Knee Surgery, Sports Traumatology, Arthroscopy

Revision rates differ across hospitals, yet patient-reported outcomes after primay knee
arthroplasty are the same: 1-year postoperative results from the prospective
multicenter cohort study, SPARK
--Manuscript Draft--

Manuscript Number: KSST-D-22-01396

Full Title: Revision rates differ across hospitals, yet patient-reported outcomes after primay knee
arthroplasty are the same: 1-year postoperative results from the prospective
multicenter cohort study, SPARK

Article Type: Original Paper

Abstract: Purpose
Wide variations in cumulative revision rates (CRRs) after primary knee arthroplasty are
well recognized. Still, whether hospital CRRs are good indicators of general surgical
quality, as defined by patients, is not clear.
Methods
This prospective study followed a cohort of primary knee arthroplasty patients (any
implant type) in three high-volume hospitals known to differ in CRRs, from
preoperatively (2016-18) to 1-year postoperatively with Oxford Knee Score (OKS),
Forgotten Joint Score, EQ-5D-5L, Copenhagen Knee ROM (range of motion) Scale,
UCLA Activity Scale, and questions concerning e.g., postoperative physiotherapy.
Also, results were compared across hospitals for patients with comparable symptoms
and radiological knee osteoarthritis.
Results
Of 1452 patients (56% inclusion), 97% responded postoperatively (90% at 1-year).
There was no hospital difference in 1-year OKS (39 ±7, p=0.1), nor when adjusted for
age, sex, BMI, baseline OKS and osteoarthritis grading. 15% of patients improved less
than Minimal Important Change (8 OKS points) without significant hospital difference
(p=0.051). Patients with equal osteoarthritis grading or baseline OKS had similar
results across hospitals (1-year OKS and willingness to repeat surgery, p>0.087)
except for the Kellgren-Lawrence grade-4 group of 64 patients (low-CRR hospital 4-6
OKS points lower). 92% of patients were willing to repeat surgery.
Conclusion
Hospitals with different revision rates had similar patient-reported outcomes after
primary knee arthroplasty in non-revised patients, supporting that revision rates are
incomplete measures of overall surgical quality. Future studies should explore if
revision rate variations may depend as much on revision thresholds and indications as
on results after primary surgery.

Keywords: knee arthroplasty; Knee replacement; epidemiology; patient-reported outcome

measures; revision rate; radiographic classification; osteoarthritis.

Powered by Editorial Manager® and ProduXion Manager® from Aries Systems Corporation
Blinded Manuscript Click here to view linked References

1 File information
2 Title: Revision rates differ across hospitals, yet patient-reported outcomes after primary knee arthroplasty are the
3 same: 1-year postoperative results from the prospective multicenter cohort study, SPARK.
4 Journal: Knee Surgery, Sports Traumatology, Arthroscopy
5 Corresponding author: anonymized.
6 Email: anonymized.

7 TITLE
8 Revision rates differ across hospitals, yet patient-reported outcomes
9 after primary knee arthroplasty are the same: 1-year postoperative
10 results from the prospective multicenter cohort study, SPARK

11 Anonymization
12 For anonymization purposes, all names that points to the origin of the study have been replaced by e.g.,
13 [Hosp. A], [Region A], [Country] and [Nationality], and references have been blinded.

14 ABSTRACT

15 Purpose

16 Wide variations in cumulative revision rates (CRRs) after primary knee arthroplasty are well recognized.

17 Still, whether hospital CRRs are good indicators of general surgical quality, as defined by patients, is not

18 clear.

19

20 Methods

21 This prospective study followed a cohort of primary knee arthroplasty patients (any implant type) in three

22 high-volume hospitals known to differ in CRRs, from preoperatively (2016-18) to 1-year postoperatively

23 with Oxford Knee Score (OKS), Forgotten Joint Score, EQ-5D-5L, Copenhagen Knee ROM (range of

24 motion) Scale, UCLA Activity Scale, and questions concerning e.g., postoperative physiotherapy. Also,

25 results were compared across hospitals for patients with comparable symptoms and radiological knee

26 osteoarthritis.

27

1
28 Results

29 Of 1452 patients (56% inclusion), 97% responded postoperatively (90% at 1-year). There was no hospital

30 difference in 1-year OKS (39 ±7, p=0.1), nor when adjusted for age, sex, BMI, baseline OKS and

31 osteoarthritis grading. 15% of patients improved less than Minimal Important Change (8 OKS points)

32 without significant hospital difference (p=0.051). Patients with equal osteoarthritis grading or baseline OKS

33 had similar results across hospitals (1-year OKS and willingness to repeat surgery, p>0.087) except for the

34 Kellgren-Lawrence grade-4 group of 64 patients (low-CRR hospital 4-6 OKS points lower). 92% of patients

35 were willing to repeat surgery.

36

37 Conclusion

38 Hospitals with different revision rates had similar patient-reported outcomes after primary knee arthroplasty

39 in non-revised patients, supporting that revision rates are incomplete measures of overall surgical quality.

40 Future studies should explore if revision rate variations may depend as much on revision thresholds and

41 indications as on results after primary surgery.

42

43 Level of evidence: Level II (Prospective comparative study).

44 Keywords

45 Knee arthroplasty, knee replacement, epidemiology, patient-reported outcome measures, revision rate,

46 radiographic classification, osteoarthritis.

47

2
48 MANUSCRIPT

49 Introduction

50 When regions or hospitals differ in cumulative revision rates (CRR) after primary knee arthroplasty (KA), it

51 leads to assumptions of differences in the quality of surgery. This is the case for variation both among and

52 within countries [21]. However, with pain relief and regaining knee function being the primary goals of KA

53 surgery, CRRs may not be the most important measure of treatment quality in the large majority of patients,

54 i.e., those, who are never revised and whose spectrum of postoperative results are not reflected in the

55 statistics [8]. This study was initiated to investigate whether [National] regional differences in CRRs from

56 1% to 5% per 2 years were a reflection of overall differences in the quality of surgery, defined as patients’

57 subjective improvement following operation [23]. By asking this question, the focus is shifted from registry-

58 assessed quality based on CCRs to quality based on patient assessment, patient-reported outcome measures

59 (PROMs) and range of motion.

