Karen Tu MD, MSc, CCFP, FCFP Scientist ICES Associate Professor DFCM, University of Toronto Active Staff Toronto Western Hospital FHT

Transcription

1 Validation of administrative data algorithms to determine population prevalence and incidence of Alzheimer's disease, dementia, MS, epilepsy and Parkinson's disease Karen Tu MD, MSc, CCFP, FCFP Scientist ICES Associate Professor DFCM, University of Toronto Active Staff Toronto Western Hospital FHT NHCC Webinar May 29, 2013 Electronic Medical Record Administrative data Linked Database (EMRALD)

2 Team Members Family Physicians Karen Tu Epilepsy Liisa Jaakkimainen Alzheimer s/dementia Debra Butt Parkinson s Disease/Parkinsonism Noah Ivers Multiple Sclerosis Neurologists Nathalie Jette Epilepsy Mark Guttman Parkinson s Disease Paul O Connor Multiple Sclerosis Psychiatrist Nathan Herrmann Alzheimer s/dementia Neuropsychologist Mary Tierney Alzheimer's/Dementia Family Physician/Geriatric Specialist Rob Lam Parkinson s Disease ICES Staff Jacqueline Young - Health Informatician Diane Green Analyst Myra Wang Research Coordinator Rob Turner Research Assistant

3 Objective Overview Develop and validate administrative data algorithms for the most common priority neurological conditions in order to be able to determine population prevalence of these conditions in Ontario and across Canada Methods A retrospective chart abstraction study to develop a reference standard using primary care physician patient records from EMRALD

4 Administrative Data Available in some form in every province and territory in Canada Hospitalizations Canadian Institute for Health Information (CIHI) Physician billings Ontario Health Insurance Plan (OHIP) Prescriptions Ontario Drug Benefit (ODB) for 65 and over only

5 Electronic Medical Record Administrative data Linked Database (EMRALD) Family physician EMR records Prescribed entity under PHIPA Administrative data holdings for the province of Ontario

6 EMRALD Geographic Distribution In this study: 83 physicians (26 clinics) Average length of time on EMR 5.6 years Patients 20 years and older: 73,003 Patients 65 years and older: 14,726

7 Random Sampling Epilepsy Dementia 73, Eligible patients 20 years and older Randomly selected for abstraction 73,003 7,500 3,404 Eligible patients 20 and older Randomly chosen for abstraction Aged 65 and older MS Targeted Sampling Parkinsonism 73,003 Eligible patients 20 years and older 73,003 Eligible patients 20 years and older 936 Abstracted all patients found based on positive hit from search terms: multiple sclerosis, MS, optic neuritis, acute transverse myelitis, demyelinating 2,319 Abstracted all patients found based on positive hit from search terms: Parkinson*, PD, paralysis agitans, cogwheel*, Lewy body

8 Multiple Sclerosis

9 Documentation of MS Evidence within the EMR % N = 237

10 Description H or 3 P in 2 YRS MS Validation Algorithms Sensitivity 91.1 ( ) Specificity 99.9 PPV 73.5 ( ) NPV ( ) Youden Kappa Prev Est % H or 3 P in 3 YRS 91.6 ( ) ( ) ( ) % H or 3 P in 4 YRS 92.0 ( ) ( ) ( ) % H or 4 P in 2 YRS 88.2 ( ) ( ) ( ) % H or 4 P in 3 YRS 90.3 ( ) ( ) ( ) % H or 4 P in 4 YRS 90.7 ( ) ( ) ( ) % H or 5 P in 2 YRS 84.0 ( ) 99.9 ( ) 82.2 ( ) 99.9 ( ) % H or 5 P in 3 YRS 86.5 ( ) 99.9 ( ) 82.3 ( ) ( ) % H or 5 P in 4 YRS 87.3 ( ) 99.9 ( ) 82.1 ( ) ( ) % H or 7 P in 20 YRS 86.1 ( ) ( ) 86.8 ( ) ( ) % N = 73,003, n = 237, prevalence = 0.32%

11 Epilepsy

12 Documentation of Epilepsy within the EMR % Remission Drugs Family Physician Only Neurologist Other Specialist Epilepsy Only Seizure Disorder Only Mixed Neither Documented By N= 95 Terminology

13 Idiopathic Metabolic- Structural Unknown Generalized Only Focal Only Mixed Unknown Neonate/ Infant Child Adolescent Adult Unknown Documentation of Epilepsy within the EMR % Etiology Seizure Type Age of Onset N= 95

