Development of a measure of health care affordability applicable in a publicly funded universal health care system.
Haggerty, Jeannie L. ; Levesque, Jean-Frederic
The accessibility principle enshrined in the 1984 Canada Health Act
(1) guaranteeing public coverage for "medically necessary
services" states that services should be provided uniformly,
unimpeded either directly or indirectly by charges or any other factors
(age, health status or financial circumstances). The publicly funded
health system is a source of national pride for Canadians, (2) but
evidence of health care inequity by socio-economic status in Canada
(3-7) suggests that even medically necessary health care is not equally
affordable. Health care affordability should be measured directly and
monitored systematically.
Health service accessibility is a function of how well the
geographic location, organization, acceptability and price of available
resources (supply) fit or interact with health need and ability of a
broad range of potential users (demand) to seek, reach, and pay for
care. (8-12) Health services are affordable if people are able to
mobilize the time and money required to obtain needed health services.
(13) Direct costs are those incurred from fees charged at the point of
care, and in the Canadian system, these are typically payments for
community-dispensed medications and non-physician health services
offered outside the hospital (e.g., physiotherapy, nutrition,
psychology, and social work). Indirect costs are non-health-care costs
related to obtaining services, such as transport, childcare or revenue
loss due to having to obtain services during working hours. Willingness
to incur direct and indirect costs varies by people's incomes as
well as factors such as perceived seriousness of the problem, perceived
quality of care, and general health efficacy.
Measures of health care affordability and economic accessibility
Health care affordability in Canada has been mostly inferred from
secondary analysis of health services use stratified by economic
quintiles, usually with health care administrative data. (14) This has
drawn attention to persisting inequities in health care access but this
measure has inherent limitations. Administrative data are, by
definition, based on realized access to services and do not capture
forgone care. Because utilization is often higher in the low
socio-economic quintiles, utilization must be adjusted by need. But
proxies of need in administrative data are themselves based on health
care utilization. Also the socio-economic strata are based on contextual
information, which provides only a proxy of socio-economic status for
individuals.
Health care surveys can measure directly access barriers and
forgone care and permit stratification by individual socioeconomic
status. The Commonwealth Fund has done international comparative surveys
of health systems, including information on the prevalence of those
forgoing medication or recommended tests and treatments because of cost.
(15) However, it is very difficult to compare Canadian provinces or the
prevalence over time because of the relatively small country samples and
the differences in sampling frames from year to year.
We recognized the need to develop a measure of health care
affordability that would be relevant to the Canadian system, when we
were conducting a series of validation studies of accessibility
measures. We had found that validated instruments are more
discriminating in urban compared to rural settings, (16) and were
conducting a mixed-method study to develop measures of health service
organizational and geographic accessibility that are equally valid for
rural and urban areas. In the exploratory qualitative phase of the
study, an unexpected finding was how the experience of accessibility
differs markedly by individuals' capacity to mobilize resources,
(17) especially in rural areas where accessing the full scope of
services requires having personal transport. A subsequent mapping of
qualitative codes to available accessibility instruments showed that
validated measures of health care affordability come from other
countries and focus almost exclusively on direct costs of care.
The objective of our study was to develop a valid and reliable
measure of health care affordability for the Canadian context based on
the people's experience. Having a complementary tool to
administrative data will help public health researchers to quantify and
monitor, at individual and population levels, the economic accessibility
of our publicly funded health system and its impact on health care
equity.
METHODS
The current study was part of a sequential mixed-method and
repeated sample study to develop patient-based measures of
accessibility. It received ethical approval from the ethical review
committees of Universite de Montreal and the Hopital Charles LeMoyne.
Development of the items
To inform the content of the instrument, we conducted 11 focus
groups in metropolitan, rural, and remote settings in Quebec. We used a
similar discussion guide in each setting to explore typical care-seeking
trajectories and responses to obstacles in the context of a minor but
urgent health problem, and for ongoing care of an existing problem. The
methods and major themes have been reported elsewhere. (17) Themes
related to health care affordability were: difficulties in accessing
services due to loss of revenue incurred by having to seek health care
during regular working hours; indirect costs due to travel and parking;
and costs of diagnostic services and other services not covered by
Medicare. Themes were expressed as codes expressed as patient
experience.
We mapped codes to items from existing instruments to see where
there were measurement gaps (available upon request). Most instruments
elicit ratings of direct visit and treatment costs, so we developed new
items to capture affordability codes. We cognitively tested the items on
convenience samples of French and English speakers with no more than
secondary school education to ensure that potential respondents
understood the statements as intended and that the response options were
relevant to their experience. We developed items simultaneously in
English and French to achieve semantic equivalence in both languages.
