Data were extracted from the Merative™ MarketScan® Commercial Claims and Encounters database (10). MarketScan contains reimbursed claims data on over 273 million unique patients since 1995. The database includes employees, dependents, and retirees receiving coverage annually under private insurance plans. No Medicaid or Medicare data are included. The database contains information regarding inpatient and outpatient claims, and prescription drug claims. It is one of the largest and longest running proprietary claims databases used for healthcare research in the U.S.

We searched all inpatient and outpatient claims in the MarketScan database for the calendar year 2021 for claims that contained an International Classification of Diseases, Tenth Revision, Clinical Modification (ICD-10) diagnosis code for EDS (Q79.6, Q79.60, Q79.61, Q79.62, Q79.63, Q79.69). These are the only codes for EDS. Because there are no current ICD-10 codes for HSD, we used the code for hypermobility syndrome (M35.7) as a surrogate. Other inclusion/exclusion criteria were age ≥ 5 years old and continuously enrolled in an insurance plan for the calendar year 2021 with prescription drug coverage included in the individual’s insurance plan. We constructed two separate cohorts: (1) patients with a diagnosis code for EDS (with or without coding for hypermobility); (2) patients with a diagnosis code for hypermobility only (no EDS diagnosis code).Additionally, each cohort was separated into adult (age ≥ 18) and child (age ≥ 5 to <18) cohorts.

All inpatient and outpatient claims that occurred during the calendar year 2021 were searched for claims with specific ICD-10 diagnosis codes that indicated chronic comorbidities. From these claims, we calculated a recent edition of the Charlson Comorbidity Index (CCI) score, which is particularly useful for claims-based research purposes (11).

We estimated the excess costs for patients in our EDS and HSD cohorts by matching each patient to one patient in the database that did not have a claim for EDS or HSD during 2021 and were continuously enrolled with prescription drug claim capture for the calendar year. A total of 14,263,020 patients met these criteria and formed the set of possible matches for our EDS and HSD patients. We matched patients on age (+/− 2 years), sex, Census division, and CCI score (+/− 1 point) using a greedy matching algorithm (12). 88.8% of the matches in the EDS cohort and 91.7% in the HSD cohort were exact. We compared mean total healthcare costs in 2021 between groups.

Excess costs for patients in the EDS and HSD cohorts were determined by matching each patient to one patient in the database that did not have a claim for EDS or HSD and comparing total costs for the calendar year. We reported descriptive statistics (mean, standard deviation [SD], median, interquartile range [IQR], and range) for total healthcare costs. Cost estimates were rounded to the nearest $100.

For patients with EDS, we examined costs for patients who had a diagnosis code for vascular type (ICD-10: Q79.63) compared to those without a vascular diagnosis coded. Separately, we examined costs for patients with EDS who had a diagnosis code for hypermobile type (ICD-10: Q79.62) and compared to HSD patients.

To estimate excess costs, we calculated the difference in mean total healthcare costs between the cohorts and their matched comparison group. We used bootstrapping to estimate 95% CIs. Specifically, we used 1,000 resamples and the percentile method to obtain CIs.

To model costs with respect to EDS comorbidities, we used generalized linear regression models. For these models, we assumed a Gamma distribution and used a log link. The Gamma distribution was selected because costs exhibited substantial skewness, with some patients having very large costs. The models contained indicators for each comorbidity, age, sex, and CCI score with age and CCI score modeled linearly. Results from the models were reported in two ways. First, we reported cost ratios (CRs) and corresponding 95% confidence intervals (CIs). The cost ratios represent the multiplicative costs for each comorbidity. Second, we reported marginal differences in costs on the dollars scale. To estimate marginal costs, we set all other variables in the model to the most common value or the median as a reference. Specifically, marginal estimates were calculated for each comorbidity after assuming no other comorbidities, age of 34 (adult cohorts) or 14 (child cohorts), female sex, and CCI score of 0 as the reference, which we refer to below as the “reference patient.”

Sample sizes for each cohort were 5,113 and 1,059 for adult and child EDS cohorts, respectively, and 4,880 and 2,427 for adult and child HSD cohorts, respectively.

A flowchart depicting how these cohorts were constructed is shown in Figure 1. In the calendar year 2021, the database contained a total of 22,828,135 patients. Diagnosis codes for EDS and HSD were found for 8,068 and 9,463 patients, respectively (<0.1% for each). When restricting the cohorts to patients age ≥ 5 years old, 0.5% of the EDS cohort and 2.9% of the HSD cohort were excluded. Then, an additional 23.1 and 20.5% for EDS and HSD cohorts, respectively, were excluded because they did not have continuous enrollment with prescription drug claim capture for the entire calendar year. In total, 76.5% of EDS patients and 77.2% of HSD patients were retained for the analysis. In contrast, for all other patients in the database, only 62.5% were retained after applying the same inclusion/exclusion criteria. The larger percentage excluded was primarily due to the continuous enrollment exclusion, which indicated that patients with EDS and HSD were more likely to have commercial insurance for the entire calendar year.

