The cost of diabetes in Canada over 10 years: applying attributable health care costs to a diabetes incidence prediction model
←
→
Page content transcription
If your browser does not render page correctly, please read the page content below
The cost of diabetes in Canada over 10 years:
applying attributable health care costs to a
diabetes incidence prediction model
Anja Bilandzic, MPH (1); Laura Rosella, PhD (1,2,3)
This article has been peer reviewed. Tweet this article
Abstract Highlights
Introduction: Our objective was to estimate the future direct health care costs due to • We created an accessible and
diabetes for a 10-year period in Canada using national survey data, a validated diabetes transparent tool to help health
risk prediction tool and individual-level attributable cost estimates. decision makers calculate future
diabetes costs.
Methods: We used the Diabetes Population Risk Tool to predict the number of new • We predicted the number of new
diabetes cases in those aged 20 years and above over a 10-year period (to 2022), using diabetes cases in Canada in those
2011 and 2012 Canadian Community Health Survey data. We derived attributable costs aged 20 years and above over the
due to diabetes from a propensity-matched case control study using the Ontario next 10 years (2011/12 to 2021/22) and
Diabetes Database and other administrative data. We calculated total costs by applying linked this with actual individual-
the respective attributable costs to the incident cases, accounting for sex, year of diag- level health care costs of diabetes.
nosis and annual disease-specific mortality rates. • By 2022, 2.16 million new cases of
diabetes are expected, correspond-
Results: The predicted 10-year risk of developing diabetes for the Canadian population ing to $15.36 billion in health care
in 2011/12 was 9.98%, corresponding to 2.16 million new cases. Total health care costs costs related to diabetes.
attributable to diabetes during this period were $7.55 billion for females and $7.81 bil- • This tool can model various risk-
lion for males ($15.36 billion total). Acute hospitalizations accounted for the greatest reduction interventions in the popu-
proportion of costs (43.2%). A population intervention resulting in 5% body weight lation; e.g. a 5% weight loss in the
loss would save $2.03 billion in health care costs. A 30% risk-reduction intervention population would save $2.03 billion
aimed at individuals with the highest diabetes risk (i.e. the top 10% of the highest-risk and a 30% risk reduction in the
group) would save $1.48 billion. group with the highest diabetes risk
would save $1.48 billion.
Conclusion: Diabetes represents a heavy health care cost burden in Canada through to
the year 2022. Our future cost calculation method can provide decision makers and
planners with an accessible and transparent tool to predict future expenditures attribut- attributable cost per incident case of dia-
able to the disease and the corresponding cost savings associated with interventions. betes in Ontario is approximately $2930 in
the first year after diagnosis and $1240 in
Keywords: diabetes, economics, attributable cost, prediction model, incidence, Canada following years.4 Recently, Rosella and
colleagues expanded upon this work to
include a greater number of direct costs in
the province, and found that the mean
Introduction Canadian health care system. In 2008, it attributable cost during eight years of fol-
was estimated that the cost of hospital low-up was $9731 for females and $10 315
The management and prevention of diabe- care, physician care and drugs for diabetes for males.5
tes remains a health priority in Canada. was $2.18 billion.2 Looking toward the
With approximately 1.96 million people future, the Canadian Diabetes Association While work has been done across Canada
living with diabetes,1 and with a growing has projected that the overall direct cost of to estimate the future economic costs of
number expected to develop the chronic diabetes will be $3.1 billion in 2020, based diabetes,3,6 most cost estimates and mod-
condition in the future, considering wide- on 3.7 million prevalent cases predicted els are complex, not transparent or not
scale strategies to curb the disease is of using a specially developed diabetes cost readily usable by health decision makers.
