Posts Tagged 'PubMed'

If you had to cancel a health database today? Evidence-based decision-making vs. Hobbesian elbowing

Trying to decide on a database cancellation can be fraught with uncertainty.  Evidence-based criteria are important but often shoved out of the way by other considerations. The culling process is a little more sophisticated than resorting to eeny, meeny, miny, moe. Basing decisions on research is laudable, but in the end it’s dollars and a Hobbesian political deftness that count for more. It’s not a case of evidence-be-damned, but rather the Ellenbogengesellschaft – the sharp-elbowed social reality – of collection development.

Consider the following databases: AMED (Allied and Complementary Medicine Database), CINAHL, EMBASE, and Global Health. If you had to cancel one of them, which would it be? The latter might get the boot from some of us, but I’m guessing most health libraries would probably push AMED overboard first. We would justify this to ourselves with the comforting assumption that the combination of the other databases provides good enough coverage of complementary and alternative medicine. CINAHL, for example, is well known for its lavish attention to allied health; PubMed offers its petal-strewn Complementary Medicine subset to smooth the way for us; and so on. And isn’t it true that in the minds of many faculty and health professionals CAM is to real medicine what holy water is to healing? That, to me, is the key issue. Always present in the background of a library’s cancellation decision is one of the guiding principles of a public service, multa docet fames (hunger teaches us many things). Regardless of what the literature tells us, this question cannot be avoided: to which database’s disappearance would the most influential library patrons object least?

Some justly claim their decisions to be reasoned and evidence-based. Others, if pressed, might have to admit that a decision can be biased, gratuitous, hasty, or obviously political. Journal and database cancellations are determined centrally in my world, and hence are mostly out of my hands. I like to think that the best reasons, and not just sharp elbows, are always brought forward before a subscription is dropped. Be that as it may, with respect to CAM resources, a Canadian study [1] has succeeded in undermining most of my notions about the quality and comprehensiveness of PubMed’s coverage of complementary and alternative medicine. I also see AMED in a new light.

AMED acclaimed
In a recently published article in Evidence-based complementary and alternative medicine, the authors’ objective was to compare a number of databases relevant to CAM. In all, they searched fifteen databases to identify CAM controlled clinical trials not also indexed in MEDLINE.

Their abstract sums things up adequately:

Searches were conducted in May 2006 using the revised Cochrane highly sensitive search strategy (HSSS) and the PubMed CAM Subset. Yield of CAM trials per 100 records was determined, and databases were compared over a standardized period (2005). The Acudoc2 RCT, Acubriefs, Index to Chiropractic Literature (ICL) and Hom-Inform databases had the highest concentrations of non-MEDLINE records, with more than 100 non-MEDLINE records per 500. Other productive databases had ratios between 500 and 1500 records to 100 non-MEDLINE records-these were AMED, MANTIS, PsycINFO, CINAHL, Global Health and Alt HealthWatch. Five databases were found to be unproductive: AGRICOLA, CAIRSS, Datadiwan, Herb Research Foundation and IBIDS. Acudoc2 RCT yielded 100 CAM trials in the most recent 100 records screened. Acubriefs, AMED, Hom-Inform, MANTIS, PsycINFO and CINAHL had more than 25 CAM trials per 100 records screened. Global Health, ICL and Alt HealthWatch were below 25 in yield. There were 255 non-MEDLINE trials from eight databases in 2005, with only 10% indexed in more than one database. Yield varied greatly between databases; the most productive databases from both sampling methods were Acubriefs, Acudoc2 RCT, AMED and CINAHL.

Not unexpectedly, in their conclusion the authors recommend a multi-database approach:

The very low overlap between … non-PubMed sources suggests the need for multiple database searching in addition to MEDLINE in order to comprehensively search for CAM controlled trials. The results indicate that of the six databases analyzed that are not focused on a specific therapy, CINAHL was the most productive, followed by AMED. The Acubriefs and Acudoc2 RCT databases were highly productive for acupuncture trials.

With budget restrictions looming, the University of Manitoba Libraries has just dumped AMED, which this study identifies as second only to CINAHL for controlled clinical trials coverage in complementary and alternative medicine. Who would have guessed that AMED would stand out in this subject area, considering how poorly it is rated in another recent study which has just been published in Physiotherapy [2]?

