Who knew healthcare could be so complicated? Snapshots from an American dataset

Just as the distribution of wealth in the United States exhibits dramatic skews – a small percent owns a disproportionate share of the total wealth – so too does the distribution of healthcare expenditures. When individuals in the US population are ranked based on their healthcare expenditures in a particular year, then it turns out that:

1.     The top 1% of individuals account for 22.8% of the total healthcare expenditures
2.    The top 5% of individuals account for 50.4 % of the total healthcare expenditures
3.    The bottom 50% account for only 2.8% of total healthcare expenditures

https://meps.ahrq.gov/data_files/publications/st497/stat497.pdf

(Healthcare expenditures refer to all payments made related to health events – either by insurer or out-of-pocket.)

The estimates are from 2014, but the trends remain quite consistent from year to year. It is true that older individuals are more likely to have higher expenditures. But even if we look only at those over 65, we will still find that a small percent has an outsize impact. There is a fractal-like consistency to the pattern: if we narrowed our search down to the top 1% in a population of 10,000, then among these 100, the top 1-5 individuals will still account for a large percent of the total.

A similar trend emerges when we look instead at the prevalence of health conditions. If we were to plot the percent of individuals in a population (y-axis) who had no health conditions (count=0 on the x axis), exactly 1 health condition (count=1), exactly 2 health conditions (count=2) and so on, we would get something like the graph to the right. About 45% of the population has no apparent health conditions; about 25% has exactly one health condition; 12% has exactly two health conditions. The percentages start to decline as the count of conditions increases, indicative of the few who have 6, 7, 8, 9 or more conditions. We are now at the tail of the distribution where healthcare costs are most likely to be concentrated.    

Because most of us in any particular year are healthy, the challenges faced by this small segment of the population can remain somewhat distant. Yet at some point in our lives – hopefully later than earlier or even better not at all: who can say – there is always a chance that we might join their ranks.

In this column, I will present visualizations of healthcare use by individuals at the tails of the cost and health condition count distributions. I started creating these visualizations while researching a publicly available dataset called the Medical Expenditure Panel Survey - MEPS for short. This is the same dataset that was used to characterize the expenditure distribution above. Aggregate trends are valuable, but it is by looking closely at individual cases that one can begin to sense what is going on. Each year MEPS collects granular data on health events for members of thousands of households across the United States. Households are chosen in the survey to represent the national demographic; each household is compensated for the time spent filling out questionnaires. To protect the identities of those surveyed, the data is anonymized before it is released to the public.

A few limitations first. While I have worked with physicians, nurses and pharmacists on research projects, I do not myself have any clinical training. I therefore cannot make any claims about whether the care received was appropriate or not. Also, I have a sense but not a detailed understanding of the arcane administrative process between insurers, hospitals, doctors' officers, and suppliers of pharmaceuticals that that leads to charges and payments in the United States – or any other country for that matter.

This piece is therefore limited to what I saw in the data. What I saw surprised me – I was somewhat naïve to the complexity of modern healthcare, and it is this complexity that I'd like to communicate through examples.

Example 1

The first example is of a 50-year old female who was #10 when it came to expenditures ($209,370) among 35,313 individuals surveyed in MEPS 2011. The figure below shows a detailed split of her healthcare use in 2011, and how it changed in 2012 (each individual stays in the survey for up to two years).

There were no emergencies or hospitalizations in 2011 or 2012. 95% of the expenditures are on pharmaceuticals: 198 prescriptions were filled in 2011. While this number includes refills, it's likely that the prescriptions are not all of the same drug, but a mix. I say this because the 42 office based visits/consultations with physicians were spread across 6-7 specialties. Each specialist was likely prescribing something different. By downloading the right data files and linking it to this particular individual you can figure out which medicines were prescribed in 2011 and 2012, the diagnoses codes for each doctor visit, how much was paid out of pocket, how much by the insurer etc. Notice how in 2012 there are more office visits/doctor consultations (57) and 4 new outpatient procedures, but the number of prescriptions goes down to 89. Expenditures drop by over $100,000 and the same person is now ranked 81 among those surveyed in 2012.  

Example 2

My second example is of an 82- year man with high blood pressure, coronary heart disease, angina pectoris, and high cholesterol. He had a heart attack at 77, was diagnosed with diabetes the year of the survey. In the figure below, I show how the 75 doctor visits this patient had in a 2-year period were spread across 12 different specialties. The number on each edge indicates how many times the patient visited a particular specialty. Int. Med is Internal Medicine; Onc is Oncology; Orth. Is Orthopedics; Gast. Is Gastroenterology; Urol. Is Urology; ENT is Ear, Nose, Throat; Rheu is Rheumatology; Derm. Is Dermatology; Opth. is Ophthalmology; Card. Is Cardiology; Neph. is Nephrology; and PCP is primary care. 