great importance. In particular, diabetes model.3 At the individual level, Goeree With the goal of preventing diabetes, a
presents a significant constraint on the and colleagues have estimated that the tool that allows decision makers to
Author references:
1. Dalla Lana School of Public Health, University of Toronto, Toronto, Ontario, Canada
2. Institute for Clinical Evaluative Sciences, Toronto, Ontario, Canada
3. Public Health Ontario, Toronto, Ontario, Canada
Correspondence: Laura Rosella, Dalla Lana School of Public Health, University of Toronto, 155 College St., Toronto, ON M5T 3M7; Tel: 416-978-6064; Email: laura.rosella@utoronto.ca
Health Promotion and Chronic Disease Prevention in Canada
Vol 37, No 2, February 2017 49 Research, Policy and Practiceestimate the economic impact of future information on the demographics, health Database (ODD) to identify new cases of
incident diabetes cases on the health care status and determinants of health of the physician-diagnosed diabetes from 01 April,
system would allow for more effective Canadian population. It is a nationally 2004, to 31 March, 2012. Three control
planning. From a program perspective, representative survey that uses a cross- subjects without diabetes were matched
being able to quantify how actions today sectional study design and is administered to each person with diabetes; they were
may shape case development and associ- on an ongoing basis, with annual data hard matched on index date (± 30 days),
ated health care spending in the future is reporting. It covers 98% of the population age (± 90 days) and the logit of the pro-
a considerable advantage in evaluating aged 12 years and older; exceptions include pensity score. This score was the pre-
strategies. The objective of this study is, people living on Indian reserves and dicted probability of developing or not
first, to estimate the future 10-year direct Crown lands, institutionalized residents, developing diabetes, calculated from a
health care costs due to new diabetes full-time members of the Canadian Forces logistic regression consisting of age, rural-
cases in Canada using national survey and people living in particular remote ity, comorbidity, geographic location and
data and individual-level attributable regions.9 The sample size for this survey neighbourhood income quintile as predic-
costs within the context of a diabetes risk was 124 929; after applying exclusion fac- tive variables.
prediction tool; and second, to apply the tors (e.g. respondents aged under 20 years
tool to two hypothetical intervention sce- and those with existing diabetes were During this eight-year follow-up period,
narios aimed at decreasing diabetes inci- excluded), the final sample size used in individual-level direct health care costs
dence in the population. analyses for this study was 90 631, repre- were tracked annually. These costs were
senting 21 598 180 people when weighted. extracted by linking various health care
Methods utilization databases and following a per-
Intervention scenarios son-level costing methodology specifically
Diabetes risk and incidence developed and validated for Ontario admin
In addition to baseline estimates (i.e. all istrative databases.13 These costs were
To estimate the predicted risk and number demographic and risk factors as outlined from the perspective of the health care
of new diabetes cases within the next 10 above), we ran two hypothetical interven- system, and included costs from inpatient
years, we used the Diabetes Population tion scenarios to examine how implement- hospitalizations, emergency department vis-
Risk Tool version 2.0. DPoRT 2.0 is an ing interventions aimed at reducing diabetes its (ED), same-day surgeries (SDS), dialy-
updated iteration of DPoRT, a predictive risk would affect the incidence of the dis- sis, oncology clinic visits, fee-for-service
algorithm developed to calculate future ease and the cost to the health care physician and non-physician services, non-
population risk and incidence of physi- system. fee-for-service physicians, prescription med
cian-diagnosed diabetes in those aged ications, laboratory, rehabilitation, complex
20 years and over. DPoRT was derived First, we modelled a nontargeted interven- continuing care, long-term care, mental
using national survey data individually tion leading to an average 5% weight loss health inpatient stays, home care services
linked to a chart-validated diabetes regis- in the population. A 5% drop in weight and medical devices. Attributable costs
try. This cohort was then used to create has a positive impact on glycemic and were calculated as the difference in cost
sex-specific survival models using base- cardiovascular health clinically10 and between those with and without diabetes.
