The questions we are looking to answer are varied:
- What kinds of perils have the highest potential cost impact given the demographic and geographic distribution of the covered population?
- Are there elements of the business strategy that make a health insurer more susceptible to climate risks?
- Does the overall health status of the existing portfolio mean a greater magnitude of risk to a changing climate? For example, do I have high proportions of people with existing respiratory diseases that might be exacerbated by climate change?
Climate change, health and healthcare usage: Understanding the starting point
Climate change can impact health both directly and indirectly. Direct impacts may include, for example, the additional use of emergency services due to injuries and exacerbation of chronic illnesses from air pollution caused by wildfires or other weather-related catastrophes, as well as mental health effects. Indirect impacts can be equally important—for example, during a flood, patients may not be able to access their usual healthcare providers, utilisation rates may actually fall and the profile of claims costs experienced may change significantly.
Depending on the country and product or benefit coverage, some insurers hold only a limited set of data on their members—some demographic variables, age, sex, address etc., and the historical claims information, which may be very rich in detail or may be relatively sparse. Some information from medical underwriting or from annual health risk assessments may be available but is rarely codified and linked with claims data. When considering the propensity for members’ health to be affected by climate change, it is important to understand the current health status profile and associated healthcare usage of the insured population. To this end, there are some tools and techniques which can be useful in classifying the covered population.
A useful starting point to building an understanding of the impact climate change might have on healthcare is to start with considering the different ways in which climate change can impact a specific service within a healthcare system for a specific group of customers, defined by health status or chronic disease profile.
We start with a population segmentation that considers chronic disease profiles for the covered population. Unlike traditional ways of looking at types of claims by procedure or diagnosis code in isolation, or segmenting by age and sex, this instead segments the population into mutually exclusive categories defined by health status. The steps are then defined as set out below:
The modelling does not attempt to assign probabilities to different climate events, but instead provides a framework for considering the impacts of different climate perils over the short and long term, according to scenarios defined by the insurer. The population segmentation and healthcare usage split can be more or less granular, according to the data that is available. However, it should be clinically meaningful and predictive of future costs and therefore the more information available about the current health status of the population the better. To estimate the longer-term effects, it is necessary to build a plausible transition matrix of the probabilities of developing chronic diseases from weather-related perils, which will necessarily rely heavily on clinical literature and judgement, rather than historical data.
Mortality can be modelled in a similar way—with the advantage that there is a substantial body of literature connecting, for example, heatwaves with excess mortality. By segmenting the population by health status and defining specific scenarios with excess temperatures for a specified length of time, the vulnerability of certain groups to heatwaves and premature death can be estimated and hence the overall likely mortality burden calculated.
We have presented a high-level framework for modelling the impact of climate change-related events on health insurers. Because of the complexity of this topic, it is difficult for health payers to conceptualise the potential impact on claims costs for health portfolios. Even once the model has been conceptualised, parameterisation is challenging. It is clear there are significant limitations, due to both the paucity of data available to estimate an insured’s health status and to the lack of generalisability from historical events to future events. There is also the difficulty in following cause through to effect, from climate change through to health burden and through to impacts on claims experience from healthcare usage or mortality for a portfolio with specific demographic, health and geographical characteristics. However, by applying a systematic approach, health insurers can attempt to estimate the relative cost and mortality impact of different climate-related perils and the long-term impact in terms of the vulnerability of their specific demographic profiles to climate-induced changes.