To estimate the national economic burden by patient age and type of RD, we took a prevalence-based approach. This approach combined the RD prevalence with per-person disease-attributable excess cost (i.e., a difference in the annual per-person costs between the RD sample and matched controls without RD). Direct medical costs were captured through an administrative claims-based analysis the indirect productivity losses for persons with RD and their caregivers, non-medical costs associated with RD, and expenditures for daily non-medical care were captured via survey that we designed and implemented– the National Economic Burden of Rare Disease Survey (hereafter, the Survey). This survey was one of the largest surveys conducted so far covering multiple communities representing 379 RDs. We mapped 379 RDs to 16 disease groups (see additional file 2) representing a body system and corresponding to the International Classification of Disease 10th Edition (ICD-10) diagnosis coding system, which was done in several steps. First, we identified the ICD-10 codes for each disease. Then, we mapped codes and added decision tree logic to the most granular (i.e., “leaf nodes”). Finally, we ran the data against the Orphanet codes to see whether there was agreement and whether we could identify any codes in Orphanet that were unidentifiable using the ICD. The entire process was reviewed several rounds by a designated technical advisor group consisting clinical experts, RD researchers, and family advocates. Figure 2 shows the cost calculation steps and data sources. The study received approval by an Institutional Review Board.
Direct Medical Costs
To quantify the annual excess medical costs associated with RD, we compared the average total cost of people with RD with that of a matched comparison group with similar characteristics including age, gender, race/ethnicity, and insurance, but without any RD, using a 10:1 control to case ratio.
Medical costs based on claims data included primary payer paid amount, patient out-of-pocket expenses (e.g., copay, co-insurance, deductibles), and any third party paid amount. We estimated the direct medical costs of RD by types of healthcare service: acute and non-acute inpatient stay; outpatient care; physician office visit; durable medical equipment (DME); other ancillary, outpatient-based drug administration; retail prescription drug use; and caregiver payments (by Medicaid). Since the Medicare 5% data do not include Part-D claims, we used commercial per-person prescription cost to impute the Medicare per-person prescription cost for each age and disease group. All costs estimates were expressed in 2019 dollars. Average direct medical cost was calculated for each of the 16 RD groups for adults and the 7 RD groups for children. We used the group average across all RDs to compare with the comparison group to derive the average RD-attributable cost, to avoid double counting.
Indirect Costs and Non-Medical Costs
We worked with broad coalition of patient advocacy organization partners to design an online survey to estimate cost due to: reduced labor market participation, productivity loss for those in the labor force, non-medical costs of RD (such as the cost of hiring professional non-medical caregivers to assist with daily living, necessary home modification costs), and disability benefits. The respondents were persons with RD. However, the family member most familiar with the health of the person with the RD could respond to the survey, if the health of the person with RD prevented accurate self-reporting, or if the person was a minor.
We took convenience sample approach and disseminated the survey to the RD communities via partner networks of more than 200 partner-patient advocacy organizations, reaching a broad range of RDs and large patient sample.
We received 3,484 responses with 1,399 being fully completed. The survey asked about the respondents’ disease and 581 RDs were reported. After removing misspellings, RD alternative names, and diseases that were not rare (e.g., cancer) or represented a protein, there were 379 unique RDs.
Additional file 3 provides the breakdown of the respondents’ self-description and shows that 57% of the respondents were people with a RD, and 41% of responses were from a family caregiver. About 28% of the responses represent children (<18 years); the rest were adults, with those above age 65 representing 14% of all responses (additional file 4). People with RDs were predominantly white (87%), followed by multi-racial individuals (4%). About 77% of people with RD had at least one caregiver (a primary caregiver) and about half had both a primary and a secondary caregiver; 23% of the RD sample did not rely on a caregiver.
Given the survey sample size, the number of RD groups, age stratification, and the number of indirect and non-medical cost components, it was not feasible to calculate reliable cost estimates for all RD groups-age strata. Where the strata sample size permitted, we calculated average costs for that strata; if strata sample size was <5 observations, we reported the average costs across all RD groups. The mapping of cost components to RD group-age strata is reported in additional file 5; cost component calculations are detailed below.
