|
|
|||
|
|
|
|
|
|
||||
|
Background |
DRG Analysis Module The DRG Analysis module starts with the DRG Overview tool, which lists all hospitals with unusual DRG coding patterns and computes two different estimates of the recovery opportunity for each hospital. Each module in the DRG Analysis compares every hospital's coding behavior with two different baselines, one based on a nationwide hospital sample and another taken from the payer's own data. Health Data Desk's flexibility allows users to compare their data to either baseline, and to switch between baselines instantaneously. 
Health Data Desk's analysis for DRG payers looks at the percentage of high-paying DRG's within certain disease groups, such as bacterial pneumonia, which is shown here. "Hospital A" has 50% of its pneumonia cases in the highest-paying DRG compared with 28% for the rest of this payer's providers. Providers with unusually high percentages of expensive DRG's are plotted above the red line. Using statistics, Health Data Desk calculates that the chance of this happening due to chance is less than 3 in 10,000! The DRG Ratio tool considers the distribution of DRG assignments within each hospital of concern. The tool identifies pairs of DRGs for which upcoding from one to another is possible. It then compares the proportion of patients assigned to higher paying DRG's in each hospital with the two different baselines Hospitals that exceed the percentage of higher-paying DRG's in the baselines are marked and listed in the DRG Overview list. With the help of medical and coding experts, we have identified key ICD9 codes that can be used to force a claim into a higher-paying DRG. Our Misspecification tool detects hospitals with very high percentages of these key ICD9 codes. The DRG Trend tool looks for changes in DRG coding patterns over time. A hospital is considered suspicious if its coding behavior changes but its patient mix does not. 
For DRG payers Health Data Desk's case-adjusted logistic regression analysis uses patient demographic and diagnosis data to predict whether each claim should be a high or low-paying claim. We then compare our prediction to the actual DRG. If a provider has many high-paying claims that our model says should be low-paying, that provider will appear above the red line in the plot. Here our model finds that "Hospital J" may be miscoding up to 30% of their pneumonia cases to a higher-paying DRG. Within each DRG family, the DRG Model Exception tool uses a proprietary case-mix-adjusted statistical model to predict the appropriate DRG of the claim based on factors such as length of stay, patient demographics, and coding variables. After calculating a predicted DRG for each claim, we can compare the model's predictions to the actual DRG assignment. We flag those cases where our model predicts a lower-paying DRG than was actually assigned. It also identifies hospitals that have higher percentages of these "model exceptions." Because our statistical model is independent of the coding software used to assign the DRG and uses other factors besides ICD-9 codes, it is less likely to be fooled by improper ICD-9 coding practices. 
One of Health Data Desk's strengths is its ability to "drill down" into the data to isolate small groups of unusual claims. This plot shows the percentage of high-paying DRG's for all physicians in "Hospital A." The highlighted doctor has 6 out of 7 patients (or 86%) in the highest-paying pneumonia DRG. The blue line represents "Hospital A's" overall percentage; the green line is the same percentage for all of this payer's claims. The Drill Down feature and Demographics Summaries tool compare the patient characteristics of selected hospitals with the baseline characteristics, for each DRG pair or family. The Claim Viewer displays detailed patient records for hospitals with anomalous behavior. |
|
|||
|
|||||
|
|
|||||