60

61 The first publication from this prospective cohort study, the SPARK study (“Variation in patient Satisfaction,

62 Patient-reported outcome measures, radiographic signs of Arthritis, and Revision rates in Knee arthroplasty

63 patients in three [National] regions”) reported baseline data on 1452 patients in three high-volume university

64 hospitals that were known to differ in CRRs [1]. It revealed how patient demography, patient-reported

65 anxiety and depression, incidence of knee replacement, implant choice and radiographic classification of

66 knee osteoarthritis varied somewhat across the hospitals, while preoperative PROMs (main outcome Oxford

67 Knee Score (OKS)) did not. The majority of hospital differences at baseline were paradox to well-known

68 revision risk factors. For example, one low-revision hospital [Hosp. A] used more unicompartmental

69 implants (49%) than the others ([Hosp. B] 14% (low-CRR) and [Hosp. C] 22% (high-CRR)) (table 1).

70 Overall, the study was unable to identify baseline characteristic differences that could offer a meaningful

71 explanation of the persistent differences in CRRs across the three administrative regions, as represented by

72 one large-volume hospital each.

3
73

74 The present study investigated whether these hospital differences in registry-based quality (CRR) reflected

75 hospital variations in the quality of surgery, as experienced by patients. This was done by comparing

76 postoperative PROM results in the SPARK cohort during 1-year follow-up after primary KA. Also, we

77 determined if patients with a certain degree of radiological knee osteoarthritis (OA) or OKS, respectively,

78 reported comparable postoperative OKS result and “willingness to repeat surgery” across hospitals.

79 Patients and methods

80 From September 2016 and 16 months forward, patients scheduled for primary knee arthroplasty of any type

81 were invited to participate in this prospective cohort study in three high-volume centers across [Country].

82 The three hospitals had revision rates that were representative of their particular region; [Region A] and [B]

83 (represented by [Hosp. A] and [Hosp. B], respectively), had relatively low CRRs (e.g., 2.2% and 1.0% per 2

84 years in 2015) whereas the Capital Region ([Hosp. C] in [City]) had high CRRs (5.0% per 2 y. in 2015) [23].

85 Patients were followed with PROM-sets from preoperatively to 1-year postoperatively.

86

87 Participants

88 Patients scheduled for primary knee replacement of any type (total (TKA), medial (MUKA) or lateral

89 (LUKA) unicompartmental, or patellofemoral (PFA) knee arthroplasty) were eligible if they were accessible

90 by email. In the last 6 months of the inclusion period, also patients without an email address were included,

91 and they received questionnaires by letter. Only the 1452 patients who provided PROM-data prior to surgery

92 were enrolled in this study (89% of those initially enrolled and contacted, 56% of all operated). The inclusion

93 process, exclusion criteria, inclusion rate and representativeness of the sample are all documented in the

94 aforementioned publication [1]. The reporting of the study followed the STROBE guidelines.

95

4
 96 Data collection

97 Questionnaires were sent out by emails with unique links before surgery and 6 weeks and 3, 6 and 12 months

98 after surgery. As 53 patients had an operation in both knees during the study period and a substantial number

99 had bilateral knee trouble, we sought to avoid confusion by clearly addressing “right” or “left knee” along

100 with the specific current follow-up time in each email. In case of no reply, two reminders were sent out 3

101 days apart. Additionally, at the 1-year follow-up, patients who failed to answer electronically were sent a

102 printed questionnaire along with a pre-paid envelope to their home address. Information regarding surgical

103 procedures was prospectively collected from the routine registrations made by surgeons immediately after

104 surgery. Any missing or erroneous surgical data were meticulously identified through patient charts and

105 corrected, and whenever possible, follow-up questionnaire sequences were resumed in time.

106

107 For patients who went through subsequent revision surgery (removal, exchange, or addition of any implant

108 material), PROM collection was ceased on the day of revision. Revision rates and indications fluctuate even

109 in large populations, thus the relatively few patients that would be revised during year 1 in this sample could

110 not be expected to be representative of recent years’ practice, and they were therefore not the main object of

111 the study. However, their baseline, immediate postoperative and revision data were collected and reported.

112 All revisions were attributed to the primary hospital, no matter which hospital (public or private) performed

113 the revision. Minor surgery, e.g., wound debridement or manipulation under anesthesia did not prompt

114 cessation.

115

116 Radiographic classification of knee osteoarthritis

117 Radiographic severity of knee OA in posteroanterior weight bearing radiographs had been graded (blindly)

118 in 1051 of patients (all available, from TKA and MUKA patients only) according to Kellgren-Lawrence (K-

119 L) classification (0-4, where 4 is most severe) and Ahlbäck score (0-5, 5 most severe) [1, 2, 12]. Moreover,

120 13 experienced KA surgeons had performed 17,767 “head-to-head” comparisons to rank the radiographs

121 from the mildest (no. 1051) to the most severe case (no. 1) based on heuristics and clinical experience and

122 free of traditional classification systems [17]. For details, we refer to previous publications [1, 17].

5
123

124 Patient-reported outcomes

125 Both absolute and change scores of the primary outcome, Oxford Knee Score (OKS, 0-48 version) were

126 recorded [6, 10] and the proportion of patients reaching Minimal Important Change (MIC) of 8 points,

127 indicating an important improvement to the average patient, was compared across hospitals [3, 11, 16, 22].

128 Copenhagen Knee ROM Scale (CKRS), an illustration-based questionnaire, assessed patient-reported range

129 of motion (ROM) with flexion ranging from 0-6 (6 max) and extension from 0-5 (5 max) [14]. Also, as

130 reported in CKRS validation studies, “flexion deficit” (i.e., flexion scores <5) would identify 95% of patients

131 with passive flexion below 100° (sensitivity) and 81% of patients with flexion above 100° (specificity) [14].

132 Likewise, “extension deficit” (i.e., extension scores <4) would identify 78% of patients with passive

133 extension above 10° and 70% of patients with extension better that 10°. The knee-specific PROMs were

134 preceded by a global knee anchor question, “How is your knee at the moment?” answered on a Visual

135 Analogue Scale (VAS) from 0-100 (100 best), the generic EQ-5D-5L, EQ-VAS [24], along with a question

136 concerning frequency of use of any type of analgesics against knee pain with answer options; “More than

137 once daily”, “Once daily”, “More than once weekly”, “More than once monthly” or “Rarely or never”.

138

139 Beside these questions, that were included in every pre- and postoperative PROM set, further questions or

140 PROMs were added at varying time points. The Forgotten Joint Score (FJS) [4] and UCLA Activity Scale

141 (UCLA) [15] were used in all sets from 3 months postoperatively, and UCLA also at baseline. At 6 months,

142 patients were asked whether they had received physiotherapeutic assistance in rehabilitation after hospital

143 discharge. At 3, 6 and 12 months, patients were asked about overall satisfaction, “How satisfied are you with

144 the overall experience of the operation and its result?” with 5 answer options (Likert boxes, one neutral).

