Health financing and economic evaluation: cost-effectiveness and DALYs

In short

Health financing is how a country raises money for health, pools it to spread risk, and spends it to buy services; economic evaluation is how it decides which services are worth buying when the pool is small. The two meet in a single question: what health gain does one more rupee of public money buy, and from whom is that rupee taken? India matters here because about two-fifths of its health spending still comes out of patients' pockets at the point of care, making financial protection — not just disease control — a core public health goal. The topic is examined in MPH health economics and health systems papers, in UPSC/state public health service and NHM programme-officer selections, and in technical rounds at HTAIn, NHSRC, state health societies, the World Bank/WHO/UNICEF, PHFI, IHME-linked groups, ICMR institutes and health-insurance and HEOR teams of pharmaceutical companies and CROs.

Explained

Health financing: functions, sources and mechanisms

• Three functions (WHO framework): revenue collection (where money comes from), pooling (combining prepaid funds so the healthy subsidise the sick and the rich the poor), and purchasing (paying providers for services).
• Sources: general taxation (direct and indirect), social health insurance contributions (payroll), voluntary private insurance premiums, out-of-pocket (OOP) payments, external/donor funds, and employer-provided care.
• Health system archetypes: Beveridge model — tax-financed, publicly delivered (UK NHS); Bismarck model — mandatory social insurance funds (Germany); national health insurance (single public purchaser buying from private providers, e.g. Thailand's UCS); OOP-dominant mixed systems — which is what much of India's outpatient care still is.
• Prepayment and pooling are what reduce risk. OOP is the most regressive and least protective mechanism: it is paid when the household is least able to pay.
• Demand-side financing gives purchasing power to the user (insurance, vouchers, conditional cash transfers such as Janani Suraksha Yojana); supply-side financing funds facilities directly (budgets, salaries, grants under the National Health Mission).
• Fiscal space for health can be created by economic growth, re-prioritising health in the budget, earmarked taxes (tobacco and alcohol "sin taxes"), better efficiency, and external funds.

Health financing in India: the current picture

• National Health Accounts (NHA), produced by NHSRC for the MoHFW using the WHO SHA 2011 framework, is the official source. The latest estimates (2021-22) put total health expenditure at roughly 3.8% of GDP, government health expenditure at about 1.84% of GDP and 48% of total health spending, and OOP at about 39% of total health expenditure — down from 69.4% in 2004-05. Figures are revised with each NHA round, so quote the year with the number.
• National Health Policy 2017 targets: public health expenditure 2.5% of GDP (target year 2025, not yet met), states to spend more than 8% of their own budgets on health, and two-thirds of resources to primary care.
• Centre-state split: health is a State subject; states finance most service delivery, the Centre adds centrally sponsored schemes (National Health Mission, with state co-contribution normally 40%, and 10% for north-eastern and Himalayan states), plus PM-ABHIM for health infrastructure and 15th Finance Commission health grants routed through local bodies.
• Publicly funded insurance: Ayushman Bharat PM-JAY — ₹5 lakh per family per year for secondary and tertiary inpatient care, for the poorest ~40% of households identified from SECC deprivation criteria, with a top-up cover for all citizens aged 70 and above (Ayushman Vay Vandana, from 2024); paid by case-based package rates in the Health Benefit Package (HBP 2022); implemented by the National Health Authority with State Health Agencies. CGHS for central government employees, ESIC for organised-sector workers below the wage ceiling, state schemes (Aarogyasri in Telangana/Andhra, MJPJAY in Maharashtra, CMCHIS in Tamil Nadu, Karunya in Kerala) and employer and retail private insurance.
• Known gap: PM-JAY covers hospitalisation, while most Indian OOP is spent on outpatient care and medicines — so insurance alone does not remove impoverishing expenditure. Free essential medicines and diagnostics programmes and Jan Aushadhi generic outlets address that side.
• Ayushman Arogya Mandir (the renamed Health and Wellness Centres) deliver the expanded primary care package — the main supply-side arm of the UHC push.

Provider payment mechanisms and strategic purchasing

• Strategic purchasing = deciding actively *what* to buy (benefit package), *from whom* (provider empanelment and quality), and *how to pay* — rather than passively reimbursing bills.
• Key market failures justifying public action: information asymmetry, supplier-induced demand, moral hazard (insured patients and providers using more care; controlled by co-payments, deductibles, gatekeeping, prior authorisation), adverse selection (the sick buy insurance; countered by mandatory or automatic enrolment and community rating), externalities of infectious disease and vaccination, and merit/public goods such as surveillance.

