A sequential and adaptive experiment to increase the uptake of long-acting reversible contraceptives in Cameroon
Last registered on December 06, 2018

Pre-Trial

Trial Information
General Information
Title
A sequential and adaptive experiment to increase the uptake of long-acting reversible contraceptives in Cameroon
RCT ID
AEARCTR-0003514
Initial registration date
October 30, 2018
Last updated
December 06, 2018 12:57 PM EST
Location(s)

This section is unavailable to the public. Use the button below to request access to this information.

Request Information
Primary Investigator
Affiliation
The World Bank
Other Primary Investigator(s)
PI Affiliation
HGOPY, Cameroon
PI Affiliation
UC San Diego
PI Affiliation
University of Exeter
PI Affiliation
George Washington University
PI Affiliation
Stanford University
Additional Trial Information
Status
In development
Start date
2018-12-03
End date
2020-12-31
Secondary IDs
Abstract
Cameroon exhibits a high and non-decreasing level of maternal mortality (roughly 600 per 100,000 live births), partially related to its relatively high total fertility rate (roughly 4.6). Survey evidence furthermore suggests that a significant fraction of these pregnancies is unwanted or considered mistimed by the mother, especially among females aged 15-19. Despite this, the rate of utilization of family planning (FP) is low: e.g. only 48% of sexually active unmarried women use any form of (modern) contraception, or MC, and even then, it is primarily condoms. The use of LARCs (long-acting reversible contraceptives, i.e. the IUD and implant) is less than 1% according to the most recent Demographic Health Survey.

We propose to use an integrated behavioral science approach to increase the take-up of both SARCs (short-acting reversible contraceptives, i.e. the pill and injectable) and especially LARCs among reproductive-age females in Cameroon, including adolescents who may be unmarried and/or nulliparous. In addition to decreasing maternal mortality and undesired pregnancies, indirect effects for the community will include: increased welfare from reduced side effects that arise due to current one-size-fits-all FP counseling; healthier children due to improved birth spacing; and increased human capital formation both for children and for young (often school-aged) potential mothers.

We propose to conduct the study at HGOPY for a duration of 12 months. We hope to do so by providing tablets to each of five nurses that conduct FP counseling to women at the hospital. The tablets contain a counseling “app” (or decision-support tool or a job-aid) that was jointly developed by professionals from HGOPY, the World Bank, and the Ministry of Health. We propose an individually-randomized experiment, where the clients will be offered randomly varying discounts for the modern contraceptive methods they wish to adopt. We also propose to experiment with certain aspects of the “app” to improve its effectiveness – both for the client and for the nurse. More details on the experimental design are provided below.
External Link(s)
Registration Citation
Citation
Athey, Susan et al. 2018. "A sequential and adaptive experiment to increase the uptake of long-acting reversible contraceptives in Cameroon." AEA RCT Registry. December 06. https://www.socialscienceregistry.org/trials/3514/history/38507
Experimental Details
Interventions
Intervention(s)
The proposed study is randomized controlled trial (RCT) to be conducted at HGOPY. As the hospital’s interest is in improving the quality of family planning services it provides and increasing the demand for family planning services sought at HGOPY, the unit of intervention is a client presenting with a request for family planning services. As such, individuals will be randomly assigned to the interventions being evaluated under this study, which are described below.

1. Randomly varying the prices the individual clients face by offering discounts: at the end of each counseling session, which is conducted by a nurse using a tablet-based app (which can be described as a decision-support tool or a job aid), the health provider will offer the client a discount from the regular prices charged by HGOPY for long-acting and short-acting methods (LARCs and SARCs). We aim to investigate the following questions using this intervention:

a. Is the demand for (modern) contraception among females responsive to changes in prices?
b. Is there a discontinuity in demand at “price = 0”? Is the elasticity different for adolescents vs. adults?

The prices (or discounts) for LARCs and SARCs will be independently randomized in a 5x2 factorial design.

