New algorithmic approach for easier and faster extended disability status scale calculation

Introduction

The clinical evaluation of disability in multiple sclerosis (MS) remains the cornerstone outcome measure in everyday practice. Although sophisticated, the traditional paper and pencil method for EDSS calculation (pEDSS) and its functional systems (FS) are the most widely used clinical outcome measures.^1,2 It was first introduced by John Kurtzke in 1955 as the Disability Status Scale (DSS), ³ then modified in 1983 as the EDSS, ⁴ and after that, a set of scoring rules (that depend on the results of a standard neurological examination called Neurostatus) were introduced in 1990. Neurostatus consists of three steps: (a) a standard neurological examination, including the assessment of ambulation (the Neurostatus subscores); (b) the assessment of the FS scores based on these subscores; and (c) the calculation of the EDSS step, depending on the FS scores, and the estimation of walking range and the need for walking support or assistance in everyday activities, evaluated with the ambulation score (AS). It includes detailed definitions and rules to clarify the scoring standards, but it also increases the complexity and time of assessments. ¹

A new algorithmic approach (aEDSS) was introduced for easier and faster EDSS practice. It consists of a set of color-coded tables and algorithms that facilitate and accelerate the estimation of neurological examination results, including the neurostatus subscores, FS scores, AS scores, and the final EDSS step. In this prospective study, we test the hypothesis that this new algorithmic approach (aEDSS) could be an easier and faster method and achieve results comparable to those of the traditional paper and pencil EDSS (pEDSS). ⁵

Materials and methods

The aEDSS algorithm

The aEDSS algorithm is a new method of visualization based on the rules provided in the “Neurostatus” for the standardized assessment of Neurostatus subscores, FS scores, the AS, and the final EDSS step (except for the visual and bowel and bladder subscores, which do not need correction in this new method). It consists of three sets of color-coded tables and algorithms.

The functional subscore algorithms

It consists of six subsets: visual, brainstem, pyramidal, cerebellar, sensory, bowel, and bladder.

Visual

Estimate the best corrected visual acuity in both eyes and follow the diagram in Figure 1 to get the final visual score from 0 to 4 without the need for correction (this is a significant difference from the traditional paper and pencil method, as it is easier and faster to calculate and will yield similar results).

OPEN IN VIEWER

Figure 1.

The visual score (corrected).

Brainstem

Calculate the seven subscores as follows: Green is 1, Yellow is 2, Orange is 3, Red is 4, and Brown is 5. The largest of those numbers is the Brainstem Score (Table 1).

OPEN IN VIEWER

Table 1.

The brain stem score (1–5).

Pyramidal

Examine the strength of all muscle groups and determine the lowest grade according to the British Medical Research Council (BMRC) and its distribution (Figure 2).

OPEN IN VIEWER

Figure 2.

The pyramidal score.

Cerebellar

Follow the diagram in Figure 3 to get the final cerebellar score.

OPEN IN VIEWER

Figure 3.

The cerebellar score.

Sensory

The sensory score is divided into superficial and deep sensations, as illustrated in Figures 4 and 5. The final sensory score is the higher of the two subscores; however, if the superficial sensation subscore is 4 and the deep sensation subscore is 4, the final score is 5, and if the superficial sensation subscore is 5 and the deep sensation subscore is 5, the final score is 6.

OPEN IN VIEWER

Figure 4.

Superficial sensation algorithm.

OPEN IN VIEWER

Figure 5.

Deep sensation algorithm.

Bowel and bladder (corrected from 1 to 5)

Calculate the four subscores as follows: green is 1, yellow is 2, orange is 3, and red is 4. The largest of those numbers is the bowel and bladder score. If the patient scores 4 in urinary urgency, incontinence, and bowel dysfunction, the score is 5. The final bowel and bladder score ranges from 0 to 5, without the need for correction (this is another difference from the traditional paper and pencil method, making it easier and faster to calculate and yielding almost the same results) (Table 2).

OPEN IN VIEWER

Table 2.

The bowel and bladder score corrected (1–5).

Cerebral FS and ambulation

Cerebral FS and ambulation were done in the same manner as in pEDSS.

Interpretation of the EDSS steps based on the seven FS scores

We can complete this task in three steps: first, determine the largest number of the seven FS scores and denote it with the symbol N; the final score will be N, N + 0.5, or N + 1. The second step is to identify how often the largest number is repeated (1, 2, or ≥ 3). The last step is to determine if the other FS scores (other than those taking the largest number) are 0, 1, or more than 1, as shown in Figures 6 and 7.

OPEN IN VIEWER

Figure 6.

Interpretation of the 7 FS to get the final function score.

OPEN IN VIEWER

Figure 7.

Examples of interpretation of the 7 FS to get the final function score.

One exception to the above rules is that if I have five FS scores of grade 2, the final score will be 5, not 3.5.

The final EDSS steps algorithm (based on the seven FS scores and the AS)

EDSS steps 1.0 to 5 are based on measures of impairment in seven FS, while EDSS steps 4.5 to 9.5 are based on walking and activities of daily living. If the patient is ambulant for less than 500 m but more than 300 m (score 4.5–5 on walking), we assess the seven FSs, and the higher number is the final EDSS score (see Figure 8).

OPEN IN VIEWER

Figure 8.

The final EDSS algorithm.

Results

The patient's characteristics are listed in Table 3. The median, range of FS grades, and final EDSS score calculated by pEDSS and aEDSS are all summarized in Table 4. Outstanding agreements were found between pEDSS and aEDSS regarding the total EDSS and FSs (Kappa > 0.81 for all) (Table 5). The mean time needed for calculation of the EDSS by using the aEDSS (16 ± 2.67 min) is significantly shorter than that for the pEDSS (31 ± 4.3 min), P < 0.0001, Figure 9.

