Validating a probability score (AI) based

Clinical Prediction Model using symptoms,

signs to predict diagnosis in Low Back Pain
Presenter- Dr Ajinkya Laxman Pingale
Junior Resident
ELMCH Lucknow
 Introduction
• Lower back ache is a significant cause of morbidity, especially in the elderly age groups. Hayashi et al 2004
showed that low back pain caused by fragility fractures in osteoporosis patients has been found to have the
most significant impact on mortality among the various kinds of fractures that occur in the elderly. This study
also showed that lower back ache is the most common symptom in Japanese patients above the age of 65
years signifying the high burden of disease in the elderly [1].

• It has been shown that in adolescents, the overall risk of low back pain is similar to adults, with
prevalence rates as high as 70% to 80% by 20 years of age (Jones et al 2005) [2].
• The causes are heterogenous, and the clinical parameter evaluation is the most important
component for the physicians while undertaking a workup of these patients. The clinical
parameters often help to differentiate the causes of low backache and serve as a handy tool in these
patients. However, one clinical parameter is not sufficient in itself to characterize these patients.
Therefore, assessment of multiple clinical parameters often becomes necessary to arrive at a
diagnosis that currently relies on radiological and investigational markers [3].

• The evaluation of patients with LBA using a clinical parameter-based risk scoring system is an
important modality in the system where advanced laboratory and radiological investigations are
not available. These clinical parameters can help to arrive at a diagnosis earlier by saving time and
resources. Timely initiation of therapy can help improve quality of life.
• Hence, we conducted this study with the primary objective of arriving at a Risk
Prediction Score using the probability distributions of patients based on their
clinical parameters which would help differentiate the patients with low back ache
according to the underlying etiology.
 Materials and method
• Study setting and Duration

• The study was conducted in the Department of Orthopedics at the Era's Medical College and
Hospital, Lucknow, UP, India over a duration of 24 months (February 2021 to February 2023).

• Sample size/Inclusion and exclusion criteria

• The inclusion criteria were all patients with lower back aches who presented in the Department.
Exclusion criteria were morbidly sick, moribund patients requiring ICU admission. Overall 155 patients who
have deemed it for inclusion and were willing to give written informed consent were taken up for analysis.
• Approval from the institutional ethics committee was taken before the start of the study and written informed
consent was taken from all the patients before inclusion in the study.

• Study flow

1. Patients were evaluated based on the inclusion/exclusion criteria. were included in the study after written
informed consent.

2. A comprehensive baseline demographic and clinical parameter assessment was done for these patients using
standard protocols for clinical evaluation.
3. An evaluation via the gold standard investigations (HLA B27, BMD studies, and MRI LS Spine) was done to
arrive at a final diagnosis for these patients and clinical parameters were evaluated against these findings.

4. A probability score for each of these clinical parameters was evaluated for each of the primary diagnoses.
• Total number of events seen in the study was taken and assessed for distribution across each category which
was then used to arrive at a standardized P score across all three categories (after normalizing the sample
distribution across groups) as well as in comparison to the gold standard measures.

1. A univariate analysis using the spearman correlation Rho coefficient was performed after codifying the
primary diagnosis in the study. Osteoporosis was kept as the baseline category and allocated code 0 while
Seronegative spondylitis was given a code 1 and Disc degenerative disease was allotted code 2.

2. A multivariate analysis using linear regression for the coded diagnosis was performed and significant factors
were identified.
3. Based on the distribution of probabilities and results from the univariate/multivariate analysis, a risk score
with the significant factors was calculated and patients have distributed basis the combined score.

4. The results were analyzed basis of the statistical analysis plan and compiled.
 Result
• It was seen that out of 155 patients with lower back aches presenting in the department in the study period,
most of the patients had disc degenerative disease (47.10%) followed by Osteoporosis (29.68%) and
Seronegative spondylitis (23.23%). The average age was highest in osteoporosis patients (63.91±15.27 years)
compared to the Disc degenerative disease group (45.78±9.73 years) and Seronegative spondylitis (SS) group
(23.19±3.66 years) (P<0.0001, F=5.118). There were 86.11% of male patients in the Seronegative spondylitis
(SS) group which was higher compared to the other two groups (P=0.0003)
• An evaluation of the clinical profile was done for all the patients. It was seen that in the seronegative
spondylitis (SS) group of patients, all (100%) patients had polyarthralgia and 97.22% had morning stiffness
>30 minutes followed by modified Schoeber's test positive seen in 91.67% of patients. 86% of these patients
had Faber's test positive while 94.44% of the patients had a response to NSAIDs present. In the osteoporosis
group of patients, polyarthralgia was seen in 76% of patients and X-ray suggestive of OP findings was seen in
69.57% of patients. Small joint involvement was seen in 45.65% of patients.
• In the Disc degenerative disease (DDD) group of patients, 94.52% had radiculopathy followed by a history of
heavy lifting seen in 79.45% of patients. 78% of these patients also had Bragard's sign positive and SLRT was
seen to be positive in 75.34% of patients with Disc degenerative disease .
• A probability score for each parameter was evaluated and results have been summarized in Table 3. We found
that the P scores for radiculopathy, heavy weightlifting, SLRT positive, and Bragard's sign were very high in
the Disc degenerative disease group. The P score for the small joint involvement was highest in the
osteoporosis group. P Scores for Morning stiffness, FABER's test, and Modified Schoeber's test were highest
in the ankylosing spondylitis group. P scores for polyarthralgia, Response to NSAIDs, and X-ray findings
were suggestive of group findings. This meant that combined P scores were higher for Seronegative
spondylitis and OP and therefore, could be used effectively to exclude Disc degenerative disease in the
patients.
• We compared the P scores to the gold standard investigation for each group and found that when compared to
HLA B27 which is the gold standard for ankylosing spondylitis, modified Schoeber's test, Faber's test, and
morning stiffness >30 mins had the highest/comparable P scores. Only small joint involvement had a higher P
score of 0.7241 when compared to bone mineral studies in Osteoporosis. Radiculopathy (0.98571), history of
heavy weightlifting (0.89231), and Bragard's sign positivity (0.89063) were closest to MRI in disc degenerative
diseases.

