EVIDENCE ABOUT DIAGNOSTIC TESTS

Min H. Huang, PT, PhD, NCS

Diagnostic Tests

Test threshold and treatment threshold Help focus the exam in a particular body

region or system. Identify potential problems that require

referral to other health care providers. Assist in the diagnostic classification (i.e. a

specific practice pattern). Diagnostic tests MUST be reliable and valid.

Study Credibility

Appraisal of evidence begins with assessment of research validity.

Higher levels of validity indicate greater confidence that there is a lack of bias.

Lists of specific questions to ask (Table 10-1 in the textbook).

Specific Questions to ask

Can the research questions or hypotheses be tested with the research design

Did the investigators compare results from the diagnostic test to results from a “gold standard” diagnostic test

Were all subjects evaluated with the comparison diagnostic test

Were the individuals performing and interpreting each test’s result unaware of the other test’s results (i.e. were they masked, or blinded)

Specific Questions to ask

Did the investigators include subjects with all levels of stages of the condition being evaluated by the measure of interest

Did the investigators confirm their findings with a new set of subjects

Did the study use appropriate statistical analysis methods for reliability and validity Correlation coefficients Face, construct, criterion, concurrent validity

Were p values or C.I. significant?

Study Results

Sensitivity (SnNout) Specificity (SpPin) Positive predictive value (PPV) Negative predictive value (NPV) Likelihood ratios (LR):

reflect a diagnostic test’s ability to provide persuasive information

LR + = Sn/(1-Sp) LR – = (1-Sn)/Sp

Receiver Operating Characteristic Curves (ROC) a graphic way to evaluate different thresholds of a

test

Loong et al. (2003).

Sn = 24/30 = 80%

Sp = 56/70 = 80%

PPV = 24/38 = 63%

NPP = 56/62 = 90%

Figure 10-7: A Receiver Operating Characteristic (ROC) Curve for an Imperfect but Useful Test

Likelihood Ratio Nomogram

Use a nomogram to calculate posttest probability, i.e. the probability that the patient/client has the condition after a test result is obtained.

LR+ = 1-2, LR- = 0.5-1.0 negligible change in pretest probability

http://www.cebm.net/index.aspx?o=1043

EVIDENCE ABOUT CLINICAL MEASURES

Min H. Huang, PT, PhD, NCS

Clinical Measures

Are NOT used to label or classify a diagnosis or practice pattern

Quantify and/or describe a patient’s impairments in a standardized fashion

Distinguish among different levels of severity of a problem

Instruments must have reliability, validity, responsiveness

Study Credibility

SAME process as diagnostic tests Refer to questions in Table 10-2 Clinical measures MUST be validated in

patient populations with different diagnoses

Study Results

Reliability and validity are confirmed by correlation coefficients.

Responsiveness is commonly assessed by Minimal detectable change (MDC): the amount of

change that just exceeds the standard error of measurement

Standardized response mean (SRM): the ratio between the mean change score and the standard deviation of the change scores; reflect the change over time

Considerations for Implementing the Evidence into Practice

Test or measure should be available, practical and safe in the setting

Test or measure should have demonstrated performance on similar patient/clients

Can pretest probabilities be estimated for the patient/client

Patient/client’s preferences and values

Review

Most useful diagnostic tests and clinical measures have demonstrated reliability and validity

Reliability is shown through statistical tests of relationships among repeated test results

Validity is demonstrated through statistical tests or comparison to the gold standard

Responsiveness is measured MDC or SRM

IMPACT OF PAIN REPORTED DURING ISOMETRIC

QUADRICEPS MUSCLE STRENGTH TESTING IN PEOPLE

WITH KNEE PAIN:

DATA FROM THE OSTEOARTHRITIS INITIATIVE

DANIEL L. RIDDLE, PAUL W. STRATFORD

Min H. Huang, PT, PhD, NCS

Introduction

Common clinical assumption Impairments in body structure or function (e.g.

pain) can impact limitations in activities and participation (e.g. physical function)

Limitations of previous research NO large scale studies available Does pain affect muscle strength?

1 study: Yes

1 study: No

Relationships between Domains of the ICF Model

Purpose

Whether the relationship between maximal isometric quad strength (X1) and functional status (Y1,Y2,….Y5) was influenced by pain during isometric testing (X2)

The extent to which pain during testing (X1) affected quad strength (Y1), or other functional tests (Y2, Y3, Y4, Y5)

Purpose

Model 1 (Initial):

Physical Function (Y) = β1 Strength (X1) + covariates + ε

Model 2 (Full):

Physical Function (Y) = β1 Strength (X1) + β2 Pain (X2) + β3 Strength (X1) × Pain(X2) + covariates + ε

Y

X

Y

X

Purpose

Model 3 (No interaction)

Physical Function (Y) = β1 Strength (X1) + β2 Pain (X2) + covariates + ε

Y

X

Y

X

Model 1

Y

X

Model 2

Purpose – Class Discussion

Method

Participants (n=1,344) Unilateral knee pain

Verbal Numerical Rating (VNR) > 3

WOMAC pain >1

Outcome variables WOMAC physical

function

20-m walk

400-m walk

5 times sit to stand

Independent variables MAX Quad strength Pain during Quad

strength testing

Multiple regression models Model 1 Model 2 Model 3

95% CI of β excludes 0

Method – Class Discussion

Results

Pain did NOT modify or confound any of the outcome variables: 400-m walk, 20-m walk, chair stand, WOMAC – Physical Function.

Results

Table 6. MODERATE or SEVER pain during testing was WEAKLY associated with reduced STRENTGH, but mild pain was not.

Results – Class Discussion

Discussion

Pain during maximal isometric Quad strength tests did not affect the construct validity of the tests

Isometric Quad muscle strength and functional status relationship is NOT affected by reports of pain during testing

Discussion – Class Discussion

Limitations

Were the samples representative of the population treated?

Measurement of pain? No psychometric properties reported.

Muscle strength measured by dynamometer – Is it applicable to clinical settings using MMT?

No Hypotheses; No power estimate – finishing expedition?