Available via license: CC BY 4.0
Content may be subject to copyright.
PHYSICAL
THERAPY
RESEARCH
This work is licensed under a Creative
Commons Attribution International License.
©2024 Japanese Society of Physical Therapy
ORIGINAL ARTICLE
A Clinical Prediction Rule for Predicting the Health-related
Quality of Life after 5 Months in Patients with Knee
Osteoarthritis Undergoing Conservative Treatment
Shunsuke Y, PT, PhD1,2,3, Tetsuya A, PT, PhD4, Shigeharu T, PT, PhD3,5,
Yu I , PT, PhD3,6, and Ryo T, PT, PhD3
1)Department of Rehabilitation, Taira Hospital, Japan
2)Department of Environmental Medicine and Public Health, Faculty of Medicine, Shimane University, Japan
3)Graduate School of Humanities and Social Sciences, Hiroshima University, Japan
4)Department of Physical Therapy, Faculty of Health and Medical Sciences, Tokoha University, Japan
5)School of Rehabilitation, Kanagawa University of Human Services, Japan
6)Department of Human Science, School of Human Science, Kibi International University, Japan
ABSTRACT. Objective: This study aimed to derive a clinical prediction rule (CPR) that can predict changes
in health-related quality of life at 5 months for patients with knee osteoarthritis (KOA) undergoing conserva-
tive treatment. Methods: Patients with KOA receiving physical therapy and exercise therapy at an outpatient
clinic were included in this study. The basic characteristics, medical information, and motor function test
results were recorded at baseline. The primary outcome measure was the change in the Japan Knee Osteoar-
thritis Measure (JKOM) 5 months after the baseline measurement. A decision tree analysis was performed
with the basic characteristics, medical information, and the motor function test results as the independent
variables and the changes in the JKOM after 5 months (≥8 in the improved groups) as the dependent variable.
Results: Analyzed data from 87 patients. The variables included the visual analog scale score, bilateral KOA,
5-m walk test, JKOM, and body mass index. Six CPRs were obtained from the terminal nodes. Accuracy val-
idation of the model for the entire decision tree revealed an area under the receiver operating characteristic
curve of 0.87 (validation data: 0.83), a positive likelihood ratio of 2.6, and a negative likelihood ratio of 0.1.
Conclusion: This CPR is an inspection characteristic that can exclude the possibility of the occurrence of an
event based on a negative result. However, since the results of this study represent the rst process of utilizing
the CPR in actual clinical practice, its application should be kept in mind.
Key words: Clinical prediction rule, Knee osteoarthritis, Machine learning, Health-related quality of life
Introduction
The prevalence of knee osteoarthritis (KOA) is
increasing in aging populations, and the number of patients
with KOA in developed countries is predicted to double by
20311). The healthcare expenditure of patients with osteoar-
thritis is estimated to be four times higher than that of those
without osteoarthritis. Furthermore, healthcare expenditure
tends to increase as the patient ages2). Therefore, pressure
on medical costs is inevitable unless we develop highly
effective treatment interventions and optimize treatment
choices through evaluation. The rst choice of treatment
for patients with KOA is follow-up with conservative
Received: April 19, 2024
Accepted: October 9, 2024
Advance Publication by J-STAGE: November 28, 2024
doi: 10.1298/ptr.E10296
Correspondence to: Shunsuke Yamashina. Department of Environ-
mental Medicine and Public Health, Faculty of Medicine, Shimane
University, 89-1 Enya, Izumo, Shimane 693-8501, Japan
e-mail: yamashina.s@med.shimane-u.ac.jp
(Phys Ther Res 00: 00–00, 0000)
Advance Publication by J-STAGE: November 28, 2024
ptr
Physical Therapy Research
2189-8448
2189-8448
Japanese Society of Physical Therapy
ptr.e10296.oa
10.1298/ptr.E10296
XX
XX
XX
XX
19April2024
2024
9October2024
XX2024
Yamashina, et al.
treatment, such as pharmacotherapy, exercise therapy, and
patient education3,4). Several guidelines recommend the
implementation of exercise therapy and patient education
as it is associated with a low risk of side effects and inter-
vention costs5–8). Identifying cases that could benet from
exercise therapy in advance may facilitate the attainment
of effective therapeutic outcomes while lowering the effect
of side effects.
Health-related quality of life (HRQOL) assessments
have been used to determine the effectiveness of KOA treat-
ment in clinical practice. Among these, the Japan Knee Osteo-
arthritis Measure (JKOM) is an HRQOL scale that reects
the social and cultural background of Japan9). HRQOL is an
important indicator that is used to determine whether total
knee arthroplasty (TKA) should be performed in patients
with KOA. Notably, compared with that of patients who
have not undergone TKA, a decline in HRQOL and daily
living functions is observed postoperatively in patients who
have undergone TKA10). A lower preoperative HRQOL leads
to poorer HRQOL at 6 months postoperatively11). HRQOL
assessment plays a role in the decision to perform surgical
treatment in patients with KOA. Moreover, it has a negative
effect on the postoperative period. Thus, improving HRQOL
should be a goal of conservative treatment.
