Patient-reported outcome measures used in patients undergoing total knee arthroplasty.

Patient-reported outcome measures (PROMs) are being used increasingly in total knee arthroplasty (TKA). We conducted a systematic review aimed at identifying psychometrically sound PROMs by appraising their measurement properties. Studies concerning the development and/or evaluation of the measurement properties of PROMs used in a TKA population were systematically retrieved via PubMed, Web of Science, Embase, and Scopus. Ratings for methodological quality and measurement properties were conducted according to updated COnsensus-based Standards for the selection of health Measurement INstruments (COSMIN) methodology. Of the 155 articles on 34 instruments included, nine PROMs met the minimum requirements for psychometric validation and can be recommended to use as measures of TKA outcome: Oxford Knee Score (OKS); OKS-Activity and Participation Questionnaire (OKS-APQ); 12-item short form Knee Injury and Osteoarthritis Outcome (KOOS-12); KOOS Physical function Short form (KOOS-PS); Western Ontario and McMaster Universities Arthritis Index-Total Knee Replacement function short form (WOMAC-TKR); Lower Extremity Functional Scale (LEFS); Forgotten Joint Score (FJS); Patient's Knee Implant Performance (PKIP); and University of California Los Angeles (UCLA) activity score. The pain and function subscales in WOMAC, as well as the pain, function, and quality of life subscales in KOOS, were validated psychometrically as standalone subscales instead of as whole instruments. However, none of the included PROMs have been validated for all measurement properties. Thus, further studies are still warranted to evaluate those PROMs. Use of the other 25 scales and subscales should be tempered until further studies validate their measurement properties. Cite this article: Bone Joint Res 2021;10(3):203-217.

We summarized available patient-reported outcome measures (PROMs) used in a total knee arthroplasty (TKA) population, and identified those with high quality by evaluating their measurement properties.

Nine instruments and six subscales in two instruments met the minimal requirements for psychometric validation, and can be recommended to use as measures for TKA outcome.

However, none of the included PROMs have been validated for all nine measurement properties.

Further studies are warranted to assess the measurement properties of existing instruments, especially for content validity, cross-culture validity, and structure validity.

We undertook a comprehensive review under the latest COnsensus-based Standards for the selection of health Measurement Instruments (COSMIN) methodology.

We also used meta-analysis to summarize results quantitatively.

Our strict selection criteria meant that only a TKA population could be assessed.

Introduction

Total knee arthroplasty (TKA) has been performed for more than 40 years. It is regarded as efficacious treatment for end-stage knee arthritis, capable of improving quality of life by reducing pain and ameliorating long-term knee function.[1] Thanks to optimization of surgical methods and prosthetic designs, the ten-year survival of knee prostheses exceeds 90%.[2] However, the proportion of dissatisfied TKA recipients remains > 10%.[3,4]

TKA is an elective procedure, so patient-reported outcome measures (PROMs) are crucial to assess how well this type of intervention serves the patients’ goals rather than strictly objective measures alone.[5]

PROMs have advantages over objective measurements because they: 1) largely eliminate clinicians’ biases and measure health status accurately from the patients’ perspective;[6] 2) enable better detection of what patients account for, and help to address possibly modifiable factors;[7] 3) aid the possibility of follow-up of patients regardless of their direct attendance; and 4) facilitate decision-making for surgical procedures.[8,9]

In clinical practice and research, it is critical (but difficult) to opt for psychometrically sound rather than frequently used PROMs for certain purposes. In this context, systematic reviews were published in 2016 and 2017 evaluating the measurement properties of PROMs in knee arthroplasty population, using COnsensus-based Standards for the selection of health Measurement Instruments (COSMIN) methodology.[10,11] Since then, many studies assessing PROMs in TKA have been conducted. Moreover, the COSMIN methodology for systematic review of PROMs has been developed further. This methodology has: 1) established specific and comprehensive guidelines for evaluating content validity;[12] 2) updated criteria for good measurement properties, risk of bias checklist,[13] approach for grading the quality of evidence, and synthesizing the overall rating; and 3) formulated recommendation standards for selection of PROMs.[14] For these reasons, we conducted a systematic review following the updated COSMIN methodology to establish a comprehensive quality assessment of PROMs for TKA.

Methods

Search strategy

We systematically searched for studies reporting the measurement properties of PROMs used in the TKA population in PubMed, Web of Science, Embase, and Scopus from commencement to 8 March 2020. To find eligible studies, we used keywords from three terms: 1) patient-reported outcome measure; 2) measurement properties (including validity, reliability, internal consistency, measurement error, responsiveness, and minimal clinically important difference (MCID)); and 3) total knee arthroplasty population. The full search strategy for PubMed is shown in detail in the Supplementary Material.

Inclusion and exclusion criteria

Articles were considered eligible if published as full texts in English and if they detailed the development or evaluated the measurement properties of PROMs used in TKA. Articles were excluded if they: 1) did not report one of the nine measurement properties or MCID; 2) did not focus on patients undergoing (or who had undergone) TKA; 3) were not published in the English language; or 4) were not full reports (e.g. only abstracts were available) because they were unlikely to contain sufficient information.

Study selection and data extraction

After removal of duplicate articles, two reviewers (SZ and XC) screened titles and abstracts independently, and identified eligible articles. Then, full manuscripts were extracted and screened for final inclusion. Discrepancies between the reviewers were resolved by discussion. The bibliographies of all selected full-text articles were screened to retrieve additional citations.

The data extraction was undertaken by two reviewers (SZ and XC) independently. First, the characteristics of included PROMs were extracted from development studies, questionnaires, and user manuals. Second, results from included studies for evaluating the methodological quality of studies and measurement properties, as well as descriptive data on feasibility and interpretability, were recorded on a standardized form.

