In the revision, just as in the original version, though the authors do not definitively rule out cost shifting, they do not necessarily interpret their findings as evidence of it. As I wrote in my reaction to the original version, we should not either. I provided several reasons, among which were this:
Hospitals were aware in advance of their risk of being penalized. Those that looked like they would be may have invested more in quality improvements. For some, those improvements didn’t translate into avoiding the penalty, but they did increase the value those hospitals were delivering. Private payers may have been willing to pay more for that additional quality. It’s possible they’re even willing to pay more for investments in quality that haven’t yet translated into actual improvements.
[Rick] Sherlock [president and CEO of the Association of Air Medical Services] explained that reimbursements from Medicare and Medicaid do not cover the cost of providing services. So charges to private patients, he told the legislators, must make up that difference.
In case you need a fresh illustration of third degree price discrimination (or "segmentation" in b-school lingo), check out the cross-country variation in Spotify's monthly subscription price. pic.twitter.com/vfItpiqPxC
Is Spotify charging higher prices in Denmark because consumers in Vietnam will only pay $2.50 per month? Do you think the Danes think this too? That would be cost shifting and it’s what the air ambulance industry wants us to believe about their private prices with respect to Medicare/Medicaid ones. It’s what the hospital industry wants us to believe too. But if it’s cost shifting in those sectors, why not in the case of Spotify? (Hint: None of this is cost shifting. All of it is price discrimination.)
Most Americans will struggle with low back pain (LBP) at some point in their life. Current clinical guidelines recommend noninvasive, conservative treatment first — physical therapy, chiropractic care, exercise, etc. — before moving to surgical treatments or opioids. Not only does this save the patient from more aggressive intervention, it may save money overall.
But saving money overall is not the same thing as saving the patient money. Repeated visits to the physical therapist or chiropractor can be expensive to patients, as copayments add up. They’re also time consuming. Patients in plans with more narrow networks of those kinds of providers might have to spend more time finding and traveling to one.
A recent paper by Kathleen Carey and colleagues studied these two components of a patient’s insurance coverage — cost sharing and network — and how they affect LBP treatment decisions. They hypothesized that a broader provider network would be associated with a higher likelihood of initially choosing conservative treatments for LBP. They also predicted increased out-of-pocket (OOP) costs would be associated with a lower likelihood of initially choosing conservative treatments for LBP. (The authors hold positions with the Departments of Health Law, Policy, & Management and Biostatistics at Boston University School of Public Health; the Department of Family Medicine at Boston Medical Center; OptumLabs; and the University of Pittsburgh School of Health and Rehabilitation Sciences.)
Each analysis in the paper included four sub-analyses based on particular plan characteristics: plan type, co-pay, deductible, and health savings account (HSA) or health reimbursement account (HRA) enrollment. The reference groups for each sub-analysis were, respectively: POS plan (a managed care plan that blends characteristics of both HMO and PPO plans), zero co-pay, zero deductible, and no HSA or HRA.
Four plan types were included, representing a range of provider networks: point of service (POS), exclusive provider organization (EPO), health maintenance organization (HMO), and preferred provider organization (PPO).
The study population included adults with private insurance and a new diagnosis of LBP during 2008-2013. Patients were excluded if they had a prior LBP diagnosis, prior back procedures, or an LBP diagnosis that would not typically be treated with conservative treatment.
Physical therapist v. PCP
Over 82,000 patients were included in the physical therapist versus PCP analysis and less than three percent chose a physical therapist as their initial treatment provider. The authors compared patients with all four plan types and found those with a PPO plan — the least restrictive plan with a wide range of providers and out-of-network coverage options — were the most likely to initially see a physical therapist. Those with an EPO plan — one of the most restrictive plans with in-network coverage only and required referrals for specialists — were the least likely.
Regarding OOP costs, the researchers found that as co-pays or deductibles increased, patients became less likely to initially see a physical therapist. Patients with an HRA were also less likely to initially see a physical therapist while patients with an HSA were more likely.
Overall, these findings support the authors’ hypotheses. A less restrictive provider network and lower OOP costs made it more likely patients would seek LBP treatment from physical therapists before PCPs.