Previous studies have shown that exercise therapy and
pharmacotherapy for patients with KOA improve HRQOL
to some extent. A report using JKOM, the Western Ontario
and McMaster Universities Arthritis Index (WOMAC),
Short-Form 36-Item Health Survey (SF-36), and visual
analog scale (VAS) as outcomes, reported that 8-week exer-
cise therapy resulted in an improvement of 8–14 points,
5.21–15.62 points, 4.36–7.91 points, and 14.61–23.04 mm
in outcome scores, respectively12). However, some patients
have shown poor response to exercise therapy13). Lack of
clarity regarding which patients can and cannot benet from
exercise therapy can lead to inappropriate treatment choices,
which may result in unnecessary medical costs. Therefore,
it is important to clarify the criteria for exercise therapy
application.
The clinical prediction rule (CPR) is a useful tool for
predicting treatment responses. Previous studies have inves-
tigated the use of CPR to rule out knee fractures and sup-
port the diagnosis of KOA14,15). In other studies, patients with
KOA who were ≤69 years of age, had knee exion muscle
strength of ≤0.36 Nm/kg, and had a VAS score of ≥33 mm
were less likely to benet from usual exercise therapy and
could be predicted to be less physically activity16). These
studies have indicated that multiple pieces of information
can be combined to predict specic events. The use of CPR
in patients with KOA is benecial in that it facilitates the
screening of patients likely to benet from exercise therapy,
thereby establishing the criteria for the application of exer-
cise therapy. On the other hand, CPR has not been used to
predict HRQOL in patients with KOA receiving conserva-
tive treatment. Thus, this study aimed to derive a CPR that
predicts the HRQOL at 5 months for patients with KOA
undergoing conservative treatment.
Methods
Ethical statement
This study was approved by the Research Ethics Com-
mittee of Tokoha University and adhered to the tenets of
the Declaration of Helsinki (approval number: 2022-501H).
Written informed consent and assent were obtained from all
participants prior to enrollment.
Study design
The study was conducted as a multicenter, eight-center
study. The study design was a prospective cohort study. The
observation period was set at 5 months. This was based on
the fact that the Japanese reimbursement system for mus-
culoskeletal rehabilitation usually sets the period of cover-
age at 5 months. However, rehabilitation may be performed
beyond 5 months in some cases.
Participants
Patients with KOA receiving physical therapy and
exercise therapy at an outpatient clinic who met the fol-
lowing criteria were included in this study: age, 50–90
years; Kellgren-Lawrence classication (K-L classica-
tion) categories 1–4; and patients capable of walking ≥10
m by themselves. Patients receiving special therapies or
treatment with special equipment (determined based on the
details of exercise therapy and physical therapy to avoid
major differences in treatment content), patients with neu-
rological disorders (affecting lower-limb muscle strength
and walking), and patients with cognitive impairment
(affecting the results of the self-administered question-
naires) were excluded.
Variables
With reference to variables from previous studies16),
the following characteristics were recorded at baseline: sex,
age, body mass index (BMI), and exercise habits (exercis-
ing at least twice a week for at least 30 minutes at a time
for at least 1 year, which entailed exercises such as walking
and strength training). Data regarding the following medi-
cal information were also recorded: the severity of KOA (as
assessed using the K-L classication), the severity of pain
(as assessed using the visual VAS score, maximum pain at
rest and during activity in the past week), bilateral KOA,
pharmacological treatment, duration of physical therapy
intervention (within 1 month, 1–5 months, 6–11 months, and
≥12 months), physical therapy intervention time (minutes
per week), physical therapy start date (0: The period from
the date of diagnosis to that of starting physical therapy was
less than 1 month, 1: The period from the date of diagnosis
to that of starting physical therapy was at least 1 month) and
the Self-Efcacy for Rehabilitation Outcome Scale (SER)
A Clinical Prediction Rules for Patients with Knee Osteoarthritis
score. The results of the following motor function tests were
also recorded: knee joint extension muscle strength on the
fault side, knee joint exion muscle strength on the fault
side, knee joint extension range of motion on the fault side,
knee joint exion range of motion on the fault side, ve-
times stand and sit test (FTSST), 5-m fastest walk time
(as assessed using the 5-m walk test; 5mWT), and HRQOL
(as assessed using JKOM). The K-L classication category
was higher for patients with bilateral KOA, and the side with
more pain was dened as the fault side.
