Determinants of Health

Why this matters in health economics

The instinct of every health system is to equate "health" with "healthcare" and to fight for health by funding more treatment. Yet the evidence of a century of public-health scholarship is that medical care, for all its value to the individual who needs it, is only a modest contributor to how long and how well a whole population lives. The larger share is produced upstream — by the conditions in which people are born, grow, work, live, and age. If you commission and evaluate as though healthcare were the main lever on population health, you will spend a rising budget for shrinking returns and wonder why the health of the poorest does not move.

This is a resourcing question, not a philosophical one. A director who understands the determinants of health can see why a housing programme, a school-meals policy, or a tobacco regulation may buy more health than an extra clinic — and can make that case with the same economic rigour applied to a drug. It also reframes what a health system is accountable for. The steep, unbroken relationship between social position and health — richer means healthier at every step of the ladder, not merely at the bottom — means that health is being manufactured, for good and ill, by decisions taken in finance ministries, planning departments, and boardrooms that never think of themselves as health actors.

The stakes are equity and money together. Because the causes of ill health are socially patterned, so are its costs: the same conditions that make people sick also make them poor, unemployed, and expensive to a health system late in the day. Naming those causes is the first step to acting on them upstream, where action is often cheaper and fairer. This chapter owns the causes — what produces health and its unequal social patterning. How that unequal result is measured and judged for fairness belongs to Chapter 3.4 — Equity; the two are a matched pair, causes and consequence.

Core concepts

The determinants of health. The determinants of health are the full set of factors that influence how healthy an individual or population is: genetics and biology, individual behaviour, medical care, and — dominating the rest at population scale — the social, economic, commercial, and environmental conditions of life. In Alan Williams' "plumbing diagram" of health economics (see Chapter 1.1 — Introduction to Health Economics), this is "box A": what influences health, the source of the flow that everything else in the diagram tries to value and manage. This chapter is the economics of box A.

Social determinants of health. The social determinants of health are the conditions of daily life — income, education, employment and job security, housing, food, early-childhood experience, and social inclusion — together with the structural drivers (the distribution of money, power, and resources) that shape them. They are not a soft addendum to "real" clinical medicine; they are, for a population, the principal production inputs of health. The World Health Organization's framing puts it bluntly: the conditions in which people live and work are the chief cause of health inequalities between and within countries.

The social gradient in health. The single most robust finding in this field is the social gradient: health improves stepwise as social position rises, across the entire range, not only for the destitute. The Whitehall Study of British civil servants — none of them poor, all in stable office jobs — found that mortality rose steadily with each step down the employment grade, an observation that reshaped the field and is closely associated with Michael Marmot. The gradient matters economically because it rules out simple explanations: if only absolute deprivation mattered, the middle of the ladder would look like the top. It does not.

Fundamental cause theory. Fundamental cause theory explains why the link between social position and health persists even as the specific diseases and risk factors change over time. Because higher-status people command flexible resources — money, knowledge, power, connections, prestige — they can deploy whatever protects health in any given era, so tackling one disease mechanism at a time never closes the gap; a new mechanism takes its place. For a health economist this is a warning: a narrow intervention aimed at one proximate risk may leave the underlying cause, and therefore the inequality, untouched.

Commercial determinants of health. The commercial determinants of health are the ways that private-sector activity — the products, marketing, supply chains, lobbying, and business models of industries such as tobacco, alcohol, ultra-processed food, and gambling — shapes health, often adversely and often at a profit that is privatized while the health costs are socialized. This is the externality logic of Chapter 1.3 — Market Failure applied to whole industries: the market price of a harmful product rarely carries its full health cost. Naming commercial determinants makes visible an actor that "social determinants" language can leave anonymous.

Environmental determinants. The physical environment — air and water quality, the built environment that makes activity easy or impossible, green space, housing quality, and neighbourhood safety — is a determinant in its own right. This chapter treats the environment as a cause of current health; the economics of climate change and planetary boundaries as a driver of future health costs is owned by Chapter 4.3 — Climate and Planetary Health Economics (cross-reference rather than repeat).

The life-course perspective. The life-course approach holds that exposures accumulate and that critical periods — especially early childhood and the first thousand days — cast long shadows over adult health. Economically, this creates long lag structures: investment in the early years may not show a health return for decades, which standard short-horizon return-on-investment tools systematically undervalue (a problem shared with prevention generally, see Chapter 3.2 — Health Policy).

