Design, and study setting
A mixed methodology was employed: (i) a prospective observational study to investigate incidences and the nature of reporting medication errors; (ii) a social network analysis of patterns of interactions within a hospital medication system.
This study was undertaken within an inpatient department of a 90-bed district hospital in Thailand, which included male and female general medical wards, an intensive-care unit (ICU), and a private medical ward. The average occupancy rate was ~ 60 inpatients during the study period.
In this hospital, the flow of the medication process began with (i) prescribing (physicians) by a hand-written entry on to the order sheet, (ii) pre-transcribing (nurses) by electronic scanning of the order sheet and transferring its information on to the pharmacy computer terminal, (iii) transcribing and appropriate labels pertaining to the medicine (pharmacists), (iv) pre-dispensing (pharmacist assistants) by filling the medicine pack as per label, and the pharmacist verifying accordance between the pack contents, the label, and the prescription, (v) dispatching the dispensed medications (pharmacists) to the appropriate ward (vi) pre-administration (nurses) to verify a match between the medication and the corresponding order sheet (vii) the medication administered (nurses) to the patient.
Characteristic of participants
During the study period, 65 employees worked within the inpatient medication system: 7 physicians (1 medical specialist in internal medicine, and 6 junior physicians), 5 pharmacists, 44 nurses, 5 pharmacist assistants, and 4 unskilled ancillary workers.
Data collection and analysis
To understand social networks in the medication system, this study required data on both reports about medication errors, and the consultation networks. Reports of these medication errors were collected at every stage of the medication process [17] and were graded as A to I as defined by the National Coordinating Council for Medication Error Reporting and Prevention (NCCMERP) [18]. Such error reports were collected for the 1-month period through the hospital spontaneous reporting system [19]. Staff were encouraged to report any event to the risk management committee. Reporters were anonymised using a code according to the protocol of data collection on consultation networks. On completing the data gathering, the reports were analysed using descriptive statistics.
Data on consultation networks in the medication system were collected from all 65 hospital participants by a structured face-to-face interview in a designated room to recall the previous months data collection period, in which data on medication errors were obtained. Two main questions were asked: (i) whether the participant consulted others about obstacles to medication use and medication-related problems. When the participant had consulted with others, they then provided the information about frequency of consultations based on recall over the last month; and (ii) who was consulted. (Additional file 1) Interview questionnaires were adapted from previous studies [15, 20] and validated by three experts in social network analyses and hospital medication systems. In this study, an ‘item objective congruence (IOC) index’ greater than 0.5 indicated an acceptable questionnaire content [21]. The information derived from the interview was then transformed to social network data (sociogram). Each pair of a relationship indicated a directed link from an informant to the consulting person in the network.
Social network data were analysed using UCInet version 6 [22] which provided both mathematical and visual analyses of network relationships as a sociogram representing complex intercommunication among hospital staff. Mathematic social network measures include the degree of centrality (in and out), and betweenness centrality. In-degree is a measure of the number of links from other staff directed to an informant, while out-degree indicates the number of links through which the informant sought consultations with other network members [15] or in lay terms, it indicates the extent to which he/she receives queries, or asked for opinion/advice. In this study, where the topic is about medication, those receiving more queries are believed to be expert about medication or are trustworthy within a network by other staff. Out-degree is the extent to which the informant seek advice from other staff. In this study, the questions were limited to the obstacles to medication use or medication-related problems; therefore, those who have out-degree possibly notice potential events, or have been involved in medication-related problems, or may hesitate to continue the process of medication use. Betweenness centrality was the degree of shortest path of the consultation seeking between all staff passing through the informant, calculated as the number of direct links from all staff of one group to all staff in another group to a specified individual. In lay terms, betweenness centrality is the extent that an individual influences the communication of other staff within the network. By passing the queries from one to another, those with high betweenness centrality seem to be a bridging person. An individual who provides a link between two different clusters demonstrates high betweenness centrality and is recognized as a ‘bridger’. Without a bridger, the network is disconnected. Therefore, finding bridgers is of importance because they represented those who often transfer information between groups and provide valuable opportunities for innovation [23].
Factors affecting the reporting of medication errors were analysed using linear regression [24] with the forward selection method using IBM SPSS Statistics version 23. Individual characteristics, in-degree, out-degree, and betweenness centrality were selected as the variables that influence reporting of medication errors. All statistical analyses were performed at a significance level of < 0.05.