Postpartum family planning in Rwanda: a cost effectiveness analysis

Dr. Wall,
 
Thank you for your thorough review and attention dated April 8, 2019. We have carefully checked the comments and revised the manuscript accordingly. Our responses are given in a point by point manner below.
 
Sincerely,
 
Author Team
 
Abstract
Reviewer comment (RC) 1:
Please define the ‘postpartum’ time period (for example, different sources may define this period as 1 year, 2 years, etc.).
Response (R) 1:
Thank you for providing the opportunity to clarify. For this analysis the postpartum time period is two years, reflected in the time horizon value. This has been further defined in the methods for additional clarification.
 
RC2:
“$2.8 million US$ per year can be saved if LARC is included as a contraceptive choice across all health centers in Rwanda…” Does this mean if LARCs (defined here as injectables and implants) were included at the few (primarily Catholic) clinics where these methods are not offered? Injectables and implants are already offered at the majority of government clinics in Rwanda.
R2:
Thank you for pointing this out. This supplemental perspective was moved out of the abstract – it was originally provided in for the original audience of this research but is not a primary result and is thus now only represented in the body of the paper. We’ve also adjusted the cost savings to reflect more conservative estimates.
This value was calculated using the current population level, birth rate, and percentage of unplanned births in Rwanda. For one woman, the use of a LARC method during the two year postpartum period saves $18.73. If we consider the total estimated unplanned pregnancies of this population conceived before the recommended interval, and estimate 25% of these can be transitioned to the use of LARC methods, approximately $951,000 per year can be saved be saved averting pregnancies with LARC methods.
12.6 million = 2018 population of Rwanda
37% unplanned pregnancies = reflected by median pregnancy timing (50% conceived before recommended interval)
32.23/1,000 = 2018 birth rate
3.88 = 2016 births per woman
203,049 births x .25 x $18.73 = $951,000 per year
 
Introduction
RC3:
In addition to the WHO citation, I suggest the authors cite the Rwandan MOH stance on the importance of PPFP (http://ec2-54-210-230-186.compute-1.amazonaws.com/wp-content/uploads/2018/02/Govt.-of-Rwanda-FP2020-Commitment-2018-Update.pdf).
R3:
This has been included.
 
RC4:
Reference 5 is from 1975 and does not pertain to Rwanda.
R4:
This reference has been removed.
 
RC5:
Reference 4 is not referenced correctly.
R5:
The missing title and hyperlink for this source have been added.
 
RC6:
I do not see either data point (76% or 26%) given in reference 6.
R6: Thank you for pointing this error. The references have been renumbered and this statement accurately reflects the source.
 
RC 7:
A 26% unmet need in postpartum periods (if the authors are referring to postpartum periods, as the sentence implies) is incorrect. The number is much higher in Rwanda where the unmet need in the postpartum period is closer to 50%. See: https://www.k4health.org/sites/default/files/Rwanda%202010%20DHS%20Reanalysis%20for%20PPFP_Final_0.pdf
R7:
The link provided by the reviewer was no longer working but infer that they were directing here: https://www.k4health.org/toolkits/ppfp/rwanda-2010-dhs-reanalysis-ppfp. These results were pulled from Rwanda’s 2010 Demographic Health Survey (DHS). This analysis used the 2015 DHS (reference 13, page 96). Much consideration was given to the value stated for unmet need. A range of reported PPFP unmet need values exists in the literature, however, we selected the value most specific to the population focus for this analysis: postpartum, married women in Rwanda. An element that may contribute to this discrepancy is the difference between unmet need for spacing versus limiting. The literature supports that there is a 26% unmet need for spacing and a 25% unmet need for limiting. Overall this results in a 51% unmet need for the general population. Our analysis focused specifically on spacing in the postpartum period.
1          Moore Z, Pfitzer A, Gubin R, Charurat E, Elliott L, Croft T. Missed opportunities for family planning: an analysis of pregnancy risk and contraceptive method use among postpartum women in 21 low- and middle-income countries. Contraception 2015; 92: 31–9.
 
RC8:
It needs to be mentioned that while faith-based health facilities may not provide family planning in Rwanda, such health facilities must have a family planning post or point of access nearby and refer patients to those posts (which is mentioned in reference 9 and here: https://pdf.usaid.gov/pdf_docs/PA00HQSV.pdf).
R8:
Thank you for highlighting the importance of the inclusion of this information. This detail has been added.
 