The marginal contribution of medical care. A recurring, sobering theme in public health is that historical gains in life expectancy owed more to sanitation, nutrition, housing, and safer work than to clinical medicine, and that the marginal contribution of additional healthcare to population health is modest compared with the social determinants. This is not an argument against medical care, which is decisive for the individual who is ill; it is an argument about where the next unit of health-producing spend has the highest yield.

Population health and the prevention paradox. Population health takes the health of a whole group, and its distribution, as the unit of concern. Geoffrey Rose's prevention paradox captures a hard economic truth of acting on determinants: a preventive measure that brings large benefit to a population often offers little to each participating individual, so population-wide strategies that shift the whole risk distribution can achieve more than targeting high-risk individuals — yet are harder to sustain politically because no single beneficiary feels saved.

Confounding and the causation problem. Because determinants cannot be randomized — you cannot allocate people to poverty — most evidence is observational, and confounding is the central threat: the association between a determinant and an outcome may be produced by a third factor linked to both. Distinguishing a determinant that causes ill health from one that merely correlates with it is the discipline's hardest methodological problem, and the tools for it belong to Chapter 2.3 — Health Econometrics (cross-reference).

Best practices

1. Draw the whole production function of health, not just the healthcare slice. Before you argue for more treatment capacity, map where the health of your population actually comes from — income, housing, work, education, environment, and behaviour, alongside medical care. Making the upstream inputs visible reframes the budget question from "how much healthcare?" to "where does the next unit of health-producing spend yield most?" and often exposes non-health budgets as the higher-yield lever.

2. Design for the gradient, not only the bottom. Because health tracks social position across the whole range, a programme aimed solely at the poorest leaves most of the socially-patterned burden untouched — much of it sits in the large middle of the distribution. The pragmatic response, sometimes called proportionate universalism, is to act universally but with intensity scaled to need, capturing the whole gradient rather than a tail. State explicitly which you are doing, because the two buy different amounts of population health.

3. Name the commercial actor when a determinant is a product. When ill health is being manufactured and sold — tobacco, alcohol, ultra-processed food, gambling — "lifestyle" and "behaviour" language hides the producer and shifts responsibility onto the consumer. Treat these as commercial determinants and analyse them with the externality and market-failure toolkit of Chapter 1.3 — Market Failure: the private price omits the health cost, which is why fiscal and regulatory levers (owned by Chapter 3.2 — Health Policy) exist.

4. Cost the long lag honestly, and choose tools that can see it. Determinant interventions — early-years support, housing, education — often return health decades later, and a standard short-horizon return-on-investment calculation will score them as poor simply because its window closes before the benefit arrives. Extend the time horizon, model the life-course accumulation, and be explicit that discounting distant benefits is a value choice with equity consequences, not a neutral technicality (see Chapter 2.1 — Economic Evaluation).

5. Separate causation from correlation before you commit money. A determinant that merely correlates with ill health is a false target; spend on it and health will not move. Insist that a proposed upstream intervention rests on a credible causal story — ideally supported by natural experiments, difference-in-differences, or other quasi-experimental evidence (Chapter 2.3 — Health Econometrics) — and be candid about residual confounding rather than presenting an association as if it settled the case.

6. Prefer the population strategy where the risk is widely distributed, but count its costs. Following Rose, shifting the whole distribution of a risk factor can prevent more cases than chasing high-risk individuals, because most cases arise from the many people at modest risk, not the few at high risk. But the prevention paradox means the benefit to each person is small and often unfelt, so build the political and evaluative case for population measures deliberately, and pair them with targeted action where a concentrated high-risk group also exists.

7. Put health into non-health decisions — adopt a "health in all policies" stance. Because health is produced across government and the economy, embed health impact into the decisions of housing, transport, planning, education, and fiscal policy, rather than leaving health to the health ministry alone. This is the Health in All Policies approach; its economic rationale is that the cheapest place to prevent much ill health is in a sector that does not currently price it in.

8. Use health impact and equity-impact assessment as a standing discipline. For any major policy or investment — a road, a benefit change, a plant closure — assess the likely health and health-inequality consequences before the decision, not after the harm. A structured assessment forces the upstream causal chain into view, gives non-health decision-makers a shared language, and creates an audit trail for accountability.