14 Epilepsy Validation Algorithms Description H or 2 P in 2 YRS H or 2 P (>30 day separation) in 2 YRS H or 2 P in 3 YRS H or 2 P (>30 day separation) in 3 YRS H or 3 P in 2 YRS H or 3 P (>30 day separation) in 2 YRS H or 3 P in 2 YRS H or 3 P (>30 day separation) in 3 YRS H or 4 P (>30 day separation) in 2 YRS H or 4 P (>30 day separation) in 3 YRS H or 5 P (>30 day separation) in 2 YRS H or 5 P (>30 day separation) in 3 YRS Sensitivity 85.3 ( ) 81.1 ( ) 86.3 ( ) 82.1 ( ) 76.8 ( ) 73.7 ( ) 76.8 ( ) 73.7 ( ) 68.4 ( ) 70.5 ( ) 64.2 ( ) 64.2 ( ) Specificity 99.2 ( ) 99.4 ( ) 99.2 ( ) 99.4 ( ) 99.5 ( ) 99.8 ( ) 99.5 ( ) 99.7 ( ) PPV 58.3 ( ) 65.3 ( ) 58.2 ( ) 65.0 ( ) 68.2 ( ) 79.5 ( ) 68.2 ( ) 77.8 ( ) 81.3 ( ) 81.7 ( ) 82.4 ( ) 82.4 ( ) NPV ( ) ( ) 99.7 ( ) 99.7 ( ) 99.7 ( ) 99.6 ( ) 99.6 ( ) 99.5 ( ) 99.5 ( ) Youden Index Kappa Estimated Prevalence % % % % % % % % % % % % N = 7500, n = 95, prevalence = 1.3%

15 Parkinsonism

16 Parkinson s Disease vs Atypical/Secondary Parkinson s within the EMR Overall Parkinsonism n = 231 (0.3%) 17% 83% Parkinson's Disease (n=191) Atypical/Secondary Parkinson's (n=40) Drug-induced PKM (15) Lewy body disease (10) Stroke or CVA/vascular PKM (4) Huntington s PKM (3) PKM, not specified (5) Progressive supranuclear palsy (2) Multiple system atrophy (1) Corticobasal degeneration (1) Multiple reasons (1)

17 Characteristics of Parkinsonism Patients Characteristics Parkinsonism n=231 Parkinson s Disease n=191 Secondary Parkinson s n=40 65 years (n, %) 191 (82.7%) 162 (84.8%) 29 (72.5%) Age in years (mean ± SD) 74.0 (10.2) 74.5 (9.6) 71.7 (12.1) Male (n, %) 139 (60.2%) 121 (63.4%) 18 (45.0%) Neurologist (n, %) 154 (66.7%) 135 (70.7%) 19 (47.5%) 1 Anti-PD Drug (n, %) 180 (77.9%) 165 (86.4%) 15 (37.5%) Symptom Onset Date (n, %) 115 (49.8%) 97 (50.8%) 18 (45.0%) Diagnosis Date (n, %) 137 (59.3%) 119 (62.3%) 18 (45.0%) PD medications: levodopa (carbidopa/levodopa, levodopa/benserazide, levodopa/carbidopa/entacapone) MAO-B inhibitors (selegiline, rasagiline), dopamine agonists (pergolide, bromocriptine, pramipexole, ropinirole), COMT inhibitors (entacapone, tolcapone) N = 73,003, n = 231, prevalence = 0.3%

18 20 Years Parkinsonism Validation Algorithms (30 Day Skip) Algorithm Sensitivity Specificity PPV NPV Youden Kappa Prevalence Estimate H 25.9% ( ) 100.0% ( ) 75.9% ( ) 99.8% ( ) % H or P 82.3% ( ) 99.6% ( ) 42.5% ( ) ( ) % H or 2 P in 1 YR 71.6% ( ) 73.8% ( ) % 2 P in 1 YR 70.3% ( ) ( ) 79.5% ( ) % H or 2 P in 2 YR 72.8% ( ) 72.2% ( ) % 2 P in 2 YR 71.6% ( ) ( ) 77.2% ( ) % H or 2 P in 3 YR 72.8% ( ) 71.0% ( ) % 2 P in 3 YR 72.0% ( ) 75.9% ( ) % H or 3 P in 1 YR 62.5% ( ) 100.0% ( ) 80.6% ( ) % 3 P in 1 YR 58.6% ( ) 100.0% ( ) 88.3% ( ) ( ) % N = 73,003, n = 231, prevalence = 0.3%

