Abstract
Uncertainty is the central element in insurance. This article examines how insurers evaluate and price risks that can present very high levels of uncertainty, and which many underwriters regard as especially hazardous in insurance terms. These are the risks associated with new medical devices, new pharmaceutical products and others substances for human consumption, such as food additives. Insurance is likely to be needed for these products both during their research and development phases, including insurance for clinical trials, and also once the device, drug or other substance gains approval and is in regular use.
The article examines the types of insurance that are available to cover these risks, the organizations that provide insurance and how the insurance is organised. It discusses the basic principles that insurers use to price insurance before considering the difficulties presented by novel and complex risks generally. The article concludes with a description of the techniques that insurers employ to analyse and price the particular risks that are our subject and a discussion of how underwriters seek to overcome the special problems associated with them.
Insurance is a contract upon speculation …. Lord Mansfield in
Introduction
Uncertainty is the central element in insurance. People and firms that insure exchange the uncertainty of suffering random and potentially devastating losses for relatively small and predictable payments in the form of insurance premiums. Insurers deal with the uncertainties of the risks they take on by spreading them across large pools of similar risks and also by spreading them over time.
This article concerns risks that present a very high level of uncertainty for insurers, namely those associated with new medical devices, pharmaceutical products and others substances for human consumption, such as food additives. Insurance may be needed both during the research and development phase, including clinical trials, and once the device, drug or other product gains approval and is in use.
This article is structured in four main parts. We begin by looking at the types of insurance that are available to cover these risks, who provides insurance and how it is organized. We then discuss the general principles that insurers use to price insurance before considering the particular problems presented by novel and complex risks generally. We conclude by looking at the techniques that insurers employ to assess, underwrite and price the particular risks that are our subject.
What sort of insurance is used, and who provides it?
Various ‘generic’ forms of insurance are available for the risks under discussion. They include product liability insurance, covering claims against manufacturers and suppliers in respect of defects or dangers in drugs, medical devices and the like and professional indemnity insurance (or errors and omissions (E&O)). The latter covers claims, typically alleging negligence, against professional firms and individuals who carry out research, design and development work or give advice or treatment. Medical malpractice insurance is a (more or less self-explanatory) variant for professionals in the medical sphere, covering either individuals personally or their employers, such as hospitals. If employees suffer harm through contact with toxic or otherwise hazardous products or substances, then employers’ liability or workers’ compensation insurance might be involved. Other possibilities include claims met by environmental liability insurers (e.g. following a release of toxins that contaminate land or water) or directors’ and officers’ (D&O) insurers, in cases where claimants target the directors of guilty firms personally rather than the firm itself. Such claimants might be the victims who suffered physical harm or, much more likely, investors who lost money when a scandal-hit medical or pharmaceutical company’s share price collapsed.
Often these and other lines of insurance cover are packaged, combined or tailored to meet the needs of a particular type of business or activity, such as insurance for sponsors, contract research organizations (CROs), medical universities and investigators conducting clinical trials. For example, regulatory rules require that volunteers who suffer harm in UK clinical trials should receive compensation on a ‘no-fault’ basis, so insurers have to adapt their policies to ensure that the insurance responds appropriately.
We should not assume that insurance is always provided by an ordinary insurance company or in the usual way. The insurance buying practices of large corporations, in particular, are very different. First, the size and capital strength of large firms means that they can meet most losses from their own resources should they wish to do so and need insurance only for catastrophes – very large individual losses or accumulations of losses from one source – or in cases where insurance is required by law. Second, most major corporations (including all the major pharmaceutical firms) own or make use of at least one ‘captive’ insurance company, established specifically to finance at least some of the risks generated by its ‘parent’ company or companies. There are many types of captive insurer, including the (self-explanatory) ‘single parent captive’ (usually a wholly owned subsidiary) and the ‘group captive’, which provide insurance cover for a number of firms, normally in the same industry. There are well over 5000 captive insurers worldwide, mostly located in tax havens. Some pharmaceutical firms rely on the size of their balance sheet to pay very substantial losses and carry no insurance at all for some risks (including sponsorship of clinical trials), perhaps having insurance only for extreme loss events or in cases where insurance is compulsory.