145 These answers were expected to be influenced by experiences related to hospital service, kindness of

146 caretakers, etc. [5], so to achieve a more specific satisfaction measure, patients were asked about

147 “willingness to repeat” at 1-year follow-up: “Suppose you could turn back time: now that you know the

148 result, would you still choose to have a knee replacement?” and 5 answer options were given: “Yes,

6
149 certainly” or “yes, probably”, “I don’t’ know”, “No, probably not” and “No, absolutely not”. Patients always

150 had an opportunity to write free text or to contact the first author directly by email or phone.

151

152 Implants, perioperative care, and follow-up routines

153 Due to the observational nature of the study, we did not interfere with normal hospital routines regarding

154 e.g., analgesics, aftercare, or choice of implants. Inevitably, this has introduced some bias as these

155 confounding factors were inseparable from the hospital variable (and were thus deliberately not adjusted for).

156 Each hospital used a different selection of cemented, uncemented and hybrid implants that had all been on

157 the market for at least 10 years and had proven good survivorship in registries [24]. The predominant implant

158 systems were NexGenTM (Zimmer Biomet), PFCTM Sigma (DePuy Synthes), TriathlonTM (Stryker), OxfordTM

159 Mobile Bearing and ZUKTM (Zimmer Biomet) and AvonTM (Stryker).

160

161 In all three hospitals, tranexamic acid, glucocorticoids and prophylactic antibiotics (dicloxacillin in [Hosp.

162 C], cefuroxime in [Hosp. A and B]) were administered preoperatively. Periarticular infiltration of local

163 anesthetics was given intraoperatively. The postoperative oral analgesics of choice were paracetamol, non-

164 steroid anti-inflammatory drugs (NSAID) and opioids for up to 4 weeks postoperatively. The mean length of

165 stay for TKA patients in [Hosp. A], [Hosp. B] and [Hosp. C] was 2.4, 1.4 and 2.2 nights, respectively, in

166 2017 [25]. For MUKA patients, numbers were 0.6, 1.3 and 0.7 nights.

167

168 Preparatory training with physiotherapists (crutch walking, stairclimbing etc.) was part of the

169 multidisciplinary patient seminar taking place approximately 2 weeks before surgery in all three centers.

170 Additionally, patients in [Hosp. C] and [Hosp. B] were trained postoperatively by a physiotherapist during

171 the hospital stay. After discharge, Copenhagen patients were routinely offered free of charge supervised

172 physiotherapy (typically more than 10 sessions). By contrast, in [Hosp. A] and [Hosp. B], a physiotherapist

173 screened patients 2-6 weeks after discharge to identify those with poor progress and offered them

174 physiotherapy if needed. If after 6-8 weeks, results were still not satisfactory, patients were referred to the

175 surgeon for a follow-up visit. All [Hosp. A] patients met the surgeon for a routine clinical follow-up

7
176 examination after 3 months, while in [Hosp. A and B], only those with abnormal findings on a 1-year

177 postoperative knee radiograph (and TKA patients only) were seen by a surgeon.

178 Statistics

179 The primary outcome was 1-year OKS, subsequently OKS change. Secondary endpoints were other PROMs

180 and ROM as measured by CKRS. All research questions were explored from the null hypothesis perspective

181 that there was no difference among hospitals; analyses compared all three hospitals, and overall p-values

182 were reported. [Hosp. A] was chosen as the reference hospital because it was in-between the two other

183 hospitals with respects to geography, degree of urbanization and revision rates. All p-values were two-sided

184 with alpha level 0.05. Standard deviations were presented as “(± SD)”. For power analysis considerations,

185 we kindly refer to the baseline article [1].

186

187 Tabular data were analyzed by Chi-square test (with Monte-Carlo correction for expected cell counts < 5),

188 and Clopper-Pearson confidence intervals were provided when relevant. Non-parametric methods (Kruskal-

189 Wallis or Wilcoxon/Mann-Whitney U test) were used for most ordinal measures (UCLA Activity Scale,

190 patient satisfaction, willingness to repeat, use of analgesics and radiographic OA severity classifications

191 (Ahlbäck and K-L)), whereas the outcomes OKS, FJS, EQ-5D, CKRS and global knee anchor were

192 compared using parametric statistical methods (analysis of variance (ANOVA) or t-test) [16]. For ease of

193 reading, some ordinal outcomes were presented as dichotomous variables, yet only when detailed ordinal

194 analysis provided no valuable additional information, and only P-values based on full ordinal comparisons

195 were reported. Multiple linear regression analysis was made for 1-year OKS and OKS change. Here,

196 analyses were made for both Ahlbäck and K-L scales, and since the radiographic classification method did

197 not change the overall result, conference intervals (CI) were reported based on one score only (Ahlbäck). To

198 study patients with poor results, dichotomization of change scores was made at 8 OKS points, which has

199 been identified in [Nationality] KA patients as the Minimal Important Change (MIC), i.e. the minimal

200 change score considered to be an “important improvement” by the notional, average patient, based on the

8
201 predictive modelling approach [11, 22]. As 1-year change scores were unavailable in revision patients, the

202 analysis was repeated with imputed results. Use of imputation is clearly marked and described in the text.

203

204 All observations were treated as independent data, though a minority of answers (7%) came from patients

205 included twice (e.g., first left and subsequently right knee surgery) [20]. All analyses were carried out in R

206 (RStudio) [19]. Data collection and Case Report Forms (CRF) were handled by Procordo Software Aps,

207 [City].

208 Results

209 Patient population and response rates

210 The main characteristics of participating patients are outlined (table 1). For in-depth characteristics and

211 sample representativeness, we refer to the publication of baseline data [1]. During the first postoperative

212 year, 3 patients decided to leave the study, 7 died and 9 were not contacted for follow-up due to errors, e.g.,

213 wrong laterality or email address. 1414 patients (97%) responded at least once postoperatively (table 2). At 1

214 year, 1307 patients (90%) replied, and non-responders were equally distributed among hospitals (p=0.4).

215

216 Revision surgery was conducted in 28 patients (1.9%) during the first postoperative year; 2 (0.6%) in [Hosp.

217 A], 4 (2.0%) in [Hosp. B] and 22 (2.4%) in [Hosp. C] (p=0.1). Details regarding time of revision, indication

218 for revision and latest postoperative OKS score are provided (table 3). Revision due to deep infection was

219 done in 13 cases; 1 (0.3%) in [Hosp. A], 1 (0.5%) in [Hosp. B] and 11 (1.2%) in [Hosp. C] (p=0.4), and due

220 to other causes than deep infection in 15 cases; 1 (0.3%) in [Hosp. A], 3 (1.5%) in [Hosp. B] and 11 (1.2%)

221 in [Hosp. C] (p=0.4).

