Web-based textual analysis of free-text patient experience comments from a survey in primary care

Patient Comments and Patient Experience Ratings Are Strongly Correlated With Emergency Department Wait Times

BACKGROUND AND OBJECTIVES

Hospitals and clinicians increasingly are reimbursed based on quality of care through financial incentives tied to value-based purchasing. Patient-centered care, measured through patient experience surveys, is a key component of many quality incentive programs. We hypothesize that operational aspects such as wait times are an important element of emergency department (ED) patient experience. The objectives of this paper are to determine (1) the association between ED wait times and patient experience and (2) whether patient comments show awareness of wait times.

METHODS

This is a cross-sectional observational study from January 1, 2019, to December 31, 2020, across 16 EDs within a regional health care system. Patient and operations data were obtained as secondary data through internal sources and merged with primary patient experience data from our data analytics team. Dependent variables are (1) the association between ED wait times in minutes and patient experience ratings and (2) the association between wait times in minutes and patient comments including the term wait (yes/no). Patients rated their "likelihood to recommend (LTR) an ED" on a 0 to 10 scale (categories: "Promoter" = 9-10, "Neutral" = 7-8, or "Detractor" = 0-6). Our aggregate experience rating, or Net Promoter Score (NPS), is calculated by the following formula for each distinct wait time (rounded to the nearest minute): NPS = 100* (# promoters - # detractors)/(# promoters + # neutrals + # detractors). Independent variables for patient age and gender and triage acuity, were included as potential confounders. We performed a mixed-effect multivariate ordinal logistic regression for the rating category as a function of 30 minutes waited. We also performed a logistic regression for the percentage of patients commenting on the wait as a function of 30 minutes waited. Standard errors are adjusted for clustering between the 16 ED sites.

RESULTS

A total of 50 833 unique participants completed an experience survey, representing a response rate of 8.1%. Of these respondents, 28.1% included comments, with 10.9% using the term "wait." The odds ratio for association of a 30-minute wait with LTR category is 0.83 [0.81, 0.84]. As wait times increase, the odds of commenting on the wait increase by 1.49 [1.46, 1.53]. We show policy-relevant bubble plot visualizations of these two relationships.

CONCLUSIONS

Patients were less likely to give a positive patient experience rating as wait times increased, and this was reflected in their comments. Improving on the factors contributing to ED wait times is essential to meeting health care systems' quality initiatives.

Sharing Patient Praises With Radiology Staff: Workflow Automation and Impact on Staff

OBJECTIVE

This study aims to develop and evaluate a semi-automated workflow using natural language processing (NLP) for sharing positive patient feedback with radiology staff, assessing its efficiency and impact on radiology staff morale.

METHODS

The HIPAA-compliant, institutional review board-waived implementation study was conducted from April 2022 to June 2023 and introduced a Patient Praises program to distribute positive patient feedback to radiology staff collected from patient surveys. The study transitioned from an initial manual workflow to a hybrid process using an NLP model trained on 1,034 annotated comments and validated on 260 holdout reports. The times to generate Patient Praises e-mails were compared between manual and hybrid workflows. Impact of Patient Praises on radiology staff was measured using a four-question Likert scale survey and an open text feedback box. Kruskal-Wallis test and post hoc Dunn's test were performed to evaluate differences in time for different workflows.

RESULTS

From April 2022 to June 2023, the radiology department received 10,643 patient surveys. Of those surveys, 95.6% contained positive comments, with 9.6% (n = 978) shared as Patient Praises to staff. After implementation of the hybrid workflow in March 2023, 45.8% of Patient Praises were sent through the hybrid workflow and 54.2% were sent manually. Time efficiency analysis on 30-case subsets revealed that the hybrid workflow without edits was the most efficient, taking a median of 0.7 min per case. A high proportion of staff found the praises made them feel appreciated (94%) and valued (90%) responding with a 5/5 agreement on 5-point Likert scale responses.

CONCLUSION

A hybrid workflow incorporating NLP significantly improves time efficiency for the Patient Praises program while increasing feelings of acknowledgment and value among staff.

Analysis of YouTube content on substance use disorder treatment and recovery

BACKGROUND AND AIM

Emerging literature suggests the role of social media in substance use disorders (SUD). This study aimed to explore the content of YouTube videos for persons on SUD treatment/recovery, describing the users' exposure and engagement metrics and understanding viewers' perspectives.

