Economic outcomes: Method for implementa5on

Transcription

1 Economic outcomes: Method for implementa5on Philippe Beutels Centre for Health Economics Research & Modelling Infec<ous Diseases, University of Antwerp, Belgium 17-18th October 2016

2 General cost categories Direct costs: Health care costs for diagnosing, trea<ng and following up cases (in and out of hospital) Personal costs (eg, transport) Indirect costs Time costs (oren termed produc<vity costs) costs = (number of cases) x (cost / average case)

3 Es5ma5ng health care system and pa5ent costs Claims or insurer administra5ve databases (+) all costs (not just hospital), but no copayments (- ) iden<fying cases (linkage), and selec<ng only relevant medical acts 3

4 Retrospec<ve matched case- control cost study in an insurer database: example IPD <5 years of age > 65 years of age Direct health care costs ( ) cases controls Time (months) IPD diagnosis Time (months) 4 Creemers et al, J Applied Sta<s<cs 2011

5 Es5ma5ng health care system and pa5ent costs Claims or insurer administra5ve databases (+) all costs (not just hospital), but no copayments (- ) iden<fying cases (linkage), and selec<ng only relevant medical acts Centralised hospital databases (+) diagnosis oren ICD code based, no linkage required (- ) only hospital costs; oren mul<ple diagnoses, may be entered by non- trea<ng staff 5

6 Age specific hospitalisa<on costs for primary diagnosis influenza (all hospitals in Belgium, 2011) Beutels P et al, 2013

7 Distribu<on of in- hospital costs, primary diagnosis influenza Length of stay important, but also pharma and medical deliveries Cost category Description HOSP Hospitalization cost, cost of staying at the hospital BPMR DELIVER PHARMA IMPL Blood Plasma, Mother s milk, radio-isotopes Medical deliveries Pharmaceutical products Implantations CM&NM Clinical microbiology and nuclear medicine Beutels P et al, 2013

8 Es5ma5ng health care system and pa5ent costs Claims or insurer administra5ve databases (+) all costs (not just hospital), but no copayments (- ) iden<fying cases (linkage), and selec<ng only relevant medical acts Centralised hospital databases (+) diagnosis oren ICD code based, no linkage required (- ) only hospital costs; oren mul<ple diagnoses Restrospec5ve/cross sec5onal survey (+) all costs, including copayments; poten<ally also costs of non- care seeking pa<ents (- ) reliability of diagnosis; recall bias 8

9 Es5ma5ng health care system and pa5ent costs Claims or insurer administra5ve databases (+) all costs (not just hospital), but no copayments (- ) iden<fying cases (linkage), and selec<ng only relevant medical acts Centralised hospital databases (+) diagnosis oren ICD code based, no linkage required (- ) only hospital costs; oren mul<ple diagnoses Restrospec5ve/cross sec5onal (+) all costs, including copayments; poten<ally also costs of non- care seeking pa<ents (- ) reliability of diagnosis; recall bias Prospec5ve data collec5on (cohort, trial) (+) poten<ally accurate case defini<on, different perspec<ves can be included (- ) costly and <me consuming 9

10 Likely easiest to collect invoice data in GIHSN hospitals Obtain invoice amounts per pa<ent- admission from accountancy/billing department at each site : Total charge to pa<ent and their insurer Hotel costs (and LOS) Medica<on costs Diagnos<c costs Interven<on costs (eg surgery) Days per type of ward (IC, general ward) ILI/flu status Available detail depends on health system and hospital - Local availability depends on health care system (may need per- diem es<mate to apply to LOS)

11 Opportuni5es for analysis Rela<vely easy to assess cost charges in terms of : Differences per pa<ent/ admission for ILI, flu (Lab confirmed) and ICD- based à poten<al valida<on of cost es<ma<ons at the country level through ICD Differences over <me, and between study sites/countries à arer adjus<ng for infla<on (using Consumer Price Indices) and purchasing power (using PPP) Differences between pathogens (eg RSV) Correla<ons between LOS and charges à poten<al valida<on if only LOS is available Correla<ons between severity/strains and the above à understand inter- season variabiity in costs per pa<ent 11

12 More difficult as part of GIHSN Insurer database analysis requires linkage, and collabora<on of large insurer (eg na<onal health insurance), but poten<ally interes<ng comparisons all costs, incl follow up arer separa<on in such databases: Flu (strains) versus ILI, flu vs other pathogens pregnant study women (with flu, ILI, other pathogens) vs other matched pregnant women Na<onal or regional centralised hospital database analysis (based on ICD) requires existence of a centralised database and access to it, and is a separate study altogether, mainly interes<ng to assess the representa<veness of study sites

13 Retrospec5ve surveys? Follow- up pa<ents (survivors) to survey their experience arer separa<on(health care costs but may also include indirect costs of impaired produc<vity) : - Labour intensive and costly, hard to reach pa<ents, hospital costs likely more important than other costs in hospitalised pa<ents, poten<al recall bias.

14 General limita5ons remain Sites may not be representa<ve for a region/country in terms of caseload, treatment prac<ce and costs Privacy/ethical approval? Commercial interests/compe<<on between hospitals? Difficult to survey follow- up costs arer separa<on (but less influen<al for burden of disease and cost- effec<veness analysis)

15 Conclusion Opportunity to collect informa5on on hospital charges to inform APributable cost of illness, BoD, cost- effec5veness Poten5al opportunity to also collect info on non- hospital costs and indirect costs for hospitalised pa5ents Representa5veness is an important limita5on Check out WHO guide and manual on cost burden es5ma5on and CEA for flu

16 Resource use from a retrospec5ve ILI survey (n=2250) COM, ILI (n=1107) AMB, ILI (n=1116) AMB, FLU (n=429) % of respondents who took medica<on: against fever 28% 42% 50% against pain 30% 35% 38% an<- inflammatory 14% 29% 32% an<bio<cs 4% 45% 38% an<- virals 5% 8% 15% against cough 41% 48% 52% against sore throat 38% 39% 39% nose spray 45% 45% 45% other 9% 16% 12% Number of medical visits: NA 1.6 (1) [0-19] 1.6 (1) [0-13] GP consults NA 1.1 (1) [0-6] 0.97 (1) [0-4] GP home visits NA 0.3 (0) [0-6] 0.4 (0) [0-6] specialist consults and home visits NA 0.2 (0) [0-12] 0.2 (0) [0-7] % of consulta<ons with GP NA 84% 86% Bilcke et al, Plos ONE 2014

17 Es5ma5ng direct costs : hospital costs These are accessible in many countries Drummond et al, 1997/2015

18 Delphi panels Ask es<mates from experts (preferably clinicians) Several rounds: Pick a number confront with other experts es<mates End in consensus Probability of disease stages, treatment descrip<ons à apply unit costs Problems Subject to bias (experts are oren paid) Atypical caseload Representa<ve choice of experts? Best prac<ce is not always current prac<ce Use only as a last resort; if nothing else is available