222

223 1-year PROM results

224 Patient satisfaction and willingness to repeat surgery were independent of hospital (table 4). OKS at 1-year

225 follow-up did not differ significantly among patients in the three hospitals (39 ± 7, p=0.1) (fig. 1a). This

9
226 remained the case when adjusted for age and sex, also when further adjusted for baseline OKS and EQ-VAS

227 and for the variables that differed among hospitals preoperatively, i.e., BMI, self-reported level of

228 “anxiety/depression” (an EQ-5D-5L item), and radiographic classification (Ahlbäck or K-L). 1-year OKS

229 Change score was lower in [Hosp. A] (+1.6 in [Hosp. B], CI 0.07-3, and +1.3 in [Hosp. C], CI 0.2-2,

230 respectively) (fig. 1b). This conclusion was partly changed with adjustment for age, sex and baseline OKS

231 (+1.0 in [Hosp. B], CI -0.3-2, +1.1 in [Hosp. C], CI 0.2-2), and when further adjustments were made for

232 BMI, EQ-VAS, self-reported anxiety and depression and radiographic classification, there were no

233 significant differences among hospitals ([Hosp. B] CI -0.4-3, [Hosp. C] CI -0.1-2, p>0.2). In [Hosp. A], 19%

234 of patients did not reach the MIC of 8 OKS points at 1 year, as did not 13% of [Hosp. B] and 14% of [Hosp.

235 C] patients (p=0.051). To fairly take into account the uneven distribution of revised patients at 1-year, all the

236 28 revision patients, that had been excluded from the latter analysis, were now assigned an imputed

237 (hypothetic) change score < 8 points, i.e., they were added to the group of patients not reaching MIC, and the

238 analysis was repeated: the proportions of patients not reaching MIC were now 20, 15 and 16%, respectively,

239 in the three hospitals (p=0.2). Alternatively, when comparing “last available” postoperative OKS change

240 score in the 1414 patients, who had answered at least one postoperatively (97% of participants including 17

241 of the 28 revision patients), there was no significant difference among hospitals either (21, 16 and 16%,

242 p=0.07). Some hospital variance was noted in knee extension, in physiotherapeutic assistance during

243 rehabilitation and, as reported, in implant choice [4], and [Hosp. B] patients gained more in general health

244 (EQ-VAS, p<0.001).

245

246 Fig. 1 (a & b) Oxford Knee Score at 1 year: A) absolute score, and B) change score with marking of minimal important change
247 (MIC) = 8 points.
248

249 PROM scores during the first postoperative year

250 When OKS was studied over time, patients in [Hosp. C] had better OKS at 6 weeks after surgery; 27.7 ± 7

251 compared to [Hosp. A] (25.6 ± 8) and [Hosp. B] (26.1 ± 7, p=0.001, unadjusted) (fig. 2a), and fewer [Hosp.

252 A] patients had change scores below MIC (66 vs. 72/76%) at 6 weeks. No similar pattern was noted in the

253 other PROMs at 6 weeks (or later), and the hospital difference was further nuanced when MUKA and TKA

10
254 patients were studied separately (fig. 2b). From 3 months, no hospital differences were observed in absolute

255 OKS (table 5 displays the total sample). Through the entire study, OKS differed between TKA and MUKA

256 patients in the total sample, e.g., 1-year OKS was 38.7 and 40.3, respectively (CI 0.6-3).

257

258 Fig. 2 (a & b) Oxford Knee Score during the first postoperative year in A) all patients, and in B) TKA and
259 MUKA patients separately. Whiskers denote mean ± 2 x std. error of the mean.
260

261 Hospital variation in results for comparable patients

262 When patients were grouped by preoperative Ahlbäck or K-L classification, neither willingness to repeat

263 surgery, 1-year OKS or “last postoperative OKS” differed among hospitals (p=0.09-1) (fig. 3). An exception,

264 though, was the “K-L 4” group of 64 patients: here, the 17 [Hosp. A] patients had 4-6 points lower 1-year

265 OKS (CI 0.04-11) and 4-6 points lower “last postoperative OKS” (CI 0.03-10). Remembering that [Hosp. A]

266 patients were more often treated with unicompartmental implants, it should be mentioned that in the “K-L 4”

267 group, the only 2 [Hosp. A] patients with MUKA implants had 1-year OKS scores 37 and 40. When all

268 patients were grouped by baseline OKS (0-20, 21-30 or 30-48) the three aforementioned outcomes did not

269 differ among hospitals (P=0.2-0.5) (total sample displayed in fig. 4). In [Hosp. A], where unicompartmental

270 implants were twice as frequently used, patients had better 1-year knee extension, but when adjustments for

271 baseline motion were made, 1-year extension and flexion were not associated to hospital. However, in the

272 total sample, after adjustments for baseline motion, MUKA was associated to a larger increase in flexion

273 (+0.34 CKRS points, p<0.001) when compared to TKA at 1 year, but not to better extension (p=0.3) (fig. 5).

274 Finally, willingness to repeat surgery was independent of hospital for patients in the same 1-year OKS group

275 (10-point intervals) (p=0.2-0.8).

276

277 Fig. 3 (a & b) Willingness to repeat surgery at 1 year postoperatively grouped by Kellgren-Lawrence
278 classification of preoperative knee OA and hospital, displayed as a) counts and b) proportion of patients.
279
280

281 Fig. 4 (a & b) Willingness to repeat surgery at 1 year postoperatively as a function of Oxford Knee Score at
282 baseline displayed as a) counts, and b) proportions of patients (total sample).
283

11
284 Fig. 5 (a & b) Patient-reported a) flexion and b) extension in the total sample, grouped by implant type
285 (MUKA or TKA only), assessed with Copenhagen Knee ROM Scale. Whiskers denote mean ± 2 x std. error
286 of the mean. Based on validation studies, flexion “4” corresponds to mean 101°, “5” to 121°, and 6 to 131°.
287 In extension, “3” refers to mean 7°, “4” to 5°, and “5” to 1°.
288
289

290 Hospital and regional revision rates

291 During the study period, the [Nationality] Knee Arthroplasty Register reported an (unexplained) decrease in

292 variation in hospital and regional 2-year CRR (table 6) [26].