AMED def-amed
Researchers at the University of Sydney compared the comprehensiveness of indexing the reports of randomized controlled trials of physiotherapy interventions by eight bibliographic databases (AMED, CENTRAL [Cochrane], CINAHL, EMBASE, Hooked on Evidence, PEDro, PsycINFO and PubMed). The results in a nutshell? PEDro indexed 99% of the trial reports, CENTRAL indexed 98%, PubMed indexed 91%, EMBASE indexed 82%, CINAHL indexed 61%, Hooked on Evidence indexed 40%, AMED indexed 36% and PsycINFO indexed 17%.

Poor AMED comes a cropper here, outclassed as it is by a free resource like PEDro (a name almost as silly as Acubriefs, which sounds like the latest offering from Stanfield’s Ltd., Canada’s self-proclaimed, one-and-only “Underwear Company”). A library might feel quite justified in cancelling its subscription after reading about AMED’s poor coverage of physiotherapy research.

Given these contrasting evaluations of a database’s effectiveness, the question arises as to which evidence will have the most weight in the decision to cancel or retain? Our pair of studies illustrate how difficult it can be to play the database shuffle in making cancellation decisions. When budgets are tight and sacrifices must be made librarians are always ready to lend a hand, but attached to those hands should be sharp elbows.


References

1. Cogo E, Sampson M, Ajiferuke I, Manheimer E, Campbell K, Daniel R, Moher D. Searching for controlled trials of complementary and alternative medicine: a comparison of 15 databases. Evid Based Complement Alternat Med. 2009 May 25. PubMed PMID: 19468052. DOI 10.1093/ecam/nep038.

2. Moseley AM, Sherrington C, Elkins MR, Herbert RD, Maher CG. Indexing of randomised controlled trials of physiotherapy interventions: a comparison of AMED, CENTRAL, CINAHL, EMBASE, hooked on evidence, PEDro, PsycINFO and PubMed.Physiotherapy. 2009 Sep;95(3):151-6. Epub 2009 Apr 23. Review. PubMed PMID: 19635333.

Photo credit: CC licensed flickr photo by fabbio:  http://flickr.com/photos/fabiovenni/240530154/

Bugs in the redesigned PubMed? Trouble with Auto Suggest and Title searching

bug-on-screen
On Friday my colleague Tania Gottschalk and I were looking forward to the pleasurable task of introducing a group of library staff to the new, improved PubMed. These are people we know well and work with every day. So they were in a tolerant mood as we began to stumble about the new PubMed interface, trying to point out the most important features and to differentiate the substantial from the merely cosmetic changes.

What we thought would be a breeze turned out very differently. After my experience in that lab I’m ready to put on my old T-shirt from the 90s, with the slogan “Smash Forehead on Keyboard to Continue.”

Preliminary philosophical reflections
Now, I’m an easy-going type, and the continental drift that has randomly rearranged PubMed’s major menu items is no big deal. Think of all the wrists around the planet getting more exercise as cursors are sent on expeditions to rediscover vital menus and links scattered across the vast white sea of the introductory PubMed page.

One could argue that the apparent complexity and randomness – I would not go so far as to say capriciousness – of the PubMed redesign must have emerged in a way that did not depend on finely-tuned details of the system: variable parameters appear to have changed spontaneously without affecting the underlying programming code, in a sort of self-organized criticality. In short, little is where it used to be, which wouldn’t necessarily be a bad thing if one had a sense of the fundamental rationale.

Why, for example, are the tutorials separated from the FAQs? Why are Clinical Queries and Single Citation Matcher called a “resource” in Advanced Search and a “tool” on the main page? Why have the ready-to-hand tabs been replaced with the scattered remnants of their algorithmic detonation?

Philosophers have argued for centuries on what constitutes the quiddity or “whatness” of a thing as opposed to its haecceity or “thisness.” Librarians, equally inclined to make much of a muchness, cannot help but quibble about a drop-down menu here or a radio button there, a menu before or a link after, this option or that feature. Nor can we stop ourselves from applying our hermeneutic of suspicion to the new PubMed.

But it is not my purpose in this post to vent on what I dislike about the changes to my favourite database. There is also a great deal that I like. Let me return to Friday’s lab session on the PubMed redesign. Brief and informal as it was, our little demo gone demonic gave rise to two apparent bugs in PubMed’s programming, something more substantial than its desultory interface tweaks.

Auto Suggest
Our first difficulty came to light with Auto Suggest. This feature, I would suggest, is not all that it should be. In our little training demo we were working with the search terms “low carbohydrate diet” and “carbohydrate restricted diet.” In our prep for the session, we noticed that after typing the former, PubMed auto-suggested the latter with “or.” What a handy way to show how PubMed could help the novice search aided by suggestions about alternative search terms. As luck would have it, when we had PubMed up on the big screen during our lab session, this happy combination failed. No “or” appeared after Tania typed the first phrase. Not even after multiple attempts.