The design and tailoring of insurance products for firms that insure are often carried out not by the insurance companies that underwrite the risks themselves but by insurance brokers. These are often specialist firms, working, for example, in the field of life sciences or specialist divisions of major international insurance brokers.
Finally, many of the organizations that produce drugs, medical devices, food additives and the like or sponsor or conduct research (including clinical trials) are international in scope. They may have assets, activities, research operations and customers in many countries worldwide. Regulations on providing insurance, the extent to which it is compulsory and who may provide it, vary enormously from one country to another. This means that insurance programmes for multinational businesses can be very complex, often combining a ‘patchwork quilt’ of local insurance policies with an overlying ‘blanket’ provided by a master policy to ensure that cover worldwide is reasonably even and consistent.
Pricing insurance: General principles
Generally, insurance underwriters must decide, in negotiation with those they insure, how wide the insurance cover is going to be and also, of course, the price. The starting point for pricing insurance is to arrive at a
To set the risk premium, insurers need to estimate both the frequency and the severity of the losses that the insurance is likely to generate: that is, assess the probability of the insurance policy recording a claim in any given year and also estimate how big the claim is likely to be if it does arrive. The origin of this information is generally claims data, usually from the insurer’s own past experience, but supplemented in some cases by publicly available sources, such as general statistics on mortality in the case of life insurance. Sometimes insurers have huge amounts of data at a very fine level of detail. Motor insurance is a case in point, where major insurers typically have claims data from several million insured drivers. This can be analysed to establish the relative riskiness of different types of vehicle, drivers of different ages, use of the vehicle in different geographical areas and the impact of many other factors. Actuarial techniques are then employed to determine the weight that should be given to different factors and the effects of their being combined, enabling a premiums to be finely calibrated to the characteristics of each policyholder and the vehicle they drive.
Life insurance is similarly ‘probabilistic’. Continuous studies of human mortality enable insurers to accurately predict the risk of death at any given age. Similarly rich claims data enable them to make appropriate adjustments in premiums for ‘non-standard’ lives – people who are more likely to die prematurely as a consequence of their lifestyle, occupation or medical history.
Businesses that generate a relatively large and continuous stream of losses and claims can sometimes be priced, at least in part, on their own past loss experience. Typically, this is done by analysing the severity and frequency of losses generated by the firm over the last three to five years and by making projections about future loss patterns and, hence, the price for insurance. This technique (known as experience rating) can be applied fully only in the case of very large businesses. However, even for medium-sized firms, past claims experience will play at least some part in pricing the risk, resulting in a reduction in the normal rate of premium if the claims record suggests that it is a ‘good’ risk of its type or a premium loading in the opposite case.
Problems of insuring new and unusual risks
Many things can disrupt this, apparently straightforward, process of statistical observation and actuarial modelling. For example, there are potential problems of moral hazard. In the context of insurance, this can be described simply as the possibility that those who buy insurance will become careless as a consequence of being insured, making the risk worse. If fact, moral hazard has been shown to be particularly complex and problematic in the case of liability insurance, 1 this being the main form of insurance which concerns us. A further problem is adverse selection. This phenomenon, lucidly described in a celebrated article, 2 arises in insurance when insurers, usually as a result of mis-pricing, attract a higher, and statistically significant, proportion of poor risks than is found in the general population.
