Pubmed2: searching Pubmed for articles on public health and data science

1 Motivation

This article was inspired by a recent meeting to discuss public health data science. It outlines a method for rapidly searching the Pubmed database and analysing retrieved abstracts. We are interested in rapidly assessing the literature on the extent of discussion or application of data science in public health research and practice.

2 Method

We used the RISmed package which provides an R interface to the Pubmed API to extract and analyse the most recent 12,000 abstract retrieved via a very non-specific search strategy. We used cluster analysis (topic modelling) to group and classify abstracts. We also searched the abstracts for the terms data science and big data and conclude there is currently a very small literature on data science in public/ population health and there is a potential large research agenda to help us understand how we can apply emerging data management and analytic techniques in modern public health practice.

2.1 Load libraries

Firstly we’ll load the relevant R libraries.

2.2 Send search

Next, we’ll query the Pubmed API via the RISmed package. We’ll use a broad search strategy.

res1 <- EUtilsSummary("data + science, population + health", 
                      type = "esearch", 
                      db = "pubmed",
                      datetype = "pdat",
                      retmax = 12000,
                      mindate = 2005, 
                      maxdate = 2016)

The query sent to Pubmed is ((“EPJ Data Sci”[Journal] OR (“data”[All Fields] AND “science”[All Fields]) OR “data science”[All Fields]) AND (“population”[MeSH Terms] OR “population”[All Fields] OR “population groups”[MeSH Terms] OR (“population”[All Fields] AND “groups”[All Fields]) OR “population groups”[All Fields]) AND (“health”[MeSH Terms] OR “health”[All Fields])) AND 2005[PDAT] : 2016[PDAT] which retrieves 1.434410^{4} Pubmed entries. The query string shows that data science is not a current MESH heading. I have restricted the download to 12,000 entries in the interests of time and limitations on the Pubmed API.

fetch <- EUtilsGet(res1, type = "efetch", db = "pubmed")

2.3 Create data frame

abstracts <- data.frame(title = fetch@ArticleTitle,
                        abstract = fetch@AbstractText, 
                        journal = fetch@Title,
                        DOI = fetch@PMID, 
                        year = fetch@YearPubmed)
## ensure abstracts are character fields (not factors)
abstracts <- abstracts %>% mutate(abstract = as.character(abstract))
abstracts %>%
  head()

abstracts %>%
  group_by(year) %>%
  count() %>%
  filter(year > 2013) %>%
  ggplot(aes(year, n)) +
  geom_point() +
  geom_line() +
  labs(title = "Pubmed articles with search terms `data science` & `population health` \n2015-2016", hjust = 0.5,
       y = "Articles")

The total number of abstracts was 11987

2.4 Word cloud

cloud <- abstracts %>%
  unnest_tokens(word, abstract) %>%
  anti_join(stop_words) %>%
  count(word, sort = TRUE) 

Joining, by = "word"

  
cloud %>%
with(wordcloud(word, n, min.freq = 10, max.words = 1000, colors = brewer.pal(8, "Dark2")), scale = c(8,.3), per.rot = 0.4)

2.5 Create bigrams

2.5.1 Bigram wordcloud

bigrams_united %>%
  with(wordcloud(bigram, n, max.words = 1000, random.order = FALSE, colors = brewer.pal(9, "Set1"), scale = c(8, 0.3)), per.rot = 0.4)

2.5.2 Journal wordlcoud

cloud3 <- abstracts %>%
  select(journal) %>%
  group_by(journal) %>%
  count(sort = TRUE)
cloud3 %>%
  with(wordcloud(journal, n, min.freq = 10, random.order = FALSE, max.words = 80, colors = brewer.pal(9, "Set1")), rot.per = .6)

