@B_CP_Antunes great to see you following #pos2015 & soon hopefully you'll be along with your work on Portuguese IPOS

— Irene Higginson (@ij_higginson) February 7, 2015

@superspesh @DrFlissMurtagh @ClareEllisSmith Was a great presentation. Really worthwhile project #POS2015

— Becky Baines (@Becbaines) February 6, 2015

#POS2015 presentations will be available on YouTube channel https://t.co/TYsb36033q & materials on POS website, http://t.co/OKX0pxqJvf #hpm

— CSI Palliative Care (@CSI_KCL) February 6, 2015

#POS2015 closes with @DrFlissMurtagh, thanks to everyone who presented & attended we hope you found the day(s) interesting & enjoyable. #hpm

— CSI Palliative Care (@CSI_KCL) February 6, 2015

#POS2015 Clinical message: use POS to drive improvements in clinical care. Research message: know what your research question is

— Becky Baines (@Becbaines) February 6, 2015

#POS2015 Take home messages: team have really enjoyed engaging and hearing views from all.

— Becky Baines (@Becbaines) February 6, 2015

#POS2015 Health economic analyses: hard to measure benefits and summarise in economic terms.

— Becky Baines (@Becbaines) February 6, 2015

#POS2015 Can join in with other studies running round the country. Eg. IPOS validation study.

— Becky Baines (@Becbaines) February 6, 2015

#POS2015 Using POS for first time in research: consider capacity, try to not rush pts, try not to second guess pt data.

— Becky Baines (@Becbaines) February 6, 2015

#POS2015 How can we present data to funders, trustees etc?

— Becky Baines (@Becbaines) February 6, 2015

#POS2015 Bottom line: what is the purpose of your data collection? May not need to be entered onto spreadsheet if main focus is clinical

— Becky Baines (@Becbaines) February 6, 2015

#POS2015 Can use qualitative thematic analysis to look at free text IPOS data too. But seems to be mainly used for clinical work

— Becky Baines (@Becbaines) February 6, 2015

#POS2015 Answer seems to be: pull data out of clinical data base (eg. Crosscare, Systm1), make it anonymous, then load into Excel.

— Becky Baines (@Becbaines) February 6, 2015

#POS2015 Discussion about how we could use SystmOne and its reporting system.

— Becky Baines (@Becbaines) February 6, 2015

#POS2015 SPSS is easier for statistical data handling, but Excel better for producing graphs.

— Becky Baines (@Becbaines) February 6, 2015

#POS2015 Now mentioning multivariate analysis in passing...think we're looking like frightened rabbits, so quickly moving on! :-)

— Becky Baines (@Becbaines) February 6, 2015

#POS2015 There are other ways of looking at effect sizes. Is it just significant, or can we use tests to show how much effect there is?

— Becky Baines (@Becbaines) February 6, 2015

#POS2015 Ouch. Statistics brain strain alert: some tests need to use means rather than medians.

— Becky Baines (@Becbaines) February 6, 2015

#POS2015 then look at mean scores over time, over 4 capture points. Show change over time.

— Becky Baines (@Becbaines) February 6, 2015

#POS2015 Excel really helpful for calculating mean scores for groups of patients

— Becky Baines (@Becbaines) February 6, 2015

#POS2015 How to detect changes in scores over time? Compare mean scores over eg. 4 visits.

— Becky Baines (@Becbaines) February 6, 2015

#POS2015 Next calculate frequencies. Easy to put into bar charts on excel.

— Becky Baines (@Becbaines) February 6, 2015

#POS2015 Need a measure of spread: can use standard deviation.

— Becky Baines (@Becbaines) February 6, 2015

#POS2015 Number scores are not linear. Eg Moderate is 2. Overwhelming is 4. Overwhelming is not twice as bad. For ordinal data, use median.

— Becky Baines (@Becbaines) February 6, 2015

#POS2015 What to do with data: first look at measures of average - mean and median.

— Becky Baines (@Becbaines) February 6, 2015

Final session of #POS2015 group split into analysing data & referencing structure, using for first time in research & health economics #hpm

— CSI Palliative Care (@CSI_KCL) February 6, 2015

#POS2015 Can set up pt scores on Spreadsheet, then use to make graphs. Easier to read quickly.

— Becky Baines (@Becbaines) February 6, 2015

#POS2015 Total score: if it goes over 20, can be said to have "moderate needs"

— Becky Baines (@Becbaines) February 6, 2015

#POS2015 Look at one item to start with. Start small. Could also use 'sub scale' scores eg. Symptoms, psychological scores, practical issues

— Becky Baines (@Becbaines) February 6, 2015

#POS2015 Total score: if it goes over 20, can be said to have "moderate needs"

— Becky Baines (@Becbaines) February 6, 2015

#POS2015 Look at one item to start with. Start small. Could also use 'sub scale' scores eg. Symptoms, psychological scores, practical issues

— Becky Baines (@Becbaines) February 6, 2015

#POS2015 Practical examples of looking at pt data over time.