RC9:
“Some denominations…” Is this analysis comparing scaling up injectables and implants in the relatively few clinics which do not offer family planning due to religious reasons, versus the current situation in which those clinics do not offer any family planning (including OCPs and condoms), but instead refer women to nearby health posts?
R9:
Thank you for pointing out the important distinction between accessibility and uptake. This distinction wasn’t sufficiently clear in our analysis. We have reframed it to make it more clear that we are talking about uptake. Rwanda provides a suite of contraceptive options at its health facilities, and nearby health posts for facilities managed by some denominations. The purpose of this analysis was to illustrate the cost effectiveness of PPFP methods specifically for women in the postpartum period to facilitate birth spacing, an objective in alignment with Rwanda MOH’s interests to scale up PPFP.
 
RC10:
From reading the introduction, I think the concept of the paper is potentially problematic. The authors make the case that faith-based organizations do not provide family planning methods and then say “Thus, attention to potential cost savings with the inclusion of these methods at all of Rwanda’s health facilities warrants further attention.” The vast majority of government facilities in Rwanda are already providing the full range of contraceptive options (acknowledging occasional method stockouts and limited supplies of IUDs which have lower demand and are not modeled here). The faith-based organizations all now have nearby health posts that provide family planning and to which the faith-based organizations refer to (as well as community health workers who now provide some methods like the injectable in the community). Are the authors arguing that the faith-based (namely Catholic) organizations should themselves have to provide all family planning methods even though the nearby posts provide such methods? To what facilities are the authors suggesting that LARC (defined here as implants and injectables) be added? Additionally, the comparator in this analysis is offer of OCPs and condoms, which the Catholic facilities are also not providing.
R10:
Thank you for providing an opportunity to clarify this point. The authors do not suggest the LARC methods be added, as most facilities provide access either directly or through health posts. The focus of this analysis was to provide information on the cost effectiveness of the four highest utilized method types to inform the ministry of health’s future planning and public health messaging campaign focus to improve health outcomes in the postpartum period. Our inclusion of the statement: “Thus, attention to potential cost savings with the inclusion of these methods at all of Rwanda’s health facilities warrants further attention,” is not a focus of our analysis, rather an information point included in the context of areas which depend on faith-based facilities for contraceptive access and thus dependent on health posts. No research on usage rates comparing PPFP usage or contraceptives in general specific to faith-based facilities and their corresponding health posts versus government health clinics exists and is an area for future research.
 
RC11:
The rationale for looking at postpartum periods specifically as it is modelled is not totally clear – are the authors saying that postpartum provision of family planning services is lower than for non-postpartum family planning (I certainly recognize that demand is lower and thus unit need during postpartum periods is higher than for non-postpartum periods, but this article seems to be comparing different service provision scenarios)?
R11:
The postpartum period has been identified as an opportunity to continue the country’s trend towards improved reproductive, maternal, and neonatal health (RMNCH). Postpartum family planning (PPFP) is one of the most effective methods to improve RMNCH outcomes and prevent unintended or closely spaced pregnancies following childbirth.[2,4–6] One in two maternal deaths can be prevented in high-risk groups with appropriate access and use of family planning services.[4]
2.         Ankomah A. Myths, misinformation, and communication about family planning and contraceptive use in Nigeria. Open Access J Contracept. 2011;Volume 2: 95. doi:10.2147/OAJC.S20921
4.         Winikoff B, Sullivan M. Assessing the role of family planning in reducing maternal mortality. Stud Fam Plann. 18: 128–43. Available: http://www.ncbi.nlm.nih.gov/pubmed/3617120
5.         Byrne A, Morgan A, Soto EJ, Dettrick Z. Context-specific, evidence-based planning for scale-up of family planning services to increase progress to MDG 5: health systems research. Reprod Health. 2012;9: 27. doi:10.1186/1742-4755-9-27
6.         High Impact Practices in Family Planning (HIPs). Immediate postpartum family planning: A key component of childbirth care. [Internet]. Washington, DC; 2017. Available: https://www.fphighimpactpractices.org/wp-content/uploads/2017/10/Immediate-PPFP.pdf
 
RC12:
This article is comparing provision of LARC (injectables+implant) along with OCPs+condoms versus OCPs and condoms alone. Again, I don't think that any facilities in Rwanda are only providing OCPs+condoms.
R12:
The purpose of this model was to compare the two most commonly utilized longer-acting methods with the two most commonly utilized shorter acting methods in Rwanda. This is stated in the Abstract and last paragraph of the Introduction. The “Non-LARC only” arm of the model is needed to provide a comparator by which to determine the cost effectiveness of the longer acting options. It does not represent one type of facility versus another.
 
