National Health Insurance
Costing and Long-term Modelling of NHI
Download full PDF Document
The purpose of this series of policy briefs on National Health Insurance (NHI) and the related IMSA web-site is to put in the public domain material and evidence that will progress the technical work of developing a National Health Insurance system in South Africa. This includes tools for costing NHI and evidence on where savings could be achieved in moving to a future mandatory system with universal coverage.
This policy brief deals with the basics of costing and pricing healthcare from an actuarial perspective. The brief highlights some of the issues and debates that need to be considered in the long-term modelling of diseases and costs including the quadruple burden of disease in South Africa.
The terms “cost” and price” are often used loosely and mean different things to different professions. It is important always to be clear on which term is meant. Actuaries like to differentiate between the terms “cost” and “price” as follows and as illustrated below:
-
The total cost is the total amount needed across all eligible beneficiaries to provide access to a defined package of benefits in a particular delivery setting.
-
The price is the amount incurred by contributors to the system and may be set according to other criteria, like equity and affordability. For example, the price may be expressed as a contribution table or as a percentage of income, subject to certain maximums.
Figure 1 : Generic Costing and Pricing of Healthcare
-
The price charged by providers to funders in the health system will be referred to as the reimbursement rate and it may be determined in a variety of ways, from fee-for-service to per visit or capitation (pre-paid, typically on a monthly basis).
The generic costing and pricing model above is applicable to medical scheme and bargaining council schemes, to provider capitation (or other reimbursement methodologies), to determining the Risk Equalisation Fund tables and to mandatory health insurance.
For National Health Insurance, the need is to determine a table of amounts or a formula relative to some definition of income, payable by defined contributors and incorporating income cross-subsidies and an equitable Government subsidy. The total amount needed would be determined from historic data, population projections, other inputs and assumptions. The total amount would be set to cover a defined package of benefits for the entire population (or a phased target population), with an expected dispensation of efficiency in healthcare delivery. There would be loadings for administration and managed care costs but probably no loading for solvency as money in would equal money out.
The elements of costing and pricing and some of the issues to be considered are summarized below. The sequence usually starts with a defined package of benefits, then works up from the bottom of the costing part of Figure 1 and then across to determine the price to be charged to contributors.
Benefit package: the package of minimum benefits needs to be defined at the outset as the detail will influence the choice of data and the design of the costing study. There may be iterations to determine alternative benefit packages when affordability is assessed in the final step. The form of rationing expected in the package has a large influence on the study design and needs to be articulated at the outset.
Raw price: a suitable source of data is identified and data extracted and tested for reasonability (this is often the longest phase of costing as further data extracts may be needed when anomalies are identified). As healthcare has a strong seasonal pattern, data must cover at least one complete calendar year and up to three years of data is ideal (any longer and trends are unlikely to be valid as benefits may have changed). If data is gathered from multiple sources, a lot of effort needs to be put into ensuring that data definitions are identical and that the results can be validly combined.
While it is true that total cost is a function of utilisation and the unit price (reimbursement rate to practitioners), it is often better to work with the total amount per beneficiary per month (pbpm). Utilisation has a strong pattern by age and gender but many researchers are not aware that average cost also has patterns by age and gender. The average cost patterns may be quite “lumpy” by comparison to utilisation and are thus difficult to use for modelling. An analogous situation occurs with admission rates and average length-of-stay (ALOS) in hospital data where ALOS is very lumpy and it is advisable to combine them and work with total bed days.
Healthcare financing is a rationing problem and rationing will occur somewhere in the system. We generally consider four parties that could do the rationing and each uses different tools, as shown in the examples below:
-
Government: by means of budget constraints; by long queues (at clinics or for getting certain elective surgery); by availability (limited ICU beds or surgical beds); and by denial (no dialysis after a certain age and no resuscitation of very low-birth-weight babies).
-
Health funders (like medical schemes or NHI): by means of limits, co-payments, deductibles and thresholds or by means of volume (like one pair of spectacles every two years).
-
Patients and their families: by means of affordability (choice to have private insurance, level of savings account or degree of out-of-pocket spending) or conscious choice (choice to refuse care in terminal illness).
-
Doctors: by prognosis typically but also by affordability in some cases (differential treatment or prescribing based on patient income).
In principle, we should try to move rationing to doctors firstly, together with their patients, rather than having accountants and bureaucrats make healthcare rationing decisions.
The box above shows the key pricing relationships. Exposure is usually calculated as “beneficiary months” which is the number of months that each beneficiary is exposed to making a potential visit or claim in the data. For example, if a person was in the risk pool for the whole year that would be 12 months of exposure but someone joining at the beginning of the last quarter of the year would have only 3 months of exposure.
