Effect of a multifaceted social franchising model on quality and coverage of maternal, newborn, and reproductive health-care services in Uttar Pradesh, India: a quasi-experimental study

BACKGROUND
How to harness the private sector to improve population health in low-income and middle-income countries is heavily debated and one prominent strategy is social franchising. We aimed to evaluate whether the Matrika social franchising model-a multifaceted intervention that established a network of private providers and strengthened the skills of both public and private sector clinicians-could improve the quality and coverage of health services along the continuum of care for maternal, newborn, and reproductive health.


METHODS
We did a quasi-experimental study, which combined matching with difference-in-differences methods. We matched 60 intervention clusters (wards or villages) with a social franchisee to 120 comparison clusters in six districts of Uttar Pradesh, India. The intervention was implemented by two not-for-profit organisations from September, 2013, to May, 2016. We did two rounds (January, 2015, and May, 2016) of a household survey for women who had given birth up to 2 years previously. The primary outcome was the proportion of women who gave birth in a health-care facility. An additional 56 prespecified outcomes measured maternal health-care use, content of care, patient experience, and other dimensions of care. We organised conceptually similar outcomes into 14 families to create summary indices. We used multivariate difference-in-differences methods for the analyses and accounted for multiple inference.


FINDINGS
The introduction of Matrika was not significantly associated with the change in facility births (4 percentage points, 95% CI -1 to 9; p=0·100). Effects for any of the other individual outcomes or for any of the 14 summary indices were not significant. Evidence was weak for an increase of 0·13 SD (95% CI 0·00 to 0·27; p=0·053) in recommended delivery care practices.


INTERPRETATION
The Matrika social franchise model was not effective in improving the quality and coverage of maternal health services at the population level. Several key reasons identified for the absence of an effect potentially provide generalisable lessons for social franchising programmes elsewhere.


FUNDING
Merck Sharp and Dohme Limited.


Introduction
Over the past few decades, India's maternal mor tality ratio has declined substantially from 437 deaths per 100 000 livebirths in 1992-93 to 167 deaths per 100 000 livebirths in 2011-13. 1,2 Despite these improve ments, maternal health still requires urgent attention. India is the second largest contributor to the global burden of maternal deaths, accounting for 15% of all maternal deaths. 3 Maternal mortality remains high in Uttar Pradesh, India's most populous state, with the most recent estimate of maternal mortality at 285 deaths per 100 000 livebirths. 2 The Indian Government has had some success in increasing facility births. 4 How ever, concerns about quality of care and the capacity of the public sector to meet the increased demand for institutional deliveries need to be addressed. 5 Whether the private sector can be harnessed to improve health is at the forefront of ongoing debates in India and internationally. 6,7 India's private healthcare sector is extensive and diverse. It ranges from sophisticated tertiary hospitals, which provide medical care of an international standard, to unqualified rural healthcare providers and alternative systems of medicine. Most registered doctors work in the private sector, which is often the first point of contact for a substantial proportion of the population. 8-10 Evidence on the most effective strategies to improve the quality of private sector services remains scarce. 7,11,12 Regulation of the private sector in India has proved challenging and alternative strategies that encourage private providers to raise standards are required.
One prominent strategy is social franchising, an organi sational model that applies the principles of commercial franchising for socially beneficial ends. Social franchises Effect of a multifaceted social franchising model on quality and coverage of maternal, newborn, and reproductive health-care services in Uttar Pradesh, India: a quasi-experimental study Sarah Tougher, Varun Dutt, Shreya Pereira, Kaveri Haldar, Vasudha Shukla, Kultar Singh, Paresh Kumar, Catherine Goodman, Timothy Powell-Jackson Summary Background How to harness the private sector to improve population health in low-income and middle-income countries is heavily debated and one prominent strategy is social franchising. We aimed to evaluate whether the Matrika social franchising model-a multifaceted intervention that established a network of private providers and strengthened the skills of both public and private sector clinicians-could improve the quality and coverage of health services along the continuum of care for maternal, newborn, and reproductive health.

