CIHI s Population Grouping Methodology: Beyond Predicting Costs Yvonne Rosehart Canadian Institute for Health Information October 12, 2017 yrosehart@cihi.ca cihi.ca @cihi_icis
CIHI s Population Grouping Methodology Day surgery, clinic and ED visits Inpatient stays Inpatient mental health stays LTC and CCC stays Physicians visits 2
POP grouper outputs Predicts for study period and 1 year into the future Clinical profile 226 health conditions Health profile group Functional status Health services accessed Demographics (age, sex, SES, geography) Predictive indicators Cost weights Number of primary care visits Number of ED visits Probability of admission to LTC 3
POP grouper outputs Predicts for study period and 1 year into the future Clinical profile 226 health conditions Health profile group Functional status Health services accessed Demographics (age, sex, SES, geography) Predictive indicators Cost weights Number of primary care visits Number of ED visits Probability of admission to LTC 4
Developing predictive models of selected service use Models Linear regression Logistic regression Predictor variables User types modelled separately 226 health conditions and influential 2-way interactions Age Sex Health system non-users Health system users with 0 health conditions Health system users with 1+ health conditions 5
Predicting future service needs 6
Predicted use of primary care services by health profile groups Number of predicted visits to family physician 16 14 12 10 8 6 4 2 0 Average number of predicted visits Z098 Z000 S053A C056A J001C Q077 H053A K053C Q079 S030C P001A D051C J055C S051A F055A B050A A013C F051A Q001 N004 G032C F001A C054A F052A R014 F053C E005A B030C K050C R021 S031C E011A R005 A033A O050C O030C R007 A039A H033C E007C H030C A036C F037A G031A S004C P002A J031A H035C K002C H004A R071 J033A S003A E014C H031C Q011 I002C E010C Q071 D001A 239 health profile groups (ranked from lowest to highest number of predicted visits) Source 2010 2011 and 2011 2012 Alberta data used in methodology development. 7 Provincial average
Predicted use of primary care services by health profile groups Health profile group Percent of cases Average number of pred. visits to FP Health system non-users 10.9% < 1 Open wound 0.9% 2 Other viral infection 0.2% 2 Otitis media 0.6% 2 Health system user without health conditions 4.5% 2 Acute myocardial infarction with heart failure + comorbidities 0.1% 12 Diabetes/hypoglycemia with PVD/other chronic vasc Dx + comorbidities 0.1% 13 Respiratory failure with heart failure 0.03% 13 Pulmonary hypertension 0.1% 13 Heart failure with CAD/arrhythmia + comorbidities 0.3% 15 Source 2010 2011 and 2011 2012 Alberta data used in methodology development. 8
Capacity planning for primary care: Gap analysis Province A B C D E Average number of visits 3.96 3.97 4.05 3.84 4.15 3.32 Average predicted visits 4.32 4.06 4.23 4.19 4.04 Standardized Utilization Index (SUI) (average actual visits/provincial average) Standardized Needs Index (SNI) (average predicted visits/provincial average) Need Use Gap (SUI SNI) 1.00 1.02 0.97 1.05 0.84 1.09 1.03 1.07 1.06 1.02-0.09-0.01-0.10-0.01-0.18 Source 2010 2011 and 2011 2012 Alberta data used in methodology development. 9
Profiling urban/rural variation in ED service In rural areas, 45% of people are visiting the ED more than we would expect, based on their health needs. Compare actual and predicted ED visits Rural Urban Total Actual < Predicted 17% 32% 29% Actual Predicted 39% 50% 48% Actual > Predicted 45% 18% 23% Source 2010 2011 and 2011 2012 Alberta data used in methodology development. 10
Predicting use of primary care and ED services among people with diabetes 6 7 8 10 11 0.4 0.4 0.6 0.7 0.8 1.0 1.2 1.4 1.5 1.7 12 12 13 11 14 14 16 2.2 18 18 0 1 2 3 4 5 6 7 8 9 10 14 15 19 20 24 Number of additional chronic conditions Source 2010 2011 and 2011 2012 Alberta data used in methodology development. 3.3 4.3 Median number of predicted visits to... Family medicine physician Emergency department
LTC admission indicator Ventile POP grouper prediction Number of people Average predicted number of admissions Average actual number of admissions Actual Proportion of admissions 1 153K 63 93 0.3% 2 153K 133 52 0.1% 3 153K 173 82 0.2% 4 153K 201 125 0.3% 17 153K 1,922 1,923 5.4% 18 153K 2,643 3,066 8.6% 19 153K 4,263 6,063 17.0% 20 153K 18,994 18,845 52.7% Source 2010 2011 and 2011 2012 Alberta data used in methodology development. 12 53% of LTC admissions in the top 5%
Use of population grouper at points of care Pred. # of FP visits FY2017 Prob. Admit to LTC FY2017 Pred. # of ED visits FY2016 Pred. # of ED visits FY2017 13
cihi.ca yrosehart@cihi.ca @cihi_icis