Identifying disability level in multiple sclerosis patients in a U.S.-based health plan claims database

Abstract

Aims

In clinical trials, disability progression in multiple sclerosis (MS) is measured by the Kurtzke expanded disability status scale (EDSS), which is not captured in routine clinical care in the U.S. This study developed a claims-based disability score (CDS) based on the EDSS for assigning MS disability level in a U.S. claims database.

Methods

This retrospective cohort study of patients with MS in the U.S., utilized adjudicated health plan claims data linked to electronic medical records (EMRs) data. Patients were identified between 1 January 2012 and 31 December 2016 and indexed on the first date of MS diagnosis. The CDS was developed to assign disability level at baseline using claims and ambulatory EMR records observed over the 1-year baseline period. All-cause healthcare costs were assessed by baseline disability level to validate the CDS.

Results

In total, 45,687 patients were identified in claims (full sample) and 1,599 linked to EMR (core sample). Over half of patients in both samples were classified with mild disability at baseline. Adjusted healthcare costs in patients with moderate and severe disability were 15% (p<.0001) and 20% higher, respectively, than in patients with mild disability at baseline in the full sample. Disease-modifying therapy (DMT) costs accounted for 89%, 82%, and 78% of outpatient pharmacy costs in patients with mild, moderate, and severe disability, respectively.

Conclusions

The CDS is the first claims-based measure of MS disability utilizing data from EMR. This novel measure advances the opportunity to examine outcomes by disability accumulation in the absence of standard markers of disease progression. Although formal validation of the CDS was not possible due to lack of available EDSS in the EMR, the economic burden results align with prior publications and show that healthcare costs increase with increasing disability. Future validation studies of the CDS are warranted.

Keywords:

• Multiple sclerosis
• health care costs
• disability level
• expanded disability status scale
• disease-modifying therapy

JEL CLASSIFICATION CODES:

• I19
• I30

Background

Multiple sclerosis (MS) is the most common cause of neurological disability in adults of working age, and is characterized by inflammation, demyelination, gliosis, and axonal destruction throughout the central nervous system^1,2. In 2016, an estimated 2.2 million people worldwide had MS, corresponding to a prevalence of 30.1 cases per 100,000 population³. In the U.S., the incidence of MS is estimated to be 2.0 cases per 100,000 person-years in men and 3.6 in women⁴. MS is a very costly chronic disease based on prior analyses using privately insured claims databases, with direct and indirect costs (including costs of disease-related absences from work, short- and long-term disability) that approximately comprised 77% (range 64–91%) and 23% (range 9–36%) of total healthcare costs, respectively^5,6.

MS disability progression is most commonly measured by the expanded disability status scale (EDSS) in clinical trials⁷. The EDSS is a clinician-based instrument that includes assessment of seven functional systems (plus “other”) including pyramidal, cerebellar, brainstem, bowel and bladder, sensory, visual, and cerebral. The score is reported in half point increments from 0 (with no disability due to MS) to 10 (death due to MS) based on neurological and physical examinations⁸. Other important EDSS cut-offs include: 6 (requires a walking aid such as a cane or crutch), 7 (restricted to wheelchair), 8 (restricted to bed/chair), and 9 (confined to bed). Even while the EDSS is the gold standard in measuring disability in the clinical trial setting, it is not typically used in routine clinical care in the US. As such, it is not reliably found in secondary data sources used for health-outcomes research, such as EHRs or administrative claims, making it difficult to capture the economic burden of advancing MS disease. As such, this is still an evidence gap, with limited literature on methods to assign or approximate MS disability levels in administrative healthcare claims data. A prior study by Munsell et al. developed a claims-based measure as an indicator of disease status based on healthcare costs⁹, however, the results were not corroborated with clinical measures of disability, such as the EDSS.

