The cintent of reseach paper

General practice / family medicine (GP/FM) has become a clinical field that is not only expected to provide high-quality primary health care, but also to contribute to medical education¹⁾. Furthermore, the World Health Organization has recommended increasing the opportunities for medical students to experience primary health care in the real world²⁾. On the other hand, medical schools across the globe are currently undergoing curriculum reforms in accordance with the global standards of the World Federation for Medical Education³⁾. According to the Japanese version of these standards, GP/FM is highlighted as an important clinical discipline, which means that it is a mandatory subject for all medical students⁴⁾. Despite recognition of the importance of GP/FM, the opportunities for medical students to participate in GP/FM clerkships are still inadequate around the world. It was reported that 50 out of 259 universities (19%) had no or very brief GP/FM clerkships, even in Europe where GP/FM is further developed⁵⁾. In order for GP/FM clerkships to be improved into undergraduate medical education, it is necessary to clarify the impacts of GP/FM clerkships⁶⁾. Two important systematic reviews were reported in 2015, discussing the impact of GP/FM clerkships on undergraduate medical education^7,8). Turkeshi and colleagues reviewed quantitative studies around the world, whilst Park and colleagues reviewed quantitative and qualitative studies in the United Kingdom^7,8). A Japanese quantitative study was reported in 2020⁶⁾, using the evaluation items of the Model Core Curriculum for Medical Education in Japan⁹⁾. However, no studies have been reported that analyzed qualitative data quantitatively. Reflective writing in medical education has been introduced in many medical schools worldwide¹⁰⁾. Analyzing the reflective writing of medical students may be useful for further understanding the impact of clinical clerkships on medical students. However, qualitative judgment in appraising reflective writing may be confounded by reader bias¹¹⁾. Text mining can extract keywords (frequent words) from large samples of reflective writing in an efficient and objective manner. Text mining can also identify relationships between extracted words¹²⁾.

The purpose of the present study was to clarify which aspects of GP/FM clerkships the students deemed to be most valuable and which experiences they felt were particularly useful, using text mining to analyze reflective writing.

Participants and setting

Throughout the academic year 2018-2019, clinical clerkships accounted for 50 weeks in both the fifth and sixth years at Fukushima Medical University. The GP/FM clerkship is a mandatory placement undertaken by fifth-year students for 5 days. The study subjects were fifth-year students (n = 125) in the academic year 2018-2019. The settings were three GP/FM clinics: Hobara Central Clinic and Kitakata Centre for Community and Family Medicine, both located in rural areas; and Hoshi Yokozuka Clinic, which serves an urban population. Family doctors certified by the Japan Primary Care Association work at these clinics and teach medical students. Students were dispatched in groups of 1 to 2 for their GP/FM clerkship.

Contents of GP/FM clerkships

Table 1 shows objectives for the GP/FM clerkship based on the Model Core Curriculum for Medical Education in Japan⁹⁾. Table 2 shows the standard schedule. It consisted of outpatient care and home visits. During outpatient care, the students performed history taking and vital sign measurements of patients, and then observed consultations by family doctors and received feedback on history taking, diagnostic reasoning, and patients’ backgrounds. When participating in home visits, the students assisted the family doctors but also interacted with family members and other health professionals such as home-visiting nurses and home helpers. Their daily reflections were facilitated by family doctors on site. Joint reflections were conducted by university faculty via video conference.

Table 1. The objectives in GP/FM clerkships based on the Model Core Curriculum for Medical Education in Japan

Table 2. The standard clerkship schedule

Data source

On the final day of their GP/FM clerkship, students completed a reflection sheet. We collected free descriptions on the aspects of the GP/FM clerkship that the students deemed to be most valuable and which experiences they felt were particularly useful via this reflection sheet. The actual question in Japanese was as follows: 「実習で特に印象に残ったことおよびそこから得た学びについて自由に記載してください」

Statistical analysis

Text mining was used to extract frequent words (nouns) from the students’ free descriptions. In the Japanese language, some verbs contain a noun (e.g., the verb 訪問するis a noun訪問 with the verbalizing suffix する). In the present study, any verbs containing a noun were separated into their noun and verb components. A co-occurrence network map was created to illustrate the relationships between the most frequently occurring words. Co-occurrence refers to how many times high-frequency words appear in the text in proximity to other high-frequency words¹²⁾. This relationship is called an association, which is calculated numerically as a figure between 0 and 1^12,13). A co-occurrence network map was used to visualize how keywords group together throughout the entire text, with connecting lines marked with numerical values in the map indicating association strength¹²⁾. All analysis was performed using KH Coder 3.0 (http: //khcoder.net/en/index.html), a free downloadable multilingual text-mining program developed by Koichi Higuchi, Ritsumeikan University, Japan¹⁴⁾.

Analysis was done in Japanese and translated into English for the report. In the process of interpreting the results of text mining along with the co-occurrence network map, it was considered necessary to refer to the original context to determine the validity, so original text and extracted words are listed in the Results section.

Ethics approval

Ethics approval was obtained from the Fukushima Medical University Human Research Ethics Committee, approval number #30153.

