Patient acceptable symptom state and treatment failure threshold values for work productivity and activity Impairment and EQ-5D-5L in osteoarthritis

Patient-reported outcome measures (PROMs) have increasingly been advocated to assess treatment effect from patient’s perspective in the clinical setting [1]. However, interpreting and communicating numeric PROMs values in a clinically relevant manner can be challenging since these values may not correlate with a patient’s perceived improvement and well-being [2, 3]. To address this, minimal clinically important difference (MCID)–defined as the smallest change in PROM scores that patients perceive as beneficial–was introduced [4]. While MCID measures how much improvement is needed for patients to feel “better,” it doesn’t provide insights on whether patients are satisfied (feel “good”) about their current status [5]. In other words, a meaningful improvement in a PROM does not necessarily reflect a desirable state, especially if a patient was in a “terrible” state to begin with [6]. In response, concepts of “patient acceptable symptom state” (PASS) [7] and “treatment failure” (TF) [8] have been introduced. PASS is the threshold above which patients will consider themselves “well” and satisfied with treatment [5], while TF is the threshold below which patients consider their symptoms to be unsatisfactory to a degree that they consider the treatment has failed [8].

While previous studies established PASS [2, 3, 7, 9‐12] and TF [2, 10, 12] thresholds for PROMs assessing symptoms among patients with osteoarthritis (OA), less attention has been paid to generic PROMs measuring general health-related quality of life (HRQoL). Work Productivity and Activity Impairment: Specific Health Problem (WPAI:SHP) and EQ-5D-5L are two PROMs that are advocated for assessing HRQoL, including work and activity limitations in OA [13, 14]. A recent study reported that across five common instruments to measure work impairment, WPAI was the instrument preferred by participants [13]. EQ-5D is a simple self-administered questionnaire which is commonly used in the OA context and is the most commonly collected PROM in the Swedish National Quality Registers [15]. Although a few studies estimated the PASS thresholds for EQ-5D-3L in OA [16‐19], to our knowledge, only one recent study has estimated this for EQ-5D-5L in OA [20] and there is no reported PASS threshold for WPAI in OA or any other condition. Furthermore, while patient education, self-management, and exercises are recommended as core first-line treatments for all persons with OA, all previous (EQ-5D) PASS thresholds were estimated among people undergoing surgical treatment which is recommended as the last resort for a minority of people with severe signs and symptoms of OA [21]. This implies that current PASS thresholds might not be applicable for general OA population. More importantly, to our knowledge, no previous study has reported TF thresholds for either EQ-5D or WPAI in any population. Combination of PASS and TF thresholds can aid to determine the scores representing an acceptable or failed post-treatment outcome. To facilitate meaningful interpretation of the values reported for EQ-5D-5L and WPAI, the present study aimed to establish PASS and TF thresholds for these PROMs among participants of a digital first-line self-management program for OA.

This is a secondary analysis of register data obtained from consecutive participants of a digitally delivered self-management program for hip and knee OA, known as Joint Academy®, described in details elsewhere [22, 23]. In short, inspired by the Swedish first-line face-to-face management program for OA (known as “Better management of patients with OsteoArthritis” which exists as a National Quality Register), the digital program was introduced in Sweden in 2016 and is targeted toward exercise, physical activity, and education delivered by a smartphone application. It contains video lectures on OA, physical activity, and self-management as well as individualized exercises and a possibility to chat asynchronously with a physical therapist during the treatment. The program is covered by the national healthcare system in Sweden.

Of 16,640 eligible participants, we excluded 2007 (12.1%) individuals who did not respond to anchor questions at any follow-up. There were differences in the baseline characteristics of participants included and those excluded, with the latter being older, with a higher proportion of hip OA and having poorer PROMs scores than those included (Table 1). A total of 14,633 individual aged 24–94 years with mean (SD) age 64.1 (9.1) years and 75.5% females were included. Of included participants, 14,607 and 2707 provided 3- and 12-month responses, respectively. It should be noted that the smaller sample size at 12-month follow-up was mainly due to the study time-frame. That is, most participants didn’t reach their 12-month follow-up when data were extracted in January 2022 (e.g., 9199 individuals enrolled between February and September 2021). For participants included in the study, the mean (SD) hypothetical and experience-based EQ-5D-5L index scores were 0.84 (0.17) and 0.82 (0.11) at baseline, respectively. The corresponding figures for WPAI–work and –activity impairments were 24.3 (24.8) and 39.3 (23.7). The correlation coefficients between PROMs and anchor questions were generally ≥ 0.35 with higher values at 12- than 3-month follow-up (Table 2).