Establishing the level of MS disability progression in real-world evidence studies presents researchers with a challenge, as the majority of available claims and electronic medical record (EMR) databases do not routinely contain EDSS score. Indeed, per our initial estimate, less than 1% of EMR records contained the EDSS number. To overcome this limitation, this study developed a claims-based disability score (CDS) to characterize disability status using the eight functional systems that comprise the EDSS using EMR data linked to adjudicated healthcare claims, with the guidance of physicians who specialize in the treatment of MS. This methodology leverages both claims and narrative EMR information to enhance prior efforts that have used claims only and allows MS patients characterization by disability level in real world settings.

Methods

This was a retrospective cohort study of patients with MS in the U.S. identified between 1 January 2012 and 31 December 2016. Disability level was characterized by observation of proxies for ambulation/functional status in claims (e.g. wheelchair claim, diagnosis of urinary incontinence), and vitals/problem lists in the EMR (e.g. visual disturbance, tactile sensation). Patients were stratified into approximate disability levels, consistent with their observed characteristics during the 1-year pre-index period and as aligned with the underlying continuum of disability as represented in the EDSS. Demographic information was collected at index date, and included age, sex, geographic region, and payer type. Clinical characteristics were collected during the 1-year pre-index period, and included the Charlson comorbidity index (CCI), common comorbidities, MS-related medication use, MS relapse, and MS-related symptoms and secondary conditions. MS relapse was defined as having an inpatient hospitalization with a primary diagnosis of MS or having a claim for an oral or intravenous corticosteroid within seven days after an MS-related outpatient visit¹⁰. The study utilized secondary data from the IQVIA PharMetrics Plusⁱ database which was linked to IQVIA’s Ambulatory EMR (AEMR) data.

Patient selection

The study consisted of two main study cohorts: the full sample, which comprised all eligible MS patients identified in PharMetrics Plus; and the core sample, which included MS patients identified in PharMetrics Plus who linked to AEMR data.

The full sample included patients with at least two outpatient claims (≥30 days apart) with a diagnosis code for MS or ≥1 inpatient claim with a primary diagnosis of MS observed between 1 January 2012 and 31 December 2016. The date of the first MS diagnosis occurring during this window was the index date. Patients had ≥1 year of continuous health plan enrollment with pharmacy and medical benefits prior to and after the index date, were ≥20 years of age on the index date and had no data quality issues in the claims database. The core sample consisted of a subset of patients from the full sample who linked to the AEMR database and had at least one diagnosis record of MS or evidence of disease-modifying therapy (DMT) in the AEMR between 1 January 2011 and 31 December 2017. Patients in the core sample were also required to have at least one visit record in the AEMR database within one year prior to the index date to assess pre-index disability.

Development of EDSS-based disability criteria

The CDS was developed in a series of steps, with guidance from four medical experts (MEs); three of whom are physicians specializing in MS treatment, and one of whom is a medical director at the sponsoring institution. As a first step, an initial list of broad terms related to MS symptomology and the EDSS criteria was developed. These terms were then searched for in two tables of the AEMR database: the problem list and the vitals tables and the list was further refined based on the data. The problem list table contains all diagnoses recorded by a physician during a clinical encounter, while the vitals table contains data on patient vitals, timed walk results, and orders for durable medical equipment (DME).

After finalizing the list, terms were mapped to EDSS and functional system score (FSS) categories which included: bowel and bladder, brainstem, cerebellar, cerebral, pyramidal, sensory, visual function, other. Three MEs then independently assigned a severity score to each term, based on their clinical opinion. Scores ranged from 1 (mild) to 3 (severe). In cases of disagreement between MEs, the final severity score for a term was established based on a majority ruling.

Next, terms for diagnoses were mapped to International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM), International Classification of Diseases, Tenth Revision, Clinical Modification (ICD-10-CM), terms related to procedures and DME were mapped to Current Procedural Terminology (CPT). Three MEs independently assigned a severity score to each clinical code, based on code descriptions and their clinical opinion. As previously, a majority ruling determined the final score for each code in cases of disagreement.