2. Randomly varying whether the “app” recommends one method at a time vs. “status quo”

Here, we would like to test, as closely as possible, the paradigm shift in FP counseling – from discussing all methods and letting the client state the method she would like to discuss to the “app” recommending a method to discuss first – based on the information elicited from the client during the session. Therefore, we would like to randomize, initially with equal probability, between two regimes (it is possible that we test four regimes, adding two and three choices at a time):

i. Status quo: The “app” displays all MC methods (up to four) that have not been ruled out by the client or medical eligibility up to that point. The choices are presented as unranked (i.e. as if equally suitable for the client) and the nurses will be trained to provide basic information on all choices (in order of the methods displayed, which may be randomized as well) and then allow the client to indicate which method they would like to discuss. From there, the procedure is the same (go to cue card, etc. and then adopt or choose another method and repeat – until a decision is made).
N.B. The preferred ranking will be known to the researchers, i.e. the app will still rank methods, but simply not display those rankings to the nurse counselor.

ii. Top recommendation: This is how the “app” currently works: it will display the MC deemed most suitable for the client given her preferences and ask her if she would like to hear about it. If she says ‘no,’ or if she says ‘yes’ but ends up not wanting to adopt the method, the next highest ranked recommendation will be displayed and the same process will be repeated until a final decision is reached.

The experiment, with longer-term follow-up, will shed light on what form of counseling works better to improve client welfare and satisfaction, as well as reducing the likelihood of discontinuations and unintended pregnancies.

3. Randomizing between two methods when there is a tie in rank:

The “app” bases its recommendations on the clients’ preferences regarding (a) side effects, and (b) how long she’d like to wait before becoming pregnant. Methods that the client does not want to consider and that do not meet medical eligibility criteria are then eliminated from the ranked list of recommendations.

When there is a “tie” between two methods, the tie is broken using a random number generator. Experimental evidence on the acceptability of LARCs for clients presenting to adopt SARCs (or people who are not sure about the method they would like to adopt) is scant, especially in the developing world. Randomization to break ties to recommend one method ahead of another will address this issue.
Intervention Start Date
2019-01-14
Intervention End Date
2020-06-30
Primary Outcomes
Primary Outcomes (end points)
The primary outcomes in this study are:

1. The client adopted a LARC or not, and
2. The client adopted a modern contraceptive method (LARC or SARC) or not.
Primary Outcomes (explanation)
These outcomes can be immediately measured using de-identified data from the tablets used by nurses.
Secondary Outcomes
Secondary Outcomes (end points)
The (individual-level) secondary outcomes in this study are "client experiences at HGOPY, satisfaction with their adopted contraceptive method, the (dis)continuation of their chosen method, and unintended pregnancy.

The (facility-level) secondary outcomes are:

1. The total number of counseling sessions conducted per week (or month),
2. The total number of modern contraceptives administered per week (or month), and
3. The ratio of LARCs to SARCs administered per week (or month)



Secondary Outcomes (explanation)
Individual-level secondary outcomes will be collected through follow-up surveys and can be analyzed experimentally.

Facility-level secondary outcomes will be aggregated through individual-level counseling session data from the tablets and will be analyzed non-experimentally (perhaps using an event-study type of analysis).
Experimental Design
Experimental Design
The three interventions described above will be randomized independently of each other at the individual level. So, all three interventions will be conducted simultaneously.

The data will be analyzed in batches using contextual multi-armed bandit (MAB) algorithms to update the random assignment probabilities in this adaptive experimental design.
Experimental Design Details
Not available
Randomization Method
Randomization by the tablet used by the nurse/health provider.
Randomization Unit
Individual client
Was the treatment clustered?
No
Experiment Characteristics
Sample size: planned number of clusters
1 hospital
Sample size: planned number of observations
No less than 1,500 and no more than 4,000 individual clients.
Sample size (or number of clusters) by treatment arms
equally distributed within each intervention initially, but then updated monthly based on the analysis of each batch.
Minimum detectable effect size for main outcomes (accounting for sample design and clustering)
Pending collection of baseline data to be fed into simulations of MAB algorithms to determine statistical power.
IRB
INSTITUTIONAL REVIEW BOARDS (IRBs)
IRB Name
COMITE INSTITUTIONNEL D'ETHIQUE DE LA RECHERCHE POUR LA SANTE HUMAINE (CIERSH)
IRB Approval Date
2018-11-09
IRB Approval Number
780/CIERSH/DM/2018