Oh well, it gave us our first opportunity to use that time-honoured escape clause of the harried presenter: “Moving right along …” I have tried to repeat the same thing again today, but Auto Suggest seems to be working again.

Is this a real bug in the PubMed code, or could it have had something to do with the congested entrails of our proxy server? I’m not sure about this one. But when my turn came to discuss field searching in Advanced Search, a truly nasty surprise was waiting.

Advanced Search: Search by Author, Journal, Publication Date, and more
This function of Advanced Search looks like a great way to introduce casual users of PubMed to searching within fields. I wanted to compare a simple, across-the-board two-word search to looking for each of the terms in the title field only. I often use the latter method (using the [TI] delimiter in the command line) just to reconnoitre unfamiliar terrain, as a quick-and-dirty way to find some relevant articles. I had never tried this method before.

Title search in PubMed

My chosen example was a search employing the two keywords “doctors” and “torture.” The basic, unadvanced search “doctors torture” yields 353 results and produces the following details log:

(“physicians”[MeSH Terms] OR “physicians”[All Fields] OR “doctors”[All Fields]) AND (“torture”[MeSH Terms] OR “torture”[All Fields])

No surprises there. Before using any Advanced Search procedure I tried my standard method of adding the Title field tag to each term just to make sure that PubMed would produce some results. There were 70 hits with this search (the field tags prevent any mapping):

doctors[Title] AND torture[Title]

Then (as illustrated above) I attempted the same search using Search by Author, Journal, Publication Date, and more in Advanced Search. I selected the Title field for each of the first two search boxes. I made sure “All of these (AND)” is selected. I entered my two terms in their respective locations.

Result? Fail.

Error #1: PubMed ignores the first term, takes only the second term (in this case, “torture”), and puts into a search box with “[Title]” directly appended.

Error #2: In doing so, PubMed leaves Advanced Search and displays the orphan term in the search box on its main search page.

Error #3: Having reverted pointlessly to the main page, PubMed fails even to perform the search for the one term it has placed in the search box there, leaving the user marooned.

Error #4: For no obvious reason except perhaps Schadenfreude, PubMed adds a peculiar message below the search box: “If you are trying to search, please enter a term.”

Title search result with message

A true bug?
Since Friday’s embarrassment, I have been able to reproduce this error consistently, using a variety of terms: “men” and “women,” “blood” and “pressure,” “aerobic” and “non-aerobic.”

I have employed various methods to ensure that other factors are not interfering, such as clearing the entire history, clearing any previous search terms, and even clearing my browser cache and restarting PubMed. Nothing will make this feature work with the Title field.

Here is another annoyance. If have performed a previous search and have not cleared everything out of the search box, searching Search by Author, etc. as above adds that orphaned second term to my existing search statement – without warning.

After my embarassing experience with advanced field searching, I felt like the guy in Hell who doesn’t know where to tell someone to go.

Other field searches work
As I investigated further into field searching in Advanced Search, I did not encounter the same errors ANDing together two Author names (e.g., Smith & Jones), or two terms as Text Word. Even as implausible a search as entering “English” and “French” into the Language fields causes the database to cough up 11,855 bilingual articles.

As far as I can determine, this PubMed bug is confined to the Title search.

Needless to say, there was much tittering around the lab as my colleague and I “moved right along” from that bit of pother. I’m curious to know if the PubMed people are aware of this bug, and if so, whether that is one of the causes of the seeming delay in transitioning to the redesigned interface. It will be interesting to see if others have endured the same sort of PubMed redesign tribulations.

Photo credit: cc licensed flickr photo by TaranRampersad: (flickr.com/photos/knowprose/101872870/)

Hard-to-reach, hard to research

manhole-ladder
From a public health perspective, who are the hard-to-reach and how can we find research articles about them?

I was asked this question by a manager in the Winnipeg Regional Health Authority. He was looking for research specifically on immunization programs for the hard-to-reach as part of the WRHA’s ongoing preparation for the expected H1N1 epidemic. In order to assist my client, how was I to construct a search strategy in PubMed and other databases that would gather disparate materials together without a comprehensive catch-all subject heading?

Finding appropriate literature is just one of many challenges associated with conducting research on hard-to-reach populations. In the first place, how do we identify and sample certain groups of individuals for health research? Undoubtedly some populations are particularly vulnerable and difficult to contact. Other populations may be defined by characteristics such as ethnicity or sexual preference that are not recorded in routinely available data sources. Yet while the need for research on the hard-to-reach is pressing, a comprehensive definition of this population is lacking, and hence it is difficult to get one’s bearings.