At 3-month follow-up, 6128 (42.0%) participants reported their current state as satisfactory, while 613 (4.2%) considered the treatment had failed (Fig. 1). Corresponding proportions at 12-month follow-up were 48.2% and 2.8%, respectively. Participants with a satisfactory symptom state reported better PROMs scores (i.e., higher EQ-5D-5L and lower work/activity impairments) than others in both follow-ups (Fig. 2). The PASS thresholds for the hypothetical EQ-5D-5L were 0.92 (95% CI 0.91, 0.92) and 0.91 (0.91, 0.921) at 3- and 12-month follow-ups (Table 3). Corresponding figures were 0.87 (95% CI 0.87, 0.87) and 0.87 (0.86, 0.87) for the experience-based EQ-5D-5L at these time points. The TF thresholds for hypothetical/experience-based EQ-5D-5L were 0.77 (95% CI 0.76, 0.78)/0.77 (0.76, 0.77) at 3-month and 0.75 (0.73, 0.77)/0.75 (0.73, 0.77) at 12-month follow-ups (Table 3). The PASS thresholds for WPAI–work and WPAI–activity impairments were 16 (95% CI 15, 16) and 26 (26, 26), respectively, at 3-month follow-up (Table 4). Similar PASS thresholds were estimated at 12-month follow-up. The TF thresholds for work impairment were 29 (95% CI 28, 31) and 33 (29, 39) at 3- and 12-month follow-ups, respectively. Corresponding figures for activity impairment were 50 (95% CI 49, 51) and 49 (46, 52), respectively (Table 4).

Our subgroup analysis showed that the PASS/TF thresholds of EQ-5D-5L for participants with severe pain were 0.03 to 0.04/0.10 to 0.15 units lower than those with mild pain at baseline (Table 3). For WPAI–work impairment, participants with severe pain had 8–10/15–19 points higher PASS/TF thresholds compared with those with mild pain. Corresponding differences for the WPAI–activity impairment ranged between 8 and 12 for the PASS thresholds and between 16 and 19 for the TF thresholds. Our complete cases analysis (n = 2681) showed that 40.5 and 48.2% of participants reported their current state as satisfactory at 3- and 12-month follow-ups, while 2.8% considered the treatment had failed at both follow-ups. We obtained almost identical PASS/TF thresholds among those with complete responses (Table A1 in appendix). Comparing knee vs. hip OA subgroups suggested that while the PASS thresholds were generally comparable between two groups, there were differences in the TF thresholds where hip OA patients tended to consider poorer PROMs as acceptable (Tables A2 and A3 in appendix). The TF/PASS thresholds were comparable for individuals aged ≤ 65 years and those older than 65 years (Table A4 in appendix).

In this study, we estimated the PASS and TF thresholds for EQ-5D-5L and WPAI among a large cohort of persons with knee or hip OA participating in a digital self-management program. Our results showed that at 3 and 12 months following participation in the digital program, around 42–48% of participants considered their current state as satisfactory, while 2–4% considered the treatment had failed. The PASS thresholds for EQ-5D-5L ranged between 0.87 and 0.92 and the TF thresholds ranged between 0.75 and 0.77, with higher PASS thresholds for the Swedish hypothetical than experience-based value set. The estimated PASS and TF thresholds for WPAI were 16–26 and 29–50, respectively, with higher thresholds for WPAI–activity than WPAI–work impairments. While we failed to detect any difference in our estimates across follow-up time, the baseline pain severity had significant effects on the estimated thresholds with those with more severe pain at baseline being prepared to accept poorer PROMs.

To our knowledge, only one previous study reported PASS thresholds, ranging from 0.68 to 0.85 for EQ-5D-5L among people undergoing total hip or knee replacement in Canada [20]. Consistent with our finding they also reported variation in the PASS thresholds according to the EQ-5D-5L value set. In the present study, the Swedish hypothetical value set was associated with 0.04- to 0.05-unit higher PASS thresholds compared with the experience-based EQ-5D-5L value set. This is in contrast with the results from Cooper et al. [32] reporting higher PASS thresholds for the Swedish experience-based than the UK hypothetical EQ-5D-3L value set among persons with chronic arthritic diseases. While, the narrower range and higher mean EQ-5D scores for experience based compared with hypothetical-based value sets are well-documented [33‐35], the mean Swedish experience-based EQ-5D-5L values are lower than the Swedish hypothetical-based values for mild health states [25]. Given that the most participants in our sample reported no to moderate problems across all dimensions of EQ-5D-5L, the mean hypothetical values were higher than the experience-based values. For instance, at 3-month follow-up, the proportions of participants with no to moderate problems ranged from 92.7% for pain/discomfort to 99.3% for self-care which resulted in 13,234 (90.6%) of participants being in a health state with larger hypothetical than experience-based EQ-5D-5L scores.