2.5.3 Extract abstracts containing the phrase ‘data science’

 g <- abstracts[grepl("data science", abstracts$abstract),]
 g1 <- g$DOI %>% list
 abstracts <- abstracts %>% 
   mutate(DOI  = as.character(DOI)) 
 
 abstracts[abstracts$DOI %in% g1[[1]],] %>%
   select(title, journal, DOI) %>%
   knitr::kable()

	title	journal	DOI
49	MO-FG-207B-01: Thorax/Lung.	Medical physics	28048692
50	MO-FG-207B-02: Breast.	Medical physics	28048031
52	MO-FG-207B-03: Brain.	Medical physics	28047352
53	MO-FG-207B-04: Respond to Therapy.	Medical physics	28046682
54	MO-FG-207B-00: State-of-the-Art in Radiomics in Radiology and Radiation Oncology.	Medical physics	28046637
257	Applying Multiple Data Collection Tools to Quantify Human Papillomavirus Vaccine Communication on Twitter.	Journal of medical Internet research	27919863
4237	Exploiting big data for critical care research.	Current opinion in critical care	26348424
4985	Spatial and temporal epidemiological analysis in the Big Data era.	Preventive veterinary medicine	26092722
5953	Strategic transformation of population studies: recommendations of the working group on epidemiology and population sciences from the National Heart, Lung, and Blood Advisory Council and Board of External Experts.	American journal of epidemiology	25743324
6433	OpenHealth Platform for Interactive Contextualization of Population Health Open Data.	AMIA ... Annual Symposium proceedings. AMIA Symposium	26958160

NA
NA

2.5.4 Extract abstracts containing the phrase ‘big data’

  
 g <- abstracts[grepl("big data", abstracts$abstract),]
 g1 <- g$DOI %>% list
 abstracts <- abstracts %>% 
   mutate(DOI  = as.character(DOI)) 
 
 abstracts[abstracts$DOI %in% g1[[1]],] %>%
   select(title, journal, DOI, year) %>%
   knitr::kable()

	title	journal	DOI	year
49	MO-FG-207B-01: Thorax/Lung.	Medical physics	28048692	2017
50	MO-FG-207B-02: Breast.	Medical physics	28048031	2017
52	MO-FG-207B-03: Brain.	Medical physics	28047352	2017
53	MO-FG-207B-04: Respond to Therapy.	Medical physics	28046682	2017
54	MO-FG-207B-00: State-of-the-Art in Radiomics in Radiology and Radiation Oncology.	Medical physics	28046637	2017
253	Crowdsourcing Precision Cerebrovascular Health: Imaging and Cloud Seeding A Million Brains Initiative™.	Frontiers in medicine	27921034	2016
658	Clinical chemistry in higher dimensions: Machine-learning and enhanced prediction from routine clinical chemistry data.	Clinical biochemistry	27452181	2016
1555	Scaling up health knowledge at European level requires sharing integrated data: an approach for collection of database specification.	ClinicoEconomics and outcomes research : CEOR	27358570	2016
1606	Social Media and Population Health Virtual Exchange for Senior Nursing Students: An International Collaboration.	Studies in health technology and informatics	27332439	2016
1896	Translation in Data Mining to Advance Personalized Medicine for Health Equity.	Intelligent information management	27195185	2016
1956	Community Vital Signs: Taking the Pulse of the Community While Caring for Patients.	Journal of the American Board of Family Medicine : JABFM	27170802	2016
3383	Integration of molecular pathology, epidemiology and social science for global precision medicine.	Expert review of molecular diagnostics	26636627	2015
3695	Routinely collected data as a strategic resource for research: priorities for methods and workforce.	Public health research & practice	26536502	2015
3951	Latest developments in allergic rhinitis in Allergy for clinicians and researchers.	Allergy	26443244	2015
4237	Exploiting big data for critical care research.	Current opinion in critical care	26348424	2015
4754	NA	Journal of the American Medical Informatics Association : JAMIA	26174867	2015
4834	Epidemiology research in rheumatology-progress and pitfalls.	Nature reviews. Rheumatology	26150125	2015
5953	Strategic transformation of population studies: recommendations of the working group on epidemiology and population sciences from the National Heart, Lung, and Blood Advisory Council and Board of External Experts.	American journal of epidemiology	25743324	2015
6212	Using networks to combine "big data" and traditional surveillance to improve influenza predictions.	Scientific reports	25634021	2015
7407	Sensor, signal, and imaging informatics: big data and smart health technologies.	Yearbook of medical informatics	25123735	2014
7408	Big Data Usage Patterns in the Health Care Domain: A Use Case Driven Approach Applied to the Assessment of Vaccination Benefits and Risks. Contribution of the IMIA Primary Healthcare Working Group.	Yearbook of medical informatics	25123718	2014
7593	Big data for population-based cancer research: the integrated cancer information and surveillance system.	North Carolina medical journal	25046092	2014
7684	Big data in health care: using analytics to identify and manage high-risk and high-cost patients.	Health affairs (Project Hope)	25006137	2014
9021	Privacy-by-Design: Understanding Data Access Models for Secondary Data.	AMIA Joint Summits on Translational Science proceedings. AMIA Joint Summits on Translational Science	24303251	2013
9159	Prevention and management of noncommunicable disease: the IOC Consensus Statement, Lausanne 2013.	Clinical journal of sport medicine : official journal of the Canadian Academy of Sport Medicine	24169298	2013
9845	Transforming epidemiology for 21st century medicine and public health.	Cancer epidemiology, biomarkers & prevention : a publication of the American Association for Cancer Research, cosponsored by the American Society of Preventive Oncology	23462917	2013