After identifying the two study populations, the AEMR records and claims for each patient were assessed for the occurrence of the terms/billing codes occurring during the 1-year pre-index period. The CDS at baseline was defined using the following hierarchy:

• Severe: Defined as having ≥1 EDSS-related symptom with severity score = 3 in any functional system;

• Moderate: Defined as having ≥1 EDSS-related symptom with severity score = 2 in any functional system, or having ≥2 functional systems with severity score = 1;

• Mild: Defined as having only one EDSS-related symptom with severity score = 1 or having no EDSS-related symptoms observed during the measurement period.

For patients with additional years of follow-up available, the disability level was re-assessed at year 1 and year 2 post-index. Values could not regress to a lower disability level. In other words, if a patient’s disability level was considered moderate at baseline, the patient could not be considered mild at year one.

Finally, as an informal validation of the CDS, total healthcare costs accrued during the 1-year post-index period was calculated for each patient in the full sample to assess whether the expected trend of cost increasing with increasing disability level was observed. Total healthcare costs included medical and outpatient pharmacy costs, including DMT infusions. The proportion of outpatient pharmacy costs attributable to DMTs was also assessed. In patients with additional years of follow-up, healthcare costs were calculated in each year after disability level was established. Unadjusted costs were calculated by summing the allowed amount on all claims occurring during the 1-year post-index period. All costs were adjusted to 2018 dollars using the healthcare component of the Consumer Price Index.

Statistical methods

Chi-square tests were used to compare categorical demographic and clinical variables by disease severity. For continuous measures, t-test and analysis of variance (ANOVA) were used to make comparisons by disease severity.

A generalized linear model with gamma distribution and log link function^11,12 was developed to assess differences in cost by disability level. The outcome was total all-cause healthcare costs accrued over the fixed one-year post-index period. Independent variables in the model included: baseline disability level, age category, sex, region, CCI, pre-index use of DMTs, pre-index occurrence of MS relapse, and the log of pre-index total all-cause healthcare cost. Wald chi-square tests were used to compare costs between patients with mild versus moderate or severe disability. The adjusted mean healthcare costs in patients with mild, moderate, and severe MS disability were estimated using the recycle prediction method¹³ and the standard error of the means were calculated using bootstrapping with 1,000 repetitions¹⁴.

Results

A total of 45,687 and 1,599 eligible patients were included in the full sample and core sample, respectively (Figure 1). Patients in the full sample had a mean age of 49 and 76% of the study population was female (Table 1). Most reflected mild disability at index (mild: 66%, moderate: 22%, severe: 13%). The core sample had similar baseline characteristics as the full sample (mean age: 49 years; 78% female) with a slightly higher proportion of patients having moderate disability per the CDS (28%) and lower proportion having mild disability (58%) (Table 2). The remaining results reported are for the full sample.

Figure 1. Patient attrition.

Table 1. Characteristics of the full sample.