The HRA/HSA analysis was particularly interesting. Carey, et al. posited the incentives and financial responsibilities for each account differ. HRAs are employer-funded, meaning patients have less financial liability. The authors speculated that patients with HSAs, to which employees also contribute, may have more at stake.
Chiropractor v. PCP
Over 115,000 patients were included in the chiropractor versus PCP analysis. More than 30 percent chose a chiropractor as their initial treatment provider. The authors again compared patients with all four plan types and patients with a PPO plan were the most likely to initially see a chiropractor. Those with an HMO plan — another very restrictive plan with in-network coverage only and required referrals for specialists — were the least likely.
There was little association between co-pays and provider choice. Interestingly, deductibles were found to be associated with an increased likelihood of initially seeing a chiropractor (except for deductibles over $1500). Patients with an HRA were a bit more likely to initially see a chiropractor but an HSA had no effect.
These findings do not support the researchers’ hypotheses as clearly as those from the physical therapy analysis. A less restrictive provider network still made it more likely for a patient to initially choose a chiropractor over a PCP, but the OOP cost analysis was less conclusive. Their findings on deductibles do not support their theory because patients who paid more OOP in the form of deductibles were more likely to see a chiropractor before a PCP for LBP than those who didn’t.
Carey, et al. found that both provider network and individual financial responsibility can affect a patient’s likelihood to seek conservative therapy for LBP before visiting a PCP. (The authors caution these findings are associations and do not demonstrate causation.) However, the impact of provider network was much more apparent. If a patient had an insurance plan with a bigger provider network, such as a PPO plan, she was more likely to choose physical therapy or chiropractic care before visiting her PCP.
Understanding how insurance policies impact a patient’s treatment choices is helpful for both patients and providers to ensure each patient receives the highest value care possible within his individual plan limitations, be it provider availability or cost.
It’s also important for patient safety. Evidence suggests that patients who seek conservative treatment for LBP first through a physical therapist or chiropractor have lower odds of both early and long-term opioid use. In light of the current opioid crisis, reducing the number of patients who use opioids for LBP when there are better, more conservative treatments available is critical. (The authors recently published a separate paper on this, specifically studying this relationship in comparison to patients who see their PCP first.)
Lastly, these findings are helpful for insurance plan design as companies move towards value-based coverage. Ideally, insurance companies should encourage patients to choose conservative, lower cost treatments for LBP first whenever possible before turning to surgical treatments or opioids. One way to do this is to make physical therapy, chiropractic care, and the like easily accessible and available at low or no cost to the patient so she and her provider choose them first in line with clinical guidelines.
As usual, the study (and some of the stories) lacked some important context and caused more worry than was warranted. There are specific reasons that this cycle is unlikely to end.
If it’s artificial, it must be bad.
People suspect, and not always incorrectly, that putting things created in a lab into their bodies cannot be good. People worry about genetically modified organisms, and monosodium glutamate and, yes, artificial sweeteners because they sound scary.
But everything is a chemical, including dihydrogen monoxide (that’s another way of saying water). These are just words we use to describe ingredients. Some ingredients occur naturally, and some are coaxed into existence. That doesn’t inherently make one better than another. In fact, I’ve argued that research supports consuming artificial sweeteners over added sugars. (The latest study concludes the opposite.)
It’s true that no one “needs” soda. There are a million varieties, and almost none taste like anything in nature. Some, like Dr Pepper, defy description.
But there are many things we eat and drink that we don’t “need.” We don’t need ice cream or pie, but for a lot of people, life would be less enjoyable without those things.
None of this should be taken as a license to drink cases of soda a week. A lack of evidence of danger at normal amounts doesn’t mean that consuming any one thing in huge amounts is a good idea. Moderation still matters.
Scientists need to publish to keep their jobs
I’m a professor on the research tenure track, and I’m here to tell you that the coin of the realm is grants and papers. You need funding to survive, and you need to publish to get funding.
As a junior faculty member, or even as a doctoral student or postdoctoral fellow, you need to publish research. Often, the easiest step is to take a large data set and publish an analysis from it showing a correlation between some factor and some outcome.
As long as the culture of science demands output as the measure of success, these studies will appear. And given that the news media also needs to publish to survive — if you didn’t know, people love to read about food and health — we’ll continue to read stories about how diet soda will kill us.
Prestigious institutions and the press
To do the kinds of analyses described here, you need large data sets that researchers can pore over. Building the data set is the hardest part of the work.