Assessment of the measurement properties of PROMs

The measurement properties of PROMs were evaluated under the updated COSMIN methodology,[12-14] which required the following consecutive procedures: evaluation of the methodological quality and measurement properties of single studies; qualitative and quantitative summary of the results of each instrument; and grading the quality of evidence and selecting instruments.

Evaluation of the methodological quality of individual studies

The methodological quality of studies was assessed using the COSMIN risk of bias checklist, which consists of ten 'tick boxes' in accordance with development studies and nine measurement properties. Each tick box includes 3 to 35 items, which are rated as “very good”, “adequate”, “doubtful”, or “inadequate” against standards. Two reviewers (YW and MY) completed the corresponding tick box per article independently. Besides, the overall rating of the methodological quality was based on the worst rating within each tick box.[13]

Evaluation of measurement properties for individual studies

We used the COSMIN methodology to assess content validity.[12] COSMIN set the criteria for good content validity from three aspects: 1) relevance (i.e. all items in a PROM should be relevant for the construct of interest within a specific population and context of use); 2) comprehensiveness (i.e. no key aspects of the construct should be missing); and 3) comprehensibility (i.e. the items should be understood by patients as intended). The rating for the content validity of each included PROM was conducted for a development study, content validity studies, and the instrument itself separately against the criteria. For the latter, the English version of included instruments was reviewed by English-fluent TKA-expert authors independently. Other measurement properties of the included instruments were evaluated according to the criteria described in Supplementary Table i.[14] The results of each study were rated qualitatively as “sufficient” (+), “insufficient” (–), or “indeterminate” (?). All results were evaluated by two reviewers (YW and MY) independently, and a third party (WQ) was consulted if consensus could not be reached.

Qualitative syntheses of the results of each instrument

The results for each measurement property from single studies per instrument were summarized qualitatively (i.e. + / – / ± / ?). An overall “sufficient” (+) or “insufficient” (–) rating was given if > 75% of results were concurrent. An “inconsistent” (±) rating was given if no rating exceeded 75% and no appropriate explanation for inconsistency could be given. An “indeterminate” (?) rating was given only if all single study results were indeterminate.

Quantitative syntheses of the results of each instrument

For PROMs with more than two available results on internal consistency, test–retest reliability or construct validity (i.e. Cronbach’s α, intraclass correlation coefficients (ICCs), and Pearson correlation coefficients) were also pooled statistically in a meta‐analysis. Weighted means and 95% confidence intervals were calculated for Pearson correlation coefficients against 36-Item Short-Form Health Survey questionnaire (SF-36; the most commonly used and validated comparator for evaluating construct validity).[15] Correlation between included scales with SF-36 pain and function subscales (convergent validity) should be higher than those with SF-36 mental and emotional subscales (divergent validity) by a minimum of 0.10. For test–retest reliability, ICCs were combined based on estimates derived from a Fisher transformation.[16] Weighted means and range of results were reported for Cronbach’s α. All analyses were undertaken with use of Stata v14.0 (StataCorp, USA).

Grading the quality of evidence for each instrument

The quality of evidence was graded for each property per instrument using Grading of Recommendations Assessment, Development and Evaluation (GRADE) approach.[14] The quality of evidence was downgraded from “high” by considering four factors: bias risk, inconsistency, imprecision, and indirectness, and eventually judged as “high”, “moderate”, “low”, or “very low” (Supplementary Table ii).

Formulating recommendations for selection of instruments

As COSMIN suggested, each PROM was placed in a recommendation category (A to C) according to its overall ratings and quality of evidence. PROMs with sufficient content validity and at least low-quality evidence for sufficient internal consistency were placed in category A, which meant they could be recommended for use and results obtained with these PROMs could be trusted. PROMs with high-quality evidence for an insufficient measurement property were placed into category C and should not be recommended for use. PROMs were placed in category B if they were categorized neither in A nor in C, which required further research to demonstrate their measurement properties.[14]

Results

Selection and characteristics of studies

In total, 155 articles were ultimately selected from 5,145 references (Figure 1). As a result, 34 PROMs were evaluated and their characteristics are presented in Table I.[17-50] Among them, the Oxford Knee Score (OKS; 39 articles),[42] Western Ontario and McMaster Universities Arthritis Index (WOMAC; 33 articles),[48] Knee Injury and Osteoarthritis Outcome Score (KOOS; 18 articles),[25] New Knee Society Scoring System (14 articles),[32] and Forgotten Joint Score (FJS; 12 articles)[20] were the most commonly evaluated instruments, with over ten included articles.

Fig. 1

Flowchart of the article selection process.

Table I.

The characteristics of included patient-reported outcome measures.

