Madagascar is one of the least developed countries globally [13]. The country has a maternal mortality rate of 392 deaths per 100,000 live births, an under-5 mortality rate of 71 per 1000 live births, and an average life expectancy of 64 years [6]. Over 40% of health care expenses are paid out-of-pocket [14]. With over 80% of the population living in extreme poverty, financial barriers are a major hindrance to accessing health care [6,8].

This study draws on data from selected health facilities in 7 regions of Madagascar (see Multimedia Appendix 1), where extreme poverty rates range between 78% and 95% [8]. During the intervention, the region was hit by severe drought and famine, putting over 2 million people into acute food insecurity [9]. Simultaneously, the second wave of the COVID-19 pandemic affected the economy severely, leading to a 7% decrease in GDP and a nationwide recession [7,15].

This study analyzed a mobile money–based conditional cash transfer intervention, aimed at increasing health care utilization; it was implemented from February 2021 to July 2022. The inclusion criteria for patients to benefit from the intervention were (1) presenting for care at a participating facility during the intervention timeframe, (2) belonging to one of the eligible patient groups (detailed below), and (3) registering on the digital mTOMADY platform with a mobile money account ( support was available at the health facilities at point of care). The intervention covered 80% of patients’ expenses for medications and medical consumables through conditional cash transfers. Patients qualified for the intervention if they sought care for (1) acutely life-threatening conditions, (2) accidents or injuries, (3) pregnancy, childbirth, or postpartum care, or (4) pediatric care. The decision to include an eligible patient in the intervention lied with the treating health care provider, and patients could decline participation at any point.

The costs covered by the intervention were limited to medical consumables and medication due to donor requirements. It did not cover consultation fees, laboratory fees (excluding laboratory consumables), hospitalization, or indirect expenses (eg, transportation costs).

Facilities could express their interest in participating with the implementing NGO, which selected facilities based on previous collaborative experience and geographical location, prioritizing those in underserved regions with high poverty rates. The NGO staff members were employed at each participating facility to support the administration of the intervention.

Once a patient completed treatment, the health facilities submitted claims for reimbursement through a digital platform for health care–related payments developed and provided by the German-Malagasy NGO, mTOMADY [16]. The claims were filed by a local employee, usually an administrative clerk at the health facility. Claims contained patient sociodemographic data (eg, age, gender, family size), medical information (diagnoses, symptoms, quantities of medication received, and consumables used), and cost data (eg, prices of each type of medication and medical consumables). All claims underwent scrutiny for accuracy by a team of registered doctors at the implementing NGO’s central administrative level. Clarification was requested by this team in case of inconsistencies or missing data. Health facilities were reimbursed for approved claims using mobile money. Participating facilities received a small bonus payment (approximately USD 0.50) for each approved claim.

We used a convergent mixed-methods design for this study, in which qualitative data were analyzed based on the findings from the quantitative analysis [17]. We analyzed qualitative data to explain the differences in the impact of the conditional cash transfer intervention on health care utilization at participating facilities.

Ethical approval for all components of the study was obtained from the University of Heidelberg Ethics Committee (Heidelberg, Germany) under the registration number: S-982/2021. In addition, we obtained formal approval from the district health office, a regional subdivision of the Malagasy Ministry of Health, in each district in which data were collected. For all secondary analyses of patient data, the ethics committee waived additional informed consent as all data were anonymized before being forwarded to the research team. For primary data collection for qualitative interviews, written informed consent was obtained from each participant prior to the interview. All identifying data were removed from the interviews during transcription, and all interviews were pseudonymized before analysis. Participants received no compensation for their participation in this study.

We referred to the STROBE (Strengthening the reporting of observational studies in epidemiology) and SRQR (Standards for reporting qualitative research) reporting guidelines in the preparation of this manuscript [22,23].

We obtained register data from 11 out of 15 health facilities; 4 facilities refused to share their register data. Multimedia Appendix 2 contains details on these 11 facilities.

For the qualitative data, we analyzed data from 64 interviews, which lasted between 30 minutes and 1.5 hours, with an average interview duration of 47 minutes. Ten interviews were conducted with project implementation staff, 22 with health care providers, 17 with beneficiaries, 9 with nonbeneficiaries, and 6 with policymakers.

Overall, health care utilization decreased significantly across most facilities after the end of the intervention. The end of the intervention had a more pronounced effect on patient numbers per month in the long term, which decreased significantly in 39 of 55 (70%) ITS models across separate facilities and patient groups. In comparison, the point effect in patient numbers only showed a significant decrease in 40% (22/55) of all models. Figure 1 illustrates total patient numbers per month at each facility.

The facility with the most pronounced decrease in patient numbers after the end of the intervention was Facility 10, a private facility located in an urban area. Overall, the negative effects of the end of the intervention were most pronounced in maternity, pediatric, and surgical care; whereas, outpatient consultations were the least affected by the end of the intervention.