The K-L classication was determined by an ortho-
pedic surgeon using radiographic images and rated on a
ve-point scale based on the presence of osteophytes and
deformities. The scores ranged from 0 to 4, and category
4 was considered the most severe17). SER uses a dedicated
patient-oriented questionnaire comprising 12 items scored
on a scale of 0 to 10, with a score of 0 for “not at all con-
dent” and 10 for “very condent.” The total score ranged
from 0 to 12018). The knee joint muscle strength on the fault
side was measured using a handheld dynamometer (Tas
F-1; Anima, Tokyo, Japan), with the sensor xed using a
band. The patients were seated with the knee joint at 90°
exion. The maximum isometric contraction of the knee
joint extension and exion muscles was measured in this
position. The distance from the center of the knee joint to
the sensor was dened as the lever arm length. The values
were normalized to the lever arm length and body weight
(Nm/kg). The measurements were acquired twice, and the
better of the two results was selected as the representative
value. The test–retest intraclass correlation coefcient (ICC)
for this measurement method ranged from 0.85 to 0.9219).
In addition, the inter-rater ICC was 0.9320). The knee joint
range of motion on the fault side was measured to determine
the maximum range of motion of the knee joint in other
motions. The values were recorded using a goniometer in
5° increments. The measurements were acquired twice, and
the better of the two results was selected as the representa-
tive value. The test–retest ICC for this measurement method
ranged from 0.81 to 0.96 21). Additionally, the inter-rater ICC
was 0.72–0.8022). FTSST was performed by instructing the
patients to stand up from a chair and sit down as rapidly as
they could. This process was repeated ve times, and the
time required to complete ve repetitions was measured.
The upper limbs were crossed in front of the chest. The
patients were seated both at the beginning and at the end of
the test. The chair had a backrest and was 42 cm high. The
measurements were acquired twice, and the better of the two
results was selected as the representative value. Measure-
ments were recorded using a stopwatch. The test–retest ICC
for this measurement method was 0.9323). Additionally, the
inter-rater ICC was 0.8124). The 5mWT was performed by
measuring the time taken to walk a 5-m walking path at a
fast pace. The walking path had a 3-m reserve path on each
side. Shoes were worn while walking. The measurements
were acquired twice, and the better of the two results was
selected as the representative value. Measurements were
recorded using a stopwatch. The test–retest ICC for this
measurement method was 0.8325). Additionally, the inter-
rater ICC was 0.8424). The JKOM uses a patient-oriented
questionnaire comprising 25 items on pain and stiffness,
daily living, usual activities, and health status9). A score of 0
corresponds to “no pain, no difculty,” whereas a score of 4
corresponds to “severe pain, great difculty.” The total score
ranged from 0 to 100, with higher scores indicating more
severe symptoms.
Outcome
The primary outcome measure was the change in
HRQOL. The JKOM was used to assess HRQOL in the
present study. A change in the JKOM of ≥8 points from the
baseline measurement was dened as improvement, whereas
a change of <8 points was dened as non-improvement.
Previous studies have shown that the change in JKOM in
KOA patients who received 8 weeks of exercise therapy
ranged from 8 to 14 points12). Although the study period was
different in this study, the lower limit of 8 points achieved in
8 weeks was adopted as the cutoff value. This was because an
8-point change in JKOM results in changes in pain and QOL.
Patients with a JKOM of <8 points at baseline were excluded.
Exercise therapy implemented during conservative treatment
mainly comprised a joint range of motion exercises, muscle
strengthening, stretching, and cardiac exercises.
Statistical analysis
The patients were classied into two groups based
on the change in JKOM after 5 months (those with a score
of ≥8 points were classied into the improved group and
those with a score of <8 points were classied into the
non-improved group). Next, the descriptive statistics of sex,
age, BMI, exercise habits, K-L classication, VAS, bilateral
KOA, pharmacological treatment, duration of physical ther-
apy intervention, physical therapy intervention time, phys-
ical therapy start date, SER, knee extension strength on the
fault side, knee exion strength on the fault side, knee exten-
sion range of motion on the fault side, knee exion range of
motion on the fault side, FTSST, 5mWT, total JKOM score
variables, and the test of difference were conducted. Statisti-
cal analysis was performed using t-tests and chi-square tests.
These variables were used as the independent vari-
ables, while the change in JKOM after 5 months (≥8 and <8
points in the improved and non-improved groups, respec-
tively) was used as the dependent variable, and a decision
tree analysis was performed. Non-improvement was dened
as the occurrence of an event (coded as 1), and CPR was
derived from the nodes of the decision tree. A binary tree
was used for the analysis. Branching was terminated if the
number of cases was <10 after branching. K-fold cross-
validation (k = 10) was used to verify the accuracy of the
overall decision tree model. K-fold cross-validation divides
the data into K pieces, one of which is the validation data, and
Yamashina, et al.
the remaining K-1 pieces are the training data, which evaluate
the percentage of correct answers; the method is to train K
pieces of data one by one for K trials to become the validation
data, and then take the average of their accuracy26). The data
were split into training and validation data, and 10 trials were
performed. The area under the receiver operating characteris-
tic curve (AUROC), classication error rate, sensitivity, spec-
icity, positive and negative predictive values, and positive
and negative likelihood ratios were calculated as indices of
accuracy. All statistical analyses were performed using JMP
Pro Version 17 (SAS Institute Japan, Tokyo, Japan).