Economic evaluation: types and the common framework

• Definition: comparative analysis of alternative courses of action in terms of both their costs and their consequences. "Comparative" and "both" are what distinguish it from a simple costing study.
• Types:
  1. Cost-minimisation analysis (CMA) — outcomes proven equivalent; choose the cheaper option.
  2. Cost-effectiveness analysis (CEA) — outcome in natural units (deaths averted, cases detected, mmHg reduced); result as cost per unit of effect.
  3. Cost-utility analysis (CUA) — outcome in QALYs gained or DALYs averted, so quantity and quality of life combine and unlike programmes become comparable.
  4. Cost-benefit analysis (CBA) — both costs and benefits in money (willingness to pay, human capital, value of a statistical life); gives net benefit and can compare health with non-health investment.
  5. Cost-consequence analysis — costs and a disaggregated list of outcomes, left for the decision-maker to weigh.
• Partial evaluations (not full economic evaluations): burden-of-disease studies, cost-of-illness studies, simple costing, outcome description.
• Design elements that must be stated (the reference case): perspective, comparator, time horizon, discount rate, costing method, outcome measure, model structure, sensitivity analysis, equity considerations.
• Perspective: health system / payer (direct medical costs only) versus societal (adds patient OOP, transport, caregiver and productivity costs). India's HTAIn reference case asks for the societal perspective as the base case (because OOP, travel and productivity losses are large in India), with a health-system/payer analysis presented alongside.
• Costs: direct medical (drugs, staff, consumables, capital annuitised), direct non-medical (travel, food, lodging), indirect/productivity (work-time lost; human-capital or friction-cost method), and intangible (pain, usually captured in utilities rather than costed).
• Micro-costing (bottom-up, item by item) versus gross/top-down costing (facility expenditure ÷ output). Joint costs are allocated by step-down allocation. Indian unit costs are available from the CHSI (Costing of Health Services in India) database developed by PGIMER Chandigarh with the DHR.
• Discounting: future costs and health effects are converted to present value; 3% per year for both costs and outcomes is the Indian and WHO-CHOICE convention, with 0% and 5% in sensitivity analysis.
• Modelling: decision trees for short, one-off events; Markov cohort models (health states, transition probabilities, cycle length, half-cycle correction) for chronic disease; microsimulation and dynamic transmission models where infection risk depends on who else is treated or vaccinated — a static model understates vaccine value because it ignores herd protection.
• Uncertainty: one-way and multi-way deterministic sensitivity analysis (tornado diagram), scenario analysis, and probabilistic sensitivity analysis (Monte Carlo over input distributions) displayed on the cost-effectiveness plane and summarised as a cost-effectiveness acceptability curve (CEAC). Value of information analysis asks whether more research is worth funding.
• Budget impact analysis answers affordability (total cost to the payer over 1-5 years) — a separate question from cost-effectiveness. Something can be highly cost-effective and still unaffordable.

DALYs, QALYs and the burden-of-disease logic

• DALY (disability-adjusted life year) — one DALY is one lost year of healthy life. DALY = YLL + YLD. It measures burden, so fewer is better; interventions are judged by DALYs averted.
• YLL (years of life lost) = deaths × remaining life expectancy at the age of death, taken from a standard reference life table (GBD 2019/2021 uses a normative table with life expectancy at birth of about 88.8 years; earlier rounds used about 86 years) so that a death at a given age counts the same everywhere.
• YLD (years lived with disability) = cases × disability weight × duration. Incidence-based YLD (incidence × DW × average duration) is the GBD classic; prevalence-based YLD (prevalence × DW) is used in GBD from 2010 onwards and is better for chronic conditions.
• Disability weights run from 0 (full health) to 1 (equivalent to death) and come from large population surveys using paired comparisons — they are not severity judgements by clinicians, and they attach to the health state, not to the person.
• Dropped features: GBD 1990 used 3% discounting and non-uniform age weighting (valuing young-adult years more). Both were removed from GBD 2010 onwards; DALYs are now undiscounted and un-age-weighted in burden reporting, though economic evaluations still discount future DALYs averted.
• Comorbidity adjustment prevents the weights of co-occurring conditions summing above 1 (multiplicative combination).
• QALY (quality-adjusted life year) = Σ (years in a state × utility of that state), utility 1 = full health, 0 = dead, negative values possible for states worse than death. More is better.
• Utility measurement: indirect via generic preference-based instruments — EQ-5D-5L (with an Indian value set now available), SF-6D, HUI — or direct via time trade-off (TTO), standard gamble (SG) (the only one grounded in expected-utility theory) and visual analogue scale (not a choice-based method, so strictly not a utility).
• DALY versus QALY: DALYs are health-gap, population and burden oriented, built on epidemiological data and expert/survey weights, and dominate global health and WHO/GBD work; QALYs are health-expectancy, patient-preference oriented and dominate HTA in the UK (NICE), and now feature in Indian HTA alongside DALYs.

