Subscribe to DSC Newsletter

Being obese is worse than smoking. Do you agree? How do you measure costs?

This discussion is about how to correctly assess costs and benefits of smoking versus obesity to society, and where to get reliable data sources, and how to properly clean data. It is also about healthcare policies and the new health insurance regulations, sometimes referred to as Obamacare.

To decide whether smoking is more dangerous than obesity, how would you proceed? How do you compute all the costs:

  • Sick days,
  • Extra healthcare expenditures,
  • Increased health insurance costs for everyone including employers,
  • Increased costs to fly or drive a car (more fuel required),
  • Need to provide larger more expensive seats in theaters and restaurants and thus accomodate fewer patrons,
  • Depletion of food sources at a faster rate,
  • Time spent on research to fight obesity rather than on constructive projects,
  • Second-hand smoke,
  • Fires caused by cigarettes,
  • Fire alarms mandatory in new houses for everybody (a side effect of smoking)
  • How do you compute costs over lifetime, as $1,000 in 2012 is worth less than $1,000 in 2030 (need for econometric models)? And how do you factor in the fact that health expenditures costs grow even faster than regular inflation (although in my opinion it will violently deflate in a few years)?

Potential benefits of smoking and obesity:

Some will argue that those seeking self-destruction would do so even if cigarettes and big fat hamburgers were illegal, possibly using illegal products instead, which would be even worse (remember the prohibition and the alcohol black market). Also, as cigarette smokers die before retirement, they save tons of money to social security. They also allow scientists do research on many different cancers and heart problems - research that would otherwise be more limited without all the smokers and the obese. Smoking also provide huge taxes to state governments, and cigarette manufacturing employs thousands of workers.

And some would even say that smoking is a natural selection process that put some downward pressures on over-population, due to shorter lifetimes of smokers. Because it is self-inflicted, it is not as bad as famines, wars, diseases, or road accidents.

Health insurance issues:

Finally, a question about health insurance. Should an insurer ask how many hamburgers per week you eat, to determine your premium? We already ask for cigarettes (by why not for alcohol or illegal drugs?) And what about smokers who switch to tar-free cigarettes (e-cigarettes) or people eating fat-free hamburgers?

To summarize, how to gather and blend all the data, analyse it, incorporate financial gains - not just losses associated with these bad habits - and come up with strategies to improve the situation - strategies that involve multiple, competing government agencies? And how do you measure improvement (lift)?

Views: 479

Comment

You need to be a member of AnalyticBridge to add comments!

Join AnalyticBridge

Comment by Ellen R. Lis on June 17, 2012 at 5:00pm

In principle, performing correct statistical analyses on correctly collected and cleaned data could permit an accurate assessment of the relative financial costs of obesity vs. smoking on society.  Determining the correct data to collect, or use, depends upon the scope and nature of the analysis question(s) asked.  


The broad question you propose is whether smoking or obesity is worse in terms of financial impact to society. Next, you provide a list of societal impacts that have financial components for data collection and analysis.  Let's look at the first societal impact you listed: Sick Days. Let's say we had a reliable source of Sick Days data from all American employers. Are the obesity/smoking characteristics of people associated with these Sick Days the only relevant ones to consider?  What is financial impact of, say, sick days contributed by millions of unemployed adults?  These people are likely to be sicker longer than those who can afford medical treatment - with financial implications.   People in the unemployed category include stay-at-home parents, people who cannot find employment, and many retired elderly.  That's a large fraction of the American population.  A lot of these people end up in emergency rooms, costing the health system more than if they were receiving regular healthcare.  Is the relevance of the Sick Days question intended only for employers and their affiliated health plans?  What is the relevance of sick days reported for obese children, or obese/smoking teenagers, who report sick from school? Why shouldn't their sick days be included in the analysis?

