Changes in urologic research from a new perspective: Text mining analysis of publication topics

Abstract

Purpose

This study aimed to explore the trends in research keywords and topics in the field of urology based on text mining over the recent decades. The investigation looked into changes in frequent subject keywords and the trends in prevailing research topics, as reflected in representative urology journals over recent decades.

Materials and Methods

A total of 27,129 bibliographic documents were collected from four urology journals, including European Urology, Journal of Urology, BJU International, and World Journal of Urology. The study then examined the changes in the most frequent author keywords over the decades. Moreover, structured topic modeling was employed to identify twenty prevailing research topics in urology and to examine their trends across different periods.

Results

The study observed consistently increasing patterns in author keywords and topics related to the prostate and oncology. Conversely, research fields such as pediatrics, male infertility, voiding dysfunction, and cancer biology exhibited a downward trend in urology. Potential factors or reasons underlying these trends were further discussed in this study.

Conclusions

This exploratory study uncovered major research topics in the discipline of urology. The findings of this study depict the domain of urology research in recent decades, providing insights for both researchers and clinicians seeking to better understand the research trends in the discipline.

Graphical Abstract

INTRODUCTION

According to the American Urological Association, urology is a branch of healthcare that deals with diseases of the male and female urinary tract (kidney, ureters, bladders, and urethra). It also encompasses the male reproductive organs responsible for fertility (penis, testis, scrotum, and prostate) [1]. Within urology, there exist diverse subspecialties such as pediatric urology, urologic oncology, kidney transplantation, male infertility, urinary tract stones, female urology, and neurourology, each encompassing various research areas. Publication serves as a significant indicator for evaluating the quality of scientific research in a specific discipline. An increasing number of journals dedicated to urology have emerged, and the total count of published articles has been growing each year. Nevertheless, despite the increased productivity in urology research, there have been few attempts to gather and analyze research trends in urology.

Text mining relies on natural language processing techniques to analyze large-scale textual data, enabling the discovery of knowledge and interesting patterns [2]. It serves as a compelling tool for exploring specific research domains and tracing research trends. By examining the content of publications in a certain field, text mining allows us to uncover the prevailing themes of scholarly communications and their evolution over time. Text mining has been utilized to analyze research trends in various medical research fields. For instance, Kim and Delen [3] applied text mining to analyze articles from leading medical informatics journals to identify the primary subject areas within the field and track their changes over time. Similar endeavors utilizing text mining have investigate research topics and trends across diverse health science areas, such as emergency medicine [4], intensive medicine care [5], inflammatory bowel disease [6], drug abuse [7], and more [8].

However, there have been relatively fewer attempts in the discipline of urology to investigate research topics and trends using text mining on large datasets. In urology, prior studies that investigated research areas relied on manual classification of research fields [9]. Oher studies analyzed trends in authorship in the field [10,11,12,13]. Few studies have adopted text mining methods to holistically understand research topics in the domain of urology. In this study, we aimed to understand the field of urology comprehensively by investigating the trends of research keywords and topics based on text mining over the past decades.

The following research questions guided the investigation: What are the changes of the most frequent subject keywords in urology research in recent decades? What are the prevailing topics in urology research in recent decades? What are the trends of those prevailing topics over time in urology in recent decades?

MATERIALS AND METHODS

1. Data collection

To address the research questions, the study collected bibliographic records from four highly cited journals in the discipline of urology: European Urology, Journal of Urology, BJU International, and World Journal of Urology. We excluded journals such as Journal of Endourology, Journal of Pediatric Urology, and The Prostate, as they primarily focus on specific subspecialties within urology. Including these journals could have created a misleading impression that the topics they cover represent a disproportionately large share of the broader field of urology. The data collection process utilized the Scopus database (https://www.scopus.com) in February 2023. The following filters were applied to retrieve bibliographic records from Scopus: (1) year range – “from 2000 to 2022”; (2) document type – “article”; (3) language – “English” only; and (4) source type – “journal”. Any articles lacking an abstract were further removed from the dataset. In total, 27,129 bibliographic documents were collected from the selected journals: 4,028 from European Urology , 12,650 from Journal of Urology, 6,886 from BJU International, and 3,565 from World Journal of Urology. As the study aims to investigate research trends over time, the dataset was divided into five periods, as outlined in Table 1. Each period spans 5 years, except for Period 5, which covers 3 years. Two sets of text corpora were generated from the collected bibliographic records, including (1) the list of author keywords and (2) the text corpus consisting of titles and abstracts. The collected text underwent the following pre-processing steps: all text was changed to lowercase, and copyright information in abstracts was removed. Additionally, subheadings in structured abstracts, such as “Objective,” “Purpose,” “Patients and Methods,” “Results,” “Conclusion,” and several others, were eliminated. For topic extraction, stop-words were removed from the text and then stemming was applied before conducting topic modeling.

