Due: November 5, 2013 Team: Eikenberry, Gamble, Hinkle, Liu, Zhang Page 1 of 17 Team: Steven Eikenberry, Dale Gamble, Jessica Hinkle, Run Liu, Jinliang Zhang ACG5075 MBA Section 4993 Professor: Joost Impink Managerial Accounting, CASE #1 A. Explore the difference between the amounts charged (charges) and the cost (costs). You may treat the difference as a discount. Does the discount depend on the state the hospital operates in? Yes, the discount depends on the state the hospital operates in. A more detailed analysis is explored below. Example 1: In Chart 1 (Above), the average discounts for surgery procedure 460 (SPINAL FUSION EXCEPT CERVICAL W/O MCC) were compared by state with the national average of $67,790. Chart 1 indicates that average discounts can vary dramatically from one state to another for Procedure 460. For example, the average discount in California is approximately six times higher than those in Vermont or Hawaii. Next, each state’s average discount was statistically compared to the nationwide distribution of discounts to determine how well each state matched the national distribution. Note that the zvalues calculated for Table 1 (Below) depend on more than just the average discount for each state. The z values also depend on how many hospitals (n) reported data in that state. The more data points for that state, the less likely the calculated average discount could be a statistical anomaly. Utilizing the zvalue takes this concern into account and allows us to determine how probable it would be for a random sample of size n to have a state average this high above (or below) the national average. These probabilities are reflected in the column labeled ‘.05 α Test’ which reports whether the particular state has less than a 5% chance of matching the nationwide distribution, either on the ‘High’ or the ‘Low’ side. If there is a greater than 5% chance of matching the nationwide distribution, the state is listed as ‘In Range’. $0 $20,000 $40,000 $60,000 $80,000 $100,000 $120,000 $140,000 MD VT HI RI ME DE MA MI NY OR ND IA ID AR WI UT MN CT NH MT AL KS WV NE OH MS KY NC VA SD AK OK MO GA WY National Average IN LA TN NM DC IL PA WA SC TX AZ FL NV NJ CO CA Chart 1: Average Discount For Procedure 460, by State
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
Due: November 5, 2013
Team: Eikenberry, Gamble, Hinkle, Liu, Zhang Page 1 of 17
Team: Steven Eikenberry, Dale Gamble, Jessica Hinkle, Run Liu, Jinliang Zhang
ACG5075 MBA Section 4993 Professor: Joost Impink
Managerial Accounting, CASE #1
A. Explore the difference between the amounts charged (charges) and the cost (costs). You may treat the difference as a discount. Does the discount depend on the state the hospital operates in?
Yes, the discount depends on the state the hospital operates in. A more detailed analysis is explored below.
Example 1:
In Chart 1 (Above), the average discounts for surgery procedure 460 (SPINAL FUSION EXCEPT CERVICAL W/O MCC) were compared by state with the national average of $67,790.
Chart 1 indicates that average discounts can vary dramatically from one state to another for Procedure 460. For example, the average discount in California is approximately six times higher than those in Vermont or Hawaii.
Next, each state’s average discount was statistically compared to the nationwide distribution of discounts to determine how well each state matched the national distribution. Note that the z-‐values calculated for Table 1 (Below) depend on more than just the average discount for each state. The z-‐values also depend on how many hospitals (n) reported data in that state. The more data points for that state, the less likely the calculated average discount could be a statistical anomaly.
Utilizing the z-‐value takes this concern into account and allows us to determine how probable it would be for a random sample of size n to have a state average this high above (or below) the national average. These probabilities are reflected in the column labeled ‘.05 α Test’ which reports whether the particular state has less than a 5% chance of matching the nationwide distribution, either on the ‘High’ or the ‘Low’ side. If there is a greater than 5% chance of matching the nationwide distribution, the state is listed as ‘In Range’.
