Title: Estimating costs of diagnosis and treatment for lung cancer using linked longitudinal data
1Estimating costs of diagnosis and treatment for
lung cancer using linked longitudinal data
A seminar presented at the British Columbia
Cancer Agency, Vancouver, Canada 1 December 2006
- Jim Butler
- Australian National University
2Overview
- Some issues in methods
- Study 1 Treatment costs
- Study 2 Diagnosis costs (in progress)
3Some issues in method
Time period of observation
4Censoring
- Costs accumulate over time to a defined event
(e.g. death) - Censored observations arise from subjects who are
not followed to the defined event- still alive
at end of study- lost to follow-up during study - Apply survival analysis to cost data
- Bias may arise if underlying assumptions are not
satisfied
5- Key assumption in applying Kaplan/Meier
methodcensoring is independent of time-to-event
(independent censoring) - ? non-censored subjects are representative of all
subjects - ? censored subjects are neither a relatively
high-risk sub-population nor a relatively
low-risk sub-population
6Probability that time to event T exceeds any
given value t is given by
s1, s2, observed failure times rs
risk of failure at time s
7- Need a set of unbiased estimators of the set of
hazards rs - Even if independent censoring characterises
time-to-event analysis, it may not characterise
cost-to-event analysis - e.g. low-cost subjects may be more likely to
drop out - Extent of problem is an empirical matter
- Less likely to be a problem with analyses using
population-based data (e.g. people leaving
region/province/country are the source of
drop-out)
8Study 1
SURVIVAL AND TREATMENT COSTS OF A
POPULATION-BASED SAMPLE OF STAGE IIIb/IV
NON-SMALL CELL LUNG CANCER (NSCLC) PATIENTS
ONBEST SUPPORTIVE CARE (BSC) Erich KliewerJim
ButlerAlain Demers Sri Navaratnam Grace
Musto Coreen Hildebrand
9Objectives
- For stage IIIb/IV NSCLC patients in BSC phase of
treatment describe their - chemotherapy treatment prior to BSC
- survival since start of BSC
- treatment since start of BSC
- costs of treatment since start of BSC
10Study Population
- Patients diagnosed with stage IIIb/IV NSCLC in
Manitoba between March 1997-June 2000 and who had
survived at least 28 days since last chemotherapy
(BSC phase). - If stage IIIb they had to have either pleural
effusion or supraclavicular lymph node metastsis. - n150
11(No Transcript)
12Lung cancer cases
13Age at diagnosis - BSC
14Survival
15Survival by stage
16Treatment summary numbers of items
17Average cost per patient-month from date of last
chemotherapy(CAD, current prices, undiscounted)
18Average cost per patient-month by month of
follow-up
19Average cost per patient-month over various
periods of follow-up(CAD, current prices,
undiscounted)
20Average cost per patient-month from date of last
chemo vs BSC date(CAD, current prices,
undiscounted)
21Lifetime treatment costs (LTC) were estimated by
applying Kaplan-Meier monthly survival
probabilities to average cost per month. The
formula with monthly discounting is
22Lifetime treatment costs(CAD)
23Study 2
ANALYSIS OF WAITING TIMES AND COSTS FOR THE
DIAGNOSIS OF NON-SMALL CELL LUNG CANCER Winson
Cheung Steve Welch Jim ButlerErich Kliewer Alain
Demers Grace Musto Sri Navaratnam
24Objectives
- To assess the timeliness of the diagnosis of
non-small cell lung cancer (NSCLC) in Manitoba
and to evaluate the variables that affect these
waiting times. - To quantify the costs involved in the diagnosis
of NSCLC in Manitoba. - To correlate data collected by chart review with
information from the Manitoba Health
administrative databases.
25Methods
- Patients diagnosed with NSCLC from January 1,
1996 to December 31, 2000 were identified using
the Manitoba Cancer Registry. - Information on demographics, diagnostic and
staging tests, timeline of investigations, and
outcomes were collected by systematic,
retrospective chart review for 543 patients. - Similar information was obtained from the
Manitoba Health administrative databases for 472
of the patients.
26Demographic Diagnostic Characteristics
Gender n 543
Female Male 40.5 (220) 59.5 (323)
Age at Diagnosis n 541
lt40 40-60 61-80 gt80 Median 1.3 (7) 21.1 (114) 65.6 (355) 12.0 (65) 68.6 yrs
27Demographic Diagnostic Characteristics
Stage at Diagnosis n 530
I II III IV 10.0 (53) 6.5 (35) 35.0 (185) 48.5 (257)
Method of Diagnosis n 541
FNA Bronchoscopy Mediastinoscopy Thoracotomy Other 30.5 (165) 38.8 (210) 5.9 (32) 7.9 (43) 16.9 (91)
28Waiting time intervals
Waiting Times n median 25th to 75th ile
Symptom to 1st CXR 1st CXR to 1st CT 1st CT to Diagnosis 328 346 346 26 days 10 days 18 days 7 to 62 days 1 to 31 days 6 to 36 days
Symptom to Diagnosis 469 72 days 41 to 124 days
29Factors affecting waiting time
30Cost Analysis for Diagnosis of NSCLC
Test or Service Subtotal Cost Mean Cost per Pt (n 464)
Total (excluding inpatient costs) 191,784 413.30
Total (including inpatient costs) 3,019,435 6,507.40
31Costs of diagnosis by type of service
Test or Service Mean Cost per Patient (n 464)
Diagnostic 263.10
Staging 26.00
Physician Visits 121.10
Inpatient 6,094.10
Others 3.10
TOTAL 6,507.40