A German expert group consisting of neurologists, health care administrators, psychologists, and MS nurses developed the first version of the Multiple Sclerosis Health Resource Utilization Survey (MS-HRS) at the University Hospital Dresden in 2009. Following intensive feedback from physicians, nurses, and patients, a second more time-efficient version was created in 2016. The requirements from clinical practice, such as time-saving use in everyday clinical practice, updatable pricing, and avoidance of double documentation, were decisive for the development of the digital version, which is browser-based and can be used across most devices (Figure 1). The time-saving use was realized through an adaptive questionnaire structure by presenting only the necessary items electronically.

The current digital German and translated English versions of the MS-HRS are in line with the recommendations on the core elements of a standardized resource questionnaire and can be completed in 10 to 15 minutes [14].

Health economic analysis can be performed from different perspectives with the societal perspective being the broadest [15]. For the practical evaluation of costs, three steps are considered, namely identification of resource consumption, quantification of resource use, and valuation of resources [15].

Information on resource consumption was obtained from studies, guidelines, textbook knowledge, administrative and accounting data, and expert opinions [4,8,16,17].

Costs were classified into direct medical, direct nonmedical, and indirect costs, according to German health technology assessment guidelines. Keeping multiple use cases in mind, we divided the tool into a core and an additional set of items (Table 1). The core elements are constitutive for the health resource utilization model and sufficient to capture the scope of resource utilization, whereas the additional questions foster the understanding of resource utilization.

We used the most accurate method of microcosting defined as “direct enumeration and costing out of every input consumed in the treatment of a particular patient” to quantify resources (Table 2) [18]. Recall periods of resource utilization have been determined to avoid over- and underreporting. Longer recall periods lead to telescoping effects, in which events occurring outside the time frame are mistakenly included in the intended period [19]. Furthermore, we took into account that more frequent events and those that are less salient are less likely to be recalled accurately over a long period. In conclusion, the recall period may differ with respect to the nature of cost items but should never exceed 12 months. We recommend intervals for the assessment between 3 and 6 months. It is noticeable that unified recall periods across all items of the survey made the questionnaire more intelligible for patients. In our results, we report all calculated costs as per quarter.

Evaluations conducted from societal perspectives are intended to reflect societal opportunity costs, which are equal to market prices in perfectly competitive markets. Nevertheless, markets in the health care sector are imperfect owing to statutory regulations. Hence, societal opportunity costs have to be approximated in most cases. Therefore, data from Bock et al were used wherever available (Table 3) [20].

Owing to a lack of data, few valuations were calculated from existing values. Visiting a nurse was rated with the lowest monetary value of €16.42. Furthermore, psychologist contact was monetarily valued as visiting a psychotherapist. The valuation for other specialists (eg, psychiatrist, urologist, and optician) was calculated as the mean of the given examples. Societal opportunity costs for investigations were approximated from the physicians’ fee [21].

To calculate annual costs for DMTs, we used defined daily dose net costs multiplied by 365 days [22]. Net costs account for statutory manufacturer discounts as well as pharmacy discounts (Table 4). The work productivity loss (absenteeism, early retirement, and presenteeism) was calculated using the human capital approach [23]. Absenteeism was defined as not showing up for work, whereas presenteeism was defined as reduced work productivity due to health problems. The loss resulted from the total number of lost hours multiplied by the average salary per hour. Any hour not worked was considered as lost. Data for this calculation were obtained from official statistics [24].

In 2011, 233 working days were used for the calculation, and the average number of working hours was 1406.2 hours [25]. Additionally, the average hourly labor cost was €29.90 [20]. The monetary work productivity loss due to sick leave was calculated as follows:

Occupational disability was calculated as the product of daily wage, average number of working days, and percentage of disability, with a maximum of €42,045.38 per year. The maximum hours dedicated to informal care was set to 60 hours per week according to German policies.

To ensure comparability of cash flows, prices from different periods were adjusted to the 2011 price level using the general price index for the national economy. This applies to patients’ self-reported medications, investments, medical consumables, and DMTs. Consumption of the remaining resources was valued with prices from the year 2011 to generate comparable costs within our validation population. For other purposes, more recent values may be derived by applying a conventional cost inflation of 2%.

All costs are reported in Euro. In 2011, the Euro to US dollar annual average exchange rate was equal to 1.392.

