Psychometric Properties of the Patient-Specific Functional Scale in Patients with Low Back Pathology: A Systematic Review and Meta-Analysis

Abstract

Purpose: This systematic review and meta-analysis identifies, critically appraises, synthesizes, and meta-analyses the reported psychometric properties of the Patient-Specific Functional Scale (PSFS) in patients with low back pain or pathology. Method: The MEDLINE, Embase, PubMed, and Google Scholar databases were searched from their inception to September 2019. We included prospective measurement studies that reported on the psychometric properties (reliability, validity, responsiveness) of the PSFS in people with low back pain or pathology. We followed the COnsensus-based Standards for the selection of health Measurement INstruments 2018 guideline for systematic reviews. We performed both quantitative and qualitative syntheses in which the results were summarized on the basis of the reported measurement properties and study quality. Results: Ten eligible studies were included. The pooled PSFS reliability measure was excellent (intra-class correlation coefficient = 0.89; 95% CI: 0.75, 0.95). Validity measures displayed correlations that ranged from −0.47 to 0.69 when compared with other patient-reported outcome measures (PROMs) or other tests. Eight studies had assessed the responsiveness of the PSFS. Effect sizes reported were large (≥ 0.91). Conclusions: The PSFS is a reliable, valid, and responsive PROM for patients with low back pain or pathology.

The 2011 Canadian Community Health Survey defined activity-limiting musculoskeletal injuries as “the result of an incident that occurred in the past 12 months that was severe enough to limit normal daily activities for at least one day.” These injuries pose a major economic burden of $19.8 billion to Canadian society as a result of direct and indirect health care costs (decreased productivity, disability, or premature death).¹ In 2009–2010, the 12-month prevalence of activity-limiting musculoskeletal injuries of the lower back and spine alone was nearly 2% in the general population.¹ This corresponds to 548,000 injuries among individuals (315,000 men, 233,000 women) aged 12 years or older.¹

These estimates highlight and emphasize the need to develop innovative and efficacious rehabilitation intervention programmes for musculoskeletal injuries of the lower back and spine. The design, conduct, and assessment of such interventions must involve psychometrically proven (reliable, valid, responsive) methods of quantifying intervention effectiveness.² Research and clinical settings have extensively reported on the use of patient-reported outcome measures (PROMs), clinician-reported outcome measures, and performance-based tests.^3–6

PROMs capture a set of self-reported information from the patients’ perspective that offers insights into their perception of the nature of the reported symptoms, the degree of physical and mental impairment, and functional changes (improvement and deterioration).⁷ Moreover, PROMs facilitate individualized patient goal setting, improve clinical decision making by allowing clinicians to choose the most suitable treatment option, and enable clinicians to reassess patients to examine the degree to which an intervention was effective.⁷

One such PROM is the Patient-Specific Functional Scale (PSFS), developed by Stratford and colleagues with the aim of examining functional change mainly among patients with musculoskeletal conditions.⁸ When a clinician administers the PSFS, the patient is required to identify up to five important activities (with two additional options) that they are unable to perform or have difficulty with as a result of their current condition. Then the patient is asked to rate each identified activity on a scale ranging from 0 to 10, where 0 = unable to perform activity and 10 = able to perform activity at preinjury level. At follow-up, the patient is presented with the previously identified limiting activities and their corresponding ratings (0–10) and is asked to re-rate each activity on the basis of their perception of their current function.⁸

When assessing the efficacy of musculoskeletal – lower back and spine – rehabilitation interventions or attempting to quantitatively account for functional changes in patients’ status, it is crucial that the PROM (the PSFS in this case) used be reliable, valid, and responsive.² Horn and colleagues, whose review included only three studies on the measurement properties of the PSFS among patients with lower back or spine pathologies, concluded that there were gaps in the research literature regarding the reliability, validity, and responsiveness parameters of the PSFS.⁹ Given the increase in the number of newly published measurement studies (10) on this topic, an up-to-date review, meta-analysis, and critical appraisal of the quality of available evidence, using the COnsensus-based Standards for the selection of health Measurement INstruments (COSMIN) 2018 guidelines to incorporate the latest available evidence, is warranted.