293 Discussion

294 National knee arthroplasty (KA) registries offer revision rates as simple and readily obtainable outcomes of

295 knee replacement surgery [8, 18]. They are fast and efficient means to detect poor performance of implants,

296 techniques, hospitals or even surgeons, but they offer little information about treatment results in the far

297 majority of patients who are not revised [8, 21]. Outcome of surgery is not a yes-no question, but rather a

298 wide spectrum ranging from a satisfied patient with a perfectly functioning prosthesis to an ill, infected

299 patient in definite need of revision surgery. Outcome evaluation should reflect this reality in order to measure

300 and ultimately improve the quality of surgery. In this epidemiological study, patients in three high-volume

301 KA centers with a history of very different cumulative revision rates (CRR) were followed from pre- to 1-

302 year postoperatively with the specific purpose of measuring and comparing the quality of treatment as

303 experienced by patients. Patients in the three hospitals were found to have comparable results after primary

304 KA when measured with well-established PROMs, patient satisfaction and willingness to repeat surgery.

305 This contradicts the conclusion that could be drawn from implant survival data alone where high hospital

306 CRR’s are generally considered to represent poorer surgical outcomes.

307

308 On most parameters, the patients improved similarly across hospitals over time in the first year after surgery.

309 1-year PROM results and willingness to repeat surgery were similar across the three hospitals with minor

310 exceptions that were largely levelled out when confounding baseline characteristics were adjusted for. The 6-

311 months mean OKS of 37.0 was in line with results from other countries, e.g., the OKS of 37.7 reported from

12
312 the National Joint Registry for England and Wales in 2015 [13]. There were insignificant variations in the

313 proportions of patients reaching minimal important change (MIC) of 8 OKS points and we are unable to say

314 whether the individual patients with poor progress would have had any benefit from revision surgery. [Hosp.

315 C] patients had higher 6-weeks postoperative OKS suggesting a faster initial recovery. The more frequent

316 use of physiotherapy in [Hosp. C] probably does not offer a valid explanation to this difference as no

317 differences were observed in other measures, e.g., flexion and extension measured by CKRS. Also, when

318 results were stratified by implant type, a different pattern was noted (fig. 2 b).

319

320 Importantly, we found that patients with comparable levels of knee OA (as measured by either baseline OKS

321 or radiographic OA classification) reached comparable outcomes (1-year OKS) and had comparable degree

322 of willingness to repeat surgery across the three hospitals. This suggests an overall uniformity in treatment

323 quality. In the whole sample there were differences in outcomes between MUKA and TKA patients, but

324 although [Hosp. A] used unicompartmental implants twice as often as the other two hospitals, these implant-

325 related differences were not directly visible in the overall comparison of hospitals. An exception, though,

326 was a tendency of better ROM in [Hosp. A]. Yet, these results were only partly significant, and it must be

327 kept in mind that clinically relevant differences in patient-reported ROM were not quantified as part of the

328 CKRS scale validation [14]. The larger gain in 1-year flexion among MUKA patients in general (+0.3 CKRS

329 points after adjustment for baseline flexion) should correspond to approximately 5 degrees based on

330 development studies, but the clinical relevance of this difference is yet to be determined [14].

331

332 Strengths and limitations

333 We found an observational study to be the most favorable design to explore the clinical reality behind the

334 wide variations in [Nationality] KA revision rates. With this design, however, conclusions about casual

335 relations, e.g., between implant choice and outcomes, cannot be drawn. It is a limitation that the study was

336 conducted in three hospitals and not whole regions, thus, results do not necessarily reflect the circumstances

337 on a regional level. Despite our intention to invite practically all primary KA patients, the participation rate

338 was limited to approximately 56% of the patients operated in that period of time (62% in [Hosp. A] and

13
339 [Hosp. C], and 37% in [Hosp. B] [1]. Though sample patients were close to being representative of the whole

340 KA population with respects to age, sex and implant types [1], we did not analyze the socioeconomic

341 distribution among hospitals, and comorbidity information was derived from the EQ-questionnaires alone [7,

342 9].

343

344 It can be considered to be an important strength that 89% of patients responded prior to surgery and there

345 was a very low loss to follow-up; 90% of participants responded at 1 year, and 97% responded at least once

346 postoperatively. Information about subsequent revision surgery was considered complete. Patients who were

347 revised participated only until the day of revision surgery, thus 1-year results were unknown in these 28

348 patients, but efforts (transparent imputations) were made to compensate for this lack of information in

349 comparative analyses.

350

351 The historical hospital differences in CRR have not been confirmed in this sample of patients. Nor was there

352 any detectable hospital difference in the proportion of patients revised due to deep infection or other causes.

353 A natural next step would be to explore revision thresholds and indications in a nationwide study to reach a

354 critical number of participants and evaluate their benefit from revision surgery. The lack of significant

355 differences in hospital revision rates in the current sample was expected due to small sample size (28

356 revision patients) but, importantly, during the study period, differences in CRR have diminished on both

357 hospital and regional levels in [Country] (table 6). This may represent a random variation or a general

358 tendency. In the study hospitals, it cannot be ruled out that KA surgeons’ revision thresholds and patterns

359 may have been affected by awareness of the ongoing SPARK study focusing specifically on surgical results

360 and revision patterns. As surgeon staffs have been largely stable and no other significant changes have been

361 made, it seems unlikely that the surgical quality of primary KA has changed at uneven pace in the three

362 hospitals between past years and the study period.

363

14
364 Conclusions

365 The SPARK study has shown that patient-reported outcomes 1 year after primary knee arthroplasty are

366 comparable in three high-volume centers that have differed in revision rates for a decade. It follows, that

367 hospital variance in revision rates does not necessarily reflect differences in the overall quality of primary

368 surgery, and an attempt to rank the general quality of treatment in knee replacement centers based on implant

369 survival data alone might have led to false conclusions. The same concerns should be raised when comparing

370 revision rates among nations. Further studies focusing specifically on revision patients should determine

371 whether patients are offered revision surgery on the same clinical grounds across regions and hospitals, and

372 moreover to which extent patients benefit from revision surgery that is not motivated by deep infection. Such

373 studies followed by discussions about revision indications and techniques might serve the patients with poor

374 results as much as the ongoing attempts to refine primary knee replacement surgery.

375 Statements and Declarations

376 Competing Interests: Authors have reported no conflicts of interests.

377

378 REFERENCES
379 1. [Anonymized] (2022) Hospital variation in revision rates after primary knee arthroplasty: A matter of

380 patient selection? Baseline data from 1452 patients in the prospective multicenter cohort study,

381 SPARK. Submitted

382 2. Ahlbäck S, Rydberg J (1980) Röntgenologisk klassifikation och undersökningsteknik vid gonartros.

