When richer families pay more for worse care: Evidence from childbirth

In rural Uttar Pradesh and Bihar, newborn mortality is 60% higher in private health facilities than public facilities, even though families with superior socioeconomic characteristics use private care. Do private facilities harm babies, or do they receive more families expecting difficult births? This article shows private facilities do cause worse outcomes for similar patients because fee-for-service payments create incentives to perform unnecessary, harmful interventions.

In India, the proportion of births delivered at health facilities has expanded rapidly in the past two decades. By 2020, over 90% of births occurred in hospitals or clinics rather than at home. This shift was meant to reduce maternal and newborn deaths by providing access to skilled medical care. Yet, in rural areas of the states of Uttar Pradesh (UP) and Bihar, the rate of newborn death remains stubbornly high – above that of any nation in 2020 except Afghanistan, Pakistan, and Nigeria.

In these two states, looking separately at the risk of newborn death among babies born in public health facilities and private facilities presents a puzzle (Verma and Cleland 2022). In private facilities, 51 babies per 1,000 births die in their first month. In public facilities, this figure is 32, nearly 40% lower. This gap persists even though families using private care pay on average Rs. 20,000 for delivery compared to Rs. 1,000 at public facilities. Mothers who give birth in private facilities are wealthier, taller, less likely to be underweight, more literate, and from more privileged caste groups. By every observed measure, private facility patients should have better outcomes. Instead, outcomes for these patients are much worse.

In principle, it could be that private facilities are seeing the hard cases, driving private mortality higher. Families expecting difficult births might seek out what they expect to be better care in private facilities, or public facilities might refer them to avoid worse statistics. In Coffey et al. (2025), I along with my co-authors, tested whether measured risk factors explain the gap by comparing mortality within groups of similar predicted risk and accounting for differences in wealth, caste, education, and maternal health. Contrary to the idea that more fragile patients choose private facilities, observable patient differences do not explain the gap; in fact, controlling for them widens it. But families might still be sorting on factors surveys do not capture – anticipated complications, previous experiences, health signals that providers and patients notice but surveys miss. In new research (Franz 2025), I examine whether private facilities cause worse newborn outcomes or whether apparent differences arise from unobserved differences in the families who use them. I show that the patterns of care at private facilities differ and these facilities do, in fact, generate higher mortality risk for similar patients.

Methodology and data

Identifying whether private facilities cause worse outcomes, rather than merely attracting riskier patients, requires comparisons where facility choice varies for reasons unrelated to individual health or expected complications. My study uses two complementary approaches.

The first addresses the problem that families expecting complications might select into private facilities. It compares mortality rates across villages with different public-private facility birth mixes. If families expecting complications were sorting into private facilities, we would see higher mortality among private facility births within any given village. But the village's overall mortality rate would not depend on what fraction of births occur in each facility type – the risks would average out. But if one type is harmful, villages with more births in that type will have higher mortality.

The second approach additionally addresses the problem that places with less access to public facilities might have worse underlying health. It examines births near district borders, where administrative boundaries cause otherwise similar families to differ in their choice of public or private facility. For instance, a village on one side of the border may be a short distance to its in-district facility. A neighbouring village on the other side of the district border may be much further from its in-district facility. The strategy compares mortality between births on either side of the borders and identifies the public-private mortality effect.

I analyse data from the National Family Health Surveys (NFHS) of 2015-2016 and 2019-2021, covering around 77,000 facility births in rural UP and Bihar (International Institute for Population Sciences, 2017 and 2021). The surveys record where each birth occurred, whether the baby survived the first month, and detailed household and maternal characteristics.

Findings

Both methodological approaches outlined above reach the same conclusion: Private facility birth causes substantially higher newborn mortality. The effect is large, over 25 additional deaths per 1,000 births. This translates to over 110,000 lives saved each year by public facilities in rural UP and Bihar. Extrapolating this effect linearly, private facilities adopting the practices used in public facilities would prevent an additional 37,000 newborn deaths annually. This reduction would make nearly 20% of the progress India needed in 2020 to meet UN Sustainable Development Goal 3.2 for neonatal mortality.

First, I discuss the results of the village-composition strategy, which looks at the relationship between neonatal mortality and the fraction of births that take place in private facilities rather than public. This relationship is positive and statistically significant: for every 10-percentage point increase in a village's private facility birth share, mortality rises by about 3 per 1,000. This is evidence that giving birth in a private facility type itself increases risk. The strength of the relationship increases after controlling for household wealth, maternal education and health, caste, religion, sanitation, and other village characteristics. Villages with more private facility births are actually wealthier and have better health indicators, meaning this comparison works against finding a harmful effect of private birth.

The second approach, the district border analysis, provides complementary evidence using a different source of variation. I focus on borders where adjacent districts differ in their rates of public facility use. As shown in Figure 1, these rates can vary widely for even adjacent districts. At these borders, families living just a few kilometers apart face different costs of using public facilities – for instance, distance to the nearest public facility in the district might increase or support from community health workers might decline.

Figure 1. Variation in fraction of district's facility births born in public health facilities

Source: NFHS-4 and NFHS-5.

Notes: (i) The figure displays a map of district borders in the states under study, Uttar Pradesh and Bihar. (ii) The region color shows the share of a district's facility births that occur in public facilities. (iii) Larger fractions are drawn in darker shades. (iv) The fraction of the district born in public facilities is based on survey-weighted births aged 1-59 months that were born in a health facility to women living in rural areas of UP and Bihar at the time of interview.

