Osteoarthritis (OA) is the most frequent joint disease, characterized by progressive articular cartilage loss that results in joint pain and functional impairment, which impacts the ability to perform daily-life activities. The prevalence of this type of disease is very high affecting 4% of the general population worldwide based on radiological diagnosis, and up to 20% in the case of specific population groups, such as women over 60 years [1]. Knee OA is a heavy burden for patients and the society as a whole. International studies have estimated that knee and hip OA constituted 0.7% of all disability adjusted life years (DALY) lost in 2010, a 40% increase with respect to 1990 [1]. Eighty-three percent of DALY lost due to OA are due to OA of the knee [2].

Knee OA entails a substantial impact on health related quality of life (HRQL) [3, 4]. HRQL is generally considered to incorporate the evaluation of functioning status as well as the patient’s perception of their emotional functioning and social role. Since patients’ responses vary greatly in the face of identical stressors, such as pain, HRQL is a crucial outcome measure [5]. The dimensions of HRQL most affected by knee OA are those related to physical activity and self-efficiency [6]. It seems that knee OA has a greater impact on the physical aspects of HRQL in the case of women, whereas men report worse scores on psychological-related scales [7]. Besides, HRQL predicts future inpatient and outpatient health care utilization and mortality in patients diagnosed of OA [8]. Therefore, measures of HRQL are important not only for assessing the burden of the disease or the results of any intervention, but also for helping informed decision-making in the allocation of often limited health resources [4].

In the case of knee OA, there are several specific tools to measure HRQL, some of which have been adapted and validated for the Spanish setting, such as the Western Ontario and McMaster Universities Osteoarthritis Index (WOMAC) –a useful questionnaire for the assessment of OA of the lower limb [9, 10]–, the “Knee Society Clinical Rating System” (KSS) [11, 12], or the Knee Injury and Osteoarthritis Outcome Score (KOOS) [13]. Other questionnaires like the “Oxford Knee Score” (OKS) have not had their psychometric properties validated for our setting. The OKS is a brief, 12-item, self-reported scale developed to measure the impact of total knee replacement surgery on the perception of HRQL by patients [14], and its scores and outcome interpretations have been slightly modified throughout the years it has been in use [15]. It is reported to be amongst the most sensitive, responsive, reliable, and valid patient-reported questionnaire for knee conditions [16]. It has been adapted and validated into Italian [17], Dutch [18], Chinese and Singapore English [19], German [20], French [21], Japanese [22], Portuguese [23], Korean [24], Persian [25], Greek [26], Spanish in Colombia [27], Arabic language [28], and Finnish language [29]. Owing to its good psychometric properties, it has been favorably compared to other widely used tools in different languages that are more difficult to administer [30, 31]. Although the OKS has been adapted to Spanish for Spain, its psychometric properties have not been assessed in the Spanish population setting. As far as we know, only the Dutch and Finnish language adaptations of OKS have been validated in a prospective manner, similarly to the original work by Dawson et al. [18, 29], whereas its factor structure has not been confirmed in any of its adapted versions.

In Spain, knee OA implies an enormous burden of illness for the people who suffer from it and for the whole society [32], and is therefore worthy of being measured. There are new instruments that serve this purpose, such as the OKS; however, for a questionnaire to be useful in culturally different areas with different languages, it must not only be translated into the new language but also adapted to account for any different or new cultural characteristics. The adaption must then be validated as the original version was. This work tackles the study of the psychometric properties of the OKS in its Spanish-adapted version, including its reliability, validity, and responsiveness.

A total of 397 patients were included: 158 in Bizkaia, 158 in Madrid, and 81 in Tenerife. Of them, 36.8% were recruited at primary care, 55.2% at traumatology, and 8.0% at rheumatology consultations. The mean time elapsed since diagnosis was 61.6 months (CI 95%: 55.6–67.6 months). Women comprised 69.8% (CI 95%: 65.3–74.3%) of the sample, with an average age of 71.4 years (CI 95%: 70.5–72.3 years).

In terms of the knee affected by OA, in 27.7% (CI 95%: 23.3–32.1%) of cases it was the right knee, in 30.0% (CI 95%: 25.5–34.5%) the left knee, and in 43.3% (CI 95%: 38.4–48.2%) both knees.