Decision rules and thresholds

• Compute the ICER between each option and the next-best non-dominated option after ranking options by cost.
• Strong (absolute) dominance: an option that costs more and does less is discarded. Extended dominance: an option whose ICER is higher than that of a more effective option is discarded; what remains is the efficiency frontier.
• Four quadrants of the cost-effectiveness plane: more effective & cheaper (adopt — "dominant"), more effective & costlier (judge against the threshold), less effective & cheaper (judge — disinvestment), less effective & costlier (reject).
• Threshold (λ, willingness to pay per DALY averted or QALY gained): the old WHO-CHOICE rule of thumb — under 1× GDP per capita "highly cost-effective", 1-3× "cost-effective" — has been withdrawn as a decision rule because it is not grounded in the actual health budget. Indian empirical work suggests an opportunity-cost threshold well below one GDP per capita (commonly estimated in the range of roughly 0.3-1× per-capita GDP); HTAIn does not publish a single fixed Indian threshold, so results are usually reported against a range.
• League tables rank interventions by cost per DALY averted to allocate a fixed budget, but are valid only if perspective, discounting, year of costs and population are comparable.
• Equity: CEA is indifferent to *who* gains. Extended cost-effectiveness analysis (ECEA) adds financial-risk protection and distributional consequences by income quintile; distributional CEA (DCEA) and equity weighting make the trade-off explicit. This matters directly for PM-JAY-type schemes whose purpose is partly financial protection.
• Institutional home in India: HTAIn (Health Technology Assessment in India) under the Department of Health Research, MoHFW, with a Medical Technology Assessment Board, a technical appraisal committee and regional resource hubs; it published an Indian reference case / HTA manual and feeds decisions on package rates, device procurement and programme choices.

Key points

• Health financing functions - revenue collection, risk pooling, purchasing; only prepayment and pooling give financial protection.
• OOP expenditure - payment at the point of care; the most regressive source, about 39% of India's total health expenditure in NHA 2021-22 (69.4% in 2004-05).
• NHP 2017 targets - public health spending 2.5% of GDP, states above 8% of their own budgets, two-thirds of funds to primary care.
• PM-JAY - ₹5 lakh per family per year, inpatient secondary/tertiary care, case-based package rates under HBP 2022, 70+ covered separately.
• Strategic purchasing - choosing what to buy, from whom and how to pay; opposite of passive reimbursement.
• Moral hazard vs adverse selection - over-use once insured vs the sick self-selecting into insurance; co-payments and gatekeeping address the first, mandatory enrolment the second.
• Full economic evaluation - compares at least two options on both cost and consequence; cost-of-illness studies are only partial.
• CEA / CUA / CBA - natural units / QALYs or DALYs / money for outcomes.
• ICER - extra cost per extra unit of health; always calculated incrementally against the next-best non-dominated option.
• DALY = YLL + YLD - a burden measure; interventions are valued by DALYs averted.
• Disability weight - 0 to 1, survey-derived, attached to a health state, not to a person.
• GBD post-2010 - no age weighting and no discounting in burden estimates; economic evaluations still discount at 3%.
• QALY - years × utility; utilities from EQ-5D-5L (Indian value set), TTO or standard gamble.
• Discounting at 3% - the Indian/WHO convention for costs and outcomes, with 0% and 5% tested in sensitivity analysis.
• Dominance and extended dominance - remove options that are costlier and less effective, or have a higher ICER than a more effective option.
• Threshold λ - the old 1-3× GDP per capita WHO rule is withdrawn; Indian estimates sit well below one GDP per capita and are reported as a range.
• Budget impact analysis - affordability to the payer; separate from and additional to cost-effectiveness.
• Catastrophic health expenditure - SDG 3.8.2 uses OOP above 10% (and 25%) of household consumption; the capacity-to-pay definition uses above 40% of non-food consumption.