383 Lakartidningen 77:2091–2093, 2096

384 3. Beard DJ, Harris K, Dawson J, Doll H, Murray DW, Carr AJ, Price AJ (2015) Meaningful changes

385 for the Oxford hip and knee scores after joint replacement surgery. J Clin Epidemiol Nuffield

386 Department of Orthopaedics, Rheumatology and Musculoskeletal Sciences, Botnar Research Centre,

387 University of Oxford, Headington, Oxford OX3 7LD, UK. Nuffield Department of Orthopaedics,

388 Rheumatology and Musculoskeletal Sciences, Botnar Research C 68:73–79

15
389 4. Behrend H, Giesinger K, Giesinger JM, Kuster MS (2012) The “forgotten joint” as the ultimate goal

390 in joint arthroplasty: validation of a new patient-reported outcome measure. J Arthroplast Department

391 of Orthopaedic Surgery, Kantonsspital St Gallen, St Gallen, Switzerland. 27:430-436.e1

392 5. Clement ND, Macdonald D, Burnett R, Simpson A, Howie CR (2017) A patient’s perception of their

393 hospital stay influences the functional outcome and satisfaction of total knee arthroplasty. Arch

394 Orthop Trauma Surg Department of Orthopaedics and Trauma, The Royal Infirmary of Edinburgh,

395 Little France, Edinburgh, EH16 4SA, UK. nickclement@doctors.org.uk. University of Edinburgh,

396 Little France, Edinburgh, EH16 4SB, UK. Department of Orthopaedics and Trauma, The Royal

397 137:693–700

398 6. Dawson J, Fitzpatrick R, Murray D, Carr A (1998) Questionnaire on the perceptions of patients about

399 total knee replacement. J Bone Jt Surg Br University of Oxford, England, UK. 80:63–69

400 7. Dunbar MJ, Robertsson O, Ryd L (2004) What’s all that noise? The effect of co-morbidity on health

401 outcome questionnaire results after knee arthroplasty. Acta Orthop Scand Department of

402 Orthopaedics, Lund University Hospital, SE-221 85 Lund, Sweden. Michael.dunbar@ort.lu.se

403 75:119–126

404 8. Goodfellow JW, O’Connor JJ, Murray DW (2010) A critique of revision rate as an outcome measure:

405 re-interpretation of knee joint registry data. J Bone Jt Surg Br Botnar Research Centre, Nuffield

406 Orthopaedic Centre, Headington, Oxford OX3 7LD, UK. jgoodfellow@btconnect.com 92:1628–1631

407 9. Harcourt WG, White SH, Jones P (2001) Specificity of the Oxford knee status questionnaire. The

408 effect of disease of the hip or lumbar spine on patients’ perception of knee disability. J Bone Jt Surg

409 Br Robert Jones and Agnes Hunt Orthopaedic Hospital, Oswestry, Shropshire, UK. 83:345–347

410 10. Harris K, Dawson J, Gibbons E, Lim CR, Beard DJ, Fitzpatrick R, Price AJ (2016) Systematic review

411 of measurement properties of patient-reported outcome measures used in patients undergoing hip and

412 knee arthroplasty. Patient Relat Outcome Meas Nuffield Department of Orthopedics, Rheumatology

413 and Musculoskeletal Sciences. Nuffield Department of Population Health (HSRU), University of

414 Oxford, Oxford, UK. 7:101–108

415 11. [Anonymized publication] Subject: MIC values Oxford Knee Score and Forgotten Joint Score

16
416 12. Kellgren JH, Lawrence JS (1957) Radiological Assessment of Osteo-Arthrosis. Ann Rheum Dis

417 16:494–502

418 13. Liddle AD, Pandit H, Judge A, Murray DW (2015) Patient-reported outcomes after total and

419 unicompartmental knee arthroplasty. Bone Joint J 97-B:793–801

420 14. Mørup-Petersen A, Holm PM, Holm CE, Klausen TW, Skou ST, Krogsgaard MR, Laursen MB,

421 Odgaard A (2018) Knee Osteoarthritis Patients Can Provide Useful Estimates of Passive Knee Range

422 of Motion: Development and Validation of the Copenhagen Knee ROM Scale. J Arthroplasty

423 Churchill Livingstone 33:2875-2883.e3

424 15. [Anonymized publication] Subject: UCLA Activity Scale.

425 16. Murray DW, Fitzpatrick R, Rogers K, Pandit H, Beard DJ, Carr AJ, Dawson J (2007) The use of the

426 Oxford hip and knee scores. J Bone Jt Surg Br Nuffield Department of Orthopaedic Surgery

427 University of Oxford, Headington, Oxford, UK. david.murray@noc.anglox.nhs.uk 89:1010–1014

428 17. [Anonymized publication] Subject: Paired comparisons, radiological osteoarthritis.

429 18. Price AJ, Alvand A, Troelsen A, Katz JN, Hooper G, Gray A, Carr A, Beard D (2018) Knee

430 replacement. Lancet Nuffield Department of Orthopaedics, Rheumatology and Musculoskeletal

431 Science, Nuffield Orthopaedic Centre, Oxford, UK. Electronic address:

432 andrew.price@ndorms.ox.ac.uk. Nuffield Department of Orthopaedics, Rheumatology and

433 Musculoskeletal Science, Nuffie 392:1672–1682

434 19. RCoreTeam RStudio version 1.1.463. https://www.R-project.org/. R: A Language and Environment

435 for Statistical Computing.

436 20. Robertsson O, Ranstam J (2003) No bias of ignored bilaterality when analysing the revision risk of

437 knee prostheses: Analysis of a population based sample of 44,590 patients with 55,298 knee

438 prostheses from the national Swedish Knee Arthroplasty Register. BMC Musculoskelet Disord 4:1

439 21. Robertsson O, Ranstam J, Lidgren L (2006) Variation in outcome and ranking of hospitals: an

440 analysis from the Swedish knee arthroplasty register. Acta Orthop Department of Orthopedics, Lund

441 University Hospital, Lund, Sweden. otto.robertsson@ort.lu.se 77:487–493

442 22. Terluin B, Eekhout I, Terwee CB, De Vet HCW (2015) Minimal important change (MIC) based on a

17
443 predictive modeling approach was more precise than MIC based on ROC analysis. J Clin Epidemiol

444 Elsevier Inc 68:1388–1396

445 23. The [National] Knee Arthoplasty Register. Annual Report 2016.