METHODS

We generated a set of 10 key phrases to search on YouTube. Eighty eligible videos were analyzed using a mixed-methods approach. Content analysis of all videos and thematic analysis of 30 videos were done using the three most viewed videos from each key phrase. The reliability of videos was assessed using a modified DISCERN. The total number of views, likes, dislikes, and comments were noted and created engagement metrics. The linguistic analysis of viewers' comments was done to assess their perspectives.

RESULTS

Sixty-three (78.8%) videos were from the US, and 59 (73.8%) were intended for persons or families with substance misuse. Persons in recovery uploaded 23 (28.7%) videos. We identified five themes - reasons for using drugs, symptoms of addiction, consequences of drug use, how to stop drug use, and expressed tone in the language. The positivity and relative positivity ratios were highest for videos developed by persons in recovery. There was a negative correlation between the relative positivity ratio and content fostering internalized stigma. Words with negative emotional experiences dominated the viewers' comments.

CONCLUSION

YouTube content on SUD treatment and recovery is popular and revolves around the biopsychosocial understanding of addiction. There is an urgent need for a language policy and regulation of non-scientific content.

Deciphering Latent Health Information in Social Media Using a Mixed-Methods Design

Natural language processing techniques have increased the volume and variety of text data that can be analyzed. The aim of this study was to identify the positive and negative topical sentiments among diet, diabetes, exercise, and obesity tweets. Using a sequential explanatory mixed-method design for our analytical framework, we analyzed a data corpus of 1.7 million diet, diabetes, exercise, and obesity (DDEO)-related tweets collected over 12 months. Sentiment analysis and topic modeling were used to analyze the data. The results show that overall, 29% of the tweets were positive, and 17% were negative. Using sentiment analysis and latent Dirichlet allocation (LDA) topic modeling, we analyzed 800 positive and negative DDEO topics. From the 800 LDA topics—after the qualitative and computational removal of incoherent topics—473 topics were characterized as coherent. Obesity was the only query health topic with a higher percentage of negative tweets. The use of social media by public health practitioners should focus not only on the dissemination of health information based on the topics discovered but also consider what they can do for the health consumer as a result of the interaction in digital spaces such as social media. Future studies will benefit from using multiclass sentiment analysis methods associated with other novel topic modeling approaches.

Leveraging Free-Form Comments to Assess and Improve Patient Satisfaction

This study employed a text-analysis methodology to identify themes within patient comments and measure the relationship of those themes to patient satisfaction. Using these findings, a spreadsheet tool was created to allow a large sample of comments to be readily analyzed. The tool was validated using patient comment data provided by the Family Medicine Residency of Idaho. The tool gives clinicians the ability to easily analyze patient comments and identify actionable measures of patient satisfaction. Additionally, this tool will allow researchers to reduce vast sets of comment text into numerical data suited for quantitative analyses.

Exploring Content Relationships Among Components of a Multisource Feedback Program

INTRODUCTION

A new multisource feedback (MSF) program was specifically designed to support physician quality improvement (QI) around the CanMEDS roles of Collaborator , Communicator , and Professional . Quantitative ratings and qualitative comments are collected from a sample of physician colleagues, co-workers (C), and patients (PT). These data are supplemented with self-ratings and given back to physicians in individualized reports. Each physician reviews the report with a trained feedback facilitator and creates one-to-three action plans for QI. This study explores how the content of the four aforementioned multisource feedback program components supports the elicitation and translation of feedback into a QI plan for change.

METHODS

Data included survey items, rater comments, a portion of facilitator reports, and action plans components for 159 physicians. Word frequency queries were used to identify common words and explore relationships among data sources.

RESULTS

Overlap between high frequency words in surveys and rater comments was substantial. The language used to describe goals in physician action plans was highly related to respondent comments, but less so to survey items. High frequency words in facilitator reports related heavily to action plan content.

DISCUSSION

All components of the program relate to one another indicating that each plays a part in the process. Patterns of overlap suggest unique functions conducted by program components. This demonstration of coherence across components of this program is one piece of evidence that supports the program's validity.

The Framework for Responsible Research With Australian Native Plant Foods: A Food Chemist's Perspective

Australia is a rich source of biodiverse native plants that are mostly unstudied by western food science despite many of them being ethnofoods of Australian Indigenous people. Finding and understanding the relevant policy and legal requirements to scientifically assess these plants in a responsible way is a major challenge for food scientists. This work aims to give an overview of what the legal and policy framework is in relation to food chemistry on Australian native plant foods, to clarify the relationships between the guidelines, laws, policies and ethics and to discuss some of the challenges they present in food chemistry. This work provides the framework of Indigenous rights, international treaties, federal and state laws and ethical guidelines including key legislation and guidelines. It discusses the specific areas that are applicable to food chemistry: the collection of plant foods, the analysis of the samples and working with Indigenous communities. This brief perspective presents a framework that can be utilized by food chemists when developing responsible research involving plant foods native to northern Australia and can help them understand some of the complexity of working in this research area.