9. Measure the right unit: population health and its distribution. If you track only the average, you can improve the mean while the gradient steepens — a common and perverse result of interventions that the already-advantaged adopt fastest. Report outcomes stratified by social position from the outset, so that a "success" that widens inequality is visible as the mixed result it is (the measurement of that distribution is owned by Chapter 3.4 — Equity).

10. Keep behaviour in its place — a mechanism, not a verdict. Individual behaviours (smoking, diet, activity) are real determinants, but they are themselves patterned by social and commercial conditions, so treating behaviour as free-floating personal choice both misreads the cause and tends to widen inequality by helping those best placed to change. Locate behaviour within its social context, and route the design of behavioural interventions through Chapter 4.1 — Behavioural Economics rather than defaulting to exhortation.

Questions to discuss with your team

1. If most of our population's health is produced outside healthcare, what share of our attention and budget honestly reflects that — and what would it take to shift some upstream? This question confronts the gap between what the evidence says produces health and where a health organization actually spends its energy, which is overwhelmingly on treatment. The honest discussion starts by admitting the structural reasons for the imbalance: budgets, mandates, and accountability are built around healthcare, and upstream returns are slow and accrue to other sectors. The team should test a concrete case — a housing, early-years, or employment intervention — and ask whether it could plausibly buy more health per unit of spend than a marginal clinical service, and if so, what stops the organization acting on that. A good answer resists both complacency ("not our job") and fantasy ("we'll redirect the hospital budget to social policy"); it identifies the realistic levers an organization actually holds — its own commissioning, its convening power, its influence on partner sectors — and names what it would stop doing to free the effort.

2. Where we are about to invest in an upstream determinant, how confident are we that we have a cause rather than a correlation — and what would change our mind? Determinant evidence is mostly observational, so the risk of spending on a factor that merely travels with ill health, without causing it, is real and expensive. The team should interrogate the causal story behind a specific proposed intervention: is there quasi-experimental or natural-experiment evidence, a plausible mechanism, and a dose–response pattern, or only a cross-sectional association that could be driven by a common cause such as poverty itself? It helps to name the most likely confounder explicitly and ask how it has been ruled out. An honest answer will rarely reach certainty — determinants cannot be randomized — so the goal is calibrated confidence and a pre-agreed way to test the intervention as it rolls out, rather than either paralysis or false certainty. The team should also be candid that the causal bar for upstream action is often held higher than for a new drug, and ask whether that asymmetry is justified.

3. When we act on the determinants of health, are we closing the gradient or unintentionally widening it — and how would we know? Interventions that rely on people to opt in, understand information, or afford a change tend to be taken up first and fastest by the already-advantaged, so a programme can improve the average while the gap between top and bottom grows. The team should ask, for a real initiative, who is most likely to benefit and whether the design reaches the whole social gradient or only its more resourced part. This requires committing upfront to measure outcomes stratified by social position, not just in aggregate, and being willing to call a gradient-widening "success" a partial failure. An honest answer accepts a genuine tension: the most cost-effective option on average may be the least equitable, and the team must decide, transparently, how it trades population gain against fair distribution — a trade-off whose measurement is owned by Chapter 3.4 — Equity.

4. When a determinant is a product being sold to our population, are we willing to name the commercial actor — and act on it — or do we retreat into the language of "lifestyle"? Much of the socially-patterned burden a health organization carries is manufactured and marketed: tobacco, alcohol, ultra-processed food, and gambling produce ill health at a profit that is privatized while the costs fall on the public purse. The team should test whether its own framing quietly blames the consumer — "poor choices", "unhealthy lifestyles" — and thereby lets the producer disappear from the analysis. The harder discussion is what the organization can actually do about a commercial determinant it does not regulate: it may not hold the fiscal or licensing levers, but it can name the actor, refuse to co-brand with it, decline its sponsorship, and add its voice to the regulatory case (the levers themselves are owned by Chapter 3.2 — Health Policy). An honest answer treats these products with the externality logic of Chapter 1.3 — Market Failure — the price omits the health cost — rather than as a matter of individual willpower, and it is candid about the political friction of confronting an industry that funds jobs, sport, or the organization's own partners.