In the qualitative parts of the survey the patient admits to having walking and vision limitations. It must not have been easy keeping up with so many doctor visits. Imagine setting up the appointments – 75 appointments in 2 years implies nearly 2 appointments every 3 weeks – driving to the clinics at the appointed time, experiencing waits, filling up multi-page forms, answering the same questions again and again, remembering all the medications prescribed. Even if there is a spouse or family member to help, being a patient seems like a demanding part-time or full-time job.

My 83-year old uncle who lives in India once joked that nowadays there is a doctor who can treat the top half of your index finger, another one for the middle section of the finger and so on. He was parodying the proliferation of multi-specialty hospitals in India. I found myself agreeing with him. But I also realized that specialization is an inevitable consequence of how quickly medical knowledge has expanded over the last hundred years. It is unreasonable to expect a single doctor to master all of it. And so when faced with a situation beyond her expertise, the primary care physician – the family doctor or the general practitioner, who is often the first point of contact and who knows the patient's medical history the best – refers the patient to a specialist. When a patient has just one condition, a primary care-specialist pair can work well as a team. But when a patient is seeing 11 other specialists, making sense of the changes in symptoms, the numerous diagnostic test results and medications can be a challenge, even if electronic medical records capture all this information (they often don't).  

Example 3

This leads me to my third example, which gives us a physician's perspective. I am borrowing it from a short article published in the New England Journal of Medicine. Its author, Matthew Press, is a primary care physician (PCP) at the Massachusetts General Hospital in Boston. In the article, titled "Instant Replay: A Quarterback's View of Care Coordination", Press describes the case study of a 70-year old patient whom he calls Mr. K:

"M. K's care was fairly straightforward— I was the only doctor he saw regularly — until the day he came into my office with flank pain and fever. A CT scan of his abdomen revealed a kidney stone — and a 5-cm mass in his liver, which a subsequent MRI indicated was probably a cholangiocarcinoma…Over the 80 days between when I informed Mr. K. about the MRI result and when his tumor was resected, 11 other clinicians [a urologist, a surgeon, a hematologist, a neurologist, a lab technician, a gastroenterologist, an interventional radiologist, an oncologist, a cardiologist, a pathologist, and a social worker] became involved in his care, and he had 5 procedures and 11 office visits (none of them with me). As the complexity of his care increased, the tasks involved in coordinating it multiplied…In total, I communicated with the other clinicians 40 times (32 e-mails and 8 phone calls) and with Mr. K. or his wife 12 times. At least 1 communication occurred on 26 of the 80 days, and on the busiest day (day 32), 6 communications occurred." (Excerpt and figure from Press 2014)

Press kept a running list of all the tasks – emails, phone calls, visits etc related to Mr. K's care. This led to the figure above and an animation on the journal website. It's worth watching how how the figure starts empty but fills up over 80 days. Luckily, Mr. K recovered and responded well to the treatment provided. After the 80-day burst of encounters things settled down. Press goes on to note that:

"Patients can be harmed when the many moving parts of their care are out of sync. We owe it to them to coordinate the care we provide and prevent this type of medical error. For example, on day 32 of Mr. K.'s care, a Friday, I noticed some new electrolyte abnormalities on laboratory tests done before an interventional radiology procedure. First I called the cardiologist who had seen Mr. K. earlier that week, after I learned from the electronic medical record (EMR) that he had prescribed a new antihypertensive. Then I called Mr. K. to arrange to have his electrolytes rechecked, which had to be done at an outside laboratory because by then it was the weekend (this took two calls to the laboratory — one to schedule and one for the results). On Sunday, I had Mr. K. change medications and on Monday asked the interventional radiology nurse practitioner to recheck the labs again before the procedure (two more calls). On day 36, she did, and the electrolytes had normalized."

Because Press had research duties, he worked part time and was the personal doctor for small number of patients. A relatively low workload gave him the time to meticulously track Mr. K's needs. In contrast, a typical doctor who works full time oversees the care of hundreds of patients. Press' article suggests that an overburdened physician will have much less time to coordinate the many dimensions of an individual's care, increasing the risk of serious errors.

Medications and Hospitalizations

There are two other situations in which modern healthcare can get complex. Just as a large number of specialists can become involved in a patient's care, so a patient may be prescribed a large number of unique medications: the former is actually a reliable indicator of the latter. In some datasets I've noticed patients who had been prescribed 15 or more unique drugs. There was one person who was on a staggering 30 unique medications! The trend is common enough to now have it own Wikipedia page and is called polypharmacy.