Table 1 shows the detailed results of the ITS models. The step change values indicate the immediate difference in patient numbers at the end of the intervention, whereas the slope change describes the change in patient numbers per month over time.

The qualitative interviews elucidated 7 key factors that differed between facilities where the ending of the intervention had a significant impact on health care utilization and those where it did not. These factors spanned across 3 levels: the community level, the facility level, and the level of the implementing NGO. Table 2 summarizes these factors, as determined in the qualitative analysis of in-depth interviews with key implementation stakeholders, sorted by factors and facilities, with and without significant changes in health care utilization after the end of the conditional cash transfer intervention.

Our study aimed to describe the impact of a mobile money–based conditional cash transfer intervention on health care utilization in a humanitarian setting, as well as to identify key factors that influenced the success of the intervention. Overall, our study showed a significant decline in health care utilization following the ending of a mobile money–based conditional cash transfer intervention in southern Madagascar. Data from 11 facilities showed a marked reduction in patient visits, particularly in maternity, pediatric, and surgical services, with outpatient consultations being the least affected. The qualitative findings highlighted key factors influencing intervention outcomes across three levels. These factors included community outreach, facility staff motivation, ease of the administrative processes of the intervention, digital literacy of facility staff, and the presence of NGO staff at facilities.

Our analysis showed that health care utilization decreased at most facilities when the conditional cash transfer intervention ended. This ties in with other findings from Madagascar, which demonstrated an increase in health care utilization when user fees were abolished [24]. This change was most pronounced for 3 patient categories: maternity patients, pediatric patients, and patients requiring surgery. Given that surgical care is particularly expensive and likely to cause catastrophic health expenditure in SSA [25-27], this is not surprising. Our findings indicate that patients might forgo necessary surgical care in the absence of financing methods for their care. The evidence on the costs of pediatric care is less comprehensive. However, several studies from SSA suggest that the costs associated with pediatric care for asthma, pediatric surgery, and pneumonia could be catastrophic for households [27-29]. Evidence on the impact of partial cost coverage or fee reduction for pediatric health services is however lacking.

Evidence on similar interventions using mobile money-based cash transfers for health care utilization in humanitarian settings is severely limited. Contrary to previous findings [30], our study revealed a significant decrease in health care utilization following the end of a cash transfer intervention. However, our study analyses a conditional cash transfer intervention, whereas previous studies examined unconditional cash transfers [30,31], in which the money may have been used for other purposes. Our findings do however align with evidence from nonhumanitarian settings, in which health care utilization increased in the presence of targeted cash transfer interventions [32,33], especially for maternity services [34]. Cash-transfer interventions in nonhumanitarian settings, especially in combination with other interventions such as health education or text-based reminders, have been shown to improve risky sexual behavior [35], and are perceived as useful in improving the adherence and management of patients with TB [36]. It must be noted that most cash transfer interventions used physical currency, limiting the comparability of our findings [30-34].

Despite an overarching impact of the conditional cash transfer intervention on health care utilization, the impact was not significant across all facilities. We identified 7 key factors that may explain these differences across facilities.

At the community level, differences in community outreach and thus the intervention-related knowledge of the population across communities was decisive. For this intervention community outreach was only conducted in certain communities. In facilities where the NGO implemented community outreach, this was done via CHWs and radio campaigns. CHWs are vital for improving health care-related knowledge, access to health care and health outcomes in the communities they serve [37,38]. As we did not interview CHWs for this sample, we cannot further elucidate the reasons behind the differences in community outreach they conducted; however, given the evidence on factors influencing CHW performance, including training, supervision, remuneration and workload [38,39], it stands to reason that similar factors were decisive in our setting.

At the facility level, facility motivation and mindset were key determining factors of the success of the intervention. Provider motivation has been described as a key factor in the implementation and adoption of new practices [38,39]. It is noteworthy that the described intervention may have impacted provider motivation and thus either positively or negatively reinforced preexisting motivation levels. For example, the intervention, leading to an increase in patient numbers at the facility, may have increased provider workload which, depending on its perceived manageability, has been identified to be a key determining factor of provider motivation [39]. Equally, the intervention required health care providers to take on additional tasks and may have negatively impacted provider motivation [39].

Bonus payments received for each claim filed, were perceived very differently across facilities. At some facilities where all staff members received benefits through bonus payments, they were perceived as motivating factors. This ties in with evidence from other contexts, which have shown that bonus payments can enhance provider motivation and potentially quality of care [40,41]. This effect is however conditional on how the distribution of bonus payments was perceived among health care providers. Bonus payments perceived as unjust or unfair decrease provider motivation [41-44 ,].