	title	journal	DOI	year
257	Applying Multiple Data Collection Tools to Quantify Human Papillomavirus Vaccine Communication on Twitter.	Journal of medical Internet research	27919863	2016
566	Prospective functional classification of all possible missense variants in PPARG.	Nature genetics	27749844	2016
658	Clinical chemistry in higher dimensions: Machine-learning and enhanced prediction from routine clinical chemistry data.	Clinical biochemistry	27452181	2016
718	Extracting PICO Sentences from Clinical Trial Reports using Supervised Distant Supervision.	Journal of machine learning research : JMLR	27746703	2016
804	Comparison of Approaches for Heart Failure Case Identification From Electronic Health Record Data.	JAMA cardiology	27706470	2016
868	Assessing methods for generalizing experimental impact estimates to target populations.	Journal of research on educational effectiveness	27668031	2016
1401	Predicting suicides after outpatient mental health visits in the Army Study to Assess Risk and Resilience in Servicemembers (Army STARRS).	Molecular psychiatry	27431294	2016
1579	Cardiac image modelling: Breadth and depth in heart disease.	Medical image analysis	27349830	2016
1732	Call for a Computer-Aided Cancer Detection and Classification Research Initiative in Oman.	Asian Pacific journal of cancer prevention : APJCP	27268600	2016
1831	The Importance of Computer Science for Public Health Training: An Opportunity and Call to Action.	JMIR public health and surveillance	27227145	2016
1913	Validating Machine Learning Algorithms for Twitter Data Against Established Measures of Suicidality.	JMIR mental health	27185366	2016
2142	Objective Assessment of Physical Activity: Classifiers for Public Health.	Medicine and science in sports and exercise	27089222	2016
2176	Do Staphylococcus epidermidis Genetic Clusters Predict Isolation Sources?	Journal of clinical microbiology	27076664	2016
3039	Automated Outcome Classification of Computed Tomography Imaging Reports for Pediatric Traumatic Brain Injury.	Academic emergency medicine : official journal of the Society for Academic Emergency Medicine	26766600	2016
3142	A land use regression model for ambient ultrafine particles in Montreal, Canada: A comparison of linear regression and a machine learning approach.	Environmental research	26720396	2016
3168	Single-cell analysis of targeted transcriptome predicts drug sensitivity of single cells within human myeloma tumors.	Leukemia	26710886	2015
3754	Combining Search, Social Media, and Traditional Data Sources to Improve Influenza Surveillance.	PLoS computational biology	26513245	2015
4432	Implications of Cardiovascular Disease Risk Assessment Using the WHO/ISH Risk Prediction Charts in Rural India.	PloS one	26287807	2015
4498	Thirty years of artificial intelligence in medicine (AIME) conferences: A review of research themes.	Artificial intelligence in medicine	26265491	2015
4660	Lung necrosis and neutrophils reflect common pathways of susceptibility to Mycobacterium tuberculosis in genetically diverse, immune-competent mice.	Disease models & mechanisms	26204894	2015
5022	RAIRS2 a new expert system for diagnosing tuberculosis with real-world tournament selection mechanism inside artificial immune recognition system.	Medical & biological engineering & computing	26081904	2015
5170	Short-term Mortality Prediction for Elderly Patients Using Medicare Claims Data.	International journal of machine learning and computing	28018571	2015
5249	Using EHRs for Heart Failure Therapy Recommendation Using Multidimensional Patient Similarity Analytics.	Studies in health technology and informatics	25991168	2015
5411	Past and current use of walking measures for children with spina bifida: a systematic review.	Archives of physical medicine and rehabilitation	25944500	2015
5511	Mapping chemical structure-activity information of HAART-drug cocktails over complex networks of AIDS epidemiology and socioeconomic data of U.S. counties.	Bio Systems	25916548	2015
7058	The genetic interacting landscape of 63 candidate genes in Major Depressive Disorder: an explorative study.	BioData mining	25279001	2014
7420	Visualization and unsupervised predictive clustering of high-dimensional multimodal neuroimaging data.	Journal of neuroscience methods	25117552	2014
8662	NeuCube: a spiking neural network architecture for mapping, learning and understanding of spatio-temporal brain data.	Neural networks : the official journal of the International Neural Network Society	24508754	2014
8865	High-throughput neuro-imaging informatics.	Frontiers in neuroinformatics	24381556	2014
9273	A Machine Learning-Based Analysis of Game Data for Attention Deficit Hyperactivity Disorder Assessment.	Games for health journal	26196929	2013
9430	e-Labs and the stock of health method for simulating health policies.	Studies in health technology and informatics	23920562	2013