Chart 1: Average Discount For Procedure 460, by State
Due: November 5, 2013
Team: Eikenberry, Gamble, Hinkle, Liu, Zhang Page 2 of 17
Table 1: Average Discounts and Probabilities of Matching National Distribution for Procedure 460, by State
State
State Average Discount
Sample Size Z-‐Value P-‐Value .05 α Test
AK $64,894 3 -‐0.111 0.46 In Range AL $49,516 27 -‐2.101 0.02 Low AR $44,078 16 -‐2.098 0.02 Low AZ $80,770 23 1.377 0.92 In Range CA $126,411 96 12.706 1.00 High CO $107,869 28 4.692 1.00 High CT $45,322 18 -‐2.109 0.02 Low DC $73,002 4 0.231 0.59 In Range DE $24,840 3 -‐1.646 0.05 In Range FL $88,429 91 4.356 1.00 High GA $67,527 45 -‐0.039 0.48 In Range HI $18,757 4 -‐2.169 0.02 Low IA $42,586 17 -‐2.299 0.01 Low ID $43,482 11 -‐1.783 0.04 In Range IL $73,695 55 0.969 0.83 In Range IN $68,369 39 0.080 0.53 In Range KS $49,650 16 -‐1.605 0.05 In Range KY $55,427 16 -‐1.094 0.14 In Range LA $69,594 27 0.207 0.58 In Range MA $29,195 21 -‐3.913 0.00 Low MD $2,385 27 -‐7.518 0.00 Low ME $23,694 7 -‐2.581 0.00 Low MI $37,934 41 -‐4.229 0.00 Low MN $44,713 22 -‐2.394 0.01 Low MO $66,622 30 -‐0.141 0.44 In Range MS $54,555 15 -‐1.134 0.13 In Range MT $49,474 9 -‐1.216 0.11 In Range NC $61,265 32 -‐0.817 0.21 In Range ND $40,751 3 -‐1.036 0.15 In Range NE $53,004 14 -‐1.224 0.11 In Range NH $49,456 7 -‐1.073 0.14 In Range NJ $95,408 31 3.402 1.00 High NM $71,097 9 0.219 0.59 In Range NV $91,818 14 1.989 0.98 High NY $38,565 51 -‐4.617 0.00 Low OH $54,241 53 -‐2.182 0.01 Low OK $65,426 21 -‐0.240 0.41 In Range OR $40,518 19 -‐2.630 0.00 Low PA $74,667 56 1.139 0.87 In Range RI $19,509 7 -‐2.826 0.00 Low SC $79,125 26 1.279 0.90 In Range SD $63,024 10 -‐0.333 0.37 In Range TN $69,813 33 0.257 0.60 In Range TX $79,442 119 2.812 1.00 High UT $44,552 11 -‐1.705 0.04 In Range VA $62,017 39 -‐0.798 0.21 In Range VT $17,322 4 -‐2.233 0.01 Low WA $75,631 29 0.934 0.82 In Range WI $44,254 23 -‐2.497 0.01 Low WV $52,319 6 -‐0.838 0.20 In Range WY $67,571 4 -‐0.010 0.50 In Range
Due: November 5, 2013
Team: Eikenberry, Gamble, Hinkle, Liu, Zhang Page 3 of 17
Table 1 (previous page) shows that 22 of the 50 states (51 if including DC) have less than a 5% chance of matching the national distribution, six on the high side and sixteen on the low side. The six high states are highlighted in red.
