29.06.2013 Views

Evaluating Patient-Based Outcome Measures - NIHR Health ...

Evaluating Patient-Based Outcome Measures - NIHR Health ...

Evaluating Patient-Based Outcome Measures - NIHR Health ...

SHOW MORE
SHOW LESS

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

that nothing relevant was lost to subsequent<br />

examination. Time restrictions curtailed the<br />

total number of articles reviewed to 209, with a<br />

decision taken one month into reading, to time<br />

order the remaining publications, giving precedence<br />

to articles published after 1990. However,<br />

all 247 articles were skim read and important<br />

articles published in the 1980s and 1970s were<br />

transcribed. The qualitative analysis involved<br />

transcribing key points and summary statements<br />

from the 209 articles into files under the following<br />

topics: general issues and concepts, selection<br />

criteria, validity, reliability, responsiveness,<br />

acceptability, feasibility, utility, comparison<br />

of instruments, numerical properties<br />

and weights.<br />

An additional benefit of this stage was the identification<br />

of an extensive list of possible search terms<br />

for the next phase of the literature search. Possible<br />

search terms were sought from the full text of<br />

articles and not just keywords.<br />

Electronic searching (step 5)<br />

The electronic literature search was used to achieve<br />

two objectives. It would validate the first phase of<br />

the literature review strategy, by cross checking how<br />

many of the 223 articles (24 of the 247 references<br />

were either books, book chapters or unpublished)<br />

previously obtained, appeared in the results of the<br />

electronic search. The inspection revealed that<br />

58% (130/223) of articles were represented in the<br />

electronic literature search. This low figure is<br />

consistent with Chalmers et al (1992), who found<br />

that MEDLINE searching only retrieved half of the<br />

relevant studies, with those missed actually contained<br />

within MEDLINE but inaccurately indexed<br />

or described by the author or the coding procedure.<br />

More importantly, the electronic search<br />

provided a substantial number of abstracts to<br />

review in order to identify any publications that<br />

provided new dimensions or additional perspectives<br />

to issues already uncovered in the first stage<br />

of the literature review.<br />

The electronic search did not initially have any date<br />

restrictions and went back as far as the databases<br />

would allow. However, the actual review of abstracts<br />

was limited to 1991–1996 in order to capture only<br />

up to date information. The electronic literature<br />

search was limited to English introducing a<br />

selective bias. An attempt was made to reduce the<br />

total number of records retrieved by only searching<br />

in title and keywords but this was found to be too<br />

narrow and risked missing many references.<br />

<strong>Health</strong> Technology Assessment 1998; Vol. 2: No. 14<br />

The electronic search was carried out in MEDLINE<br />

(1966–1996), CINAHL (Cumulative Index of<br />

Nursing and Allied <strong>Health</strong>, 1982–1996), PsychLIT<br />

(1974–1996), Sociofile (1974–1996), Econlit<br />

(1969–1996), all of which were on the University<br />

of Oxford’s electronic reference library (ERL) and<br />

accessed using Winspirs/Silverplatter software via<br />

the University network. Additionally, the EMBASE<br />

database (1990–1996) was accessed using the Bath<br />

Information & Data Services (BIDS). As the BIDS<br />

and the ERL databases are assessed via two different<br />

pathways it was not possible to perform a<br />

combined search.<br />

The electronic search strategy combined the term<br />

‘patient-based outcome*’ and its synonyms with<br />

related methodological terms to retrieve only<br />

publications that looked at methods of evaluating<br />

patient-based outcome measures. The search terms<br />

used after refinement of an original set are shown<br />

in Box 6. The terms in search one were combined<br />

using the ‘or’ connector and the same done for<br />

search two. The results of the two searches were<br />

then combined using the ‘and’ connector.<br />

BOX 6 Electronic search terms<br />

Search one: retrieval of all records using the terms<br />

<strong>Patient</strong>-based outcome *<br />

<strong>Health</strong> status Subjective health status<br />

<strong>Health</strong> status indicator * <strong>Health</strong> status assessment<br />

Quality of life Disability scale<br />

<strong>Health</strong>-related Performance status<br />

quality of life<br />

Functional status<br />

Search two: retrieval of all records using the terms<br />

Methodol *<br />

Effect size<br />

Psychometric * Sensitivity to change<br />

Validity Reproducibility<br />

Reliability Acceptability<br />

Responsiveness Utility measure *<br />

Combine search one and two with ‘and’ connector<br />

* = truncation symbol<br />

The original list of search terms was run in the<br />

MEDLINE database and refined by eliminating<br />

terms that retrieved a high number of falsepositives.<br />

This was done by reviewing a sample of<br />

the records retrieved from individual search terms<br />

and estimating the number of false hits. The<br />

exclusion of terms was then discussed and verified<br />

at project meetings. A summary of the excluded<br />

terms is shown in Table 5.<br />

67

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!