Therefore, the objectives of this systematic review were to identify, critically appraise, synthesize, and meta-analyse the measurement properties of the PSFS in patients with low back pain, low back pathology, or both.

Methods

Guidelines and protocol registration

We followed the COSMIN 2018 guidelines for systematic reviews of PROMs.^10–11

Eligibility criteria

We included prospective measurement studies, published in the English language and in peer-reviewed journals, that had recruited patients with low back pain or pathology and that reported on at least one psychometric property (reliability, validity, responsiveness) of the PSFS.^5–6

Exclusion criteria

Studies with no data on the measurement properties of the PSFS among patients with low back pain or pathology were excluded from this systematic review,^5–6 as were posters and conference abstracts.^5–6

Information sources

We searched four electronic databases – MEDLINE, Embase, PubMed, and Google Scholar – from inception to September 2019, using the following keywords: (reliability OR consistency OR validity OR responsiveness OR calibration OR validation OR agreement OR minimal detectable change OR clinically important difference OR psychometric properties OR measurement properties) AND (patient specific functional scale) AND (musculoskeletal conditions) OR (low back conditions) OR (low back pain) OR (acute low back pain) OR (chronic low back pain). Additional studies were identified using the reference lists of the included studies.

Study selection

Two independent reviewers (GN and PB) performed the systematic electronic searches in all the databases. GN and PB also identified and removed duplicate studies. The independent reviewers then screened the titles and abstracts and identified the full-text studies. The final stage consisted of carrying out independent full-text reviews to determine which final articles were eligible to be included. In case of disagreement, a third reviewer – the most experienced (JCM) – provided consensus through discussion.

Data extraction

Two independent reviewers (SR and PHA) carried out the data extraction. The extracted data included the author, year, study population, setting, sample size, interval measurement evaluated, and intervention detail.^10–11 On the basis of the COSMIN guidelines for systematic reviews, information concerning the criterion and construct validity, reliability and measurement error, and responsiveness were extracted.^10–11 When the data were insufficient, GN contacted the original authors by email and requested further data.

To quantify reliability, reported as intra-class correlation coefficients (ICCs), we considered ICCs < 0.40 to have poor reliability; 0.40 to < 0.75, fair to good reliability; and ≥ 0.75, excellent reliability.¹² To quantify validity outcomes, Pearson or Spearman correlation coefficients were selected. The strength of a correlation coefficient was categorized using the guide supplied by Evans: 0.00−0.19 = very weak, 0.20−0.39 = weak, 0.40−0.59 = moderate, 0.60−0.79 = strong, and 0.80−1.00 = very strong.¹³ Measurement error statistics included standard error of measurement (SEM), calculated as SD × √1 – ICC, and minimal detectable change (MDC), calculated as SEM × 1.96 × √2.¹² To quantify responsiveness and to allow clinical decision making, we used the benchmark effect size (ES) scores suggested by Cohen: trivial, < 0.20; small, ≥ 0.20 to < 0.50; moderate, ≥ 0.50 to < 0.80; and large, ≥ 0.80.¹⁴

COSMIN guidelines for systematic reviews of patient-reported outcome measures

To summarize the quality of the available evidence, the included measurement studies were appraised by two independent authors using the COSMIN 2018 guidelines for systematic reviews of PROMs.^10–11 First, we used the COSMIN 2018 risk-of-bias checklist and rated each study as very good, adequate, doubtful, or inadequate.^10–11 To examine the overall rating of the quality of evidence for a specific measurement property, the lowest rating among the studies was considered.^10–11

Second, we evaluated the results of individual studies on a measurement property against the updated criteria for good measurement properties.^10–11 This involved evaluating the results of the included articles as sufficient, insufficient, or indeterminate.^10–11