Analyzing the numbers on hundreds of thousands of people isn’t child’s play. But gathering the data is much more expensive and time-consuming.
Because of this, a few universities produce a disproportionate amount of the research on these topics. They also tend to be the universities with the most resources and the most recognizable names. Because they’re also usually prestigious, they attract more researchers and more funding to build bigger and newer data sets.
They also get more media attention because of having access to more researchers, prestige and funding. If the research is coming out of a super-respected institution, it must be important.
Lather. Rinse. Repeat.
We still don’t understand the limitations of observational studies
No matter how many times you stress the difference between correlation and causation, people still look at “increased risk” and determine that the risk is causing the bad outcome. For reporting on hundreds of thousands of people, observational studies are generally the only realistic option. With very few exceptions, they can tell us only if two things are related, not whether one is to blame for the other (as opposed to randomized control trials).
With respect to diet sodas, it’s plausible that the people who tend to drink them also tend to be worried about their weight or health; it could be a recent heart attack or other health setback that is causing the consumption rather than the other way around. But you shouldn’t assume that diet sodas cause better health either; it could be that more health-conscious people avoid added sugars.
Many of these new observational studies add little to our understanding. At some point, a study with 200,000 participants isn’t “better” than one with 100,000 participants, because almost all have limitations — often the same ones — that we can’t fix.
Dr. John Ioannidis wrote in a seminal editorial: “Individuals consume thousands of chemicals in millions of possible daily combinations. For instance, there are more than 250,000 different foods and even more potentially edible items, with 300,000 edible plants alone.”
And yet, he added, “much of the literature silently assumes disease risk” is governed by the “most abundant substances; for example, carbohydrates or fats.” We don’t know what else is at play, and using observational studies, we never will.
(Observational research is still the best way to study population-wide risk factors; sophisticated techniques like regression discontinuity can even create quasi-randomized groups to try to get closer to understanding causality. Too few employ such techniques.)
Moreover, too many reports still focus only on the relative risk and not on the absolute risk. If a risk increases by 10 percent, for example, that sounds bad. But if the baseline risk is 0.1 percent, that 10 percent increase winds up moving the baseline to only 0.11 percent.
It would probably be a public service if we stopped repeating a lot of this research — and stopped reporting on it breathlessly. If that’s impossible, the best people can do is stop paying so much attention.
Nambi Ndugga is a Policy Analyst with Boston University’s School of Public Health. She tweets at @joerianatalie.
What follows is a summary of articles that look at the effects on in-utero exposure to a variety of stressors on future health outcomes. First, some definitions:
The fetal origins hypothesis (FOH) states that disruptions to fetal nutrition have significant effects on future health outcomes. David Barker, who is credited with the development of the theory, determined that nutritional mismatch between the intrauterine and postnatal environments could result in adverse health outcomes in later life. Barker’s original work looked at the relationship between mid to late gestation undernutrition and programmed cardiovascular and metabolic illness in later life. Current research expands on FOH to include other prenatal pathways and postnatal outcomes. These have informed the development of the Developmental Origins of Health and Disease hypothesis (DOHAD).
DOHAD was formalized in the early 2000s through the formation of the International Society for Developmental Origins of Health and Disease. The hypothesis states that the period from conception to early childhood is critical to an individual’s immediate and future health. DOHAD expands on FOH by first recognizing the limits of nutrition as a marker of prenatal pathways to postnatal outcomes. It then includes other physical, environmental, and psychological stressors on the fetus and their effects on future health outcomes. Currently, both DOHAD and FOH are interchangeably used to describe the effects of prenatal growth and development on future health outcomes. For the remainder of this post, I will use FOH or simply “fetal origins.”
The Brenner hypothesis adheres to the DOHAD hypothesis but has a narrower scope. It posits that individuals with a reduced nephron number (number of nephrons per kidney) have an increased risk of future non-communicable disease. Reduced nephron numbers are associated with intrauterine growth retardation that results in low birth weight and premature birth, both major indicators of future non-communicable diseases, including chronic kidney disease.