124 students participated in the study; their baseline characteristics are shown in Table 3. The total number of sentences extracted from their free reflections was 321 and the total number of all words, including verbs and other parts of speech, was 10,627. The average number of words written by one student was 137.5 (SD 72.2) and the average number of sentences was 2.5 (SD 1.1). Table 4 shows the top 20 most frequent words. The top five frequently occurring words were patient, home-visit, medical practice, medical care, and family. Figure 1 shows the co-occurrence network map linking each frequent word. The strongest co-occurrence relationship was recognized between home-visit and medical practice (Jaccard coefficient 0.66). A co-occurrence relationship was also recognized between home-visit and family (Jaccard coefficient 0.2). A translation of typical sentences, including extracted words, is shown in Table 5.

Table 3. Student baseline characteristics

Table 4. List of frequent words

Fig. 1. The co-occurrence network map between each frequent word.

A community is a “part that is relatively strongly connected to each other.” The KH coder automatically detects the community, groups it, and shows the results in different colors. Words that do not form groups with other words are shown in white. Words in the same community are connected with solid lines, while words in different communities are connected with dashed lines. The number on the line is the Jaccard coefficient. This coefficient is calculated between 0 and 1, and the closer it is to 1, the stronger the relationship.

Table 5. Translation of typical sentences containing extracted words

Amongst the top 20 most frequent words, home-visit, family, history taking, home, nursing, life, community, and outpatient were considered the words indicative of specific meanings. Other words were considered too general to reveal useful data. Home-visit was the most frequent word except for patient, and it appears frequently in the description of home visit (see typical sentences: A, B, C and D in Table 5). In a previous study conducted among the same participants as the present study, home visit was the item with the highest self-evaluation score at the end of the GP/FM clerkship⁶⁾. In this respect, the results of the student self-evaluation and the reflective writing are in agreement. family, home, and life appear frequently in the description of the patient’s psychological / social background (see typical sentences: A, B and C in Table 5). Nursing and community appear frequently in the description of team-based health care (see typical sentences: C and D in Table 5). In the previous study, it was suggested that psychological/social background and team-based health care are difficult to learn during the clerkships at teaching hospitals⁶⁾; therefore, we considered it important that medical students were able to learn such skills and valued the GP/FM clerkship experience. History taking and outpatient appear frequently in the description of active participation in outpatient care (see typical sentences: E in Table 5). A systematic review in the UK reported the benefits of GP/FM clerkships, including deeper exploration of psycho-social factors and social determinants of health, exposure to multidisciplinary learning skills, and increased feedback on history taking and physical examinations compared to that experienced in hospitals⁸⁾. Therefore, we find that our results, obtained by the text mining of students’ reflective writing, are consistent with the findings obtained by conventional methods.

In the co-occurrence network map, the strongest co-occurrence relationship was found between home-visit and medical practice, because the expression home-visit medical practice (訪問診療) was often used (see typical sentences: A, B and C in Table 5). It is noteworthy that a co-occurrence relationship was recognized between home-visit and family. By referring to the students’ answers, it emerges that medical students may learn the necessity of care for patients’ families, as well as the patients themselves, in a home care setting (see typical sentences: A, B and C in Table 5). It is possible that students learned the importance of responding to the different needs of each family and caring for all aspects of life through multidisciplinary collaboration, in a home care setting. A study at another university in Japan reported that medical students learn through home care practice that home care can also reduce the anxiety and burden of patients’ families¹⁵⁾.

The current study has some limitations that should be addressed. First, it was based at a single educational institution. However, it is noteworthy that the data were obtained from three different GP/FM clinics located in urban and rural areas. Second, because the text data are student-reported outcomes, there might have been some self-reporting bias. It is possible that the students made positive statements out of gratitude and consideration for the family doctors they interacted with. Third, reflective writing was used to explore the impact of GP/FM clerkships on medical students in the present study, but there is insufficient evidence to use reflective writing as an assessment tool¹⁶⁾. As a future improvement based on the second and third limitations, if students are guaranteed that the doctors they interact with will be masked from their reflective answers, and if they are anonymously and freely written, this study design may be an effective tool in obtaining answers as honest as possible. Fourth, as the exposure to GP/FM was only for 5 days, it is unclear whether the impact is sustainable. Fifth, we were unable to measure the average number of outpatients or home care patients whom a single medical student encountered in the morning or afternoon. The volume of encounters may have affected the impact of the students’ reflections.

Data suggest that medical students may learn the necessity of care for the family as well as the patient in a home care setting. Using text mining to analyze student reflective writing may be useful for further understanding the impact of clinical clerkships on medical students.

We thank Professor Aya Goto of the Center for Integrated Sciences and Humanities at Fukushima Medical University, for her useful comments on the conception and design of the study.

We thank Doctor Tomoo Hidaka of the Department of Hygiene and Preventive Medicine at Fukushima Medical University, for his useful comments on the analysis and interpretation of the data.

We thank Chisato Kubo of the Office for Gender Equality Support at Fukushima Medical University, for inputting the data.

We are indebted to Assistant Professor Maham Stanyon of the Center for Medical Education and Career Development at Fukushima Medical University, for her English editing of the manuscript.

We also wish to thank the Scientific English Editing Section of Fukushima Medical University.

The authors declare no conflicts of interest.