Instrument	Reference to the first article	Year initially published	Construct	Recall period	Number of items	Number of domains	Response options	Scoring	Range of score	Interpretation of the score	Numbers of articles included
American Academy of Orthopaedic Surgeons (AAOS) Hip and Knee Core Scale	Johanson et al [17]	2004	Impact on quality of life of hip / knee problem	Past week	7	3	5-, 6-, 7-point ordinal response	Sum up and convert to a metric of 0 to 100	0 to 100	100 for the best possible score	3
Anterior Knee Pain Scale (AKPS)	Kujala et al [18]	1993	Anterior knee pain symptoms	Not mention	13	1	3 to 4-, 5-point ordinal response	Sum up	0 to 100	0 for the worst possible score	2
Activity Scale for Arthroplasty Patients (ASAP)	Domzalski et al[51]	2010	High level functioning and participating for arthroplasty recipients	Today	10	1	4-point Likert response	Sum up	10 to 40	10 for the worst possible score	2
Core Outcome Measures Index knee (COMI-knee)	Impellizzeri et al [19]	2016	Lack of clear description	Item dependent	6	5	5-point Likert response and NRS	Take the average of score (mean) for each domain	0 to 10	0 for the best possible score	1
Forgotten Joint Score (FJS)	Behrend et al [20]	2012	Ability to forget the artificial joint	Not mention	12	1	5-point Likert response	Sum up and convert to a metric of 0 to 100	0 to 100	0 for the worst possible score	13
High-Activity Arthroplasty Score (HAAS)	Talbot et al [21]	2010	Variation in functional ability after arthroplasty	Not mention	4	1	4-, 5-, 6-, 7-point ordinal response	Sum up	0 to 18	0 for the worst possible score	3
High-Flexion Knee Score (HFKS)	Na et al [22]	2012	Knee status in the high-function range	Not mention	9	2	5-point Likert response	Sum up	9 to 45	9 for the worst possible score	1
Intermittent and Constant Osteoarthritis Pain (ICOAP)	Hawker et al [23,24]	2007	Constant and intermittent pain experienced by hip / knee OA patients	Past week	11	2	5-point Likert response	Sum up and convert to a metric of 0 to 100 for each subscale and total score	0 to 100	0 for the best possible score	9
Knee Injury and OsteoarthritisOutcome Score (KOOS)	Roos et al [25]	1998	Symptoms and function disability in knee injury and osteoarthritis patients	Past week	42	5	5-point Likert response	Sum up and convert to a metric of 0 to 100 for each subscale	0 to 100	0 for the worst possible score	18
Knee Injury and OsteoarthritisOutcome Score Joint Arthroplasty (KOOS, JR)	Lyhman et al [26]	2016	Knee health status before and after knee arthroplasty	Past week	7	1	5-point Likert response	Sum up and convert to a metric of 0 to 100	0 to 100	0 for the worst possible score	4
Knee Injury and OsteoarthritisOutcome Physical functionShort form (KOOS-PS)	Perruccio et al [27]	2008	Physical function states that represent the progression of physical disability from early to late knee OA	Past week	7	1	5-point Likert response	Sum up and convert to a metric of 0 to 100	0 to 100	0 for the best possible score	8
12-item short forms Knee Injury and OsteoarthritisOutcome (KOOS-12)	Gandek et al [28]	2018	Difficulties experienced by knee OA patients by measuring knee specific pain, function and quality of life	Past week	12	3	5-point Likert response	Sum up and convert to a metric of 0 to 100	0 to 100	0 for the worst possible score	2
Activities of Daily Living Scale of the Knee Outcome Survey (KOS-ADLS)	Irrgang et al [29]	1998	Symptoms and functional limitations experienced during activities of daily living	Past 1 to 2 days	14	2	6-point Likert response	Sum up and convert to a metric of 0% to 100%	0% to 100%	0 for the worst possible score	4
Knee Pain Questionnaire (KPQ)	Boeckstyns et al [30]	1987	Knee pain	Not mention	10	1	Dichotomous response	The sum of positive and unanswered questions	0 to 10	0 for the best possible score	2
Knee Surgery Perception Questionnaire (KSPQ)	Levinger et al [31]	2014	Discrepancy between patients’ expectations and actual functional abilities preceding to knee arthroplasty surgery	Current	Part A: 20Part B: 20	5	6-point Likert response	Part A&B: Sum up,Discrepancy score: Subtract the score of Part A from Part B	Part A&B: 0 to 120Discrepancy score: -120 to 120	Part A&B: 0 for the worst possible scoreDiscrepancy score: the lowest score for low expectation from surgery	1
2011 Knee Society Scoring System (KSS 2011)	Noble et al [32]	2011	Health status of the knee by evaluating pain relief, functional abilities, satisfaction, and fulfillment of expectations	Not mention	34 (7 objective, 27 subjective)	3 subjective domains	5-, 6-, 7-point ordinal response	Sum up for each subscale	Satisfaction subscale: 0 to 40Expectation subscale: 3 to 15Functional activities subscale: 0 to 100	The lowest score for the worst possible score for each subscale	15
Adjusted 2011 Knee Society Scoring System (KSS-A)	Dinjens et al [33]	2016	The same as KSS 2011	Not mention	25 (5 objective, 20 subjective)	5 (1 objective, 4 subjective)	3-, 5-, 6-, 7-point ordinal response and NRS	Sum up and convert to a metric of 0% to 100% for each subscale	0% to 100%	0 for the worst possible score	1
Knee Society Scoring System short form (KSS short form)	Scuderi et al [34]	2015	The same as NKSS	Not mention	10	3	3-, 5-, 6-point ordinal response and NRS	Sum up for each subscale	Lack of clear description	Lack of clear description	1
Lower-Extremity ActivityScale (LEAS)	Saleh et al [35]	2005	Lower limb activity	Not mention	18	1	1 point per checked item		1 to 18	1 for the worst possible score	2
Lower Limb Functional Scale(LEFS)	Binkley et al [36]	1999	Lower limb function	Today	20	1	5-point Likert response	Sum up	0 to 80	0 for the worst possible score	3
Lequesne Algofunctional Index for the Knee (Lequesne)	Lequesne et al [37]	1987	Severity of knee OA	Not mention	11	3	2-, 3-, 5-, 7-point ordinal response	Sum up and convert to six level of severity for knee OA	0 to 24	0 for the best possible score	3
Lysholm Knee Scoring Scale (Lysholm)	Lysholm et al [38]	1982	Lack of clear description	Not mention	8	1	3-, 4-, 5-, 6-point ordinal response	Sum up	0 to 100	0 for the worst possible score	2
Modified Forgotten Joint Score (MFJS)	Robinson et al [39]	2018	Ability to forget the artificial joint	Not mention	10	1	5-point Likert response	Sum up and convert to a metric of 0% to 100%	0% to 100%	0 for the worst possible score	1
Computer-Adaptive Test for Hip and Knee OA (OA-CAT)	McDonough et al [40]	2009	Function, Pain and Disability associated with hip / knee OA	Past month	15	3	5-point ordinal response	Logit scores transformed to T-scores (Mean: 50, SD 10)			3
OsteoArthritis of Knee and Hip Quality of Life Scale (OAKHQOL)	Rat et al [41]	2006	Quality of life affected by hip / knee OA	Past 4 weeks	43	5	11-point Likert response	Sum up and convert to a metric of 0 to 100	0 to 100	0 for the worst possible score	4
Oxford Knee Score (OKS)	Dawson et al [42]	1998	Knee arthroplasty outcomes by measuring problems for knee arthroplasty recipients	Past 4 weeks	12	1	5-point Likert response	Sum up	0 to 48	0 for the worst possible score	39
Oxford Knee Score–Activity and Participation Questionnaire (OKS-APQ)	Dawson et al [43]	2014	Knee arthroplasty outcome by measuring physical activities and social participation	Past 4 weeks	8	1	5-point Likert response	Sum up and convert to a metric of 0 to 100	0 to 100	0 for the worst possible score	1
Patient Administered Questionnaires Knee (PAQ-knee)	Mancuso et al [44]	2012	Knee condition by measuring pain, function, psychological wellbeing, and satisfaction	Item dependent	29	1	4-, 5-point ordinal response and NRS	Sum up 18 of the 29 questions	0 to 100	0 for the best possible score	1
Patient’s Knee Implant Performance (PKIP)	Lewis et al [45]	2014	Patients' functional performance of their arthroplasty knee.	Past week	24	4	5-, 6-, 11-point ordinal response	Sum up and convert to a metric of 0 to 100	0 to 100	0 for the worst possible score	2
Tegner Activity Scale	Tegner et al [46]	1985	Activity level	Current	10	1	1 point per checked item		0 to 10	0 for the worst possible score	3
University of California Los Angeles (UCLA) activity score	Amstutz et al [47]	1984	Activity level	Current	10	1	1 point per checked item		1 to 10	1 for the worst possible score	7
Western Ontario and McMaster Universities Arthritis Index (WOMAC)	Bellamy et al [48]	1988	Discomfort and disability in hip/knee OA patients	Past 48 hours	24	3	5-point Likert response, VAS, or NRS	Sum up for each subscale	Dependent on response opinions	0 for the best possible score	34
Western Ontario and McMaster Universities Arthritis Index-Total Knee Replacement function short form (WOMAC-TKR)	Libes et al [49]	2013	Knee specific function	Past 48 hours	7	1	VAS	Sum up and convert to a metric of 0 to 100	0 to 100	0 for the best possible score	1
Work, Osteoarthritis or joint-Replacement Questionnaire (WORQ)	Kievit et al [50]	2014	Impact on work ability of knee problems	Past week	13	1	5-point Likert response	Sum up and convert to a metric of 0 to 100	0 to 100	0 for the worst possible score	2