	Full sample patients		Patients with 1-year post-index eligibility
			Mild disability		Moderate disability		Severe disability		p Value
	(N = 45,687)		(n = 29,972)		(n = 9,916)		(n = 5,799)
Age group (n, %)
20–34 years	5,595	12.2%	3,914	13.1%	1,095	11.0%	586	10.1%	<.0001
35–44 years	10,505	23.0%	7,351	24.5%	2,084	21.0%	1,070	18.5%
45–54 years	14,901	32.6%	9,951	33.2%	3,219	32.5%	1,731	29.9%
55–64 years	12,367	27.1%	7,575	25.3%	2,924	29.5%	1,868	32.2%
65+ years	2,319	5.1%	1,181	3.9%	594	6.0%	544	9.4%
Age (continuous)
Mean	48.5		47.7		49.4		50.9		<.0001
SD	11.1		10.9		11.2		11.8
Median	49		48		50		52
Gender (n, %)
Male	10,985	24.0%	7,082	23.6%	2,527	25.5%	1,376	23.7%	.0007
Female	34,702	76.0%	22,890	76.4%	7,389	74.5%	4,423	76.3%
Geographic region (n, %)
Northeast	13,055	28.6%	8,606	28.7%	2,828	28.5%	1,621	28.0%	<.0001
Midwest	14,377	31.5%	9,562	31.9%	2,996	30.2%	1,819	31.4%
South	12,441	27.2%	7,944	26.5%	2,893	29.2%	1,604	27.7%
West	4,992	10.9%	3,261	10.9%	1,065	10.7%	666	11.5%
Unknown	822	1.8%	599	2.0%	134	1.4%	89	1.5%
Payer type (n, %)
Commercial/self-insured	43,212	94.6%	28,517	95.1%	9,420	95.0%	5,275	91.0%	<.0001
Medicaid	1,420	3.1%	825	2.8%	286	2.9%	309	5.3%
Medicare	734	1.6%	402	1.3%	153	1.5%	179	3.1%
Unknown	321	0.7%	228	0.8%	57	0.6%	36	0.6%
Index year (n, %)
2012	32,111	70.3%	21,013	70.1%	6,981	70.4%	4,117	71.0%	<.0001
2013	4,822	10.6%	3,409	11.4%	904	9.1%	509	8.8%
2014	3,267	7.2%	2,109	7.0%	717	7.2%	441	7.6%
2015	3,480	7.6%	2,293	7.7%	714	7.2%	473	8.2%
2016	2,007	4.4%	1,148	3.8%	600	6.1%	259	4.5%
Charlson comorbidity index (CCI) (n, %)
0	31,982	70.0%	22,839	76.2%	6,256	63.1%	2,887	49.8%	<.0001
1	7,737	16.9%	4,673	15.6%	1,926	19.4%	1,138	19.6%
2	3,613	7.9%	1,673	5.6%	1,017	10.3%	923	15.9%
3	1,229	2.7%	431	1.4%	406	4.1%	392	6.8%
4+	1,126	2.5%	356	1.2%	311	3.1%	459	7.9%
Mean	0.5		0.4		0.7		1.1		<.0001
SD	1.1		0.9		1.2		1.6
Median	0		0		0		1
Common comorbid conditions (n, %)
Chronic pain/fibromyalgia	5,588	12.2%	2,625	8.8%	1,848	18.6%	1,115	19.2%	<.0001
Diabetes	3,875	8.5%	2,113	7.0%	974	9.8%	788	13.6%	<.0001
Dyslipidemia	12,793	28.0%	7,600	25.4%	3,162	31.9%	2,031	35.0%	<.0001
Hypertension	12,402	27.1%	7,142	23.8%	3,130	31.6%	2,130	36.7%	<.0001