Formulas and examples

Example 1 — DALYs for a district road-injury burden (undiscounted, no age weighting). District population 2,00,000. In one year: 200 deaths at mean age 25, standard remaining life expectancy 61.6 years; 1,500 survivors with moderate long-term disability, DW 0.27, average duration 2 years.

• YLL = 200 × 61.6 = 12,320 years.
• YLD = 1,500 × 0.27 × 2 = 810 years.
• DALYs = 12,320 + 810 = 13,130.
• Burden rate = 13,130 / 2,00,000 × 1,000 = 65.7 DALYs per 1,000 population per year. YLL is 93.8% of the total — a young-death-dominated burden, typical of injuries.

Example 2 — ICER and dominance across three screening strategies.

• C is removed by strong dominance. If λ = ₹2,20,000 per DALY averted (of the order of one per-capita GDP, about ₹2.2 lakh in 2024-25), both A and B fall below λ and B is chosen.
• NMB of B over A = (400 × 2,20,000) − 1,80,00,000 = 8,80,00,000 − 1,80,00,000 = +₹7.0 crore.
• Note how the average CER misleads: by average CER, A looks best; the correct rule is incremental.

Example 3 — Discounting a three-year programme cost at 3%. Costs ₹10,00,000 at the start of each of years 0, 1 and 2.

• Year 0: 10,00,000 / 1.03⁰ = 10,00,000
• Year 1: 10,00,000 / 1.03 = 9,70,874
• Year 2: 10,00,000 / 1.0609 = 9,42,596
• PV = ₹29,13,470 against an undiscounted ₹30,00,000 — discounting has reduced the figure by 2.9%. If 1,000 DALYs are averted in each of those years, discounted DALYs = 2,913.47, so the ICER (₹1,000 per DALY) is unchanged here because the time profiles match; it changes whenever costs fall early and benefits late — the usual case for prevention and vaccination.

Example 4 — Cost-utility with QALYs. New regimen: utility 0.75 for 10 years = 7.50 QALYs, lifetime cost ₹6,24,000. Standard care: utility 0.60 for 8 years = 4.80 QALYs, cost ₹3,00,000.

• ΔE = 7.50 − 4.80 = 2.70 QALYs; ΔC = 6,24,000 − 3,00,000 = ₹3,24,000.
• ICER = 3,24,000 / 2.70 = ₹1,20,000 per QALY gained.
• NHB at λ = ₹2,20,000: 2.70 − (3,24,000 / 2,20,000) = 2.70 − 1.47 = +1.23 QALYs — adopt.

Example 5 — Financial protection for one household. Monthly consumption ₹12,000, of which food ₹6,500; monthly OOP on a chronic illness ₹1,500.

• OOP share of total consumption = 1,500 / 12,000 = 12.5% → exceeds the 10% SDG 3.8.2 threshold, so catastrophic; below 25%.
• Capacity to pay = 12,000 − 6,500 = ₹5,500; OOP share = 1,500 / 5,500 = 27.3% → below the 40% threshold, so not catastrophic on that definition.
• The same household is counted catastrophic under one definition and not the other — always state which threshold and denominator were used.

Where it is used

• HTAIn and the Department of Health Research — appraisals that feed PM-JAY package rates, device and diagnostic procurement and programme design; HTA analyst and health economist posts.
• National Health Authority and State Health Agencies — package costing, fraud and claims analytics, provider payment design, empanelment economics.
• NHSRC, NHM state programme management units and State Health Resource Centres — National Health Accounts, programme implementation plans, district health action plans, unit-cost studies and public expenditure reviews.
• ICMR institutes, PGIMER, AIIMS, PHFI, IIPH campuses, NIE — burden-of-disease and DALY estimation, cost-effectiveness of TB, HIV, cervical-cancer HPV vaccination, dialysis and NCD screening programmes.
• Pharmaceutical and device HEOR / market access teams and CROs — cost-effectiveness and budget-impact dossiers, EQ-5D-based utility studies, pricing and reimbursement submissions.
• Insurers, TPAs and actuarial teams; World Bank, WHO, Gates Foundation, Global Fund, UNICEF and state finance departments — premium and benefit-package design, investment cases, fiscal space analysis.