Methods
We did a quasi-experimental study, which combined matching with difference-in-differences methods. We matched 60 intervention clusters (wards or villages) with a social franchisee to 120 comparison clusters in six districts of Uttar Pradesh, India. The intervention was implemented by two not-for-profit organisations from September, 2013, to May, 2016. We did two rounds (January, 2015, and May, 2016) of a household survey for women who had given birth up to 2 years previously. The primary outcome was the proportion of women who gave birth in a health-care facility. An additional 56 prespecified outcomes measured maternal health-care use, content of care, patient experience, and other dimensions of care. We organised conceptually similar outcomes into 14 families to create summary indices. We used multivariate difference-in-differences methods for the analyses and accounted for multiple inference.
Findings The introduction of Matrika was not significantly associated with the change in facility births (4 percentage points, 95% CI -1 to 9; p=0·100). Effects for any of the other individual outcomes or for any of the 14 summary indices were not significant. Evidence was weak for an increase of 0·13 SD (95% CI 0·00 to 0·27; p=0·053) in recommended delivery care practices.
are networks of private providers that pay a fee to operate under contract with a common agency under a single brand. In return, the franchiser markets the brand and supports the provider to adhere to quality stand ards through training, clinical protocols, drugsupply management, and new technologies such as telemedicine. 13 In 2014, franchises reached almost 30 million people in lowincome and middleincome countries (LMICs), with most funding coming from international donors. 14 Although considerable resources are being channelled to social franchising in LMICs, evidence from rigorous studies on the effectiveness of clinical social franchising is scarce, 15-17 and this gap in knowledge urgently needs to be addressed.
In this study, we report results of an impact evaluation of the Matrika social franchise programme, implemented by World Health Partners (the franchiser) in partnership with Pathfinder International in India. We aimed to determine whether the social franchise model could improve the quality and coverage of healthcare services along the continuum of care for maternal, newborn, and reproductive health.

Study design
The Matrika programme was a complex multifaceted intervention that sought to improve maternal health primarily by leveraging the private sector. The basic approach combined various activities to encourage more women to use services and raise the quality of antenatal care, obstetric care, and family planning services (panel). The core component of the programme was the Sky social franchise network of private providers, but it was also recognised that the capacity of, and linkages with, the public sector would need to be strengthened if the programme was to have an effect. The intervention was implemented in three districts (Kannuaj, Kanpur Nagar, and Kanpur Dehat) of the Indian state of Uttar Pradesh between September, 2013, and May, 2016. Our process to understand the theory of change for Matrika began with a meeting between the evaluation team and the implementing partners in December, 2013. 18 Over the course of implementation, we collected data on imple mentation and updated the theory of change to reflect adaptations (appendix).
The study was done in the three intervention districts and three neighbouring districts (Auraiya, Etawah, and Fatehpur), with a combined population of 13·7 million and facility births ranging from 51% to 62%. 19,20 The study districts were demographically similar to the rest of the state, according to the data from the Indian Census 2011 (appendix). The private market for maternal health care in the study area was largely made up of small, individually owned hospitals and clinics located in urban and periurban areas. Most facilities were owned by doctors formally qualified in allopathic medicine and, to a lesser extent, providers of ayurveda, yoga and naturopathy, unani, siddha, and homoeopathy (or AYUSH).
The study comprised a quasiexperimental impact evaluation, a process evaluation, and a costing and financial sustainability analysis. The study received ethics approval from the Indian Council of Medical Research (HMSC/2014/10/HSR), Public Healthcare Society in India (10/Nov/2013), and the London School of Hygiene & Tropical Medicine (London, UK; 8610). Women gave written informed consent to participate in

Research in context
Evidence before this study We identified four systematic reviews that examined the effect of social franchise programmes in low-income and middle-income countries, the most recent of which was published in 2016. We updated the search in MEDLINE to March 15, 2017, with MeSH terms and keyword searches using the term "social franchising" and found one new study. Overall, the methodological rigour of studies was very poor, as indicated by the fact that no studies met the inclusion criteria in the Cochrane systematic literature review. The most recent review identified one randomised controlled trial and eight non-randomised controlled studies. Only two studies stand out for their methodological rigour. The first study examined a social franchising programme involving community health-care workers in Myanmar and showed that it increased treatment of childhood diarrhoea with oral rehydration solution containing zinc. The second study evaluated a social franchising programme for paediatric care in the state of Bihar (India) and found no measurable population effect on appropriate treatment for childhood diarrhoea or pneumonia. The remaining studies showed mixed results of the effect of social franchising on measures of use and quality of care. Social franchises have been documented in more than 40 countries, and in 2014, at least 90 social franchise programmes were in existence.

Added value of this study
To our knowledge, this study is one of very few rigorous studies examining the effect of a social franchising programme in health. We showed that the multifaceted social franchise programme did not increase facility births, or any other dimension of care, as measured by a large number of secondary outcomes across the continuum of care for maternal, newborn, and reproductive health. Reasons for why the intervention did not work provide potentially generalisable lessons for social franchising programmes elsewhere.

Implications of all the available evidence
Our results place a higher burden of proof on governments and donors looking to invest in social franchising for maternal health. The design of future social franchising programmes should take account of past failures.