These district boundaries create discontinuous changes in facility use: Crossing from a district with lower private facility use to one with higher private use, births shift by about 8 percentage points toward private facilities. Figure 2 shows that, at these same borders, neonatal mortality increases by 11 per 1,000. If we attribute the entire mortality shift to the shift towards private facility birth, this implies private facility birth increases neonatal mortality by about 120 deaths per 1,000 births. This estimate is uncertain, with a ‘confidence interval’ ranging from 5 to 240 per 1,000, but the direction is clear and consistent with the village-composition analysis.

Figure 2. Crossing from a district with less private facility birth to a district with more, mortality becomes more likely

Source: NFHS-4 and NFHS-5.

Notes: The figure displays the results of linear regressions using the estimated mean-squared-error-minimising bandwidth. It presents neonatal morality as a function of distance from the border. For every pair of adjacent districts, observations from the district with a lower fraction born in private are on the left side of the plots, with negative distances; the districts with a greater fraction born in private are on the right, with positive distances. The plotted points are a weighted binscatter of deciles on each side of the cutoff. Observations are births aged 1-59 months that were born in a health facility to women living in rural areas of UP and Bihar at the time of interview, with a top-tercile (9.5 percentage points) cross-border difference in the fraction of rural births in private facilities. Regressions are survey-weighted with standard errors clustered at the village level.

Mechanism: Over-provision of harmful care

I present evidence that private facility birth is more dangerous because private providers more often perform harmful interventions (Kumar et al. 2025). Patients do not know what medical care is best, so they rely on providers' recommendations (McGuire 2000). Private providers charge fees per service and so have incentive to recommend and perform interventions regardless of medical necessity (Srivastav et al. 2023). For many conditions and in many contexts, this provider-induced demand may merely be wasteful; for newborns in rural India, these interventions are dangerous.

In this setting, potentially harmful practices include washing babies soon after birth (Varma, Khan and Hazra 2010), using electric warmers that may lack power or be improperly calibrated or used (Datta et al. 2017), suctioning airways with a plastic tube (Kumar, Kumar and Basu 2019), feeding babies before breastfeeding (Maria et al. 2022), and administering unnecessary antibiotics (Agarwal et al. 2021). Each of these increases the risk of death for the baby, by lowering body temperature, introducing pathogens, or actively damaging organ systems. No data exist on which specific interventions occur at each birth, but I look at the first step of any of these interventions: separation of the mother and baby.

Three patterns establish separation as the reason why private facility birth is more dangerous than public. First, if interventions explain the effect, then separations should occur more in private facilities and less in public facilities. Returning to the regression discontinuity evidence, at district borders where the use of private facility increases, separation rates jump from 25% to 35%. In other words, health workers at public facilities are less likely to keep a child away from its mother, whereas at private facilities health workers are more likely to interrupt mother-child contact immediately following birth. This difference in separation is a candidate explanation.

Second, if interventions explain the effect, then in places where public facilities and private facilities separate at similar rates, mortality should be similar. Testing this with a village-level regression of the public-private mortality gap on the public-private separation gap, I find this is the case. In villages where public and private facilities separate mothers and babies similarly often, the mortality gap disappears. The public advantage appears only where public facilities separate mothers and babies less than private.

Third, if interventions did not explain the effect, then the mortality pattern should not depend on whether the baby was separated. Looking at mortality outcomes separately for births that were separated from their mother and those that were not, the mortality relationship reverses within each group. Among births that were not separated, villages with more private facility births have slightly lower mortality, as expected given their wealthier populations. Among births that were separated, the same pattern holds. The overall mortality gap exists only because private facilities separate more often, and separation is associated with substantially higher mortality risk.

Policy implications

This research establishes that private facility birth causes higher mortality and identifies separation-enabled interventions as the mechanism. But it leaves open several questions critical for policy. Would training private providers on the dangers of unnecessary intervention reduce separation rates? The problem may not be knowledge. Separation and its associated interventions may persist because they generate revenue and appear to patients like active, valuable care (Das et al. 2016, Siddiqui, Nair and Coﬀey 2023). Could informing families about mortality risks shift demand away from facilities that separate? Perhaps these families value other attributes of private facilities – shorter wait times, more attentive and respectful service, cleaner facilities, or they simply do not believe public facilities could provide better care.

The most direct policy intervention would be stricter regulation of private healthcare quality. However, existing regulations in UP and Bihar are weakly enforced, and separation practices are difficult to monitor without direct observation of care. Fee-for-service payment structures are deeply embedded in India's private healthcare system, and moving toward bundled payments or salary-based compensation (a standard remedy for overutilisation in healthcare systems elsewhere) would require fundamental restructuring. The evidence here clarifies the problem and its mechanism, but finding effective policy solutions remains an urgent challenge.

Understanding why families continue to choose private facilities despite worse outcomes requires examining both supply and demand. It is possible that while some characteristics of private versus public facilities, such as crowdedness, are easily observable to patients and family decision-makers, the relatively rare outcome of child death is not. To answer this question requires a better understanding of the determinants of health than extant data sources can provide – especially, data that combine facility audits to document which specific interventions occur during separation, household surveys to understand how families evaluate facility quality, and choice experiments to measure how much families value different facility attributes. Evidence like this could inform targeted interventions to reduce the 37,000 preventable deaths occurring each year in private facilities across these two states.