Interview questions

• Q: What makes a study a full economic evaluation? A: It compares two or more alternatives and examines both their costs and their consequences; a cost-of-illness or burden study does neither comparatively.
• Q: Why is the ICER used instead of the average cost-effectiveness ratio? A: Decisions are about moving from one option to the next, so what matters is the extra cost for the extra health gain; average ratios hide that and can rank options wrongly.
• Q: How is a DALY computed? A: DALY = YLL + YLD, where YLL is deaths multiplied by standard remaining life expectancy at the age of death and YLD is cases multiplied by a disability weight and duration.
• Q: Difference between a DALY and a QALY? A: A DALY measures health lost (a gap from an ideal, fewer is better, epidemiology- and survey-weight based); a QALY measures health gained (years weighted by preference-based utilities, more is better).
• Q: Does GBD still use age weighting and discounting? A: No — both were dropped from GBD 2010 onwards; burden is reported undiscounted and un-age-weighted, although economic evaluations still discount future DALYs averted, usually at 3%.
• Q: Why discount health benefits at all? A: Because people and budgets have positive time preference and the opportunity cost of money compounds; not discounting benefits while discounting costs biases analyses towards programmes with very distant effects.
• Q: What threshold would you use for India? A: There is no single official Indian threshold; the old WHO 1-3× GDP per capita rule has been withdrawn, Indian opportunity-cost estimates sit well below one per-capita GDP, so results should be presented across a range with a CEAC.
• Q: PM-JAY pays by package rates — what behaviour does that encourage, and how is it policed? A: Case-based payment encourages shorter stays and cost control but also up-coding, unnecessary admissions and case-splitting; it is policed by pre-authorisation, claims analytics, audits and de-empanelment.
• Q: India's OOP share has fallen, yet households still report distress spending. Why? A: Publicly funded insurance covers hospitalisation, while most OOP is on outpatient care, medicines and diagnostics, and indirect costs such as travel and lost wages are not covered at all.
• Q: An intervention is highly cost-effective but the state says no. Is that irrational? A: Not necessarily — cost-effectiveness is about value for money, affordability is about the total budget impact, and feasibility, equity and absorptive capacity are separate criteria.
• Q: When must a dynamic transmission model be used instead of a static one? A: Whenever an intervention changes the force of infection — vaccination, treatment-as-prevention, vector control — because a static model misses herd protection and understates value.
• Q: What is extended cost-effectiveness analysis and why does India care about it? A: ECEA adds financial-risk protection and the distribution of health and money gains across income groups, which is central when a scheme's explicit aim is protecting poor households from catastrophic spending.
• Q: Which perspective gives the larger cost estimate and why? A: The societal perspective, because it adds patient out-of-pocket, transport, caregiver and productivity costs to health-system costs.

Common mistakes

• Reporting an average CER (or a cost per case treated) and calling it an ICER, and failing to remove dominated and extended-dominated options before computing ICERs.
• Saying "a DALY gained" — DALYs are a burden and are averted; QALYs are gained. Reversing the sign of the better option is a guaranteed mark lost.
• Adding disability weights for co-morbid conditions so the total exceeds 1, instead of combining them multiplicatively.
• Using age weighting or discounting in a GBD-style burden estimate, or quoting the 1990 formula as current practice.
• Using local period life expectancy instead of the standard reference life table for YLL, which makes a death in a poor state count for less than the same death elsewhere.
• Discounting costs but not health effects, or mixing cost years without inflating to a common base year and currency.
• Treating cost-effectiveness as proof of affordability, and skipping budget impact and sensitivity analysis altogether.
• Double counting productivity loss both in the cost side and inside the utility/disability weight, and shifting perspective midway through a study.
• Quoting a stale NHA percentage without its reference year, or stating India's public health spending has reached 2.5% of GDP — it has not.

Related topics

Health systems in India: primary, secondary and tertiary care · Health policy analysis and planning · Global health: WHO, the SDGs and the International Health Regulations · Epidemiology: measures of disease burden and risk · Biostatistics and decision modelling · Health management information systems and monitoring and evaluation