Title: An analysis of access to end-of-life care for adults dying of cancer in Nova Scotia
1An analysis of access to end-of-life care for
adults dying of cancer in Nova Scotia
- Meaghan OBrien
- June 6, 2006
2Outline
- Purpose
- Databases
- Study subjects
- Concentration 1 Culture in research
- Concentration 2 EOL data quality
- Concentration 3 Trends in the place of death
of cancer patients - Strengths and limitations
- The future
3Purpose
- Identify variables associated with place of
cancer deaths in NS to help assess level of
equality in access to EOL/PC services
4Databases
Cape Breton PC Program Database
Capital Health PC Program Database
NS Vital Statistics
NS Cancer Centre OPIS
Study Subjects
NS Cancer Registry
1996 and 2001 Census Data
Statistics Canada Postal Code Conversion File
5Study Subjects
All NS residents who died of cancer according to
death certificates, 1994-2003 (n 25,127)
Excluding
Those 19 years of age or younger (n 63)
Those whose cancer diagnosis was not known prior
to death, i.e. death certificate only cases (n
930)
Those who died out of province (n 150)
Those with missing place of death information (n
1097)
22,886 final study subjects 6,151 died out of
hospital 16,735 died in hospital
6Concentration 1 A sociological critique of
ecological measures of culture
- No universally accepted definition of culture
- Leads to difficulties in research
- Cultural safety issues also arise
- Potential method for developing ecological
cultural variables presented
7Ecological cultural variable creation
Subjects postal code of usual residence at time
of death or most recent residential postal code
from OPIS
Cut-points for EA/DAs developed for each
cultural variable (ex. 10, 50)
Census EA/DA
Subjects assigned cultural labels based on
the cultural description developed for their
EA/DA
8Likelihood of out-of-hospital death
Increased Decreased Not significant
gt10 Immigrant gt 10 unemployed gt50 Aboriginal
gt10 Mixed Black gt 50 not high school graduates gt 10 religion other than Catholic, Protestant or none
Median household income upper quintile gt10 Non-black Visible minority
gt10 Non-official language
gt10 French
gt50 single occupant dwellings
From multivariate logistic regression analysis
with year of death, sex, age and tumour group
also in the model to control for potential
confounding
9- Strengths
- Information in existence
- Available nation-wide
- Weaknesses
- Conceptually narrow definition
- Undercount in census
- Describes communities of residence not
individuals - Cultural safety and acceptance of method could be
a problem
10Concentration 2 - Assessment and improvement of
end-of-life and palliative care datasets
- Increasing interest in EOL/PC, meaning increasing
of EOL/PC datasets - Quality data is essential
- Little published to guide data quality
improvement in EOL/PC dataset development
11Concentration 2
From literature and NS EOL dataset development, 9
data quality concepts described with examples
- Data acceptance
- Subject completeness
- Service completeness
- Data field completeness
- Coding constancy
- Data field accuracy
- Validity
- Reporting constancy
- Timeliness
12Concentration 3 Trends in the place of death
of cancer patients in Nova Scotia, 1994-2003
- Based on Burge F, Lawson B, Johnston G. (2003)
Trends in the Place of Death of Cancer Patients.