Finally, the overall result was assessed using the modified Grading of Recommendations, Assessment, Development, and Evaluation approach to evaluate the quality of evidence.^10–11 This approach considers the risk of bias, inconsistency, imprecision, and indirectness to provide a rating of the quality of the evidence (high, moderate, low, or very low),^10–11 where high quality = “We are very confident that the true measurement property lies close to that of the estimate of the measurement property”; moderate quality = “We are moderately confident in the measurement property estimate: the true measurement property is likely to be close to the estimate of the measurement property, but there is a possibility that it is substantially different”; low quality = “Our confidence in the measurement property estimate is limited: the true measurement property may be substantially different from the estimate of the measurement property”; and very low quality = “We have very little confidence in the measurement property estimate: the true measurement property is likely to be substantially different from the estimate of the measurement property”.^10–11

Summary

A meta-analysis of the correlation of the ICCs (reliability) was performed using R, (Version 3.6.3; R Foundation for Statistical Computing, Vienna, Austria) using a random effects model, and 95% CIs were displayed for the correlation coefficients. A qualitative synthesis and assessment of methodological quality was also performed for the remaining measurement properties (i.e., validity and responsiveness).^3–6

Results

Study selection

Our search identified 198 articles after duplicates were removed. We then carried out title and abstract screening, resulting in 44 articles selected for full-text review. Of these, 10 individual measurement studies were deemed eligible for this systematic review and meta-analysis.^15–24 A flowchart of the selection process is displayed in Figure 1.

Figure 1 .

Flowchart of study selection.

Study characteristics

The characteristics of the 10 studies are displayed in Table 1. Five studies reported on reliability estimates of the PSFS, 6 reported on validity estimates, and 8 reported on responsiveness estimates. These studies were conducted between 1995 and 2018 and included 1,769 patients with low back pain, pathology, or both.^15–24 Study size ranged from 48 to 831 patients.

Table 1 .

Summary of Studies Addressing the Psychometrics of the PSFS

Study	Sample	Setting	Sample size	Measurement property assessed	Treatment	Retest interval or follow-up
Abbott & Schmitt23	Patients (mean age 34 [SD 17] y) with neck, back, upper extremity, or lower extremity problems, undergoing episode of care with physiotherapist	University of Otago School of Physiotherapy Clinics	N = 250 (gender not reported)	Concurrent, convergent, discriminant validity; responsiveness	Rehabilitation programme	At 2nd, 6th, and/or discharge follow-up visits
Cleland et al.21	Patients (mean age 70 [SD 8] y) with lumbar spinal stenosis	Wilford Hall Medical Center, Brooke Army Medical Center	N = 55 (43% women)	Test–retest reliability, construct validity, responsiveness	PT treatment	6 wk
Costa et al.22	Patients (mean age 44 [SD 17] y) with new episodes of acute LBP doing PT	PT clinics	N = 99 (58% women)	Test–retest reliability, responsiveness	PT treatment	24 h (reproducibility) and 2 wk or at discharge (responsiveness)
Hall et al.24	Patients (mean age 49.7 [SD 15.9] y) with LBP	Community	N = 831 (56% women)	Responsiveness	Treatment ranged from NSAIDs vs. placebo NSAIDs to exercise, motor control exercises, and manipulation therapy	Directly after treatment (“short term”) and several wk post-treatment (“mid-term”)
Lehtola et al.15	Aged 18–65 y with acute, subacute, or chronic LBP	Public and private PT clinics	N = 78 (68% women)	Construct validity, responsiveness	PT treatment; no specific protocol mentioned	After PT treatment (no specific time frame provided)
Maughan & Lewis16	Chronic (> 4 wk) LBP	PT clinics	N = 48 (67% women)	Test–retest reliability, responsiveness	Patient education programme delivered by physiotherapists	5 wk after treatment
Pengel et al.17	Subacute (6 wk–3 mo) LBP	University	N = 156 (56% women)	Responsiveness	Conservative treatments, non-specified	6 wk
Riebel et al.18	Military personnel with MSK pain > 2 wk	Military facility	N = 85 (20% women)	Construct validity	PT treatment, not specified	6 wk (or at end of treatment)
Sharma et al.19	MSK pain (with LBP)	Hospital and university	N = 104; n = 48 with low back pain (gender not reported)	Test–retest reliability, construct or concurrent validity, responsiveness	Treatment not specified	2 wk
Stratford et al.20	LBP (non-chronic)	Outpatient PT department	N = 63 (gender not reported)	Test–retest reliability, concurrent validity	PT treatment, no specific protocol mentioned	2 wk

PSFS = Patient-Specific Functional Scale; PT = physical therapy; LBP = low back pain; NSAIDS = non-steroidal anti-inflammatory drugs; MSK = musculoskeletal.