Before creating my search strategy, I had to do some research of my own in order to get a handle on exactly the groups that I would be investigating.

Defining the hard-to-reach

What exactly is meant by ‘hard-to-reach’ is a matter of some debate. The term is inconsistently applied. It will sometimes be used to refer to minority groups, such as immigrants, LGBT people, or the homeless; it can be used to refer to ‘hidden populations’, groups of people who do not wish to be found or contacted, such as illegal drug users or gang members; at other times it may refer to broader segments of the population, such as the elderly, or young people, or people with disabilities. In the service context, hard-to-reach often refers to the underserved, certain minority groups, those slipping through the social safety net, and those who are deemed to be ‘service resistant’.

Yet another term used in this context is ‘hidden populations’, meaning those who are hidden from the point of view of research sampling. Hidden populations may also actively seek to conceal their group identity, as for example in the case of sniffers, injection drug users, LGBT people who are in the closet, sexually active teens, etc.

The hard-to-reach are also called the ‘seldom heard’. The use of this term indicates that these are people who do not have a collective voice and are often under-represented in consultation and involvement activities about developing services. ‘Hard-to-reach’ suggests that there is something that prevents their engagement with services. ‘Seldom heard’ emphasizes the responsibility of agencies to reach out to excluded people, ensuring that they have access to social care services and that their voices can be heard.

One of the chief difficulties in defining the hard-to-reach is the unintentional imputation of a homogeneity among distinct groups that does not necessarily exist. Or it may imply that the problem is one within the group itself and not within the approach. Attempts at categorization can have a stigmatizing effect. Hard-to-reach audiences have, with varying degrees of prejudice, been called obstinate, recalcitrant, chronically uninformed, disadvantaged, have-not, illiterate, dysfunctional, and information poor.

Associative and Nonassociative

In addition to these various ways to categorize the hard-to-reach, we can distinguish between ‘associative’ hard-to-reach populations, such as people at risk of AIDS, and ‘nonassociative’ hard-to-reach populations: those whose members do not normally have contact with other members.

Nonassociative populations share two primary characteristics. The first is demographic. There is no effective centralized information about them, and a large proportion of their members do not know each other. The second is that their members share characteristics or attributes that make it important for health and human services to have information about them to inform service planning, policy, and delivery. In addition to these features, they are often low-frequency populations, and they might be subject to stigma of various kinds.

There has been a large amount of research on associative hard-to-reach populations, those whose members are socially networked with each other and form a community (with literally thousands of studies being done on populations at risk of HIV and AIDS, such as injection drug users) – but there have been very few rigorous studies of nonassociative populations, particularly those that are less in the public eye, such as shut-ins.

The central focus of my search strategy was to gather together information precisely on these nonassociative populations. I was looking for groups defined by individual attributes (such as health or social status) where there is often no overriding reason for within-population socializing and where a substantial proportion of population members do not have strong social links with other members and, indeed, might even resist such contact. These hard-to-reach groups must be taken into account in immunization planning. No effective H1N1 prevention strategy can exclude them.

After much effort I decided to include the following in my search strategy:

  1. The homeless, the marginally housed, street people, and sex trade workers
  2. Shelter residents (including women and youth)
  3. Inmates in the correctional system (the incarcerated, parolees, the recently released, and those in half-way homes)
  4. Persons with serious and persistent mental health issues, including dementia or addiction
  5. Housebound persons (cystic fibrosis, arthritis), shut-ins, and the disabled
  6. The linguistically isolated (people with communication impairments, recent immigrants who are not fluent in English or French)
  7. Selected recipients of Family Services and Housing (employment income assistance, government housing, children in care)
  8. Miscellaneous nonassociative groups (transients, the uninsured, the socially isolated)

PubMed Search Strategy

Here, finally, is the strategy I employed for my PubMed search:

(“Immunization”[MAJR] OR vaccinat*[TI] OR immuniz*[TI] OR immunis*[TI] OR “Immunization Programs”[MAJR] OR “Immunization Schedule”[MAJR] OR “Influenza, Human/prevention and control”[MAJR] OR “unvaccinated population”)

