ICU triage: a challenge and an opportunity

In a well-publicized snapshot of the challenge of ICU triage, Chang and colleagues wrote:

Critical care services can be life-saving, but many patients admitted to intensive care units (ICUs) are too sick or, conversely, not sick enough to benefit. Intensive care unit overutilization can produce more costly and invasive care without improving outcomes.

Emphasis added. Hyder provides an interesting critique to which Chang and Shapiro respond. In this post, I shall consider over-utilization by those “not sick enough to benefit”: 23.4% of the 808 patients admitted to the UCLA Medical Center in the study by Chang et al. This over-utilization provides both a challenge and a win-win opportunity (better outcomes at lower cost) if we can meet the challenge.

In a forward-looking vision, which some may regard as optimistic, Anesi et al wrote:

In the year 2050 we will unambiguously reimburse healthcare based on value, and so there is good reason to suspect that we will have targeted and reduced many services that provide little or no benefit to patients…

It can be argued that ICU over-utilization, on average, provides no overall benefit, while significantly increasing costs. Gooch and Kahn observed that US spending on critical care represents nearly 3% of GDP, while:

In contrast, the United Kingdom spends only 0.1% of its gross domestic product on critical care services, with no evidence of worse patient outcomes and similar life expectancies as in the United States. Although there are many differences between these 2 countries, one significant difference is intensive care unit (ICU) bed supply. The United States has 25 ICU beds per 100 000 people, as compared with 5 per 100 000 in the United Kingdom. As a result, ICU case-mix differs substantially. In the United Kingdom, the majority of ICU patients are at high risk for death, whereas in the United States, many patients are admitted to the ICU for observation.

As observed by Halpern, these differences come at a significant cost in the US:

The number of intensive care unit (ICU) beds in the United States has continued to increase over the last 3 decades, as have ICU utilization rates and costs, and this despite the lack of any federal, regional, or critical care society mandates to justify these increases. Some experts believe that the increase in the number of ICU beds has led to inappropriate use of these beds by patients who are either too healthy or too sick to benefit from intensive care. This may in part explain the stable national ICU occupancy rate of approximately 68% between 1985 and 2010 and suggests that ICU utilization has simply risen to meet the increased number of beds.

Emphasis added. I shall consider here only ICU usage by patients too healthy to benefit. Although the economics behind reducing ICU over-utilization by “those not sick enough to benefit” appears simple, the underlying cause is in fact likely complex.

icu-costs-fig-1

This one appears easy: lower costs and potentially better outcomes

At the same time, I recall several caveats, well known to health economists, but important in planning and communication:

  1. We expect ICUs to be available when needed, including for emergencies and disasters,
  2. ICUs have high fixed costs,
  3. Decision-making is critical: incremental costs of adding capacity become fixed costs in the future.

Chris Sampson recently reviewed a study aimed at overconsumption or misconsumption (a consequence of over-utilization). The authors of that paper suggest that “cultural change might be required to achieve significant shifts in clinical behaviour.” Chris laments that this study did not ‘dig deeper’; here we aim to dig deeper in one specific area: ICU triage for patients “not sick enough to benefit.” More questions than answers at this stage, but hopefully the questions will ultimately lead to answers.

I begin by stepping back: economic decisions frequently involve compromises in allocating scarce resources. Decisions in health economics are frequently no different. How scarce are ICU resources? What happens if they are less scarce? What are the costs? Increasing availability can frequently lead to increased utilization, a phenomenon called “demand elasticity”. For example, increasing expressway/motorway capacity “can lead to increased traffic as new drivers seize the opportunity to travel on the larger road”, and thus no reduction in travel time. Gooch and Kahn further note that:

The presence of demand elasticity in decisions regarding ICU care has major implications for health care delivery and financing. Primarily, this indicates it is possible to reduce the costs of US hospital care by constraining ICU bed supply, perhaps through certificate of need laws or other legislation.

I offer a highly simplified sketch of how ICU over-utilization by those “not sick enough to benefit” is one driver of a vicious cycle in ICU cost growth.

icu-costs-fig-2

ICU over-utilization by patients “not sick enough to benefit” as a driver for ICU demand elasticity

Who (if anyone) is at fault for this ICU vicious cycle? Chang and Shapiro offer one suggestion:

For medical conditions where ICU care is frequently provided, but may not always be necessary, institutions that utilize ICUs more frequently are more likely to perform invasive procedures and have higher costs but have no improvement in hospital mortality. Hospitals had similar ICU utilization patterns across the 4 medical conditions, suggesting that systematic institutional factors may influence decisions to potentially overutilize ICU care.

Emphasis added. I note that demand elasticity is not in itself bad; it must simply be recognized, controlled and used appropriately. As part of a discussion in print on the role of cost considerations in medical decisions, Du and Kahn write:

Although we argue that costs should not be factored into medical decision-making in the ICU, this does not mean that we should not strive toward healthcare cost reduction in other ways. One strategy is to devise systems of care that prevent unnecessary or unwanted ICU admissions—given the small amount of ICU care that is due to discretionary spending, the only real way to reduce ICU costs is to prevent ICU admissions in the first place.