NRS, numerical rating scales; OA, osteoarthritis; VAS, visual analogue scale.

Methodological quality and rating for measurement properties

The methodological quality and rating for content validity is displayed in Supplementary Table iii. All included studies were of “doubtful” or “inadequate” methodological quality except for Patient’s Knee Implant Performance (PKIP),[45] which had an “adequate” rating for instrument development. Around half of PROMs were given an “indeterminate” rating for all three aspects of development procedures and lack of content validity studies, so content validity could be rated based only on the reviewers’ rating about the instruments themselves. Supplementary Table iv shows the methodological quality and rating for the remaining measurement properties for each included study.

Overall rating and quality of evidence for included PROMs

Table II presents the qualitatively summarized ratings for the measurement properties of the included instruments. None of the PROMs reported overall ratings for all nine measurement properties, due to no rating being given for cross-cultural validity. Ratings for measurement invariance and criterion validity were reported for three and four instruments, respectively. We also analyzed quantitatively eight instruments with > 2 available results in internal consistency, reliability, or construct validity (Table III). Most results coincided except for the test–retest reliability results of Intermittent and Constant Osteoarthritis Pain (ICOAP) and Function,[23] and the sports and recreation activities subscale in the KOOS—with better rating under quantitative analyses.

Table II.

Overall qualitative rating and quality of evidence for measurement properties of each instrument.

