Aims Methods Results · • Discharge Planning starts on hospital day 1 • Family Education...
1
Aims Background Methods Results Results Acknowledgements Conclusions 1) Compare post- and pre-GFP patient groups characteristics and outcomes. 2) Apply a predictive model developed from two years of Geriatric Fracture Program (GFP) patients to pre-GFP patients to account for patient complexity in analyzing the outcomes of the UC Davis GFP. • >300,000 geriatric hip fractures per year in the United States, incidence expected to increase 1,2 • High rates of complications, leading to prolonged hospitalizations 3 . • $9-15bn/yr of inpatient costs 4 . • $26,000-$35,000 average hospitalization cost alone References 1: Morris, A. H., Zuckerman, J. D., & AAOS Council of Health Policy and Practice, USA. American Academy of Orthopaedic Surgeons. (2002). National Consensus Conference on Improving the Continuum of Care for Patients with Hip Fracture. The Journal of Bone and Joint Surgery. American Volume, 4, 84. 2: Cummings, S. R., Rubin, S. M., & Black, D. (1990). The future of hip fractures in the United States. Numbers, costs, and potential effects of postmenopausal estrogen. Clinical Orthopaedics and Related Research, 252, 163-6. Kates, S. L. (2016). Hip fracture programs: are they effective?. Injury, 47. 3: Braithwaite, R. S., Col, N. F., & Wong, J. B. (2003). Estimating Hip Fracture Morbidity, Mortality and Costs. Journal of the American Geriatrics Society, 51, 3, 364-370. 4:Kates, S. L. (2016). Hip fracture programs: are they effective?. Injury, 47. • Programs of coordinated care have been developed with markedly improved outcomes 4 . • In January 2014, UCDMC introduced the multi- disciplinary Geriatric Fracture Program (GFP). • Retrospective chart review of 2012-2013 (pre-GFP, n=119) and 2014-2015 (post-GFP n=174) with the following data abstracted: • A predictive LOS (PLOS) model was created by a multivariate regression analysis with post-GFP data. • The model was retroactively applied to the pre-GFP group to assess improvements. • A threshold of PLOS + 1.5 days as a clinically- relevant cutoff for estimating if the GFP could have improved each patients actual LOS Variable Parameter estimate P-value Time to Surgery (Each midnight) + 0.14 0.0016 * ASA Score (one point increase) + 0.12 0.0689 CCI < 4 - 0.02 0.8058 Age (each decade > 82) + 0.002 0.5590 Gender (Female) - 0.04 0.5522 Cohorted on D14 (Ortho ward) - 0.08 0.2319 Initial INR <1.5 - 0.18 0.0844 No delirium - 0.19 0.0082 * Pre-GFP 2012-2013 (N=119) Post-GFP 2014-2015 (N=174) P-value Age 81.2 ± 8.4 82.0 ± 7.9 0.4585 Sex 83 F (69.8%) 118 F (67.8%) 0.7264 CCI < 4 89 (74.8%) 133 (76.4%) 0.7466 ASA Score 2: 17 (14.3%) 3: 71 (59.7%) 4: 31 (26.1%) 2: 18 (10.7%) 3: 117 (69.2%) 4: 34 (19.5%) 0.2862 Time to Surgery (Midnights) 0: 3 (2.5%) 1: 87 (73.1%) 2+: 29 (23.6%) 0: 8 (4.6%) 1: 113 (64.9%) 2: 53 (30.6%) 0.5331 INR (initial) INR < 1.5: 109 (91.6%) INR < 1.5: 149 (85.6%) 0.1221 Delirium 27 (22.7%) 74 (42.5%) 0.0004 * Ortho Ward 66 (55.5%) 100 (57.5%) 0.7332 Complications (not delirium) 42 (35.3%) 31 (17.8%) 0.0007 * Length of Stay 7.8 ± 6.0 5.9 ± 3.1 0.0023 * Delay to surgery (>2 midnights) 26 (21.9%) 52 (29.9%) 0.1264 • Actual LOS and complications significantly declined after initiation of the GFP; delirium was detected much more commonly • Using the PLOS model, 49.5% of patients in the pre-GFP group would have had decreased LOS under the GFP management (Figure 1). • Length of stay is a useful proxy for both quality of care, complications, and cost effectiveness in our geriatric fracture program. • This type of modeling is novel in this population and important for QI focus and hospital resource allocation. • Clinically modifiable variables that significantly impacted LOS included delirium prevention and decreasing time to surgery. • Our predictive model indicates that if the GFP was