Moving toward AI-assisted decision-making: Observation on clinicians' management of multimedia patient information in synchronous and asynchronous telehealth contexts

BACKGROUND

Artificial intelligence (AI) intends to support clinicians' patient diagnosis decisions by processing and identifying insights from multimedia patient information.

OBJECTIVE

We explored clinicians' current decision-making patterns using multimedia patient information (MPI) provided by AI algorithms and identified areas where AI can support clinicians in diagnostic decision-making.

DESIGN

We recruited 87 advanced practice nursing (APN) students who had experience making diagnostic decisions using AI algorithms under various care contexts, including telehealth and other healthcare modalities. The participants described their diagnostic decision-making experiences using videos, images, and audio-based MPI.

RESULTS

Clinicians processed multimedia patient information differentially such that their focus, selection, and utilization of MPI influence diagnosis and satisfaction levels.

CONCLUSIONS AND IMPLICATIONS

To streamline collaboration between AI and clinicians across healthcare contexts, AI should understand clinicians' patterns of MPI processing under various care environments and provide them with interpretable analytic results for them. Furthermore, clinicians must be trained with the interface and contents of AI technology and analytic assistance.

User Reviews of Depression App Features: Sentiment Analysis

Background

Mental health in general, and depression in particular, remain undertreated conditions. Mobile health (mHealth) apps offer tremendous potential to overcome the barriers to accessing mental health care and millions of depression apps have been installed and used. However, little is known about the effect of these apps on a potentially vulnerable user population and the emotional reactions that they generate, even though emotions are a key component of mental health. App reviews, spontaneously posted by the users on app stores, offer up-to-date insights into the experiences and emotions of this population and are increasingly decisive in influencing mHealth app adoption.

Objective

This study aims to investigate the emotional reactions of depression app users to different app features by systematically analyzing the sentiments expressed in app reviews.

Methods

We extracted 3261 user reviews of depression apps. The 61 corresponding apps were categorized by the features they offered (psychoeducation, medical assessment, therapeutic treatment, supportive resources, and entertainment). We then produced word clouds by features and analyzed the reviews using the Linguistic Inquiry Word Count 2015 (Pennebaker Conglomerates, Inc), a lexicon-based natural language analytical tool that analyzes the lexicons used and the valence of a text in 4 dimensions (authenticity, clout, analytic, and tone). We compared the language patterns associated with the different features of the underlying apps.

Results

The analysis highlighted significant differences in the sentiments expressed for the different features offered. Psychoeducation apps exhibited more clout but less authenticity (ie, personal disclosure). Medical assessment apps stood out for the strong negative emotions and the relatively negative ratings that they generated. Therapeutic treatment app features generated more positive emotions, even though user feedback tended to be less authentic but more analytical (ie, more factual). Supportive resources (connecting users to physical services and people) and entertainment apps also generated fewer negative emotions and less anxiety.

Conclusions

Developers should be careful in selecting the features they offer in their depression apps. Medical assessment features may be riskier as users receive potentially disturbing feedback on their condition and may react with strong negative emotions. In contrast, offering information, contacts, or even games may be safer starting points to engage people with depression at a distance. We highlight the necessity to differentiate how mHealth apps are assessed and vetted based on the features they offer. Methodologically, this study points to novel ways to investigate the impact of mHealth apps and app features on people with mental health issues. mHealth apps exist in a rapidly changing ecosystem that is driven by user satisfaction and adoption decisions. As such, user perceptions are essential and must be monitored to ensure adoption and avoid harm to a fragile population that may not benefit from traditional health care resources.

Understanding patient reviews with minimum supervision

Understanding patient opinions expressed towards healthcare services in online platforms could allow healthcare professionals to respond to address patients' concerns in a timely manner. Extracting patient opinion towards various aspects of health services is closely related to aspect-based sentiment analysis (ABSA) in which we need to identify both opinion targets and target-specific opinion expressions. The lack of aspect-level annotations however makes it difficult to build such an ABSA system. This paper proposes a joint learning framework for simultaneous unsupervised aspect extraction at the sentence level and supervised sentiment classification at the document level. It achieves 98.2% sentiment classification accuracy when tested on the reviews about healthcare services collected from Yelp, outperforming several strong baselines. Moreover, our model can extract coherent aspects and can automatically infer the distribution of aspects under different polarities without requiring aspect-level annotations for model learning.