Another key factor was previous NGO interventions at the same facilities, which led to confusion at the facilities with the changes in the procedures of the new intervention. Given the sizeable percentage of public services that are delivered by NGOs in SSA [45], such negative effects of past NGO interventions on future interventions should be examined more closely.

Given the digital nature of the conditional cash transfer intervention, a lack of digital literacy at the facility level was a key factor hampering the intervention. This finding ties in with findings from other studies across SSA [46-51]. In the face of rising digitalisation of health care delivery, governments, NGOs, and multilaterals, should invest in improving the digital skills of health care providers to eliminate a key barrier to the implementation of digital health interventions in the future.

Two factors emerged as decisive at the implementing NGO level. First, NGO staff placed at each facility was perceived as an important success factor for the intervention. NGO staff alleviated health care provider workloads by supporting the intervention’s administration [52]. Despite this positive effect, the fact that NGO staff were paid directly by the NGO, may have reinforced perceived inequalities in remuneration and thus furthered pre-existing frustrations [43].

Another noteworthy finding of our study is that facilities which appeared more successful were facilities, where the intervention was implemented later, and which thus benefited from more established and well-running processes. This echoes previous findings on public health interventions, which identified “implementation infrastructure,” including robust administrative systems as a key success factor [40,41]. It is vital to consider these challenges for future, similar interventions, for example by incorporating piloting phases into implementation plans to establish robust administrative processes early on.

A key challenge for cash transfer interventions, especially in humanitarian settings, is sustaining the increases in health care utilization after financial support concludes. Based on the findings from our study we propose several strategies to enhance sustainability in similar interventions.

First, strengthening local health financing systems is essential for sustaining service utilization. Future interventions could partner with local governments, private, or non-profit health insurance schemes to transition conditional cash transfers into more permanent financial assistance for vulnerable populations. Developing mechanisms for partial subsidies through national health funds or community-based health insurance schemes can support continued access to essential services without relying solely on external funding.

Second, we observed that the increased health care utilization during the intervention period partly stemmed from targeted community outreach and health education efforts. Sustained CHWs in health promotion can maintain community awareness and perceived value of health care services. Future interventions should invest in CHW training and create resources to enable communities to make informed health decisions, encouraging health care utilization even without direct financial incentives.

Finally, the data analyzed in this study contained several indications that the intervention had a positive impact on the trust expressed by patients in health care providers and the health system. This was particularly observed at facilities where health care providers helped patients in signing up for the intervention, for example by allowing them to use their ID cards to obtain a SIM card, or where they supported patients who were struggling with copayments. This improved trust is crucial as it can lead to sustained improvements in health care-seeking behavior even after financial incentives end. Future implementers should prioritise trust-building measures—such as transparent communication, continuous outreach, and patient-centred care—to cultivate a health care environment where patients feel supported and motivated to seek timely medical assistance.

We conducted an ITS analysis without a control group. Given the public health emergencies in southern Madagascar in recent years, which impacted health care utilization in the region, we opted against a historic control group. In 2020 the COVID-19 pandemic heavily impacted the Malagasy population and health system [7]. In 2018-2019, a severe measles epidemic [49,53] and in 2017, a plague outbreak affected the island [52,54]. Choosing control data from health facilities in a different part of the country was equally unreliable. Even though the health system is weak across Madagascar, the south of the island has particular challenges. Geographical access barriers are particularly pronounced [8], and facilities, human resources, and equipment are particularly scarce [10]. Equally, extreme poverty is more pronounced in the study regions than in the rest of the country [8,55]. Lastly, the severe drought and subsequent famine were specific to the study region and did not affect other parts of the country [9].

Four out of fifteen health facilities refused to share their facility registers with us for quantitative analysis. As facilities did not need to give a reason for refusing to share their data, we cannot eliminate the possibility that there is thus a selection bias in our final sample. However, we included facilities with varying levels of impact of the intervention on health care utilization, suggesting that the overarching results were not skewed. Our qualitative data draws on a convenience sample of facilities, which includes most but not all facilities for which a quantitative analysis was conducted, meaning we may not have captured all factors that influenced the success of the intervention.

Further, our qualitative data were collected in Malagasy and translated into English, which may have introduced translation errors. However, all data were translated by a trained interpreter with previous experience in qualitative research and spot-checked randomly for consistency between recording and translation by a native Malagasy speaker.

In conclusion, our study shows that a mobile money–based conditional cash transfer intervention led to an increase in health care utilization in a humanitarian setting, particularly in surgical and pediatric cases. This indicates that similar interventions could be useful for mitigating the effects of future humanitarian crises on health care utilization. Several factors influenced the success of the intervention. Designers and implementers of future, similar interventions should proactively mitigate these factors. Additional researchers should be invested in filling the remaining evidence gaps regarding the use of mobile money–based cash transfer interventions in humanitarian settings.