Discussion

This paper proposes an adapted methodology that can be used to measure access to paediatric medicines, based on the principles embedded in SDG indicator 3.b.3. This novel methodology could be an important tool for policy-makers and programme managers in identifying major barriers to access and developing appropriate policies to improve access to medicines for children. In adapting the methodology, two proposed core sets of paediatric medicines were established for children of different ages, taking into account their specific health needs and age-appropriate formulations. Careful approaches were taken to create the NUNT—a novel parameter—which enables affordability calculations across ages. The adapted methodology was successfully applied to data from three individual countries, providing proof of concept of this methodology.

With no reliable method for measuring access to paediatric medicines having been established yet, the child-specific methodology presented in this paper can provide guidance to others aiming to study access to medicines for children. The use of a single methodology and core set of medicines to express access to medicines will allow for intercountry comparability of the SDG indicator. Another important advantage of such a standardised tool is its ease of use. By predetermining which medicines and formulations should be surveyed, by providing the typical NUNT, and demonstrating how accessibility should be calculated, this method only requires countries to collect the facility data and some additional inputs. Yet, standardisation can also be viewed as rigidity, which is inherent to any tool that uses a single core set for global reference. Local guidelines that recommend use of other active ingredients or formulations than those in the core set could lead to skewed outcomes. Therefore, this standardised method incorporates some flexibilities, allowing for several formulations or active ingredients from the same therapeutic class to be interchanged (ie, antiepileptics, antimalarials, etc). This allows countries to apply this method to their national situation. Additionally, we recognise that the proposed core set should be subject to regular updates, in accordance with updates to the WHO EMLc and international treatment guidelines.

On closer examination of the case studies of Burundi, China and Haiti, the widespread inaccessibility seen in the results seemed to stem from unavailable rather than unaffordable medicines, for both LPMs and OBs. A recent systematic review on children’s medicines identified 14 studies that reported on the availability of children’s medicines and found a median availability of 38.1% and 24.2% for LPMs and OBs in the public sectors and of 35.9% and 21.1% in the private sectors, respectively.56 With that, the unavailability of child medicines detected in the present case studies is in line with the results of the systematic review. The same systematic review identified 11 studies that reported on the affordability of medicines, based on the number of days’ wages of the LPGW. In the public sector, affordability was 83.6% and 48.5% for LPMs and OBs, with 72.2% and 68.8% in the private sector. The results of this systematic review emphasise the need for a method that combines the two dimensions into a single indicator, as separate evaluation of these elements overestimates actual access to medicines for the patient. Beyond that, some of the studies included in the systematic review included unrepresentative samples of medicines (eg, studies focused on a single disease area or studies simply failing to consider child-appropriate formulations such as oral liquids or appropriate medicine strengths), again confirming the need for a standardised methodology to measure access to child medicines.

Before this methodology can, however, be applied on a widespread scale, several steps must be undertaken to further validate the methodology and examine the uncertainties introduced through our adaptations of the tool. First, the proposed core sets of medicines for school-aged children (not shown) should be validated through expert consultation. Additionally, the robustness of the adapted methodology with regard to the NUNT will need to be tested as it is an important variable when calculating affordability. The NUNT was determined based on recommended dosages and duration of treatment prescribed in international guidelines, which were often expressed as ranges. This generates some uncertainty when converting to a single NUNT. Also, determining an NUNT in many cases involved transformation of weight-based to age-based dosing through weight-to-age charts, introducing further uncertainties. The WHO provides international weight-for-age charts for boys and girls until the age of 545 and ages 5–10 years,46 but no international charts are available for children above the age of 10. Therefore, Dutch growth diagrams were used to approximate median weights of children 10–12 years.47 Initial comparison of international and Dutch growth charts shows that differences, if any, are small and will likely have had no significant impact on the NUNT. Furthermore, the NUNT is a single number used to represent an entire age group. How big the uncertainties with regard to the NUNT are and whether a single NUNT is indeed sufficiently representative for an entire age group should become clear in sensitivity analyses. Additionally, the case studies now performed were on a subset of the complete core set for young children, limited by the small number of age-appropriate medicines that had been surveyed in the three case study countries. Sensitivity analyses should also be performed to determine the minimum number of medicines required for a reliable measure of accessibility. To perform meaningful sensitivity analyses, more data on child medicines is needed than was available for the present case studies.