People with academic or work backgrounds in many areas related to healthcare can contribute diverse and valuable input as to which subsets of a Sick Days data set can be used to contribute substantially toward answering your broad question. One might bring up the fact that there are many reasons, not related to obesity and smoking, that explain why obese and/or smoking people take sick days. What potentially collectible characteristics of a Sick Day could be used to make the source data for a Sick Days Analysis more meaningful? A diagnosis code, procedure code, data on medications/treatments prescribed for each Sick Days episode, and data identifying the subject as obese and/or smoking, for starters. This data can be collected when the obese/smoking subject obtains medical treatment during the Sick Days episode. In processes associated with medical treatment, the data can be, and often is, collected.  However, given that tens of millions of people in America do not seek medical treatment for Sick Days episodes due to lack of health insurance or funds, the relevance of our answer to the Sick Days question, based on medical treatment-dependent data alone, may be too low to be generalized to society.  There is no mandated requirement that data related to obesity and smoking be collected, in a standard way, across medical providers. Online surveys could be used to collect some of the data needed to identify people who smoke and/or are obese.  But linking that information to medical data on them would require extra steps - which should include subject consent to use his/her medical records for the analyses.

If we had reliable and complete medical treatment data on every obese/smoking person sampled over an analysis period of interest, and if we could separate obesity-associated Sick Days data from smoking-associated Sick Days data by some algorithm, and we found something of correlative value, we still have not established a causal link between obesity and Sick Days expenses, or between smoking and Sick Days expenses.  Obesity and smoking, of themselves, are not often reported as reasons (diagnoses) for a Sick Days episode.  Obesity could be explicitly indicated in the data as diagnostic for patients taking sick time to undergo the following treatments:

(1) Weight reduction at a clinic where the protocol followed involves a long term stay at the clinic

(2) A clinical nutritionist-monitored, outpatient weight reduction program where weight and body measurements related to obesity are taken every few weeks over an extended period

(3) Bariatric surgery, plus follow up appointments aimed at improving diet/exercise habits

(4) Liposuction

Doctors may attribute smoking or obesity as a contributing factor for a different diagnosis, and recommend steps for the patient to quit smoking and lose weight as part of the treatment for that diagnosis.  For example, an overweight person who is diagnosed with hypertension (high blood pressure) may be asked to lose weight, since weight loss can help reduce high blood pressure.

Others from the public health arena might contend that the biggest negative financial impacts associated with obesity and smoking have less to do with whether a subject is currently smoking/obese, and more to do with that subject's risk of contracting chronic diseases affecting the heart, lungs, body circulation, and sugar metabolism (diabetes).  Identifying those populations could be tricky if chronic disease risk is related to the length of time a subject has been obese and/or smoking.

Reliable data for analysis of the financial effects of smoking on expensive chronic conditions could be made available if:

(1) A law were passed that required every purchase of cigarettes to be associated with a unique, nationwide, and permanent identifier of the smoker consuming the cigarettes,  

(2) The data entry error rate for these identifiers is very low,

(3) Cigarette purchase data were made publicly available for analysis,

(4) The unique, nationwide person identifier were also required in publicly available (and otherwise deidentified!) patient medical records,

(5) The reliability of the patient medical record data (from a data management perspective) is good, and

(6) The data collected in (1) through (5) is consistently collected, in a manner that guarantees ethical public use, while maintaining patient privacy, throughout the entire analysis period.

This list is a list of very big 'ifs'. We're still a long way from meeting any of the above data management criteria.  Plus, all of the above data management requisites would need to be in effect for at least the number of years established by other research to put a smoking person in a high risk group for later health complications. That number of years could vary from smoker to smoker, as future research-based insights on genetic risk factors for smoking-related complications are gained during the analysis period.

Data on smoking status and obesity (body mass index) are beginning to be collected standardly as a result of many physician practices and hospitals adopting electronic health record (EHR) technology.  This technology is certified for consistent storage of this and other data important for management of population health.  The EHR adoption trend was recently spurred by the availability of "Meaningful Use" government incentives under the HITECH provisions of the American Recovery and Reinvestment Act of 2009.  Along with adoption of EHR technology, healthcare providers must demonstrate they are using the technology to collect and store certain data, including smoking status and body mass index, to qualify for Meaningful Use incentives. Many healthcare organizations have received these incentives.

Smoking status data stored in EHRs for patient care purposes is potentially easier to obtain than similar data derived from cigarette purchase data. However, there is a catch: the smoking status data currently collected via "Meaningful Use" EHRs is simply a coded value with no quantitive information on how long the person has been smoking (http://healthcare.nist.gov/docs/170.302.g_smokingstatus_v1.0.pdf).  A code representing one of the following smoking statuses is entered:

current every day smoker
current some day smoker
former smoker
never smoker
smoker, current status unknown
unknown if ever smoked

Cigarette purchase data would be much more temporally grounded, and potentially more useful than smoking status for the analysis, but only if it can be reliably linked to patient care data. In a Meaningful Use EHR, the smoking status data is already linked.