Osteoarthritis	19,582	42.9%	10,553	35.2%	5,934	59.8%	3,095	53.4%	<.0001
Smoking or history of smoking	3,962	8.7%	2,183	7.3%	1,013	10.2%	766	13.2%	<.0001
Thyroid disease	5,826	12.8%	3,473	11.6%	1,471	14.8%	882	15.2%	<.0001
MS-related medication use (n, %)
Any disease-modifying therapy (DMT) use	21,487	47.0%	14,701	49.0%	4,427	44.6%	2,359	40.7%	<.0001
Alemtuzumab	1	0.0%	0	0.0%	1	0.0%	0	0.0%	.3191
Daclizumab	1	0.0%	0	0.0%	1	0.0%	0	0.0%	.3184
Dimethyl fumarate	148	0.7%	113	0.8%	26	0.6%	9	0.4%	.0709
Fingolimod	1,332	6.2%	815	5.5%	330	7.5%	187	7.9%	<.0001
Glatiramer acetate	8,322	38.7%	5,626	38.3%	1,774	40.1%	922	39.1%	.0907
Interferon b-1a	8,434	39.3%	6,043	41.1%	1,582	35.7%	809	34.3%	<.0001
Interferon b-1b	2,168	10.1%	1,477	10.0%	457	10.3%	234	9.9%	.8308
Mitoxantrone	27	0.1%	9	0.1%	10	0.2%	8	0.3%	.0002
Natalizumab	2,210	10.3%	1,224	8.3%	591	13.3%	395	16.7%	<.0001
Pegylated interferon b-1a	11	0.1%	9	0.1%	1	0.0%	1	0.0%	.7987
Teriflunomide	28	0.1%	25	0.2%	3	0.1%	0	0.0%	.0452
Rituximab	119	0.6%	53	0.4%	36	0.8%	30	1.3%	<.0001
No DMT use	24,200	53.0%	15,271	51.0%	5,489	55.4%	3,440	59.3%	<.0001
Baclofen	5,514	12.2%	2,270	7.7%	1,834	18.5%	1,410	24.3%	<.0001
Relapse
Patients with a relapse in inpatient or outpatient setting (n, %)	5,995	13.1%	2,640	8.8%	1,853	18.7%	1,502	25.9%	<.0001
Number of relapses in all patients
Mean	0.4		0.2		0.5		1.1		<.0001
SD	1.9		0.9		1.8		4.1
Median	0.0		0.0		0.0		0.0
MS-related symptoms and secondary conditions (n, %)
Anxiety	4,973	11.0%	2,593	8.8%	1,428	14.4%	952	16.4%	<.0001
Depression	7,650	17.0%	3,909	13.3%	2,085	21.0%	1,656	28.6%	<.0001
Eye symptoms	1,518	3.4%	601	2.0%	539	5.4%	378	6.5%	<.0001
Fatigue	10,783	23.9%	5,165	17.6%	3,359	33.9%	2,259	39.0%	<.0001
Fractures	142	0.3%	48	0.2%	56	0.6%	38	0.7%	<.0001
Irritable bowel syndrome	906	2.0%	474	1.6%	281	2.8%	151	2.6%	<.0001
Muscular weakness	3,504	7.8%	0	0.0%	2,330	23.5%	1,174	20.2%	<.0001
Osteoporosis	1,544	3.4%	709	2.4%	455	4.6%	380	6.6%	<.0001
Pyramidal symptoms	5,720	12.7%	0	0.0%	4,111	41.5%	1,609	27.7%	<.0001
Sensory problems	9,249	20.5%	4,644	15.8%	2,994	30.2%	1,611	27.8%	<.0001
Sleep disturbance	4,732	10.5%	2,362	8.0%	1,447	14.6%	923	15.9%	<.0001