Methodological quality and measurement properties of the studies

In this study, we used a κ coefficient to assess the degree of agreement in assessment of methodological quality between the two reviewers (GN and SL; κ coefficient = 0.90; 95% CI: 0.85, 0.95). The methodological quality of the 10 studies is provided in Table 2. Most reliability studies were rated as adequate, whereas validity and responsiveness studies were rated as very good.

Table 2 .

Quality of Studies on the Measurement Properties of the PSFS: COSMIN 2018 Guidelines for Systematic Reviews

Study	Country	Reliability			Validity			Responsiveness
		No. of patients	RoB*	Result (rating)†	No. of patients	RoB*	Result (rating)†	No. of patients	RoB*	Result (rating)†
Abbott & Schmitt23	New Zealand				250	Very good (discriminative)	+	250	Very good (construct)	+
									Very good (discriminative)	+
Cleland et al.21	United States	20	Adequate (test–retest	−	35	Very good (construct)	+	55	Very good (construct)	+
			Adequate (measurement error)	−					Very good (gold standard)	+
									Doubtful (before and after)	?
Costa et al.22	Australia	50	Adequate (test–retest)	−	99	Very good (construct)	?	99	Very good (construct)	?
									Very good (gold standard)	+
Hall et al.24	New South Wales, Australia, or Auckland, New Zealand							1,662	Very good (construct)	?
									Very good (gold standard)	+
									Very good (before and after)	?
Lehtola et al.15	Finland				78	Very good (construct)	+	78	Very good (gold standard)	−
Maughan and Lewis16	United Kingdom	25	Adequate (test–retest)	+				23	Very good (gold standard)	+
			Adequate (measurement error)	+
Pengel et al.17	Australia							156	Adequate (construct)	?
									Doubtful (before and after)	?
Riebel et al.18	United States				85	Very good (construct)	−
Sharma et al.19	Nepal	36	Very good (test–retest)	+	104 (48 LBP)	Very good (construct)	+	64	Very good (gold standard)	+
			Very good (measurement error)	+		Very good (discriminative)	+
Stratford et al.20	Canada	63	Doubtful (test–retest)	+	63	Very good (criterion)	+
			Doubtful (measurement error)	+
Pooled or summary result (overall rating)		Low-quality evidence‡§

^*Methodological quality of the included studies using the COSMIN Risk of Bias checklist. Each study was rated as very good, adequate, doubtful, or inadequate quality.

^†Criteria for good measurement properties: + = sufficient; − = insufficient; ? = indeterminate.

^‡The quality of evidence was downgraded by one level because of risk of bias (i.e., doubtful).

^§The quality of evidence was downgraded by one level because of inconsistency.

PSFS = Patient-Specific Functional Scale; COSMIN = COnsensus-based Standards for the selection of health Measurement INstruments; RoB = risk of bias; LBP = low back pain.

Patient-Specific Functional Scale reliability estimates (meta-analysis)

Five studies (Table 3) were pooled to assess the test–retest reliability estimates of the PSFS in patients with low back pain or pathology.^16,19–22 The pooled results indicated excellent correlation coefficients (five studies; 194 patients; ICC = 0.89; 95% CI: 0.75, 0.95; between-study heterogeneity [I²] = 88%). The decision to meta-analyse was based on a homogeneous sample (all patients with low back pathology) and on the fact all five reported test–retest reliability.

Table 3 .

