Data Analytics 101: Understanding and Using Data

Transcription

1 Data Analytics 101: Understanding and Using Data Sponsored by 1915 N. Fine Ave #104 Fresno CA Phone: (559) Fax: (559) Program Handouts Wednesday, June 8 Track Two 10:25 am - 11:25 am 2016 CHIA Convention & Exhibit Speaker Deborah Neville, RHIA, CCS-P California Health Information Association, AHIMA Affiliate

2 California Health Information Association California Health Information Association Data Analytics 101: Understanding and Using Data June 8, 2016 Deborah Neville, RHIA, CCS P Director Revenue Cycle, Coding & Compliance Elsevier d.neville@elsevier.com ELSEVIER Elsevier, Inc. 245 Peachtree Center, Avenue NE, Suite 1900 Atlanta, GA Disclaimer This material is designed and provided to communicate information about clinical documentation, coding, and compliance in an educational format and manner. The author is not providing or offering legal advice but, rather, practical and useful information and tools to achieve compliant results in the area of clinical documentation, data quality, and coding. Every reasonable effort has been taken to ensure that the educational information provided is accurate and useful. Applying best practice solutions and achieving results will vary in each hospital/facility and clinical situation. California Health Information Association, AHIMA Affiliate 1

3 Goals/Objectives Learn to differentiate between data sources to identify and select pertinent data elements to generate reportable information based on user need. Understand how to compile and manipulate data in preparation for data analysis. Recognize different methods of data presentation using graphs, tables, spreadsheets, or other methods and types of analysis that are most aligned with the desired presentation. Reports Using Reported Data. There are nearly 1 million Americans who visit the emergency room each year because of dental pain at a cost that runs into the hundreds of millions. Among more than 4,000 patients in the study, scans for uncertain reasons happened more than 83 percent of the time, while scans for inappropriate reasons happened 11 percent of the time. Only slightly over 5 percent of these patients were screened for appropriate reasons, the study found. California Health Information Association, AHIMA Affiliate 2

4 Vs. Reports From Your Data?. There are nearly 100 patients discharged from this hospital who visited the emergency department each year because of COPD at a cost that runs into the hundreds of thousands. The study sample selected all patients who were listed on the hospital s surgical elective schedule during the research period. Finally, a total of 297 patients agreed to participate in the study the incidence of immediate post operative and thirtyminute later pressure ulcers was 9.8% and 5.1%, respectively. The Value of Data Treatment Electronic Information Savings Data presentation helps users to compile information to improve patient outcomes, develop best practices, prevent adverse reactions, and ensure continuity of care. Electronic health records provide numerous sources of information one true source for each type of data should be agreed on. Data compiled from a variety of sources will present information for strategic planning, cost containment, and identification of efficiencies. California Health Information Association, AHIMA Affiliate 3

5 What is Quality Data? In its updated Data Quality Model, the American Health Information Management Association (AHIMA) revised its Data Quality Standards, listing the following ten elements of data quality: Accuracy Accessibility Comprehensiveness Consistency Currency Definition Granularity Precision Relevancy Timeliness Source: Data Quality Management Model (Updated). AHIMA 2015 What is Quality Data? Data Accuracy: data free of identifiable errors. In coding, this can be such things as making sure characters are not dropped such as a leading zero. Data Accessibility: The level of ease and efficiency at which data are legally obtainable, within a well protected and controlled environment Data Comprehensiveness: The extent to which all required data within the entire scope are collected, documenting intended exclusions Data Consistency: The extent to which the healthcare data are reliable, identical, and reproducible by different users across applications Data Currency: The extent to which data are up to date; a datum value is up to date if it is current for a specific point in time, and it is outdated if it was current at a preceding time but incorrect at a later time. Important when wanting to compare data of different time periods. California Health Information Association, AHIMA Affiliate 4

6 What is Quality Data? Data Definition: The specific meaning of a healthcare related data element Data Granularity: The level of detail at which the attributes and characteristics of data quality in healthcare data are defined. How specific can one get? Data Precision: The degree to which measures support their purpose, and/or the closeness of two or more measures to each other Data Relevancy: The extent to which healthcare related data are useful for the purposes for which they were collected. For example, if something is only collected some of the time, it is not very useful to the denominator. Data Timeliness: The availability of up to date data within the useful, operative, or indicated time Process for Converting Data into Useful Information California Health Information Association, AHIMA Affiliate 5

7 Defining the Problem and Selecting Data The first step is selecting the correct data. In order to select the correct data the user must be able to: Identify the problem Frame the question to be answered Know what data are available to help answer the question or investigate the problem Defining the Problem and Selecting Data The nature of the problem or question being asked will determine which data source(s) are most likely to contain the answer. Knowing the scope of problem or the purpose for asking the question will help to set the parameters (what and who to include or exclude) and whether a focused or random sample will suffice, or if all of the cases must be included. California Health Information Association, AHIMA Affiliate 6