Patients with relapsing-remitting MS (RRMS) were recruited in two prospective noninterventional multicenter studies conducted in Germany [26,27]. In that context, assessments of medical history and other general data, such as the Expanded Disability Status Score (EDSS), were performed by the treating neurologist. The current analyses are limited to patients with an EDSS of 0-6.0 to ensure sufficiently large subpopulations. Approval for both studies was obtained from independent local ethics committees, and all patients provided written informed consent for the collection of data [26,27].

Practicability was determined by the response rate of patients completing the MS-HRS. In addition, we compared the characteristics of completers and noncompleters to avoid selection bias. For reliability and validity analyses, we focused on the core elements of the survey. As we did not intend to evaluate DMTs at this step, costs for DMTs were not part of our methodological evaluation, which focused on psychometric properties.

For reliability analysis, test-retest reliability of the total costs in a group of stable patients over 3 months was estimated. Stable patients were defined as those not having relapse or increase in the EDSS during the 3 months of the retest period and another 3 months prior to the assessment. Thresholds for intraclass correlation coefficient (ICC) were applied as recommended (ICC<0.5: poor reliability; 0.5≤ICC<0.75: moderate reliability; 0.75≤ICC<0.9: good reliability; and ICC>0.90: excellent reliability) [28]. Guttman lambda 2 and lambda 6 are reported for the monetarized standardized items of the health resource utilization model with respect to the heterogeneous structure, providing a lower bound estimate of the consistency of the pricing approach. In an ordinary setting for test construction, desired levels would lay above 0.7. As our survey is not based on latent constructs but instead on real costs, consistency analysis is not part of the primary evaluation of reliability, but it provides additional insights for the model and the cost components.

Repeated expert consensus meetings of health economists, psychologists, and neurologists were conducted to secure face validity in terms of consistency and completeness.

Construct validity was assessed as convergent and discriminant validity via correlations with associated patient-reported outcomes (PROs) and known-group analyses by the EDSS (ranges: 0-1.0, 1.5-2.5, 3.0-4.0, and 4.5-5.5). The selected PRO measures were the EuroQol-5 Dimensions (EQ-5D), UK Neurological Disability Scale (UKNDS), and Patient Reported Outcome Indices for MS (PRIMUS) [29-32]. We expected to find significantly higher costs in groups with higher EDSS scores and correlations above 0.40 for convergent validity. As we did not intend to evaluate DMTs, the Treatment Satisfaction Questionnaire for Medication (TSQM) should present correlations clearly below 0.3 for discriminant validity [33].

Continuous values are reported as arithmetic mean and SD. Ordinal values are reported as median and IQR. One-way random ICCs were used to calculate estimates for the test-retest reliability. All other correlations were calculated with Spearman rank correlation coefficients. Kruskal-Wallis H tests were conducted for known group analyses. Mann-Whitney U tests with adjustments by Bonferroni correction for multiple tests were applied for pairwise comparisons. All reported P values were compared to an alpha error level of 5%. No imputations were made for missing values.

In total, 2207 of 2388 patients completed the questionnaire at baseline and were therefore included in the validation process. The study population had a mean age of 41.73 (SD 10.19) years and was mostly female (1609/2207, 72.90%) (Table 5). Employed patients (1347/2207, 61.03%) were working predominantly full time (794/1347, 58.95%).

In terms of relapses within the previous year, active (1015/2207, 45.99%) and nonactive (1192/2207, 54.01%) patients were balanced in the population. Participants reported a mean disease duration since diagnosis of 7.54 (SD 6.11) years and a mean EDSS of 2.43 (SD 1.57).

Majority of patients stated that they used direct medical services in the past 3 months (Table 7). In contrast, less than every second patient (44%) had indirect medical costs and 16% had direct nonmedical costs. Besides DMTs, indirect costs were the main cost drivers ahead of direct medical and direct nonmedical costs.

Patients were mainly treated in private practices (2068/2207, 85%) and less often during inpatient hospital stays (138/2207, 6%) and day care admissions (65/2207, 3%). However, the highest cost was for inpatient treatments (€315.06 [SD €1587.09]), followed by consultations in the primary sector (€209.87 [SD €292.73]) and day admissions in hospitals (€32.22 [SD €311.68]). Total costs per quarter averaged €2462 (SD €3650; median: €631) without DMTs and €7126 (SD €3697; median: €5871) with DMTs. Therefore, annual costs for mild-to-moderate RRMS ranged between €9528 (SD €14,603) without DMTs and €28,203 (SD €14,808) with DMTs.