Summary of the Reliability Properties of the PSFS

Study	Type of reliability	No. of patients	Reliability estimates
Cleland et al.21	Test–retest	20	ICC 0.59, 95% CI: 0.2, 0.81 SEM 1.03 points MDC90% 2.4 points
Costa et al.22	Test–retest	50	ICC 0.85, 95% CI: 0.77, 0.90
Maughan & Lewis16	Test–retest	25	ICC 0.91 SEM 0.5 points
Sharma et al.19	Test–retest	36	ICC 0.89, 95% CI: 0.78, 0.94 SEM 0.63 points MDC95% 1.46 points
Stratford et al.20	Test–retest	63	ICC 0.97 SEM 0.41 points

PSFS = Patient-Specific Functional Scale; ICC = intra-class correlation coefficient; SEM = standard error of measurement; MDC = minimal detectable change.

Patient-Specific Functional Scale validity estimates

We found seven studies that had assessed the validity of the PSFS (Table 4).^15,18–23 We did not meta-analyse our validity estimates despite the homogeneous sample because the PSFS estimates were compared with a wide range of PROMs. Instead, we produced a qualitative synthesis.

Table 4 .

Summary of the Validity Properties of the PSFS

Study	Type of validity	Validity estimate (range, min–max)
Abbott & Schmitt23	Discriminative (N = 250): PSFS vs. ODI	0.62
Cleland et al.21	Construct (N = 35): PSFS vs. GRoC	0.69
Costa et al.22	Construct (N = 99)
	PSFS vs. FRI	−0.53
	PSFS vs. RMDQ	−0.51
	PSFS vs. NRS	−0.45
Lehtola et al.15	Construct (N = 78)
	PSFS vs. RMDQ	0.63
	PSFS vs. VAS	0.59
Riebel et al.18	Construct (N = 85)
	PSFS vs. painful patterns	0.41 (0.00−0.70)
	PSFS vs. SFMA	0.22 (−0.21−0.58)
Sharma et al.19	Construct and discriminative (N = 100)
	PSFS vs. ODI	−0.47
	PSFS vs. NPRS	−0.32
Stratford et al.20	Criterion (N = 63): PSFS vs. RMDQ	0.55−0.74

PSFS = Patient-Specific Functional Scale; ODI = Oswestry Disability Index; GRoC = Global Rating of Change; FRI = Functional Rating Index; RMDQ = Roland–Morris Disability Questionnaire; NRS = numerical rating scale; VAS = visual analogue scale; SFMA = Selective Functional Movement Assessment; NPRS = Numeric Pain Rating Scale.

Abbott and Schmitt (very good quality; sufficient properties) compared PSFS scores with Oswestry Disability Index (ODI) scores and reported strong (positive) correlations (r = 0.62) among 250 patients with low back pain (lower extremity problems).²³

Cleland and colleagues (very good quality; sufficient properties) compared PSFS scores with Global Rating of Change (GRoC) scores and indicated strong (positive) correlations (r = 0.69) among the 35 patients with lumbar spinal stenosis.²¹

Costa and colleagues (very good quality; indeterminate properties) compared PSFS scores with Functional Rating Index, Roland–Morris Disability Questionnaire (RMDQ), and numerical rating scale scores and found moderate (negative) correlations (r = −0.45 to −0.53) among 99 patients with lumbar spinal stenosis.²²

Lehtola and colleagues (very good quality; sufficient properties) compared PSFS scores with visual analogue scale (VAS) scores (r = 0.59; moderate positive correlation) and with RMDQ scores (r = 0.63; strong positive correlation) among 78 patients with low back pain.¹⁵

Riebel and colleagues (very good quality; insufficient properties) compared PSFS scores with Selective Functional Movement Assessment (SFMA) scores (r = 0.22; weak positive correlation) and with a number of painful patterns (r = 0.41; moderate positive correlation) among 85 patients with musculoskeletal pain in their lumbopelvic region.¹⁸

Sharma and colleagues (very good quality; sufficient properties) compared PSFS scores with ODI scores (r = −0.47; moderate negative correlation) and with the Numeric Pain Rating Scale (NPRS) scores (r = −0.32; weak negative correlation) of 104 patients (48 with low back pain).¹⁹

Stratford and colleagues (very good quality; sufficient properties) compared PSFS scores with RMDQ scores and found moderate to strong (positive) correlations (r = 0.55–0.74) among 63 patients with low back pain.²⁰

Patient-Specific Functional Scale responsiveness estimates

We identified eight studies that assessed the responsiveness of the PSFS (Table 5). Again, we did not meta-analyse our responsiveness estimates despite homogeneous samples because various responsiveness estimates were reported. Instead, we produced a qualitative synthesis.