8 Types of Data Administrative Financial Clinical Data Sources: Definition A data source simply describes the location or setting from which the data is generated. Four common activities are considered when evaluating hospital data. They are: Registration or patient access Clinical assessment Treatment or other services Billing and reimbursement Billing/Claims & Payment Treatments & Therapies Registration Clinical Documentation California Health Information Association, AHIMA Affiliate 7

9 Each Source Generates Unique Data (examples) Registration/Patient Access Patient demographics insurance information MPI Clinical Documentation Medical necessity criteria Patient outcomes and best practices Adverse reactions, complications Protocols Public health disease profiling Quality reporting Each Source Generates Unique Data (examples) Treatments and Therapies Protocols Effectiveness & efficiency of medical vs, surgical intervention Patient response and compliance Billing/Claims and Payment Quality measure reporting Services & supplies Revenue (coverage and contracting) Cash flow and accounts receivable Case mix index MS DRGs vs. AP DRGs, complications and comorbidities, SOI and ROM California Health Information Association, AHIMA Affiliate 8

10 Compiling and Manipulating Data Compiling data simply means to gather the data that are considered pertinent and necessary to provide the desired result from the appropriate source. Data must often be manipulated so that the data provide the desired result. Manipulation means applying actions to the data, such as using formulas; including or excluding data; using various presentation methods, such as pivot tables, etc. Compiling and Manipulating Data Example: A productivity report is needed to compare Dr. A to Dr. B. Some data elements to consider are diagnosis and/or procedure codes; time period; account numbers; medical record numbers; number and type of visits; costs and revenue associated with visits; or other data elements. California Health Information Association, AHIMA Affiliate 9

11 Manipulating Data Mathematical knowledge and formulas are important when determining what is needed, such as: What is the total? (count, sum) How often? (frequency distribution, percentages) What s normal? (averages, standard deviation) How do I compare? (common healthcare statistics) Mathematical formulas and calculations must be correct. If using a spreadsheet or table with numbers, when copying formulas, make sure they are still calculating the correct information from the correct columns. Example of Compiling & Analyzing Data California Health Information Association, AHIMA Affiliate 10

12 Data Analysis and Spreadsheets Data analysis is the process of: organizing, presenting, and calculating data in an understandable way so that an end user can evaluate data points which result in usable, meaningful information. Are you getting the results you expect to see? Is what you are reporting accurate? Data Analysis and Spreadsheets: Results Are you getting what you expect to see and is what you are reporting accurate? In this previous example, if you didn t know the number of patients, how useful would this bar graph be? California Health Information Association, AHIMA Affiliate 11

13 Data Analysis and Spreadsheets: Some Questions What else might be missing to make this more useful? How about a time period? If codes were used to select this data, which ones were they? Should you list the provider s name or some type of other identifier that is less recognizable? Did you compare the correct providers? Some may have similar names. Did you include all providers? What setting(s) was this data taken from? (e.g. the office, discharges, any patient with a diagnosis of asthma or those with a principal diagnosis) Data Analysis: Trends, Patterns, Events That Affect Data Seasonal effects (senior citizen s annual migration to Florida, flu season, influx of tourists during summer or during ski season, or winter break) Services (adding a new service or discontinuing one, new equipment, opening a new clinic or satellite site) Staffing Issues (new batch of residents start in July, vacations, layoffs, strikes, shortages for other reasons) Construction/Renovations (wings or units closed for remodeling, beds added or closed) Disasters (hurricanes, major floods, blizzard of the century, 9/11, epidemic) Other (poor ranking in a major journal, new managed care contract with major employer, major revision to the DRGs, Relative Weights (RW) or payment system, closing or opening of a major employer.) California Health Information Association, AHIMA Affiliate 12

14 Data Analysis and Spreadsheets: Uses Spreadsheets are presented as tables and commonly used for calculations. For instance: Cost per unit Accounting Calculation of interest rates and return on investment Case Mix Analysis Human Resources salary and bonus calculations Creation of dashboards to summarize data analysis Audits case selection and results tracking Productivity tracking and reporting Spreadsheet Tools Operation Tool Rationale Rearrange the data so that the cases appear in discharge date order instead of medical record number order. Sort Review only the female patients in a group of cases. Examine data to find cases that are unusual, compared to the rest of the data. Obtain a quick snapshot of the statistics that describe the total charges for cases in DRG 234. Filter Pivot Table Data Analysis Organizes data in the order specified by the user Allow the user to review rows containing only the selected values in a particular column Count occurrences of specific observations, so unusual volumes of occurrences stand out. Data Analysis is a module in Excel that provides menudriven statistical analysis choices. California Health Information Association, AHIMA Affiliate 13