Emergency care and the patient experience: Using sentiment analysis and topic modeling to understand the impact of the COVID-19 pandemic

The COVID-19 pandemic has presented many unique challenges to patient care especially in emergency medicine. These challenges result in an altered patient experience. Patient experience refers to the cumulative impression made on patients during their medical visit and is measured by a standardized survey tool. Patient experience is considered a key measure of quality of care. The volume of survey data received makes it difficult to spot trends and concerns in patient comments. Topic modeling and sentiment analysis are well documented analytic techniques that can be used to gain insight into patient experience and make sense of vast quantities of data. This study examined three periods of time, pre, during and post-COVID-19 first wave in order to identify key trends in sentiment and topics related to patient experience. Previously collected, anonymized Press Ganey (PG) survey data was used from three northeastern emergency department that make up an academic emergency department. Data was collected for three contiguous time periods: Pre-COVID-19 (12/10/2019- 3/10/2020), During COVID-19: (3/11/2020–6/10/2020), and Post-first wave COVID-19 (6/11/2020- 9/10/2020). Preprocessing of the data was carried out then a sentiment label (i.e., positive, negative, neutral, mixed) was assigned by the tool. These labels were used to assess the validity of Press Ganey labels. Next, a topic modeling approach from machine learning was used to analyze the contents of the patient comments and uncover concerns and perceptions of patient experiences. Themes that emerged from the analysis of patient comments included concerns over personal safety and exposure to the virus, exclusion of family from decision making and care and high levels of scrutiny over systems issues, care, and treatment protocols. Topic modeling showed shifting priorities and concerns throughout the three periods examined. Prior to the pandemic, patient comments were largely positive and focused on technical expertise and perceptions of competence. New topics and concerns that patients reported relevant to the pandemic were identified during-COVID-19. Comments on systems issues regarding processes to limit viral spread and concerns over family/visitor restrictions were dominant. Although there was evidence of praise and appreciation of the efforts of staff there was also a high level of scrutiny of the processes encountered during the emergency visit. Sentiment analysis and topic modeling offer a unique method for organizing and analyzing the shifting concerns of patients and families. Suggestions of interventions are made to address these evolving concerns. The automation of analysis using artificial intelligence would allow for rapid and accurate analysis of patient feedback.

Using Text Mining Techniques to Identify Health Care Providers With Patient Safety Problems: Exploratory Study

BACKGROUND

Regulatory bodies such as health care inspectorates can identify potential patient safety problems in health care providers by analyzing patient complaints. However, it is challenging to analyze the large number of complaints. Text mining techniques may help identify signals of problems with patient safety at health care providers.

OBJECTIVE

The aim of this study was to explore whether employing text mining techniques on patient complaint databases can help identify potential problems with patient safety at health care providers and automatically predict the severity of patient complaints.

METHODS

We performed an exploratory study on the complaints database of the Dutch Health and Youth Care Inspectorate with more than 22,000 written complaints. Severe complaints are defined as those cases where the inspectorate contact point experts deemed it worthy of a triage by the inspectorate, or complaints that led to direct action by the inspectorate. We investigated a range of supervised machine learning techniques to assign a severity label to complaints that can be used to prioritize which incoming complaints need the most attention. We studied several features based on the complaints' written content, including sentiment analysis, to decide which were helpful for severity prediction. Finally, we showcased how we could combine these severity predictions and automatic keyword analysis on the complaints database and listed health care providers and their organization-specific complaints to determine the average severity of complaints per organization.

RESULTS

A straightforward text classification approach using a bag-of-words feature representation worked best for the severity prediction of complaints. We obtained an accuracy of 87%-93% (2658-2990 of 3319 complaints) on the held-out test set and an F1 score of 45%-51% on the severe complaints. The skewed class distribution led to only reasonable recall (47%-54%) and precision (44%-49%) scores. The use of sentiment analysis for severity prediction was not helpful. By combining the predicted severity outcomes with an automatic keyword analysis, we identified several health care providers that could have patient safety problems.

CONCLUSIONS

Text mining techniques for analyzing complaints by civilians can support inspectorates. They can automatically predict the severity of the complaints, or they can be used for keyword analysis. This can help the inspectorate detect potential patient safety problems, or support prioritizing follow-up supervision activities by sorting complaints based on the severity per organization or per sector.