Whilst moral hazard and adverse selection are important, bigger problems for insurers can arise from the absence or thinness of suitable loss data on which to base premiums or, in cases where such data exist, the possibility that it may be unreliable as a predictor of future losses and claims. Loss data may be absent or thin when the risk in question is novel in some way or the number of people wishing to insure is too small to provide a solid statistical base. The relationship between the severity and frequency of losses can be an aggravating factor. Some lines of insurance (sometimes labelled ‘catastrophe classes’ by insurers) tend to generate infrequent but, typically, very large losses rather than a steady stream of relatively small claims. Results on an insurance portfolio consisting of risks of the former type will tend to be more volatile than the latter. Insurance is underpinned by the
Existing data, even, when they are apparently rich, can be unreliable for a number of reasons. For example, the effects of inflation, or changes in costs and prices generally, may mean that the size of past claims does not give an accurate guide to the cost of similar future claims. More seriously, shifts may be taking place in the underlying probabilities of loss, so that past claims do not provide a good guide to either the frequency or severity of future losses. A global pandemic, for example, could (at least in theory) bring about a major shift in human mortality, and the effects of climate change could radically alter the pattern of weather-related insurance claims (e.g. those arising from floods or windstorms). Significant shifts in the probability of loss can also be brought about by changes in the law, which can make claims for accident compensation or other liability suits easier to bring, or by advances in science, including medicine or technology. The latter can have a number of effects on the incidence and size of claims. For example, the discovery of a link between exposure to a toxic substance and a particular disease or illness may enable victims to identify and target a liable defendant (and their insurers) for the first time, and medical advances that preserve the lives of accident or disease victims can actually increase the cost of compensation claims they make because their treatment may now be both more expensive and prolonged.
A further key element that increases uncertainty is the timing of insurance claims in relation to their originating cause. In insurance parlance, claims, and the lines of business that generate them, can be either ‘short tail’ or (relatively) ‘long tail’. In the former case, the time span between the event that generates the insurance claim and its being made and eventually settled is relatively short. This is the case for most forms of property insurance and also for motor insurance, where little time passes between the negligent act that results in a vehicle accident, the accident itself and the insurance claim that results from it. Long tail lines of insurance – nearly all of which are forms of liability insurance – are those where there is an extended time gap between the originating cause of the insurance claim, its being made and the final settlement. Classic examples include insurance claims in respect of latent and/or gradually developing diseases and claims for gradual pollution and contamination. From the first category, and most notorious of all, are the many hundreds of thousands of insurance claims that have arisen from exposure to asbestos. The World Health Organization estimates that each year there are around 107,000 deaths worldwide resulting from exposure to asbestos. In the United Kingdom, there are still around 5000 deaths each year, even though the import and commercial use of asbestos ceased many years ago. It is well known that diseases resulting from asbestos exposure typically manifest themselves a very long time after the fibres are first inhaled. Injury claims can take a long time to settle, especially if they become the subject of a court dispute, with the result that the ultimate compensation payment to an asbestos victim, or their family, is often made by an insurer that underwrote and priced the risk anything up to 50 years ago.
The longer the potential delay in receiving and settling claims, the greater the uncertainty for the insurer. This uncertainty can arise not only from things already mentioned, such as potential changes in the law or scientific advances that make claims easier to substantiate or greater in amount, but also changes in social attitudes or in the availability of alternative compensation (such as social security benefits) either of which may increase people’s propensity to claim. Other imponderables include future fluctuations in rates of inflation and interest rates, each of which can affect the size of insurance claims and the value offsetting investment income that insurers expect to generate.
Insuring the unknown – How is it done?
It should be obvious from what has been said that insurance covering the risks with which we are mainly concerned – those arising from new drugs, new forms of treatment and medical devices, new food additives and the like – are likely to have many, if not all, of the characteristics that generate the uncertainties mentioned above. These include lack of historic loss data, potential for major shifts in the probability of loss and significant potential for long-tail claims, with all their attendant problems. In some cases, such as ‘first into man’ clinical studies, the uncertainty is likely to be particularly acute. A final, very significant feature is the marked potential, especially in the case of pharmaceuticals and food additives, for mass (tort) claims brought by very large groups of victims who suffer harm from the products, sometimes long after first using them. For example, use of the anti-inflammatory drug Vioxx (Rofecoxib), marketed by Merck on gaining Food and Drug Administration approval in 1999 and estimated to have been prescribed for about 80 million people, may have resulted in as many as 140,000 heart attacks or strokes. 3 This generated tens of thousands of individual claims and around 190 class actions, eventually settled by Merck at a total cost of around US$ 6 billion.