Table 5 .

Summary of the Responsiveness Properties of the PSFS

Study	Type of responsiveness	Responsiveness estimates
Abbott & Schmitt23	ES (Cohen d; N = 250)	d = 1.53
Cleland et al.21	ES (N = 55)	0.91
	External responsiveness	0.84
	GRI	r = 1.75
	AUC	2.00
	MCID	1.34
Costa et al.22	ES (N = 99)	0.95
	Correlation of change scores with GPE	r = 0.34
Hall et al.24	LBP (short term):
	ES (n = 831)	1.70
	SRM (n = 831)	1.11
	Correlations with GPE	r = 0.61
	Low activity limitation (short term):
	ES (n = 453)	1.55
	SRM (n = 453)	1.11
	Correlations with GPE	r = 0.60
	High activity limitation (short term):
	ES (n = 378)	2.16
	SRM (n = 378)	1.12
	Correlations with GPE	r = 0.65
	LBP (mid-term):
	ES (n = 820)	1.80
	SRM (n = 820)	1.11
	Correlations with GPE	r = 0.65
	Low activity limitation (mid-term):
	ES (n = 452)	1.74
	SRM (n = 452)	1.81
	Correlations with GPE	r = 0.61
	High activity limitation (mid-term):
	ES (n = 368)	2.16
	SRM (n = 368)	1.04
	Correlations with GPE	r = 0.70
Lehtola et al.15	ES (n = 78)	1.02
	SRM	0.97
Maughan & Lewis16	MCID (n = 23)	2.3
	AUC	0.86
Pengel et al.17	ES (n = 156)	1.6 (1.4−1.8)
	GRI	r = 2.1
Sharma et al.19	MIC (n = 64)	2.00
	AUC	0.83 (0.74, 0.91)

PSFS = Patient-Specific Functional Scale; ES = effect size; GRI = Guyatt Responsiveness Index; AUC = area under the receiver operating characteristic curve; MCID = minimally clinically important difference; GPE = Global Perceived Effect; LBP = low back pain; SRM = standardized response mean; GRoC = Global Rating of Change; MIC = minimally important change.

Abbott and Schmitt (very good quality; sufficient properties) reported a Cohen’s d of 1.53 in 250 patients with low back pain,²³ consistent with a large ES.²³

Cleland and colleagues (very good quality; sufficient properties) reported an ES of 0.91, external responsiveness of 0.84, Guyatt Responsiveness Index (GRI) of 1.75, area under the receiver operating characteristic curve (AUC) of 2.00, and a minimal clinically important difference (MCID) of 1.34 among 55 patients with lumbar spinal stenosis.²¹

Costa and colleagues (very good quality; sufficient properties) reported a large ES of 0.95 with the Global Perceived Effect (GPE) and a correlation of 0.34 with GRoC in 99 patients with lumbar spinal stenosis.²²

Hall and colleagues (very good quality; sufficient properties) reported an ES of 1.55−2.16, standardized response mean (SRM) of 1.04−1.81, and correlations with GPE scores that ranged from r = 0.60−0.70 at short- and mid-term follow-ups in patients with low back pain.²⁴

Lehtola and colleagues (very good quality; insufficient properties) indicated ES and SRM of 1.02 and 0.97, respectively, in 78 patients with low back pain, indicating large ESs.¹⁵

Maughan and Lewis (very good quality; sufficient properties) reported an MCID of 2.3 points and an AUC of 0.86 in 23 patients with low back pain.¹⁶

Pengel and colleagues (adequate or doubtful quality; indeterminate properties) indicated an ES of 1.6 and GRI r of 2.1 among 156 patients with low back pain.¹⁷

Sharma and colleagues (very good quality; sufficient properties) reported the minimally important change of 2.00 points and AUC of 0.83 among 64 patients with low back pain.¹⁹

Discussion

This review synthesized the current evidence regarding the psychometric properties of the PSFS in patients with low back pain or pathology. On the basis of the current evidence, the PSFS displayed (1) excellent reliability estimates (low-quality evidence) and (2) moderate to strong construct validity with the ODI, GRoC, VAS, and RMDQ (very good quality evidence; sufficient measurement properties). In addition, on the basis of four very good quality studies with sufficient measurement properties, the PSFS was identified as a responsive outcome measure to assess functional change in patients with low back pain or pathology.

