Doctors Keep Workers Home: Vietnam Study Links Clinic Staffing to Migration

When a young worker in rural Vietnam weighs whether to leave home for a job in Ho Chi Minh City or an industrial province on the coast, the decision may hinge on something few migration models have bothered to measure: how well staffed the local clinic is. A new longitudinal study drawing on nearly a decade of national household survey data suggests that the density of medical personnel in a province is one of the more reliable predictors of whether working-age residents stay put or pack their bags for employment elsewhere. The finding, published in BMC Health Services Research, adds a health-systems dimension to one of the most consequential questions in development economics, and it arrives with a twist that complicates the intuitive story: not every form of health infrastructure holds people back.

The research team, led by Thang T. Vo of the Health and Agricultural Policy Research Institute at the University of Economics Ho Chi Minh City, set out to test a genuinely two-sided hypothesis. On one hand, reliable local clinical services could act as a retention force, giving households a reason to remain where their health needs can be met. On the other hand, formal care capacity, particularly hospital beds, might release families from caregiving obligations that tie them to a place, freeing a member to migrate for work. In other words, health services could either anchor people or unshackle them, and which effect dominates is an empirical question that has been largely unanswered in developing countries, where most of the world’s internal migration now takes place.

To answer it, the researchers assembled an unusually rich dataset. They linked four consecutive two-year transitions of the Vietnam Household Living Standards Survey, spanning 2010 to 2018, with province-level health and socioeconomic indicators covering all 63 Vietnamese provinces. The resulting analytical file contained 42,150 person-transition observations among individuals aged 15 to 60, of which 41,344 complete cases entered the main models. Migration was defined conservatively: a person counted as a work-related out-migrant only if a baseline household member left the household for employment during the following two years. That definition captures the practical, household-anchored reality of labor mobility in Vietnam, where one member often departs while the rest of the family stays behind.

The statistical architecture of the study reflects the nested structure of the data. The authors used multilevel mixed-effects logistic regression with random intercepts at both the household and province levels, allowing them to separate individual and family circumstances from the broader local environment. Fixed covariates included individual and household characteristics, provincial socioeconomic controls, and year fixed effects to absorb national shocks and secular trends. The headline result was an adjusted odds ratio: each one-standard-deviation increase in medical-staff density was associated with roughly 12 percent lower odds of subsequent work-related migration (AOR 0.883, 95 percent confidence interval 0.813 to 0.960). Translated into probabilities, that same one-standard-deviation increase corresponded to a difference of 0.564 percentage points in the adjusted migration probability, with a confidence interval running from minus 0.817 to minus 0.312 points. In a country where internal labor migration reshapes entire provincial economies, an effect of that size, applied across millions of working-age adults, is far from trivial.

But the beds told a different story. District-level bed density was positively associated with migration (AOR 1.087, 95 percent CI 1.023 to 1.154), and commune-level bed density showed a similar pattern (AOR 1.063, 95 percent CI 1.007 to 1.121). This is precisely the caregiving-release mechanism the authors had hypothesized: when inpatient capacity is available nearby, families can delegate the care of elderly or chronically ill relatives to formal institutions, and a working-age member becomes free to chase wages elsewhere. Yet the authors are careful about how far this interpretation can be pushed. In sensitivity analyses that varied model specifications, the bed-density associations proved less stable than the medical-staff finding, and the authors describe the evidence for caregiving release through bed capacity as weaker and specification-dependent. The contrast between the two infrastructure types is itself informative: staffing, which reflects ongoing clinical relationships and continuity of care, appears to anchor people, while physical capacity, which substitutes for family labor, may do the opposite.

Health insurance emerged as a third, and arguably the most socially consequential, signal. Individuals who owned health insurance were substantially less likely to migrate for work in the following two years (AOR 0.722, 95 percent CI 0.594 to 0.879), a nearly 28 percent reduction in odds. The authors resist a single explanation, and for good reason. Insurance coverage in Vietnam is often tied to formal employment, so the association may partly reflect that insured individuals already hold stable jobs and therefore have less reason to move. But it may also reflect genuine mobility frictions: when benefits are difficult to use outside one’s home province, an insured worker who migrates risks losing access to subsidized care. The finding thus doubles as an argument for benefit portability, a policy lever that could decouple health security from geographic immobility.

The study also probed how these associations vary across the income distribution, and one interaction stood out. The negative association between pharmaceutical-staff density and migration grew stronger as household income rose, and it remained statistically significant after Benjamini-Hochberg correction for multiple testing (interaction p = 0.007, q = 0.043). That pattern hints at a mechanism rooted in household health capital: better-off families, who can afford medications and have more assets to protect, may be particularly sensitive to the availability of pharmaceutical expertise when deciding whether staying home is compatible with managing a family member’s illness. Meanwhile, alternative indicators the authors tested, including measures of service use and diagnostic capacity, did not reach significance, underscoring that the migration-health link is not a generic effect of any medical presence but something specific to the clinical workforce.

The authors are explicit about the limits of their design. These are observational, conditional associations, not causal effects, and the negative medical-staff association, while the most consistent finding across alternative specifications, was not invariant to every modeling choice. Unobserved provincial characteristics, from labor-market vibrancy to amenity value, could correlate with both health staffing and migration propensities. Reverse causation is also conceivable: provinces that people leave may struggle to recruit and retain doctors, so migration could partly drive staffing rather than the reverse. The biennial structure of the survey, with health conditions measured before migration outcomes, mitigates but does not eliminate these concerns. What the study offers is not proof but a disciplined, nationally representative description of a pattern that theory predicted could exist in either direction.

Why does this matter beyond Vietnam? Internal migration is the demographic engine of most developing economies, and policymakers routinely treat it as a function of wages, jobs, and education. This study argues that the local health system belongs in that equation, both as an anchor and as a release valve. The policy implications cut in two directions at once. Strengthening local clinical capacity, continuity of primary care, and community and long-term-care support could help provinces retain working-age residents, which matters for places worried about hollowing out. Simultaneously, making health-insurance benefits portable across provinces would remove a friction that may currently trap workers in places with fewer economic opportunities. Health policy and migration policy, the study suggests, are not separate portfolios but two views of the same household calculus, and any government that ignores one while engineering the other is likely to be surprised by the outcome.

There is also a quieter, more human reading of the numbers. Behind each odds ratio is a family deciding whether it can afford to let a son or daughter leave when a grandmother needs dialysis, or whether a mother will stay because the commune health station can manage her diabetes. As Vietnam’s population ages and its internal labor market continues to integrate, the interplay between caregiving and mobility will only intensify. The Vietnamese researchers, funded by the University of Economics Ho Chi Minh City, have produced a template that other developing countries can follow: link household panels to local health-system data, model the hierarchy explicitly, and test competing mechanisms rather than assuming a single direction. The result is a rare piece of evidence showing that the geography of doctors, beds, and insurance cards quietly shapes the geography of opportunity, one household decision at a time.

Subject of Research: The association between local health service delivery and labor out-migration in Vietnam

Article Title: Health service delivery and subsequent labor out-migration in Vietnam: a longitudinal multilevel study

Article References: Health service delivery and subsequent labor out-migration in Vietnam: a longitudinal multilevel study. (n.d.). https://doi.org/10.1186/s12913-026-15829-9

Image Credits: AI Generated

DOI: 10.1186/s12913-026-15829-9

Keywords: Vietnam, labor migration, health service delivery, health workforce, health insurance, caregiving, hospital beds, multilevel analysis, VHLSS, health systems, internal migration, BMC Health Services Research