19 Impact of time interval requirement for 2 nd physician billing Algorithm Sensitivity Specificity PPV > 20 No time interval requirement between physician billings 2 P in 1 YR (no skip) 72.8% ( ) 72.8% ( ) 30 day skip between physician billings 2 P in 1 YR 70.3% 79.5% (30 day skip) ( ) ( ) ( ) >65 No time interval requirement between physician billings 2 P in 1 YR 75.0% 99.7% 76.6% (no skip) ( ) ( ) ( ) 30 day skip between physician billings 2 P in 1 YR 72.9% 99.8% 82.8% (30 day skip) ( ) ( ) NPV 99.7% ( ) 99.6% ( ) Youden Kappa Prevalence Estimate % % % % 20 years and older; N = 73,003, n = 231, prevalence = 0.3% 65 years and older; N = 14,159, n = 192, prevalence = 1.3%

20 Impact of Drugs on Parkinsonism Algorithm 1 Rx 1 Rx 1 Rx AND 1 P 1 Rx AND 1 P 1 Rx AND 1 P +/- 6 months 1 Rx AND 1 P +/- 6 months 2 P in 1 YR OR 1 Rx and 1 P 2 P in 1 YR OR 1 Rx and 1 P 2 P in 1 YR OR 1 Rx and 1 P +/- 6 months 2 P in 1 YR OR 1 Rx and 1 P +/- 6 months Sensitivity 70.7% ( ) 81.3% ( ) 66.4% ( ) 76.6% ( ) 65.5% ( ) 75.5% ( ) 78.0% ( ) 81.8% ( ) 77.6% ( ) 81.3 ( ) Algorithms Specificity 99.6% ( ) 98.3% ( ) 100.0% ( ) 99.8% 100.0% ( ) ( ) 99.7% ( ) 99.7 ( ) 20 years and older; N=73003, n=231; prevalence=0.3% 65 years and older; N=14159, n=192, prevalence=1.3% PPV 35.7% ( ) 39.5% ( ) 83.7% ( ) 84.5% ( ) 86.4% ( ) 87.3% ( ) 75.4% ( ) 78.1% ( ) 76.9% ( ) 80.0 ( ) NPV 99.7% ( ) 99.7% ( ) Youden Kappa Prevalence Estimate % % % % % 99.7% ( ) % % 99.8% ( ) % 99.7 ( ) % %

21 Parkinsonism Validation Algorithms 20 Years vs 65 Years Comparison Algorithm Sensitivity Specificity PPV NPV Youden Kappa Prevalence Estimate No Drugs > 20 >65 2 P in 1 YR 2 P in 1 YR With Drugs > 20 2 P in 1 YR OR 1 Rx +/- 6 months 1 P >65 1 Rx AND 1 P in 1 YR 70.3% ( ) 72.9% ( ) 77.6% ( ) 75.5% ( ) ( ) 99.8% ( ) 79.5% ( ) 82.8% ( ) 76.9% ( ) 87.3% ( ) 99.6% ( ) 99.7% ( ) % % % % 20 years and older; N = 73,003, n = 231, prevalence = 0.3% 65 years and older; N = 14,159, n = 192, prevalence = 1.3%

22 Dementia

23 Alzheimer s Disease vs Other Dementias within the EMR 20 years and older n = 60 (0.8%) 65 years and older n = 150 (4.4%) 38% Alzheimer's Disease (n=37) 51% Alzheimer's Disease (n=73) 62% Other Dementias (n=23) 49% Other Dementias (n=77)

24 100 Alzheimer s Disease and Dementia Documentation in the Patient Chart ( 65 years and older) 80 % Geriatrics Neurology Psychiatry Multiple Dementia Specialists Documentation By Inpatient Note Or Other Specialist FP Only Drugs Only Alzheimer's Disease (n=73) Dementia (n=77) Alzheimer's Disease or Dementia (n=150) N = 3404, n = 150, prevalence = 4.4%

25 65 Years Dementia Validation Algorithms (30 Day Skip) Algorithm Sensitivity Specificity PPV NPV Youden Kappa Prevalence Estimate H H or 1 P H or 2 P in 1 YR H or 2 P in 2 YRS H or 2 P in 3 YRS H or 3 P in 1 YR H or 3 P in 2 YRS H or 3 P in 3 YRS 23.3% ( ) 85.3% ( ) 74.7% ( ) 76.0% ( ) 76.7% ( ) 54.0% ( ) 61.3% ( ) 62.7% ( ) 99.6% ( ) 94.3% ( ) 98.6% ( ) 98.4% ( ) 98.4% ( ) 99.3% ( ) 99.1% ( ) 99.1% ( ) 74.5% ( ) 41.0% ( ) 71.3% ( ) 68.7% ( ) 68.5% ( ) 77.9% ( ) 76.7% ( ) 77.0% ( ) 96.6% ( ) 99.3% ( ) 98.8% ( ) 98.9% ( ) 98.9% ( ) 97.9% ( ) 98.2% ( ) 98.3% ( ) % % % % % % % % N = 3404, n = 150, prevalence = 4.4% N =

