docx (54)

Chapter 13 Research and Data Analysis Valerie Watzlaf, PhD, RHIA, FAHIMA Real-World Case 13.1 Researchers were interested in assessing the relationship between obesity and breast cancer recurrence and fatality in postmenopausal African-American and Caucasian women with primary breast cancer. Data was collected on women with primary breast cancer and included the following variables: ? Age ? Age at diagnosis of breast cancer ? Weight ? Height ? Data of diagnosis of breast cancer ? Menopausal status ? Diagnosis and coding of tumor (histopathology and topography) ? Stage of tumor ? Size of tumor ? Number of positive lymph nodes ? Estrogen receptor analysis ? Progesterone receptor analysis ? Site of distant metastasis ? First course of treatment (surgery, radiation, chemotherapy) ? Additional treatment ? Five-year recurrence and survival rates Recurrence and survival status were determined by reviewing the cancer registry follow-up data and medical record information across the multiple healthcare sites involved in this study. The cancer registries were accredited by the American College of Surgeons and used active follow-up on all cancer patients. Postmenopausal status was determined as subjects older than 55 years. In those subjects younger than age 55, determination was made by consulting the cancer registry data, the medical record, and physician’s office records. Premenopausal patients and patients whose menopausal status could not be determined from the data were excluded from the study. Body mass index (BMI) was based on height and weight collected from the medical record or cancer registry at the date of diagnosis only. Values greater than 27 were considered to indicate obesity. The effect of weight changes during the follow-up period was not evaluated. Real-World Case Discussion Questions 1. How was obesity determined in this group of patients? In this group of patient’s obesity was determined by several factors that were pulled not only from the patient’s medical records, but the cancer registry. The body mass index (BMI) was based on height and weight collected from the patient’s medical record or the cancer registry at the date of diagnosis only. Any values greater than 27 were considered obesity. 2. Why is it important to collect the other variables when studying the relationship between obesity and breast cancer recurrence? The reason it is important to collect the other variables when studying the relationship between obesity and breast cancer recurrence is because excess body weight has been known to be the link to increase the risk of premenopausal breast cancer, and evidence of obesity is associated with women diagnosed with early stage breast cancer. 3. Postmenopausal status was determined as subjects older than 55 years. What may be a better way to determine whether the women in this study were postmenopausal? A better way to determine if these subjects were post-menopausal would be to check their hormone levels 4. What type of study design is this? If it was found that obesity and breast cancer recurrence were related what type of study could be conducted to test whether changes in diet may make a difference in the recurrence of breast cancer in women who have had primary breast cancer? A cohort study or observational study. There is a study that they are doing called a “Breast Cancer Weight Loss (BWEL) Study”. This is the study that can be conducted to test whether changes in diet may make a difference in the recurrence of breast cancer in women who have had primary breast cancer. 5. This study did not examine pre-menopausal women. It did not look at other races besides African American. Is this a limitation? Why or why not? There is no limitation, studies are based on self-report and is a classic measure for race and is still the most reliable source and also has been the subject of much debate. Real-World Case 13.2 A descriptive research study was performed to investigate the completeness of the ICD-10-CM coding system in capturing public health diseases when compared to ICD-9-CM. In order to do this, the infectious and reportable public health conditions—such as avian flu, smallpox, anthrax, and such—were examined first by reviewing each state department of health’s website to determine which diseases are required to be reported. Once this list was developed, it was supplemented with the CDC national reportable disease listing. The final list of public health reportable infectious diseases included all the reportable infectious diseases by state as well as those required by CDC. This list was supplemented with two other areas that are very pertinent to public health—the top 10 causes of mortality: ? Accidents ? Alzheimer’s disease ? Cerebrovascular disease ? Diabetes mellitus ? Influenza ? Lower respiratory disease ? Nephritis ? Septicemia ? Heart disease ? The top five malignant neoplasms And the classification of death and injury resulting from terrorism list, including 10 major categories as follows: ? Terrorism involving explosion of marine weapons ? Destruction of aircraft ? Other explosions and fragments ? Fires ? Firearms ? Nuclear weapons ? Biological weapons ? Chemical weapons ? Terrorism other specified ? Sequelae of terrorism A total of 248 public health disease categories were developed. When coding the diseases, several more codes and descriptions were listed so that the number of codes far exceeded the 248 disease categories. A website was then developed so that all of the public health diseases and descriptions could be easily accessed by the researchers and the focus group members. For example, when organizing the reportable disease list on the website, every disease was categorized alphabetically. When the specific reportable disease was accessed, a spreadsheet with each of the ICD-9-CM and ICD-10-CM codes could be easily viewed. This was extremely useful for reviewing the codes, rankings, explanations for using a specific ranking, and so forth. Although the list of 248 disease categories is not exhaustive of all public health diseases, it was believed to provide an adequate number to make comparisons between the two coding systems. The 248 public health diseases were then coded using both ICD-9-CM and ICD-10-CM so that comparisons between the two coding systems could be made. The research coder for this study has a master’s of science degree in information science and is a registered health information administrator (RHIA) and has taught coding for more than 20 years. She was also trained and educated on the ICD-10-CM coding system through AHIMA’s online ICD-10-CM coding seminars. The research assistant, who performed data entry and assisted in some of the ICD-10-CM coding, has a master’s of science degree in health information systems and was also trained and educated on the ICD-10-CM coding system. All final codes were approved by a research coder. Quality checks for final codes were performed by a secondary investigator, who has a doctorate in public health and is an RHIA and certified coding specialist (CCS); and also by the principal investigator, who has a doctorate in epidemiology and is an RHIA. Comparison tables that describe the specificity of the coding for ICD-9-CM and ICD-10-CM for each of the public health diseases were developed. A ranked score was assigned to each public health disease for both the ICD-10-CM and ICD-9-CM coding systems. The ranking was determined by comparing the number of codes, level of specificity, and ability of the code description to fully capture the diagnostic term. The ranked or ordinal scale consisted of the following: 5 = Diagnosis is fully captured by the code(s) (all codes, specificity, description is found) 4 = Diagnosis is almost fully captured by the code(s) (minor detail is missing) 3 = Diagnosis is partially captured by the code(s) (moderate detail is missing) 2 = Diagnosis is less than partially captured by the code(s) (major detail is missing) 1 = Diagnosis is not captured by the code(s) (codes, specificity, description is not found) The ranking scale was developed by the research team and reviewed and approved by the focus group members. All assigned rankings were also reviewed and approved by the research team and by all focus group members. Researchers do acknowledge that there was some subjectivity involved in the assignment of the rankings. Once all rankings were assigned, a focus group that included seven experts in ICD-9-CM, ICD-10-CM, and public health convened. Two of the focus group members have medical degrees, two are working on their doctorates in public health and have extensive education and training in coding, and three have coding credentials and have worked in the coding field for more than 10 years. The purpose of the focus group was to review and examine the information accumulated from the study and provide feedback and recommendations regarding where changes need to be made in the ICD-10-CM system. Therefore, the focus group examined the rankings and made changes. The researchers reviewed and discussed all comments from the focus group, clarifying any questions, and then made the appropriate changes to the rankings and code descriptions. In the analysis of all the public health diseases, such as reportable diseases (p < 0.001), top 10 causes of death (p < 0.001), and those related to terrorism (p < 0.001), it was found that the overall rankings for disease capture for ICD-10-CM were significantly higher than the rankings for ICD-9-CM. In this example the p value is a statistic that demonstrates statistical significance. It is computed by running statistical tests to determine if the differences between ICD-9-CM and ICD-10-CM rankings were real or due to chance. If the p value is less than 0.05, the differences seen are not due to chance and it demonstrates that what was found in this study is real. (Watzlaf et al. 2007). Watzlaf V.J.M., J.H. Garvin, S. Moeini, and P. Anania-Firouzan. 