retroactively applied to the 2012-2013 patients nearly half would could have had a predicted LOS at least 1.5 days shorter. • There were decreased complications, excluding delirium, and length of stay after the GFP was applied to similar patients. • The apparent increase in delirium is likely an effect of the increased effort placed on nursing reporting of Confusion Assessment Method (CAM) scores mandated by the GFP. Discharge ED Table 1: Unadjusted Regression Analysis of LOS by Demographics and Clinical Characteristics. *statistical significance Table 2. Demographic and Clinical Characteristics of Geriatric Fracture Program Patients. *statistical significance Emergency Department • Quick Diagnosis to initiate GFP protocol • Fascia Iliaca (IF) Block • Consult Ortho/ Med Pre-op Admission • Optimize for OR • Medicine Team • Discharge Planning starts on hospital day 1 • Family Education Provided Post Op Care • PT/OT – early/often • Orthopaedic Nursing Care • Delirium prevention • Discharge Planning OR – ideally on day 1 • Orthopaedic surgery allows immediate weight-bearing • Anesthesia is spinal versus general and repeat IF block Figure 1: Predictive Length of Stay Model Applied to pre-GFP Patients with PLOS+1.5 Days set as Threshold for Clinical Relevance • Charlson Comorbidity Index (CCI) • Time to surgery (TtoS) • Delays to surgery (DTS) • Delirium • Other Complications • LOS • INR • ASA score • Ortho Ward • Demographics -2 0 2 4 6 8 10 12 14 16 18 Difference in Actual LOS –Predicted LOS (days) Affiliations 1: UC Davis School of Medicine, Sacramento, CA 95817 2: Department of Orthopaedics, UC Davis Medical Center, Sacramento, CA 95817 3: Quality and Safety, UC Davis Medical Center, Sacramento, CA 95817 Primary Project Mentors Philip Wolinsky MD and Garin Hecht MD Predictive Modeling for a Geriatric Hip Fracture Program as a Method of Assessing Outcomes Parker Goodell 1 MPH, Garin Hecht 2 MD, Trevor Shelton 2 MD, Christina Slee 3 MPH, and Philip Wolinsky 2 MD Patients > +1.5d of predicted LOS Patients < +1.5d of predicted LOS Clinical Threshold +1.5 days
Transcript of Aims Methods Results · • Discharge Planning starts on hospital day 1 • Family Education...
`Aims
Background
Methods Results
Results
Acknowledgements
Conclusions1) Comparepost- andpre-GFPpatientgroups
characteristicsandoutcomes.2) Applyapredictivemodeldevelopedfromtwo
yearsofGeriatricFractureProgram(GFP)patientstopre-GFPpatientstoaccountforpatientcomplexityinanalyzingtheoutcomesoftheUCDavisGFP.
• >300,000geriatrichipfracturesperyearintheUnitedStates,incidenceexpectedtoincrease1,2
• Highratesofcomplications,leadingtoprolongedhospitalizations3.
• $9-15bn/yr ofinpatientcosts4.• $26,000-$35,000average
hospitalizationcostalone
References1:Morris,A.H.,Zuckerman,J.D.,&AAOSCouncilofHealthPolicyandPractice,USA.AmericanAcademyofOrthopaedicSurgeons.(2002).NationalConsensusConferenceonImprovingtheContinuumofCareforPatientswithHipFracture. TheJournalofBoneandJointSurgery.AmericanVolume, 4,84.2:Cummings,S.R.,Rubin,S.M.,&Black,D.(1990).ThefutureofhipfracturesintheUnitedStates.Numbers,costs,andpotentialeffectsofpostmenopausalestrogen. ClinicalOrthopaedics andRelatedResearch, 252,163-6.Kates,S.L.(2016).Hipfractureprograms:aretheyeffective?. Injury, 47.3: Braithwaite,R.S.,Col,N.F.,&Wong,J.B.(2003).EstimatingHipFractureMorbidity,MortalityandCosts. JournaloftheAmericanGeriatricsSociety, 51, 3,364-370.4:Kates,S.L.(2016).Hipfractureprograms:aretheyeffective?. Injury, 47.
• Programsofcoordinatedcarehavebeendevelopedwithmarkedlyimprovedoutcomes4.
• InJanuary2014,UCDMCintroducedthemulti-disciplinaryGeriatricFractureProgram(GFP).
• Retrospectivechartreviewof2012-2013(pre-GFP,n=119)and2014-2015(post-GFPn=174)withthefollowingdataabstracted:
• ApredictiveLOS(PLOS)modelwascreatedbyamultivariateregressionanalysiswithpost-GFPdata.
• Themodelwasretroactivelyappliedtothepre-GFPgrouptoassessimprovements.