Total knee replacement surgery had been previously performed in 18.1% (CI 95%: 14.3–21.9%) of cases. The average Charlson’s comorbidity index was 0.8 points (CI 95%: 0.7–0.9), and mean BMI was 29.7 (IC 95%: 29.2–30.2).

Validity

With regards to the validity of the construct, a unidimensional structure was found in the EFA with a single factor that explained 55.5% of variance (KMO = 0.946, Bartlett’s test of sphericity χ² = 2597, 66 degrees of freedom, p < 0.001). All factor loadings were >0.50, and commonalities were >0.40 except for questions 4 and 8 (Table 2).

	EFA 1 factor
Variable	Factor loading	Commonality
OKS 1	0.761	0.579
OKS 2	0.751	0.564
OKS 3	0.774	0.599
OKS 4	0.572	0.327
OKS 5	0.796	0.634
OKS 6	0.773	0.597
OKS 7	0.670	0.449
OKS 8	0.605	0.365
OKS 9	0.839	0.705
OKS 10	0.767	0.589
OKS 11	0.788	0.620
OKS 12	0.794	0.630

EFA Exploratory factor analysis

OKS Oxford Knee Score

The results of the CFA (Fig. 1) showed excellent fit indices: (a) the RMSEA was 0.076, that is <0.08; and (b) the CFI and TLI were 0.981 and 0.977, respectively, exceeding the benchmark of 0.95. Factor loadings were all statistically significant (P < 0.001), ranging from 0.58 to 0.86 (Fig. 1).

The validity of known groups, which measures the discriminatory capacity of the questionnaire, can be checked in Table 3, where mean scores and 95% CI of the OKS are shown for the various terciles of the WOMAC and EQ-5D-5 L scales distribution. Differences between the three groups are clearly observed in the OKS scores, with average changes of 5.6 and 11.9 points per tercile.

	OKS score for the lower tercile of the questionnaire distribution (CI 95%)	OKS score for the middle tercile of the questionnaire distribution (CI 95%)	OKS score for the top tercile of the questionnaire distribution (CI 95%)
WOMAC pain	31.4 (30.0–32.7)	20.5 (19.5–21.6)	13.8 (12.7–14.9)
WOMAC rigidity	29.7 (28.2–31.3)	22.1 (20.8–23.4)	14.4 (13.3–15.4)
WOMAC impairment	32.0 (30.8–33.3)	20.1 (19.1–21.0)	13.1 (12.1–14.0)
EQ-5D-5 L utility	13.6 (12.5–14.6)	21.1 (20.0–22.2)	31.1 (29.7–32.4)
EQ-5D-5 L VAS	15.7 (14.0–17.3)	21.5 (20.3–22.7)	27.1 (25.3–28.9)

OKS Oxford Knee Score

WOMAC Western Ontario and McMaster Universities Osteoarthritis Index

VAS Visual Analogue Scale

Table 4 shows the correlations between the OKS scores and WOMAC domains or EQ-5D-5 L VAS and utilities. Due to the different ways in which scores are presented on the scales, negative correlations with the WOMAC and positive ones with the EQ-5D-5 L were to be expected. All associations were strong except for rigidity on the WOMAC scale, whose correlation was at the limit of the required threshold, and the EQ-5D-5 L VAS.

	OKS	WOMAC pain	WOMAC rigidity	WOMAC impairment	EQ-5D-5 L Utility	EQ-5D-5 L VAS
OKS	_
WOMAC pain	−0.745*	_
WOMAC rigidity	−0.641*	0.643*	_
WOMAC impairment	−0.849*	0.799*	0.719*	_
EQ-5D-5 L utility	0.756*	−0.623*	−0.504*	−0.706*	_
EQ-5D-5 L VAS	0.530*	−0.472*	−0.361*	−0.516*	0.547*	_

*p < 0,001

OKS Oxford Knee Score

WOMAC Western Ontario and McMaster Universities Osteoarthritis Index

Higher scores indicate better health condition in the OKS and EQ-5D-5 L, and opposite in the case of the WOMAC

The Spanish version of the OKS questionnaire is a reliable, sensitive to changes, valid tool to measure HRQL in patients that suffer from knee OA. Given the extraordinarily high response rate, it also is a well accepted questionnaire.