446 24. https://euroqol.org/wp-content/uploads/2019/09/EQ-5D-5L-English-User-Guide_version-3.0-Sept-

447 2019-secured.pdf.

448 25. The [National] Knee Arthoplasty Register. Annual Report 2018.

449 26. The [National] Knee Arthoplasty Register. Annual Report 2021.

450

451

18
Figure Click here to access/download;Figure;Fig1.PNG
Figure Click here to access/download;Figure;Fig2.PNG
Figure Click here to access/download;Figure;Fig3.png
Figure Click here to access/download;Figure;Fig4.png
Figure Click here to access/download;Figure;Fig5.PNG
Table

File information
Title: Revision rates differ across hospitals, yet patient-reported outcomes after primary knee arthroplasty are the
same: 1-year postoperative results from the prospective multicenter cohort study, SPARK.
Journal: Knee Surgery, Sports Traumatology, Arthroscopy
Corresponding author: anonymized.
Email: anonymized.

Table 1. Baseline characteristics of participants

Value/count %
Patients (n) 1452 100
Age (years) (mean [median] ± SD) 68.0 [69] ± 9
Male sex (n) 659 45
Body Mass Index (kg/m2) (mean [median] ± SD) 29 [28] ± 5
Patients per hospital, all implants (UKAa) 1452 (393)a 100 (27)
[Hosp. A] 321 (157) 22b (49)c
[Hosp. B] 202 (28) 14b (14)c
[Hosp. C] 929 (208) 64b (22)c
Radiographic severity of knee osteoarthritis (n total) 1051 (100)
Kellgren-Lawrence classification ≥ 2 (n) 987 94
≥ 3 (n) 851 81
Ahlbäck classification ≥ 2 (n) 704 67
≥ 3 (n) 305 29
Patient characteristics. For complete baseline results per hospital, we refer to the baseline publication [4].
Hospitals are named short for simplicity: “[Hosp. A]” = [anonymized] in Region [A], “[Hosp. B]” = [anonymized] in Region [B]
and “[Hosp. C]” = [anonymized] in Capital Region [C].
a) UKA denote the proportion of patients who had a unicompartmental implant inserted (medial, lateral or patellofemoral).

Proportions were in concordance with those reported in the [National] Knee Arthroplasty Register.
b) Hospital proportion of SPARK sample. c) UKA proportion of local hospital sample.

Table 2. Questionnaire response rates

Baseline 6 weeks 3 months 6 months 1 year Any postop.
Responding patients (n) 1452a 1147b 1237 1241 1307 1414
Available patients (n) 1452a 1296b 1435 1433 1417 1443
Revised/dead patients (n) 0/0 9/1 15/2 17/2 28/7 9/1
Response rate
per available patients (%) (100) a 89 86 87 92 98
per 1452 baseline responders (%) 100 a 79 89 86 90 97
Days from surgery (mean [median] ± SD) -29 [-18] ± 32 39 [38] ± 7 87 [84] ± 14 179 [176] ± 14 368 [359] ± 27 -
a) For response rates at inclusion, we refer to publication of baseline data [4]. b) The 6-week questionnaire was delayed and thus not
sent to the first 146 included patients. Non-responders at 1-year were further contacted by letter.

1
Table 3. Characteristics of patients who were revised during the first postoperative year
Patient group n Male sex Age BMI Implant type
(%) (n (%)) years (mean ± SD) kg/m2 (mean ± SD) (TKA/MUKA/other)
No revision 1424 (98) 642 (45) 68.0 ± 9 28.9 ± 5 1039/328/57
Revision 28 (2) 17 (61) 66.4 ± 10 26.9 ± 4 20/8/0
P-value - 0.1 0.4 0.008a 0.7
Revision time [Hosp. A] [Hosp. B] [Hosp. C] Total sample
n = 321 n = 202 n = 929 n = 1452
0 - 6 w. N 1 1 7 9
Indication A A A, A, A, A, A, B, B 7A, 2B
Last OKS before revision - - - -
6 w. - 3 mo. N 0 0 6 6
Indication A, A, A, A, B, C 4A, 1B, 1C
Last OKS before revision A: 34,39,NA,NA. B: 25. C: 28. Mean: (28)
3 - 6 mo. N 1 1 0 2
Indication C C 2C
Last OKS before revision 10 20 Mean: 15
6 – 12 mo. N 0 2 9 11
Indication C, C A, A, B, C, C, C, C, C, C 2A, 1B, 8C
Last OKS before revision 11,35 A: 45,18. B: 26. C: 16,28,29,31,32,34. Mean: 28
Total Revisions during year 1 2 4 22 28
Indications 1A, 1C 1A, 3C 11A, 4B, 7C 13A, 4B, 11C
Revision rate in sample (%) 0.6 2.0 2.4 1.9
95% CI (%) 0.7 - 2 0.5 - 5 2-4 1–3
Indication: Revision due to A=deep infection, B=fracture or liner dislocation, C=other cause. “Last OKS before revision” =
Patient’s last postoperative Oxford Knee Score before revision. NA= Missing (not available). a) BMI in revision group: CI -0.6-(-3).