5. Which determinant levers do we actually control, which belong to other sectors, and how honestly are we counting the ones we merely hope someone else will pull? The most powerful determinants — income, housing, employment, the built environment — sit in departments a health organization does not run, so a plan that depends on them is a plan that depends on other people's decisions and budgets. The team should sort a proposed strategy into what it directly controls (its own commissioning, estate, hiring, and convening power), what it can influence through partnership, and what it can only advocate for, and be realistic about how much health each tier can buy. This is the practical test of a "health in all policies" stance: it is easy to endorse in principle and hard to resource when the benefit accrues to another sector's ledger while the cost falls on yours. An honest answer resists both extremes — the fantasy that the health budget can fix housing, and the fatalism that nothing outside the clinic is our business — and it names the shared-accountability mechanism (a joint plan, a pooled budget, a health-impact assessment) that would make cross-sector action more than a wish.

6. Are our evaluation horizons long enough to see the returns from acting on determinants, or is our appraisal method quietly biased toward treatment? Determinant interventions — early-years support, housing retrofit, employment schemes — often return health decades later, so a standard short-horizon return-on-investment tool will score them as poor simply because its window closes before the benefit arrives, systematically favouring the fast, visible payback of treatment. The team should examine the time horizon and discount rate baked into its own appraisals and ask whether they are defensible or merely conventional, because those two choices can decide the result before any evidence is weighed. It helps to model the life-course accumulation explicitly and to show the answer under more than one horizon, so the sensitivity is visible rather than hidden. An honest answer admits that discounting distant health benefits is a value choice with equity consequences — it downgrades the health of future and younger people — not a neutral technicality, and it distinguishes a genuinely weak intervention from one that is only weak inside too short a window (see Chapter 2.1 — Economic Evaluation).

In practice: a health economics example

Scenario: a fictional post-industrial middle-income country in Central Europe, here called the Republic of Talvara, where a regional health authority confronts a stark health divide in a former coal-and-steel region.

The Kolnik Valley Regional Health Authority in Talvara oversees a deindustrialized region where the old mines and steelworks closed a generation ago. Male life expectancy in the valley's poorest districts trails the national capital by close to a decade, and the regional hospital is overwhelmed by early cardiovascular disease, chronic lung conditions, and the health consequences of long-term unemployment and heavy drinking. The regional director's instinct, and the political pressure, is to fund a new cardiology unit. Before committing, she asks the authority's health economist to assess whether that is the highest-yield use of a fixed regeneration grant that could, unusually, be spent across sectors.

The economist begins by drawing the region's health production function rather than its hospital demand. The excess mortality, she shows, is not primarily a shortfall of cardiology capacity; it is the downstream signature of upstream conditions — mass long-term unemployment since the plant closures, damp and poorly-heated post-war housing, a food environment dominated by cheap ultra-processed products, aggressive alcohol retailing, and air quality still degraded by domestic coal burning. These map onto the social, commercial, and environmental determinants directly. She stratifies outcomes by district and confirms a steep social gradient within the valley itself, not merely a poor-versus-rich contrast, which tells her that a programme aimed only at the most deprived estate would miss most of the burden.

She is disciplined about causation. The association between unemployment and ill health could run either way, or be driven by a common cause, so she leans on quasi-experimental evidence from comparable industrial closures elsewhere and on the plausibility of mechanisms — income loss, psychosocial stress, the collapse of routine — rather than presenting the raw correlation as proof (the methods are those of Chapter 2.3 — Health Econometrics). For the housing and air-quality strands, where natural-experiment evidence on retrofitting and clean-air zones is stronger, her causal confidence is higher, and she says so, grading her recommendations by the strength of the evidence behind each.

She then models three broad uses of the grant against a long horizon. The cardiology unit would treat cases well but leave the manufacture of new cases untouched, and its benefits, though real, would flow disproportionately to those who present and comply — risking a widening of the gradient. A housing-retrofit-and-clean-air package would act on environmental determinants with reasonable causal support and a plausible medium-term return, but its health benefit arrives slowly and would be undervalued by a five-year return-on-investment tool, so she extends the horizon and flags the discounting choice explicitly. A "health in all policies" strand — coordinating with the employment, planning, and licensing authorities on job schemes, alcohol-outlet density, and a shift away from domestic coal — is the most upstream and potentially the highest-yield, but rests on the softest evidence and on levers the health authority does not directly control.