CMAJ - Dependent variable - place of death
- In hospital (includes deaths in acute care
hospital beds, transitional care beds and nursing
home beds in a nursing home that occupies the
same location as a hospital) - Out of hospital
13Variables
ENVIRONMENT POPULATION
HEALTH
OUTCOMES
CHARACTERISTICS
BEHAVIOUR
- Health care
- system
- Distance to
- cancer care
- External
- environment
- Year of death
- Predisposing
- characteristics
- - Demographics
- Age
- Sex
- -Social structure
- -Health beliefs
- Ecological
- cultural
- variables
- Enabling
- resources
- Personal
- Nursing home
- resident
- Community
- Income
- Region of
- province
- Need
- Tumour
- group
- Time from
- diagnosis
- to death
- Time from
- initial
- registration
- in a PCP
- to death
- Personal health
- practices
- Use of health
- services
- Medical
- oncology
- Palliative
- radiation
- Palliative care
- program
- Location of
- death
- Consumer
- satisfaction
14Definitions
- Nursing home residence place of death or place
of usual residence at time of death matches the
name of a nursing home on an extended list of
nursing homes in NS - Systemic therapy receipt of chemotherapy in
last 12 months of life - Palliative radiation - definition based Johnston
et al. (2001) given a palliative intent code by
treating radiation oncologist or if lt10
fractions were administered in last 9 months of
life
15Method of analysis
- Statistical software (SAS)
- Univariate and multivariate logistic regression
analysis used to identify odds of dying out of
hospital over time - Most parsimonious model of out-of-hospital death
developed using multivariate logistic regression
analyses
16Predictors of out-of-hospital death
- Population characteristics
- Predisposing characteristics
- - Demographics
- Female (OR1.3, CI 1.2-1.3, vs male)
- 75-84 yrs (OR1.5, CI 1.3-1.9, vs 20-44 yrs)
- 85 years (OR2.4, CI 2.0-2.9, vs 20-44 yrs)
- - Social structure and health beliefs
- Immigrant cmty (OR1.2, CI 1.1-1.3)
- Enabling Resources
- - Personal
- Nursing home residence (OR12.3, CI 9.3-16.2)
- - Community
- Upper quintile MHI (OR1.2, CI 1.1-1.3, vs lowest
quintile)
17Predictors of out-of-hospital death
- Population characteristics
- Need
- Survival
- 61-120 days (OR2.2, CI 1.9-2.4, vs lt60 days)
- 121 days (OR2.6, CI 2.4-2.8, vs lt 60 days)
- Tumour group
- Breast (OR1.2, CI 1.03-1.3 vs lung)
- Colorectal (OR1.2, CI 1.1-1.4, vs lung)
- Prostate (OR1.1, CI 1.00-1.3, vs lung)
- CB PCP
- 17-45 days in program (OR1.4, CI 1.01-1.9 vs
lt16 days) - 46-124 days in program (OR1.4, CI 1.03-2.0 vs
lt16 days - 125 days in program (OR1.7, CI 1.2-2.3 vs lt16
days) - CH PCP
- 17-45 days in program (OR2.0, CI 1.6-2.5 vs lt16
days) - 46-124 days in program (OR2.3, CI 1.8-2.9 vs
lt16 days) - 125 days in program (OR2.1, CI 1.7-2.7 vs lt16
days)
18Predictors of out-of-hospital death
- Health Behaviour
- Use of health services
- Referral to
- CB PCP (OR1.5, CI 1.2-1.8)
- CH PCP (OR1.1, CI 1.0-1.2)
Systemic therapy, Aboriginal and French were not
significant in univariate analysis.
19Predictors of out-of-hospital death
- Population characteristics
- Predisposing characteristics
- - Social structure and health beliefs
- Non-official language cmty (OR0.8, CI 0.7-1.00)
- Enabling Resources
- - Community
- Cape Breton County (OR0.7, CI 0.6-0.7, vs
Halifax County) - All other NS counties (OR0.7, CI 0.7-0.8, vs
Halifax County) - Health behaviour
- Use of health services
- Palliative radiation (OR0.9, CI 0.8-0.9)
20- Variables significant in univariate but not
multivariate analysis
Variable Crude OR and CI Correlated with
Year of death 1996 1.2 (1.0-1.3) 1998 1.2 (1.0-1.4) 2003 1.1 (0.9-1.2)
Rural or urban residence 0.8 (0.8-0.9) Region, income and distance to cancer center
Distance to cancer center 21-100km 0.8 (0.8-0.9) 101km 0.8 (0.7-0.8) Region
Mixed Black 1.1 (1.0-1.3)
Visible minority other than Black 1.4 (1.1-1.6) Immigrant
Religion other than Catholic, Protestant or none 1.4 (1.0-1.9) Immigrant
Majority not high school graduates 0.8 (0.8-0.9) Income and immigrant
Majority single occupant dwellings 1.2 (1.0-1.3) Sensitivity analysis
High unemployment 0.8 (0.7-0.9) Income
21Strengths
- Creation of several new variables
- Nursing home residence
- Systemic therapy
- Palliative radiation
- Addition of CB PCP database
- Co-operation with researchers in other provinces
- Increased attention to cultural safety
- Creation of detailed data quality and study
methods - Data and results reviewed by stakeholders
- Results presented to peers at 4 conferences
22Limitations
- Data limited to what was available from existing
databases - Absence of preferred place of death data
- Lack of data on intensity of hospital use at EOL
and type of hospital unit (PC, ICU, ..) - Only deals with inequality, not inequity
23The future
- Co-operation amongst researchers, database
administrators and stakeholders - Linkage to
- Hospital separations database
- Physician claims
- Home care data
- Drug data
- Other PCP databases
- Addition of other variables
- Begin to assess whether inequities exist
24Questions, concerns, ideas?