2
Table 4. Patient-reported outcomes at 1-year follow-up.
Hospital
[Hosp. A] [Hosp. B] [Hosp. C]
Total sample P
(Low rev. rate) (Low rev. rate) (High rev. rate)
Patients at baseline (n) 1452 321 202 929
Implant type (n (%)) <0.001
TKA 1059 (73) 164 (51) 174 (86) 721 (78)
MUKA 336 (23) 129 (40) 25 (12) 182 (20)
PFA 50 (3) 23 (7.2) 3 (1.5) 24 (2.6)
LUKA 7 (1) 5 (1.6) 0 (0.0) 2 (0.2)
Oxford Knee Score (OKS)
1 y. (n=1306) 39 [41] ± 7 38.1 [40] ± 8.3 39.1 [41] ± 7.2 39.2 [41] ± 7.2 0.09
Last available postop. (n=1414) 38 [40] ± 8 37.5 [40] ± 8.7 38.7 [40] ± 7.5 38.5 [40] ± 7.8 0.1
Change (n=1307) 15 ± 8 14.3 ± 8.7 15.9 ± 7.8 15.7 ± 8.0 0.04
OKS change < MIC (8 points) (total no. of patients in analysis (%))
1 y. (n = 1307) 195 (15) 56 (19) 25 (13) 114 (14) 0.051
1 y. imputed* (n = 1335) 223 (17) 58 (20) 29 (15) 136 (16) 0.2
Last available postop. (n = 1414) 237 (17) 66 (21) 31 (16) 140 (16) 0.07
Overall assessments
Willing to repeat surgery (%) 0.1
"Yes, certainly" 1005 (77) 211 (73) 150 (80) 644 (77)
"Yes, probably" 200 (15) 46 (16) 26 (14) 128 (15)
"I don’t know" 52 (4) 14 (4.9) 6 (3) 32 (3.9)
"No, probably not" 32 (2.5) 12 (4.2) 3 (1.6) 17 (2.0)
"No, absolutely not" 17 (1.3) 5 (1.7) 2 (1.1) 10 (1.2)
“Satisfied” or “very satisfied” (%) 1125 (86) 238 (83) 161 (87) 726 (87) 0.6a
Global knee anchor 1 y. (0-100) 80 ± 21 78 ± 24 81 ± 21 80 ± 19 0.08
Change 51 ± 26 50 ± 29 51 ± 26 51 ± 25 0.8
Forgotten Joint Score, 1y 60 ± 27 59.1 ± 29 59.7 ± 25 60.1 ± 26 0.9
Patient-reported knee range of motion (CKRS units b)
Flexion 1y. 5.4 [6] ± 0.8 5.41 [6] ± 0.76 5.30 [5] ± 0.76 5.34 [5] ± 0.77 0.3
Deficit (CKRS 0-4) (n (%)) 165 (13) 32 (11) 21 (11) 112 (13) 0.5
Change 0.48 [0] ± 1 0.57 [0] ± 1.2 0.55 [0] ± 1.2 0.43 [0] ± 1.1 0.2
Extension 1 y. 4.1 [4] ± 0.7 4.24 [4] ± 0.65 4.10 [4] ± 0.61 4.12 [4] ± 0.68 0.02
Deficit (CKRS 0-3) (n (%)) 161 (12) 29 (10) 24 (13) 108 (13) 0.4
Change c 0.7 [1] ± 1 0.73 [1] ± 1.0 0.72 [1] ± 0.9 0.64 [1] ± 1.0 0.6
UCLA (Activity) 1 y. 6.0 [6] ± 2 5.8 [6] ± 1.9 6.0 [6] ± 1.8 6.0 [6] ± 1.9 0.5
Change 1.2 [1] ± 2 1.0 [1] ± 1.9 1.3 [1] ± 1.9 1.3 [1] ± 1.9 0.06
EQ-VAS 1 y. 79 ± 18 78 ± 20 82 ± 15 79 ± 18 0.08
Change 17 ± 23 16.1 ± 24 24.3 ± 24 16.3 ± 22 <0.001
EQ-5D-5L Index 1 y. 0.81 ± 0.2 0.80 ± 0.17 0.83 ± 0.14 0.82 ± 0.14 0.04
Change 0.22 ± 0.2 0.20 ± 0.18 0.23 ± 0.15 0.22 ± 0.17 0.049
Daily use of analgesics against knee pain (n (%)) 166 (13) 41 (14) 22 (12) 103 (12) 0.4a
d
Supervised physiotherapy in rehabilitation (n (%)) 702 (73) 115 (51) 92 (70) 495 (81) <0.001
When no unit is noted, means ± SD [and medians] are provided. *) “1y. imputed”: Here, all 28 revised patients are assumed to be in
the group with OKS change < MIC (8 points). a) Patient satisfaction is dichotomized for presentation, but P-value refers to tests of
all 5 ordinal answer options. b) CKRS: With Copenhagen Knee ROM Scale, patients rate flexion from 0 (unable) to 6 (full flexion
ability), and extension from 0 (unable) to 5 (full extension or slight hyperextension). c) Only the last 699 patients included in this
analysis due to delay of scale development. d) Only the last 966 patients were asked about physiotherapy.

3
Table 5. Development of main PROMs over time after surgery (all hospitals).
Baseline 6 weeks 3 months 6 months 1 year
(preop.)
Oxford Knee Score (OKS) 23 ± 7 27 ± 8 34 ± 8 37 ± 7 39 ± 7
OKS change score - 3.6 ± 8 10 ± 8 14 ± 8 15 ± 8
OKS change score < MIC (8 points) (n (%)) - 788 (69) 462 (36) 262 (21) 195 (15)
Global knee anchor (0-100) 28 ± 18 60 ± 21 71 ± 22 76 ± 21 80 ± 21
Range of motion (Copenhagen Knee ROM Scale (CKRS) units)
Flexion 4.9 ± 1.2 4.5 ± 1.1 5.0 ± 0.9 5.3 ± 0.8 5.4 ± 0.8
Deficit (CKRS 0-4) (n (%)) 416 (29) 525 (46) 317 (25) 188 (15) 165 (13)
Extension 3.5 ± 0.9 3.5 ± 0.7 3.9 ± 0.8 4.0 ± 0.7 4.1 ± 0.7
Deficit (CKRS 0-3) (n (%)) a 340 (49) a 336 (42) a 246 (27) a 202 (17) 161 (12)
Forgotten Joint Score - - 43 ± 25 53 ± 26 60 ± 27
UCLA Activity Scale 4.7 [4] ± 2 - 5.4 [5] ± 2 5.8 [6] ± 2 6.0 [6] ± 2
Daily use of analgesics against knee pain (n (%)) 854 (59) 870 (76) 498 (39) 274 (22) 166 (13)
EQ-5D VAS 61 ± 22 71 ± 18 76 ± 17 78 ± 18 79 ± 18
EQ-5D-5L Index 0.59 ± 0.2 0.70 ± 0.1 0.76 ± 0.1 0.79 ± 0.1 0.81 ± 0.2
When no unit is noted, means, ± SD and [medians] are provided.
a) In CRKS extension, total n is increasing during the study due to concomitant scale development.

Table 6. 2-year cumulative revision rates in study hospitals and according regions
2-year CRR (%)
Pre-study period Study period
Hospital (Region) Mean 2011-13 2016 2017 2018
[Hosp. A] (Region [A]) 1.9 (2.5) 3.2 (2.2) 4.5 (2.0) 3.0 (1.9)
[Hosp. B] (Region [B]) 1.6 (1.5) 2.4 (2.6) 2.9 (3.4) 1.7 (3.9)
[Hosp. C] (Capital Region [C]) 5.6 (4.7) 3.3 (2.8) 3.1 (3.0) 3.8 (3.9)
Figures from the [National] Knee Arthroplasty Register. Bold figures denote the highest cumulative revision rate (CRR) of each
year.

4
Supplementary Material

Click here to access/download

Supplementary Material
STROBE Checklist SPARK 1 & 2.doc