Computer-assisted textual analysis of free-text comments in the Swiss Cancer Patient Experiences (SCAPE) survey

Background

Patient experience surveys are increasingly conducted in cancer care as they provide important results to consider in future development of cancer care and health policymaking. These surveys usually include closed-ended questions (patient-reported experience measures (PREMs)) and space for free-text comments, but published results are mostly based on PREMs. We aimed to identify the underlying themes of patients’ experiences as shared in their own words in the Swiss Cancer Patient Experiences (SCAPE) survey and compare these themes with those assessed with PREMs to investigate how the textual analysis of free-text comments contributes to the understanding of patients’ experiences of care.

Methods

SCAPE is a multicenter cross-sectional survey that was conducted between October 2018 and March 2019 in French-speaking parts of Switzerland. Patients were invited to rate their care in 65 closed-ended questions (PREMs) and to add free-text comments regarding their cancer-related experiences at the end of the survey. We conducted computer-assisted textual analysis using the IRaMuTeQ software on the comments provided by 31% (n = 844) of SCAPE survey respondents (n = 2755).

Results

We identified five main thematic classes, two of which consisting of a detailed description of ‘cancer care pathways’. The remaining three classes were related to ‘medical care’, ‘gratitude and praise’, and the way patients lived with cancer (‘cancer and me’). Further analysis of this last class showed that patients’ comments related to the following themes: ‘initial shock’, ‘loneliness’, ‘understanding and acceptance’, ‘cancer repercussions’, and ‘information and communication’. While closed-ended questions related mainly to factual aspects of experiences of care, free-text comments related primarily to the personal and emotional experiences and consequences of having cancer and receiving care.

Conclusions

A computer-assisted textual analysis of free-text in our patient survey allowed a time-efficient classification of free-text data that provided insights on the personal experience of living with cancer and additional information on patient experiences that had not been collected with the closed-ended questions, underlining the importance of offering space for comments. Such results can be useful to inform questionnaire development, provide feedback to professional teams, and guide patient-centered initiatives to improve the quality and safety of cancer care.

Quantifying Experiences with Telepractice for Aphasia Therapy: A Text Mining Analysis of Client Response Data

Measures of satisfaction following treatment for aphasia have been limited. The challenge associated with reduced verbal output among many persons with aphasia (PWA) has reportedly been a key reason measures of treatment satisfaction have been limited. A novel approach to measure treatment satisfaction is the use of content analysis (CA), which uses the presence of certain words, themes, or concepts to explore outcomes such as treatment satisfaction particularly among individuals who generate limited output. CA utilizes responses and response patterns to assign meaning to client responses. The aim of this study was to use CA to measure posttreatment satisfaction with a telepractice approach for aphasia treatment. Seventeen PWA received 12 treatment sessions over a 6-week period. At the conclusion of the treatment, CA was utilized to explore patient satisfaction with this treatment approach. The participants reported an overall positive sentiment for the telepractice approach. Two primary topics emerged which were healthcare provider and healthcare delivery, where text analysis revealed discussion of these topics to be centered around being "helpful" and "being effective." This study demonstrated that CA can be an effective approach for determining satisfaction with aphasia treatment particularly among PWA with limited verbal abilities.

Analyzing free-text survey responses: An accessible strategy for developing patient-centered programs and program evaluation

Despite widespread availability of yoga in the Veterans Health Administration (VA), it remains unclear how to best evaluate yoga programs. This is particularly problematic for programs aimed at veterans with mental health concerns, as evaluation typically focuses narrowly upon mental health symptom severity, even though program participants may have other health-related priorities. We analyzed responses to free-text questions on 237 surveys completed by veterans with mental health concerns enrolled in a yoga program at six VA clinics in Louisiana to characterize veteran participants' experiences with yoga. Qualitative analysis resulted in 15 domains reflecting veterans' individual health-related values and priorities. We use results to illustrate the potential for analysis of free-text responses to reveal valuable insights into patient experiences, demonstrating how these data can inform patient-centered program evaluation. The approach we present is more accessible to those responsible for decision-making about local programs than conventional methods of analyzing qualitive evaluation data.