26 65 Years Dementia Validation Algorithms Impact of 30 Day Skip and Drugs Algorithm Sensitivity Specificity PPV NPV Youden Kappa Prevalence Estimate Impact of 30 day skip between physician billings H or 3 P in 2 YRS 65.3% ( ) 98.8% ( ) 71.0% ( ) 98.4% ( ) % H or 3 P in 2 YRS (30 day skip) 61.3% ( ) 99.1% ( ) 76.7% ( ) 98.2% ( ) % Impact of adding medications H or 3 P in 2 YRS OR 1 Rx H or 3 P in 2 YRS OR 1 Rx (30 day skip) 80.7% ( ) 79.3% ( ) 98.7% ( ) 99.1% ( ) 74.7% ( ) 80.4% ( ) 99.1% ( ) 99.0% ( ) % % N = 3404, n = 150, prevalence = 4.4%

27 Dementia Validation Algorithms 20 Years vs 65 Years Comparison Algorithm Sensitivity Specificity PPV NPV Youden Kappa Prevalence Estimate No Drugs > 20 H or 3 P in 2 YRS 55.0% ( ) 99.8% 73.3% ( ) 99.6% ( ) % >65 H or 3 P in 2 YRS 61.3% ( ) 99.1% ( ) 76.7% ( ) 98.2% ( ) % With Drugs > 20 H or 3 P in 2 YR OR 1 Rx 75.0% ( ) 99.8% 78.9% ( ) 99.8% % >65 H or 3 P in 2 YR OR 1 Rx 79.3% ( ) 99.1% ( ) 80.4% ( ) 99.0% ( ) % 20 years and older; N = 7500, n = 60, prevalence = 0.8% 65 years and older; N = 3404, n = 150, prevalence = 4.4%

28 Summary of Optimal Algorithms Condition Algorithm Multiple Sclerosis H or 5 P in 2 YRS Epilepsy H or 3 P* in 2 YRS Parkinsonism 2 P* in 1 YR 2 P* in 1 YR OR 1 Rx and 1 P +/- 6 months Parkinsonism 2 P* in 1 YR 1 Rx AND 1 P in 1 YR Dementia H or 3 P* in 2 YRS H or 3 P* in 2 YRS OR 1 Rx Dementia H or 3 P* in 2 YRS H or 3 P* in 2 YRS OR 1 Rx Sensitivity 84.0% ( ) 73.7% ( ) 70.3% ( ) 77.6% ( ) 72.9% ( ) 75.5% ( ) 55.0% ( ) 75.0% ( ) 61.3% ( ) 79.3% ( ) Specificity ( ) 99.8% ( ) ( % ( ) 99.8% 99.8% 99.1% ( ) 99.1% ( ) PPV 82.2% ( ) 79.5% ( ) 79.5% ( ) 76.9% ( ) 82.8% ( ) 87.3% ( ) 73.3% ( ) 78.9% ( ) 76.7% ( ) 80.4% ( ) NPV 100.0% ( ) 99.7% ( ) 99.6% ( ) 99.7% ( ) 99.6% ( ) 99.8% 98.2% ( ) 99.0% ( ) Youden Index Prevalence Kappa Estimate Cohort Prevalence 0.3% 0.3% Adult (20+) 1.2% 1.3% Adult (20+) % 0.3% % 0.3% Adult (20+) % 1.3% % 1.3% Elderly (65+) % 0.8% % 0.8% Adult (20+) % 4.4% % 4.4% Elderly (65+)

29 Rate per 100 Annual Multiple Sclerosis, Epilepsy, Parkinsonism, and Dementia Age-Adjusted Cumulative Prevalence Rate for Adults 20 Years of Age Year MS Epilepsy Parkinsonism Dementia

30 Rate per 1000 Multiple Sclerosis, Epilepsy, Parkinsonism, and Dementia Age-Adjusted Incidence for Adults 20 Years of Age Year MS Epilepsy Parkinsonism Dementia