— Becky Baines (@Becbaines) February 6, 2015

#POS2015 Next up: Analysing POS data.

— Becky Baines (@Becbaines) February 6, 2015

#POS2015 @Becbaines @CSI_KCL re POS in Africa refs: http://t.co/3C8SDxp5XK & http://t.co/fbDZHqfAai

— Irene Higginson (@ij_higginson) February 6, 2015

@Becbaines @ij_higginson @DrFlissMurtagh #POS2015 Pretty complex and you need to be a a bit of an expert by the sounds of it! :-)

— Becky Baines (@Becbaines) February 6, 2015

#POS2015 For Trials, even quasi-experimental ones, need fixed time points. Think also about change over time, eg HIV stable vs CKD

— Becky Baines (@Becbaines) February 6, 2015

#POS2015 Ideally would want to assess Sx at times of deterioration in condition. But worry is that it would be burdensome. Ethical issue

— Becky Baines (@Becbaines) February 6, 2015

#POS2015 In research: how to capture at regular points? Often having time points is a practical way of solving this.

— Becky Baines (@Becbaines) February 6, 2015

#POS2015 Tool is servant of clinician. Clin need most important. But problem is, how to remember to do it? Electronic systems could prompt?

— Becky Baines (@Becbaines) February 6, 2015

#POS2015 Q from @DrFlissMurtagh: collect data at fixed time points or in response to change in clinical condition?

— Becky Baines (@Becbaines) February 6, 2015

#POS2015 72 team members in Africa looked at value of routine data. Advocacy, reporting to finders, demonstrate good, boost psychosocial etc

— Becky Baines (@Becbaines) February 6, 2015

#POS2015 Can set targets for improvement. ENCOMPASS study. Set challenge to reduce Sx score to less than 2 by time pt X

— Becky Baines (@Becbaines) February 6, 2015

#POS2015 Hospices in South Africa need outcome data to demonstrate 5 star status. Use routine data in box plots to show change over time.

— Becky Baines (@Becbaines) February 6, 2015

#POS2015 @learnhospice @Becbaines re use of longitudinal POS data to: demonstrate care, showing improvement over time during #hospice care

— Irene Higginson (@ij_higginson) February 6, 2015

#POS2015 African guidance on outcome measurement for research or clinical audit. Do you have link @CSI_KCL?

— Becky Baines (@Becbaines) February 6, 2015

palliative care research. missing data inevitable. guidance on handling in MORECare http://t.co/GJVzpKNPTI #POS2015

— josie dixon (@superspesh) February 6, 2015

#POS2015 'Missingness': one category MCAR (completely at random)...forgot!

— Becky Baines (@Becbaines) February 6, 2015

#POS2015 minimise by training and feedback and quality control of data entry.

— Becky Baines (@Becbaines) February 6, 2015

@Becbaines any thoughts on how useful trajectories r? #POS2015 @ij_higginson

— Sarah Russell (@learnhospice) February 6, 2015

#POS2015 Missing data: too ill to answer? Didn't want to answer? Pt died? Lost?

— Becky Baines (@Becbaines) February 6, 2015

@ij_higginson: #POS2015 @Becbaines day care referred to by Richard Harding POS scores in #palliative day care http://t.co/QiVrBb2oHo”TY

— Sarah Russell (@learnhospice) February 6, 2015

#POS2015 Can look at graphs of individual trajectories. Some stable, some variable, some increasing.

— Becky Baines (@Becbaines) February 6, 2015

#POS2015 @Becbaines day care referred to by Richard Harding POS scores in #palliative day care http://t.co/pTxMhLjsSA

— Irene Higginson (@ij_higginson) February 6, 2015

#POS2015 Murtagh et all 2011. Stage 5 CKD showed increased burden of symptoms as death gets nearer (by looking backwards at data)

— Becky Baines (@Becbaines) February 6, 2015

@ShazzaHudson make sure u look at @Becbaines tweets from #POS2015

— Sarah Russell (@learnhospice) February 6, 2015

@Becbaines #POS2015 u r on fire Becky! Ty 4 all your tweets

— Sarah Russell (@learnhospice) February 6, 2015

#POS2015 Bit of brain strain here now...'multilevel modelling'. Any help anyone?

— Becky Baines (@Becbaines) February 6, 2015

#POS2015 Group scores of patients can give population messages over time.

— Becky Baines (@Becbaines) February 6, 2015

#pos2015 Richard Harding when analysing POS scores can use group or individual scores. #palliative #hpm

— Irene Higginson (@ij_higginson) February 6, 2015

#POS2015 Example of looking at effectiveness of Palliative Day Care. (Anyone get the reference? I missed it)

— Becky Baines (@Becbaines) February 6, 2015

#POS2015 Can use box plots to demonstrate how median scores and IQRs change over time.