The results of our systematic review agree with the findings of the review by Horn and colleagues, which included three measurement (reliability) studies and two validity and responsiveness studies.⁹ That review concluded that the evidence (up to May 2010) indicated that the PSFS had demonstrated sufficient reliability, validity, and responsiveness in the patient population with low back pain or pathology.⁹ Moreover, Horn and colleagues used a checklist developed by Terwee and colleagues to assess the quality of the included studies. Our review builds on the review by Horn and colleagues in that it (1) provides more estimates of comprehensive reliability (five studies), validity (seven studies), and responsiveness (eight studies) and includes six additional measurement studies and (2) provides ratings of the quality of the evidence using COSMIN 2018 guidelines for systematic reviews of PROMs.

Five studies were pooled to assess the test–retest reliability estimates of the PSFS in patients with low back pain or pathology.^16,19–22 This process yielded a pooled ICC estimate of 0.89 (95% CI: 0.75, 0.95). However, between-study heterogeneity was high (I² = 88%). A possible explanation for this high level might be that the studies included patients with varied low back pathology (patients with acute, non-chronic, or chronic low back pain; lumbopelvic pain; lumbar spinal stenosis). As recommended by COSMIN 2018, we downgraded the level of evidence by one level because of inconsistency). However, it is important to note that the lower boundary of the 95% CI of our pooled ICC estimate was 0.75; an ICC score of 0.75 indicates excellent reliability.

Evidence from the majority of the included studies indicated that the PSFS had moderate to strong correlations with other PROMs and other tests. However, we noted that Riebel and colleagues had indicated that the correlation between the PSFS and SFMA was 0.22,¹⁸ and Sharma and colleagues found the correlation between the PSFS and NPRS to be −0.32.¹⁹ The study by Riebel and colleagues considered the SFMA a performance-based test, with its psychometric properties dependent on the clinician’s level of experience and scoring method used.¹⁸

Moreover, using PROMs as surrogates for a performance-based test or vice versa is suboptimal and inappropriate.^6,25 The study by Sharma and colleagues reported a weak correlation of −0.32,¹⁹ perhaps because the study sample included adults experiencing musculoskeletal pain (low back, knee, shoulder, neck, elbow), and 48% experienced low back pain. It is possible that the multiplicity of musculoskeletal pain (high heterogeneity) could have contributed to such low correlations.

Implications for practice

On the basis of the evidence we synthesized, we propose that the PSFS has high test–retest reliability and construct validity and can be used as a PROM for assessing functional change among patients with low back pain and pathology. When assessing the extent to which an intervention was effective, clinicians may consider changes ranging from 1.34 to 2.3 points on the PSFS to be clinically important deteriorations or improvements.

This review had two potential limitations. First, we did not include studies that were not published in English; however, this was consistent with our registered protocol. Second, we did not include any results from the grey literature, abstracts, or theses.

Conclusion

We synthesized low-quality evidence demonstrating that the PSFS is a reliable PROM as well as very good quality evidence with sufficient measurement properties to indicate that the PSFS is a valid and responsive outcome measure in patients with low back pain or pathology.

Key Messages

What is already known on this topic

The Patient-Specific Functional Scale (PSFS) is a self-reported outcome measure that assesses functional change among patients with musculoskeletal conditions. There is sparse evidence concerning the measurement properties of the PSFS among patients with low back pain or pathology.

What this study adds

The PSFS demonstrated high test–retest reliability and construct validity measures and can be used as a patient-reported outcome measure to assess functional change in patients with low back pain or pathology.