Table 1. Number of bibliographic documents each period.

Period	Year	No. of documents
1	2000–2004	6,380
2	2005–2009	6,867
3	2010–2014	6,564
4	2015–2019	4,785
5	2020–2022	2,533

2. Data analysis

Multiple analysis methods were employed to respond to the research questions. First, we tallied the occurrences of author keywords within each period. Author keywords typically encompass the most crucial terms, highlighting the content or topic of each article, as directly assigned by the authors themselves. We extracted author keywords that occurred at least 10 times within each period to identify trends in popular terms across the periods. Since author keywords were not chosen from controlled vocabularies, we manually re-categorized those terms that represented the same concept.

Second, we explored prevailing topics and/or themes from the selected journals based on text mining. Titles and abstracts were chosen for topic analysis, which provide richer information about the papers beyond author keywords. Abstracts usually include research purposes, themes, methods, results, and other summary information. Topic modeling was employed as a method to detect latent key topics or concepts from the text corpus, consisting of titles and abstracts. Topic modeling is an unsupervised machine learning method that can be used to uncover latent topics underlying a large set of text documents (citation). For topic modeling, we utilized the R “stm” package (https://cran.r-project.org/web/packages/stm/index.html). We tested and evaluated different numbers of topics (i.e., k=15; k=20; k=25; k=30; k=35) to identify a reasonable model that can explain the field of urology appropriately. In this paper, we reported a model of twenty topics, which was deemed to be comprehensive enough to cover most prevailing research topics in urology. More importantly, we investigated the patterns of topic probabilities over time to trace the trends of research topics between 2000 and 2022.

RESULTS

1. Most frequent author keywords (RQ 1)

We investigated the most frequent keywords appearing in the selected journals. The analysis of author keywords holds significance since these subject terms were directly chosen by the authors, who have the most accurate comprehension of the article’s content. We extracted all author keywords that appeared 10 times or more within each period. Then, we further regrouped the terms indicating the same concept. Table 2 lists the top twenty terms in each period. Across all periods, the most popular keyword terms included “urinary bladder”, “prostate cancer”, “prostate”, “prostatic neoplasms”, and others. We observed differences in frequent terms between early and later periods. In Periods 1 and 2, the top three terms consistently were “urinary bladder (1st)”, “prostate (2nd)”, and “prostatic neoplasms (3rd)”, from Period 3, “prostate cancer” emerged as the first, while “urinary bladder” dropped to lower rank. From Period 3, “robotics” was ranked highly, 8th in Period 3, 4th in Period 4, and 7th in Period 5. Additionally, “magnetic resonance imaging” was highly ranked in Periods 4 and 5. Notably, “COVID-19” was ranked at 35th in Period 5, revealing its influence on urology research.

Table 2. Top 20 author keywords in each period.

