A metrologist’s perspective on psychometric methods
L R Pendrill
Research Institutes of Sweden, Metrology,
Eklandagatan 86, 41261 Göteborg (SE),
2 PMhealth June 2017
L R Pendrill
Variation in Primary Care
Indicators
Potential causes of
variation:
•
Disease prevalence
•
How doctors diagnose
•
How data coders
interpret diagnoses
•
….
Extracts from OECD 2017:
[http://www.oecd.org/health/health-systems/health-data-governance.htm]
“Strong commitment of countries to make better use of health data, to foster international cooperation
in health research and ultimately to improve health system performance and outcomes for people”
“Health data necessary to improve quality, safety and patient-centeredness of health care services
and to support scientific innovation, discovery and evaluation of new treatments and to redesign
4 PMhealth June 2017
4
Traceability => measurements can be compared
•
Under both repeatability and reproducibility
•
Even different measurement quantities
Uncertainty => declared measurement quality
•
’Fit-for-purpose’
•
Quantified risks of decision errors
Metrology – quality-assured measurement
This gives processes and products which have:
•
Interoperability
6 PMhealth June 2017
n i i n Ax
x
n
t
u
1 2 % 68 , 11
1
Delivering calibration
Delivery 1: Calibration: how much does my weight really weigh?
8 PMhealth June 2017
My
health?
n i i n Ax
x
n
t
u
1 2 % 68 , 11
1
Delivering calibration
Delivery 1: Calibration: how healthy/ill aim I?
0,8 units ± 0,2 units
Object: Health
1,0 units =>
0,8 units
My
health?
Delivering calibration
Delivery 1: Calibration: cognitive ability?
10 PMhealth June 2017
12 PMhealth June 2017
Reorienting health systems to become
more people-centred
Invest in measures so that health systems deliver what matters most
to people.
Too often:
• we only rely on measures of what health systems do, and how
much they cost,
14 PMhealth June 2017
Historically, NHS focused on measuring ‘inputs’ such as
attendances, hospital admissions, and waiting times.
•
Easy to measure,
•
but fail to capture whether patient’s care was good or
bad, or even clinically effective.
•
No substitute for measuring actual outcomes as well
as costs involved over full cycle of care for patient’s
problem.
•
Recent efforts to capture patient “experience” are
useful, but not same as outcomes
Key challenge:
•
defining outcomes that matter for each condition,
•
how to measure them.
•
Thus far: efforts bottom up and different across organizations and geography.
Pressing need to develop standardized sets of outcomes by condition:
•
to enable comparison and learning,
•
put in place infrastructure and tools needed to collect and measure them across entire
system
16 PMhealth June 2017
Recommendations of IFCC Task Force on Impact of Laboratory
Medicine on Clinical Management and Outcomes (TF-ICO) include:
• “Effective collaboration with clinicians
• Determination to accept patient outcome and patient experience
as primary measure of laboratory effectiveness.”
18 PMhealth June 2017
“Large number of tools available to measure
person-centred care, but no agreement about
which tools most worthwhile.
No ‘silver bullet’ or best measure covers all
aspects of person-centred care.