Margins and adjustments: the raw data may need to be adjusted, depending on the quality and applicability of the data extract. One of the most common adjustments is to estimate the “incurred but not reported” claims (IBNR) if the data has been extracted very soon after the calendar year end. Medical scheme claims, for example, may be submitted up to four months after the event and there may be disputes about the amount payable. By delaying the extraction of data for five or six months after year end, the claims are usually fully “run-off” and no estimate needs to be made. Other adjustments may be due to sharp changes in the quality of coding or data submission during a period and sometimes a decision is taken to use only the patterns from the latter part of the time period. These adjustments require intimate knowledge of the data and considerable judgement is needed to ensure that the adjusted data is valid for the purpose of the study. Removing what some people call “outliers” (very large claims) is never advisable in healthcare data as the very nature of the data is some very expensive extreme events. Having a few of these in the data is in fact normal and the larger the data set the more predictable are these large cases.
The size of the data set extracted needs to take into account the minimum risk pool sizes in the box on the right. These are very much minimum sizes and the larger the risk pool the more stable the results will be. The most recent work on costing Prescribed Minimum Benefits in medical schemes was done with 49.8 million beneficiary months of data or effectively 4.2 million lives worth of data for the calendar year.
Demographic correction: it is rare to be able to obtain raw data for precisely the group that needs to be costed. It is much more likely that data from several sources will be used and then adapted for use on the expected target market. Major errors can be made if the demographic structure of the data is not taken into account and the number blindly applied to a population with a different demographic structure. As age and gender are the primary risk factors in cost, it makes sense to always do the costing work by at least age and gender. In some cases, using province is also useful as there are major differences in hospitalisation rates between the provinces. HIV strongly influences need for healthcare and Policy Brief 43 showed how this differs by province.
New benefit package correction: often the benefit package that is to be mandatory is being changed as part of the study. Taking data collected when a benefit is voluntary and converting it to the expected utilisation reimbursement rate and thus raw price when it is mandatory is a difficult exercise and requires experience and judgment. Typically, usage will be higher once a benefit is included in a minimum package but the extent of the change is difficult to forecast. It is simpler to take generous benefits and determine what the raw price should be if limits, deductibles or co-payments are applied, although all of these will still alter provider and beneficiary behaviour from that observed in historic data.
Inflation to period of use: the diagram overleaf describes the problem of using historic data to predict cost in the future. If data is extracted in Q3 2009, then the true inflation from mid-2008 to mid-2009 will already be known. An estimate of inflation will need to be made from mid-2009 to the middle of the period of use, mid-2010. Inflation is usually calculated and estimated separately for different components of the benefits, like hospitals, medicines and visits. It is crucial to isolate any changes in the demographics of the risk pool from the price effects in calculating the historic inflation.
Figure 2: Benefit, Contribution and Data Cycle in Medical Schemes
Adjustments: the price of any negotiated healthcare delivery contracts and contracts for administration and managed care need to be explicitly taken into account. These contracts need to have been finalized before the costing is completed in order to be certain as to what needs to be charged. There are some spectacular examples of insolvency where a fund promised certain contributions to members first, before concluding contracts with providers, only to find that no providers would contract at the rate used in the calculations. This is a particular danger for NHI in that contribution levels have been promised to be lower than for existing medical schemes yet no negotiations with providers have yet been entered into.
Other adjustments in the total cost may be for anticipated investment earnings and there may be loadings for a liquidity buffer or solvency margin. Healthcare expenditure has a very particular seasonal shape by month of the year. Summer months and months with many holidays (the April or May Easter holidays) have much lower claims while the winter months have higher claims. This means that a buffer of unspent funds will build up in the first quarter of the year, there may be over-spending in the middle of the year and the final quarter of the calendar year is typically light. Unsuspecting trustees or managers can over-react and enhance benefits based on Q1 results only to find themselves under financial pressure in the following two quarters.
The single most critical assumption to be made in NHI costing and pricing will be the level of efficiency that might be achieved. There is a perception that the private sector is inefficient and that somehow the public sector can use the same funds more efficiently. This central assumption is typically offered without evidence and it is unwise to anticipate any improvement in efficiency (resulting in a lowering of costs) unless there is hard evidence on the table.
Spread total cost as a price to be charged using allowable rating factors: medical schemes may use income but not age as a factor in setting prices to be charged. Separate adult and child rates are permitted in the voluntary environment to encourage families to enrol children. Under NHI, the group of people who will become contributors will need to be tightly defined, as will the definition of how contributions are to be calculated and the definition of income. It is always possible in setting the prices to be charged to socially-engineer the table to provide relief for vulnerable groups. In medical schemes, elderly pensioners or lower income workers are typically favoured in restricted schemes. It is not usually feasible to do so in open schemes because of the potential anti-selection if only some open schemes follow this route. This is where Government needs to play a role in regulating acceptable cross-subsidies that all open schemes must implement. Under NHI the definition of who is to contribute and who will be exempt from contributing (including definitions related to citizenship, employment, age and income) is critical before any price can be determined.
|