Consequências na alimentação de crianças órfãs após a morte materna: uma investigação por meio de softwares de mineração de texto

Objetiva-se descrever as consequências no aleitamento e na alimentação que terão as crianças órfãs menores de cinco anos em decorrência da morte materna, aplicando-se softwares livres de mineração de texto. Estudo transversal com base em artigos publicados nos repositórios PubMed e BIREME nos temas de morte materna e crianças órfãs. Foram selecionados dez artigos publicados entre 2005 e 2015, de acesso livre, nos quais foram lidos apenas o título ou o resumo e que cumpriam com os critérios. Os arquivos de texto definiram o corpus para análise de conteúdo semiestruturado. Palavras-chave foram incluídas para a mineração. A análise do corpus foi feita com TagCrowd e Textalyser para encontrar os termos mais e menos frequentes, AntConc e Voyant Tools, para extrair palavras-chave na análise de contexto. Foram analisadas 67.642 palavras em dez textos semiestruturados. Os termos CHILDREN (827) e DEATH (821) foram os mais frequentes, e os menos frequentes foram BREASTFEEDING (10) e NUTRITION (4). Foram encontradas 44 concordâncias para o termo raiz BREAST* e 25 para a palavra NUTRITION em orações como: “crianças órfãs têm o aumento de risco de mortalidade por falta de amamentação, e são mais susceptíveis às infecções”. As sentenças de concordância apontam que a mudança no aleitamento materno conduz a uma nutrição pobre, o que deixa o recém-nascido exposto a infecções, aumentando o risco de morte. O processamento de texto com as ferramentas livres foi rápido e permitiu extrair informações úteis e compreensíveis; a análise dos dez artigos mostrou as consequências na alimentação da criança após a morte materna, tendo efeito na morbidade e mortalidade infantil.

Top 10 global educational topics in stroke: A survey by the World Stroke Organization

AIM

As part of a program of work to develop an educational strategy and implementation plan for the World Stroke Organization, we conducted a survey of World Stroke Organization members (health professionals, laypersons (Stroke Support Organizations)) to identify their potential educational needs.

METHODS

We developed a questionnaire to identify priority educational needs in consultation with the World Stroke Organization Education Committee. The World Stroke Organization invited all individual members and associated Stroke Support Organizations to complete the questionnaire via a web-based survey. Survey responses were supplemented by questionnaires emailed directly to key persons in Stroke Support Organizations and information from semi-structured telephone interviews, where necessary. The questionnaire asked respondents to prioritize topics in diagnosis, management of acute stroke, stroke care services, stroke rehabilitation, and stroke prevention. Free-text responses were assessed with word cloud.

RESULTS

The online survey was completed by 264 respondents from 60 countries; 19.1% were from low- and middle-income countries, 59% were stroke specialist physicians, 28% allied health professionals or nurses, 9% Stroke Support Organizations, 4% general physicians. Fifteen Stroke Support Organizations from 11 countries responded to the emailed survey. Seven Stroke Support Organizations' members were interviewed by telephone; one was interviewed in-person. We highlight the two highest priority topics in each of the five questionnaire domains.

CONCLUSION

The 10 priority topics were all applicable in a low- or middle-income setting: setting up and delivering stroke diagnosis, treatment, rehabilitation and prevention services, and emphasized the most basic elements of care. The survey participants have identified a number of key topics that merit inclusion in stroke teaching materials and courses, especially those aimed at practitioners working in resource-limited settings.

Public Perception on Healthcare Services: Evidence from Social Media Platforms in China

Social media has been used as data resource in a growing number of health-related research. The objectives of this study were to identify content volume and sentiment polarity of social media records relevant to healthcare services in China. A list of the key words of healthcare services were used to extract data from WeChat and Qzone, between June 2017 and September 2017. The data were put into a corpus, where content analyses were performed using Tencent natural language processing (NLP). The final corpus contained approximately 29 million records. Records on patient safety were the most frequently mentioned topic (approximately 8.73 million, 30.1% of the corpus), with the contents on humanistic care having received the least social media references (0.43 Million, 1.5%). Sentiment analyses showed 36.1%, 16.4%, and 47.4% of positive, neutral, and negative emotions, respectively. The doctor-patient relationship category had the highest proportion of negative contents (74.9%), followed by service efficiency (59.5%), and nursing service (53.0%). Neutral disposition was found to be the highest (30.4%) in the contents on appointment-booking services. This study added evidence to the magnitude and direction of public perceptions on healthcare services in China’s hospital and pointed to the possibility of monitoring healthcare service improvement, using readily available data in social media.

