Benchmarking & Clinician Profiles

Assigned Reading

Use data balancing to benchmark clinicians (use instructor's last name as password) Read►

Presentations

• Session overview YouTube►
• Use data balancing to benchmark clinicians Slides► YouTube►

Assignments

Start with a summary page.  Start each question in a separate page, sheet, or file. In the summary page write statements comparing your work to answers given or videos.  For example, "I got the same answers as the Teach One video for question 1." 

Question 1: In the following question, use SQL to analyze the data.  Assume that we have followed two clinicians, Smith and Jones, and constructed the decision trees in Figure 1. Data► Atoosa's Excel► Pooja's Video► SQL Code►

Figure 1:  Practice Patterns of Dr. Jones and Smith

• What is the expected length of stay for each of the clinicians?
• What is the expected length of stay for Dr. Smith if he were to take care of patients of Dr. Jones?
• What is the expected length of stay for Dr. Jones if he were to take of patients of Dr. Smith?

Question 2:  The following data report patients with 10 Diagnostic Related Groups and 3 HCC indices cared for by a clinician and his peer group.  Data►

•  Regress "Cared for by Dr. Smith" on the HCC and other patient charcteristics.  What type of patients are more likely to be cared for by Dr. Smith.  Regression Results►
• Using regression results, identify which variables are likely to be in the Markov Blanket of the variable "Cared for by Dr. Smith".
•  Use SQL to determine if the clinician is more efficient than his peer group. SQL► Mai's Teach One►

Question 3: The following data show the variation in diabetes in select counties across United States.  Using stratified covariate balancing report the impact of access to supermarkets on diabetes after controlling for other variables. Data►

1. Check that all variables are positively and monotonely related to prevalence of diabetes in the county. Monotone?►
2. Assign a binary variable to each variable in such a manner that when the variable is 1, diabetes is more likely.
3. Drop from analysis covariates that are not parents on Markov Blanket of diabetes.  Accomplish this task using the following steps:
   • Regress diabetes on all variables (with no interaction terms in the model), identify variables that are signficant predictors of diabetes and have a large effect size
   • Do a second regression, verifying that no interaction terms that involve the signficant variables are predictors of diabetes (have a statistically signficant and large effect size). Include on the list of parents of Markov Blanket, any variable whose interactions is predictive of diabetes.
4. Calculate the impact of access to food sources on diabetes, while controlling for other variables.   Accomplish this task by stratifying the variables identified as parents in the Markov Blanket, then switch the distribution of controls (low-diabetic counties) with distribution of cases (high diabetic counties).
5. Report overlap and impact of food access on diabetes.

Question 4:  In synthetic cases we predict the probability of an event from the relevant data.  In this question you are asked to create a synthetic case where insufficient data exists.  You are asked to predict 6-month mortality rate for 80 year nursing home residents with walking and toileting disabilities but no other disabilities.  There are not 30 such cases in the data.  You are asked to predict what this value would have been based on 2 nearest data points. 

• Verify that there are less than 30 cases in the database of 80 year olds who only cannot walk and toilet. 
• Verify that all variables lead to increased mortality. If not, re-name the variables so when the variable is assigned the value of 1, it has higher probability of mortality.
  • Excessive Match:  Add one additional disability to the list until you find 30 cases.  These additional cases are excessive matches in the sense that these cases have additional disabilities that increase probability of mortality. For example, when you add unable to urine to 1 then you are including cases that also have urine incontinence. For another example, setting unable to eat to 1 will examine patients who cannot walk, toilet or eat.  For each excessive match calculate the number of cases and the probability of being dead in 6 months.
  • Partial Match:  Remove one of the disabilities of the patient and calculate number of cases and average probability of mortality in 6 months.  In this example we are looking at 80 year old who cannot walk or toilet.  Remove toilet and see if you get 30 cases.  Calculate number of cases you can match.  Calculate probability of mortality of these cases in 6 months.
• Calculate the probability of mortality for the synthetic case as average of the minimum of excessive matches and maximum of partial matches.
• SQL►

Use the following dictionary of variables to create a header for the data. Data► Adel's Teach One►

Question 5:  The following data show the recovery from various disabilities in two nursing homes.  Two sets of data are presented.  The first set shows the disabilities of the patients at admission to the nursing home, using variables that start with "u", standing for "unable".  The recovery from the disabilities is also shown in variables that start with "r".  Compare the performance of these two nursing homes using distribution switch method.  In particular, switch the distribution for age, gender, and 9 disabilities on admission.   The outcome of interest is the number of disabilities recovered from (variable shown as nRecovery).  Use synthetic method to estimate outcome for cases not present in both nursing homes. Which nursing home has better outcome for its own residents?  What happens if residents at nursing home A were cared for at nursing home B, which nursing home would have better outcomes?  What will happen if the reverse happens?  Data► SQL & Answer►

Question 6: The following data provide the length of stay of patients seen by Dr. Smith (Variable Dr. Smith=1) and his peer group (variable Dr. Smith = 0).  Balance the data by propensity to seek care from Dr. Smith.  This involve first predicting probability of a patient type utilizing services of Dr. Smith; then weighting the data inversely proportional to the probability of using Dr. Smith.  Note that patients cared for by Dr. Smith and by his peer group will have a different set of weights.  The net results of weights is that patients cared for by Dr. Smith and his peer will have the same rate of various diseases.  (a) Graphically show that the weighting procedure you followed results in same set of patients treated by either Dr. Smith or his peer.  (b) Report the un-confounded impact of Dr. Smith on length of stay. 

• Data►
• Lina's Solution►
• Ayesha Bibi's Teach One Slides► Python► HTML►

More

1. Practice profiling PubMed► 
2. Importance of risk adjustment in measuring performance in primary care PubMed►