• Univariate analysis for clinical parameters in the ascertaining primary cause of LBA was done and we
saw that Polyarthralgia, small joint involvement, response to NSAIDs, and Xray findings were negatively
correlated with the diagnosis suggesting that the occurrence was lowest in Disc degenerative disease group when
compared to the Seronegative spondylitis and OP groups. Faber test, Modified Schoeber's test, and morning
stiffness >30 mins were associated with Seronegative spondylitis mostly and therefore, the results were
significant on a P score analysis and association analysis while being ambiguous (statistically not significant) on
a correlation matrix
• . Heavy weightlifting, Radiculopathy, SLRT, and Bragard's sign were seen to be positively correlated with the
diagnosis as suggested by the fact that their prevalence was highest in Disc degenerative disease patients. The
results were significant statistically.

• A multivariate linear regression model was plotted using significant factors on univariate analysis
and it was seen that Poly arthralgia, radiculopathy, morning stiffness >30 mins, Bragard's sign, and Faber's
test were significant clinical parameters even on multivariate analysis for their respective primary diagnosis.
Polyarthralgia was statistically significant in terms of excluding Disc degenerative disease. Radiculopathy
and Bragard's test were significant clinical parameters for Disc degenerative disease. Morning Stiffness >30
mins and Faber's test were significant parameters for the seronegative spondylitis patients.
• Based on the distribution of probabilities and results from the univariate/multivariate analysis, a risk score
with the significant factors was calculated and patients have distributed basis the combined score. The final
risk score equation was -

• (Risk score = 0.693 - 0.166 x Polyarthralgia + 0.862 x Radiculopathy + 0.47 x Morning Stiffness + 0.111
x Bragard sign - 0.167 x Faber's test) (where the presence of a clinical parameter was accorded a score of
1)

• Basis the above equation, we found that (Figure 5) –

1. Disk degenerative disease had the highest risk score of 1.3 and above.

2. Seronegative spondylarthritis was associated with a risk score of 0.8-1.1

3. Osteoporosis had the lowest risk score for the given clinical parameters – 0.4-0.7.
 Discussion
• The importance of clinical factors in the diagnosis and management of lower back aches has been highlighted
in studies across the world. Karoli et al (2022) showed that seronegative spondyloarthritis is characterized by
inflammatory back pain, radiographic sacroiliitis, excess spinal bone formation, and a high prevalence of
HLA–B27.[4] The severity of arthralgia, stiffness, and limited flexibility varies widely among patients and
over the course of illness. They also showed that 72% were males and 28% were females. The mean age of
participants was 46 ± 12 years. These findings were in line with what was seen in our study.
• Chester J et al (2022) have shown that compression of the disc with bulging in disc degenerative disease is
often associated with radiculopathy into the posterior leg and dorsal foot. This is one of the most important
findings in such patients [5].
• Rotta et al (2020) have shown that small joint involvement is common in osteoporotic patients with pre-
existing inflammatory arthritis. Osteoporosis is known to impact the vertebral column and long bones mostly
and manifests as a fracture, concomitant arthritis like rheumatoid Arthritis (RA), psoriatic arthritis (PsA), and
seronegative Spondylarthritis (SS) often lead to small joint involvement in these patients.
• Naidu et al (2015) worked on a clinical criterion for early diagnosis of AS. The key clinical features which
were a part of the proposed criteria for inflammatory back pain in young to middle-aged adults with chronic
back pain were Morning stiffness of at least 30 minutes duration, Improvement of back pain with exercises
but not with rest,
• The study highlighted the role of clinical parameters in arriving at a risk-scoring system that will help ensure
timely management and avoid the unnecessary need for high-end investigations for confirming the diagnosis.
Polyarthralgia, Radiculopathy, Morning Stiffness, Bragard sign, and Faber's test were the significant
clinical parameters in multivariate analysis. There have been attempts at clinical feature-based modeling
before, as per various international studies.
• None of the clinical models developed thus far can sufficiently predict and differentiate the risk of various
causes of lower back aches. Therefore, the current model provides a good starting point for further evaluation
based on clinical findings and thus, needs to be tested and validated further for more optimization of the
same.
Thank You
 Conclusion
• We concluded that disc degenerative disease was the most common cause of lower back aches in patients
visiting our center. We ascertained a clinical prediction score to diagnose the patients with low back aches.
The score relied on 5 major significant clinical parameters seen on multivariate analysis – Polyarthralgia,
Radiculopathy, Morning stiffness >30 min, Bragard sign, and Faber's test. The equation for the score was –

• Risk score = 0.693 - 0.166 x Polyarthralgia + 0.862 x Radiculopathy + 0.47 x Morning Stiffness +
0.111 x Bragard sign - 0.167 x Faber's test.

• Using this score, it was ascertained that the disk degenerative disease had the highest risk score of 1.3 and
above. Seronegative spondylarthritis was associated with a risk score of 0.8-1.1 and Patients with
osteoporosis had the lowest risk score for the given clinical parameters of 0.4-0.7.