Prediction of the Clinical Course of Low-Back Trouble Using Multivariable Models

A. K. Burton, K. M. Tillotson

Research output: Contribution to journalArticle

48 Citations (Scopus)

Abstract

The inability to predict outcome in patients with low-back pain seriously impedes clinical trials and leads to inappropriate or unnecessary treatment. This prospective study investigated the value of multivariable mathematical models to predict the 1-year clinical course of 109 patients with low-back trouble (LBT). Discriminant analysis was used to determine predictive models for outcome groups at 1 month, 3 months and 1 year. The variables selected in the analyses were subsets of 29 items from a clinical interview at presentation. These included anamnesis features of the first episode as well as symptomatic details and results from clinical tests for the current spell. The derived models successfully discriminated outcome groups with estimates of sensitivity and specificity ranging from 63 to 99%. When models from one set of patients were tested for predictive accuracy by the application of them to a different set, nonrecovery and satisfactory improvement were predicted with a 76-100% success rate. The results affirmed the importance of considering combinations of interrelated variables for prediction and discrimination in LBT. This work has demonstrated that outcome can be predicted successfully by the use of mathematic models based just on presentation data. The ability to determine homogenous groups in respect to outcome is seen as an important aid to therapeutic research; further work will enable refinement of these models for general clinical use and for incorporation into computer-based interview systems.

Original languageEnglish
Pages (from-to)7-14
Number of pages8
JournalSpine
Volume16
Issue number1
DOIs
Publication statusPublished - 1 Jan 1991

Fingerprint

Interviews
Therapeutic Human Experimentation
Mathematics
Discriminant Analysis
Low Back Pain
Theoretical Models
Clinical Trials
Prospective Studies
Sensitivity and Specificity
Therapeutics

Cite this

@article{5d61ba1a95db46b38876b701eafe8018,
title = "Prediction of the Clinical Course of Low-Back Trouble Using Multivariable Models",
abstract = "The inability to predict outcome in patients with low-back pain seriously impedes clinical trials and leads to inappropriate or unnecessary treatment. This prospective study investigated the value of multivariable mathematical models to predict the 1-year clinical course of 109 patients with low-back trouble (LBT). Discriminant analysis was used to determine predictive models for outcome groups at 1 month, 3 months and 1 year. The variables selected in the analyses were subsets of 29 items from a clinical interview at presentation. These included anamnesis features of the first episode as well as symptomatic details and results from clinical tests for the current spell. The derived models successfully discriminated outcome groups with estimates of sensitivity and specificity ranging from 63 to 99{\%}. When models from one set of patients were tested for predictive accuracy by the application of them to a different set, nonrecovery and satisfactory improvement were predicted with a 76-100{\%} success rate. The results affirmed the importance of considering combinations of interrelated variables for prediction and discrimination in LBT. This work has demonstrated that outcome can be predicted successfully by the use of mathematic models based just on presentation data. The ability to determine homogenous groups in respect to outcome is seen as an important aid to therapeutic research; further work will enable refinement of these models for general clinical use and for incorporation into computer-based interview systems.",
keywords = "Lowback pain, Multivariable models, Outcome, Prediction",
author = "Burton, {A. K.} and Tillotson, {K. M.}",
year = "1991",
month = "1",
day = "1",
doi = "10.1097/00007632-199101000-00002",
language = "English",
volume = "16",
pages = "7--14",
journal = "Spine",
issn = "0362-2436",
publisher = "Lippincott Williams and Wilkins",
number = "1",

}

Prediction of the Clinical Course of Low-Back Trouble Using Multivariable Models. / Burton, A. K.; Tillotson, K. M.

In: Spine, Vol. 16, No. 1, 01.01.1991, p. 7-14.

Research output: Contribution to journalArticle

TY - JOUR

T1 - Prediction of the Clinical Course of Low-Back Trouble Using Multivariable Models

AU - Burton, A. K.

AU - Tillotson, K. M.

PY - 1991/1/1

Y1 - 1991/1/1

N2 - The inability to predict outcome in patients with low-back pain seriously impedes clinical trials and leads to inappropriate or unnecessary treatment. This prospective study investigated the value of multivariable mathematical models to predict the 1-year clinical course of 109 patients with low-back trouble (LBT). Discriminant analysis was used to determine predictive models for outcome groups at 1 month, 3 months and 1 year. The variables selected in the analyses were subsets of 29 items from a clinical interview at presentation. These included anamnesis features of the first episode as well as symptomatic details and results from clinical tests for the current spell. The derived models successfully discriminated outcome groups with estimates of sensitivity and specificity ranging from 63 to 99%. When models from one set of patients were tested for predictive accuracy by the application of them to a different set, nonrecovery and satisfactory improvement were predicted with a 76-100% success rate. The results affirmed the importance of considering combinations of interrelated variables for prediction and discrimination in LBT. This work has demonstrated that outcome can be predicted successfully by the use of mathematic models based just on presentation data. The ability to determine homogenous groups in respect to outcome is seen as an important aid to therapeutic research; further work will enable refinement of these models for general clinical use and for incorporation into computer-based interview systems.

AB - The inability to predict outcome in patients with low-back pain seriously impedes clinical trials and leads to inappropriate or unnecessary treatment. This prospective study investigated the value of multivariable mathematical models to predict the 1-year clinical course of 109 patients with low-back trouble (LBT). Discriminant analysis was used to determine predictive models for outcome groups at 1 month, 3 months and 1 year. The variables selected in the analyses were subsets of 29 items from a clinical interview at presentation. These included anamnesis features of the first episode as well as symptomatic details and results from clinical tests for the current spell. The derived models successfully discriminated outcome groups with estimates of sensitivity and specificity ranging from 63 to 99%. When models from one set of patients were tested for predictive accuracy by the application of them to a different set, nonrecovery and satisfactory improvement were predicted with a 76-100% success rate. The results affirmed the importance of considering combinations of interrelated variables for prediction and discrimination in LBT. This work has demonstrated that outcome can be predicted successfully by the use of mathematic models based just on presentation data. The ability to determine homogenous groups in respect to outcome is seen as an important aid to therapeutic research; further work will enable refinement of these models for general clinical use and for incorporation into computer-based interview systems.

KW - Lowback pain

KW - Multivariable models

KW - Outcome

KW - Prediction

UR - http://www.scopus.com/inward/record.url?scp=0026098579&partnerID=8YFLogxK

U2 - 10.1097/00007632-199101000-00002

DO - 10.1097/00007632-199101000-00002

M3 - Article

C2 - 1825895

AN - SCOPUS:0026098579

VL - 16

SP - 7

EP - 14

JO - Spine

JF - Spine

SN - 0362-2436

IS - 1

ER -