Her recommendation is not to abandon the cardiology need but to split the grant: a smaller clinical investment to meet unavoidable current demand, a housing-and-air package where the causal case is strongest, and a modest, well-evaluated employment-and-alcohol strand pursued jointly with other sectors, with all outcomes tracked stratified by district so that any widening of the gradient is caught early. She is explicit about the discomfort: the most politically saleable option (the unit) is not the highest-yield for population health, the highest-yield option (upstream regeneration) is the hardest to attribute and the slowest to pay off, and the authority is being asked to spend health money to move levers that other sectors own. The value of the analysis is not a single clean number but a defensible, transparent allocation that treats the valley's ill health as something being produced, and therefore preventable, rather than merely something to be treated.

Four sector lenses

Startup

A young digital-health or public-health venture is well placed to attack a single determinant with a sharp, measurable proposition — a service that improves housing warmth, a platform that connects unemployed people to work and support, a tool that reshapes a local food environment. Its advantage is focus and the ability to build measurement in from day one; its danger is claiming a health return it cannot yet substantiate, because determinant effects are slow, confounded, and hard to attribute to one product. A credible start-up in this space is unusually honest about its causal evidence and its lag structure, chooses a proximate, measurable intermediate outcome, and resists the temptation to market a correlation as a cure. It should also recognize that acting on a determinant often means influencing a system it does not control, which is a harder sell to investors than a self-contained clinical product.

Small business

An established small provider — a general-practice partnership, a single community pharmacy, a care home, or a small supplier of home-adaptation or food services — meets the determinants of health daily in its own caseload, and unlike a start-up it is not chasing growth but sustaining a steady, self-funded operation embedded in one place. Its distinctive asset is local knowledge and trust: it knows which streets, employers, and housing stock produce its sickest patients, and it can act on the determinants at a human scale — a referral into a benefits-advice or warm-homes service, a link with a local employer, a food-bank partnership — without the overhead of a system-wide programme. Its constraint is thin margins and no slack: acting upstream costs staff time it is not paid for, because the payment model rewards the clinical contact, not the housing letter, and the health return accrues years later and often to another budget. A realistic small provider therefore picks a small number of high-yield, low-cost links to local social infrastructure, documents the need it sees to strengthen the case for others to fund the upstream work, and resists being asked to shoulder structural problems that only a payer or government can move.

Enterprise

A large provider, insurer, or integrated system increasingly sees the determinants of health on its own balance sheet: socially-driven illness arrives as expensive, late-stage demand, and in capitated or population-budget models the organization carries that cost. This gives an enterprise a genuine economic incentive to invest upstream — in housing, social prescribing, or employment support — but also a temptation to skim the determinants that are cheap to address and ignore the structural ones. At scale, an enterprise can pool data to build a real population-health production function and can act as an anchor institution, using its own purchasing and employment to improve local conditions. Its accountability challenge is that upstream returns are slow and cross-sector, so boards under annual pressure must be shown the long-horizon and equity case, not just the year-one figure.

Government

A ministry or national authority is the only actor that owns the full set of levers — fiscal, regulatory, and structural — that move the social, commercial, and environmental determinants, and it is where "health in all policies" either lives or dies. Its distinctive power is to price externalities (through taxation of harmful products), to regulate the commercial determinants, and to shape income, housing, and education directly; its distinctive difficulty is that health sits in one department while most determinants sit in others, so coordination and shared accountability are the binding constraints. Government also carries the equity duty most explicitly: because determinants are socially patterned, government action is often the only thing that can bend the whole gradient rather than the tail. Its horizon should be the longest of the three sectors, matching the life-course lags, and its hardest task is sustaining upstream investment across electoral cycles that reward visible treatment over invisible prevention.

Common failure modes

• Equating health with healthcare. Treating the health budget as the health-production budget and pouring marginal money into treatment while the upstream causes go unaddressed. Fix: map the full production function of health and ask where the next unit of spend yields most, including non-health budgets.

• Targeting only the poorest and missing the gradient. Designing for the bottom of the distribution when most of the socially-patterned burden sits across the whole range. Fix: act with proportionate universalism — universal reach, intensity scaled to need — and design for the gradient.

