SlideShare une entreprise Scribd logo
1  sur  16
Télécharger pour lire hors ligne
HealthCare Quality
 Improvement Solutions




© 2012 by HealthCare Quality Improvement Solutions, LLC
• What is hypothesis testing?
                     A quantitative method for answering questions and
                      determining whether potential factors significantly effect
                      process performance




                                                                                   2
© 2012 by HealthCare Quality Improvement Solutions, LLC
• Primary purpose from a quality improvement
           perspective:
                     Determine whether the outcome of interest is produced
                      by a similar or dissimilar process




                                                                              3
© 2012 by HealthCare Quality Improvement Solutions, LLC
Hospital                             Hospital
  Potential                                               Performance                          Performance
   Factors                      Discharge                                    Discharge
                                 Rx ACEI                     75%              Rx ACEI             75%
Physician                  Similar Processes                            Dissimilar Processes

                                                             72%                                  95%



                                                             75%                                  71%


                                                             77%                                  60%


                           P-Value 0.567                                   P-Value 0.001
                                                                                                             4
© 2012 by HealthCare Quality Improvement Solutions, LLC
Similar Processes                                Dissimilar Processes
         • Critical few factors will                      • Critical few factors will
           not be identified                                be identified
         • Redesign the process                           • Focus quality
                                                            improvement on the
                                                            critical few factors




                                                                                        5
© 2012 by HealthCare Quality Improvement Solutions, LLC
• Question to be answered:
                     Is the defendant innocent or guilty?
         • The defendant is presumed
           innocent until proven guilty
                     In hypothesis testing this is known
                      as the null hypothesis and is
                      denoted H0
                     H0: Defendant is innocent
         • The plaintiff asserts that the
           defendant is guilty
                     In hypothesis testing this is known
                      as the alternate hypothesis and is
                      denoted HA
                     HA: Defendant is guilty

                                                             6
© 2012 by HealthCare Quality Improvement Solutions, LLC
• Potential factor:
                     Arrival day of week
         • Question:
                     Is there a significant difference in Pneumonia Antibiotic
                      Timing (Median) between patients that arrive on a
                      Weekday vs. Weekend?
         • Hypotheses:
                     H0: Weekday Median = Weekend Median
                            The antibiotic administration process is similar for weekdays
                             and weekend
                               – Arrival day of week does not significantly effect Pneumonia
                                 Antibiotic Timing
                     HA: Weekday Median                  Weekend Median
                            The antibiotic administration process is dissimilar for weekdays
                             and weekend
                               – Arrival day of week does significantly effect Pneumonia
                                 Antibiotic Timing and is among the Critical Few Factors

                                                                                                7
© 2012 by HealthCare Quality Improvement Solutions, LLC
• The null hypothesis H0:
                     Asserts there is no difference among factors
         • The alternate hypothesis HA:
                     Asserts that there is a difference among factors




                                                                         8
© 2012 by HealthCare Quality Improvement Solutions, LLC
Legal Process
 • The legal process is not
   perfect                                                                                     Truth
                                                                                   Innocent       Guilty
             Innocent defendants can be
              found guilty by the jury                    Jury
                                                                     Innocent      Correct        Incorrect

                     In hypothesis testing this is       Decision   Guilty        Incorrect      Correct
                      known as a Type I Error
                        – Rejecting the null
                                                                                Data Analysis
                          hypothesis when it is true
             Guilty defendants can be                                                   Actual State
              found innocent by the jury                                           H0 True        H0 False
                     In hypothesis testing this is                  Accept H0     Correct        Type II
                      known as a Type II Error                                                    Error -
                                                      Decision
                        – Accepting the null                         Reject H0     Type I         Correct
                          hypothesis when it is false                              Error -



© 2012 by HealthCare Quality Improvement Solutions, LLC                                                       9
• The P-Value is the chance of making a Type I
           Error if H0 is rejected
         • Decision Criteria:
                     If the P-Value is less than or equal to - reject H0
                     If the P-Value is greater than - accept H0