2.6 Create document term matrix

abstracts %>%
  unnest_tokens(word, abstract) %>%
  anti_join(stop_words) %>%
  count(DOI, word, sort = TRUE) %>%
  cast_dtm(DOI, word, n) ->
  abstracts1

Joining, by = "word"

3 Topic modelling (cluster analysis of abstracts)

library(topicmodels)
abs_lda <- LDA(abstracts1, k = 10, control = list(seed = 1234))
abs_lda_td <- tidytext:::tidy.LDA(abs_lda)

3.1 Plot terms by topic

3.2 Classify documents

abs_lda_gamma <- tidytext:::tidy.LDA(abs_lda, matrix = "gamma")
abs_class <- abs_lda_gamma %>%
  group_by(document) %>%
  top_n(1, gamma) %>%
  ungroup() %>%
  arrange(gamma)
abs_class %>%
  sample_n(6)

abstracts %>%
  group_by(journal) %>%
  count(sort = TRUE) %>%
  filter(n >=40) ->top40
abstracts %>%
  left_join(top40) %>%
  filter(!is.na(n)) %>%
  rename(document = DOI) %>%
  left_join(abs_class) %>%
  filter(!is.na(topic)) %>%
  ggplot(aes(document, factor(topic))) +
  geom_jitter(aes(colour = factor(topic)), size = 1, width = 0.1) +
  facet_wrap(~journal) +
  theme(strip.text.x = element_text(size = 8), axis.text.x = element_blank(), axis.ticks.x = element_blank(), axis.text = element_text(size  =10)) +
  scale_color_viridis(discrete = TRUE, option = "C") +
  theme(panel.background = element_rect(fill = "aliceblue")) +
  labs(title = "Top 15 journals",
       subtitle = "Documents by topic",
       x = "Time")

NA