RC13:
One of the most cost-effective methods of contraception (with very low commodity costs), the copper IUD, is not included in this analysis. This LARC method has relatively low uptake and relatively few providers are trained to insert this method, particularly postpartum. I would strongly encourage the authors to include increasing the provision of the copper IUD as a scenario in their modelling.
R13:
Thank you for this point. As stated in the Abstract and Introduction, only the most common methods currently used by postpartum women in Rwanda were included in this analysis in order to more accurately model contraceptive usage and thus realistic cost. Very low acceptance of IUDs currently exists (<0.07%) (Rwanda DHS 2016). Despite the copper IUD’s cost effectiveness, cultural acceptance has not yet occurred and is thus not relevant for this analysis. An investment in current attitudes and behavior change initiatives would be required before this is applicable in the Rwandan context and could relevant information to shift method use.
 
RC14:
The authors may be better served by reconceiving the research question to compare scenarios where uptake of postpartum family planning (including the IUD) is increased (via demand generation and IUD skills training), recognizing that implant, injectable, condom, and OCP provision are already being provided in most clinics.
R14:
The purpose of this research document is to provide cost information for currently utilized PPFP methods. It does not aim to project costs for future training and uptake of methods not currently accepted.
 
Methods
RC15:
Table 1 specifies a 3% discount rate applied to costs, but this is not mentioned in the methods. Were the effects also discounted? The table should clarify that this is 3%/year, if in fact the discount rate was applied annually. The column heading says ‘estimate 2 years (USD)’ so this is unclear.
R15:
Table 1 Key Input Parameters values are recorded with respect to the two year time horizon of the analysis. Yes a 3%/year discount rate was applied.
 
RC16:
I disagree with combining injectables and implants and naming them ‘LARC’. Typically, LARCs are defined as the IUD and implant.
R16:
This analysis defined longer-acting methods as those not requiring intervention by the user on a daily or per sexual encounter basis.
 
RC17:
Age of consideration in the analysis is 12-49, but the model parameters come from DHS data which uses a 15-49 age range.
R17:
Thank you for pointing out this discrepancy. An age adjustment has been made to match the age provided in the DHS data.
 
RC18:
Breastfeeding postpartum women (the majority of postpartum women breastfeed in Rwanda) are not eligible for injectables per WHO guidelines until they are 6 weeks postpartum. Is this considered in the model?
R18:
Thank you for highlighting this point. For this model we used the higher dollar value to illustrate the more conservative perspective, representing the full 104 weeks ($22.15), as opposed to 98 weeks ($20.87) for this method. This representation of cost varies in error in favor of the non-LARC methods in the model, and thus if the injection supplier, type, and cost stays the same, it makes LARC methods more cost saving then shown. Reports have indicated that supply chains for the current injection method may not be reliable, thus, we selected the more conservative cost to be more inclusive of supplier and injection type changes.
 
RC19:
POPs are available to postpartum women who are breastfeeding but COCs are recommended for non-breastfeeding women beginning at 3 weeks. Have these timing issues and differences in POP versus COC commodity costs been considered?
R19: Thank you for this point. Per WHO, “The number of users of combined orals in any situation almost always exceeds the number of users of progestin-only orals. In addition, many service providers do not have clear guidelines about how to counsel women who might be eligible to use progestin-only orals—this makes it more difficult to accurately estimate the demand for POPs.” http://apps.who.int/medicinedocs/documents/s21863en/s21863en.pdf
For this reason, we did not differentiate between types of oral contraceptives. Effectiveness rates between the two are very similar.
 
RC20:
If ‘45% of women do not use contraception postpartum’ as stated, how is this represented in the decision tree? In the no-LARC arm, 47% of women do not contracept. But in the LARC+non-LARC arm, only 23% of women do not contracept.
R20:
The three nodes stemming from the “LARC+Non-LARC” arm are weighted. LARC (0.48) +Non-LARC (0.29) + No Contraception (0.23) = 1, or 100%; the same method applied to the “Non-LARC Only” arm. Decision nodes must equal 1; thus, as defined by our parameters of the family methods included in this analysis, values were weighted based on usage prevalence.
 
RC21:
How do the authors arrive at the estimate that 48% of women will select LARC (per the decision tree)?
R21:
See Response 20.
 
RC22:
The methods are very underdeveloped. It would be helpful if the authors describe: the process for analyzing for the most influential variables in the model, the sensitivity analyses, how the cost ranges were selected in the sensitivity analysis, and how method discontinuations were handled in the model. What methods were women who discontinued condoms, pills, or no contraception assumed to take up?
R22:
Thank you for highlighting an area that needed further explanation. The process for analyzing the most influential variables in the model was determined by a tornado diagram, which compares the relative importance of variables.
Sensitivity analyses were processed through the use of “chance” nodes in TreeAge Pro, which provides the functionality of sensitivity analyses within the software. The cost ranges were selected in the sensitivity analysis and are as stated in the second paragraph of the methods which states: “Input ranges were set to 50–150% of original value.”
In regard to method discontinuations in the model, discontinuation rates and subsequent uptake probabilities are provided in Table 1. Discontinuation was modeled after 1 year of use (1 year postpartum) with consideration for effectiveness tapering (effectiveness is maintained for 6 months following missed injection date). Additionally, limitation 3 provides further detail of the consideration given to discontinuation: “Tapering effectiveness upon discontinuation of injection method – It is reported that the injection method continues to provide protection for nine months following the last injection. This phenomenon is modeled through an altered discontinuation rate if injection follow-up stopped at nine months postpartum.”
 
