Slides - Navid Ghaffarzadegan

Why Are Patients TreatedDifferently?A Conceptual Framework forClassifying Practice Variation inMedicineNavid Ghaffarzadegan (navidg@gmail.com)Erika G. MartinUniversity at Albany, State University of New YorkGhaffarzadegan and Martin (2011) 1Introduction• Navid Ghaffarzadegan, PhD candidate in Public Policy• Decision and Policy Sciences– How people make decisions– How decisions influence government performance.• Medical Decision Making, health policy implications• Practice variation in Medicine• Approach: Behavioral decision makingGhaffarzadegan and Martin (2011) 2Introduction• The concept of practice variation• Evidence: cesarean section surgeries,cancer diagnosis and treatments,cardiovascular care, pneumonia diagnosis,pediatric services, psychiatric services,etc.• Public administration/policy relevanceGhaffarzadegan and Martin (2011) 31

Introduction• Common causes of practice variation– Disease characteristics.– Variation in finance structures.– Physician characteristics.– Patient characteristics.– Regional and organizational factors.– These factors ultimately cause variation throughdecision making factors, thus it is important tounderstand cognitive limitations in medical decisionmakingGhaffarzadegan and Martin (2011) 4The practice variation projectCurrent presentationCurrent stage inthe project1. Literature Review:Reviewing the literatureof MDM, public health,and decision making onpractice variation2. Modeling:Building a simulationmodel of physicians’decision making for c-section3. Data Replication:Investigating themodel’s capabilities toreplicate real world data5. Policy Analysis:Using the calibratedmodel to investigateeffects of several policyrecommendations4. Model Improvement:Improving andcalibrating the modelwith dataGhaffarzadegan and Martin (2011) 5Problem Definition• This study:– Focuses on decision making determinant ofpractice variation.• Main questions:– How has practice variation been studied?– How can we make sense of the literature?• Proposes an analytic framework ofbehavioral decision making determinantsof practice variation.Ghaffarzadegan and Martin (2011) 62

Methods• Literature Review• Searched PubMed for "practice variation," whichidentified 333 articles.• Narrowed articles to those published in coremedical journals.• Supplemented our search with articles from asimilar keyword search in PsychINFO andGoogle Scholar, and the Dartmouth Atlas ofHealth Care.• Finally, we tracked key citations from the articlesreviewed. Our final sample included 75 articles.Ghaffarzadegan and Martin (2011) 7Methods• The open coding revealed two keyfindings:– First, several behavioral decision makingphenomena play important roles in practicevariation.– Second, there is a conceptual differenceacross studies of practice variation in howresearchers categorize, describe, and studypractice variation.Ghaffarzadegan and Martin (2011) 8FrameworkTable (1): Different forms of practice variation among medically similar patientsWithin–patientWithin–physicianUnreliabilityInconsistency in how a physiciandiagnoses and treats a single patient overtime.Between–physicianDisagreementDifferent styles of practice anddisagreement among physicians onhow to diagnose and treat a specificcase.Between–patientBiasTreating patients within a practicedifferently based on variation inpatients’ preferences, and/or physicians’bias toward a group of patients.Interactive effectsDifferent doctors treat different groupsof patients with different preferences,and different groups of patients choosedifferent doctors.Ghaffarzadegan and Martin (2011) 93

Within–Physician/Within–PatientVariation: Unreliability• when physicians are inconsistent about decisionsthat they make for the same patient over time.• The amount of unreliability may vary acrossphysicians.• Einhorn’s study: Medical pathologists reviewingbiopsy slides.• Kirwan et al.: British rheumatologists reviewingpatient vignettes.Ghaffarzadegan and Martin (2011) 10Within–Physician/Within–PatientVariation: Unreliability• Unreliability is produced during two differentstages of judgment:– information acquisition– information processing• Importance of how information is representedand the ease of accessing critical pieces ofinformation influence decision making accuracy.• Initial judgment, confirmation bias (Kostopoulouet al. 2009).Ghaffarzadegan and Martin (2011) 11FrameworkTable (1): Different forms of practice variation among medically similar patientsWithin–patientWithin–physicianUnreliabilityInconsistency in how a physiciandiagnoses and treats a single patient overtime.Between–physicianDisagreementDifferent styles of practice anddisagreement among physicians onhow to diagnose and treat a specificcase.Between–patientBiasTreating patients within a practicedifferently based on variation inpatients’ preferences, and/or physicians’bias toward a group of patients.Interactive effectsDifferent doctors treat different groupsof patients with different preferences,and different groups of patients choosedifferent doctors.Ghaffarzadegan and Martin (2011) 124

Between–physician/Within–patient Variation: Disagreement• when different physicians vary in their diagnosisand treatment for a single patient.• These studies control for patient characteristics.• Way et al.: psychiatrists reviewing tapedinterviews of patients at urban psychiatricemergency services.– considerable disagreement.– Disagreement varies in diagnosing different diseases.Ghaffarzadegan and Martin (2011) 13Between–physician/Within–patient Variation: Disagreement• Gonzalez-Vallejo et al.:– Acceptable level of agreement across pediatricianson their judgment of the probability of acute otitismedia– low agreement on the type of antibiotic.• Follow up study, Sorum et al.:– physicians place different weights on availableinformation cues when making a diagnosis.– Have different decision making models to analyzedata and recommend treatments.– No cross-national differences.Ghaffarzadegan and Martin (2011) 14Between–physician/Within–patient Variation: Disagreement• Epstein and Nicholson (2009): c-section surgery.– variation in risk adjusted cesarean surgery rate inNew York and Florida.– Residency programs a minimal explanation.– More variation within an area.– 30% of variation in risk-adjusted cesarean rates isdue to non-observable physician related factors.– Do not converge over time to a community norm orstandard.Ghaffarzadegan and Martin (2011) 155