The validity of the OKS was assessed from different perspectives, although its “apparent” validity has not been tested since it is an adaptation. Discriminatory or known-groups validity seems adequate since the outcome score differs greatly in subjects with very different scores on the WOMAC or the EQ-5D-5 L scales. Additionally, it does not show significant ceiling or floor effects that compromise such discriminatory capacity, as has been previously pointed out in other adaptations [17, 20, 21, 29].

The convergent validity of the tool seemed appropriate. Correlations of the OKS adapted-version with the specific scales of the WOMAC or the generic scales of the EQ-5D-5 L were stronger than those found between the original version and other generic tools that measure HRQL [14, 30]. In the case of adaptations of the OKS to other languages, like Portuguese [23] or German [20], these correlations were similar or slightly stronger. This way of measuring convergent validity offered better results for the OKS than those reported in other questionnaires like the Spanish version of the KSS [12].

Construct validity of the OKS was also studied. Its factorial structure sustains the unidimensionality of the questionnaire. In the EFA, all items were found to consistently saturate the same factor and showed higher values than the English version [14]. Acceptable values of RMSEA were obtained during the CFA, and TLI and CFI were optimal [44]. Although the possibility of disaggregating pain and impairment components from the OKS has been proposed, this unifactorial structure seems to be the most solid one [53], which is supported by this outcome.

Internal consistency was better than for the original scale in the inclusion period (Cronbach’s α =0.92 vs. 0.87) [14]. The test-retest reliability was very high and the obtained values, measured through the ICC, allowed to qualify the tool as reliable [40].

The discriminatory capacity of the questionnaire was adequate, which accounts for its ability to distinguish between individuals in different situations, but the tool can also be used to study the perception changes of a single person’s situation, which means that its evaluative capability is adequate [54]. In fact, this tool was designed for that purpose and the outcome of this study supports this type of use. The effect size of the change for “moderate” positive changes was similar to the WOMAC but slightly lower than the set benchmark of 0.8. In the validation assessment of the original version, the observed effect size of the change after surgery was 2.1 [14], which is only comparable to really substantial improvements (effect size =1.6) since it was tested in patients that had undergone knee replacement. The evaluative capacity was greater to detect positive than negative changes, as is the case with other questionnaires [55], although even in the case of negative changes the capacity of this tool is similar or higher than the WOMAC, and greater than the generic EQ-5D-5 L questionnaire.

The MCID was 6.1 in the case of subjects that had experienced moderate improvement. Values of MCID between 3 and 5 points had been proposed for the validation of the original version [15] and confirmed in subsequent studies [56], although these studies only included subjects that had undergone knee replacement surgery. The MCD_95% was 4.38, which represents the lowest score change (at the particular patient level) that is not the result of measurement error of the instrument, and can be understood as the lowest bound of real change, although it may not indicate clinical significance [50]. The ratio between the MCID and MDC_95% was higher than 1, indicating that the MCID can be discriminated clearly from measurement error.

This work has some limitations. The studied subjects may not be representative of the national population. Patients were included from different regions and at different stages of the disease evolution, although we did not record or classify the knee OA severity of each patient. Besides, there are intrinsic limitations to the methodology used to assess the psychometric characteristics (the classical test theory), with its assumptions and weaknesses, but the validation process has been complemented with a CFA specific for categorical data that scrutinizes such deductive assumptions using statistical analysis [42].

The outcome of this study allows for proposing the application of the OKS in those situations where the original version has been used, such as measuring HRQL improvement after knee replacement surgery [30] or studying surgery-related factors [57–59], but also to discriminate between patients in different clinical situations and to appreciate their evolution with time in view of its capacity to detect “moderate” improvements in patients.

The Spanish adaptation of the OKS questionnaire is a valid tool for assessing the perception of HRQL of patients suffering from knee OA. It is well accepted by patients and shows psychometric properties that support its usefulness both for the assessment of a patient’s condition and its subsequent evolution. Its comparative utility is quite similar to that of tools that have been extensively used after their adaptation, like the WOMAC questionnaire. The incorporation of this type of tools in usual clinical practice will allow for appraising, in a valid and reliable way, the patient’s self-perception of HRQL as well as the outcome of health interventions addressed at them.