Table 2. Characteristics of the core sample.

	Core sample patients		Patients with 1-year post-index eligibility
			Mild disability		Moderate disability		Severe disability		p Value
	(N = 1,599)
			(n = 931)		(n = 442)		(n = 226)
Age group (n, %)
20–34 years	160	10.0%	105	11.3%	38	8.6%	17	7.5%	<.0001
35–44 years	358	22.4%	241	25.9%	76	17.2%	41	18.1%
45–54 years	524	32.8%	288	30.9%	169	38.2%	67	29.6%
55–64 years	489	30.6%	268	28.8%	132	29.9%	89	39.4%
65+ years	68	4.3%	29	3.1%	27	6.1%	12	5.3%
Age (continuous)
Mean	49.3		48.2		50.4		51.3		<.0001
SD	10.7		10.6		10.7		10.6
Median	50		49		51		54
Gender (n, %)
Male	352	22.0%	201	21.6%	105	23.8%	46	20.4%	.5376
Female	1,247	78.0%	730	78.4%	337	76.2%	180	79.6%
Geographic region (n, %)
Northeast	764	47.8%	447	48.0%	225	50.9%	92	40.7%	.0003
Midwest	347	21.7%	222	23.8%	86	19.5%	39	17.3%
South	332	20.8%	184	19.8%	79	17.9%	69	30.5%
West	156	9.8%	78	8.4%	52	11.8%	26	11.5%
Payer type (n, %)
Commercial/self-insured	1,528	95.6%	903	97.0%	414	93.7%	211	93.4%	.0115
Medicaid	14	0.9%	7	0.8%	5	1.1%	2	0.9%
Medicare	57	3.6%	21	2.3%	23	5.2%	13	5.8%
Index year (n, %)
2012	1,277	79.9%	750	80.6%	357	80.8%	170	75.2%	.0123
2013	129	8.1%	87	9.3%	23	5.2%	19	8.4%
2014	89	5.6%	48	5.2%	28	6.3%	13	5.8%
2015	66	4.1%	29	3.1%	20	4.5%	17	7.5%
2016	38	2.4%	17	1.8%	14	3.2%	7	3.1%
Charlson comorbidity index (CCI) (n, %)
0	1,127	70.5%	718	77.1%	287	64.9%	122	54.0%	<.0001
1	249	15.6%	141	15.2%	73	16.5%	35	15.5%
2	132	8.3%	52	5.6%	48	10.9%	32	14.2%
3	50	3.1%	14	1.5%	16	3.6%	20	8.9%
4+	41	2.6%	6	0.6%	18	4.1%	17	7.5%
Mean	0.6		0.3		0.7		1.2		<.0001
SD	1.2		0.8		1.3		1.8
Median	0		0		0		0
Common comorbid conditions (n, %)
Chronic pain/fibromyalgia	189	11.8%	74	8.0%	72	16.3%	43	19.0%	<.0001
Diabetes	140	8.8%	61	6.6%	45	10.2%	34	15.0%	.0001
Dyslipidemia	494	30.9%	245	26.3%	167	37.8%	82	36.3%	<.0001
Hypertension	440	27.5%	221	23.7%	141	31.9%	78	34.5%	.0003
Osteoarthritis	688	43.0%	348	37.4%	237	53.6%	103	45.6%	<.0001
Smoking or history of smoking	145	9.1%	67	7.2%	44	10.0%	34	15.0%	.0008
Thyroid disease	243	15.2%	132	14.2%	64	14.5%	47	20.8%	.0403
MS-related medication use (n, %)
Any disease-modifying therapy (DMT) use	938	58.7%	576	61.9%	256	57.9%	938	58.7%	.0039
Dimethyl fumarate	2	0.2%	1	0.2%	0	0.0%	1	0.9%	.2863
Fingolimod	49	5.2%	23	4.0%	12	4.7%	14	13.2%	.0004
Glatiramer acetate	365	39.0%	221	38.4%	109	42.7%	35	33.0%	.2045
Interferon b-1a	393	42.0%	257	44.7%	98	38.4%	38	35.8%	.0956
Interferon b-1b	84	9.0%	51	8.9%	24	9.4%	9	8.5%	.9523
Mitoxantrone	1	0.1%	0	0.0%	1	0.4%	0	0.0%	.3867
Natalizumab	85	9.1%	41	7.1%	28	11.0%	16	15.1%	.0149
Rituximab	3	0.2%	2	0.2%	1	0.2%	0	0.0%	1.0000
No DMT use	661	41.3%	355	38.1%	186	42.1%	120	53.1%	.0002
Baclofen	260	16.3%	95	10.2%	97	22.1%	68	30.1%	<.0001
Relapse
Patients with a relapse in inpatient or outpatient setting (n, %)	255	15.9%	102	11.0%	87	19.7%	66	29.2%	<.0001
Number of relapses in all patients
Mean	0.4		0.2		0.5		0.9		<.0001
SD	1.5		1.2		1.5		2.2
Median	0.0		0.0		0.0		0.0
MS-related symptoms and secondary conditions (n, %)
Anxiety	178	11.2%	84	9.1%	62	14.1%	32	14.2%	.0065
Depression	317	19.9%	151	16.3%	108	24.6%	58	25.7%	<.0001
Eye symptoms	55	3.5%	14	1.5%	23	5.2%	18	8.0%	<.0001
Fatigue	415	26.1%	181	19.5%	140	31.9%	94	41.6%	<.0001
Fractures	6	0.4%	1	0.1%	5	1.1%	0	0.0%	.0157
Irritable bowel syndrome	30	1.9%	14	1.5%	10	2.3%	6	2.7%	.4059
Muscular weakness	139	8.7%	0	0.0%	91	20.7%	48	21.2%	<.0001
Osteoporosis	67	4.2%	23	2.5%	28	6.3%	16	7.1%	.0002
Pyramidal symptoms	229	14.4%	0	0.0%	164	37.4%	65	28.8%	<.0001
Sensory problems	323	20.3%	147	15.8%	118	26.9%	58	25.7%	<.0001
Sleep disturbance	167	10.5%	70	7.5%	61	13.9%	36	15.9%	<.0001