AND

(hard-to-reach OR “seldom heard” OR “hidden population” OR “hidden populations” OR homeless OR homelessness OR  “Homeless Persons”[MAJR] OR “Transients and Migrants”[MAJR] OR “Housing”[MAJR] OR “Prostitution”[MAJR] OR “sex trade workers” OR migrant OR vagrants OR “street worker” OR “street workers” OR “street people” OR “street youth” OR “street kids” OR “street children” OR “street involved” OR “unstable housing” OR shelters OR “shelter residents” OR (marginally[TIAB] AND housed[TIAB]) OR under-housed OR “marginalized population” OR “marginalized populations”

OR “Emigrants and Immigrants”[MAJR] OR “Refugees”[MAJR] OR “recent immigrants” OR “recent immigrant” OR “undocumented immigrant” OR “undocumented immigrants” OR “illegal immigrant” OR “illegal immigrants” OR emigres

OR “Vulnerable Populations”[MAJR] OR “Poverty”[MAJR] OR “Poverty Areas”[MH] OR “Social Class”[MH] OR “Socioeconomic Factors”[MH] OR “Urban Population”[MH] OR welfare OR underserved OR “underserved areas” OR “high-risk inner-city” OR socially-at-risk OR “at-risk population” OR “at-risk populations” OR slum OR slums OR ghetto OR ghettos OR favela OR favelas OR “low-socio-economic” OR disadvantaged OR low-income OR impoverished OR uninsured OR underinsured

OR “Prisoners”[MAJR] OR “Prisons”[MAJR] OR incarcerated OR incarceration OR “corrections facility” OR “correctional facilities” OR  “correctional population” OR prisoners OR probation OR probationers OR probationary OR parole OR parolees OR inmates OR “half-way house” OR “half-way houses”

OR “Drug Users”[MAJR] OR “Substance Abuse, Intravenous/psychology”[MAJR] OR “drug users” OR addicts OR addicted OR “drug addicts” OR “substance abuser” OR “substance abusers”

OR deaf[TIAB] OR “Hearing Impaired Persons”[MAJR] OR “Social Isolation”[MAJR] OR “low literacy” OR “language isolated” OR (linguistically[TIAB] AND isolated[TIAB]) OR homebound OR shut-in OR shut-ins OR “disabled persons”[MH]

OR “Mentally Ill Persons”[MAJR] OR “mentally ill”)

References

1. Brackertz N. Who is hard to reach and why? ISR working paper 2007. SISRQ/EL 06.07Institute for Social Research (Australia) [online]. Available from: http://www.sisr.net/publications/0701brackertz.pdf

2. Southern DA, Lewis S, Maxwell CJ, Dunn JR, Noseworthy TW, Corbett G, Thomas K, Ghali WA. Sampling ‘hard-to-reach’ populations in health research: yield from a study targeting Americans living in Canada. BMC Med Res Methodol. 2008 Aug 18;8:57. PubMed PMID: 18710574

3. Stewart M, Makwarimba E, Barnfather A, Letourneau N, Neufeld A. Researching reducing health disparities: mixed-methods approaches. Soc Sci Med. 2008 Mar;66(6):1406-17. Epub 2008 Jan 14. PubMed PMID: 18248867.

4. Thompson S, Phillips D. Reaching and engaging hard-to-reach populations with a high proportion of nonassociative members. Qual Health Res. 2007 Nov;17(9):1292-303. PubMed PMID: 17968045.

Improving PubMed: parerga and paralipomena

pubmedskeleton
A few weeks ago I posted about ways to improve PubMed in terms of both functionality and interface. At our “table talk” about this issue at the Canadian Health Libraries Association conference there were many wish list candidates and just as many pet peeves – more than could be stuffed into a single blog post. Some of the more incidental items not included in the top five – we can dignify them by calling them the parerga and paralipomena – are still of interest. I’ve dusted off the dog days hair from a nice selection. It should make for some undemanding beach reading in this season of sunshine, the Zen of tanning, and the sound of one synapse snapping.

For serious PubMed acolytes, satori has been an ever-receding prospect. In May we were told that change is coming to the PubMed interface; and the wait continues. The suspense is killing us, although just today I hear from the Krafty Librarian that enlightenment will come in September. Any moment now the new academic year will be sharking round the corner and sloping in fast. Soon the library will be filled with the slapping of flip-flops, the rattle of back pack straps, and the incessant digital smoke-signalling of smart phones as our students return to their colleges and coffee shop queues like Deleuzian nomads. (It must be my age, but teaching PubMed these days feels rather like trying to give a workshop on needlepoint at a skateboarding convention.)