Similarly, the average state discounts for Procedure 470 (MAJOR JOINT REPLACEMENT OR REATTACHMENT OF LOWER EXTREMITY W/O MCC) were found to vary widely:
It was then determined that 28 of the 50 states have less than a 5% probability of matching the national distribution. Six of the states have improbably high discounts, and twenty-‐two have improbably low discounts. Again, the six high states are highlighted in red in Table 2; note that these are the same states that were indicated as having high discounts in Table 1, as seen on the following page:
$0
$10,000
$20,000
$30,000
$40,000
$50,000
$60,000
$70,000
$80,000
MD
VT
MA
MT ND RI
MI
ME
WV HI
MN
DE
NY ID
CT
UT
IA
SD
OR
WI
AR
KY
KS
OH
NC
WY
NE
MO
NH
DC
IN
OK
GA
VA
AK
NM
WA
National Average
TN
LA
PA
IL
AL
AZ
MS SC
CO
TX
FL
NV NJ
CA
Chart 2: Average Discount For Procedure 470, by State
Due: November 5, 2013
Team: Eikenberry, Gamble, Hinkle, Liu, Zhang Page 4 of 17
Table 2: Average Discounts and Probabilities of Matching National Distribution for Procedure 470, by State
State
State Average Discount
Sample Size Z-‐Value P-‐Value .05 α Test
AK 35945.62 7 -‐0.170 0.43 In Range AL 40129.94 53 0.795 0.79 In Range AR 27045.70 28 -‐2.292 0.01 Low AZ 40190.03 49 0.782 0.78 In Range CA 68788.17 235 19.880 1.00 High CO 45028.52 44 2.071 0.98 High CT 24051.72 29 -‐3.000 0.00 Low DC 34465.29 6 -‐0.308 0.38 In Range DE 22905.26 5 -‐1.352 0.09 In Range FL 52867.21 151 7.828 1.00 High GA 34965.45 81 -‐0.944 0.17 In Range HI 21457.87 11 -‐2.204 0.01 Low IA 24277.00 33 -‐3.147 0.00 Low ID 23934.49 12 -‐1.947 0.03 In Range IL 39576.49 110 0.904 0.82 In Range IN 34595.39 74 -‐1.034 0.15 In Range KS 28498.44 44 -‐2.473 0.01 Low KY 27333.74 48 -‐2.918 0.00 Low LA 38727.72 62 0.402 0.66 In Range MA 17698.81 56 -‐6.140 0.00 Low MD 1317.38 43 -‐9.832 0.00 Low ME 20850.02 20 -‐3.085 0.00 Low MI 20665.65 89 -‐6.580 0.00 Low MN 22464.89 49 -‐4.361 0.00 Low MO 32181.24 63 -‐1.748 0.04 In Range MS 40661.36 30 0.719 0.76 In Range MT 18229.52 12 -‐2.766 0.00 Low NC 29246.40 80 -‐3.058 0.00 Low ND 18703.99 6 -‐1.908 0.03 In Range NE 30619.23 21 -‐1.306 0.10 In Range NH 34311.62 13 -‐0.476 0.32 In Range NJ 62802.39 60 8.124 1.00 High NM 37329.20 23 -‐0.033 0.49 In Range NV 60848.54 21 4.435 1.00 High NY 23747.22 141 -‐6.766 0.00 Low OH 29003.33 126 -‐3.951 0.00 Low OK 34835.12 51 -‐0.788 0.22 In Range OR 24957.31 31 -‐2.893 0.00 Low PA 38777.25 135 0.617 0.73 In Range RI 19375.31 10 -‐2.375 0.01 Low SC 43355.07 43 1.592 0.94 In Range SD 24519.72 15 -‐2.083 0.02 Low TN 38068.41 63 0.188 0.57 In Range TX 47996.88 224 6.513 1.00 High UT 24081.10 27 -‐2.889 0.00 Low VA 35036.75 64 -‐0.815 0.21 In Range VT 14809.95 6 -‐2.303 0.01 Low WA 37397.22 45 -‐0.027 0.49 In Range WI 25123.71 63 -‐4.070 0.00 Low WV 21080.97 27 -‐3.535 0.00 Low WY 29776.63 11 -‐1.061 0.14 In Range
Due: November 5, 2013
Team: Eikenberry, Gamble, Hinkle, Liu, Zhang Page 5 of 17
As noted above, the same six states (California, Colorado, Florida, Nevada, New Jersey, and Texas) have improbably high discounts for both Procedures 460 and 470. In statistical terms, this means the same states had high ‘z-‐values’ in both data sets. On closer inspection, it becomes apparent that many of the states with low discounts for Procedure 460 also have low discounts for Procedure 470. We can use statistics to determine how well each state’s average discount for Procedure 460 correlates to that state’s discount for Procedure 470. Chart 3 shows this correlation:
As shown in Chart 3, the Coefficient of Determination (R2) between discounts for Procedures 460 and 470 for a particular state is 0.8103; this means that over 80% of the variation in the Procedure 470 discount can be attributed to the variation in the Procedure 460 discount in that same state.