The mean (standard deviation (SD)) 1-year total unadjusted all-cause healthcare costs ranged from $45,574 (39,908) in patients with mild disability to $60,567 (61,880) in patients with severe disability at baseline, with a significant associated observed between cost and disability level (p<.0001). Healthcare costs were largely driven by outpatient pharmacy costs, most of which was due to the cost of DMTs. In patients with mild disability, 89% of outpatient pharmacy costs were due to DMTs (DMT costs: $30,555 (27,984)). In patients with moderate disability, DMTs accounted for 82% of outpatient pharmacy costs (DMT costs: $30,212 (28,320)). Finally, DMT costs accounted for 78% of outpatient pharmacy costs in patients with severe disability at baseline (DMT costs: $27,251 (28,968)).

Adjusted total annual healthcare costs (mean ± SE) were highest among severe patients: $56,206 ± 87, followed by moderate: $53,488 ± 83, and mild: $46,691 ± 72 (Figure 2). Comparison across disability level showed that adjusted total annual healthcare costs were 15% (p≤.0001) and 20% (p<.0001) higher in moderate and severe than mild patients, respectively.

Figure 2. Adjusted healthcare costs by baseline disability level in the full sample (n = 45,687). The patterned portion of the pharmacy costs indicates costs associated with DMT use.

Discussion

To our knowledge, this is the first published study to develop a measure of MS disability in claims data that proxies the EDSS, which is considered the gold standard disability assessment and endpoint in clinical trials. Key medical expertise in using EMR narrative and claims data to approximate disability level was central in this process including alignment on potential proxy indicators, a priori, followed by rigorous review of the resulting groupings by internal and external MEs. Overall, higher disability level per the CDS was associated with increased total all-cause 1-year post-index healthcare cost, for both unadjusted and adjusted cost analyses. The increase in total cost at higher disability levels was due to increases in both medical and outpatient pharmacy costs.

While pharmacy costs increased between patients with mild and moderate disability and leveled-off, medical costs increased steadily from one disability level to the next. Jones et al. which quantified the relationship between increased disability and health care resource utilization, quality of life, work productivity, and health care costs in patients with MS also found increased disability was associated with increased health care resource utilization and costs. The finding of this study using the proxy measure of MS disability was consistent with finding from the study by Jones et al. in which EDSS score was used to measure disability¹⁵. Similarly, Kobelt et al. reported that costs for severely disabled individuals were more than twice those for persons with relatively mild disease¹⁶. Compared to an individual with an EDSS score of 2, costs for an individual with an EDSS score of 8 are almost three times as high¹⁷. In addition, an observational, cross-sectional multi-country study which collected information on 16,808 patients’ characteristics, disease type, use of resources and loss of resource and used EDSS to stratify patients by disease level also showed the relationship between costs and disability, that costs increased on average fivefold between mild and severe MS¹⁸.

While total healthcare costs increased, the proportion of pharmacy costs attributed to DMTs decreased with increasing CDS, indicating a need for more treatment options in patients with severe disability. Studies have shown that current DMTs are not clinically effective in patients with EDSS ≥ 6.5¹⁹. A survey of 26 neurologists with expert knowledge of MS in Europe, North America and New Zealand found that nearly 60% of surveyed neurologists preferred to stop DMTs at an EDSS between 6.0 and 8.0. Some indicated that third-party reimbursement is a factor in discontinuing DMTs at certain levels of EDSS²⁰. Most recently, a consensus paper by the MS coalition concluded that access to DMTs should not be limited by level of disability²¹, which is a shift from prior recommendations, possibly due to the approval of newer therapies indicated for a more progressed patient population²².