An important strength of this child-specific methodology is the use of an existing, formally approved tool as starting point which was adapted to suit the needs of children. Core concepts used in the adapted methodology and its data requirements are therefore in line with conventional methods and data collection tools. However, through this approach our methodology also inherits some of the limitations of the original 3.b.3 indicator methodology. Particularly, weighting for regional burden of disease when calculating access at the facility level as done in the original methodology raises several concerns. For one, the methodology assigns equal weights to medicines that are used to treat the same disease and thus counts the burden of this disease multiple times. To illustrate, the basket of medicines includes both oral rehydration salts and zinc sulphate for diarrheal diseases, whereas only retinol was selected for measles/vitamin A deficiency. This leads to disproportionate weighting for actual burden of disease when calculating access at the facility level. Disproportionality is also a concern for antibacterial medicines, which use may be overrepresented by using GHE code 20, a code that is linked to all infectious and parasitic diseases. Although a proxy for this GHE code was used in the present study (GHE code 370 for ‘other infectious diseases’), additional analyses should demonstrate how different weighting approaches affect the results. Additionally, the quality of the underlying GHEs data is unclear, especially because these data may be more difficult to obtain for children than for adults. Lastly, arguments can be made that the current approach of weighting for burden of disease is undesirable because it implies that some medicines are more important than others, even though all medicines in the basket are essential medicines and should always be accessible.

On a similar note, expressing affordability as a function of a poverty line instead of the LPGW wage has been used previously,57 but a measure combining the NPL and LPGW wage as is used in the original 3.b.3 indicator has yet to prove itself. This is particularly relevant because it seems that somewhat less medicines were unaffordable in the present case studies than what was observed using the LPGW wage alone.56 Further testing of the proposed child-specific methodology should include several scenarios for weighting for burden of disease and calculating affordability, which could lead to further adaptations of the methodology.

Since no facilities met the benchmark of 80% in our case study countries, the overall SDG indicator 3.b.3 was by definition 0% in all countries. Through this benchmarking approach relevant differences in access between countries and sectors were lost (eg, access in Burundi was better with a mean facility score of 40.3% vs 22.3% and 22.2% in China and Haiti, respectively). Additionally, the detail required for identifying the major obstacles in accessibility is also missing when the SDG indicator is reported as a single outcome. This highlights that disaggregated data on a facility and medicine level is vital in understanding the drivers of inaccessibility to medicines, particularly when the indicator value reflects a suboptimal level of access. We recommend that the indicator should therefore be reported in both a composite and disaggregated form.

To provide first evidence of the child-specific tool that we developed, we were limited to the use of historical datasets. In selecting suitable datasets for the case studies it was observed that only a small number of age-appropriate medicines are being surveyed in low-income and middle-income countries.58 The WHO/HAI datasets used for the present case studies were selected for their quality of data and relatively high inclusivity of age-appropriate medicines, yet they still included a modest sample of child-appropriate medicines. Further analyses on a dataset with a higher number of age-appropriate medicines are thus required, which may need to be collected prospectively. Although the relevance of the findings to the current situation of Burundi, China and Haiti is limited because of the older data, the aim of providing proof of concept of the adapted methodology was achieved nonetheless. Finally, the individual facility data that support the findings of this study are not publicly available, but aggregated data per medicine and country can be obtained from the HAI website.51 The aggregated data are sufficient to allow initial comparison of our methodology to previously existing tools.