Du and Kahn also argue for careful cost-effectiveness analyses, such as that supported by NICE in the UK:

These programs limit use of treatments that are not cost-effective, taking cost decisions out of the hands of physicians and putting them where they belong: in the hands of society at large… We will achieve real ICU savings only by encouraging a society committed to system-based reforms.

Emphasis added. One can debate “taking cost decisions out of the hands of physicians”, though Guidet & Beale‘s and Capuzzo & Rhodes‘s argument for more physician awareness of cost might provide a good intermediate position in this debate.

Finally, increasing ICU supply (that is, ICU beds) in response to well-conceived increases in ICU demand is not in itself bad; ICU supply must be able to respond to demands imposed by disasters or other emergencies. We need to seek out novel ways to provide this capacity without incurring potentially unnecessary fixed costs, perhaps from region-wide stockpiling of supplies and equipment, and region-wide pools of on-call physicians and other ICU personnel. In summary, current health-related literature offers a wide-ranging discussion of the growing costs of intensive care; in my opinion: more questions than answers at this stage, but hopefully the questions will ultimately lead to answers.

Credits

Variations in NHS admissions at a glance

Variations in admissions to NHS hospitals are the source of a great deal of consternation. Over the long-run, admissions and the volume of activity required of the NHS have increased, without equivalent increases in funding or productivity. Over the course of the year, there are repeated claims of crises as hospitals are ill-equipped for the increase in demand in the winter. While different patterns of admissions at weekends relative to weekdays may be the foundation of the ‘weekend effect’ as we recently demonstrated. And yet all these different sources of variation produce a singular time series of numbers of daily admissions. But, each of the different sources of variation are important for different planning and research aims. So let’s decompose the daily number of admissions into its various components.

Data

Daily number of emergency admissions to NHS hospitals between April 2007 and March 2015 from Hospital Episode Statistics.

Methods

A similar analysis was first conducted on variations in the number of births by day of the year. A full description of the model can be found in Chapter 21 of the textbook Bayesian Data Analysis (indeed the model is shown on the front cover!). The model is a sum of Gaussian processes, each one modelling a different aspect of the data, such as the long-run trend or weekly periodic variation. We have previously used Gaussian processes in a geostatistical model on this blog. Gaussian processes are a flexible class of models for which any finite dimensional marginal distribution is Gaussian. Different covariance functions can be specified for different models, such as the aforementioned periodic or long-run trends. The model was run using the software GPstuff in Octave (basically an open-source version of Matlab) and we have modified code from the GPstuff website.

Results

admit5-1

The four panels of the figure reveal to us things we may claim to already know. Emergency admissions have been increasing over time and were about 15% higher in 2015 than in 2007 (top panel). The second panel shows us the day of the week effects: there are about 20% fewer admissions on a Saturday or Sunday than on a weekday. The third panel shows a decrease in summer and increase in winter as we often see reported, although perhaps not quite as large as we might have expected. And finally the bottom panel shows the effects of different days of the year. We should note that the large dip at the end of March/beginning of April is an artifact of coding at the end of the financial year in HES and not an actual drop in admissions. But, we do see expected drops for public holidays such as Christmas and the August bank holiday.

While none of this is unexpected it does show that there’s a lot going on underneath the aggregate data. Perhaps the most alarming aspect of the data is the long run increase in emergency admissions when we compare it to the (lack of) change in funding or productivity. It suggests that hospitals will often be running at capacity so other variation, such as over winter, may lead to an excess capacity problem. We might also speculate on other possible ‘weekend effects’, such as admission on a bank holiday.

As a final thought, the method used to model the data is an excellent way of modelling data with an unknown structure without posing assumptions such as linearity that might be too strong. Hence their use in geostatistics. They are widely used in machine learning and artificial intelligence as well. We often encounter data with unknown and potentially complicated structures in health care and public health research so hopefully this will serve as a good advert for some new methods. See this book, or the one referenced in the methods section, for an in depth look.

Credits

Chris Sampson’s journal round-up for 11th April 2016

Every Monday our authors provide a round-up of some of the most recently published peer reviewed articles from the field. We don’t cover everything, or even what’s most important – just a few papers that have interested the author. Visit our Resources page for links to more journals or follow the HealthEconBot. If you’d like to write one of our weekly journal round-ups, get in touch.