In summary, this is the first published US study using claims in combination with EMR data to develop a score measuring MS disability level, an important indicator of MS progression, using all eight functional systems of the EDSS. In alignment with prior research, healthcare costs steadily increase with increasing CDS level. The CDS advances the opportunity to examine treatment patterns and outcomes by disability accumulation in the absence of standard markers of disease progression. However, this study has several limitations. First, while EDSS is routinely used in clinical trials, it is not typically collected in clinical practice; therefore, it was only available in a small portion of patients in the EMR database. Only four patients (0.3%) in the core sample had an EDSS available, thus formal validation of the CDS was not conducted. The level of disability for study patients was estimated using proxies for ambulation/functional status in claims or vitals/problem lists in the EMR. While the claims coding was expected to be specific, the use of relevant codes may not confer sensitivity in the grouping approach used in this study. Thus, these proxies and the final determination of accuracy of the disability designation were established in conjunction with clinical ME input and review. These proxies were intended to categorize patients solely for the research purposes in this study. They do not represent a clinical designation and may be challenged by external stakeholders. The CDS is a static measure of patient disability and may inaccurately designate a severity level to patients who are in the midst of a relapse during the CDS measurement period. Despite these limitations, we believe this methodology captured our intention to assess the viability of utilizing real-world data sources to identify patient disability level. Second, the ambulatory EMR includes only data captured from physicians contributing to the EMR network. Patients that have vitals measurements (e.g. visual disturbance, tactile sensation) taken by health care providers outside of the network are not captured. This could lead to a limitation of generalizability of our findings. Third, PharMetrics Plus data are sourced primarily from employer-sponsored commercial and self-insured plans, and as such underrepresent patients older than 65 years of age, specifically those with fee-for-service Medicare, thus limiting the generalizability of the study in older adults. This is an important limitation since MS is a chronic condition that is typically diagnosed in working-age adults, and ultimately leads to disability, which may then invoke Medicare coverage. Fourth, this analysis is retrospective in nature and relies upon existing coding in claims data and narrative text in EMR that were not collected for research purposes to approximate disability level. In addition, the IQVIA patient claims and EMR datasets apply a proprietary linkage methodology and as such cannot be made fully transparent. As a result, complete verification of the linkage cannot be ascertained and the possibility of error in linkage may exist. Lastly, indirect costs and direct non-medical costs associated with MS were not captured in the databases. Prior studies have found that while direct medical costs are important contributors in the earlier stages of MS, indirect costs outweigh direct medical costs in later stages^5,6. Future studies validating the CDS are warranted.

Transparency

Declaration of funding

This study was sponsored by Novartis Pharmaceuticals Corporation.

Declaration of financial/other relationships

RB, EF, and AO provide consulting services to Novartis. MG and RLW are employees of IQVIA, which received consulting fees to conduct research on this study. YD was an employee of IQVIA at the time the study was completed. LB was a postdoctoral research fellow at Novartis at the time the study was completed. WS and KJ are employees of Novartis. PR was an employee of Novartis at the time the study was completed. A peer reviewer on this manuscript has disclosed that they have previously been, and are currently, involved in research to assess the severity of MS patients using claims data (but not in EMR/EHR data). A peer reviewer on this manuscript has disclosed that they have been a speaker for several MS medication companies. The peer reviewers on this manuscript have no other relevant financial relationships or otherwise to disclose.

Author contributions

RB, EF, AO, and WS were involved in the study design and provided clinical guidance. MG, RLW, and YD designed the study, provided interpretation of the results, and drafted the manuscript. LB, PR, and KJ were involved in the study design, interpretation of results, and manuscript editing.

Previous presentations

Results from this study were presented at the 35th Congress of the European Committee for Treatment and Research in Multiple Sclerosis.

Acknowledgements

The authors thank Hsiu-Ching Chang for conducting the programming and statistical analysis.

Notes

i PharMetrics Plus is a registered trademark of IQVIA, Plymouth Meeting, PA, USA.