Sample size
The sample size of AUROC and other inspection
characteristic values were calculated using the following
method. A previous cohort study on patients with KOA
reported that 33.8%–47.9% of patients had good HRQOL
scores, daily functioning, and the ability to stand and sit27).
The ratio was set as 1–2.3, with approximately 70% and
30% in the improved (no events) and non-improved (with
events) groups, respectively. Statistical power was set at
0.85 and AUROC at 0.7. Twenty and 46 patients had to be
included in the improved and unimproved groups, respec-
tively, to achieve the required sample size.
Results
A total of 174 participants, comprising 35 men and 139
women, were eligible for inclusion in this study. Among these
174 patients, 88 met the criteria for analysis and were eli-
gible for follow-up for 5 months. Among these 88 patients,
one patient was excluded as the JKOM score was <8 at base-
line (Fig. 1). Thus, 87 patients were included in the present
study. The improved and non-improved groups comprised 28
(32%) and 59 (68%) participants, respectively. The variables
that showed signicant differences between the improved
and non-improved groups were VAS (p = 0.02), duration of
physical therapy (p = 0.02), knee extension muscle strength
(p = 0.04), 5mWT (p = 0.02), and JKOM (p = 0.002). Table 1
presents the other variables and their details.
VAS, bilateral KOA, 5mWT, JKOM, and BMI were
the variables derived from the decision tree analysis. Six
CPRs were obtained from terminal nodes (Fig. 2). CPR1
comprised a VAS score of <27 mm, which included 32%
(25/79) of the participants. CPR2 comprised a VAS score
of ≥27 mm, bilateral KOA, and 5mWT of ≥4.83 sec, which
included 11.5% (9/79) of participants. CPR3 comprised a
VAS score of ≥27 mm, bilateral KOA, 5mWT of <4.83 sec,
and JKOM of <31 points, which included 16% (13/79) par-
ticipants. CPR4 comprised a VAS score of ≥27 mm, bilateral
KOA, 5mWT of <4.83 sec, and JKOM of ≥31 points, which
included 11.5% (9/79) participants. CPR5 comprised a VAS
score of ≥27 mm, unilateral KOA, and BMI of ≥22.9 kg/m2,
which included 14% (11/79) participants. CPR6 comprised
a VAS score of ≥27 mm, unilateral KOA and BMI of <22.9
Total samples
n = 174 (Men: 35, Women: 139)
Exclusion of JKOM<8
n =1 (M: 0, W: 1)
Follow-up samples
n = 88 (M: 17, W: 71)
Drop-out samples
n = 86 (M: 18, W: 68)
Unidentified (M: 3, W: 24)
Completion of treatment (M: 11, W: 29)
Aggravation of internal disease (M: 0, W: 4)
Suffered from orthopedic disease (M: 0, W: 4)
Transitioned to care insurance (M: 1, W: 2)
Surgical treatment (M: 0, W: 1)
Decease (M: 0, W: 1)
Rejection of participation (M: 0, W: 2)
Various circumstances (M: 3, W: 1)
Analyzed
n = 87 (M: 17, W: 70)
Fig. 1. Flowchart depicting the process of selecting the
participants to be analyzed.
JKOM, Japan Knee Osteoarthritis Measure
kg/m2, which included 15% (12/79) participants. CPR1,
CPR2, CPR3, and CPR5 yielded positive results. Table 2
summarizes the positive and negative results of the CPR
(positive indicates non-improvement, and negative indicates
improvement). The accuracy validation of the model for the
entire decision tree (average of 10 trials) yielded AUROC,
classication error rate, sensitivity, specicity, positive pre-
dictive ratio, negative predictive ratio, positive likelihood
ratio, and negative likelihood ratio of 0.87 (validation data:
0.83), 0.17 (validation data: 0.18), 0.92, 0.64, 0.84, 0.78, 2.6,
and 0.1, respectively.
Discussion
This study used decision tree analysis to derive a CPR
that predicts HRQOL as assessed by the JKOM of patients
with KOA after undergoing conservative treatment for 5
months. The variables derived were VAS, bilateral KOA,
5mWT, JKOM, and BMI. The overall accuracy and test
characteristic values of the decision tree model were good.
In total, there are four positive CPRs and two negative CPRs.
