What is Epidemiology?

How likely you are going to change your health-related behavior, if your provider suggests so?

What is your weight status?

How do you prefer to wear your clothes?

Tuesday, March 1, 2011

Doctors Can Influence Patients to Lose Weight. Really?

There is an interesting health news on the MSN website from a recent study reporting findings on how doctors can influence patients to lose weight. [1] The study addressed that people would be more likely to try to lose weight if their doctors told them that they are overweight or obese. 


Based on the information it provided, I searched for the original article for this study. This article published in the Feb. 28 issue of Archives of Internal Medicine. [2] This study used 2005-2008 National Health and Nutrition Examination Survey (NHANES) data on 20-64 years old adults with greater than or equal to Body Mass Index (BMI) of 25.0. Since the data are snapshots without any temporal information, it is a cross-sectional study based on the definition from textbook. [3] The author aimed to find the influence of physician acknowledgment of patient' weight status on patient perceptions of overweight or obesity.


Here, the exposure would be whether the patients reported that their physician told them they were overweight, and the disease would be patient perceptions of overweight or obesity. They included 7790 people aged between 20-64 years from 2005-2008 NHANES data as a sample. Since they narrowed their sample within certain age range, it is a restriction sample. Because it is very hard to sample huge population, they weighted the data, so the 7790 people represent 174 million people.


They did have some confounding variables in their logistic regression model, such as age, sex, race, poverty to income ratio, marital status, education, whether the patient has a routine source of health care, and the number of physician visits in the last 12 months. But I think there might be other confounders, such as whether they have health insurance, and their occupations. I think those would also influence their perception of overweight or obesity. For example, uninsured people might not go to the doctor regularly, so the chance of being told their weight status by their physicians would be lower. Also, some job needs higher muscular-level work which might cause higher BMI (greater than 25), but people might not think they are overweight or obese, even they've been told so. 


In the end of this articles, it points out its limitations: 1. no cause and effect association can be made; 2. no information about the extent of the physician intervention; 3. Using BMI standard to define overweight would be wrong to muscular people; 4. no information about the reason for people to see a doctor.


The authors concluded that patient-reported physician acknowledgment of patients' weight status is highly associated with patient perceptions of overweight and obesity based on the odds ratios for different groups. In this case, physicians need to tell more overweight and obese patients about their weight status so that it could help encourage them to change their behaviors to lose weight and lower their risk for many diseases.


I think this is a good article based on very solid foundation. The conclusion that author addressed is responsible as they know what limitations are. So they address the conclusion very carefully without suggesting any causality relationship. I think it could alarm physicians to pay more attention on telling overweight or obese patients their weight status, since most of patients would listen to their physicians and start to change their behaviors for weight loss. 


References:

1. Amanda Gardner. Doctors Can Influence patient to Lose Weight: Studies. 2011 HealthDay. http://health.msn.com/healthy-living/articlepage.aspx?cp-documentid=100270168

2. Robert E. Post, Arch G. Mainous III, et al. The Influence of Physician Acknowledgment of Patients' Weight Status on Patient Perceptions of Overweight and Obesity in the United States. http://archinte.ama-assn.org/cgi/content/full/171/4/316


3. Gordis L. Epidemiology. 4th ed. Elsevier. 2008. ISBN: 9781416040026

How to design an epidemiology-related study?

Based on what I learned from the textbook, there are several different study designs, such as case-control study, case-cohort study, case-crossover design, and cross-sectional study. [1] let me briefly introduce them one by one. 

Firstly, what is a case-control study? Basically, we need two groups of people, one group of people with the disease (called cases), and the other don't have the disease (called controls). Then we compare these two groups and try to figure out the possible relation between an exposure and a certain disease. The major limitation of this design is recall bias between people who have the disease and try to connect the disease to the exposure intensionally.

Secondly, what is a case-cohort study? In this type of study, we basically select a certain population and follow them over certain time period. This type of study is a good fit for finding risk factors for a certain disease without recall bias, since people were followed from baseline.

Thirdly, what is a case-crossover design? Normally, this is used for studying the acute outcomes with short exposure. Each people in this study represents his/her own control. Recall bias is still the major problem for this type of study.

Fourthly, what is cross-sectional study? The main feature for this type of study is both exposure and disease are examined simultaneously for each people at one time point. This is also called prevalence study, since the determined cases of disease represents the prevalence of this disease at one time point. The major problem for this type of study is missing temporal information. This means we don't have any time series information to know whether exposure happens before the disease or the other way around. So, we could only address the association between exposure and disease, but not the causality. 

Reference:
1. Gordis L. Epidemiology. 4th ed. Elsevier. 2008. ISBN: 9781416040026

Morbidity vs. Mortality

"Morbidity" and "Mortality" are the two most important and basic concepts in epidemiology, because both of them could be used for measuring the occurrence of disease.

Firstly, what is morbidity measurement? Normally, we use incidence rate and prevalence to express the extent of a disease in populations. Here, 

Incidence rate per 1,000 =                                             
  No. of NEW cases of a disease occurring                  
              in the population during                                    
            a specified period of time               * 1,000        
       No. of persons who are at risk of                                        
          developing the disease during
                   that period of time

Prevalence per 1,000 =
 No. of cases of a disease present
in the population at a specified time * 1,000
   No. of persons in the population
           at that specified time

The relationship between incidence and prevalence could be demonstrate further by figure 1. [1] Assume that the flask is a community, and the beads in the flask represent the prevalent cases of a disease in this community. The first pic shows the baseline of prevalence, then the three pics below show three different ways that incidence could affect prevalence.

Figure 1. Relationship between incidence and prevalence.


Secondly, what is mortality measurement? Normally, mortality could be used for measurement of disease severity and how effective a treatment for a diseases is over time. There are two different types of mortality rates:

Annual mortality rate for all causes                                 
(per 1.000 population) =                                                    
  Total # of deaths from all causes in 1 year  * 1000    
  # of persons in the population at midyear                    

Disease-specific/cause-specific rate
(per 1,000 population) =
     # of deaths from a disease in 1 year     * 1000
  # of persons in the population at midyear

Case-Fatality rate (percent) =                                       
    # of individuals hying during a                                     
specified period of time after disease                                      
             onset or diagnosis                          * 100        
# of individuals with the specified disease                      
                                                                                             
Proportionate mortality (percent) =
   # of deaths from certain disease
             in a population during
              certain time period              * 100
  Total # of deaths in the population
      during the same time period

Those are some basic concepts and formulas on how to measure the occurrence of a disease that I learned from the textbook. After knowing those formulas, I think you would have more sense on how epidemiological event is measured.

Reference:
1. Gordis L. Epidemiology. 4th ed. Elsevier. 2008. ISBN: 9781416040026