Rank	Period 1 (2000–2004)	Period 2 (2005–2009)	Period 3 (2010–2014)	Period 4 (2015–2019)	Period 5 (2020–2022)
1	Urinary bladder	Urinary bladder	Prostate cancer	Prostate cancer	Prostate cancer
2	Prostate	Prostate	Prostatectomy	Prostatic neoplasms	Prostatic neoplasms
3	Prostatic neoplasms	Prostatic neoplasms	Urinary bladder	Prostatectomy	Bladder cancer
4	Kidney	Prostate cancer	Prostatic neoplasms	Robotics	Prostatectomy
5	Prostatectomy	Prostatectomy	Prostate	Urinary bladder	Magnetic resonance imaging
6	Prostate cancer	Kidney	Kidney	Bladder cancer	Uroonc
7	Laparoscopy	Laparoscopy	Renal cell carcinoma	Renal cell carcinoma	Robotics
8	Prostatic hyperplasia	Urinary incontinence	Robotics	Magnetic resonance imaging	Benign prostatic hyperplasia
9	Urinary incontinence	Renal cell carcinoma	Bladder cancer	Cystectomy	Urology
10	Bladder neoplasms	Carcinoma	Urinary incontinence	Prostate cancer-specific mortality	Renal cell carcinoma
11	Urethra	Prostate specific antigen	Carcinoma	Complications	Prostate cancer-specific mortality
12	Ureter	Nephrectomy	Complications	Nephrolithiasis	Cystectomy
13	Penis	Urethra	Laparoscopy	Biopsy	Urolithiasis
14	Prostate specific antigen	Bladder neoplasms	Cystectomy	Urinary incontinence	Complications
15	Renal cell carcinoma	Bladder cancer	Nephrectomy	Prostate specific antigen	Lower urinary tract symptoms
16	Testis	Erectile dysfunction	Prognosis	Benign prostatic hyperplasia	Nephrolithiasis
17	Nephrectomy	Complications	Prostate specific antigen	Prognosis	Urinary incontinence
18	Biopsy	Prognosis	Outcomes	Nephrectomy	Urinary bladder neoplasms
19	Urodynamics	Pediatrics	Biopsy	Lower urinary tract symptoms	Quality of life
20	Urinary diversion	Ureter	Urinary bladder neoplasms	Outcomes	Biopsy

2. Prevailing topics in urology (RQ 2)

To explore various research topics within urology, a topic modeling analysis was conducted (Table 3). The results uncovered various distinct subject areas within urology. Firstly, topics related to prostate research were identified, including T1 (prostate biopsy), T4 (lower urinary tract symptoms and erectile dysfunction [LUTS and ED]), T5 (oncologic outcomes), T6 (randomized controlled trial for LUTS and ED), T10 (radical treatment for prostate cancer), T12 (cancer biology), T13 (risk factors for prostate cancer), T15 (voiding dysfunction), and T20 (benign prostatic hyperplasia). Secondly, numerous topics represented oncology-related research, such as T1 (prostate biopsy), T2 (surgery for kidney cancer), T5 (oncologic outcomes), T10 (radical treatment for prostate cancer), T12 (cancer biology), T13 (risk factors for prostate cancers), T17 (chemotherapy), and others. Thirdly, topics relevant to andrology included T4 (LUTS and ED), T6 (randomized controlled trial for LUTS and ED), and T7 (male infertility). Fourthly, topics concerning voiding dysfunction comprised T4 (LUTS and ED), T6 (randomized controlled trial for LUTS and ED), T8 (interstitial cystitis/bladder pain syndrome), T11 (stress urinary incontinence), T14 (urethroplasty), T15 (voiding dysfunction), and others. Additionally, the topic model revealed pediatric hydronephrosis (T3) and urinary stone (T19) as distinct topics.

Table 3. Topic modeling result.