The MS-HRS represents a reliable, valid, and easy-to-administer questionnaire providing a holistic assessment of resource utilization for patients with MS. Health resources were derived from all pathways of patients with MS in an adapted health resource model for MS.

Some very respectable efforts have already been taken to research the health economic footprint of MS [2,5-7]. In Europe, Kobelt et al repeatedly assessed the costs and burden of MS in a cross-sectional survey approach, including direct, indirect, and intangible costs [6,7]. Concepts and definitions of subcosts may differ over time with respect to scope and style of reporting (eg, intangible costs where a clear line is recommended) [34]. For our model approach, we gave strong emphasis to direct and indirect costs, as done by Karampampa et al and Reese et al in their models [5,11]. We increased the depth of the assessment of indirect costs by adding a quantification of costs for presenteeism in addition to costs for absenteeism (sick leave and disability pension).

Part of our questionnaire development was a validation process that confirmed reliability and validity in a way that no previous approach did within the MS domain (Figure 2). Kobelt et al also recognized the need for validation, but a systematic approach beyond the aspects of practicability and face validity remained a task to fulfil [35]. Even though we did not develop a one-dimensional questionnaire for a latent construct, we were able to demonstrate essential psychometric qualities in the health economic assessment of MS. As of today, a multidomain open-access database of resource-use questionnaires does not contain a published instrument for MS [12]. Most of the cost assessments for MS were not developed to be published for general use but developed for application in certain cross-sectional frameworks. A very recent analysis of real-life cost outcomes underlined the rising interest in longitudinal assessment of health resource utilization [10].

The proportion of patients claiming certain health services was slightly lower in the study by Kobelt et al than in this study [6]. The lower proportion of patients taking over-the-counter medications and investments was particularly noticeable. This may be explained by the less severe disease progression beyond a RRMS profile. Taking this into account, the average annualized disease burden was within the range expected from previous publications [5,7,11,36].

Kobelt et al have recently reported precise price tags for most unit costs, and other authors have at least provided an indirect description of the valuation process and its exact results [5,9]. Differences were found for the valuation of subcosts, such as inpatient hospitalization, owing to different sources of valuation or different definitions for health resource units. In any case, precise and fully transparent reporting of per unit costs is highly recommended, especially for main cost drivers.

Event-related costing can also be considered to be of interest, but a clear separation from costs being assessed via routine data collection is necessary [5]. Otherwise, costs may double due to double assessment. We discussed whether to collect data for both sick days and missed working hours, as described by Reilly et al [16]. Furthermore, asking about sick hours would have meant switching to a recall period of 7 days, whereas the remaining questionnaire covers the last 6 months. Owing to higher usability, we decided to assess sick days only.

Claims data were considered for the assessment of criterion validity. However, there is a disadvantage that only billing-relevant data are available, preventing consideration of a societal perspective. Information on informal care, over-the-counter medication, and presenteeism and partial information on investments and medical consumables is not recorded. In addition, claims data do not provide the number of physician visits by a patient in Germany. In consequence, claims data were not suitable for the validation process.

Health economic studies often require both clinical and economic data. The MS-HRS can easily be used as part of clinical interventional and noninterventional studies to collect economic data. In a large population, we demonstrated that the questionnaire is easy to administer and has good psychometric properties. These characteristics provide the necessary prerequisites for high-quality health economic studies (eg, cost-effectiveness analyses).

A patient-centered questionnaire is subject to notable limitations. Reliable recall periods of health resource use are time-limited, especially in patients who are cognitively more affected by the disease. Furthermore, patients may not want to disclose socioeconomic and health economic information because it is considered too confidential. We did not include information about more recent therapies as we did not gather related data for validation, which will be done in follow-up studies. Beyond this, further costs may be thinkable but less likely to have an impact on a societal level (eg, crowd-funded therapies at the current level and cost-related voluntary work loss). In addition, price tags for cost components have to be updated and adapted to local levels. For patients with progressive MS and patients with severe disability, further studies have to confirm the given psychometric properties.

The MS-HRS is a promising option to measure costs precisely in cross-sectional and longitudinal settings instead of estimating them or using surrogates. Further country-wise cost weights will facilitate the transparent estimation of MS-related costs across multiple regions. The MS-HRS is available online [37].