For both PROMs, we found that the estimated thresholds were stable over time. Previous studies have reported mixed findings on the time dependency of the thresholds [17, 18, 20]. Consistent with our finding, Connelly et al. [18] and Giesinger et al. [36] reported time-constant thresholds for EQ-5D-3L. Conner‑Spady et al. [20] reported time-dependent PASS thresholds for EQ-5D-5L among people undergoing total knee replacement when using Canadian value set, while time-constant thresholds were reported when the EQ-5D-5L scores were calculated using crosswalk. Naal et al. [17] reported time-dependent PASS thresholds for EQ-5D-3L among persons with total hip arthroplasty and time-constant PASS threshold among those with total knee arthroplasty.

Our results showed that persons with more severe pain at baseline were willing to accept poorer (more severe) PROMs after participation in the digital program. While the baseline dependency of PASS/TF thresholds for EQ-5D-5L and WPAI has not previously been explored, the baseline dependency of PASS/TF thresholds for other PROMs is well documented [2, 9, 37]. This baseline dependency calls for considering the comparability of population’s characteristics when using the PASS/TF thresholds reported in the present study in other populations. In other words, while the estimated PASS/TF thresholds in the total sample can be applied in the population with similar baseline characteristics (e.g., NRS pain 5, the Swedish hypothetical/experience-based EQ-5D-5L score 0.84/0.82 and work/activity impairments 24/39), for populations with milder/more severe symptoms, the threshold values from our subgroup analysis should be used. Albeit, other factors such as type of intervention and co-existing conditions might also influence the thresholds and hence application in other population should be done with caution.

The observed tendency among participants with hip OA, compared to knee OA, to accept poorer PROMs might be due to a higher proportion of respondents considering the treatment failed in the former group (4.9% vs. 3.8% at 3 months and 3.9% vs. 2.2% at 12 months). Moreover, participants with hip OA who considered the treatment failed had poorer baseline health status than their counterparts with knee OA. For instance, while the EQ-5D-5L scores were comparable for participants with knee and hip OA who responded “no” to the treatment failure question (0.83 vs. 0.82 for the experience-based values and 0.85 vs. 0.84 for the hypothetical values), there were larger differences among those who considered the treatment failed (0.78 vs 0.73 for the experience-based values and 0.77 vs. 0.72 for the hypothetical values). Limited variations in the estimated thresholds across age groups were consistent with previous results on the EQ-5D index score PASS thresholds [16, 19]. This finding implies that the estimated PASS/TF thresholds can be applied across different age subgroups.

Estimating the first TF threshold for EQ-5D-5L and first TF/PASS thresholds for WPAI, a large sample of persons with knee or hip OA participating in a digital first-line treatment, and the use of a less biased and more precise approach to estimate the thresholds are the main strengths of the present study. Using a dichotomized anchor question reflecting patients’ own judgment on satisfaction with their symptoms was another strength of the current study [18]. However, several limitations of the study should be considered when interpreting the findings. While the correlation between anchor questions and PROMs were generally acceptable confirming their validity, the anchor-PROM correlation for the WPAI–work activity and TF anchor question at 3-month follow-up was inadequate (< 0.30) [38] which calls for caution in interpreting this threshold. The rate of satisfaction relies on the focus of the anchor question [39]. In the present study, the PASS/TF anchor questions focused on participants’ satisfaction with their current state of knee/hip symptoms and functions which are different from the focus of EQ-5D-5L (overall health-related quality of life) and WPAI (work and activity). This might have influenced the correlations and accuracy of our estimates, particularly for work impairment which has less overlap with symptoms/functions than health-related quality of life and daily activity. Although the anchor questions used in the current study are widely used to estimate PASS/TF thresholds [6, 8, 10], there is no gold standard to capture patients’ satisfaction and alternative anchor questions with different wording might result in different thresholds. The study cohort included individuals with self-reported OA who self-selected to participate in a digital program. These individuals are different from those participating in the face-to-face OA core treatment [40] as well as from those identified in routine practice [41], particularly with higher proportion of females and high educated people in the digital program. Most individuals (91%) participated in the program during the COVID-19 pandemic which could influence their health status and responses to the anchor questions and RROMs, especially 3993 (24%) individuals participating during February–December 2020 prior to initiation of COVID-19 vaccination in Sweden and hence limit the generalizability of our findings. There were some differences between those included in the analysis and those excluded because of missing responses. These might limit the generalizability of our findings. WPAI captures work impairments among people who are employed and hence the findings are not applicable to individuals who lost their jobs due to OA.

This study provides the first PASS and TF thresholds for EQ-5D-5L and WPAI among persons undergoing a digital first-line treatment for OA. These thresholds might facilitate meaningful interpretation of these PROMs among people with knee or hip OA. Our results suggest that the EQ-5D-5L and WPAI PASS/TF thresholds were stable over time and hence can be applied across different time points after first-line treatments for OA. However, observed variations by value set (for EQ-5D-5L) and baseline pain intensity might limit their generalizability and hence should be applied with great caution in other settings/populations.