Person-centred care (PCC)
20 PMhealth June 2017
Person-centred care (PCC)
OMERACT values, include:
• Trust
• Respect
• Transparency
• Partnership
• Communication
• Diversity
• Confidentiality
22 PMhealth June 2017
1. To reach consensus on which measurement properties
should be evaluated of Health‐Related Patient‐Reported
Outcomes (HR‐PROs) and how they should be defined
2. To develop standards for how these measurement
properties should be evaluated in terms of study design
and statistical analysis
COnsensus‐based Standards for the selection of
health Measurement INstruments
Measurements in PCC
Compared with traditional measures (e.g blood pressure or queueing times) in care:
• Patient assessment ≠ Professional assessment
Person-centred care (PCC)
Patient:
• More symptoms
• Greater impact on daily living
• More subjective
24
PMhealth June 2017
W P Fisher Jr. 1999
)
( hallenge
C
Measuring Man:
- Status, function of person
- Test against specifications
Man as Measurement Instrument:
- Perception of product/service
function, comfort etc
26 PMhealth June 2017
Person
Tool
Task
Health
Task - tool
Environment*
• Body structure*
• Body function*
*Five components of health:
[International Classification of Functioning, Disability and Health (ICF)]
• activity*
• participation*
A Farbrot, S Abbas, A Nihlstrand, J Dagman, R Emardson, S Kanerva & L R Pendrill , “Defining comfort for heavily-incontinent patients assisted by Health care products in several contexts”, The Simon Foundation for Continence's Innovating for Continence Conference Series, Chicago (US), April 2013
Person – centred measures
)
( eniency
L
)
(Quality
)
( Ability
)
( ifficulty
D
Ordinal data (e.g. ’Satisfaction’) – incorrect use of statistics in
traditional analyses
Item responses:
• only ordered structure
• not numerical value in mathematical sense
Statistical methods applicable to data from rating scales differ completely from
traditional methods for quantitative variables
6
3
2
1
Ordinal data
http://www.123rf.com28 PMhealth June 2017
Rasch (1961)
Measuring People
success successP
P
1
log
)
( bility
A
)
( hallenge
C
•
Correct ordinal data treatment
•
Better resolution
30 PMhealth June 2017
underestimated
MMSE Items
Less difficult
More difficult
13. Delayed recall
11. Immediate recall
1 - 10. Orientation
14 - 20.
Metrological
references
Task difficulty, δ
Difficul
ty
Mas
s
32 PMhealth June 2017
Balance as Measurement Instrument - Sensitivity (
C
)
R = C·S
+ ”additional terms”
Stimulus (
S
): Mass of weight
Response (
R
):
Mass of weight x
Balance sensitivity
Man as Measurement Instrument - Sensitivity (
C
)
R = C·S
+ ”additional terms”
Stimulus (
S
): Task challenge
Response (
R
):
R
)
( hallenge
C
( bility
A
)
e
z
z
P
R
success
34 PMhealth June 2017 34
*
*
,
,
,
,
1
log
s
i
j
j
i
success
j
i
success
P
P
j
j
j
unit
‘logistic measurement function’
j
j
*
i
j
‘common unit’ of measure
Task difficulty, δ
Set, s, of items rather than single item
S M Humphry 2011, "The Role of the Unit in Physics and Psychometrics",
Measurement: Interdisciplinary Research and Perspectives, 9(1): p. 1-24
Difficul
36 PMhealth June 2017
NeuroMet
EMPIR 15HLT04: Innovative measurements for
improved diagnosis and management of
neurodegenerative diseases
June 2016 – June 2019
Acknowledgments
The European Metrology Programme for Innovation & Research (EMPIR, Horizon2020, Art. 185) is jointly funded by the EMPIR participating countries within EURAMET (www.euramet.org) and the European Union in this EMPIR 15 HLT04 NeuroMet project (coordinator: LGC (UK))
38 PMhealth June 2017
My
health?
Delivering calibration
Delivery 1: Calibration: cognitive ability?
0,8 units ± 0,2 units
40 PMhealth June 2017
Metrological
references
Task difficulty, δ
Difficul
ty
Mas
s
“Name all three unrelated objects”
““Earlier I told you the names of
three things. Can you tell me
what those were?””
“Repeat the phrase:
‘No ifs, ands, or buts.’”
U nce rt a in Pos s ib le Proba bl e
42 PMhealth June 2017
13. Delayed recall
Less difficult
More difficult
Less able
More able
11. Immediate recall
Room for improvement
sample range
Op
tim
u
m
p
o
int
Scale range
PLOS ONE | DOI:10.1371/journal.pone.0162889 October 14, 2016
Under-estimate
44 PMhealth June 2017
• PMhealth – Swedish national workshop in psychometrics in health sciences
2017 (Kristianstad, June)
• EMPIR mini-symposium, Innovative measurements for improved diagnosis
and management of neurodegenerative diseases, (RI:SE, Göteborg, July)
• ENBIS 2017 (Naples, Sept.)
Quality-assured measurement of
perception
Measurement of
innovation
EMRP NEW04 uncertainty
Comfort Assessment
for heavy incontinent
46 PMhealth June 2017