Developing Plain language Summaries for Diagnostic Test Accuracy (DTA) reviews Penny Whiting, Clare Davenport, Mariska Leeflang, Gowri Gopalakrishna, Isabel de Salis Collaboration for Leadership in Applied Collaboration for Leadership in Applied Health Research and Care West
at is a plain language g summary? MECIR Standard d Prepare a summary of the review containing all the crucial information in plain language that will be understood by the general public.
at is a plain language g summary? MECIR Standard d Prepare a summary of the review containing all the crucial information in plain language that will be understood by the general public.
llenges for DTA PLS ethodology less familiar aired accuracy measures ccuracy measures poorly understood arge degree of heterogeneity uality difficult to communicate otentially larger target audience than PLS
ps Stage 1: Focus Grou Focus group 1: Consumers (n=8) PLS 0.1 Focus group 2: Journalists (n=9) PLS 0.2 Focus group 3: Clinicians (n=2) PLS 0.2 Stage 2: User testin ng One-on-one User testing PLS 0.3 Sta age 3: Su urvey Web-based survey: Multiple rounds PLS 0.3-0.? Stage 4: PL S PLS template and guidance
rting point: Evidence
us groups: PLS version 0.1
erical results options CODE had a sensitivity of 95%, this means that it most people with dementia would ntified by the IQCODE. It had a specificity of 89% meaning that 11% of those t dementia would also have a positive result. 1 000 people tested using the IQCODE, we would expect 612 a low score (less than 3.3) of which only 13 would have a (false negative results). It is probably less helpful for ruling gnosis of dementia. We would expect 388 people to have a ODE score (above 3.3) and of these 141 would not have a (false positive results). 3 re below shows the expected results in a tical group of a 1000 patients tested for ia using the IQCODE: Option 2 Using the IQCODE for diagnosing dementia in the general hospital setting: Of every 1000 people tested using the IQCODE, an estimated 612 will have a negative result and of these 13 will actually have dementia Of the 388 people with a positive IQCODE result, 141 will be incorrectly classified as having dementia Option 4 The IQCODE can help in 'ruling out' dementia in the general hospital setting. This means that if a person has a low IQCODE score, they are unlikely to have dementia. Yet if a person has a high score, that person
us group: Topic guide at is your overall feeling? e there sections of the document that you like? Any ctions that you don t like? oking at the two PLSs side by side, how do they mpare? at are your first impressions of the 4 [results] sentation options?
us group: results Risk of Bias So to say that, We cided they had a low isk of bias, if you ve got no idea what it eans, there s no point in having it in there. I assumed that that sort of implied that the studies aren t done by drug companies or aren t funded by them in some way, but I wasn t really but I don t really know.
us group: results ve test result I think there are often ases where a positive result is actually bad news, so the word positive is questionable Accuracy measures Yeah, you could use the 95% and the 11% on their own. Without saying specificity and sensitivity at all. If I had a test t result and was 89% something, is that good or is that bad? There s no context to explain.
nges made to PLS based on s groups ade text short and concise emoved reference to how good the test is emoved section on summary of findings ntroduced downstream consequence of test arly on (new heading) and to the flow diagram estructured flow diagram to go from left to right
0.3
r testing linician Commissioner Cochrane author Clinician ician Journalist Clinician Patient t representative ti
vey round 1 argeted dissemination road disseminationi urrently 58 responses
r feedback!
Stewart NIHR: ntal Elf Retweeted Sense about Science mportant work. Please take part in the survey if you can.the Mental Elf added, about Science @senseaboutsci prove plain language summaries! Researchers would love your k: http://bit.ly/29i82st @cochranecollab #dementia"
ke the layout of the summary Strongly Agree Agree Neutral Disagree Strongly Disagree 13% 78% 9% 0% 0% ree ee ral ree ree
w should additional ormation be included? In the main text of the summary? As footnotes As hyperlinks to the main Cochrane review 19% 29% 52% xt of the su... As footnotestes o the main C...
lusion of "consequences" in the gram helps understand the results Strongly Agree Agree Neutral Disagree Strongly Disagree 48% 38% 5% 0% 10% ree ee ral ree ree
e following would improve the gram Use of colour Going from top to bottom instead of left to right Inclusion of % Inclusion of tp, fn, tn, fn Smiley face next to good consequence, sad face next to bad consequence 36% 6% 27% 9% 21% Use of colour to bottom in.... Inclusion of % n of tp, fn, tn, fn xt to good co....
e diagram is a good way of mmarising the results of the review Strongly Agree Agree Neutral Disagree Strongly Disagree 0% 0% 0% 0% 0% ree ree tral ree ree
llenges for DTA PLS ethodology less familiar aired accuracy measures ccuracy measures poorly understood arge degree of heterogeneity uality difficult to communicate otentially larger target audience than PLS
llenges for DTA PLS ethodology less familiar aired accuracy measures ccuracy measures poorly understood arge degree of heterogeneity uality difficult to communicate otentially larger target audience than PLS
ther challenges hould we try and incorporate confidence ntervals? hat to do with reviews that include multiple ndex tests?
t steps lease fill in our survey: https://sscm.onlinesurveys.ac.uk/cochrane-pls p uture rounds of survey evelopment of final version of PLS with ccompany guidance aunch at Cochrane, October 2016
ank you! nny.whiting@bristol.ac.uk itter: @CLAHRC_West bsite: clahrc-west.nihr.ac.uk