For obesity, Meaningful Use objectives for data collection include the body mass index.  This is a measure of a patient's obesity at the time its raw data (height and weight) are collected (http://healthcare.nist.gov/docs/170.302.f.2_BMI_v1.1.pdf).  

For analyses where the length of time spent smoking, or in an obese state, are important, the above EHR data may need to be collected over several years, along with other data revealing development of associated chronic diseases.  

Even if we could obtain a representative set of longitudinal data for obese or smoking subjects to perform a Sick Days cost analysis related to chronic disease complications, we would still need to somehow control for other factors, not related to obesity or smoking, that can contribute to those
same complications. A reasonably correct set of factors would need to be identified, and control cohorts for the obese and smoking subjects be defined.


Another approach for controls in the analysis might be to construct additional cohorts of subjects matching subjects in a smoking/obese cohort in key characteristics, but receive a medical intervention known to reduce obesity, or assist the person in quitting smoking.  

Another question for cohort building: Is there a significant fraction of the American population that is obese, and smoking regularly? If that were the case, the obesity vs. smoking question may be even harder to answer with data.  Many people start a lifetime habit of smoking in their teens.  Obesity in younger Americans is on the rise today.  We need to learn whether teen smokers are likely to be obese, and vice versa.     

Compared to your broad question, it seems to be easier to determine the data requirements for answering each of the societal impact questions you list. But is this list sufficiently complete? Many more specific societal impact questions are possible. Some of them will have more relevance to the broad question than others.  If this list were produced as a result of consensus-seeking focus group meetings, involving experts from diverse health care-related fields, who can best determine which financial cost data can best inform your broad question, it might be worth the financial and human investments needed to identify, gather, analyze, and report on the required data.

As for the broad question itself: Under what conditions would a well-informed answer to the question benefit patient care or health policy decisions?  Could the answer be used to justify denial of treatment on financial grounds only?  Obesity and smoking are major health problems.  Providing financial evidence to greatly favor treatment of one problem over the other does not make a lot of sense to me.

Comment by Angela Waner on June 11, 2012 at 3:13pm

I don't believe it is possible to measure if one is "worse" then the other. 

Comment by Vincent Granville on June 11, 2012 at 3:01pm

Hi Rebecca: absolutely agree with you - that's why I mention extra health expenditures attributed to smoking (or obesity or alcohol or whatever). 

Comment by Rebecca T Barber on June 11, 2012 at 12:57pm

I would also point out that, just as not everyone who smokes get's lung cancer, not everyone who is obese has any of the other problems you list.  You not only need to compute the costs, but also the baseline prevalence.  For example, if we say that increased risk of Diabetes is related to obesity (correlationally, not causally, btw) we would need to look at the rate of diabetes in the general population, the rate in the obese population, and could only look at the costs for the percentage above the baseline.  

The same, of course, goes for smoking.  However this is something that seems to almost NEVER be done in these types of studies.  There are relative risks and absolute risks; increasing the relative risk of a disease by 50% sounds bad, but when the absolute risk started at 1% and is now 1.5%, it's hard to get overly concerned.

Comment by QUYEN KIET on June 11, 2012 at 12:35pm

This is a dangerous slippery slope that is requesting more and more intrusive lifestyle information in order to determine a person's value to society. Smoking/obesity today, but what about similarly relevant questions such as:

1. How often one reads the newspaper? informed voters are better for society

2. How far one walks on average? healthier members of society

3. What is the ratio of vegetable to meat ingested per week? healthier members of society

4. How often one goes to church service? social support reduces likelihood of depression

5. How often one disciplines their children? How often one buys toys for their children? How often one talks to their children?

And so on. Where does it end?

On Data Science Central

© 2019   AnalyticBridge.com is a subsidiary and dedicated channel of Data Science Central LLC   Powered by

Badges  |  Report an Issue  |  Privacy Policy  |  Terms of Service