A Clinical Prediction Rules for Patients with Knee Osteoarthritis
Table 1. Characteristics of follow-up, drop-out, and exclusion participants
All (n = 174) Drop-out and exclusion (n = 87)
Follow-up (n = 87)
p-value
Improved group (n = 28) Non-improved group (n = 59)
Sex Men: 35, Women: 139 Men: 18, Women: 69 Men: 7, Women: 21 Men: 10, Women: 49 0.94
Age (years) 73.1 (9.0) 72.3 (10.2) 72.3 (10.2) 74.9 (8.6) 0.22
BMI (kg/m2) 24.50 (3.82) 25.21 (3.98) 23.48 (4.32) 24.60 (3.86) 0.23
Exercise habits Yes: 63, No: 111 Yes: 26, No: 61 Yes: 15, No: 13 Yes: 22, No: 37 0.15
K-L classications I: 39, II: 60, III: 58, IV: 17 I: 17, II: 27, III: 37, IV: 6 I: 6, II: 12, III: 9, IV: 1 I: 16, II: 21, III: 12, IV: 10 0.68
VAS (mm) 48.00 (25.00) 53.70 (24.16) 50.25 (17.52) 37.61 (25.14) 0.02
Bilateral KOA Unilateral: 88, Bilateral: 86 Unilateral: 52, Bilateral: 35 Unilateral: 15, Bilateral: 13 Unilateral: 21, Bilateral: 38 0.11
Pharmacological treatment Yes: 146, No: 28 Yes: 75, No: 12 Yes: 21, No: 7 Yes: 50, No: 9 0.28
Duration of physical therapy 1: 83, 2: 28, 3: 19, 4: 44 1: 61, 2: 8, 3: 10, 4: 8 1: 11, 2: 9, 3: 3, 4: 5 1: 11, 2: 11, 3: 6, 4: 31 0.02
Physical therapy intervention time (min) 58.51 (21.93) 55.81 (19.67) 58.57 (24.90) 63.05 (22.84) 0.41
Physical therapy start date 0: 149, 1: 25 0: 80, 1: 7 0: 26, 1: 2 0: 45, 1: 14 0.06
SER (points) 86.16 (19.67) 84.01 (19.79) 85.57 (19.10) 88.04 (18.33) 0.56
Knee extension muscle strength (Nm/kg) 0.88 (0.41) 0.84 (0.41) 0.98 (0.46) 0.79 (0.35) 0.04
Knee exion muscle strength (Nm/kg) 0.49 (0.19) 0.48 (0.21) 0.47 (0.18) 0.45 (0.17) 0.60
Knee extension ROM (degree) -4.77 (4.76) -4.30 (5.54) -6.61 (4.72) -6.52 (4.67) 0.94
Knee exion ROM (degree) 132.99 (14.35) 132.21 (13.86) 128.57 (15.86) 131.53 (15.95) 0.42
FTSST (sec) 10.24 (4.58) 10.61 (5.05) 9.43 (3.12) 10.13 (4.53) 0.46
5mWT (sec) 4.03 (1.75) 3.97 (2.00) 3.59 (0.89) 4.34 (1.63) 0.02
JKOM (points) 30.14 (16.09) 30.17 (15.27) 38.11 (18.13) 26.73 (14.98) 0.002
Mean (SD). Comparison result between improved and non-improved groups, duration of physical therapy: 1, 1 month or less; 2, 2–5 months; 3, 6–11 months; 4, 12 months or more. Physical
therapy intervention time: intervention time per week (minutes). Physical therapy start date: 0, the period from the date of diagnosis to the date of the start of physical therapy is less than 1 month;
1, start of physical therapy is at least 1 month. JKOM < 8 to exclude one patient.
BMI, body mass index; K-L, Kellgren-Lawrence; VAS, visual analog scale; KOA, knee osteoarthritis; SER, Self-Efcacy for Rehabilitation Outcome Scale; ROM, range of motion; FTSST, ve-
times stand and sit test; 5 mWT, 5-m walk test; JKOM, Japan Knee Osteoarthritis Measure
Yamashina, et al.
The novelty of the present study is that it derived a
CPR that is highly accurate in predicting JKOM in conser-
vatively treated patients with KOA. Limitations of the CPR
for KOA in previous studies include a high misclassication
rate of 0.3928) and lack of cross-validation29). In this study,
a low misclassication rate (training data: 0.17, validation
data: 0.18) was conrmed and cross-validation was per-
formed. This conrms the high accuracy of the CPR derived
in this study. This derived CPR contributes to the accurate
prediction of HRQOL in patients with knee KOA in clinical
practice.
The accuracy of the entire classication derived from
the decision tree analysis was 0.87 for the AUROC (0.83
for the validation data) and 0.17 for the classication error
rate (0.18 for the validation data) for training and validation.