In the midst of these developments a group of us are tasked with rewriting the library’s PubMed handout – our draggletailed effort to put something in the hand that may have sailed over the head. We’re concerned about our editorial efforts being overtaken by those promised developments this fall, resulting in useless screenshots and instructions rendered incomprehensible in the wink of an eye. What’s more, some of us are no longer convinced that printed handouts have much significance for students steeped in the culture of me, Wii and PS3. Still, we know that some will actually read the thing and find it a useful learning tool. So we persevere. We’re also rejigging seminar presentations to minimize the yawn quotient and to postpone for as long as possible the lurch of induced slumber. Essential to this project is avoidance of the irritating monotone and cruelly prolonged torpor that mar an otherwise mediocre training session.

Naturally such preoccupations lead to the kind of critical observations that come with too intimate an acquaintance with PubMed. Librarians are its most discriminating users, and so our opinions should count for something. None of the following suggestions will be cause for much flapping in the pigeon loft. But if our ultimate goal is to improve health care for all, then it is always worthwhile to expect more of our indispensable PubMed and to dream about how its beneficial influence could be extended. From what I hear so far, September’s promised changes will not be terribly exciting.

Ah well. Here, at any rate, are the parerga and paralipomena:

Foreign-language articles, or, What’s wrong with diacritics?

We’re not dummies. Why does PubMed insist on holding our hand? Let’s see the actual title of the article in the citation – in the original language. Most of us are using sophisticated computers with heaps of memory and advanced graphics. I’m not afraid of Chinese characters or Turkish accents. Being forced to browse non-English titles in translation and coddled in their chaste square brackets is as ersatz an affair as watching a dubbed film. Why can’t PubMed citations be allowed to be themselves, umlauts and all?

Make field tags work better with parentheses

A field tag should only have to be used once with a parenthetical expression. PubMed should accept a search expression of this type:

(aboriginal* OR inuit OR metis OR “first nations” OR “native people”) [TI]

It’s much less appealing to have to type all this:

aboriginal*[TI] OR inuit[TI] OR metis[TI] OR “first nations”[TI] OR “native people”[TI]

Colour it up

PubMed should improve the use of colour to highlight items added to the Clipboard as well as (optionally) items added to a Collection. The little green number – who invented that? – is too indistinct. Much more could be done with colour. Allow optional use of colour highlighting for records by publication type (RCTs, reviews, guidelines), by selected language(s), and other limits such as population, gender, country of publication, etc.

Citation display and sorting

Let’s have sorting of results in various ways: by author, date of publication, publication type, etc. Default display order for references should be an optional setting in My NCBI. Increase the default number of citations to display from the ridiculously low 20 to at least 50 or preferably 500.

Keyboard shortcuts

A useful option in PubMed would be keyboard shortcuts for selecting articles and for other functions (e.g., select tabs, change display, send to, save to clipboard, add to a collection). Better yet: make them customizable.

Clicking on tiny, fussy little check boxes is a time-wasting, synovium-damaging exercise.  (Remember the old DOS MEDLINE days, when you could highlight a record and select it by hitting the spacebar?) Currently, adding items to the Clipboard requires a two-step process (select item and then select Send To Clipboard from a pull-down menu). This process is sadly inefficient. It would be so much more convenient if items could be added to the Clipboard with a keystroke.

Email updates in My NCBI

This can’t be too hard a programming task. Let’s allow more than one email address for email updates. This can be useful for several reasons.

When an email update is requested from a saved search, PubMed currently defaults to the email address of the person whose My NCBI account created the search. This cannot be edited. It would be vastly more efficient if PubMed provided an editing option so that email updates could be sent to one or more email addresses.

The default for the number of records sent in any one email update is much too low. It should either not be set at all or should be set higher than five. This unhelpful initial setting is easily missed, and for searches that routinely retrieve more than five records, vital citations can be passed over. Yes, the emailed updates do include a link to view all results, but I would wager that many if not most end users don’t see it or don’t bother with it. This limit is an unnecessary impediment.

My Medline?

We absolutely have to rename “My NCBI” to something – anything – else. For cynics, I suggest Never Can Be Inspiring. The positive thinkers among us may prefer Nothing Can Beat It. There could be a contest to suggest a name. On second thought, that might result in a moniker similar in cutesiness to Loansome Doc.

I ask you: what’s wrong with plain old My PubMed? And for the truly adventurous, I would dare to suggest that we drop the name PubMed altogether and go back to calling the whole thing Medline.

Five ways to improve PubMed

anatomy-mannequinWhen I think about the current state of PubMed, I am reminded of the apocryphal Irishman who, when asked the road to Dublin, replied, “If I wanted to get to Dublin, I wouldn’t start from here.” We have come a long way from the old command line interface, but in the harsh digital glare of the increasingly rapid advances of web technology all around us, PubMed is beginning to look feeble and a bit tatty.

