the distribution of promis domains among patients prior to
TRANSCRIPT
The Distribution of PROMIS Domains among Patients
prior toUndergoing Spine Surgery
Amy M. Cizik, PhD, MPH
September 27, 2017
Disclosures
• This work is supported by the National Institute of Arthritis and Musculoskeletal and Skin Diseases of the National Institutes of Health under Award Number R21AR068009
Motivation
Spine SCOAP and CERTAIN
• Surgeon-led collaborative using performance surveillance and benchmarking to deliver more appropriate, safer and higher quality surgical spine care across Washington State
• Engaging spine surgeons in the process of selecting evidence-based QI metrics for monitoring
CERTAIN Research Community
• N=157 spine fusion patients from academic (25%) and private practice (75%)
• Age = 62 years (23 – 87 years, Median 64)
• 58% female
• 45% retired
• 14% receive disability or worker’s compensation
• 78% reported having leg pain (radiculopathy)
PRO Measurement in Spine
• Oswestry Disability Index– 10 items– 6 responses (0 – 5)– Pain, Personal Care, Sexual
Function, Driving, Sleeping
• PROMIS Domains Recommended– Physical Function (PF)– Depression (DEP)– Sleep Disturbance (SLEEP)– Pain Interference (PAININ)
• PROMIS Domains Measured– Depression (DEP)
– Anxiety (EDANX)
– Sleep Disturbance (SLEEP)
– Pain Interference (PAININ)
– Emotional Support (FSE)
– Instrumental Support (CCC and SS)
Data Collection Modalities
141
13
00
20
40
60
80
100
120
140
160
Paper Online Phone
Spine Fusion BL Survey Completion Type
94
31
00
20
40
60
80
100
120
140
160
Paper Online Phone
30-90 Day Survey Completion Type
Specific Aim
• Determine the extent to which domains are correlated and understand the patterns of correlation among patients undergoing lumbar fusion
Data Analysis
• Descriptive statistics
• Pearson’s correlation coefficients (r) was used to assess the strength of correlation between domains.
• Principal-component analysis (PCA) was conducted to identify dimensions and to describe relationships between variables within and across PROMIS domains – Oblique rotation was used to allow for correlation
between components
– The number of components was selected by the minimum number of components required to cumulatively explain 75% of the variation observed
Anxiety and Depression
Emotional and Instrumental Support
Pain Interference and Sleep Disturbance
t-score Distributions
PROMIS DOMAINSAMPLE
MEAN
SAMPLE
SDIQR
Anxiety 55.5 9.1 [48.5, 63.1]
Depression 53.0 9.5 [41.0, 60.5]
Sleep Disturbance 52.0 3.6 [49.7, 54.5]
Instrumental Support 55.0 8.8 [49.5, 63.3]
Emotional Support 54.6 8.2 [48.9, 62.0]
Pain Interference 67.1 5.5 [63.9, 71.3]
Pain NRS Leg 5.9 2.5 [4.0, 8.0]
Pain NRS Low Back 6.5 2.1 [5.0, 8.0]
Correlation Heat Map
0.69
0.42
0.41 0.37
-0.34
-0.30
Principal Component Analysis
PC1 PC2 PC3 PC4
Anxiety 0.89 0.07 0.02 0.05
Depression 0.90 -0.02 -0.07 -0.08
Sleep Disturbance 0.00 0.01 0.01 0.98
Instrumental Support 0.11 -0.10 0.90 0.02
Emotional Support -0.29 0.12 0.76 -0.02
Pain Interference 0.16 0.58 0.22 -0.25
Pain Intensity (Leg) -0.12 0.88 -0.13 -0.01
Pain Intensity (Back) 0.26 0.71 0.09 0.07
Proportion of Variance 0.23 0.21 0.19 0.13
Cumulative Proportion 0.23 0.44 0.63 0.76
PC=principal component
Summary
• Barriers to web-based/mobile technology in private practice clinics
• Anxiety and Depression domains explain almost 25% of the variation in baseline PROMIS domain characteristics in patients about to undergo lumbar spine fusion surgery
• Pain domains explains an additional 20% of the variation of all the baseline domains measured
Next Steps• Responder Analysis• Latent class analysis to identify clusters of domains grouped
together within patient subtypes• Association between clusters and non-response using logistic
regression models accounting for repeated testing• Sensitivity analyses
– Repeat the latent class analysis on the raw percentage change of the primary outcome measure (ODI) to examine the robustness of variables that define latent classes
– Examine different degrees of change in the primary outcome beyond the 15% threshold, for example PASS
• Missing data analysis to describe and characterize enrolled participants who do not provide further outcome measures due to attrition using multiple imputation
Acknowledgements• Katie Odem-Davis, PhD
• Jesse Fann, MD, MPH
• Jeffrey G. Jarvik, MD, MPH
• Richard A. Deyo, MD, MPH
• David R. Flum, MD, MPH
• Danielle Lavallee, PhD, PharmD
• Ellie Brewer, MPH
Contact Information