Studies based on the 1918 Influenza Pandemic
The 1918 Influenza Pandemic served as a natural experiment to study the fetal origins hypothesis. The pandemic arrived suddenly in October 1918 and lasted until January 1919 and affected 1/3 of pregnant women and their fetuses. Using US census data, researchers found that certain outcomes (educational, professional, etc.) followed similar patterns for the affected population throughout their lifespans.
Almond and Mazumder (2005) used data from the Survey of Income and Program Participation from 1984 to 1996, to look at health outcomes of cohorts who were in utero during the 1918 Pandemic. They found that cohorts in utero during the Pandemic had more adverse health outcomes throughout the lifespan compared to the cohorts born a few months earlier or later. Specifically, the research found that later life risks of cancer, heart problems, kidney problems, hypertension, and stomach problems have a statistically significant association with those who were in-utero during the height of the Pandemic. They also found that the 1919 cohort experienced a significant increase in incidences of functional limitations, including trouble hearing (19%), speaking (35%), lifting (13%), and walking (17%).
Mazumder et al. (2010) looked at the effects of maternal exposure to infections during pregnancy and its effects on fetal programming (when the fetal environment predisposes the fetus to postnatal disease). They found that cohorts that were in-utero during the 1918 Influenza Pandemic were associated with a greater than 20% excess incidence of adult on-set cardiovascular disease, specifically ischemic heart disease. They also found that those born during the Pandemic had excess incidences of diabetes, later in life. A potential study limitation is that their analyses were based on self-reported data. However, other studies have corroborated these findings with medical records. The researchers identified potential pathways between prenatal infection exposure and adult health outcomes. It is possible that prenatal exposure to influenza increases the risk of low birth weight, a known risk factor of later heart disease and diabetes.
Fletcher (2014) used the National Longitudinal Mortality Study to look at the mortality effects of those who were in utero during the 1918 Influenza Pandemic. While the study authors did find some links between in-utero exposure to influenza and educational attainment, but they did not find statistically significant evidence linking in-utero exposure to influenza and later mortality outcomes.
A study based on prenatal radiation exposure
The CDC found that prenatal exposure to radiation can increase an individual’s risk of cancer in later life. The level of risk is dependent on the amount of radiation the fetus is exposed to and its gestational age.
Studies based on prenatal heavy metal exposure
Bailey and Fry (2014) found associations with prenatal and early-life exposure to inorganic arsenic and later in life health outcomes including, cancer, intellectual function, and other non-communicable diseases. Similar to the radiation exposure, the health outcomes are dependent on the level or extent of exposure and the gestational age of the fetus.
Dauphine et al. (2010) conducted a pilot study in Chile looking at the effects of in-utero and childhood exposure of arsenic in drinking water on adult lung function. They drew comparisons between two cities, one where the inhabitants primarily, if not exclusively, used the arsenic-contaminated municipal water supply and another where inhabitants had minimal to no exposure. The researchers found that a link between in-utero and childhood exposure to arsenic in drinking water and long-term lung function deficits and shortness of breath in adulthood. The effects measured are equivalent to those found in those who smoke throughout their adult lives.
Studies based on fasting
Xi Chen (2014) used Ramadan fasting and festival activities as natural experiments to determine the effect of under-nutrition on prenatal health and subsequent effects on later health. In-utero exposure to maternal fasting can lead to developmental adaptations that are good for the survival of the fetus in the short term, but will result in stunted fetus growth and may increase the risk of non-communicable diseases in later life. They found that exposure to maternal fasting retards the development of a specific placental enzyme needed to convert the stress hormone cortisol to its inactive form cortisone, exposing the fetus to excess stress hormone, a known in-utero influence of adult health. The excess exposure may lead to the reprogramming of Hypothalamic Pituitary Adrenal axis which is linked to type 2 Diabetes, high blood pressure, and cognitive impairment.
Ewijk et al. (2013) looked at maternal nutrition during Ramadan to determine if intermittent fasting had lasting effects on fetal physiology. They looked at birth information, and height and weight measurements of 14,120 people (12,900 Muslims and 1220 non-Muslims). They found that Muslim individuals who were conceived or were in mid to late gestation during Ramadan were both thinner and shorter than those who were not Muslim. They found that the average adult BMI of Muslims who were exposed to Ramadan in-utero was 0.32 units lower than those who weren’t. They also found that those who were exposed to Ramadan during gestation were 0.20 cm shorter than those who weren’t exposed, however this finding was not statistically significant.
