Do hybrid open access articles identified by oaDOI also include some delayed open access?

Overview hybrid OA

According to Piwowar et al. (2017a), articles were categorized as hybrid if they were “Free under an open license in a toll-access journal”. What licenses were used to identify hybrid open access?

wos %>%
    filter(oa_color_long == "hybrid") %>%
    group_by(license) %>%
    summarize(n = n()) %>%
    mutate(perc_prop = n / sum(n) * 100)

It seems that not for every hybrid open access articles a license was disclosed in the dataset indicated by NA values.

Now, let’s obtain a breakdown of hybrid open access by publisher:

wos %>%
    filter(oa_color_long == "hybrid") %>%
    group_by(publisher) %>%
    summarize(n = n()) %>%
    mutate(perc_prop = n / sum(n) * 100) %>%
    arrange(desc(n))

How did oaDOI determine the hybrid OA evidence?

wos %>%
    filter(oa_color_long == "hybrid") %>%
    group_by(evidence) %>%
    summarize(n = n()) %>%
    mutate(perc_prop = n / sum(n) * 100) %>%
    arrange(desc(n))

Delayed Hybrid OA

Licenses metadata in Crossref can also be used to identify delayed open access content. To check, let us write a helper function that a) retrieves Crossref metadata for a single DOI using the rcrossref package, and b) parses licensing metadata. delay-in-days must be zero in order indicate immediate free access.

licensing_check <- function(doi = NULL) {
   cr_tmp <- rcrossref::cr_works_(doi, parse = FALSE) %>%
        jsonlite::fromJSON()
        cr_tmp$message$license
        }
# example for delayed usage of http://www.elsevier.com/open-access/userlicense/1.0/ license
licensing_check("10.1016/j.dam.2008.06.028")

Now, let us apply the function to every hybrid open access article in the subset hybrid (via crossref license).

wos %>%
    filter(oa_color_long == "hybrid") %>%
    filter(evidence == "hybrid (via crossref license)") -> cr_dois
cr_dois

cr_df <- purrr::map(cr_dois$doi, purrr::safely(licensing_check))
tt <- purrr::map(cr_df, "result")
names(tt) <- cr_dois$doi
license_dates <- map_df(tt, `[`, c("URL", "delay-in-days"), .id = "doi") 
# backup
readr::write_csv(license_dates, "wos_cr_licensing_md.csv")

We need to check for open licenses used by oaDOI, which can be found in the source code of oaDOI.

license_patterns <- tolower(c("creativecommons.org/licenses/",
                      "http://koreanjpathol.org/authors/access.php",
                      "http://olabout.wiley.com/WileyCDA/Section/id-815641.html",
                      "http://pubs.acs.org/page/policy/authorchoice_ccby_termsofuse.html",
                      "http://pubs.acs.org/page/policy/authorchoice_ccbyncnd_termsofuse.html",
                      "http://pubs.acs.org/page/policy/authorchoice_termsofuse.html",
                      "http://www.elsevier.com/open-access/userlicense/1.0/",
                      "http://www.ieee.org/publications_standards/publications/rights/oapa.pdf"))

license_dates %>%
    mutate(URL = tolower(URL)) %>%
    mutate(hybrid_license = ifelse(grepl(paste(license_patterns, collapse = "|"), URL), TRUE, FALSE)) %>%
  filter(hybrid_license == TRUE) %>%
    filter(`delay-in-days` > 0) -> wos_delay

The following table shows freely available articles with delayed licenses.

wos_delay

Overview hybrid OA

cr_sample %>%
    filter(oa_color_long == "hybrid") %>%
    group_by(license) %>%
    summarize(n = n()) %>%
    mutate(perc_prop = n / sum(n) * 100)

Let’s see how hybrid OA is distributed over publishers.

cr_sample %>%
    filter(oa_color_long == "hybrid") %>%
    group_by(publisher) %>%
    summarize(n = n()) %>%
    mutate(perc_prop = n / sum(n) * 100) %>%
    arrange(desc(n))

How did oaDOI determine the hybrid open access evidence:

cr_sample %>%
    filter(oa_color_long == "hybrid") %>%
    group_by(evidence) %>%
    summarize(n = n()) %>%
    mutate(perc_prop = n / sum(n) * 100) %>%
    arrange(desc(n))

Now obtain Crossref licensing metadata for hybrid open access article determined by oaDOI:

cr_sample %>%
    filter(oa_color_long == "hybrid") %>%
    filter(evidence == "hybrid (via crossref license)") -> cr_sample_dois
cr_df <- purrr::map(cr_sample_dois$doi, purrr::safely(licensing_check))
tt <- purrr::map(cr_df, "result")
names(tt) <- cr_sample_dois$doi
license_dates <- map_df(tt, `[`, c("URL", "delay-in-days"), .id = "doi") 
# backup
readr::write_csv(license_dates, "cr_sample_cr_licensing_md.csv")

license_dates %>%
  mutate(URL = tolower(URL)) %>%
  mutate(hybrid_license = ifelse(grepl(paste(license_patterns, collapse = "|"), URL), TRUE, FALSE)) %>%
  filter(hybrid_license == TRUE) %>%
  filter(`delay-in-days` > 0) -> cr_delay
cr_delay 

Finally some validation: Do articles with more than one license can have at least one delayed and at least one immediate open license?

# some validation
cr_not_delayed <- license_dates %>%
  mutate(URL = tolower(URL)) %>%
  mutate(hybrid_license = ifelse(grepl(paste(license_patterns, collapse = "|"), URL), TRUE, FALSE)) %>%
  filter(hybrid_license == TRUE) %>%
  filter(`delay-in-days` < 1)
table(cr_delay$doi %in% cr_not_delayed$doi)

## 
## FALSE 
##  1205

Results

The following table shows that across both DOI samples around 70% of articles determined as hybrid open access via Crossref licensing information were tagged with open licenses, which came into effect a certain time after publication.

data_frame(Sample = c("Crossref", "WoS"), 
           `Articles in hybrid (via crossref license) subset` = c(n_distinct(cr_sample_dois$doi), n_distinct(cr_dois$doi)), 
           `Delayed licenses found` = c(n_distinct(cr_delay$doi), n_distinct(wos_delay$doi))) %>%
  mutate(`Proportion (in%)` = `Delayed licenses found` / `Articles in hybrid (via crossref license) subset`)