15 Line Graphs Line graphs: easy to view trended observations, e.g. the number of injuries over time injuries May June July Aug Sep Oct Nov injuries Bar Graphs Bar graphs often communicate comparisons of frequencies, counts, or amounts California Health Information Association, AHIMA Affiliate 14

16 Pivot Table Pivot Table 30 Base DRG Distribution for Each MDC Number of Base DRGs PRE Surgical Medical MDCs Surgical Medical California Health Information Association, AHIMA Affiliate 15

17 California Health Information Association California Health Information Association Measuring Case Mix 32 Example: Measuring Case Mix Determine your normal (baseline) case mix: Descriptive reports that help you to determine baseline are: Summaries with total counts, means, standard deviations Rankings with position number, totals, means, standard deviations Profiles with ranges (minimum and maximum values) modes, medians, percentile breakdowns, and summary statistics Case listings sorted on specific variables so that like records appear together for pattern identification. California Health Information Association, AHIMA Affiliate 16

18 Measuring Case Mix Define groups by deciding what you want to count (e.g. cases, days, dollars, etc.) by what (e.g. the By Group variables) for what, who and/or when (e.g. the For variables). Organize groups from broadest categories to specific (i.e. How many surgical cases with ALOS> High trim point (HITRIM) by MD for 2015). Countables By Group For Discharges/Visits Time period Time period Days Service/Procedure/Test Service/Procedure/Test Dollars MD MD Minutes/hour Nursing unit Nursing unit Relative weights DRG DRG Scores Quality measure MDC Payer Payer Gender Gender Diagnosis/Procedure code Diagnosis/Procedure code Age/ Age group Age/ Age group Discharge/Admit status Discharge/Admit status Cost center Cost center Revenue code Revenue code Accommodation code Accommodation code Zip code Zip code LOS LOS 34 Measuring Case Mix Manipulating Data Some ways to measure (quantify) case mix: Total # of cases (by week, month, quarter, year, MD, service, diagnosis, procedure, payor, etc.) Average daily census, number of visits or encounters Total/Average Relative Weight (DRGs, APCs, APGs) Total/Average LOS (by service, type of accommodation) Total/Average Charge/Cost (DRGs, APCs, APGs) Total/Average Payment (by payor, or payment type) California Health Information Association, AHIMA Affiliate 17

19 Measuring Case Mix Manipulating Data Manipulating data refers to the process of doing something, usually performing statistical functions, with the data. Some of the more common statistical functions performed are: Counts The number of observations or occurrences (total) in a category or group (i.e. year, DRG, MD, Service, Procedure code, etc.) Mean Average value of a measurable attribute (days, cost, charges, SIWs, etc.) for observations in a category. Arithmetic Mean: Total of all the values of the measurable attribute in a category or group and divide the total by the number of observations in the category or group. Geometric Mean: The nth root of the product of all values of the measurable attribute in a group or category with n observations. The arithmetic mean is the more widely used of the two averaging methods. However, the geometric mean is the better choice if you are working with any of the following: Untrimmed data with very divergent outliers Averaging percentages or indices rather than number of occurrences or objects Measuring the average rate of increase from one period to another 35 Measuring Case Mix Manipulating Data 36 Median The middle value in a distribution of observations above and below which lie an equal number of observations. If there are an even number of observations then the median is the arithmetic mean of the values of the two middle observations. The median is a quick way of getting a sense of the skew of a range of values in a data set. It can sometimes be used as a standard or goal(i.e. optimal LOS, cost, response time, etc.) for CQI initiatives. Mode The value which occurs most frequently in a group or category of observations. If you determine that the mode occurs frequently enough you can sometimes use it as a standard or goal (i.e. most frequent # of charts coded per day by coder A for the quarter) for CQI initiatives. California Health Information Association, AHIMA Affiliate 18