FrameworkTable (1): Different forms of practice variation among medically similar patientsWithin–patientWithin–physicianUnreliabilityInconsistency in how a physiciandiagnoses and treats a single patient overtime.Between–physicianDisagreementDifferent styles of practice anddisagreement among physicians onhow to diagnose and treat a specificcase.Between–patientBiasTreating patients within a practicedifferently based on variation inpatients’ preferences, and/or physicians’bias toward a group of patients.Interactive effectsDifferent doctors treat different groupsof patients with different preferences,and different groups of patients choosedifferent doctors.Ghaffarzadegan and Martin (2011) 16Within–Physician/Between–Patient Variation: Bias• when a single physician performs different practicesfor medically similar patients.• Discrimination?• Physicians may consider patients’ demographiccharacteristics in their diagnostic and treatmentdecisions.• For example, the prevalence of type II diabetes is50% higher among non–Hispanic blacks, compared tonon–Hispanic whites.Ghaffarzadegan and Martin (2011) 17Within–Physician/Between–Patient Variation: Bias• Racial minorities higher preference for endof-lifetreatments.• The IOM reviewed the evidence for biasand disparities in different diseases andhealth care services.– Uncertainties (e.g., communication problems),– Decision making biases.– Increase awareness.Ghaffarzadegan and Martin (2011) 186

FrameworkTable (1): Different forms of practice variation among medically similar patientsWithin–patientWithin–physicianUnreliabilityInconsistency in how a physiciandiagnoses and treats a single patient overtime.Between–physicianDisagreementDifferent styles of practice anddisagreement among physicians onhow to diagnose and treat a specificcase.Between–patientBiasTreating patients within a practicedifferently based on variation inpatients’ preferences, and/or physicians’bias toward a group of patients.Interactive effectsDifferent doctors treat different groupsof patients with different preferences,and different groups of patients choosedifferent doctors.Ghaffarzadegan and Martin (2011) 19Between–Physician/Between–PatientVariation: Interactive Effects• Tape et al.: emergency department physicians atthree sites (Illinois, Nebraska, Virginia) diagnosepneumonia.– Physicians at each site use clinical findings differently– they do so in a way that is consistent with the bestpossible decision in their location.• Within a region, different physicians may nottreat similar patients.– Heterogeneous distributions of patients acrossdoctors.• Insurance providers add to heterogeneity ofpatient distribution.Ghaffarzadegan and Martin (2011) 20Between–Physician/Between–PatientVariation: Interactive Effects• Different doctors treat different groups ofpatients with different preferences, anddifferent groups of patients choosedifferent doctors.• Building a cognitive reference class:consists of patients that they treat in theirdaily practices.• Patients cognitive reference style ofpractice PatientsGhaffarzadegan and Martin (2011) 217

Causal MechanismsCausal explanations within a patient:UncertaintyImperfect informationNo guidelinesUnclear outcome feedbackCausal explanations within a physician:Unreliability/inconsistenciesDiagnosis and treatment errorsWorkloadTime pressure, stress, moodReliance on intuitionWithin–Physician/Within–Patient Between–Physicians/Within–PatientUnreliabilityDisagreementPracticeVariationWithin–Physician/Between– Between–Physicians/Between–PatientsPatientsBiasInteractive effectsCausal explanations across physicians:Physician characteristics and valuesTrainings and skillsLevels of experienceOrganizational affiliationsCausal explanations across patients:Patient characteristics and valuesPatient preferencesHealth risksCommunity characteristicsGhaffarzadegan and Martin (2011) 22Discussions• Desired versus undesired forms ofpractice variation• Policy tools to reduce undesired practicevariation• Avenues for Future ResearchGhaffarzadegan and Martin (2011) 23Future worksCurrent presentationCurrent stage inthe project1. Literature Review:Reviewing the literatureof MDM, public health,and decision making onpractice variation2. Modeling:Building a simulationmodel of physicians’decision making for c-section3. Data Replication:Investigating themodel’s capabilities toreplicate real world data5. Policy Analysis:Using the calibratedmodel to investigateeffects of several policyrecommendations4. Model Improvement:Improving andcalibrating the modelwith dataGhaffarzadegan and Martin (2011) 248

Conclusion• From a behavioral decision making perspective,there are four conceptually distinct forms ofpractice variation: unreliability, disagreement,bias, and interactive effects.• Practice variation is undesired when it is notbased on patients’ preferences or on evidenceabout differential disease burden acrosspopulations. That can happen in the first threeforms of practice variation (unreliability,disagreement, and bias).Ghaffarzadegan and Martin (2011) 25Conclusion• Practice variation in the form of interactiveeffects can be a natural response ofphysicians who adapt to their patientpopulation.• Policies should address undesired formsof practice variation, while allowingpractice variation to occur whenappropriate.Ghaffarzadegan and Martin (2011) 26Thanks!• Email: navidg@gmail.com• The paper is available online:– http://navidg.com/research.htmGhaffarzadegan and Martin (2011) 279