Study design

We conducted a prospective and multicenter cohort study in France between 2014 and 2015.

The main inclusion criteria were having a lymphedema, being aged 7 years or more, living in metropolitan France, and being able to use Internet and email. Lymphedema patients were included in a convenience sample of 27 centres, among which 11 were public or private hospitals and 16 were individual physicians or physiotherapists working in a private practice. Each centre included all the consecutive patients who met the inclusion criteria during the study period. Because we anticipated a high drop-out rate due to the data collection burden, we ran a 1-month pre-inclusion phase. Only patients who agreed to participate at the end of the pre-inclusion phase were included in the study. All patients (or their legal representative) signed a written informed consent prior to inclusion. The study was approved by the French institutional review board (CPP Sud Mediterranee V; approval number 2014-A00716-41).

Data collection

Data collection was standardized using a paper questionnaire filled in by health care professionals and an electronic questionnaire filled in by patients on a dedicated and secured Internet website. The professionals provided demographic and clinical data, and patients reported socio-economic and clinical data, health care services utilization, cost and reimbursement, and health-related quality of life measures. Health care service utilization was collected weekly. To ensure high-quality data, we accompanied patients and professionals with initial training, real-time data quality monitoring and active email and telephone communication.

We collected the following socio-economic data: employment status and category [19], monthly net post-tax household income (using an 11-category scale), VHI and special schemes entitlements (CMUc, ACS and ALD). Questions were asked about the travel distance to seek lymphedema care and subjective unmet medical needs due to travel distance or cost. A patient had a subjective unmet medical need when she declared having forgone in the previous year at least one of the lymphedema care recommended by the current French clinical guidelines [20]: medical visits, compression medical devices or physiotherapy. The monthly income by consumption unit (CU) was calculated according to the modified equivalence scale of the Organization for Economic Co-operation and Development (OECD). Consumption units are a standardized measure of the size of a household determined as follows: the first adult has a weight of 1.0, each subsequent adult has a weight of 0.7, and each person under 18 has a weight of 0.5. During the six-month follow-up, service utilization, cost and reimbursement (by MHI and by VHI, if any) were prospectively collected for the following outpatient health care services: physician visits, non-physician services, imaging and laboratory tests, drugs, prescription and over-the-counter medical devices (bandaging, multi-layer inelastic bandaging, pneumatic compression, compression garments, sticking-plaster, wound dressing), and transportation. Inpatient care was collected but not analyzed, because (i) lymphedema patients are primarily managed in the outpatient setting, and (ii) inpatient care may have generated very high amounts of OOPP for very few patients. Non-medical cost was considered as follows: housekeeping or childcare, cosmetics, clothes, shoes, and thermal therapies. Housekeeping and childcare were considered to account for the impact on family life. We included cosmetics, clothes and shoes because lymphedema often results in significant changes in skin colour and in the circumferences of limbs [11]. Finally, lymphedema patients often rely to thermal therapies in France [20]. Two types of OOPP were estimated: after MHI reimbursement and after MHI and VHI reimbursement.

Two quality of life-related measures were assessed: utility values derived from the EQ5D-3L with the French index values[21]. In the absence of French index values, capability scores were obtained from the ICECAP-A [22] using the UK index values[23].

Missing or inconsistent data

Missing or inconsistent cost data for services with a statutory tariff were corrected using the adequate tariff, assuming no extra-billing. Regarding services without a statutory tariff (i.e., medical devices, transport and non-medical services), costs and reimbursements were checked directly with the patients by telephone.

Sample size and statistical analyses

The sample size was calculated to estimate the annual OOPP with a given precision. We hypothesized a mean OOPP equal to 200€ and a standard deviation equal to 1150€ [10]. To have a 95% confidence interval with a total width equal to 300, 230 patients were required.

Descriptive analyses were performed using frequencies and proportions for qualitative variables and mean, median, standard deviation and range for quantitative variables. Comparative univariate analyses were performed using the Student’s or Wilcoxon test for quantitative variables and the Chi² or Fisher test for qualitative variables.

To analyze the income-group differences in OOPP, we split the sample into 5 income per CU groups, each group representing one-fifth of the whole sample. Hence, the first income quintile refers to the poorest 20% and the fifth to the wealthiest 20%.

The vertical equity of OOPP was examined by describing the relationship between the ability to pay (measured by income per CU) and the amount of OOPP. We followed a standard strategy combining disaggregated and aggregated measures [24, 25]. Disaggregated evidence is presented graphically by comparing the concentration curve of income (Lorenz curve) and the concentration curve of OOPP. The Lorenz curve plots the cumulative share of the sample (from the poorest to the wealthiest) against the cumulative share of the total income they receive. The OOPP concentration curve plots the cumulative share of the sample (similarly to the Lorenz curve) against the cumulative share of OOPP. If the OOPP is strictly proportional (i.e., patients pay the same proportion of their income), then the Lorenz and OOPP concentration curves will overlap. When the poorest patients pay a higher proportion of their income (regressive payment system), then the OOPP concentration curve lies above the Lorenz curve. Aggregated measures are the Gini coefficient for income, the Kakwani index for OOPP and the Reynolds-Smolensky index. The Kakwani index K is calculated by subtracting the Gini coefficient G for income from the concentration index for OOPP, and it quantifies the extent to which OOPP depart from proportionality. The Kakwani index is negative when there are vertical inequities favouring the rich (regressive payment system) and positive when there are vertical inequities favouring the poor (progressive payment system). The Reynolds-Smolensky index measures the overall redistributive effect of OOPP on income and includes some horizontal inequities [26].