line risk factors from the survey for diabe- represents a modest and realistic weight
tes incidence. Specifically, we assessed decrease for many. This intervention would Cost calculations
the probability of physician-diagnosed reflect a large-scale change, such as a
diabetes from the interview date until cen- change in the built environment (e.g. it We developed a cost calculator to use
soring for death or end of follow-up. The has been shown that populations in highly DPoRT 2.0 incidence predictions and per-
model was developed in the Ontario walkable areas have lower overweight and patient attributable cost values to estimate
cohort and predictions from the model obesity prevalence rates11) or the imple- the direct health care costs attributable to
were validated against actual observed mentation of improved nutrition labelling. diabetes, over a future 10-year period. All
diabetes incidence in two external cohorts calculations were sex-specific, reflecting
in Ontario and Manitoba. Variables used Second, we ran an intervention scenario differences in health care use5 and per-
within its two sex-specific models include in which those in the highest-risk decile haps self-care patterns.14 The number of
a combination of hypertension, ethnicity, (i.e. those who have a 10-year risk of incident cases projected to occur each
education, immigrant status, body mass developing diabetes ≥ 22.6%) were tar- year was multiplied by the corresponding
index, smoking status, heart disease and geted for an intervention leading to a 30% per-patient annual cost, dependent on the
income. Full details on the model specifi- reduction in their risk. For example, this time since the diabetes diagnosis, and tak-
cation and validation can be found else- approach might consist of a targeted life- ing into account annual mortality rates,
where.7 The regression model can run on style intervention program or a pharma- which were generated from the age-
nationally available population health sur- ceutical intervention that has proven specific mortality rates of patients in the
veys and has been updated (DPoRT 2.0) efficacy in randomized trials.12 ODD. Mortality rates were specific to year
and used to established prevention targets of follow-up. We assumed that deaths
for diabetes.8 Attributable cost estimates occurred halfway through the year, and as
such, half of those who died contributed
For this study, we used DPoRT 2.0 to gen- To estimate future costs attributed to dia- costs to that specific year. Because the
erate incidence predictions based on the betes, we used results from a recent pro- individual costing estimates used eight
recent 2011 and 2012 Canadian Community pensity-matched cohort study.5 Briefly, years of follow-up in the analysis, it was
Health Survey (CCHS). The CCHS collects this study used the Ontario Diabetes assumed that the costs attributable to
Health Promotion and Chronic Disease Prevention in Canada
Research, Policy and Practice 50 Vol 37, No 2, February 2017individuals who contributed costs in years TABLE 1
9 and 10 after diagnosis did so at the same Health care costs attributable to diabetes, baseline scenario and two hypothetical
monetary value as year 8. As there was a intervention scenarios, Canada, both sexes, 2011/12 to 2021/22
downward tendency in health care costs
observed for the first eight years, we con- Incidence (# of 10-year overall cost
10-year riska (%)
cases, thousands) ($, billions)
ducted a sensitivity analysis whereby years
9 and 10 costs were estimated by follow- Baseline characteristics
ing a linear trend to see the effect of Overall 9.98 2156 15.36
changing the individual attributable costs
Female 8.85 1000 7.55
on the resulting cost estimates. Sex
Male 11.23 1156 7.81
Cost distribution by sector 5% weight loss in population
In order to estimate the burden of costs by Overall 8.67 1873 13.33
sector, the mean costs per health care seg- Female 7.79 880 6.64
ment over the eight years of follow-up Sex
Male 9.64 993 6.70
were converted to percentages and multi-
plied by the total costs estimated from the 30% risk reduction in highest-risk group b
cost calculator. Overall 9.02 1949 13.88
Female 8.20 927 6.97
We performed all statistical analyses using Sex
SAS version 9.4 (SAS Institute Inc., Cary, Male 9.93 1022 6.91
NC, USA). Abbreviation: $, Canadian dollars.
a
10-year risk of developing diabetes.
Results b
The highest-risk group has a 10-year risk of developing diabetes ≥ 22.6%.
The predicted 10-year risk of developing
diabetes for the Canadian population as very different from estimates assuming population, including sex-specific estimates,
a whole is 9.98%, corresponding to equal costs for years 8, 9 and 10. Because as well as region-specific costs. The ability
2 156 000 new cases between 2011/12 and the total difference was approximately to predict incident cases annually also
2021/22. The risk is higher among males $15.96 million, we determined that using allows users to calculate costs per year in
than females (11.23% vs. 8.85%), with the originally proposed costing methodol- the future and costs by year of follow-up
males representing more new cases over- ogy was appropriate. for any number of years ranging from one
all. The estimated total health care cost of to 10.