So how do insurers approach risks of this sort? There are many aspects to the science of insurance underwriting including, for example, the need to set coverage limits and retentions. We mean by ‘retention’, in this context, the amount of a given risk that an individual insurer will cover. Large risks are often divided, in various ways, among a number of insurers, with many of them writing only a fairly small ‘line’ on risks they regard as especially hazardous. Insurers also aim to build broad portfolios of insurance risks that are balanced as a whole. Insurers balance their portfolios by aiming for a spread of risks across, for example, different lines of insurance business and different geographical locations or territories. Among all these things, setting the scope of the cover (what is to be insured and what is to be excluded) and setting the price are especially important. We should remember, however, that insurance in developed countries is a highly competitive business, and buyers, whose own bargaining power is often supported by that of major international insurance brokers, have considerable influence in determining the scope and the price of insurance cover.
Limits and exclusions
Subject to the constraints mentioned above, insurers often seek to shape the cover in such a way as to exclude or limit the impact of certain risks or forms of loss, especially when they lack knowledge of them. For example, a number of insurers began to exclude claims arising from exposure to nanotubes and nanoparticles when their use started to become common, and most insurers have adopted a cautious approach since. Another common technique, employed by insurers in relation to what they perceive as heavy risks, is to impose relatively high deductibles and/or reduced limits for them – that is, require the insured to bear a significant proportion of each loss and lower the ceiling of coverage for individual claims or accumulations of claims from one event or in one period of insurance. For example, insurers are likely to impose a deductible in the region of US$5000 for what they perceive to be low-risk clinical trials but US$25,000 or more for high-risk trials. Another possibility, especially in relation to risks that may generate claims from territories where compensation levels are high and people are regarded as litigious, is the use of what are known as ‘jurisdiction clauses’ – these can deny coverage in respect of court judgements secured against the insured in specified territories (most frequently the United States and North America).
Which insurer pays? – Different policy ‘triggers’
Traditionally, liability insurance was written either on a ‘causation’ or, or more frequently, ‘occurrence’ basis, with the effect that the insurer that meets the claim would be the one on risk when the original negligent or wrongful act was committed by the policyholder (causation trigger) or when the resulting damage happened (occurrence trigger). This means that the insured is covered under the original insurance contract even if the claim is made at a much later date, when the policy may have been replaced or even cancelled altogether. This type of contract works well for instantaneous accidents, such as those which give rise to motor claims. However, it is less effective where the injury or harm is latent or occurs gradually, or both. In this case, the possibility of long-tail claims lodged many years after the policy was written makes it difficult for insurers to price the insurance.
The emergence of a number of long-tail risks, including gradually developing industrial diseases, have made these problems more acute for insurers in recent years. In response, insurers developed the so-called ‘claims-made’ trigger. Under this arrangement, the insurer that responds is the one on risk when the claim by the injured party is made against the insured. In this way, the difficulties of long-tail claims for insurers are alleviated because the time lag between writing the insurance contract and paying claims is very much shortened. Claims reserving – the process of deciding how much money needs to be set aside for future claims – becomes simpler, and the insurer is effectively able to reprice the insurance from year to year. Since losses are met by the current insurer, rather than one that was on risk at an earlier date, policyholders are able to review the adequacy of their cover annually. Again, there is little possibility of the insurer being unidentifiable or no longer in business when a claim comes in.