All analyses were performed using SAS software, version 9.2.

Strengths and limitations

This study suffers from some limitations that should be acknowledged. First, we had to exclude 28 patients for whom data on out-of-pocket payment was missing. However, we did so to improve the data quality, and there is no argument for a systematic bias. Second, we improved the representativeness of the sample by including a wide range of public and private centres throughout the country, and by including all consecutive patients who met the inclusion criteria in each centre. However we were not able to check the representativeness because of lacking data on the characteristics of the lymphedema centres and patients in France. In addition, because we did not include patients who were unable to use the Internet and email, the oldest and most disabled patients are likely to be underrepresented in our study sample. Again, we do not believe this might challenge the main findings. Finally, our analysis is focused on ambulatory care costs and it might be argued that inpatient care could generate very high out-of-pocket costs. However, we did so because inpatient admissions would have generated high costs but very low out-of-pocket costs. Indeed, out-of-pocket costs are concentrated on ambulatory care services in France for several reasons [8, 9]: inpatient care is predominantly covered by the mandatory social insurance (91% on average, vs. 63% for ambulatory services), hospital medical devices are not billed to patients, transportation from and to hospitals are covered by the mandatory social insurance, and there is no extra-billing.

Nevertheless, this is one of the only OOPP studies in France accounting for a wide array of data. Indeed, we were able to prospectively collect socio-economic and clinical data as well as unmet medical needs and reimbursements by VHI that are not available in standard retrospective analyses. The national, multicenter and primarily outpatient-based pre-screening and screening certainly made the sample representative of the whole lymphedema population. As an example, the Gini coefficient for income in our sample (0.286) is very close to that of the French general population measured in 2013 (0.294; [34]. Finally, the data quality check process allowed us to control the impact of missing and inaccurate data.

Policy implications

We show that the vertical equity principle is violated regarding the financing of outpatient lymphedema health care in France. Although cost-sharing has been implemented in most European countries, its ability to reduce the utilization of unnecessary health care remains unclear [2, 35]. According to the most widely accepted prevalence data[11], the total number of patients living with a lymphedema in France is equal to 268,000. Based on this figure and on our results, taking over the transportation OOPP (32.9 Euros per patient per month) would cost 105.8 Million Euros per year to the system. Similarly, eliminating the medical devices OOPP (26.2 Euros per patient per month) would cost 84.3 Million Euros per year. However, improving the vertical inequity does not imply eliminating OOPP. In France, various income- or disease-based mechanisms should control OOPP. However, they are not effective in protecting the poorest households from high levels of OOPP. Other budget-neutral or not options should be considered, such as adding a ceiling on OOPP [36] to the existing financing mechanisms, with or without removing the disease-based ALD exemption scheme [37]. Another option could be to implement a single co-payment for outpatient care with an annual ceiling, which may or may not be modified according to income alone [38] or to income and household composition [39].

As regards prescription medical devices specifically, OOPP are frequently explained by the difference between the statutory service and the selling price. Policy makers should therefore reconsider the possibility of charging a free selling price for medical devices. Likewise, the large amount of OOPP due to transport stresses the issue of geographical access to reference centres across the country. Possible levers for action include an improved distribution of reference centres across the territory and telehealth.

More generally, our results contribute to the debate about the consequences of cost-sharing in health care. Cost-sharing mechanisms have constantly been associated with reduced health care utilization and expenditures since the seminal RAND experiment [40]. But the consequences on health status are controversial: greater cost sharing is associated with reduced use of unnecessary care but also with reduction in utilization of high-value care such as long-term medications and preventive services [41, 42]. Our study suggests that patients facing out-of-pocket costs regarding ambulatory care for chronic oedema declare unmedical medical needs more often than the general population (31% vs. 26%, though figures are not directly comparable). Furthermore, the poorest patients declare more unmet medical needs, have lower total health care expenditures, have lower level of quality of life, and dedicated a higher share of their income to out-of-pocket costs. However, investigating whether these patients forgo necessary or unnecessary services was beyond the scope of our observational study and should be investigated using a natural experiment design.

Conclusion

This work shows that regarding outpatient health care, French lymphedema patients face significant and regressive OOPP, associated with an increased risk of unmet medical needs. Such results shed light on significant socioeconomic inequalities and bring into question the current financing arrangements of outpatient health care in France.