RD may increase the likelihood that severe functional impairment or disability will prevent them from working, or limit employment opportunities and reduces earnings. 3,4 Our survey indicated that among the working age (18-64) persons with RD, 43.8% are in the labor market, as compared to the national labor force participation rate of 63.1% among the U.S. adult population. To ensure that the early termination of employment was related to RD, we calculated labor market employment related earnings loss due to RD as the count of persons with RD, who have retired or stopped working in the past 12 months and indicated that RD played a major role in their decision, multiplied with the median annual earnings by job status (full-time versus part-time) obtained from the 2019 American Community Survey public use microdata sample. We used medians rather than averages, as medians are less likely affected by outliers. As the full-time/part-time status of persons with RD before retirement was unknown, we used the allocation of full-time to part-time job status among currently working persons with RD. Then, we calculated earnings loss due to early retirement for those who retired due to RD as a weighted average between those assumed working full-time before retirement and those working part-time before retirement.
We calculated two measures of reduced labor market productivity for those who are employed: absenteeism, (increased workdays missed due to illness), and presenteeism, (illness-related poorer work performance while on the job). We asked about the number of days in an average working month during 2019 the person with RD and the caregivers missed work or felt less productive while at work because of RD. Based on responses to these two questions and the average daily earnings calculated from the self-reported annual earnings, we calculated the productivity loss due to absenteeism by multiplying the number of days missed with the daily earnings and then annualized the total loss. Presenteeism was calculated similarly, with an adjustment factor applied to each day felt unproductive, reflecting that an unproductive day is not equivalent of a total loss of a whole day’s value. The adjustment factor was obtained from the responses to productivity self-assessment scale: i.e., on days when feel less productive, on average how productivity of the person with RD and caregivers was affected on a scale from 0 to 10, where 0 represents “not at all”, 1-3 “mildly”, 4-6 “moderately”, 7-9 “markedly”, and 10 represents “extremely”. We translated these responses into the reduced overall productivity (e.g., 0 corresponds to 100% productivity, 10 corresponds to 0% productivity or the full reduction in productivity). Daily earnings were calculated from the annual earning brackets that was applicable to the respondent in 2019 (categorical responses were converted into numerical values based on the mid-point of each earnings category: everyone who indicated earnings “less than $1,000” were assigned earnings of $500, etc.).
Additionally, RD may affect patients and the caregivers’ ability to participate in various social activities using their leisure time.The challenge of quantifying social productivity is the measuring the time forgone from social activities and in the proper valuation of the time forgone. Although one could argue that forgone leisure time visiting family and friends also creates economic loss, we focused on activities that directly involved volunteering and provide a conservative estimate of the social productivity loss. We asked about the number of hours the person with RD and the caregivers spent in a typical week before and after RD started having a significant impact, on the following social activities: performing voluntary or charity work; providing help to family/friends/neighbors unrelated to personal care or care for person with RD; participating in a political or community-based organization.
We compared self-reported volunteering hours before RD with the average national annual volunteering hours obtained from the Current Population Survey (CPS) Volunteer Supplement that measures the population’s participation in volunteer activities (2017). The national average volunteering hours are generally lower than the self-reported volunteering hours (e.g., 1.9 hours per week versus 12.1 hours per week for person with RD before RD). Therefore, we took a conservative approach in our calculations by calculating the percentage of people volunteered and average hours volunteered from CPS and multiplied with the estimated percentage productivity loss from the Survey (calculated as the difference between before and after hours divided by before hours) for the three activities combined. Productivity loss due to forgone volunteering activities was calculated as volunteering hours affected per year times $27.20, which is a dollar value per volunteering hour according to the Independent Sector.
The non-medical costs calculated included expenses of purchasing formal care (e.g., adult day care and personal aides) and necessary modification to homes, purchases of adapted motor vehicles or car modifications for accessibility, medical foods, dietary supplements, specialty clothing (e.g., compression stockings), and increased travel costs for medical visits. We also asked about healthcare services not covered by insurance such as experimental treatments, alternative or non-traditional treatments (alternative therapies, massage therapy, acupuncture), and over-the-counter drugs. We estimated cost of non-medical components and medical out-of-pocket costs by multiplying the weighted percentage of families who responded as having such expenses and the average expense per-family per-year.
To capture the overall economic burden of RDs, it is always an important policy perspective to be able to identify the extent to which individuals are transitioning into public programs, and what the potential costs to public programs are due to any specific condition/disease, particularly if these costs are avoidable. For example, the Social Security Disability Insurance (SSDI) and the Supplemental Security Income (SSI) are considered as transfer payments (i.e., a cost to one person is a benefit to another person). Therefore, these components may inform on the extent of government budgetary burden due to a specific disease. We asked respondents whether the person with RD had received SSI, SSDI, or other types of disability income, in 2019. While we estimated the average and total disability income due to RD, these costs were excluded from the overall burden estimates, as these funds could have been used for healthcare payments or non-medical expenses already captured in other cost components.