Instrument	Subscales	Content validity		Structural validity		Internal consistency		Cross-cultural validity		Measurement invariance
		Overall rating	Quality of evidence	Overall rating	Quality of evidence	Overall rating	Quality of evidence	Overall rating	Quality of evidence	Overall rating	Quality of evidence
AAOS Hip and Knee		±	Very low			?	Low
AKPS		+	Very low			?	Low
ASAP		±	Very low			?	Low
COMI-knee		+	Very low
FJS		+	Moderate	+	High	+	High
HAAS		+	Moderate			?	Moderate
HFKS		+	Very low
ICOAP	Constant pain	+	Very low			?	Moderate
	Intermittent pain	+	Very low			?	Moderate
KOOS	Pain	+	Moderate	+	High	+	High
	Function, daily living	+	Moderate	+	High	+	High
	Function, sports and recreational activities	+	Moderate	+	High	+	High
	Quality of life	+	Moderate	+	High	+	High
	Symptoms	+	Moderate			?	Moderate
KOOS, JR		+	Very low			?	Low
KOOS-PS		+	Low	+	Moderate	+	High
KOOS-12		+	Very low	+	Moderate	+	High			+	Low
KOS-ADLS		+	Very low			?	Low
KPQ		+	Very low
KSPQ		+	Very low
KSS (2011)		+	Very low	?	Moderate	?	Moderate			+	Low
KSS-A		+	Very low			?	Low
KSS short form		+	Very low	±	Moderate	?	Moderate
LEAS		+	Very low
LEFS		+	Very low	+	Low	+	Moderate
Lequesne		+	Very low			?	Very low
Lysholm		?				?	Very low
MFJS		+	Very low			?	Low
OA-CAT		?		+	Moderate	+	Moderate
OAKHQOL		+	Moderate
OKS		+	Moderate	+	High	+	High
OKS-APQ		+	Very low	+	Low	+	Moderate			+	Very low
PAQ-knee		+	Very low			?	Low
PKIP		+	Moderate	+	High	+	High
Tegner		±	Very low
UCLA		+	Very low
WOMAC	Pain, and Function	+	Very low	+	Low	+	High
	Stiffness	+	Very low			?	Moderate
WOMAC-TKR		+	Very low	+	Low	+	Moderate
WORQ		+	Moderate			?	Low
Instrument	Subscales	Reliability	Measurement error	Criterion validity	Construct validity	Responsiveness
		Overall rating	Quality of evidence	Overall rating	Quality of evidence	Overall rating	Quality of evidence	Overall rating	Quality of evidence	Overall rating	Quality of evidence
AAOS Hip and Knee								+	Moderate	?	Very low
AKPS								+	Moderate
ASAP								+	Low
COMI-knee		+	Very low	?	Very low			+	Moderate	+	Moderate
FJS		+	High	–	Low			+	High	?	Low
HAAS		+	Low	?	Low			+	Moderate
HFKS								+	Very low	+	Very low
ICOAP	Constant pain	±	Moderate	–	Very low			+	High	+	Very low
	Intermittent pain	–	Moderate	–	Very low			+	High	+	Very low
KOOS	Pain	+	Moderate	+	Low			±	High	+	Very low
	Function, daily living	+	Moderate	+	Very low			+	High	+	Very low
	Function, sports and recreational activities	±	Moderate	–	Moderate			+	High	+	Very low
	Quality of life	+	Moderate	–	Very low			+	High	+	Very low
	Symptoms	+	Moderate	–	Low			+	High	+	Very low
KOOS, JR				+	Very low	+	Very low	+	Moderate	+	High
KOOS-PS		±	Moderate					+	High	?	Low
KOOS-12						+	Very low	+	Moderate	?	Very low
KOS-ADLS		+	Low	?	Low			+	High	+	High
KPQ		–	Very low
KSPQ								+	Moderate
KSS (2011)		+	High					+	Moderate	+	Low
KSS-A								+	Moderate	?	Very low
KSS short form						+	Very low	+	Moderate	?	Very low
LEAS		+	Very low					–	Moderate	–	Moderate
LEFS		+	Very low					+	Moderate	?	Very low
Lequesne		–	Low							?	Very low
Lysholm		+	Very low	?	Very low			+	Very low
MFJS		+	Very low					+	Moderate
OA-CAT										?	Very low
OAKHQOL		–	Low					–	Low	?	Low
OKS		+	High	+	Moderate			+	High	±	High
OKS-APQ		+	Very low					+	Low	?	Very low
PAQ-knee		+	Low					+	Low	+	Low
PKIP		+	Very low					+	Moderate	+	Moderate
Tegner		+	Very low	?	Very low			+	Moderate
UCLA		+	Low	?	Very low			+	Moderate	?	Low
WOMAC	Pain, and Function	+	High	+	Low			+	Moderate	+	Low
	Stiffness	+	High	–	Low			+	Moderate	+	Low
WOMAC-TKR						+	Very low			?	Very low
WORQ		+	Moderate	+	Moderate			+	Moderate	?	Very low

sufficient; –, insufficient; ±, inconsistent; ?, indeterminate; AAOS Hip and Knee, AAOS Hip and Knee Core Scale; AKPS, Anterior Knee Pain Scale; ASAP, Activity Scale for Arthroplasty Patients; COMI-knee, Core Outcome Measures Index knee; FJS, Forgotten Joint Score; HAAS, High-Activity Arthroplasty Score; HFKS, High-Flexion Knee Score; ICOAP, Intermittent and Constant Osteoarthritis Pain; KOOS, Knee Injury and Osteoarthritis Outcome Score; KOOS, JR, Knee Injury and Osteoarthritis Outcome Score Joint Arthroplasty; KOOS-PS, Knee Injury and Osteoarthritis Outcome Physical function Short form; KOOS-12, 12-item short forms Knee Injury and Osteoarthritis Outcome; KOS-ADLS, Activities of Daily Living Scale of the Knee Outcome Survey; KPQ, Knee Pain Questionnaire; KSPQ, Knee Surgery Perception Questionnaire; KSS (2011), 2011 Knee Society Scoring System; KSS-A, Adjusted 2011 Knee Society Scoring System; KSS short form, Knee Society Scoring System short form; LEAS, Lower-Extremity Activity Scale; LEFS, Lower Limb Functional Scale; Lequesne, Lequesne Algofunctional Index for the Knee; Lysholm, Lysholm Knee Scoring Scale; MFJS, Modified Forgotten Joint Score; OA-CAT, Computer-Adaptive Test for Hip and Knee OA; OAKHQOL, OsteoArthritis of Knee and Hip Quality of Life Scale; OKS, Oxford Knee Score; OKS-APQ, Oxford Knee Score–Activity and Participation Questionnaire; PAQ-knee, Patient Administered Questionnaires Knee; PKIP, Patient’s Knee Implant Performance; Tegner, Tegner Activity Scale; UCLA, University of California Los Angeles activity score; WOMAC, Western Ontario and McMaster Universities Arthritis Index; WOMAC-TKR, Western Ontario and McMaster Universities Arthritis Index-Total Knee Replacement function short form; WORQ, Work, Osteoarthritis or joint-Replacement Questionnaire.