2007. The effectiveness of ICD-10-CM in capturing public health diseases. Perspectives in Health Information Management. 4(6).World Health Organization. 2015. http://www.who.int/en/. Real-World Case Discussion Questions 1. A descriptive research study was performed to investigate the completeness of the ICD-10-CM coding system in capturing public health diseases when compared to ICD-9-CM. In order to do this, the infectious and reportable public health conditions—such as avian flu, smallpox, anthrax, and such—were examined first. Why would it be important to examine infectious and reportable public health conditions when examining ICD-10 and ICD-9 classification systems? It is important to examine infections and reportable public health conditions when examining ICD-10 and ICD-9 classification systems to assure accuracy in capturing and reporting of public health diseases. After this review it was found that the overall rankings for disease capture for ICD-10-CM were significantly higher than the rankings for ICD-9-CM 2. Did the research team for this study have appropriate experience to conduct and participate in the study? Why or why not? I believe that the research team for this study did indeed have an appropriate experience to conduct and participate in the study. Overall results demonstrate that ICD-10-CM is more specific and fully captures more of the public health related diseases than ICD-9-CM. In the analysis of all the public health related diseases such as reportable diseases, top ten causes of death, and those related to terrorism, it was found that the overall rankings for disease capture for ICD-10-CM were significantly higher for than the rankings for ICD-9-CM 3. A ranked score was assigned to each public health disease for both the ICD-10-CM and ICD-9-CM coding systems. What did this ranked scoring system look like and do you believe that is was an objective scoring system? If not, explain why. I have my doubts about this system because some of the diseases are not captured by either system. National Health Reportable Diseases ICD-9-CM Rank ICD-10-CM Rank Explanation Chickenpox 3 5 I-10 has more codes/descriptions Ciguatera Fish Poisoning 2 5 I-10 has more codes/descriptions Gonococcal infection* 4 5 I-10 has more codes and codes related to pregnancy complications Hepatitis A acute/chronic* 4 5 Captured by both with similar descriptions but I-10 has an “other” category Hepatitis B acute/chronic* 5 4 Captured by both systems but I- 9 more specific for some categories Herpes, neonatal or genital 3 5 I-10 much more specific 4. What final results and conclusions can be made from this study? The final result and conclusion that can be made from this study is there are still improvements that need to be made before the change-over. This is needed to make a smooth transition. 5. What does the p value mean in the results section of this study? P values determined by Mann Whitney U non parameter test. Application Exercises Instructions: Answer the following questions. Construct a pie graph showing the following percentages for coding productivity among coders at Facility Y. Coder Number and Percentage of Charts Coded Per Day 1 15 14.42308 2 25 24.03846 3 12 11.53846 4 22 21.15385 5 30 28.84615 Total 104 2. State whether the following variables are nominal, ordinal, interval or ratio. Variable Type (nominal, ordinal, interval or ratio) Gender NOMINAL Race NOMINAL Patient Satisfaction Scores ORDINAL Temperature INTERVAL Weight RATIO Height RATIO Blood pressure INTERVAL 3. What trends do you see from Table 1, below? How would you summarize what the table is showing? Table 1 Demographic Characteristics of Respondents (N = 12) Characteristic Response Mean SD Age (years) 54.67 12.71 Years of experience 23.42 12.48 No. Percentage Gender Male 9 75.0 Female 3 25.0 Setting Hospital (or other facility) only 5 41.6 Private practice only 2 16.7 Both 5 41.6 Medical specialty Emergency medicine 2 16.7 Ophthalmology 1 8.3 Internal medicine/geriatrics 1 8.3 Plastic/reconstructive surgery 1 8.3 General surgery 1 8.3 Obstetrics/gynecology 1 8.3 Psychiatry 2 16.7 Family medicine 1 8.3 Hematology/oncology 1 8.3 Physical medicine 1 8.3 Previous use of electronic health records Yes 10 83.3 No 2 16.7 Exposure to ICD-10-CM/PCS Yes 3 25.0 No 9 75.0 Review Quiz Instructions: For each item, complete the statement correctly or choose the most appropriate answer. 