Tables 1 and 2 respectively indicate that 22 of the states’ average discounts for Procedure 460 and 28 of the states’ average discounts for Procedure 470 do not match the national average distribution. This means that the discounts depend heavily on the state in which the hospital operates. Chart 3 further indicates that the states are at least 80% correlated from one procedure to another; states that have high discounts for one procedure are likely to have high discounts for another, and states with low discounts for one procedure are likely to have low discounts for another.
This correlation increases to nearly 90% when comparing the probabilities (as represented by the z-‐values) that a state is higher or lower than the national average across different procedures, as shown in Chart 4, located on the following page.
Chart 3: Correlation between State Discounts for Procedures 460 and 470
Due: November 5, 2013
Team: Eikenberry, Gamble, Hinkle, Liu, Zhang Page 6 of 17
Once again, the discount for a procedure depends on the state in which the hospital operates.
The calculations above were repeated after taking outliers into account. Outliers were determined via two methods:
In the first method, any discount found to be greater than three standard deviations from the population mean was classified as an outlier. Assuming a normal distribution, this would theoretically classify only 0.3% of the samples as outliers. Instead, 18 of 1332 samples (1.4%) for Procedure 460 and 38 of 2750 samples (again 1.4%) for Procedure 470 were calculated as outliers with this approach.
A second approach of identifying outliers was also examined. This method calculated an interquartile range (Q3-‐Q1: the middle 50% of the discount population falls into this range), and classified any value greater than 1.5 times the interquartile range above the third quartile or less than 1.5 times the interquartile range below the first quartile as an outlier. This more aggressive approach classified 54 of 1332 samples (4.1%) for Procedure 460 and 113 of 2750 samples (again 4.1%) for Procedure 470 as outliers. Since this second method had greater potential to disrupt the earlier finding, it was utilized to identify and exclude the outliers from the calculations.
In both methods, all of the outliers were on the high side.
As a result, for Procedure 460, 21 of the 50 states (and DC) were found to have average discounts less than 5% probable of matching the national average, vs 22 states previously. It should be noted that if the probability were increased to 5.1% the result would have been 22 states again.
R² = 0.89248
-‐15
-‐10
-‐5
0
5
10
15
20
25
-‐10.000 -‐5.000 0.000 5.000 10.000 15.000
State 470 z-‐Value
State 460 z-‐Value
Chart 4: Correlation between State Discount z-‐Values for Procedures 460 and 470
Due: November 5, 2013
Team: Eikenberry, Gamble, Hinkle, Liu, Zhang Page 7 of 17
For Procedure 470 the number of “non-‐conforming” states actually increased to 30 from 28. The six high states were unchanged, but removing the outliers (all of which were on the high side) reduced the deviation sufficiently to cause more of the low states to fall outside the now-‐narrower 5% threshold. In any event, the conclusion that the discounts depend on the state in which the hospital operates was not affected. Similarly, the coefficient of determination between discounts for Procedures 460 and 470 reduced almost insignificantly to 0.80 (see Chart 5), while the coefficient of determination between z-‐values for the discounts decreased slightly to 0.86; the correlations remain very strong:
Chart 5: Correlation between State Discounts for Procedures 460 and 470 (excluding outliers)
Due: November 5, 2013
Team: Eikenberry, Gamble, Hinkle, Liu, Zhang Page 8 of 17
B. Explore the costs for both drgs. What is the distribution of the average cost for each drg over the hospitals? Are there differences between states? (You may ignore the amounts charged (charges) in this and following questions.)