Table III.

Quantitative summarizing measurement properties for eligible instruments.

Instruments	Subscales	Pooled Cronbach’s αMean (Range)	Pooled ICCMean (95% CI)	Pooled Pearson correlation coefficientMean (95% CI)[*]
				Bodily pain	Physical function	Role emotional	Mental health	PCS	MCS
FJS		0.94 (0.81, 0.97)	0.93 (0.92, 0.94)
ICOAP	Constant pain	0.90 (0.76, 0.97)	0.76 (0.71, 0.81)
	Intermittent pain	0.88 (0.81, 0.93)	0.73 (0.67, 0.78)
KOOS	Pain	0.88 (0.65, 0.92)	0.88 (0.86, 0.90)	0.63 (0.63, 0.64)	0.49 (0.49, 0.49)	0.17 (0.16, 0.18)	0.27 (0.27, 0.28)
	Function, daily living	0.94 (0.80, 0.96)	0.85 (0.82, 0.88)	0.62 (0.61, 0.62)	0.60 (0.60, 0.60)	0.29 (0.28, 0.30)	0.32 (0.32, 0.32)
	Function, sports and recreational activities	0.88 (0.60, 0.98)	0.74 (0.70, 0.78)	0.38 (0.37, 0.38)	0.43 (0.43, 0.44)	0.12 (0.10, 0.13)	0.16 (0.16, 0.16)
	Quality of life	0.82 (0.71, 0.92)	0.81 (0.78, 0.84)	0.53 (0.53, 0.54)	0.48 (0.47, 0.48)	0.19 (0.19, 0.20)	0.28 (0.28, 0.28)
	Symptoms	0.75 (0.56, 0.93)	0.85 (0.83, 0.88)	0.45 (0.45, 0.45)	0.37 (0.37, 0.38)	0.17 (0.17, 0.18)	0.24 (0.24, 0.24)
KOOS-PS		0.87 (0.79, 0.91)
KOOS-12	Pain	0.78 (0.75, 0.84)
	Function	0.79 (0.78, 0.82)
	Quality of life	0.82 (0.80, 0.84)
	Summary score	0.91 (0.90, 0.93)
KSS (2011)	Symptoms	0.80 (0.31, 0.96)	0.89 (0.86, 0.91)	0.35 (0.32, 0.39)	0.24 (0.21, 0.26)	0.15 (0.13, 0.16)	0.17 (0.15, 0.19)	0.36 (0.34, 0.38)	0.19 (0.18, 0.20)
	Patient satisfaction	0.88 (0.79, 0.95)	0.88 (0.85, 0.90)	0.49 (0.47, 0.51)	0.32 (0.30, 0.33)	0.24 (0.22, 0.25)	0.29 (0.27, 0.30)	0.33 (0.31, 0.36)	0.30 (0.29, 0.30)
	Patient expectations	0.88 (0.73, 1.00)	0.85 (0.82, 0.87)	0.08 (0.06, 0.11)	0.13 (0.11, 0.16)	0.12 (0.12, 0.13)	0.10 (0.08, 0.11)	0.05 (0.03, 0.08)	0.10 (0.09, 0.11)
	Walking and standing	0.80 (0.68, 0.96)	0.84 (0.80, 0.87)
	Standard activities	0.88 (0.83, 0.94)	0.86 (0.82, 0.89)
	Advanced activities	0.84 (0.73, 0.93)	0.84 (0.80, 0.87)
	Discretionary knee activities	0.79 (0.51, 0.94)	0.78 (0.72, 0.82)
	Functional activities	0.90 (0.80, 0.93)	0.92 (0.91, 0.94)	0.47 (0.46, 0.49)	0.52 (0.50, 0.54)	0.28 (0.26, 0.29)	0.32 (0.30, 0.33)	0.57 (0.56, 0.59)	0.24 (0.22, 0.25)
OKS		0.92 (0.66, 0.94)	0.93 (0.92, 0.94)	0.62 (0.61, 0.63)	0.66 (0.66, 0.67)	0.32 (0.31, 0.33)	0.32 (0.31, 0.32)	0.65 (0.64, 0.66)	0.46 (0.45, 0.48)
WOMAC	Function	0.94 (0.82, 0.98)	0.91 (0.89, 0.93)	-0.56 (-0.56, –0.55)	-0.57 (-0.57, –0.57)	-0.30 (-0.31, –0.30)	-0.26 (-0.27, –0.26)
	Pain	0.83 (0.67, 0.91)	0.88 (0.86, 0.91)	-0.57 (-0.57, –0.56)	-0.47 (-0.48, –0.47)	-0.29 (-0.30, –0.29)	-0.28 (-0.28, –0.27)
	Stiffness	0.84 (0.70, 0.91)	0.87 (0.84, 0.90)	-0.47 (-0.47, –0.47)	-0.36 (-0.37, –0.36)	-0.28 (-0.28, –0.27)	-0.19 (-0.19, –0.19)

Pearson coefficients with 36-Item Short-Form Health Survey questionnaire.

KSS (2011), 2011 Knee Society Scoring System; CI, confidence interval; FJS, Forgotten Joint Score; ICOAP, Intermittent and Constant Osteoarthritis Pain; KOOS, Knee Injury and Osteoarthritis Outcome Score; KOOS-12, 12-item short forms Knee Injury and Osteoarthritis Outcome; KOOS-PS, Knee Injury and Osteoarthritis Outcome Physical function Short form; MCS, Mental compartment summary score; OKS, Oxford Knee Score; PCS, Physical compartment summary score; WOMAC, Western Ontario and McMaster Universities Arthritis Index.