AUROC has good accuracy at ≥0.830). Furthermore, a classi-
cation error rate of approximately 0.2 has been commonly
reported31,32). These results were judged to be superior to the
reference data. Therefore, similar results are likely to be
obtained when this CPR is practiced on other samples of the
population. The positive and negative likelihood ratios for
CPR in the present study were 2.6 and 0.1, respectively. The
change in post-test probability was slight when the positive
likelihood ratio ranged from 2 to 5. The change in post-test
probability was moderate when the negative likelihood ratio
ranged from 0.1 to 0.233). Thus, a negative CPR (a VAS score
Fig. 2. Decision tree model predicting the likelihood of improvement in JKOM after 5 months. ‘Positive’ indicates
non-improvement and ‘Negative’ indicates improvement.
KOA, knee osteoarthritis; CPR, clinical prediction rule; VAS, visual analog scale; BMI, body mass index; 5mWT, 5-m walk test;
JKOM, Japan Knee Osteoarthritis Measure
Table 2. More about CPRs
Positive CPR
VAS <27 mm (CPR 1)
VAS ≥27 mm, bilateral KOA, 5mWT ≥4.83 sec (CPR 2)
VAS ≥27 mm, bilateral KOA, 5mWT <4.83 sec, JKOM <31 points (CPR 3)
VAS ≥27 mm, unilateral KOA, BMI ≥22.9 kg/m2 (CPR 5)
Negative CPR
VAS ≥27 mm, bilateral KOA, 5mWT <4.83 sec, JKOM ≥31 points (CPR 4)
VAS ≥27 mm, unilateral KOA, BMI <22.9 kg/m2 (CPR 6)
Average of 10-fold crossing. Positive CPR: non-improvement, negative CPR:
improvement.
CPR, clinical prediction rule; VAS, visual analog scale; KOA, knee osteoarthri-
tis; 5mWT, 5 m walk test; JKOM, Japan Knee Osteoarthritis Measure; BMI,
body mass index
A Clinical Prediction Rules for Patients with Knee Osteoarthritis
of ≥27 mm, bilateral KOA, 5mWT of <4.83 sec, JKOM of
≥31 points, or a VAS score of ≥27 mm, unilateral KOA, BMI
of <22.9 kg/m2) is considered to rule out the possibility of an
event occurring (to derive an improvement group). Increased
walking speed in patients with moderate to severe KOA has
reportedly led to improved QOL34). CPR with a VAS score
of ≥27 mm, bilateral KOA, 5mWT of <4.83 sec, and JKOM
of ≥31 points are likely to improve if walking speed is main-
tained, even in patients with moderate symptoms. Further-
more, a greater number of arthropathies and higher BMI affect
activity limitation and arthropathy severity35,36). In CPR with a
VAS score greater than 27 mm, unilateral KOA, and BMI less
than 22.9, the symptoms are highly likely to improve because
they are mild and there is no metabolic abnormality.
This study used physical function and medical informa-
tion, such as VAS and 5mWT, to screen patients who should
be considered more carefully for exercise therapy. We sug-
gest that screening will allow us to develop a treatment plan
for such patients with KOA that includes options other than
general exercise therapy. This may contribute to the determi-
nation of patients with KOA and the decision to apply treat-
ment options. The clinical utility of this CPR is that when
CPR is positive, specic individualized measures should be
taken, including measures other than exercise therapy. On the
other hand, when CPR is negative, treatment options (local
pharmacotherapy, self-management programs, weight con-
trol, cognitive-behavioral therapy) following versatile guide-
lines should be considered as an option37).
This study has four limitations: The rst is that 40
of the subjects who had difculty following up completed
treatment earlier than 5 months. It is possible that some
patients with KOA had improved by more than 8 points in
JKOM within 5 months and that there were more cases in
the improved group than in the sample of this study. There-
fore, the decision to apply inspection characteristic values
such as positive likelihood ratios and negative likelihood
ratios should be reserved. Second, in this study, we were
able to conduct the training and validation tasks of the CPR
but were not able to validate them with test data. Usually,
when a model is created, training data, validation data, and
test data must be prepared. Therefore, clinicians understand
this and should use the ndings of this study. To utilize the
CPR, it is necessary to collect a separate sample, validate
cross-validation, and conduct a budget impact analysis.
Third, physical therapy intervention times have not been
standardized. The possibility that the time of physical ther-
apy intervention has an interactive effect on the improve-
ment of JKOM must be considered. In the case of clinical
application, it is necessary to consider whether the subjects
are similar to those in the present study (approximately 60
minutes per week). Fourth, the guarantee of external validity
is not sufcient. The age range (73.6 years) and BMI (24.0)
of the subjects in this study are similar to those reported in
previous studies (age: prevalence increases from 70 years;
BMI: 23.0)38). However, the sex ratio in the present study
was 20% men and 80% women, while in the previous study,
it was 28% men and 72% women38), which is somewhat dif-
ferent. Despite some limitations of this study, it is signicant
that we derived a CPR that predicts HRQOL at 5 months for
patients with KOA undergoing conservative treatment. We
believe that this is the rst process to utilize this technology
in actual clinical practice and will contribute to the develop-
ment of research areas in the eld of KOA.
