"My Disease Cost More Than Yours": It Really Depends on What is Counted

In the world of those who advocate for individuals who have a particular disease or condition, one thing that is often discussed is how much a disease costs.  In my casual observation as a scientist working with advocates, sometimes it seems like we get into a discusion of "My disease is bigger than your disease!"  Or, more appropriately, "My disease costs more than your disease and therefore deserves more attention."  Not quite like a schoolyard brawl. But definitely a sense of trying to get some attention based on the magnitude of the impact.

How do we measure the magnitude of the impact?  We often talk about direct costs (or how much we spend on medical care) and indirect costs (or measures having to do with productivity).  Sometimes, it is hard enough to measure the direct costs.  Who pays what?  How do we know how much they pay? IIs that is paid what it really costs?  How does the system (in the US or elsewhere) help to make it clear whether there is much of a relationship between what is paid and what it costs?

But that is the easy side.  Measuring lost productivity is even more complicated.  A key question that has been debated in the health economics community is whether to measure the value of the individual's time and the concept of potential productivity (often referred to as the "human capital approach") or whether to measure just the productivity lost by the firm (referred to as the "friction cost approach").

The conceptual model has been discussed in the literature extensively, but there is limited literature comparing estimates using the two approaches for the same disease with the same population.  A recent study by Paul Hanly and colleagues compared the estimates of the productivity cost of breast and prostate cancer using the two approaches.

It is at this point that we see that the numbers that are used by advocates greatly depend on what is being counted.  When counting all productivity costs over a lifetime, breast cancer has a far higher impact per person in Ireland than prostate cancer.  The breast cancer cases are younger, are likely earning more, and live longer with the impact of cancer.  However, with the friction cost approach, the two conditions are responsible for nearly identical productivity losses with prostate cancer having a slightly higher value.

In both cases, the wage is used rather than total compensation.  Total compensation includes things like employer sponsored health insurance premiums, payments to retirement accounts by the employer, etc.  Perhaps these are not issues in Ireland in the same way that they are in the United States, but they do need to be considered.

The next time you see that a given disease is costing a given country some enormous number of billions of dollars per year, be careful to stop and think "what is being counted," what should be counted, and how should we count it?  I find that I'm not ever sure of the answer to the last question.  And while the answer to the first one should be clear from reading a well-written scientific article, it may not be clear from reading a popular press interpretation.  Finally, the answer to the middle question may well depend on the policy context.  Let the user of results beware.  

Response Rates of Emergency Medical Services and Mortality

A recently published article looks at the association between reduced emergency medical system (EMS) response time and the mortality outcomes of patients.  You may be asking, "Well, why does it take a study to show that?"  It would seem logical and intuitive that faster response times are associated with better outcomes.  Many municipalities and others responsible for local EMS units have spent quite a bit of time and money trying to minimize response times.  If they did not lead to better outcomes why would we be doing such a thing?

In fact, there are many things in medical care where what is intuitive is what is done and there is not a strong evidence base to support the action.  There are many in the system who are trying to change this and move us to a more "evidence-based" medicine approach, but it takes a while.

How did this study address the question in a novel manner?  Sometimes, randomized trials are appropriate.  In this case it would be completely unethical to make it take longer to respond to some people at random.  The approach used is describe in the study's abstract which can be found here.  The author, Dr. Elizabeth Wilde, points out that some studies focusing on cardiac events have shown the expected relationship but that there was little evidence outside cardiac events and what evidence there was outside cardiac events suggested no relationship.  Why might there be no relationship when the data are analyzed?  That requires us to think about incentives and to think about who knows what.  If the caller indicates a dire emergency the dispatcher can (and has an incentive to) communicate this to the EMS unit.  This is a form of triage.  The researcher working with the data later has not idea how the dispatcher communicated with the EMS unit.  So, if the dispatcher consistently triages cases in ways that make the response times for more dire cases shorter, then those cases may do better than they would otherwise.  But if the original mortality rate for those cases was high, making it a little lower will just make it similar to the mortality rate for the ess severe cases that take longer.  Then, there will be no apparent relationship between the  time of response and the morality outcomes.

Dr. Wilde found a way to use some other data--the distance from the location of the person who called for EMS services to the nearest EMS unit--as a proxy for the response time.  People have used this type of proxy (or to use the technical term, instrumental) variable before--to show things like the effectiveness of more intense treatment for heart attacks. In that case, there was a similar concern about the severity of the condition being observable to the medical care provider but not the researcher.

In the end, Dr. Wilde found that a one minute increase in response time was associated with an 8% mortality increase one day after the incident and a 17% mortality increase 90 days after the incident.

So, now we have an evidence base for efforts to improve response time.  What is the most appropriate way to do that?  That is a separate economic, political, and normative question.  It could involve technology of locating individuals.  It could involve technology for traffic control?  It could involve enforcement of traffic rules.  Or it could involve a change in norms where people are more aware of the true costs of not moving out of the way of EMS vehicles as quickly as possible.

Regardless, the study by Dr. Wilde shows that every minute can be associated with increasing the potential to save more lives.