Recommendation for selection of PROMs

Eight instruments with “sufficient” ratings for the three indispensable measurement properties were placed in category A, fulfilling COSMIN standard for use as TKA outcome measures: OKS, Lower Limb Functional Scale (LEFS), KOOS Physical function Short form (KOOS-PS), FJS, WOMAC-Total Knee Replacement function short form (WOMAC-TKR), Oxford Knee Score–Activity and Participation Questionnaire (OKS-APQ), PKIP, and 12-item short form KOOS (KOOS-12).

In addition, the pain and function subscales in WOMAC, as well as the pain, function, and quality of life subscales in KOOS, were recommended for use as standalone subscales instead of whole instrument due to the “insufficient” internal consistency of the other subscales.

University of California Los Angeles (UCLA) activity score and Lower-Extremity Activity Scale (LEAS)— one-item questionnaire with sufficient content validity—also met the standard. However, we placed LEAS in category B because of “insufficient” construct validity and responsiveness.

The remaining scales were all placed in category B and required further validation studies. No included PROM was in category C.

Feasibility and interpretability

Descriptive data on feasibility and interpretability are shown in Supplementary Table v. For the recommended PROMs stated above, OKS, function subscale in WOMAC, and the function, sports and recreational activities subscale in KOOS had studies reporting > 10% items missing. Studies of ceiling effects—generally defined as > 15% of the respondents achieving the highest possible score—revealed that such effects emerged 6 to 12 months postoperatively for WOMAC, KOOS, and their adapted versions, despite the considerable variation in results between studies.

Discussion

Our review systematically summarized 155 articles evaluating the measurement properties of 34 PROMs according to the latest COSMIN methodology. Nine PROMs (OKS, LEFS, KOOS-PS, FJS, WOMAC-TKR, OKS-APQ, PKIP, KOOS-12, and UCLA) met the COSMIN standard for recommendation of use for assessing TKA outcomes. WOMAC and KOOS were recommended for use as separate subscales, rather than a total score. The other 23 instruments, including the stiffness subscale in WOMAC, and the symptoms subscale in KOOS, had the potential for use but would need further validation studies.

The strengths of our review were its size and use of the latest COSMIN methodology. We identified 34 specific PROMs in the TKA population, which exceeded the 13 scales in the review by Harris et al (five knee scores and eight lower-limb scores).[11] Thus, with new articles published in the last five years, it is no wonder that we identified more psychometrically validated PROMs than those in OKS and OKS-APQ as revealed by Harris et al.[11] With regard to WOMAC, our results concurred with the supporting measurement properties identified by Harris et al[11] for pain and function subscales, but without evidence for sufficient structure validity for the stiffness subscale. Thus, we recommended to use pain and function subscales solely.

We also identified 68 additional articles that were not included in the review by Gagnier et al.[10] Of these, 31 documented development or assessed the psychometric properties of 11 additional PROMs (Core Outcome Measures Index Knee,[19] High-Flexion Knee Score,[22] KOOS, Joint Replacement,[26,52-54] KOOS-12,[28,55] 2011 KSS,[56-64] Adjusted 2011 KSS,[33] KSS short form,[34] LEFS,[36,65,66] Modified FJS,[39] Computer-Adaptive Test for Hip and Knee OA,[40,67,68] and UCLA[40,47,66,69-72]). Eight instruments were not selected for our review: the British Orthopaedic Association Score and Original KSS were not completely patient-reported; the Physical Activity Scale for the Elderly questionnaire, Self-Efficacy for Rehabilitation outcome scale, and Short Musculoskeletal Function Assessment Questionnaire did not aim to measure TKA outcomes; Arthritis Impact Measurement Scales, Japanese Knee Osteoarthritis Measure, and the Lower Limb Activity Profile were not evaluated for their psychometric properties in a specific TKA population. We concurred with the review by Harris et al[11] as they considered Musculoskeletal Outcomes Data Evaluation and Management System and AAOS Hip and Knee Questionnaire as the same instrument, and combined their results. However, our findings are completely different from those in a systematic review by Gagnier et al[10] in that they concluded Work, Osteoarthritis or joint-Replacement Questionnaire (WORQ) to be a promising instrument. This change could be explained by the addition of recently published literature in our study, especially on those evaluating structure validity (one of the determinants for recommendation),[28,65,73-75] as well as modifications in methodological instruments. The updated COSMIN methodology established a standard for instrument selection instead of counting positive ratings for properties.

Measurement properties related to recommendation formulation

Content validity and internal consistency were related to the quality of items and internal structure of instruments. Thus, sufficient ratings of these properties constituted the basic requirement for instrument selection.[14]

Content validity

The rating for content validity of a PROM was given based on the information on development study, content validity studies, and the instrument itself.

As for PROMs development, 11/34 included development studies that provided a clear description of representative patient involvement in elicitation of items, which was in concordance with a previous study which revealed that more than one-quarter of the development procedures lacked patient involvement.[76] Approximately one-half of 11 PROMs were developed based on a TKA population, which would degrade the quality of evidence for the other half of instruments because of indirectness. PROMs were developed to reflect disease effects on patients from their perspective. Items failed to achieve this purpose if they were generated by physicians or focus groups without a targeted population. Thus, individual interviews with specific patients was the best method to enable the relevance of items.[77] In addition, no study (except the development study for PKIP) undertook and described clearly interviews with target populations about comprehensiveness and comprehensibility of PROMs.[45] This absence led to an overall “insufficient” methodological quality and “indeterminate” rating for these two aspects.

Less than half of included instruments had available content-validity studies. Of the 30 content-validity studies, five studies asked patients[78-82] and two asked experts[83,84] about relevance, comprehensiveness, or comprehensibility. The remaining studies (> 70%) were pilot studies for comprehensibility within cross-culture adaption, which was a high proportion.