For the insurance buyer, claims-made cover has one key disadvantage. It is the risk that the insurer may decide to impose harsh terms on the renewal of a particular policy, or even refuse to renew altogether, if problems from the past begin to emerge. In an extreme case, the insured might find it impossible to buy substitute cover and be left without any protection. In practice, this problem is alleviated in two ways. First, insurers always undertake to indemnify policyholders, not only in respect of claims made against them during the period of insurance but also for claims arising from incidents, events or occurrences notified in the period of insurance, which s
Firm history and track record
It will be apparent from what has been said already that for insurers, the history, track record, reputation and, especially, past claims experience of the organizations they insure is crucial. Long-established businesses with sound reputations and good claims experience are, rather obviously, regarded as good risks, whereas newly established firms are an unknown quantity and likely to make insurers cautious and conservative in their approach and pricing. Similarly, drugs and medical devices that have been in production and use for many years pose much less risk than new products. Obviously, first into man trials of new drugs or devices generate the greatest level of uncertainty since specific claims data will be absent. However, insurers will normally have experience in relation to the
People at risk and the scale of the risk
The intended use of a product, and by whom it is likely to be used, is always a key factor for insurers. In our case, insurers are likely to be cautious with regard to what we might describe as vulnerable users, such as young children, the elderly or women in pregnancy and childbirth. A further dimension of any liability risk is the number of people who are likely to be exposed to it, so insurers will want to know, for example, turnover figures for drugs or devices in production. Similarly, in the case of clinical trials, long, large-scale studies involving many hundreds of volunteers are evidently the riskiest. Another, connected, aspect is the limit of indemnity – the amount of insurance cover – needed by the insured. This might be anything from about £1 million for a small firm or single investigator in a clinical trial to many hundreds of millions for insurance carried by the largest corporations. ABPI guidelines suggest a limit of £5 million for first into man studies and £2.5 million for others.
Territory of the risk
Firms that have substantial exposure to litigious territories pay higher premiums for all forms of liability insurance, including product liability, E&O, medical malpractice and D&O liability; perhaps three times higher in cases where there is substantial US exposure. Similarly, clinical trials using volunteers in these territories will be more expensive to insure. Major pharmaceutical firms almost inevitably have such exposure since they are likely to have customers and carry out or sponsor trials and research worldwide. In many cases, they will raise capital in the United States also and so face the prospect of D&O claims from angry activist shareholders if problems in the firm cause its share price to tumble. Smaller organizations, by contrast, may have little or no exposure to litigious jurisdictions.
Shifting risk
Risk is something that can be shifted, contractually, to others, and so insurers will always take an interest in the extent to which their insured clients have either assumed extra risk contractually, or disposed of it. For example, volunteers in clinical trials assume risk through a process of informed consent, relieving sponsors, investigators, CROs and others of some liability – though of course not all. Similarly, investigators, CROs, hospitals and individual medical staff will typically seek an indemnity from the sponsors of clinical trials to cover cases where they are held responsible but have not been negligent themselves. Insurers always take a keen interest in these waivers and indemnities and in the extent to which they can be legally enforced. Liability policies often carry a so-called ‘contractual liability exclusion’, which denies coverage in cases where the insured has assumed an unacceptably high level of extra responsibility by agreement with another.
Insurance market cycles
A final point, which applies to the pricing of all non-life insurance, should be made. This relates to the volatility of insurance markets, which go through repeated cycles of high insurance premiums and high profits (the so-called ‘hard’ markets) and cheap, readily available insurance and reduced profits (the so-called ‘soft’ markets). This ‘insurance cycle’ (or ‘underwriting cycle’), which typically spans 6–9 years from peak to peak, has generated much discussion among scholars and resulted in an extensive academic literature, including well-known articles by Cummins and Outreville and Lamm-Tennant and Weiss.4,5 Whilst many theories have been put forward, the phenomenon has never been fully explained by insurance economists. However, we can note that at the time of writing the cycle is in its soft phase, meaning that insurance, even for difficult and unusual risks, is quite easily obtainable.
Footnotes
Conflict of interest
The authors declared no potential conflicts of interest with respect to the research, authorship and/or publication of this article.
Funding
The authors received no financial support for the research, authorship and/or publication of this article.