Understanding "understanding" in Public Understanding of Science

This study examines the conflation of terms such as "knowledge" and "understanding" in peer-reviewed literature, and tests the hypothesis that little current research clearly distinguishes between importantly distinct epistemic states. Two sets of data are presented from papers published in the journal Public Understanding of Science. In the first set, the digital text analysis tool, Voyant, is used to analyze all papers published in 2014 for the use of epistemic success terms. In the second set of data, all papers published in Public Understanding of Science from 2010-2015 are systematically analyzed to identify instances in which epistemic states are empirically measured. The results indicate that epistemic success terms are inconsistently defined, and that measurement of understanding, in particular, is rarely achieved in public understanding of science studies. We suggest that more diligent attention to measuring understanding, as opposed to mere knowledge, will increase efficacy of scientific outreach and communication efforts.

Social Media as a Catalyst for Policy Action and Social Change for Health and Well-Being: Viewpoint

This viewpoint paper argues that policy interventions can benefit from the continued use of social media analytics, which can serve as an important complement to traditional social science data collection and analysis. Efforts to improve well-being should provide an opportunity to explore these areas more deeply, and encourage the efforts of those conducting national and local data collection on health to incorporate more of these emerging data sources.

Social media remains a relatively untapped source of information to catalyze policy action and social change. However, the diversity of social media platforms and available analysis techniques provides multiple ways to offer insight for policy making and decision making. For instance, social media content can provide timely information about the impact of policy interventions. Social media location information can inform where to deploy resources or disseminate public messaging. Network analysis of social media connections can reveal underserved populations who may be disconnected from public services. Machine learning can help recognize important patterns for disease surveillance or to model population sentiment. To fully realize these potential policy uses, limitations to social media data will need to be overcome, including data reliability and validity, and potential privacy risks.

Traditional data collection may not fully capture the upstream factors and systemic relationships that influence health and well-being. Policy actions and social change efforts, such as the Robert Wood Johnson Foundation’s effort to advance a culture of health, which are intended to drive change in a network of upstream health drivers, will need to incorporate a broad range of behavioral information, such as health attitudes or physical activity levels. Applying innovative techniques to emerging data has the potential to extract insight from unstructured data or fuse disparate sources of data, such as linking health attitudes that are expressed to health behaviors or broader health and well-being outcomes.

Understanding patient satisfaction with received healthcare services: A natural language processing approach

Important information is encoded in free-text patient comments. We determine the most common topics in patient comments, design automatic topic classifiers, identify comments ' sentiment, and find new topics in negative comments. Our annotation scheme consisted of 28 topics, with positive and negative sentiment. Within those 28 topics, the seven most frequent accounted for 63% of annotations. For automated topic classification, we developed vocabulary-based and Naive Bayes ' classifiers. For sentiment analysis, another Naive Bayes ' classifier was used. Finally, we used topic modeling to search for unexpected topics within negative comments. The seven most common topics were appointment access, appointment wait, empathy, explanation, friendliness, practice environment, and overall experience. The best F-measures from our classifier were 0.52(NB), 0.57(NB), 0.36(Vocab), 0.74(NB), 0.40(NB), and 0.44(Vocab), respectively. F- scores ranged from 0.16 to 0.74. The sentiment classification F-score was 0.84. Negative comment topic modeling revealed complaints about appointment access, appointment wait, and time spent with physician.

Patient perspectives on delays in diagnosis and treatment of cancer: a qualitative analysis of free-text data

BACKGROUND

Earlier cancer diagnosis is crucial in improving cancer survival. The International Cancer Benchmarking Partnership Module 4 (ICBP4) is a quantitative survey study that explores the reasons for delays in diagnosis and treatment of breast, colorectal, lung, and ovarian cancer. To further understand the associated diagnostic processes, it is also important to explore the patient perspectives expressed in the free-text comments.

AIM

To use the free-text data provided by patients completing the ICBP4 survey to augment the understanding of patients' perspectives of their diagnostic journey.

DESIGN AND SETTING

Qualitative analysis of the free-text data collected in Wales between October 2013 and December 2014 as part of the ICBP4 survey. Newly-diagnosed patients with either breast, ovarian, colorectal, or lung cancer were identified from registry data and then invited by their GPs to participate in the survey.

METHOD

A thematic framework was used to analyse the free-text comments provided at the end of the ICBP4 survey. Of the 905 patients who returned a questionnaire, 530 included comments.

RESULTS

The free-text data provided information about patients' perspectives of the diagnostic journey. Analysis identified factors that acted as either barriers or facilitators at different stages of the diagnostic process. Some factors, such as screening, doctor-patient familiarity, and private treatment, acted as both barriers and facilitators depending on the context.