The distribution for both procedures is right skewed. Most states are near the average, but in both cases there are several states far above the average. This can be seen in the histograms that follow. The average cost of each procedure across all states and hospitals is highlighted in yellow. Outliers were included in this analysis. Please see below: DRG 460 State Costs
Bin Frequency 22000 1 24000 2 26000 13 28000 14 30000 10 32000 4 34000 3 More 4 DRG 470 State Costs
Bin Frequency 12000 1 13000 8 14000 19 15000 9 16000 5 17000 2 18000 2 More 5
460 Statistics Mean 27995.48079 Standard Deviation 3623.656395 Skewness 1.02148753 Range 17333.0717 Minimum 21268.3718 Maximum 38601.4435
470 Statistics Mean 14572.47531 Standard Deviation 2075.497573 Skewness 1.466339835 Range 9451.739479 Minimum 11709.20188 Maximum 21160.94136
DRG ID Number of Procedures Total Cost of All Procedures Average Cost of Procedure 460 65997 $1,804,860,128 $27,348 470 427207 $6,119,519,851 $14,324
Due: November 5, 2013
Team: Eikenberry, Gamble, Hinkle, Liu, Zhang Page 9 of 17
Steps taken in the calculation of the above responses: 1. Calculate the total cost of each procedure for each hospital using sum product 2. Use a pivot table to group hospitals by state and aggregate the total costs and number of procedures (Pivot worksheet: in Appendix)
3. Use these aggregated numbers to get a state average (Columns D & G in Pivot Work sheet) 4. Create histograms for each 5. Run descriptive statistics on each
Additionally, the comparison chart above demonstrates that with exception of rare outliers, there are
no obvious differences in the cost of the drgs by state.
0
10000
20000
30000
40000
50000
AK
AR
CA
CT
DE
GA
IA
IL
KS
LA
MD MI
MO
MT ND
NH
NM
NY
OK
PA
SC
TN
UT
VT
WI
WY
Cost of drgs by states
average_cost_460 average_cost_470
Due: November 5, 2013
Team: Eikenberry, Gamble, Hinkle, Liu, Zhang Page 10 of 17
C. In addition to B): Does the degree of specialization impact the costs? Use the number of unique drgs for each hospital as a proxy of specialization.
The degree of specialization has little impact on costs. For both DRG460 and DRG 470, the correlation is less than .1 between these data sets and the coefficient of determination is less than 1%. (this means less than 1% of the price variance is explained by the number of unique DRGs). Outliers here were again included and even they could not create a correlation:
Steps taken in the calculation of the above response:
1. Find the average price for both 460 & 470 for each hospital using a V lookup from the data on Question B spreadsheet
2. We have more info on hospital specialization than for cost of 460 & 470, so eliminate hospitals with no data before running a correlation (this is where columns F&G have N/As … correlation data sets are J&K columns and N&O columns (In Appendix)
3. Run correlations between J&K and N&O , answers in yellow above 4. Create scatter plots between #Unique DRGs and average cost 5. Add R squared to each scatterplot to find the coefficient of determination
R² = 0.0051
0 20000 40000 60000 80000 100000 120000 140000
0 20 40 60 80 100 120
Avg Price
Unique DRGs
460 Price vs. Unique DRGs
R² = 0.00712
0
10000
20000
30000
40000
50000
0 20 40 60 80 100 120
Avg Price
Unique DRGs
470 Price vs. Unique # DRGs
Due: November 5, 2013
Team: Eikenberry, Gamble, Hinkle, Liu, Zhang Page 11 of 17
D. Do you believe there are economies of scale (positive or negative) for drgs 460 and 470 for Shands and FRMC?
There is no evidence in the data to suggest that Shands and FRMC will observe economies of scale (or diseconomies of scale) as the result of their joint venture initiative.
To determine if there are economies of scale for Procedures 460 or 470, it is necessary to correlate the costs of the procedures with the number of patients at each hospital on which those procedures are performed. Charts 6 and 7 plot the costs against the number of discharges (which serves as a measure of the number of patients treated, presumably with a small error for patients that died during treatment) for Procedures 460 and 470, respectively.
R² = 0.00506
$0
$20,000
$40,000
$60,000
$80,000
$100,000
$120,000
$140,000
0 50 100 150 200 250 300 350 400
Cost of Treatment
# Patients Treated (Discharges)
Chart 6: Cost of Procedure 460 vs # of Patients Treated
Chart 7: Cost of Procedure 470 vs # of Patients Treated
Due: November 5, 2013
Team: Eikenberry, Gamble, Hinkle, Liu, Zhang Page 12 of 17
The coefficient of determination is very low for both procedures, only half of one percent, which indicates that increasing the number of procedures performed has no discernible effect on cost. Hence, there are no economies of scale for either procedure.