Estimating the medical care costs of obesity in the United States: systematic review, meta-analysis, and empirical analysis. Value in Health Published 6th April 2016

I’m always a little wary of the “[insert disease] costs the economy $[insert big number] per year” studies. There is just too much up for debate: whether a cost can be attributed to the disease; who bears the cost; whether in fact it should be considered a cost at all. A second look through the lense of a critical review is just what these studies need. Obesity is a big deal, but there is wide variation in estimates of its cost to the US economy. This study includes a systematic review and meta-analysis looking at the medical costs of obesity estimated by studies between 2008 and 2012. Twelve studies were included in the review. The annual cost of obesity per person that was reported in the studies ranged from $227 to $7269. Wow! The pooled estimate from the meta-analysis was $1910; around $150 billion for the US as a whole. The authors looked at the methods used in the studies, but due to the variation in methods chosen and data used they weren’t able to learn that much about how this might affect estimates. The studies aren’t entirely comparable to one another. So the authors also carry out an original analysis using data from the Medical Expenditure Panel Survey to explore the impact on estimates of alternative modelling strategies. The analysis was varied by 4 age groups, 5 statistical models and 4 sets of confounders to give 80 estimates in total. The alternative statistical models didn’t make much difference, but the authors found that their extended estimating equation had the best goodness of fit. This analysis found an average cost of $1343 per person. Age groups and confounders were important. Costs were especially high in the over 65s. Older obese people have a lot of obesity-related diseases, while obese children have very few and have relatively low costs. Controlling for obesity-related disease explained away most of the incremental cost. This brings us back to the question of what should and what shouldn’t be considered a cost of the disease. What we really want to know is the counterfactual cost of the presence of obesity; what if these people weren’t obese? It remains unclear how studies might even go about defining this, let alone actually estimating it.

Introduction of a national minimum wage reduced depressive symptoms in low-wage workers: a quasi-natural experiment in the UK. Health Economics [PubMedPublished 4th April 2016

The introduction to this paper states that “no study has investigated the health effects of the UK National Minimum Wage”. That took me by surprise. So – apparently – here is the first, and it’s particularly relevant given the recent introduction of the so-called ‘National Living Wage’. The authors use data from the BHPS to test whether the increase in wages for low earners associated with the introduction of the minimum wage resulted in a positive health effect. A difference-in-differences analysis was performed using data from just before and just after the introduction of the minimum wage. Health effect is measured using the General Health Questionnaire (GHQ), which asks about current mental health problems relative to what the respondent normally feels. The intervention group was those earning less than £3.60 per hour in 1998 and between £3.60 and £4.00 per hour in 1999. There are 2 alternative control groups; one consisting those earning just above the minimum wage in 1998 and another for people whose employer did not comply. Plenty of effort is made to try and isolate the effect by incorporating physical health changes into the model and exploring the role of financial strain as a mediating effect. The results show a (statistically significant) positive impact on the GHQ. But the results aren’t quite as compelling as they might at first seem. There are a lot of exclusions that might not stand up to scrutiny, and the intervention group was made up of just 63 people. It would be good to see the analysis adapted into an economic evaluation of the policy.

An econometric model of healthcare demand with nonlinear pricing. Health Economics [PubMedPublished 4th April 2016

In Germany, health insurance is mandatory and most people receive their coverage through a public system. Between 2004 and 2013 it operated an interesting policy: the first visit to a doctor in each calendar quarter was subject to a co-payment of €10, with no copayment for subsequent visits. That’s not a lot of money for most people, but instinct would tell me that at least some people would probably avoid a single visit within a quarter and perhaps bunch-up visits if possible. This study tests that instinct. The authors develop a model of health care demand based on health shocks arriving as a Poisson process. It assumes that the co-payment increases the probability of no visit taking place and that if one does take place then this is more likely to be later in the quarter. A joint analysis of two difference-in-differences experiments is used, based on both the introduction and the repeal of the policy. The control group consists the people with private health insurance who were not affected by the policy changes. Data come from the German Socio-Economic Panel and the main analysis included over 30,000 observations. This was in part thanks to the development and successful implementation of a method to address mismatching between observation date and calendar quarters. None of the various model specifications identified a statistically significant effect of the policy on the number of doctor visits, so I suspect it won’t be reintroduced any time soon.

Diagnosing the causes of rising health-care expenditure in Canada: does Baumol’s cost disease loom large? American Journal of Health Economics Published 31st March 2016

Baumol’s cost disease is a neat idea: health care costs will rise faster than most others because health care is labour intensive and – while wages will grow in line with other industries – productivity growth cannot keep up. There’s some evidence that Baumol’s cost disease does exist, but there is less evidence about how big a deal it is compared to other non-observable drivers of rising health care expenditure. As for many other countries, Canada’s health care spending has grown at a much faster rate than the consumer price index. This new study looks at national and provincial data from Canada for 1982-2011 and decomposes the growth rate into that driven by the cost disease, technological progress and observable factors. Observable variables include population ageing, per capita income growth, economic recession and social determinants of health. The analysis uses a recently developed method, referred to as the Hartwig-Colombier test, to evaluate the impact of Baumol’s cost disease. In line with previous research, growth in per capita income is shown to be the most important driver of health care spending growth. For all provinces, the analysis finds that the cost disease is relatively unimportant. Technological progress appears to have a far greater influence, accounting for at least 31% of spending increases. Furthermore, the authors find that population ageing is not such a big concern and that the spending increases resulting from it are manageable. The implication is that if Canada wants to control spending growth then it should focus on managing the adoption of new technologies.