CONCLUSION

Factors identified in this study help to explain how existing models of cancer diagnosis (for example, the Pathways to Treatment Model) work in practice. It is important that clinicians are aware of how these factors may interact with individual clinical cases and either facilitate, or act as a barrier to, subsequent cancer diagnosis. Understanding and implementing this knowledge into clinical practice may result in quicker cancer diagnoses.

Aspiring to Unintended Consequences of Natural Language Processing: A Review of Recent Developments in Clinical and Consumer-Generated Text Processing

OBJECTIVES

This paper reviews work over the past two years in Natural Language Processing (NLP) applied to clinical and consumer-generated texts.

METHODS

We included any application or methodological publication that leverages text to facilitate healthcare and address the health-related needs of consumers and populations.

RESULTS

Many important developments in clinical text processing, both foundational and task-oriented, were addressed in community- wide evaluations and discussed in corresponding special issues that are referenced in this review. These focused issues and in-depth reviews of several other active research areas, such as pharmacovigilance and summarization, allowed us to discuss in greater depth disease modeling and predictive analytics using clinical texts, and text analysis in social media for healthcare quality assessment, trends towards online interventions based on rapid analysis of health-related posts, and consumer health question answering, among other issues.

CONCLUSIONS

Our analysis shows that although clinical NLP continues to advance towards practical applications and more NLP methods are used in large-scale live health information applications, more needs to be done to make NLP use in clinical applications a routine widespread reality. Progress in clinical NLP is mirrored by developments in social media text analysis: the research is moving from capturing trends to addressing individual health-related posts, thus showing potential to become a tool for precision medicine and a valuable addition to the standard healthcare quality evaluation tools.

Influence of Intensified Supervision by Health Care Inspectorates on Online Patient Ratings of Hospitals: A Multilevel Study of More Than 43,000 Online Ratings

BACKGROUND

In the Netherlands, hospitals with quality or safety issues are put under intensified supervision by the Dutch Health Care Inspectorate, which involves frequent announced and unannounced site visits and other measures. Patient rating sites are an upcoming phenomenon in health care. Patient reviews might be influenced by perceived quality including the media coverage of health care providers when the health care inspectorate imposes intensified supervision, but no data are available to show how these are related.

OBJECTIVE

The aim of this study was to investigate whether and how being under intensified supervision of the health care inspectorate influences online patient ratings of hospitals.

METHODS

We performed a longitudinal study using data from the patient rating site Zorgkaart Nederland, from January 1, 2010 to December 31, 2015. We compared data of 7 hospitals under intensified supervision with a control group of 28 hospitals. The dataset contained 43,856 ratings. We performed a multilevel logistic regression analysis to account for clustering of ratings within hospitals. Fixed effects in our analysis were hospital type, time, and the period of intensified supervision. Random effect was the hospital. The outcome variable was the dichotomized rating score.

RESULTS

The period of intensified supervision was associated with a low rating score for the hospitals compared with control group hospitals; both 1 year before intensified supervision (odds ratio, OR, 1.67, 95% CI 1.06-2.63) and 1 year after (OR 1.79, 95% CI 1.14-2.81) the differences are significant. For all periods, the odds on a low rating score for hospitals under intensified supervision are higher than for the control group hospitals, corrected for time. Time is also associated with low rating scores, with decreasing ORs over time since 2010.

CONCLUSIONS

Hospitals that are confronted with intensified supervision by the health care inspectorate have lower ratings on patient rating sites. The scores are independent of the period: before, during, or just after the intervention by the health care inspectorate. Health care inspectorates might learn from these results because they indicate that the inspectorate identifies the same hospitals as "at risk" as the patients rate as underperformers.

Evolution of research in health geographics through the International Journal of Health Geographics (2002-2015)

Health geographics is a fast-developing research area. Subjects broached in scientific literature are most varied, ranging from vectorial diseases to access to healthcare, with a recent revival of themes such as the implication of health in the Smart City, or a predominantly individual-centered approach. Far beyond standard meta-analyses, the present study deliberately adopts the standpoint of questioning space in its foundations, through various authors of the International Journal of Health Geographics, a highly influential journal in that field. The idea is to find space as the common denominator in this specialized literature, as well as its relation to spatial analysis, without for all that trying to tend towards exhaustive approaches. 660 articles have being published in the journal since launch, but 359 articles were selected based on the presence of the word “Space” in either the title, or the abstract or the text over 13 years of the journal’s existence. From that database, a lexical analysis (tag cloud) reveals the perception of space in literature, and shows how approaches are evolving, thus underlining that the scope of health geographics is far from narrowing.