Topic	Annotation	Proportion (%)	Top probability terms
T1	Prostate biopsy	6.66	Biopsi prostat cancer detect imag patient predict use score signific
T2	Surgery for kidney cancer	4.56	Renal nephrectomi kidney patient partial tumor function mass surgeri transplant
T3	Pediatric hydronephrosis	4.63	Patient children urinari tract year reflux obstruct month age follow-up
T4	LUTS and ED	5.35	Symptom men score function urinari sexual life erectil qualiti age
T5	Oncologic outcomes	7.22	Patient carcinoma surviv tumor stage tumour cell associ grade predict
T6	Randomized controlled trial for LUTS and ED	5.53	Group treatment patient signific random studi effect trial improv control
T7	Male infertility	2.79	Testicular patient sperm men testi test male germ varicocel case
T8	Interstitial cystitis/bladder pain syndrome	2.74	Bladder patient pain syndrom chronic pelvic urin cystiti urinari interstiti
T9	Tissue engineering/regenerative medicine	3.92	Tissu rat group muscl control model nerv smooth signific week
T10	Radical treatment for prostate cancer	5.63	Patient radic prostatectomi node cancer lymph recurr local prostat posit
T11	Stress urinary incontinence	3.60	Incontin patient urinari contin women stress implant urethr sling sphincter
T12	Cancer biology	5.67	Cell express cancer gene protein growth tumor prostat use activ
T13	Risk factors for prostate cancer	7.42	Prostat cancer men risk psa associ year age increas antigen
T14	Urethroplasty	5.92	Patient stone use urethr penil stent strictur ureter techniqu repair
T15	Voiding dysfunction	3.88	Bladder pressur detrusor rat increas effect stimul overact contract activ
T16	Clinical guidelines	7.32	Use urolog review care studi data clinic treatment cost manag
T17	Chemotherapy	4.17	Patient therapi cancer treatment chemotherapi respons surviv bladder metastat progress
T18	Surgical techniques	7.08	Complic patient laparoscop outcom surgeri postop group radic rate time
T19	Urinary stone	2.99	Stone urinari urin infect patient calcium risk oxal antibiot increas
T20	Benign prostatic hyperplasia	2.94	Prostat volum benign patient laser hyperplasia flow bph transurethr resect

LUTS, lower urinary tract symptoms; ED, erectile dysfunction.

3. Trends of research topics in urology (RQ 3)

We examined research topic trends between 2000 and 2022 (Fig. 1). The topics that exhibited an overall upward trajectory included T1 (prostate biopsy), T5 (oncologic outcomes), T10 (radical treatment for prostate cancer), T13 (risk factors to prostate cancer), T16 (clinical guidelines), T17 (chemotherapy), and T18 (surgical techniques). These topics are mostly related to prostate and oncology research. In contrast, downward trends were observed in T3 (pediatric hydronephrosis), T7 (male infertility), T8 (interstitial cystitis/bladder pain syndrome), T9 (tissue engineering/regenerative medicine), T11 (stress urinary incontinence), T12 (cancer biology), T14 (urethroplasty), and T15 (voiding dysfunction). Several topics showed more random patterns without any steady trend, for example, T2 (surgery for kidney cancer), T4 (LUTS and ED), T6 (randomized controlled trial for LUTS and ED), T19 (urinary stone), and T20 (benign prostatic hyperplasia).

Fig. 1. Estimated topic proportion changes over time.

DISCUSSION

This study explored various research subjects in urology based on text mining. We collected 27,129 bibliographic documents from four representative journals in urology. Then, we analyzed frequent author keywords and conducted topic modeling to identify prevailing research topics. To the best of the authors’ knowledge, a similar analysis of research trends has not been previously undertaken in urology.

We have identified several key observations regarding the research trends in urology. First, the analysis of author keywords and topic modeling unveiled changes in the research landscape in urology. One of the notable aspects is the consistent increase in author keywords and topics related to the prostate and oncology throughout the observed timeframe. On the other hand, research fields such as pediatrics, male infertility, voiding dysfunction, and cancer biology showed a continuous downward trend. Various factors including medical needs, scientific opportunities, and funding can be considered as the potential drivers for these research trends. First, we can consider the differences in disease burden and the resulting clinical activity, leading to varying medical needs. According to data from the Global Burden of Disease study, total cancer ranked as the second-highest cause of disability-adjusted life years, deaths, and years of life lost behind cardiovascular diseases globally. The number of new cancer cases increased from 18.7 million in 2010 to 23.6 million in 2019, marking a 26.3% increase [14]. Among these, urologic cancers, particularly prostate cancer, witnessed a significant surge in disease burden. Compared to 1990, prostate cancer had the largest increase at 169.11%, followed by kidney cancer at 154.78%, and bladder cancer at 123.34% [15]. Furthermore, when analyzed by incidence per 100,000, the incidence rate for benign prostatic hyperplasia significantly increased from 24.94 in 1990 to 31.77 in 2017 [16]. Also, we can point to disparities in research funding as one of the underlying reasons for differences in research trends. According to the analysis of National Institutes of Health (NIH) funding, approximately 62.1% of the NIH funding received by the Department of Urology was related to urologic oncology, with prostate cancer research accounting for 52.9% of that portion. In contrast, NIH funding for research related to benign prostatic hyperplasia received 5.1%, pediatric urology received 4.2%, calculi research received 3.2%, female urology received 3.1%, and male infertility received 0.7%, indicating relatively minor funding allocation to these areas [17]. The substantial high medical needs and funding in oncology and prostatic disease are likely associated with the increased clinical activity of healthcare professionals and a corresponding rise in research.