The final subscale of health care affordability was inspired by the
Picker Institute approach (18) and relevant items from the Commonwealth
Fund international health system performance surveys. (19-21) The items
are shown in Table 2.
Instrument administration
The initial questionnaire was administered in 2008 by
computer-assisted telephone interview to 750 respondents selected by
random digit dialing, 250 each from a metropolitan area (population >
100,000), a rural area (one to two hours travel time from the urban
area), and a remote area (more than three hours travel from a tertiary
care centre). This sampling frame allowed us to examine for differential
effects by geographic context. At the end of the interview, 492 accepted
to be contacted again.
For the second administration in 2009, we used a self-administered
paper questionnaire sent by mail. Of those we succeeded in contacting
after up to 10 attempts by telephone (n = 399), 368 agreed to
participate and were mailed a self-administered version of the refined
questionnaire; 316 (79%) responded.
Measurement of ability to pay
We used indicators of socio-economic status as a proxy of ability
to pay to determine whether the measure varied appropriately by this
comparator variable. In the first administration, we measured
socio-economic status by highest level of education achieved and the
person's self-perceived financial status with Likert-type responses
ranging from poor to comfortable. For the second administration, we
additionally elicited household or personal revenue, number of family
members in the household, home ownership, vehicle ownership, existence
of a retirement savings plan by a family member, and whether the person
held complementary medical insurance that covers medication and
non-physician health care services. Ability to pay was indicated by a
composite index for income status using the current poverty line cut-off
for low-income status: up to three-member families earning less than
$40,000/year combined, or four-or-more-member households earning less
than $60,000/year. For those with missing revenue information (24/316),
we imputed low-income status from the other economic indicators.
ANALYSIS
We used descriptive statistics to exclude items with > 4%
missing values either globally or by educational or economic categories.
We flagged potential variance problems due to non-normal distributions.
For construct validity, we used exploratory factor analysis to explore
which items group together, loading ([greater than or equal to] 0.5) on
a presumed underlying construct or latent variable. We used confirmatory
factor analysis using structural equation modelling to ensure that the
affordability items formed a single dimension and were distinct from
other dimensions of accessibility. (22)
We inferred reliability from the measure of internal consistency
(standardized Cronbach's alpha) using the original 5-point
frequency responses and the dichotomous scoring. We used nonparametric
and parametric Item Response Theory (IRT) analysis (TESTgraph and
Multilog) to examine the performance of individual items, to ensure that
the response options behaved as expected to achieve maximum reliability.
IRT analysis was used to ensure that each item contributed specific and
unique information and to establish the scoring for the subscale.
To examine predictive validity, we used logistic or ordinal
regression modelling to examine how individual items and the full
subscale were associated with binary indicators of difficult access. We
looked for effect modification by socio-economic status by including
interaction terms and doing stratified analysis. Regression and basic
psychometric analyses were conducted with SAS9. (23)
RESULTS
The socio-demographic, health, and health care characteristics of
the initial and repeat study samples are provided in Table 1. Those who
responded to the second questionnaire did not differ in perceived
financial status, but were more likely to have a personal physician, had
slightly higher levels of education, were more likely to have unmet
needs for health care, and seemed to have poorer health indicators. The
repeat sample is sufficiently similar to the initial sample to represent
the typical health care affordability experience of the population. The
sample also has sufficient variation to permit robust comparisons by
socioeconomic status: 57.2% (179) had employment-related private
insurance that covers medications, 48.4% (150) had insurance that covers
complementary health services such as psychology and physiotherapy and
59.6% were classified as low-income.
For the initial telephone administration, we elicited the frequency
of: 1) payment at point of care for services not covered by Medicare; 2)
loss of revenue from seeking care; and 3) indirect costs. For each, we
also elicited "what is or would be" the importance of that
cost on the decision to seek care. The percent reporting costs rarely to
very often were: 20% direct costs; 27% loss of revenue; 26% indirect
costs. Over half of the sample rated actual or anticipated costs as
being important in the decision to seek care, an effect more pronounced
in low-income respondents. However, despite various attempts at scoring,
we could not find an appropriate way to represent the items together in
a way that predicted health care use, and the items did not group on a
single factor. For the second administration, we abandoned this approach
in favour of the reported frequency of unmet needs due to cost; the
remaining results are based on responses to the self-administered
subscale in the repeat sample.