Conclusion
This study examined the derivation of a CPR that pre-
dicts HRQOL at 5 months for patients with KOA undergo-
ing conservative treatment. The accuracy of the CPR was
good for both AUROC and the classication error rate. This
CPR was an inspection characteristic that could exclude the
possibility of an event occurring due to a negative result.
However, since the results of this study represent the rst
process of utilizing CPR in actual clinical practice, its appli-
cation should be kept in mind.
Acknowledgments: We thank the facilities that cooper-
ated in this multi-center joint research.
Funding: This work was supported by the Grants-in-
Aid for Scientic Research of Japan Society for the Promo-
tion of Science (Japan), Grant Number 18K17738.
Conicts of Interest: The authors declare no conict
of interest.
References
1) Badley EM, Wang PP: Arthritis and the aging population: Projec-
tions of arthritis prevalence in Canada 1991 to 2031. J Rheuma-
tol. 1998; 25: 138–144.
2) Wang SX, Ganguli AX, et al.: Healthcare resource utilization and
costs by age and joint location among osteoarthritis patients in a
privately insured population. J Med Econ. 2017; 20: 1299–1306.
3) McAlindon TE, Bannuru RR, et al.: OARSI guidelines for the
non-surgical management of knee osteoarthritis. Osteoarthritis
Cartilage. 2014; 22: 363–388.
4) Walker-Bone K, Javaid K, et al.: Regular review: Medical man-
agement of osteoarthritis. BMJ. 2000; 321: 936–940.
5) Fernandes L, Hagen KB, et al.: EULAR recommendations for
the non-pharmacological core management of hip and knee
osteoarthritis. Ann Rheum Dis. 2013; 72: 1125–1135.
6) Zhang W, Nuki G, et al.: OARSI recommendations for the man-
agement of hip and knee osteoarthritis: Part III: Changes in evi-
dence following systematic cumulative update of research pub-
lished through January 2009. Osteoarthritis Cartilage. 2010; 18:
476–499.
7) Hochberg MC, Altman RD, et al.: American College of
Rheumatology: American College of Rheumatology 2012
recommendations for the use of nonpharmacologic and phar-
macologic therapies in osteoarthritis of the hand, hip, and knee.
Arthritis Care Res (Hoboken). 2012; 64: 465–474.
Yamashina, et al.
8) Gay C, Chabaud A, et al.: Educating patients about the benets
of physical activity and exercise for their hip and knee osteo-
arthritis. Systematic literature review. Ann Phys Rehabil Med.
2016; 59: 174–183.
9) Akai M, Doi T, et al.: An outcome measure for Japanese people
with knee osteoarthritis. J Rheumatol. 2005; 32: 1524–1532.
10) Liu L, Ishijima M, et al.: Disability for daily living is a predictor
for joint replacement in patients with end-stage knee osteoarthri-
tis. J Bone Miner Metab. 2014; 32: 192–199.
11) Desmeules F, Dionne CE, et al.: Determinants of pain, functional
limitations and health-related quality of life six months after total
knee arthroplasty: Results from a prospective cohort study. BMC
Sports Sci Med Rehabil. 2013; 5.
12) Doi T, Akai M, et al.: Effect of home exercise of quadriceps on
knee osteoarthritis compared with nonsteroidal antiinammatory
drugs: A randomized controlled trial. Am J Phys Med Rehabil.
2008; 87: 258–269.
13) Atalay SG, Durmus A, et al.: The effect of acupuncture and
physiotherapy on patients with knee osteoarthritis: A randomized
controlled study. Pain Physician. 2021; 24: E269–E278.
14) Bachmann LM, Haberzeth S, et al.: The accuracy of the Ottawa
knee rule to rule out knee fractures: A systematic review. Ann
Intern Med. 2004; 140: 121–124.
15) Altman R, Asch E, et al.: Development of criteria for the
classication and reporting of osteoarthritis. Classication of
osteoarthritis of the knee. Diagnostic and therapeutic criteria
committee of the American Rheumatism Association. Arthritis
Rheum. 1986; 29: 1039–1049.
16) Amano T: Evaluating the diagnostic accuracy of a screening tool
for low physical activity in independently ambulating adults with
knee osteoarthritis: A prospective cohort study. Physiother Res
Int. 2024; 29: e2041.