1. Any numerical value that goes from one whole number to the next whole number, such as $30,567.32, 62.596, and 18.65 is called what kind of variable? a. Discrete b. Continuous c. Simple d. Ordinal 2. When a number is assigned to a specific category such as a 1= male and 2 = female, this is what type of variable? a. Ordinal b. Interval c. Ratio d. Nominal 3. The key to building what statistical tool is to make it able to stand alone so anyone reading it can understand the information displayed? a. Graph b. Plot c. Table d. Measure of central tendency 4. Which is a simple chart used to describe qualitative, categorical, or discrete variables such as nominal or ordinal data? a. Line graph b. Pie chart c. Frequency polygon d. Bar chart 5. This chart is an example of which of the following? a. Pie chart b. Horizontal stacked bar chart c. One-variable bar chart d. Histogram 6. Which chart is similar in appearance to a bar chart but the highest ranking value is listed as the first column and the next highest ranking is second, and so on, to the lowest ranking? a. Pie chart b. Pareto chart c. Vilfredo chart d. Histogram 7. Refer to the pie chart below. Which locality has the greatest percentage of healthcare facilities in Region 1? a. Rural b. Suburban c. Urban d. It is unable to be determined from this graph 8. True or false. A line graph is a graphical device used to display continuous data and to show changes or trends of the data over time. a. True b. False 9. True or false. A histogram should be used with discrete data that is part of a frequency distribution. a. True b. False 10. True or false. Frequency polygons differ from line graphs in that frequency polygons display the entire frequency distribution (counts) of the continuous variable; a line graph plots only the specific data points over time. a. True b. False 11. True or false. A scatter chart, scatter plot, scatter diagram, or scatter graph is used to demonstrate a relationship between two variables. a. True b. False 12. True or false. A bubble chart is similar to a scatter chart except that it compares three data variables. a. True b. False Rationale: A bubble chart is similar to a scatter chart except that it compares three data variables. The larger the bubble the larger the value of the variable. 13. Coder Status as 1=advanced, 2=intermediate, 3=beginner is an example of a__________ in common data configurations of statistical software/ a. Data list b. Missing value c. Output d. Value label 14. Based on this output table, what is the average coding test score for the beginner coder? Coding test score Coder status Mean N Standard Deviation advanced 93.0000 3 5.00000 intermediate 89.5000 2 .70711 beginner 73.3333 3 6.42910 Total 84.7500 8 10.51190 a. 93 b. 6.4 c. 73 d. 90 15. What is the mean length of stay for the following data set: 10, 12, 5, 6, 7, 9, 5 a. 7.7 b. 9.7 c. 8.6 d. 7.9 16. The first step to compute the _________ will include ranking the values from lowest to highest. a. Mean b. Median c. Mode d. Standard deviation 17. What is the mode in this array of the following variables? 1,2,3,4,5,6,7,7,7,8,9 a. 2 b. 4 c. 7 d. There is no mode 18. Which statistics examine the spread of different values around the measures of central tendency? a. Measures of variability b. Measures of significance c. The range d. The variance 19. Which measure of variability is the simplest one to compute? a. Standard deviation b. Variance c. Range d. Mode 20. In which type of distribution are the mean, median, and mode equal? a. Bimodal distribution b. Simple distribution c. Non-normal distribution d. Normal distribution 21. What is a standardized unit that provides the relative position of any observation in the distribution and is also the number of standard deviations that the observed value lies away from the mean, ?? a. T-score b. Z-score c. F-test d. Chi test 22. A Z-score of 3.5 represents a score that is how many standard deviations above the mean? a. 3.5 b. 6.5 c. 68 d. 3.5 below the mean 23. What type of study design was used to explore how the use of automated coding software (computer-assisted coding [CAC]) could be used to enhance anti-fraud activities? a. Retrospective b. Prospective c. Experimental d. Descriptive 24. If a researcher was interested in whether the use of estrogen replacement therapy (ERT) caused colon cancer in postmenopausal women, they would recruit a group of women with colon cancer (cases) and another group of women without colon cancer (controls) as subjects. What is this type of study design called? a. Prospective cohort b. Historical prospective c. Descriptive d. Retrospective case-control 25. This particular process is important in effectively testing whether the specific intervention actually made a difference in the outcome of the disease in experimental studies. It is called which of the following? a. Odds ratio b. Relative risk c. Randomization d. Incidence rate