Table 1 shows characteristics and comorbidities for the EDS and HSD cohorts.

In the EDS cohort, 87.3% of adults and 72.8% of children had one or more EDS/HSD comorbid conditions, compared to 75.6% of adults and 60.7% of children in the hypermobility cohort. In the EDS cohort, 36.0% of adults and 21.4% of children had 3 or more EDS/HSD comorbid conditions, compared to 21.4% of adults and 11.9% of children in the hypermobility cohort.

Patient characteristics for those included in the sample were generally similar to those who were excluded. For all EDS patients (adults and child cohorts) who were excluded, the median age was 29 (compared to 30 for those included), 85.3% were female (compared to 84.4%) and 62.5% had a CCI score of 0 (compared to 59.6%). Similarly, for all HSD patients (adult and child cohorts) who were excluded, the median age was 27 (compared to 25), 82.4% were female (compared to 81.1%), and 71.3% had CCI score of 0 (compared to 67.0%).

Table 2 shows diagnoses found in claims during the year for the EDS cohort. These diagnoses are not mutually exclusive because patients could have claims for any of the diagnoses during the calendar year.

Table 3 shows the total healthcare costs overall, and by each claim source (outpatient, inpatient, pharmaceutical, and Emergency Department). By definition, all of the cohorts had outpatient claims. 13.3% of adults with EDS and 9.5% of children had inpatient claims, nearly two-fold compared to the HSD cohorts. The vast majority had pharmaceutical claims which is not surprising given the multi-system manifestations of the conditions. Approximately one-third of those with EDS and one quarter with HSD had ED claims.

Cardiovascular complications have been noted in all EDS subtypes due tissue fragility, however, patients with vascular type EDS are at greater risk of life-threatening events including cerebrovascular events, aneurysm, and arterial dissection/rupture. For this reason, we compared vascular type to non-vascular types. For the adult EDS cohort, Table 4 shows total healthcare costs, by each claim source, stratified by whether patients had vascular type of EDS. For the child EDS cohort, only 12 patients had a diagnosis of vascular type, and therefore, we did not conduct this analysis. Mean costs were higher for patients with vascular EDS across all categories except outpatient pharmaceutical claims. This finding is consistent with the potential catastrophic manifestations of this type.

We compared total healthcare costs between patients with hypermobile type EDS (hEDS) and HSD patients. Table 5 shows total healthcare costs for hEDS compared to HSD. Mean costs were higher for patients with hEDS. Costs were also higher across all categories except inpatient claims which were approximately the same (data not shown). This finding is consistent with the clinical observation that individuals with hEDS are likely to be more severe in terms of their clinical presentation than those with HSD due to the coexisting conditions associated with EDS.

Table 6 shows parameter estimates from fitted regression models that included comorbidities and age and sex for total healthcare costs. Multiplicative costs (cost ratios) were reported for each comorbidity along with marginal differences calculated on the dollars scale (for the reference patient).

As reported in Table 5, for the EDS and HSD cohorts, both adults and children, with any of the comorbidities had greater healthcare costs. The largest difference was found in the EDS cohorts with gastrointestinal comorbid conditions, with more than double the costs (Adults: CR = 2.04; Children CR = 2.37), which reflected a marginal cost increase of $12,400 for the reference adult patient and $10,000 for the reference child patient. Patients with HSD, both adults and children, showed similar patterns, but to a lesser degree. Age and sex had less impact on marginal costs, however, differences in the comorbidity index were significant in all cohorts, adding to costs.

The estimated excess costs for patients with EDS and HSD are shown in Table 7. In 2021, adults with EDS and HSD had total healthcare commercial insurance costs that were $21,100 and $11,600 more than the matched comparison groups without EDS and HSD. Children with EDS and HSD had total healthcare commercial insurance costs that were $17,000 and $11,000 more than the matched comparison groups without EDS and HSD.

This study has several limitations. Selection bias is a limitation of our study. MarketScan is a very large convenience sample, but we cannot say that it is nationally representative because there may be differences in patient characteristics and healthcare plans that differ from the universe of all privately insured individuals. This to our knowledge has not been assessed. All of the study cohort patients had EDS/HSD, defined as at least one outpatient claim in 2021. We do not know when patients were diagnosed. Socioeconomic variables that impact having commercial insurance coverage are not included.

Notably, most EDS patients (79.5%) in this dataset were assigned the ICD-10 code “Unspecified” (Q79.60). This limited our ability to analyze subtypes. The large percentage coded unspecified could be explained by patients with overlapping types, uncertainty about types, or one of the rare types that do not have an ICD-10 subcode. Likewise, the study selection criteria for the EDS/HSD comorbid conditions were dependent on the selection of ICD-10 codes. We do not know the year that the patient was originally diagnosed; thus, we could not determine whether costs differed for patients who were newly diagnosed compared to those receiving an established treatment course. Another limitation is that there are no published diagnostic criteria for hypermobility syndrome (M35.7) and no diagnostic codes for HSD. Finally, our findings may not be applicable to patients with noncommercial insurance.