To discern the influence of outliers on the data, the interquartile range procedure described in Section A was again utilized to identify and remove outlying data points. Outliers were removed both for the cost of treatment, and for the number of patients treated. Charts 8 and 9 show that the results are unchanged: there is no evidence of a correlation between cost of treatment and number of procedures performed. If anything, the correlation decreased with the removal of the outliers.
Chart 8: Cost of Procedure 460 vs # of Patients Treated (Outliers Removed)
R² = 0.00025
$0
$5,000
$10,000
$15,000
$20,000
$25,000
0 50 100 150 200 250 300 350 400 450 500
Cost of Treatment
# Patients Treated (Discharges)
Chart 9: Cost of Procedure 470 vs # of Patients Treated (Outliers Removed)
Due: November 5, 2013
Team: Eikenberry, Gamble, Hinkle, Liu, Zhang Page 13 of 17
Finally, given the earlier findings that discounts and costs depend heavily on the states in which the hospitals are located, the costs and patient counts for just the Florida hospitals were analyzed. Charts 10 and 11 show that while the Coefficient of Determination did increase, the resulting correlation was still insignificant at 2% and 1% respectively for Procedures 460 and 470.
In conclusion, there is no evidence in the data to suggest that Shands and FRMC will observe economies of scale (or diseconomies of scale) as the result of their joint venture initiative.
R² = 0.01997
$0
$10,000
$20,000
$30,000
$40,000
$50,000
$60,000
0 50 100 150 200 250 300
Cost of Treatment
# Patients Treated (Discharges)
Chart 10: Florida Cost of Procedure 460 vs # of Patients Treated
R² = 0.01102
$0
$5,000
$10,000
$15,000
$20,000
$25,000
$30,000
0 100 200 300 400 500 600 700 800
Cost of Treatment
# Patients Treated (Discharges)
Chart 11: Florida Cost of Procedure 470 vs # of Patients Treated
Due: November 5, 2013
Team: Eikenberry, Gamble, Hinkle, Liu, Zhang Page 14 of 17
E. Shortly address challenges that Shands and FRMC need to be manage in case they go ahead and form the joint venture. Address issues like profit/cost sharing, difference in goals (profit vs educational), compensation, et cetera.
The main challenge facing Shands is the difference in missions. Shands is a nonprofit
organization, while FRMC is profit-‐seeking. Therefore FRMC will be more naturally inclined to reduce costs & revenue. As a manager of FRMC, the joint venture would be a hard sell for me (based on no knowledge of economies of scale) because I can perform both DRGs at a lower cost than can Shands. However, 460 is a more expensive operation so if I only perform 470, my costs will be minimized. Depending on my contribution margin on 470 though, I may be less profitable even though my costs are lower.
Profit sharing is the next largest challenge. Will each hospital simply cease doing one of the
procedures and absorb all costs and profits for the one they continue to do? Or, for instance, will FRMC be sharing some profits with Shands since Shands would be agreeing to take on the more costly procedure. As mentioned in the paragraph above, if FRMC’s contribution margin is such that they make less by only performing 470, they will surely want to share some of Shands’ 460 profit, since FRMC is a profit seeker.
Yet another key difficulty faced by a Shands/FRMC joint venture would be legal approval
from both entities. To reduce potential lawsuit liabilities, each hospital doubtless follows strict protocols for many aspects of their operations. These protocols would cover admitting procedures, surgery confirmation procedures (eg, confirming the left knee rather than the right knee is to be replaced), patient discharge routines, medical record retention, billing, and many more. The lawyers for each hospital would either need to approve the use of each of the other hospital’s protocols, or endeavor to merge the two sets of protocols into a unified whole.
Due: November 5, 2013
Team: Eikenberry, Gamble, Hinkle, Liu, Zhang Page 15 of 17
APPENDIX PIVOT TABLE referenced in Question B: Column Labels
460 470
Sum of Total Cost Sum of # of Procedures Sum of Total Cost