                                                                            10
© 2012 by HealthCare Quality Improvement Solutions, LLC
• Let’s return to the Pneumonia Antibiotic Timing
           question:
                  Is there a significant difference in Pneumonia Antibiotic
                   Timing (Median) between patients that arrive on a
                   Weekday vs. Weekend?
         • Hypotheses:
                     H0: Weekday Median = Weekend Median
                     HA: Weekday Median Weekend Median
         • Level of Significance ( ) – also referred to as
           alpha
                     0.05


                                                                               11
© 2012 by HealthCare Quality Improvement Solutions, LLC
• Let’s return to the Pneumonia Antibiotic Timing
           question:
                  Is there a significant difference in Pneumonia Antibiotic
                   Timing (Median) between patients that arrive on a
                   Weekday vs. Weekend? Yes
         • Hypotheses:
                  H0: Weekday Median = Weekend Median
                  HA: Weekday Median  Weekend Median
         • Level of Significance
                     0.05




                                                              P-Value


                                                                               12
© 2012 by HealthCare Quality Improvement Solutions, LLC
P-Value




         • Investigate why it takes longer on the weekend
           to deliver the initial antibiotic.




                                                                    13
© 2012 by HealthCare Quality Improvement Solutions, LLC
• When H0 is rejected, the hypothesis test is
           considered statistically significant at the selected
             level
         • It indicates that the sample measurement is
           unlikely if the null hypothesis is true

         •   QI Perspective:
              The sample measurements are likely being
               generated by dissimilar processes
              The factor is a critical factor effecting
               performance


                                                                  14
© 2012 by HealthCare Quality Improvement Solutions, LLC
Hospital                             Hospital
  Potential                                               Performance                          Performance
   Factors                      Discharge                                    Discharge
                                 Rx ACEI                     75%              Rx ACEI             75%
 Physician                 Similar Processes                            Dissimilar Processes

                                                             72%                                  95%



                                                             75%                                  71%


                                                             77%                                  60%


                           P-Value 0.567                                   P-Value 0.001
                                                                                                         15
© 2012 by HealthCare Quality Improvement Solutions, LLC
HealthCare Quality
 Improvement Solutions




                            Robert Sutter Contact Information
                  Website: https://sites.google.com/site/robertsutterrnmbamha/




© 2012 by HealthCare Quality Improvement Solutions, LLC

Contenu connexe

Similaire à Hypothesis Testing Fundamentals

HYPOTHESIS TESTING 20200702.pptx
HYPOTHESIS TESTING 20200702.pptxHYPOTHESIS TESTING 20200702.pptx
HYPOTHESIS TESTING 20200702.pptxDr. Gururaj Phatak
 
I/O chapter 4
I/O chapter 4I/O chapter 4
I/O chapter 4Roi Xcel
 
Research methodology iii
Research methodology iiiResearch methodology iii
Research methodology iiiAnwar Siddiqui
 
MRCPsych - How To Analyse Diagnostic Test Studies (May09)
MRCPsych - How To Analyse Diagnostic Test Studies (May09)MRCPsych - How To Analyse Diagnostic Test Studies (May09)
MRCPsych - How To Analyse Diagnostic Test Studies (May09)Alex J Mitchell
 
Hypothesis testing and p values 06
Hypothesis testing and p values  06Hypothesis testing and p values  06
Hypothesis testing and p values 06DrZahid Khan
 
Is the quality of case management in a medical home associated with patient s...
Is the quality of case management in a medical home associated with patient s...Is the quality of case management in a medical home associated with patient s...
Is the quality of case management in a medical home associated with patient s...Leonard Davis Institute of Health Economics
 
Direct to Consumer vs. Organic Growth - What works?
Direct to Consumer vs. Organic Growth - What works?Direct to Consumer vs. Organic Growth - What works?
Direct to Consumer vs. Organic Growth - What works?VSee
 
MRCPsych10 - How to analyse diagnostic and prognostic studies
MRCPsych10 - How to analyse diagnostic and prognostic studiesMRCPsych10 - How to analyse diagnostic and prognostic studies
MRCPsych10 - How to analyse diagnostic and prognostic studiesAlex J Mitchell
 