But it isn’t just that PubMed’s overall appearance could use an overhaul. Many of us frequent and not-so-frequent users are increasingly dissatisfied with the usability of the interface, the presentation of the data, and the seemingly haphazard pace of development. In his keynote at the Canadian Health Libraries Association conference in Winnipeg, David Rothman reminded us once again of the opinionated Harvard PhD student, Anna Kushnir, whose rant about PubMed grabbed the attention of librarians. Although her criticisms were hasty and ill-informed, we know they reflect the frustrations of many users of the database. We shouldn’t dismiss them with knowing smiles. Kushnir’s diatribe in a prominent publication was a clear signal that change is needed. PubMed cannot remain in statu quo.

On the last day of CHLA 2009, we arranged ourselves into groups and held “Table Talks” about various issues confronting health libraries. I sat at one of two tables given the less than onerous task of coming up with ideas for improving PubMed. Our discussion was lively and the time passed quickly as we nearly fell over ourselves in our eagerness to get to Dublin.

From the many recommendations that were made, these are at the top of my wishlist:

1. Direct Export to Citation Managers

Importing references from PubMed into citation managers is a cumbersome process (display results in MEDLINE format, save results to text file, import text file into citation manager). For institutions using RefWorks, RefGrab-It helps. But wouldn’t it be nice if PubMed offered its own single-step direct citation export à la Scopus?  Even Google Scholar can do this trick. It might also get some people to actually use PubMed rather than searching it from within EndNote or Reference Manager.

2. Improve the MeSH database

Where to start. The MeSH database is stiff and laboured, with occasional outbreaks of tumid extravagance. My group all agreed that we need clearer, more intuitive visual displays of the thesaurus and subheadings. The creation of a search statement using MeSH headings needs a complete rethink. The ‘Add To’ feature for inserting MeSH terms to a search box is kludgy. Parentheses can end up skewed when AND and OR statements are added to this box. Even searching for MeSH headings is difficult and unpredictable. But worse, no one really understands it.  When I teach MeSH, my students glaze over as if I were lecturing on 12-tone music. The way PubMed presents MeSH is fussy and needlessly complex. We need a MeSH mashup.

3. Eliminate need for capitalization of Boolean operators

Does this require further elaboration? AND, OR and NOT are ridiculous holdovers from the days of vacuum tubes and punchcards. They should die now.

4. Add adjacency searching and real string searching

Adjacency searching, although chiefly of interest to librarians and hackers, would crank up the power and precision of our search strategies. And let’s allow true quoted phrase searching. Right now a phrase in double quotation marks will only be found if that phrase appears in PubMed’s phrase index. Searching for an unindexed phrase, e.g., “enhanced interrogation,”  “assessment and management,” is out of the question.

5. Simplify the creation of permanent links to PubMed records

This is my own little bugbear. Couldn’t there be a simpler way to create a permanent link to a PubMed record? How about a neat little button which, when clicked, will copy the permanent link for one or more records to the operating system clipboard for easy insertion into a document or web page.

PubMed custom filters: where the bee sucks

Work without Hope draws nectar in a sieve,
And Hope without an object cannot live.
~  Samuel Taylor Coleridge

bee-sucksI received a prompt comment from the programmer responsible for the new custom filter feature in My NCBI. I was assured that any bugs are now being corrected. Reason for hope.

I was also told that my health literacy filter was broken. It had somehow accumulated a few of those annoying smart quotes (subject for another rant at another time). I carefully retyped all the quotation marks. But the filter still doesn’t work properly. Although it tells me how many records I will find, it won’t apply the filter when I click the tab. I have to copy the code into the search box myself.

However, I did have luck with other attempts at custom filtering. I am frequently asked for Canadian content from PubMed. So I created a simple search and saved it in Custom Filters. Here is the code:

“canada”[MeSH Terms] OR canad*[All Fields]

Sidebar. You’d think that “canad*[All Fields]” would be sufficient, but that search produces fewer results than when ORing the two terms above.

Having added the new filter to my tabs, I did a general PubMed search on “diabetes.” I clicked my Canada tab and it worked. PubMed instantly created a subset, just as if I had ANDed the code to “diabetes” in the search box. From 330,831 records I was able to limit to 7,470 records with some mention of Canada. I love it, but not unconditionally.

What a pity that, as reported in the NLM Technical Bulletin of May 29, custom filters do not have the “tack” on the filter name tab – which means you can’t further refine your results by, say, clicking on the “Published in the last 5 years” tab.