OPEN IN VIEWER

Figure 9.

Comparison between aEDSS and pEDSS regarding the time (minutes) needed for calculation.

OPEN IN VIEWER

Table 3.

Patient characteristics.

Patient characteristics (n = 200)
Age (median, range), years 32 years (13–58)
Gender (female/male) 131 females/69 males
Disease duration (median, range), years 10.8 years (5.7–33.1)
Disease course (RRMS/SPMS/PPMS) 149 RRMS, 45 SPMS, 6 PPMS

RRMS: relapsing-remitting multiple sclerosis; SPMS: secondary progressive multiple sclerosis; PPMS: primary progressive multiple sclerosis.

OPEN IN VIEWER

Table 4.

The median (range) of FS grades, and total EDSS calculated by pEDSS and aEDSS.

	Visual	Brainstem	Pyramidal	Cerebellar	Sensory	Bladder	Cerebral	Total
pEDSS	0 (0–5)	0 (0–4)	2 (0–5)	0 (0–5)	0 (0–4)	0 (0–6)	0 (0–3)	3 (0–9)
aEDSS	0 (0–5)	0 (0–4)	2 (0–5)	0 (0–5)	0 (0–4)	0 (0–6)	0 (0–3)	3 (0–9)

OPEN IN VIEWER

Table 5.

Inter-rater agreements between pEDSS and aEDSS.

	Kappa	Standard error	95% CI
Total EDSS	0.89	0.023	0.84–0.93
visual	0.96	0.01	0.93–1
Brain stem	0.97	0.01	0.94–1
Pyramidal	0.96	0.01	0.94–0.99
Cerebellar	0.97	0.01	0.94–1
Sensory	0.97	0.01	0.95–1
Bowel and bladder	0.97	0.01	0.94–1

A multi-class confusion matrix was created, and the results showed that the overall accuracy (correct classification/total number of classifications) was 0.9 (180/200), or 90%, as shown in Table 6. The parameters sensitivity, specificity, and precision were calculated as true positive/(true positive + false negative), true negative/(true negative + false positive), and true positive/(true positive + false positive), respectively, as shown in Table 7.

OPEN IN VIEWER

Table 6.

The multi-class confusion matrix for pEDSS and aEDSS.

OPEN IN VIEWER

Table 7.

Sensitivity, specificity, and precision of various parameters in the multi-class confusion matrix for pEDSS and aEDSS.

	True positive	False negative	False positive	True negative	Precision	Sensitivity, Recall/True positive rates	Specificity
0	5	1	0	194	1	0.83	1
1	24	1	0	175	1	0.96	1
1.5	20	0	0	180	1	1	1
2	24	0	3	173	0.89	1	0.98
2.5	15	3	0	182	1	0.83	1
3	20	3	5	172	0.8	0.87	0.97
3.5	12	2	6	180	0.67	0.86	0.97
4	3	3	4	190	0.43	0.5	0.98
4.5	0	1	2	197	0	0	0.99
5	2	3	0	195	1	0.4	1
5.5	4	2	0	194	1	0.67	1
6	22	1	0	177	1	0.96	1
6.5	13	0	0	187	1	1	1
7	4	0	0	196	1	1	1
7.5	3	0	0	197	1	1	1
8	3	0	0	197	1	1	1
8.5	5	0	0	195	1	1	1
9	1	0	0	199	1	1	1

Discussion

EDSS is the cornerstone of the daily practice of MS. Despite this, EDSS is a bit complex and requires expert training to be able to calculate it accurately, as it consumes a lot of time in its calculation.

Several electronic EDSS calculators were developed to overcome these problems and are available as download applications.^8–10 Furthermore, some calculators are designed for patients to use directly to monitor their own clinical condition. ¹¹

Little information is known about the basis of these methods and their agreement with the traditional paper and pencil method. Most of them provide little information about the specific rules used to estimate the FS scores and EDSS steps.

Some only provide the EDSS step determination based on given FS scores and ambulation. None of these tools give a clear way to calculate the FS based on a complete neurological assessment. ⁵

In our new algorithmic approach, we provide an easy and straightforward method that relies on the same specific rules outlined in Neurostatus, but in a much simpler form, allowing for easier and faster calculation of the FS scores and the total EDSS, while simultaneously achieving excellent agreement with the traditional paper and pencil method.

Selection bias and observer bias are two systemic biases that were avoided in our study. ¹² To avoid selection bias, which occurs when the sample is not representative of the whole MS population, we attempted to involve patients with all degrees of disability (EDSS from 0 to 9) so as not to focus only on ambulant patients.

Regarding observer bias, it could result from variation between observers in administering pEDSS and aEDSS, or from an observer's reporting bias when one observer tends to underestimate or overestimate the results of either pEDSS or aEDSS. To account for observer variability, all the neurologists involved in this study were certified EDSS readers who were well-versed in administering both the pEDSS and the aEDSS. Concerning observer reporting bias, we had two neurologists on the same day, one for pEDSS and the other for aEDSS, with patients randomly assigned to do either pEDSS first with one neurologist, then shifted to the other for aEDSS, or vice versa, and the two neurologists were blind to each other's results.

Limitations: Our limitations included the difficulty of involving more patients with severe disabilities (EDSS ≥ 5.5) due to their limited mobility and visits to the clinic, so ambulation was calculated in the aEDSS in the same way as in the pEDSS, some inter-rater variability, especially at the lower end of the EDSS spectrum of scores, and presumably a lack of objective assessment of walking distance.

In conclusion, this study found that the new algorithm approach (aEDSS) is a faster, yet simpler way with excellent agreement compared to pEDSS. As a result, it can significantly enhance and facilitate the everyday practice and clinical trials that use the Neurostatus-EDSS as an outcome measure.