Almond and Mazumder (2008) used Ramadan as a natural experiment in fasting and fetal health. They looked at census data from several countries (US, Iraq, Uganda) to look at birth date, location, and disability status. They found strong associations between exposure to Ramadan and the increased risk of disability, especially intellectual disabilities. In Uganda, they found that Muslims born nine months after Ramadan (month-one exposure to Ramadan) saw a 22% increase on the mean disability rate (3.8%). The Iraqi census demonstrated similar results with exposure to Ramadan in the first month of pregnancy resulting in a 23% increase on the rate of disability relative to the mean (1.5%). In US they found a 18% increase on the rate of disability relative to the mean (6.07%).
Studies based on environmental maternal stress, including war, natural disasters, famine
Akresh, Caruso, Thirumurthy (2014) looked at the impact of armed conflict on subsequent health outcomes. The researchers compared children near the conflict sites compared to those further away. They found that children who were exposed to war in-utero experienced a subsequent 0.306 reduction in their age-adjusted height measurements, those who were exposed in early childhood (9 months after birth) experiences a 0.423 reduction in age-adjusted heights. The effects were higher in children from the losing country, in this case, Eritrea, compared to the winning country, Ethiopia. The impact increased if the intensity of the conflict was greater. The researchers used the concentration of war induced displacement, geographic proximity to fighting, and the length of time exposed to the conflict as proxies for conflict intensity.
St. Clair et al. (2005) studied the effects of intrauterine nutritional deficiency on the development of adult schizophrenia using the 1959-1961 famine in China as a natural experiment. They found that cohorts exposed to the famine had an increased risk of adult schizophrenia, those born before the famine had a 0.84% risk (1959), those born during the famine had a 2.15% risk (1960), and those born after the famine had a 1.81% risk (1961). Their findings replicated those of studies that looked at the effects of in-utero exposure to the 1944-1945 Dutch Hunger Winter which found that the birth cohort conceived in the height of the famine had a 2-fold increase in risk of Schizophrenia.
A study based on psychological stress
Persson and Slater (2016) studied how in-utero exposure to maternal stress from family ruptures affects mental health in later life. They found that exposure to severe psychological stress in utero, such as the death of a close family member has causal impacts on the development of future psychological conditions. They found that in-utero exposure to the death of a mother’s close relative has a large effect on the consumption of prescription drugs for mental health conditions. In children they found a 25% increase in the likelihood of purchasing ADHD medications, and a 24% increase in the average daily dose of ADHD medication. In adults they found an 8% increase in the likelihood of consuming prescription drugs for anxiety and depression, and a 19% and 12% increase in the average daily doses of these drugs.
A substantial body of research demonstrates that in-utero exposures to stressors influence future health outcomes. Given the prevalence of non-communicable diseases worldwide, this may provide policy makers and healthcare practitioners additional approaches to improve health.
My primary academic appointment is with the Department of Health Law, Policy and Management at Boston University’s School of Public Health. Below are recent publications from me and my colleagues from that department.
This will be a recurring post (~monthly) and content will be updated if publications I’ve missed are brought to my attention. You can find all posts in this series here.
We spend a lot of money on social programs. Studies indicate that a lot of those programs work to improve health outcomes. So, does that social spending reduce healthcare costs? Not really, at least not on a national level. But cutting social spending doesn’t help with health. As usual, it’s complicated, but there’s a lot of evidence these programs are doing good. Spending is worth it sometimes.
Even a divided America can agree on this goal: a health system that is cheaper but doesn’t sacrifice quality. In other words, just get rid of the waste.
A new study, published this morning in JAMA, finds that roughly 20 percent to 25 percent of American health care spending is wasteful. It’s a startling number but not a new finding. What is surprising is how little we know about how to prevent it.
William Shrank, a physician who is chief medical officer of the health insurer Humana and the lead author of the study, said, “One contribution of our study is that we show that we have good evidence on how to eliminate some kinds of waste, but not all of it.”
Following the best available evidence, as reviewed in the study, would eliminate only one-quarter of the waste — reducing health spending by about 5 percent.
Austin and Aaron are participants in the Amazon Services LLC Associates Program, an affiliate advertising program designed to provide a means for sites to earn advertising fees by advertising and linking to amazon.com.