these new cases is $15.36 billion. In terms of distribution of costs, the larg-
est proportion of health care spending Because this is a new cost methodology
If a population-level (small impact and goes to acute hospitalizations: approxi- that focusses on the development of inci-
large reach) intervention was put in place mately 43.2% ($6.64 billion). The second dent diabetes cases, it is difficult to com-
that resulted in an average body weight largest share is for physician costs, which pare these estimates with previously
loss of 5% in the population, the 10-year represent 21.9% ($3.37 billion) of all projected costs. Previous Canadian esti-
predicted risk of developing diabetes would costs. Prescription medications and assis- mates have used varying health care costs
drop to 8.67%, resulting in 1 873 000 cases tive devices account for 16.9% of costs associated with diabetes, and have either
developing in this time period (Table 1). ($2.60 billion); followed by home care, focussed on projected costs per year based
This reduced number of new cases would nonphysician care and long-term care
on prevalent cases3,6 or have retrospec-
cost $13.33 billion, resulting in a savings ($1.05 billion); other inpatient services
tively reported on cases that have already
of $2.03 billion when compared with ($0.88 billion); and ED, SDS and outpa-
occurred.15-17 The report Economic Burden
baseline characteristics. tient clinic services ($0.83 billion) (Figure 1).
of Illness in Canada, 2005−2008 (EBIC)
offers comprehensive cost estimates for a
In contrast, if an intervention targeting Discussion variety of conditions, including diabetes.2
those with the highest predicted risk (the
top 10% of the highest-risk group) in the Between 2011/12 and 2021/22, new cases Our cost methodology differs from that
population were carried out, the overall of diabetes are estimated to result in used in EBIC in that EBIC used prevalence-
risk of developing diabetes would be 9.02%. $15.36 billion in Canadian health care based costs while we used incidence-
This would translate to 1 949 000 new costs, almost two-thirds of which will be based costs. In addition, we estimated
cases, at a total cost of $13.88 billion spent on acute hospitalizations and physi- attributable costs; our costs represent the
(Table 1). Compared with the baseline cian services (65.1%). This study intro- difference in health care costs that are
scenario, $1.48 billion in direct health duces a novel way of estimating future directly attributable to diabetes, while
care costs would be averted. health care costs attributable to new cases EBIC only generates overall cost of illness.
of diabetes. The linkage of an incidence This is achieved by using a propensity-
When we estimated costs for years 9 and prediction model with individual-level matched cohort design.5 Finally, EBIC did
10 using a linear trend based upon years attributable costs allows for estimates to not couple these estimates with predic-
1 to 8 of observation, the results were not be derived for different segments of the tions on future cases and therefore did not
Health Promotion and Chronic Disease Prevention in Canada
Vol 37, No 2, February 2017 51 Research, Policy and PracticeFIGURE 1 attributable cost estimates did become
Distribution of total 10-year direct health care costs attributable available in the future, the cost calculation
to diabetes ($ billions), Canada, 2011/12 to 2021/22 method could easily be adapted to include
these region-specific costs.
6.64
Second, this method uses average attribut-
able costs by sex and year of follow-up.
Direct health care costs ($ billions)
As such, it cannot account for costs
averted within specific subgroups, who
may be using more or less health care
3.37 than the average. For example, in an inter-
2.60
vention aimed at a high-risk group, it is
likely that these people spend more health
care dollars than the average, but their
1.05 averted cost calculated will not reflect this
0.88 0.83
(i.e. it will be underestimated using this
method). Efforts to produce estimates that
are defined to more specific populations
Inpatient (acute Physician Prescription Home care, non- Inpatient (other) ED, SDS,
hospitalization) medication, physician care, outpatient clinic would enable more accurate estimates,
devices long-term care (dialysis,
oncology) particularly when modelling intervention
scenarios for certain target groups.
Abbreviations: ED, emergency department visits; SDS, same-day surgery.
Note: Figures have been rounded.