RC23:
What are the units of the discontinuation rates in Table 1?
R23:
Percentage (%) – the unit indicated with the column heading of probability (the likelihood or chance of an event occurring).
 
RC24:
The second paragraph of the results should be in the methods section, and expanded upon.
R24:
This sentence was moved and expanded upon.
 
RC25:
Are the ‘chance of ‘pregnancy’ estimates given per year? The denominator should be stated. Regardless of the units, I do not think that the chance or pregnancy for injectables is 0.03% while for implant it is 1%. The implant is more effective than the injectable at pregnancy prevention per CDC typical-use failure rate estimates.
R25:
Thank you for pointing out this discrepancy. The chance of pregnancy value reflects the chance of pregnancy within the two year time horizon of this analysis. Both methods provide a <1% chance of pregnancy. To conduct the analysis from a more conservative perspective, both values could reflect the 1% chance of pregnancy in future iterations.
 
RC26:
It is unclear how the distribution of method use was derived in Table 1 and Figure 1. Does this assume the current distribution of uptake in Rwanda?
R26:
See Response 20.
 
RC27:
Table 1 footnote mentions that there was some weighting of estimates, but this is not described in the methods.
R27:
See Response 20.
 
RC28:
The outcome of interest (pregnancies averted) is never mentioned in the methods and should be explicitly stated.
R28:
The pregnancies averted outcome has been explicitly stated in the methods.
 
RC29:
Similar to the issues mentioned previously, the tile for Figure 1 states “The two primary arms comparing family planning menu types,” but the non-LARC only option is not an option at government facilities.
R29:
See Response 12.
 
RC30:
Table 1 footnote states “Value established from range”. What range? What does this refer to?
R30:
The ** asterisks in the table legend is reflected in the table for the “normal vaginal delivery” row. A range of costs for a normal vaginal delivery are cited in the literature. This is discussed in the citation provided (page 26).
 
RC31:
In Table 1, how is the cost estimate for pregnancy hospitalization taken from a study on post-abortion care (reference 18)?
R31:
The referenced study provided general Rwanda health systems costs and can thus costs for other facility based care.
 
Limitations
RC32:
It is worth mentioning that the societal perspective is not included, as is recommended.
R32:
Thank you for this point this has been added to the limitations. This analysis only provided output from a health systems perspective.
 
Conclusions
RC33:
“This evaluation illustrates the inclusion of LARC methods in contraceptive options results in savings of $18.73 per pregnancy averted compared to family planning menus that exclusively include non-LARC methods for women in Rwanda for two years following birth.” “…$2.8 million US$ per year can be saved if LARC is included as a contraceptive choice across all health center.” Again, it is unclear to me whether the authors modelled adding injectable/implant options to the very few clinics which do not offer them (which would primarily be clinics which offer no family planning methods at all, and not their counterfactual scenario of OCPs and condoms). If so, did the authors also model adding OCPs and condoms to those clinics?
R33:
Thank you for bringing to our attention the important distinction between availability and uptake. This model illustrates the cost effectiveness of longer acting methods compared to shorter acting. Based on these results, there is an opportunity for economic gains if there is greater uptake of longer acting methods; increased uptake is associated with more availability. See Response 2.
 
RC34:
With consideration for the country’s low postnatal care attendance rate…” First, this sentence needs a reference. Second, what are the authors referring to? A very high percentage of Rwandan women attend postnatal infant vaccination visits.
R34:
Thank you for this important point. Infant vaccination visits are currently not integrated into postnatal care. For further details see:
Williams, Pamela, Nathalie Kayiramirwa Murindahabi, Elizabeth Butrick, David Nzeyimana, Felix Sayinzoga, Bernard Ngabo, Angèle Musabyimana, and Sabine F Musange. “Postnatal Care in Rwanda: Facilitators and Barriers to Postnatal Care Attendance and Recommendations to Improve Participation.” Journal of Global Health Reports, 2019. (in press)
 
References
RC35:
Some references are incorrect, please double-check them (e.g., reference 7 is incomplete, reference 4 is not provided correctly, the link for reference 13 and 14 are the same/duplicated, referencing blogposts (reference 22) is not appropriate and the peer reviewed literature is preferred).
R35:
Thank you for bringing this error to our attention. References 4, 7, and 22 have been corrected, and reference 14 removed.