17) Kellgren JH, Lawrence JS: Radiological assessment of
osteo-arthrosis. Ann Rheum Dis. 1957; 16: 494–502.
18) Stevens M, van den Akker-Scheek I, et al.: A Dutch translation
of the self-efcacy for rehabilitation outcome scale (SER): A rst
impression on reliability and validity. Patient Educ Couns. 2005;
58: 121–126.
19) Katoh M, Isozaki K: Reliability of isometric knee extension
muscle strength measurements of healthy elderly subjects made
with a hand-held dynamometer and a belt. J Phys Ther Sci. 2014;
26: 1855–1859.
20) Yamasaki H, Hasegawa T: The measurement of the isometric
knee extension muscle strength by hand-held dynamometer us-
ing the belt for the xation. Journal of Kochi Rehabilitation In-
stitute. 2002; 3: 7–11. (in Japanese)
21) Fritz JM, Delitto A, et al.: An examination of the selective tissue ten-
sion scheme, with evidence for the concept of a capsular pattern of
the knee. Phys Ther. 1998; 78: 1046–1056, discussion 1057–1061.
22) Matsumura M, Usa H, et al.: Age and gender differences in lower
extremity range of motion and muscle strength: A cross-sectional
study of males and females, ages 20-70. Rigakuryoho Kagaku.
2015; 30: 239–246. (in Japanese)
23) Muñoz-Bermejo L, Adsuar JC, et al.: Test-retest reliability of
Five Times Sit to Stand Test (FTSST) in adults: A systematic
review and meta-analysis. Biology (Basel). 2021; 10: 510.
24) Amano T: [Measurement error in motor function tests in healthy
young adults] Kenjyo jyakunensha ni okeru undoukinouken-
sa no sokuteigosa. Tokoha daigaku hokeniryougakubu kiyou.
2021;12:1–7. (in Japanese)
25) Amano T, Suzuki N: Minimal detectable change for motor func-
tion tests in patients with knee osteoarthritis. Prog Rehabil Med.
2018; 3: 20180022.
26) Jung Y, Hu J: A K-fold Averaging cross-validation procedure. J
Nonparametr Stat. 2015; 27: 167–179.
27) Sharma L, Cahue S, et al.: Physical functioning over three years
in knee osteoarthritis: Role of psychosocial, local mechanical,
and neuromuscular factors. Arthritis Rheum. 2003; 48: 3359–
3370.
28) Tanaka S, Nishigami T, et al.: Identifying participants with knee
osteoarthritis likely to benet from physical therapy education
and exercise: A hypothesis-generating study. Eur J Pain. 2021;
25: 485–496.
29) Deyle GD, Gill NW, et al.: Knee OA: Which patients are unlikely to
benet from manual PT and exercise? J Fam Pract. 2012; 61: E1–E8.
30) Nahm FS: Receiver operating characteristic curve: Overview and
practical use for clinicians. Korean J Anesthesiol. 2022; 75: 25–36.
31) Sheets CG, Zhang L, et al.: Ten-year retrospective study of the
effectiveness of quantitative percussion diagnostics as an indica-
tor of the level of structural pathology in teeth. J Prosthet Dent.
2020; 123: 693–700.
32) Lee JH, Jeong SN, et al.: Predictive data mining for diagnosing
periodontal disease: The Korea National Health and Nutrition
Examination Surveys (KNHANES V and VI) from 2010 to 2015.
J Public Health Dent. 2019; 79: 44–52.
33) Jaeschke R, Guyatt GH, et al.: Users’ guides to the medical liter-
ature: III. How to use an article about a diagnostic test. B. What
are the results and will they help me in caring for my patients?
JAMA. 1994; 271: 703–707.
34) Ohmine T, Demizu S, et al.: Improvement in gait speed affects
short-term improvement in activities of daily living in patients
with moderate and severe knee osteoarthritis. Prog Rehabil Med.
2024; 9: 20240002.
35) van Dijk GM, Veenhof C, et al.: CARPA Study Group: Prognosis
of limitations in activities in osteoarthritis of the hip or knee: A
3-year cohort study. Arch Phys Med Rehabil. 2010; 91: 58–66.
36) Bonakdari H, Jamshidi A, et al.: A warning machine learning al-
gorithm for early knee osteoarthritis structural progressor patient
screening. Ther Adv Musculoskelet Dis. 2021;13.
37) Bannuru RR, Osani MC, et al.: OARSI guidelines for the
non-surgical management of knee, hip, and polyarticular osteo-
arthritis. Osteoarthritis Cartilage. 2019; 27: 1578–1589.
38) Yoshimura N, Muraki S, et al.: Prevalence of knee osteoarthritis,
lumbar spondylosis, and osteoporosis in Japanese men and wom-
en: The research on osteoarthritis/osteoporosis against disability
study. J Bone Miner Metab. 2009; 27: 620–628.