Third, the model does not account for
future changes in health care spending or
allow for intervention planning or esti- for the evaluation of different policy
inflation. It is assumed that diabetes case
mates on future cost burden. options and can assist in determining how
management will remain the same through
best to move forward with chronic disease
2022 and that current models of care will
Strengths and limitations prevention activities. For example, in
continue to be applied and used in the
Canada, there are dozens of promising
This methodology has unique strengths. same way. Given the window of 10 years,
policy choices and interventions aimed at
First, the costs are based on actual healthy living being led through federal, this assumption is likely appropriate.
observed health care cost data from a pro- provincial and regional partnerships.18 Longer prediction periods would need to
spective cohort over eight years of obser- Such programs could benefit from a tool address potential changes to care and
vation. Therefore, these are not projected that would factor in context-specific popu- management.
estimates only, but instead reflect the real- lation characteristics to evaluate the most
ity of contemporary diabetes costs to the appropriate and feasible intervention Finally, our estimates do not account for
health care system. The use of attributable strategies from an economic and health the costs associated with diabetes that are
cost as a metric is also advantageous as it perspective. Further applications could not related to health care, including indi-
represents the excess cost of disease include providing information on the out- rect costs, out-of-pocket costs and costs
beyond average spending, due to the com- comes of improved treatment and disease not captured in administrative databases,
parison with the group without the dis- management strategies. Since these as well as emotional and social costs for
ease. Using total costs based only on the approaches can lengthen life and possibly patients and other caregivers. It is esti-
diseased population can overestimate the reduce costs, this information, combined mated that direct health care costs only
spending on disease and can provide with the effect on incidence, could offer account for 17% of total costs attributable
inflated evaluations.2 insight into the combination of both treat- to diabetes,3 so it is crucial to consider
ment and prevention approaches. these additional expenses in future
Second, this method is simple to apply research.
and can be used by a variety of end users. The simplicity of this model does mean
This is the aim of the tool itself—to be that several assumptions had to be made Conclusion
accessible and transparent for use within and must be acknowledged. First, the cost
applied settings, such as provincial minis- estimates are derived from a study that The goal of this work is to provide health
tries of health and regional health bodies. was based on Ontario data and thus the decision makers with a readily usable tool
Being able to model intervention scenar- attributable costs used for national esti- that will allow them to make cost estima-
ios, unique to the user’s program goals mates assume that health care spending is tions up to 10 years in the future. Health
and region, is an added benefit for health similar in other provinces and territories. planners and policy makers who focus on
planners and decision makers who seek to However, it is known that differences exist preventing diabetes at the population level
estimate the economic offsets of various across jurisdictions, including within the can use this tool to evaluate different
diabetes prevention strategies. Being able general care and management of diabetes,17 intervention strategies with customized
to estimate the cost averted, in addition to as well as in provincial coverage for services incidence and cost predictions, which will
the number of cases prevented through and products such as medications and assist them in determining the most
customized intervention strategies, allows assistive devices.19,20 If province-specific appropriate actions for the future.
Health Promotion and Chronic Disease Prevention in Canada
Research, Policy and Practice 52 Vol 37, No 2, February 2017Acknowledgements 7. Rosella LC, Manuel DG, Burchill C, 16. Dawson KG, Gomes D, Gerstein H,
Stukel TA, PHIAT-DM team. A popu- Blanchard JF, Kahler KH. The eco-
Dr. Laura Rosella is supported by a lation-based risk algorithm for the nomic cost of diabetes in Canada,
Canada Research Chair in Population development of diabetes: develop- 1998. Diabetes Care. 2002;25(8):1303-7.
Health Analytics. This work was sup- ment and validation of the Diabetes
ported by the Canadian Institutes of Population Risk Tool (DPoRT). J 17. Simpson SH, Corabian P, Jacobs P,
Health Research Operating Grant from the Epidemiol Community Health. 2011; Johnson JA. The cost of major comor-
Institute of Nutrition, Metabolism and 65(7):613-20. bidity in people with diabetes melli-
Diabetes (funding reference number 126615). tus. CMAJ. 2003;168(13):1661-7.