• AthresholdofPLOS+1.5daysasaclinically-relevantcutoffforestimatingiftheGFPcouldhaveimprovedeachpatientsactualLOS
Variable Parameterestimate
P-value
TimetoSurgery(Eachmidnight) +0.14 0.0016*
ASAScore(onepointincrease) +0.12 0.0689
CCI<4 - 0.02 0.8058Age(eachdecade>82) +0.002 0.5590
Gender(Female) - 0.04 0.5522Cohorted onD14(Orthoward) - 0.08 0.2319InitialINR<1.5 - 0.18 0.0844Nodelirium - 0.19 0.0082*
Pre-GFP2012-2013(N=119)
Post-GFP2014-2015(N=174)
P-value
Age 81.2± 8.4 82.0± 7.9 0.4585
Sex 83F(69.8%) 118F(67.8%) 0.7264
CCI<4 89(74.8%) 133(76.4%) 0.7466
ASAScore2:17(14.3%)3:71(59.7%)4:31(26.1%)
2:18(10.7%)3:117(69.2%)4:34(19.5%)
0.2862
TimetoSurgery(Midnights)
0:3(2.5%)1:87(73.1%)2+:29(23.6%)
0:8(4.6%)1:113(64.9%)2:53(30.6%)
0.5331
INR(initial) INR<1.5:109(91.6%)
INR<1.5:149(85.6%) 0.1221
Delirium 27(22.7%) 74(42.5%) 0.0004 *
OrthoWard 66(55.5%) 100(57.5%) 0.7332
Complications(notdelirium) 42(35.3%) 31(17.8%) 0.0007*
LengthofStay 7.8± 6.0 5.9± 3.1 0.0023*
Delaytosurgery(>2midnights) 26(21.9%) 52(29.9%) 0.1264
• ActualLOSandcomplicationssignificantlydeclinedafterinitiationoftheGFP;deliriumwasdetectedmuchmorecommonly
• UsingthePLOSmodel,49.5%ofpatientsinthepre-GFPgroupwouldhavehaddecreasedLOSundertheGFPmanagement(Figure1).
• Lengthofstayisausefulproxyforbothqualityofcare,complications,andcosteffectivenessinourgeriatricfractureprogram.
• ThistypeofmodelingisnovelinthispopulationandimportantforQIfocusandhospitalresourceallocation.
• ClinicallymodifiablevariablesthatsignificantlyimpactedLOSincludeddeliriumpreventionanddecreasingtimetosurgery.
• OurpredictivemodelindicatesthatiftheGFPwasretroactivelyappliedtothe2012-2013patientsnearlyhalfwouldcouldhavehadapredictedLOSatleast1.5daysshorter.
• Thereweredecreasedcomplications,excludingdelirium,andlengthofstayaftertheGFPwasappliedtosimilarpatients.
• TheapparentincreaseindeliriumislikelyaneffectoftheincreasedeffortplacedonnursingreportingofConfusionAssessmentMethod(CAM)scoresmandatedbytheGFP.
DischargeED
Table1:UnadjustedRegressionAnalysisofLOSbyDemographicsandClinicalCharacteristics.*statisticalsignificance
Table2.DemographicandClinicalCharacteristicsofGeriatricFractureProgramPatients.*statisticalsignificance
EmergencyDepartment• QuickDiagnosisto
initiateGFPprotocol• FasciaIliaca (IF)Block• ConsultOrtho/Med
Pre-opAdmission• OptimizeforOR• MedicineTeam• DischargePlanningstarts
onhospitalday1• FamilyEducationProvided
PostOpCare• PT/OT– early/often• OrthopaedicNursingCare• Deliriumprevention• DischargePlanning
OR– ideallyonday1• Orthopaedicsurgeryallows
immediateweight-bearing• Anesthesiaisspinalversus
generalandrepeatIFblock
Figure1: PredictiveLengthofStayModelAppliedtopre-GFPPatientswithPLOS+1.5DayssetasThresholdforClinicalRelevance
• Charlson ComorbidityIndex(CCI)• Timetosurgery(TtoS)• Delaystosurgery(DTS)• Delirium• OtherComplications
• LOS• INR• ASAscore• OrthoWard• Demographics
-2
0
2
4
6
8
10
12
14
16
18
Diffe
renceinActualLOS–P
redicted
LOS(days)
Affiliations1:UCDavisSchoolofMedicine,Sacramento,CA958172:DepartmentofOrthopaedics,UCDavisMedicalCenter,Sacramento,CA958173:QualityandSafety,UCDavisMedicalCenter,Sacramento,CA95817PrimaryProjectMentorsPhilipWolinsky MDandGarinHechtMD
PredictiveModelingforaGeriatricHipFractureProgramasaMethodofAssessingOutcomesParkerGoodell1 MPH,GarinHecht2 MD,TrevorShelton2 MD,ChristinaSlee3 MPH,andPhilipWolinsky2 MD
Patients>+1.5dofpredictedLOS
Patients<+1.5dofpredictedLOS
ClinicalThreshold+1.5days