• Reading a correlation as a cause. Investing in a determinant that travels with ill health but does not cause it, and seeing no health movement. Fix: demand a credible causal story and quasi-experimental evidence; name the likely confounder and how it was ruled out (Chapter 2.3 — Health Econometrics).

• Blaming behaviour and ignoring its causes. Framing socially- and commercially-driven ill health as free personal choice, which both misdiagnoses the cause and widens inequality. Fix: locate behaviour within its social and commercial context and name the commercial actor when a determinant is a sold product.

• Undervaluing the long game. Killing an early-years or housing intervention with a short-horizon return-on-investment tool whose window closes before the benefit arrives. Fix: extend the horizon to the life-course, model accumulation, and treat discounting as an explicit value choice.

• Improving the average while widening the gap. Claiming success on a mean outcome that the advantaged captured first, leaving the gradient steeper. Fix: measure outcomes stratified by social position from the start and count gradient-widening as a partial failure.

Maturity model

Checklist

• Map the full production function of your population's health — social, commercial, environmental, behavioural, and medical — not just healthcare capacity.
• Report your key outcomes stratified by social position, and check whether the gradient is narrowing or widening.
• For any upstream investment, write down the causal story and the strongest confounder, and grade the evidence behind it.
• Choose evaluation horizons long enough to capture life-course lags, and state the discounting choice explicitly.
• Decide, and state, whether an intervention is targeted or proportionately universal, and why.
• Name the commercial actor whenever a determinant is a marketed product, and analyse it as an externality.
• Identify which determinant levers your organization actually controls and which require another sector to act.
• Apply a health and equity impact assessment to major non-health decisions before they are taken.
• Route behavioural interventions through behavioural design (Chapter 4.1) rather than defaulting to exhortation.
• Cross-reference the distributional judgement to Chapter 3.4 — Equity, keeping causes and fairness distinct but linked.

Key sources

• WHO Commission on Social Determinants of Health, Closing the Gap in a Generation (2008) — the foundational global statement of the social determinants and the health gradient.
• Marmot et al., Fair Society, Healthy Lives (The Marmot Review, 2010) and its follow-ups — proportionate universalism and the gradient, developed for policy.
• Geoffrey Rose, The Strategy of Preventive Medicine — the population strategy, the prevention paradox, and shifting the risk distribution.
• Dahlgren and Whitehead's "rainbow" model of the layered determinants of health — a standard framing of the causal layers from individual to structural.
• WHO and The Lancet series on the commercial determinants of health — the analysis of industry as a driver of population health.
• Economics Network, Health Economics for Teachers — situating determinants within the wider discipline (see Chapter 1.1 — Introduction to Health Economics).

References

1. Determinants of health — Wikipedia — https://en.wikipedia.org/wiki/Determinants_of_health
2. Social determinants of health — Wikipedia — https://en.wikipedia.org/wiki/Social_determinants_of_health
3. Whitehall Study — Wikipedia — https://en.wikipedia.org/wiki/Whitehall_Study
4. Michael Marmot — Wikipedia — https://en.wikipedia.org/wiki/Michael_Marmot
5. Fundamental cause theory — Wikipedia — https://en.wikipedia.org/wiki/Fundamental_cause_theory
6. Commercial determinants of health — Wikipedia — https://en.wikipedia.org/wiki/Commercial_determinants_of_health
7. Life course approach — Wikipedia — https://en.wikipedia.org/wiki/Life_course_approach
8. Public health — Wikipedia — https://en.wikipedia.org/wiki/Public_health
9. Population health — Wikipedia — https://en.wikipedia.org/wiki/Population_health
10. Prevention paradox — Wikipedia — https://en.wikipedia.org/wiki/Prevention_paradox
11. Confounding — Wikipedia — https://en.wikipedia.org/wiki/Confounding
12. Health in All Policies — Wikipedia — https://en.wikipedia.org/wiki/Health_in_All_Policies
13. WHO Commission on Social Determinants of Health, Closing the Gap in a Generation — World Health Organization — https://www.who.int/publications/i/item/WHO-IER-CSDH-08.1
14. Economics Network — Health Economics for Teachers — https://economicsnetwork.ac.uk/health/teachers