The funders had no direct role in the anal- 8. Rosella LC, Lebenbaum M, Li Y,
ysis, interpretation, writing or submission Wang J, Manuel DG. Risk distribution 18. Pan-Canadian Public Health Network.
of this manuscript. and its influence on the population Towards a healthier Canada - 2015
targets for diabetes prevention. Prev progress report on advancing the
Conflicts of interest Med. 2014;58:17-21. federal / provincial / territorial frame
work on healthy weights. Available
The authors disclose no conflict of 9. Statistics Canada. Canadian Community from: www.phn-rsp.ca/thcpr-vcpsre
interest. Health Survey user guide: 2012 and -2015/index-eng.php#initiatives
2011-2012 microdata files. Ottawa (ON):
Authors' contributions Statistics Canada; 2013. 19. Canadian Diabetes Association,
Diabète Québec. Diabetes: Canada at
10. Canadian Diabetes Association Clinical
LR conceived the manuscript. LR and AB the tipping point: charting a new
Practice Guidelines Expert Committee.
contributed to analytic plan and interpre- path. Toronto (ON): Canadian Diabetes
Canadian Diabetes Association 2013
tation. AB prepared data and ran all analy- Association and Diabète Québec; 2011.
clinical practice guidelines for the
ses. AB drafted the manuscript and LR
prevention and management of dia- 20. Canadian Diabetes Association. The
edited and critically reviewed the final
betes in Canada. Can J Diabetes. burden of out-of-pocket costs for
content.
2013;37(suppl 1):S1-S212. Canadians with diabetes [Internet].
References 11. Glazier RH, Creatore MI, Weyman JT, Toronto (ON): Canadian Diabetes
et al. Density, destinations or both? A Association; [date unknown; cited
1. Statistics Canada. CANSIM database: 2015 Jun 15]. Available from: http://
comparison of measures of walkabil-
Table 105-0501: Diabetes, by age group w w w. d i a b e t e s . c a / C DA / m e d i a
ity in relation to transportation behav
and sex [Internet]. Ottawa (ON): /documents/publications-and
iors, obesity and diabetes in Toronto,
Statistics Canada; 2014 [modified 2016 -newsletters/advocacy-reports/burden
Canada. PLoS One. 2014;9(1):e85295.
Apr 21; cited 2015 May 01]. Available -of-out-of-pocket-costs-for-canadians
Erratum in PLoS One. 2014;9(3):
from: http://www5.statcan.gc.ca -with-diabetes.pdf
e91485.
/cansim/a26?lang=eng&id=1050501
12. Gillies CL, Abrams KR, Lambert PC,
2. Public Health Agency of Canada.
et al. Pharmacological and lifestyle
Economic burden of illness in Canada,
interventions to prevent or delay type
2005−2008. Ottawa (ON): Public
2 diabetes in people with impaired
Health Agency of Canada; 2014 [Cat
glucose tolerance: systematic review
alogue No.: HP50-1/2013E-PDF].
and meta-analysis. BMJ. 2007;
3. Somerville R. An economic tsunami: 334(7588):299.
the cost of diabetes in Canada.
13. Wodchis WP, Bushmeneva K, Nikitovic
Toronto (ON): Canadian Diabetes
M, McKillop I. Guidelines on person-
Association; 2009.
level costing using administrative
4. Goeree R, Lim ME, Hopkins R, et al. databases in Ontario. Toronto (ON):
Prevalence, total and excess costs of Health System Performance Research
diabetes and related complications in Network; 2013.
Ontario, Canada. Can J Diabetes. 2009;
14. De Melo M, de Sa E, Gucciardi E.
33(1):35-45.
Exploring differences in Canadian
5. Rosella LC, Lebenbaum M, Fitzpatrick adult men and women with diabetes
T, et al. Impact of diabetes on health- management: results from the Canadian
care costs in a population-based cohort: Community Health Survey. BMC Public
a cost analysis. Diabet Med. 2016; Health. 2013;13(1):1089.
33(3):395-403.
15. Pohar SL, Majumdar SR, Johnson JA.
6. Ohinmaa A, Jacobs P, Simpson S, Health care costs and mortality for
Johnson JA. The projection of preval Canadian urban and rural patients
ence and cost of diabetes in Canada: with diabetes: population-based trends
2000 to 2016. Can J Diabetes. 2004; from 1993−2001. Clin Ther. 2007;29
28(2):1-8. Spec No:1316-24.
Health Promotion and Chronic Disease Prevention in Canada
Vol 37, No 2, February 2017 53 Research, Policy and PracticeYou can also read