20 Measuring Case Mix Manipulating Data 37 Percentages A portion or share in relation to a whole. Percentages are an easily understandable way of indicating how much of all the observations is represented by a subgroup or portion. Percentages make it possible to compare data from different sized populations. Converting from straight frequency counts to percentages puts both populations on a scale of 1 to 100. Percentile A number that divides a range of values (from a group or category) into 100 divisions or cells (1 st percentile 100 th percentile). Every value in the group or category falls into one of the cells and can be compared to the other values in the group or category on a percentage basis (e.g. A s LOS was in the 92 nd percentile for the group) Ratios The rate or frequency at which two things occur in relation to each other. Knowing the frequency at which an event occurs in a population (i.e. 20 out of every 100 deliveries is a C section) enables you to set standards for comparison and make predictions. Measuring Case Mix Manipulating Data 38 Variance The difference between what was expected and what actually occurred. Or, the difference between two sets of values Accounting The arithmetic difference between two values. Statistical The mean of the squared differences between the value of each observation and the sample mean. Calculating variances (accounting style) is a very easy way to detect changes or deviations from what was expected and what actually happened. Standard Deviation The square root of the arithmetic mean of the squared differences (deviations) between the mean value and each of the actual values in a group Standard deviation indicates how closely related or cohesive groups of values are to each other. The higher the standard deviation, the more widely distributed the values are around the mean. Given a normal distribution of values (aka a bell curve ) about 68% of the values in the data set will occur within one standard deviation of the arithmetic mean of the group and 96% will occur within two standard deviations of the mean. California Health Information Association, AHIMA Affiliate 19

21 39 Measuring Case Mix Case Mix Index The most common way to measure case mix is by using the relative weights, which are and published for every DRG (and APC/APG): Relative weights are an accepted industry standard for identifying the expected resource use at the case level It is easy to assign a DRG (or APC/APG) to every case and get a relative weight on each one that is comparable across all patients. Once each case has a relative weight it is a simple calculation to total up all the cases relative weights and divide by the total number of cases to come up with an average relative weight for the facility This average relative weight is commonly referred to as the Case Mix Index or CMI. 40 Calculate A Case Mix Index California Health Information Association, AHIMA Affiliate 20

22 41 Monitoring: What to Watch Over Time Routinely run the same reports and compare the results to previous reports Identify changes from one set of reports to the other Calculate variances to quantify differences (changes). Some useful characteristics to compare and look for changes in are: Volume of cases by DRG, service, MD, diagnosis, or procedure Length of stay by the above Costs/charges/payments Admission/discharge patterns Comparative Analyses Comparative analyses require at least two sets of observations so that comparisons can be made. Make sure you are comparing Mcintosh apples to Mcintosh apples and not McIntosh apples to Gala apples or some other type of apples or fruit (or KNOW the differences and take them into account) 42 Array the data so that different sets of data can be viewed simultaneously. Comparison analyses make use of: Comparative summaries, rankings and profiles between the data sets with variances identified and quantified. California Health Information Association, AHIMA Affiliate 21

23 43 Example: Monitoring Case Mix 1.6 Mean CMI CMI Quarterly Mean CMI Year Mean CMI Q1 12 Q2 12 Q3 12 Q4 12 Q1 13 Q2 13 Q3 13 Q Example: Comparison 1.6 Comparison of CMI CMI CMI 2013 CMI Quarters California Health Information Association, AHIMA Affiliate 22

24 45 Possible Scenario The hospital has 200 beds and discharged 12,350 patients in 2014 and 12,132 patients in Administration has noted a decrease in the Case Mix Index recently and is wondering whether the recent shake up in the Orthopedic service is to blame. You have been asked to take a look at the orthopedic cases for the past two years and try to determine whether there are any patterns or trends on which administration can act. 46 Possible Scenario How will you identify orthopedic cases? MS DRG, by Diagnosis code, or by attending physician or surgeon? Since we are being asked to specifically look at physician behavior, and since it is the attending physician who is responsible for the patient, we will ask for the cases attributed to a specific list of attending physicians. California Health Information Association, AHIMA Affiliate 23

25 47 Possible Scenario: Next Steps Select the appropriate time period Narrow down the specific data elements: includes and excludes Males and females? All payors? Age? Narrow down the data to extract, for example: First 6 diagnoses and procedures Total charges Gender Payer Discharge disposition Account or record number (for verification) Write the request for data carefully (for yourself or someone else to run the report) Possible Scenario: Working Through the Process Organizing and cleaning the data For example, any numeric looking data that has leading zeros that need to be retained Analyze the data Look at it several different ways, such as: Was there a significant change in volume from year to year? Was there an increase or decrease in charges? Did any orthopedists who had cases in 2014 not have any cases in 2015? Report the findings Was the orthopedic service shake up to blame for the hospital s declining CMI? Display the data Show the differences that developed between the two years. A table will probably do, but a couple of vertical bar graphs would be good for visuals. 48 California Health Information Association, AHIMA Affiliate 24

26 Summary Understand what data is being requested and verify its accuracy. Use the best data source to select pertinent data elements to generate reportable information based on user need. Compile and manipulate data for the appropriate audience. Use the best data presentation for the type of data considering the use of graphs, tables, spreadsheets, or other methods. Include all relevant information. Question/Answer California Health Information Association, AHIMA Affiliate 25