Transcript Susy Independent Project - SJD Fall 2012 5221

Assessing Variables of
Health Plan Authorization
Turn-Around-Timelines
Susan J. Dickerson
Professor Edward Schumacher
HCAI 5221
29 November 2012
The Theory
Effect on Authorization TAT timelines in the State of Texas for one health plan that has several
different service delivery areas within the State.
• Authorization requests have mandated turnaround timeline standards of 3 days maximum
that have to be met.
• The health plan experienced a massive membership growth in March of 2012 which was
met with major delays in authorizations not being met within the health care services
area of this health maintenance organization in Texas.
• This is a retrospective study, and I have selected authorization requests from the March
2012 through September 2012 timeframe specifically to document the impact that my
involvement had on the TAT.
The equation:
TAT = β₀ (All DATA) + β₁*(SDA-Dallas)+ β₂ (LOB Star) *+ β₃*Month (March-Oct) +e
Reason for Study
• This topic interested me because as a corporate employee,
I was called in to assist the Texas plan as they were
experiencing delays in meeting the TAT - I have been
assisting them to ensure compliance with State and
Contract Regulations.
• The Texas Department of Insurance (TDI) monitors
compliance with TAT under the Utilization Review Agent
(URA) licensing rules as does the Health and Human
Services Commission (HHSC) who oversees the Medicaid
and the Children's Health Insurance Policy (CHIP).
• Steep sanctions (as high as 1000 per auth for every day
delay could be imposed state wide).
• The Texas plan has over 5 service areas and processes
1000-2000 authorization requests daily.
DATA
Data Statistics
ANALYSIS OF GRAPHICAL DATA,
CORRELATION TESTS AND CHI
SQUARE TESTS
• The graphical data and correlation tests are consistent with each other and
do not demonstrate a linear relationship between my three independent
variables of Service Delivery Area (SDA), Line of Business (LOB), or Month,
versus my dependent variable of TAT.
• The graphical data did not show a linear relationship, but was more vertical
and silo appearing, possibly due to the nature of my chosen independent
variables.
• The correlation tests showed no direct linear relationship between TAT and
SDA, but it did reveal a slight POS correlation with the SDA variable and a
very large NEG correlations with the Month independent variable with my
dependent variable of TAT and showed a correlation with line of business.
• Chi square tests demonstrated that there is evidence of a relationship
between my independent variables (LOB and Month) when reviewed
individually versus my dependent variable TAT.
Simple Regression LOB/TAT
The null hypothesis is that there is no relationship between LOB and TAT.
The alternative hypothesis is that there is a relationship between LOB and TAT.
Our test statistic is t=(.63-0)/0.30=2.14
We use our p value of 0.035 to reject the null hypothesis, so we conclude that there is a relationship between the LOB and TAT.
I had thought that there would be no relationship between LOB and TAT because staff is to handle auths similarly however this warranted further
investigation and the results revealed that staff who handled STAR+Plus LOB were entering the received date as the date they determined they
would authorize or deny the request instead of the date the data request was received, so this data entry error allowed for a 5.4% of variation in
TAT to be explained by LOB and these results allowed me to identify this error and request a CAP.
Simple Regression Month/TAT
The null hypothesis is that there is no relationship between Month and TAT.
The alternative hypothesis is that there is a relationship between Month and TAT.
Our test statistic is t=(-0.41-0)/0.05 = -8.23
We use our p value of 0.000000000002 to reject the null hypothesis, and conclude that there is a
significant relationship between Month and resulting TAT.
I had thought that there would definitely be a negative relationship between the Month and
resulting TAT due to the process improvements put in place. This demonstrates that 44% of the
variation in TAT is explained by the Month of each data set capturing the turn around time.
Simple Regression SDA/TAT
The null hypothesis is that there e is no relationship between SDA and TAT.
The alternative hypothesis is that there is a relationship between SDA and TAT.
Our test statistic is t=(-0.06-0)/0.39=-0.16
We use our p value of 0.88 to reject the null hypothesis, and we accept that there is slight relationship between
SDA and TAT.
I did not think that there should be a large difference in SDA TAT since they were all given the same CAP
/process improvement but wanted to see if the Dallas SDA had a more positive impact due to leadership or
follow through and there does appear to be a slight negative relationship in SDA Dallas and TAT of less than 1%.
Multiple Regression
F TEST FOR SIGNIFICANCE OF THE OVERALL MULTIPLE REGRESSION MODEL
H0=There is no linear relationship between TAT and SDA, LOB and Month.
H1: There is a linear relationship between TAT and at least one of my independent variables of LOB and Month.
F=27
The rule is to reject the null hypothesis at the alpha level of significance if F>F(k,n-k-1)
Using a 0.05 level of significance, the critical value of the F distribution with three and eighty-four degrees of freedom
is approximately 3.276. So 27 > 3.276, so you reject the null hypothesis and conclude that there is a linear
relationship between TAT and SDA, LOB or Month.
Multiple Regression Analysis
B₀=4.35 B₁=(-0.10) B₂=(0.64 ) B₃=(-0.41)
So my equation is:
TAT= 4.36 + (-0.10)(DALLAS/Non Dallas ) + (.64)(LOB Star/Star+Plus) + (-0.41)(Month - March-Oct)
The slope of TAT with SDA indicates that for LOB and Month, the TAT is estimated to decrease by 0.10. The slope of TAT with LOB indicates
that for SDA and Month , the TAT is estimated to increase by 0.64. The slope of TAT and Month indicates that for SDA and LOB, the TAT is
estimated to decrease by 0.41.
My net regression coefficients estimate the mean change in TAT per unit change in one of my independent variables (SDA , LOB, Month) while
holding constant the other two independent variables. My regression coefficients do make sense in light of my previous chi square tests and
simple regressions, in that there is a slight negative association with SDA Dallas, slight positive association with LOB Star, and significant
negative association with Month (March-Oct) independent variables upon my dependent variable.
R^2=85.18/173.18=0.49
Adjusted R^2= 1-(1-R^2) ((n-1)/(n-k-1))= 0.47
Thus, 47% of variation in TAT is explained by the multiple regression model, adjusted for number of independent variables (3) and sample size
(88). 51% still remains unexplained.
My Standard Error of the estimate is 1.02. This measures variability around the prediction line of our model. Therefore, the typical difference
between actual TAT and the predicted value of TAT is approximately 1.02 days.
Multiple Regression Analysis