March Webinar: One Million Data Points: The Link Between Well-Being, Optima...
March Webinar:  One Million Data Points:  The Link Between Well-Being, Optima...March Webinar:  One Million Data Points:  The Link Between Well-Being, Optima...
March Webinar: One Million Data Points: The Link Between Well-Being, Optima...Worksite Wellness Council of Massachusetts
 

Similaire à Hypothesis Testing Fundamentals (11)

HYPOTHESIS TESTING 20200702.pptx
HYPOTHESIS TESTING 20200702.pptxHYPOTHESIS TESTING 20200702.pptx
HYPOTHESIS TESTING 20200702.pptx
 
I/O chapter 4
I/O chapter 4I/O chapter 4
I/O chapter 4
 
Research methodology iii
Research methodology iiiResearch methodology iii
Research methodology iii
 
MRCPsych - How To Analyse Diagnostic Test Studies (May09)
MRCPsych - How To Analyse Diagnostic Test Studies (May09)MRCPsych - How To Analyse Diagnostic Test Studies (May09)
MRCPsych - How To Analyse Diagnostic Test Studies (May09)
 
Hypothesis testing and p values 06
Hypothesis testing and p values  06Hypothesis testing and p values  06
Hypothesis testing and p values 06
 
LDI Charles Leighton Memorial Lecture with Mark Chassin, MD 5_4_12
LDI Charles Leighton Memorial Lecture with Mark Chassin, MD  5_4_12LDI Charles Leighton Memorial Lecture with Mark Chassin, MD  5_4_12
LDI Charles Leighton Memorial Lecture with Mark Chassin, MD 5_4_12
 
Is the quality of case management in a medical home associated with patient s...
Is the quality of case management in a medical home associated with patient s...Is the quality of case management in a medical home associated with patient s...
Is the quality of case management in a medical home associated with patient s...
 
Direct to Consumer vs. Organic Growth - What works?
Direct to Consumer vs. Organic Growth - What works?Direct to Consumer vs. Organic Growth - What works?
Direct to Consumer vs. Organic Growth - What works?
 
MRCPsych10 - How to analyse diagnostic and prognostic studies
MRCPsych10 - How to analyse diagnostic and prognostic studiesMRCPsych10 - How to analyse diagnostic and prognostic studies
MRCPsych10 - How to analyse diagnostic and prognostic studies
 
Power point rethinking assessment
Power point   rethinking assessmentPower point   rethinking assessment
Power point rethinking assessment
 
March Webinar: One Million Data Points: The Link Between Well-Being, Optima...
March Webinar:  One Million Data Points:  The Link Between Well-Being, Optima...March Webinar:  One Million Data Points:  The Link Between Well-Being, Optima...
March Webinar: One Million Data Points: The Link Between Well-Being, Optima...
 