Internal consistency

Cronbach’s α was calculated commonly to evaluate internal consistency. However, as a prerequisite for interpreting internal consistency,[85] “sufficient” structure validity was graded only for eight instruments and six subscales in two multidimensional scales. Six studies conducted confirmatory factor analysis—regarded as the best method—to confirm the unidimensional structure of OKS, as well as the pain, function, quality of life subscales of KOOS, FJS, OKS-APQ, and PKIP.[28,43,73-86,86] Four studies undertook exploratory factor analysis (a less robust but adequate method),[32,65,87,88] but only one of them provided complete results and found the unidimensionality of LEFS.[65] We provided a “sufficient” rating for structure validity but downgraded the quality of evidence for WOMAC-TKA, KOOS-12, and KOOS-PS, as well as the pain and function subscales of WOMAC because all of their items were included in the validated subscales in KOOS.

Reliability and measurement error

Reliability and measurement error were assessed based on test–retest designs, but reliability (i.e. ICC) was more commonly calculated than measurement error. With respect to the methodological quality of test–retest designs, nearly three-quarters of studies had a “doubtful” or “inadequate” rating due to inappropriate time intervals and/or unclear test conditions. Time intervals between the two tests should be neither too short nor too long to avoid a recall bias and changes in the patient’s state, respectively. A time interval of two weeks is, in general, considered appropriate, albeit this is not standardized.[89]

Hypothesis for construct validity and responsiveness

The latest COSMIN methodology deletes all standards for formulating hypotheses, and is recommended to set or adopt hypotheses by the review teams themselves.[14] This strategy led to more evidence on hypotheses for construct validity (30/34 of included instruments) and responsiveness (26/34 of included instruments). More than 80% of included studies measured responsiveness by calculating the effect size rather than a correlation coefficient. The effect size measures the magnitude of change, but gives little information about the ability of the instrument to detect changes over time (i.e. responsiveness).[89] We rated these results as “indeterminate” because we could not formulate a hypothesis without knowing the true change.

In recent decades, conventional PROMs (e.g. WOMAC, KOOS, OKS) have been blamed for their postoperative ceiling effects (15% of the respondents achieving the highest possible score). We found that ceiling effects started to occur in WOMAC, KOOS, and their adapted versions six to 12 months after TKA (especially for pain subscales). Obvious ceiling effects were not revealed for OKS or other PROMs under this definition.

The ceiling effect is crucial because it prevents detection of further improvements in patients who have reached the highest score, which influences the discriminative power of the instrument. Improvements in surgical methods and changes in TKA recipients (e.g. increase in the number of younger patients with higher functional demands and expectations)[90] require expansion of the threshold for the best possible state defined 20 to 30 years previously.

As PROMs, scores for ceiling effects are thought to reflect patients’ perception of their health status rather than the ability to discriminate changes.[91] The goal of TKA is to achieve freedom from pain and functional satisfaction. In addition, many patients do not experience pain postoperatively, which leads to a skewed distribution of postoperative scores.

Thus, ceiling effects have a limited impact on the use of all recommended PROMs in TKA recipients overall. However, for long-term follow-up studies (or younger patients with high activity demands), instruments with a specific construct (e.g. “forgotten artificial joints”)[39] or developed for a specific population (e.g. OKS-APQ)[43] can be implemented.

Current trends

Recently, researchers have shown an increased interest in PROMs as an outcome measure for TKA. In our systematic review, though the first included PROM was introduced in 1982,[17] half of the PROMs and more than 60% of studies included were developed within the last decade. A systematic review also showed an increase in use of PROMs in TKA studies.[92]

However, the most frequently evaluated and used PROMs remain OKS, WOMAC, and KOOS, which were developed more than 20 years ago.[92] However, except for OKS, other instruments were not of sufficient quality. Also, sufficient internal consistency for subscales in KOOS and WOMAC was not demonstrated until publication of a study in 2019,[52] which highlighted the need for recommending PROMs of sufficient measurement properties to use in clinical practice and research.

In our review, no instrument was evaluated for all nine measurement properties. Also, no study of high methodological quality assessed content validity or cross-culture validity. To this extent, the recommended PROMs only met the minimal requirements for psychometric validity and still require further validation studies. Further studies are still warranted to evaluate existing PROMs, especially for a study evaluating content validity of included PROMs by interviewing patients and experts because content validity studies are the only variable determinants for this measurement property.

Additionally, we used qualitative and quantitative methods to synthesize ratings for measurement properties, and quantitative methods produced better ratings. Often, a quantitatively pooled method has a better ability to detect subtle changes than a qualitatively summarized method. Thus, we suggest using both qualitative and quantitative methods when conducting COSMIN review in the future.

Our systematic review had three main limitations. First, although we undertook exhaustive research unlikely missing any major trials, omissions might have occurred. Thus, we searched all the references of included studies manually, and included all the relevant articles. Second, we only included studies evaluating measurement properties in a TKA population. Thus, the results tested on populations combining TKA, total hip arthroplasty, or other patients were eliminated, which contributed to some PROMs and studies in previous systematic reviews not being included in our study. We believe that this strict inclusion criterion could increase the accuracy of our results in TKA patients. Finally, the recommendation formulated by our review does not necessarily mean the other 23 instruments are of “poor” quality. To some extent, they require further robust studies to evaluate their measurement properties.

In conclusion, nine PROMs and six subscales in two PROMs met the minimum requirements for psychometric validation and can be recommended for use as measures of TKA outcome. These are OKS, LEFS, KOOS-PS, FJS, WOMAC-TKR, OKS-APQ, PKIP, KOOS-12, UCLA, the pain and function subscales in WOMAC, and the pain, function, and quality-of-life subscales in KOOS. However, none of the included PROMs have been validated for all measurement properties. Thus, further studies are still warranted to evaluate those PROMs. Use of the other 25 scales and subscales should be tempered until further studies validate their measurement properties.