communication as well as multidisciplinary communication and collaboration,and improve the incentive and supervision mechanism of antibiotic prescribing behavior,and strengthen the application of big data.[Innovation and Deficiency]The innovations of this research mainly included the following three points:(1)From a comprehensive perspective of the resistance risk,patient health,price,and policy regulation effects,this study improved the previous identification of antibiotic prescribing patterns.Based on the following eight prescribing indicators: average number of medicines issued per prescription,average number of antibiotics issued per prescription,percentage of prescriptions involving antibiotics,percentage of prescriptions involving combination of antibiotics,percentage of antibiotic prescriptions involving restricted antibiotics,percentage of antibiotic prescriptions involving antibiotics in Watch and Reserve List,percentage of antibiotic prescriptions involving broad-spectrum antibiotics,percentage of antibiotic prescriptions involving parenteral administered antibiotics,three different levels of potential antibiotic prescribing patterns were systematically identified: high,medium,and low;and key intervention group were also identified.It overcomes the shortcomings of previous single measurement of antibiotic prescribing behavior,provides a comprehensive perspective for in-depth understanding of antibiotic prescribing behavior,and provides a foundation for the selection of targeted intervention physicians.(2)Following Hinkin's questionnaire development steps,the measurement tool for factors influencing antibiotic prescribing behavior of primary care physicians was developed following item generation,cognitive interview,questionnaire survey,factor analysis,and internal consistency test.The measurement tool for factors influencing antibiotic prescribing behavior was developed and validated for the first time,including 9 dimensions and 26 items,which fills the gaps of COM-B in the field of antibiotic prescribing scale development,and enriches the existing measurement tool,as well as provides an effective measurement tool for identifying the obstacles and facilitators of rational antibiotic prescribing behaviors.(3)This study established the COM-B model for antibiotic prescribing behavior of primary care physicians for the first time,including knowledge and skills,behavior regulation,social influence,and motivation,which fills the gaps in the empirical research of COM-B model in the field of antibiotic prescribing behavior,and provides evidence-based basis and effective operating mechanism for the development of targeted intervention strategies.As for research limitations,considering that the use of antibiotics in primary care is particularly serious,and the specific use between primary care and secondary or tertiary hospitals was different,therefore the empirical research in this study was mainly based on primary care settings.Future research can further validate the measurement tool and the COM-B model in secondary and tertiary hospitals,and further enrich the sample selection to verify the generalization of the research conclusion.Besides,the disease diagnosis codes in the information system of primary care institutions were not complete,thus this study did not combine specific diseases,etc.,and future studies can consider further controlling specific diseases or conditions,thereby deepening the exploration on the antibiotic prescribing behaviors.In addition,due to the unavailability of data at institutional level,this study did not explore the differences of antibiotic use among institutions,which can be further explored in the future.Moreover,based on the empirical results in this study,targeted intervention strategies were developed following the BCW steps and principles,but the specific intervention strategies have not been evaluated in conjunction with the actual practice.Future studies should refine and optimize the intervention strategy,and further evaluate the intervention effects.