— Becky Baines (@Becbaines) February 6, 2015

#POS2015 Can be done looking it baseline score for individual items...then following over time in groups.

— Becky Baines (@Becbaines) February 6, 2015

#POS2015 Look at mean and SD? Or Median and range? Need to look at the distribution. Is it normal distribution?

— Becky Baines (@Becbaines) February 6, 2015

#POS2015 Graphs can be useful. Eg. Pain gets better, anxiety worse over time.

— Becky Baines (@Becbaines) February 6, 2015

#POS2015 using individual pt data: 3 data points give trends in the individual item scores. What does total score mean in practice?

— Becky Baines (@Becbaines) February 6, 2015

#POS2015 What research questions can we answer using routinely collected data?

— Becky Baines (@Becbaines) February 6, 2015

#POS2015 Next up: Dr Richard Harding - using routinely collected data

— Becky Baines (@Becbaines) February 6, 2015

#POS2015 Can we use qualitative research to help us review the '3 main problems' data? Richer info?

— Becky Baines (@Becbaines) February 6, 2015

#POS2015 Use in Qualitative Research too. Select using POS to gain range of pts for interviews. Use in mixed methods data.

— Becky Baines (@Becbaines) February 6, 2015

#POS2015 Use POS at baseline, then after an intervention. Also need to think about how far back they can remember...eg days/weeks

— Becky Baines (@Becbaines) February 6, 2015

#POS2015 POS is validated for pt, proxy and staff. So really helpful for missing data.

— Becky Baines (@Becbaines) February 6, 2015

#POS2015 MORECare statisticians said: #hpm studies need to expect missing data. If you don't, you probably have studied wrong population

— Becky Baines (@Becbaines) February 6, 2015

#POS2015 Useful therefore to have proxy version as helps understand reasons for missing data AND impute missing values

— Becky Baines (@Becbaines) February 6, 2015

Dr Richard Harding looks at using routinely collected data and how this varies internationally #POS2015 #hpm

— CSI Palliative Care (@CSI_KCL) February 6, 2015

#POS2015 Need to know why people are missing from study...holiday of a life time in Tenerife or really ill and unable to contribute

— Becky Baines (@Becbaines) February 6, 2015

#POS2015 Pt AND proxy data useful to capture in RCT, cohort studies. Helps deal with missing data. Helps us impute.

— Becky Baines (@Becbaines) February 6, 2015

#POS2015 Cohort, case-control or cross-sectional study designs can all use POS.

— Becky Baines (@Becbaines) February 6, 2015

#POS2015 Experimental: random allocation to groups. Observational (to test hypothesis) eg. Early exposure to #hpm.

— Becky Baines (@Becbaines) February 6, 2015

#POS2015 Studies may directly collect POS data, may use POS to screen for inclusion into studies or be descriptor of issues at baseline

— Becky Baines (@Becbaines) February 6, 2015

#POS2015 can do primary or secondary research. Secondary analysis can be done using data that is already collected by other people.

— Becky Baines (@Becbaines) February 6, 2015

Using POS across different research designs, @ij_higginson discusses #POS2015 #hpm #palliative care

— CSI Palliative Care (@CSI_KCL) February 6, 2015

@becbaines thanks for tweeting #POS2015 feel like Im there!

— Debra Parker Oliver (@oliverdr) February 6, 2015

#POS2015 @ij_higginson reiterating that the study design must fit the research aims. Different designs needed for different questions

— Becky Baines (@Becbaines) February 6, 2015

#POS2015 @ij_higginson speaking now about using POS across different research design.

— Becky Baines (@Becbaines) February 6, 2015

#POS2015 OACC trying to do some implementation research on doing POS in practice. Jana Witt is research manager for OACC

— Becky Baines (@Becbaines) February 6, 2015

#POS2015 RCTs can be used in implementation trials. But expensive and tough to set up. Pragmatic trials or before and after studies useful

— Becky Baines (@Becbaines) February 6, 2015

#POS2015 So which parts of the implementation process do you want to study?

— Becky Baines (@Becbaines) February 6, 2015

#POS2015 ...inner setting (culture), individual and finally implementation process.

— Becky Baines (@Becbaines) February 6, 2015

#POS2015 PARIHS framework: evidence, context and facilitation. CFIR: intervention, outer setting (big picture, policy etc)...

— Becky Baines (@Becbaines) February 6, 2015

#POS2015 focus on team, organisation, practice change rather than symptoms, patient, health outcomes.

— Becky Baines (@Becbaines) February 6, 2015

#POS2015 Implementation science: aka knowledge transfer. Different to QI and different to research.

— Becky Baines (@Becbaines) February 6, 2015

#POS2015 Implementation science by Jana Witt. How to think about process of implementing POS.

— Becky Baines (@Becbaines) February 6, 2015