Hypothesis Testing Fundamentals

  • 1. HealthCare Quality Improvement Solutions © 2012 by HealthCare Quality Improvement Solutions, LLC
  • 2. • What is hypothesis testing?  A quantitative method for answering questions and determining whether potential factors significantly effect process performance 2 © 2012 by HealthCare Quality Improvement Solutions, LLC
  • 3. • Primary purpose from a quality improvement perspective:  Determine whether the outcome of interest is produced by a similar or dissimilar process 3 © 2012 by HealthCare Quality Improvement Solutions, LLC
  • 4. Hospital Hospital Potential Performance Performance Factors Discharge Discharge Rx ACEI 75% Rx ACEI 75% Physician Similar Processes Dissimilar Processes 72% 95% 75% 71% 77% 60% P-Value 0.567 P-Value 0.001 4 © 2012 by HealthCare Quality Improvement Solutions, LLC
  • 5. Similar Processes Dissimilar Processes • Critical few factors will • Critical few factors will not be identified be identified • Redesign the process • Focus quality improvement on the critical few factors 5 © 2012 by HealthCare Quality Improvement Solutions, LLC
  • 6. • Question to be answered:  Is the defendant innocent or guilty? • The defendant is presumed innocent until proven guilty  In hypothesis testing this is known as the null hypothesis and is denoted H0  H0: Defendant is innocent • The plaintiff asserts that the defendant is guilty  In hypothesis testing this is known as the alternate hypothesis and is denoted HA  HA: Defendant is guilty 6 © 2012 by HealthCare Quality Improvement Solutions, LLC
  • 7. • Potential factor:  Arrival day of week • Question:  Is there a significant difference in Pneumonia Antibiotic Timing (Median) between patients that arrive on a Weekday vs. Weekend? • Hypotheses:  H0: Weekday Median = Weekend Median  The antibiotic administration process is similar for weekdays and weekend – Arrival day of week does not significantly effect Pneumonia Antibiotic Timing  HA: Weekday Median Weekend Median  The antibiotic administration process is dissimilar for weekdays and weekend – Arrival day of week does significantly effect Pneumonia Antibiotic Timing and is among the Critical Few Factors 7 © 2012 by HealthCare Quality Improvement Solutions, LLC
  • 8. • The null hypothesis H0:  Asserts there is no difference among factors • The alternate hypothesis HA:  Asserts that there is a difference among factors 8 © 2012 by HealthCare Quality Improvement Solutions, LLC
  • 9. Legal Process • The legal process is not perfect Truth Innocent Guilty  Innocent defendants can be found guilty by the jury Jury Innocent Correct Incorrect  In hypothesis testing this is Decision Guilty Incorrect Correct known as a Type I Error – Rejecting the null Data Analysis hypothesis when it is true  Guilty defendants can be Actual State found innocent by the jury H0 True H0 False  In hypothesis testing this is Accept H0 Correct Type II known as a Type II Error Error - Decision – Accepting the null Reject H0 Type I Correct hypothesis when it is false Error - © 2012 by HealthCare Quality Improvement Solutions, LLC 9
  • 10. • The P-Value is the chance of making a Type I Error if H0 is rejected • Decision Criteria:  If the P-Value is less than or equal to - reject H0  If the P-Value is greater than - accept H0 10 © 2012 by HealthCare Quality Improvement Solutions, LLC
  • 11. • Let’s return to the Pneumonia Antibiotic Timing question:  Is there a significant difference in Pneumonia Antibiotic Timing (Median) between patients that arrive on a Weekday vs. Weekend? • Hypotheses:  H0: Weekday Median = Weekend Median  HA: Weekday Median Weekend Median • Level of Significance ( ) – also referred to as alpha  0.05 11 © 2012 by HealthCare Quality Improvement Solutions, LLC
  • 12. • Let’s return to the Pneumonia Antibiotic Timing question:  Is there a significant difference in Pneumonia Antibiotic Timing (Median) between patients that arrive on a Weekday vs. Weekend? Yes • Hypotheses:  H0: Weekday Median = Weekend Median  HA: Weekday Median Weekend Median • Level of Significance  0.05 P-Value 12 © 2012 by HealthCare Quality Improvement Solutions, LLC
  • 13. P-Value • Investigate why it takes longer on the weekend to deliver the initial antibiotic. 13 © 2012 by HealthCare Quality Improvement Solutions, LLC
  • 14. • When H0 is rejected, the hypothesis test is considered statistically significant at the selected level • It indicates that the sample measurement is unlikely if the null hypothesis is true • QI Perspective:  The sample measurements are likely being generated by dissimilar processes  The factor is a critical factor effecting performance 14 © 2012 by HealthCare Quality Improvement Solutions, LLC
  • 15. Hospital Hospital Potential Performance Performance Factors Discharge Discharge Rx ACEI 75% Rx ACEI 75% Physician Similar Processes Dissimilar Processes 72% 95% 75% 71% 77% 60% P-Value 0.567 P-Value 0.001 15 © 2012 by HealthCare Quality Improvement Solutions, LLC
  • 16. HealthCare Quality Improvement Solutions Robert Sutter Contact Information Website: https://sites.google.com/site/robertsutterrnmbamha/ © 2012 by HealthCare Quality Improvement Solutions, LLC