Other than the difficulty a person faces in remembering what should be taken, at what time of the day, when to refill, and why something is being taken, it made me wonder: how do clinicians and pharmacists weigh the side effects and possible interactions between all these medications? One medication in isolation can be tested in what is called a randomized controlled trial; the interactions between two, three or four medications gets harder, but nevertheless I assume they can be tested. It's mind-boggling to consider how a dozen or two dozen unique medications might interact.

The second type of situation is repeated emergency events and hospitalizations. The image below shows the sequence of events for a 42-year old man surveyed in MEPS 2011. He was hospitalized four times. The horizontal timeline marks the months of the year: J for Jan, F for Feb, and so on; PCP refers to a primary care visit. The last hospitalization happened at the end of December, the most festive time of the year.

 

As with other things we've discussed, repeated hospitalizations and emergency events are prevalent in a small percent of the population. A hospital stay and the period after discharge can be extremely disorienting. Medical issues can be aggravated by disabilities, mental health concerns, substance abuse, absence of a supporting family member, and lack of steady housing. Absent an effective plan on what will happen after leaving the hospital, it is not uncommon for patients to get readmitted within days or weeks of discharge. The sickest patients end up receiving lots of expensive hospital care – which puts them into the top 1-5% - but that care isn't good enough to avoid future hospitalizations.  

Concluding Remarks

To summarize, the healthcare expenditure distribution is highly skewed. The vast majority of the population uses little or no healthcare, which is as it should be. Medicine has made significant progress is curbing infant mortalities and infectious diseases, and this has lead to longer lifespans. For most of that lifespan we can now hope to stay healthy. But it is also true that worldwide, even in developing countries, there's a been a shift from communicable diseases to the slow burn of chronic conditions. These diseases can manifest at any age but particularly when we grow older. Along with diabetes, arthritis, depression and mental health, certain types of cancers are now considered as chronic conditions that need to be managed long term.

The speed of technological advances in medicine and incentives to aggressively market those advances have led to numerous detection and treatment options. These options are supposed to make things easier and they often do. Yet they have also contributed to the complexity of modern healthcare. Many different specialists can get involved in a single patient's care, conduct a variety of diagnostic tests, prescribe a mix of medications. Such an escalation in intensity places a high burden on the patient without necessarily guaranteeing better health. The US Centers for Disease Control (CDC) acknowledges "that as the number of chronic conditions increases, the risks of the following outcomes also increase: mortality, poor functional status, unnecessary hospitalizations, adverse drug events, duplicative tests, and conflicting medical advice".

The medical community recognizes the seriousness of the problem and its outsize impact on healthcare costs. Yet it has struggled to formulate effective responses. This is because the problem is not about tackling one disease in isolation, rather it is about how combinations of diseases collectively impact an individual. Complicating matters are socioeconomic factors and entrenched incentives which do not reward a holistic approach. Nevertheless a range of responses have emerged. I will finish the piece by briefly describing a couple.

One response has been to employ nurses and social workers visit medically complex patients on a regular basis, often in the patient's own home. Their goal is to simplify patient's care to the extent possible, help organize medications, and avoid unnecessary emergencies and hospital visits. They also address issues such as lack of employment, housing, insurance, mental health and addictions. In the process, the nurses and social workers hope to establish genuine relationships with their patients. A prominent illustration of this approach comes from Camden, New Jersey, featured here in a 13-min PBS Frontline video. The Camden initiative, which MIT is evaluating in a research study, explicitly recognizes that medical issues, especially in cities such as Camden struggling with crime and economic decline, are inextricably linked to the social context. Similar approaches are being tested across the country, from urban Houston to rural Montana.

Another interesting response that caught my attention is a conceptual framework called minimally disruptive medicine. The premise of minimally disruptive medicine is that patients can be overwhelmed by the demands that healthcare places on them. In the push to measure this and diagnose that, to try this new therapy or that new medication, what is truly important can get lost. Supporting this view are patient comments such as these:

"To keep myself healthy, I miss out on a lot of things that people my age take for granted – working fulltime, cooking, showering every day, going out to socialize" [25-year old woman from the UK] 

"There is stuff that I am SUPPOSED to do, and stuff that I actually DO.  If I did everything I am SUPPOSED to do, my life would revolve around doctors and tests and such and there wouldn't be very much left for living my life.  So I've made a bunch of choices (with the input of my family and friends, because it's important for me to have their support)." [46-year old woman in the US] 

Beyond a certain threshold the demands of healthcare, rather than make things better, can end up depleting the patient's financial, social and emotional capacity. This in turn makes it less likely that patients will follow recommended treatments. Minimally disruptive medicine asks: Is it possible to reduce the burden of treatment? Is it possible to truly listen to what the patient wants? Can the treatment be better aligned with the patient's goals in life: pursuing professional interests, spending time with family and friends, and having fun?

Perhaps I'll write a piece explaining these ongoing efforts in more detail. In the meantime, I'd love to hear what readers think.