Second, despite the steady increase in research related to prostate or oncology, our analysis revealed a consistent decline in publications related to cancer biology in general urology journals. The decline in publications related to cancer biology can be attributed to various factors. Some possible hypotheses include: (1) The diversity and increase in specialized journals imply that research on cancers can be published across various platforms. Although research related to cancers is generally increasing, these papers might be dispersed across diverse journals instead of being concentrated in general urology journal. (2) Researchers may change their criteria for selecting journals to publish their work. For instance, researchers previously involved in cancer biology might submit their research results to journals in diverse fields, leading to a decrease in publication volume in general urology journal. (3) The research trends in the field of urology are changing, potentially drawing more interest from researchers towards topics other than cancer biology. For instance, increased interest in research for clinical outcomes of medical or surgical treatment could result in a decline in publications related to cancer biology.

Finally, this is one of the few that employed text mining methods in exploring research topics and trends in the discipline of urology. Many previous studies that investigated bibliographic data in urology have mostly focused on authorships, citations and impact factors, descriptive statistics, or quantitative statistical analysis [9,10,11,12,13,18,19]. Little research has adopted the use of text mining techniques, particularly structural topic modeling, within the context of urology bibliographic analysis. Although the analyses of structured fields such as authorship, citations and impact factors provide valuable insights into understanding the field, they often fall short in revealing semantically hidden topics and the shifts of content over time. On the contrary, unstructured textual content records the primary outcomes of urology studies. Text mining directly targets this content and provides insights into what is being discussed in the literature over time. The scalability of modern text mining techniques allows processing a large size of articles efficiently. This line of text mining methods has great implications for research topics and trends analysis. It also makes its way into emerging topic identification and prediction [20]. In this study, we utilized text mining to explore key concepts and topics that highlight the content of scholarly literature in urology. More importantly, the use of structural topic modeling enabled us to analyze the abstract text, which contains a richer context, including important elements of each article, such as purposes, objectives, methods, results, implications, and other significant components.

This study is not without limitations. First, we analyzed only four journals in urology. Those four journals might not represent the entirety of urology. However, they are the oldest and most high-impact journals in the field of urology. Second, author-assigned keywords did not use controlled vocabulary. Even though we manually refined the observed terms, there are still chances of potential bias caused from unstandardized keywords. Third, topic modeling analysis was done with only titles and abstracts. Full-text data was not used for topic exploration. A recent comparative study n topic extraction from full-text of papers versus their corresponding abstracts may offer a more efficient approach. This is because abstracts can convey similar information as the full text but with significantly fewer words [21]. Despite these limitations, our research holds significance in providing a detailed analysis of trends in urological research. Such analysis is valuable as it aids researchers in allocating research resources more efficiently by identifying crucial areas of focus. It also assists in publishing papers in appropriate journals. Furthermore, engaging in research on more pivotal topics allows researchers to amplify their contributions and impact within the field.

CONCLUSIONS

In this paper, we present an exploratory study that identifies major research topics in the urology research domain. By examining the trends of research keywords and topics through text mining, this study provides a comprehensive overview of the evolution of scholarly output in urology over recent decades. These findings not only offer a deeper understanding of the current research landscape but also serve as a valuable resource for researchers. This study highlights key areas of focus and emerging trends, which can guide the development of future research projects and collaborations. By fostering a more informed approach to research and practice, we hope this study contributes to the balanced growth of urologic research and its application to clinical care. Future work could delve deeper into the factors influencing these trends to further optimize research strategies in the domain.