Using PROMIS-29 to predict Neck Disability Index (NDI) scores using a national sample of cervical spine surgery patients.

Background Context: Patient reported outcome measures (PROMs) are valuable tools for evaluating the success of spine surgery, with the Neck Disability Index (NDI) commonly used to assess pain-related disability. Recently, patient-reported outcomes measurement information system (PROMIS) has gained attention in its ability to measure PROs across general patient populations. However, PROMIS is not condition-specific so spine researchers are reluctant to incorporate it in place of common legacy measures.
Purpose: To compare the PROMIS-29 (v2.0) to the NDI and compute a conversion equation.
Study Design: This study retrospectively analyzes prospectively collected data from the cervical module of national spine registry, the Quality Outcomes Database (QOD).
Patient Sample: The QOD was queried for cervical spine surgery patients with PROMIS-29 and NDI scores. The cervical module of QOD includes patients undergoing primary or revision surgery for cervical degenerative spine diseases. Exclusion criteria included age under 18 years and diagnoses of infection, tumor, or trauma as the cause of cervical-related pain.
Outcome Measures: The outcome of interest for this study was a conversion equation from PROMIS-29 to NDI.
Methods: The PROMIS-29 includes seven 4-item domains each rated on a 5-point scale: Physical function, depression, anxiety, fatigue, sleep disturbance, ability to participate in social roles and activities (social roles), and pain interference plus one stand-alone pain intensity item. The NDI contains 10 pain-related questions scored from 0 (no pain) to 5 (most severe pain). Outcomes were collected prior to surgery and at 3- and 12-month post surgery. Patients were included in the current analysis if they had outcome data available at one or more time points. Multivariable mixed effects regression models predicting NDI scores from PROMIS-29 domains were conducted in a development data set and validated in a separate data set. Predicted NDI scores were plotted against NDI scores to determine how well PROMIS-29 domains predicted NDI. Conversion equations were created from the PROMIS-29 regression coefficients.
Results: 2,018 patients from 18 US hospitals were included (mean age=57 years (SD=12)) with 48% female, 87% Caucasian, and 11% had revision surgery. Strong correlations were found between NDI and pain interference (r=0.79), pain intensity (r=0.74), social roles (r=-0.71), physical function (r=-0.69), sleep disturbance (r=0.63), fatigue (r=0.63), and anxiety (r=0.54). Correlation between NDI and depression (r=0.49) was slightly weaker. The pattern of correlations was consistent across timepoints. Four conversion equations were created for NDI using (1) only pain interference, (2) only physical function, (3) pain interference and physical function, and (4) the five statistically significant domains of pain interference, physical function, social roles, sleep disturbance, and anxiety, plus the pain intensity item. Equations 1, 3, and 4 were the best predictors of NDI, predicting approximately 80% of NDI scores within 15 points in the validation data set. Equation 4 (NDI _% =18.897+0.855*[pain interference _raw ]-0.694*[physical function _raw ]+2.010*[pain intensity _raw ]-0.663*[social roles _raw ]+0.732*[sleep disturbance _raw ]+0.426*[anxiety _raw ]) predicted NDI most accurately with an R ² between the predicted and actual NDI scores of 0.72. Model 1 (R ²  = 0.62; NDI _% =-4.055+3.164*[pain interference _raw ])) and Model 3 (R ² =0.65; NDI%=17.321+2.543*[pain interference _raw ]-1.012*[physical function _raw ]) also had good accuracy.
Conclusions: Findings suggest accurate NDI scores can be derived from PROMIS-29 domains. Clinicians who want to move from NDI to PROMIS-29 can use this equation to obtain estimated NDI scores when only collecting PROMIS-29. These results support the use of PROMIS-29 in cervical surgery populations and underscore the idea that PROMIS-29 domains have the potential to replace disease-specific traditional PROMs.
(Copyright © 2020 Elsevier Inc. All rights reserved.)