It occurred to me that the Custom Filters feature might be having trouble with very long strings of code. I experimented with the Cochrane filter for controlled trials (see Robinson et al. Int J Epidemiol. 2002 Feb;31(1):150-3.) True nectar to the professional:

(randomized controlled trial[pt] OR controlled clinical trial[pt] OR randomized controlled trials[mh] OR random allocation[mh] OR double-blind method[mh] OR single-blind method[mh] OR clinical trial[pt] OR clinical trials[mh] OR (“clinical trial”[tw]) OR ((singl*[tw] OR doubl*[tw] OR trebl*[tw] OR tripl*[tw]) AND (mask*[tw] OR blind*[tw])) OR (“latin square”[tw]) OR placebos[mh] OR placebo*[tw] OR random*[tw] OR research design[mh:noexp] OR comparative study[mh] OR evaluation studies[mh] OR follow-up studies[mh] OR prospective studies[mh] OR cross-over studies[mh] OR control*[tw] OR prospectiv*[tw] OR volunteer*[tw]) NOT (animal[mh] NOT human[mh])

Plug this un-in-one-breath-utterable cryptogram into Custom Filters and it works perfectly. WTMI. Whom the Gods would destroy, etc.

I must apologize for the flippant remarks I made in my previous post. It must have been the nicotine. I hope that the snag with the NLM health literacy filter can be fixed. I intend to use Custom Filters frequently and uncomplainingly when all the kinks are worked out. Can we then look forward to the addition of the kitschy tack icon to the custom filter tab? That would be nectar indeed.

Where the bee sucks, there suck I.

Custom filters in PubMed. Fail.

“Filter, flavor, flip-top box!” For a moment I was excited to hear about PubMed’s new Custom Filters feature – until I actually tried using it. Yes, it’s nice to be able to store my canned search strategies in My NCBI, but let me tell you about the difference between niceness and usefulness.

pubmed-custom-filter
The good news is that I can now store search filters in My NCBI, from the very simple to the impossibly convoluted. I can even add a filter to the PubMed window as a tab that shows the number of citations retrieved by the filter search. The bad news, however, is very bad. The tab doesn’t actually work. It displays the number of results I would get if I applied the filter. But will it apply the filter? Clicking on the tab has no effect, at least in the few despairing attempts I made yesterday.

filterflavorHere is where I’m in danger of starting to sound like Sylvia Plath (Viciousness in the kitchen! / The potatoes hiss.) To apply a saved filter to a PubMed search I was in the middle of, I had to interrupt the proceedings, go into My NCBI, find my filter, click Edit to display the saved search strategy, highlight it, and copy and paste the whole thing back into the PubMed search box. Is this progress? Why go to the trouble of creating a custom filter tab just to have it sit there and show a number?

It is puzzling why Custom Filters would be called an “enhancement.” This feature amounts to little more than a method of storing search strategies – and a hamstrung one at that. It’s as if whoever programmed this must have gone to the cottage before the coding was complete. Hey you there, sunning on the dock. Please butt out, lose the swimming goggles, pick up your laptop, and give us a filter tab that works.

By the way, the first filter I saved is NLM’s own health literacy search filter.

In case you’re planning to use it, there is an error in the code and the web page has not been updated. The former MeSH heading “Prescriptions, Drug” has been replaced by “Drug Prescriptions.” (Actually, the better term to use is the broader MeSH term “Prescriptions.”)

“health literacy” OR
“health literate” OR
“medical literacy” OR
(health [ti] AND
literacy [ti]) OR
(functional [tw] AND
health [tw] AND
literacy [tw]) OR
((low-literate [ti] OR
low-literacy[ti] OR
literacy[ti] OR
illiteracy[ti] OR
literate[ti] OR
illiterate[ti] OR
reading [mh] OR
comprehension [mh]) AND
(health promotion [major] OR
health education [major] OR
patient education [major] OR
communication barriers [major] OR
communication [major:noexp] OR
health knowledge, attitudes, practice [major] OR
attitude to health[major])) OR
(comprehension [major] AND
educational status [major])OR
(family [ti] AND
literacy [ti])OR
((“drug labeling”
OR Prescriptions [mh])
AND “comprehension”)
OR “adult literacy” OR
“limited literacy” OR
“patient literacy” OR
“patient understanding”[ti] AND
english [la]


Subscribe

My Tweets

My Delicious Bookmarks

PubMed Logo

Blog Stats

  • 67,152 hits