The items developed for the second administration are shown in
Table 2. The items elicit the frequency of five cost-related unmet needs
or accessibility difficulties. As per the Picker Institute approach, we
offer Likert-type frequency response options but apply a dichotomous
scoring to each item, where 1 represents a cost barrier to needed care
and 0 represents no barrier for each of the five items. The 0 =
never/rarely cut-off showed most sensitivity to access-related
consequences.
Table 2 shows the descriptive statistics and comparison between
income groups. No items had more than 2% true missing values, but an
important proportion of respondents indicated that the service referred
to in the question was not applicable to them. For all items,
cost-related barriers were more prevalent and frequent in the low-income
than in the high-income group; only one was not statistically
significant: item 3, not getting complementary health services ([chi
square] = 4.3, 4df, p = 0.35). This item had the highest rate of
non-applicability (34%), as well as the highest frequency of problems
with affordability, even among the high-income respondents. It has the
lowest correlation with other items (range r = 0.31 to 0.39), and
consequently reduces the internal consistency of the scale, but it is
retained for conceptual reasons as it represents an important cost
barrier in Canada. The subscale shows adequate internal consistency
([alpha] = 0.77) and the items load high on a single factor.
Two overall scores can be derived: population and individual. The
population score is the average percent of respondents reporting a
cost-based lack of access based on the five individual items. The
population score is the complement of population health system coverage,
i.e., the percent of the population who do not obtain needed health
services because of cost barriers. The percent of non-coverage for our
respondents was 12.2%; 15.1% in the low-income group and 8.6% in the
higher-income group. We also show the percent reporting at least one
cost barrier.
The individual score is the sum of the dichotomously coded items.
It ranges from zero to five, where a high score means lower health care
affordability and is interpreted as the accumulated cost barriers to
medically necessary services. The overall mean of the subscale is 0.49
(SD = 0.96), with both median and mode equal to zero. Despite being
skewed toward zero (skewness = 1.53), the entire range from 0 to 5 was
used in our sample. Most importantly, the mean was different for low-
vs. high-income respondents, 0.59 vs. 0.36 respectively, and was
statistically significant (t = -3.58, p = 0.0004).
To determine how to interpret the "not-applicable"
response options (e.g., "I do not take any medications"), we
compared results where this was considered as missing (n = 175 complete
observations) and as "never" (n = 308 complete observations).
Interpreting "not applicable" as "never" slightly
reduced the between-item Pearson correlation coefficients, the
item-total correlations, and the internal consistency ([alpha] = 0.81 to
0.75). It increased slightly the loadings on a single factor but did not
change the general conclusions. Furthermore, IRT analysis showed that
the probability of endorsing the non-applicable response option
overlapped mostly with the probability of endorsing "never".
So, for the final scoring, "not applicable" was considered
equivalent to "never".
Finally, IRT analyses indicated that the 5-point Likert response
options could be collapsed into a 3-point option. For future
applications, we would recommend offering: never or rarely; sometimes;
often or very often.
Consequences of cost-related barriers to care
To examine the predictive validity of our health care affordability
subscale on access difficulties, we used four indicators that had been
evoked in the qualitative phase: nuisance, unmet needs, emergency room
use, and problem aggravation. Nuisance relates to having to restart the
care-seeking trajectory at least once after encountering impediments;
43% of respondents reported experiencing a nuisance occasionally or
sometimes, and 7% often or very often. The other consequences are more
significant, although they occur less frequently: 22% reported having an
unmet need after abandoning the care process altogether; 12% reported
using the hospital emergency room for health system reasons; and 8%
reported an aggravation of their health problem because of delays in
getting care.
As shown in Table 3, each unit increase in the sum of cost-related
barriers results in an increased likelihood of these consequences
occurring, after controlling for health status. Indeed, the magnitude of
association is stronger in models that do not control for health status,
reflecting the well-known gradient between poor health and
socio-economic status. The predicted probability for all but emergency
room use is stronger in low-income than in high-income respondents. The
association between number of economic barriers and consequences is
strongest for unmet needs and problem aggravation. The Somer's D
can be interpreted similarly to risk difference as the difference in the
probability of experiencing a consequence for those with at least one
economic barrier vs. those with no economic barriers. In the low-income
population, the Somer's D is 0.47 for unmet needs and 0.73 for
problem aggravation.
DISCUSSION
We have developed a direct measure of health care affordability
that reliably and validly captures the cost-related barriers to health
services in Canada. Scores reflecting problems with health care
affordability are significantly and strongly associated with indicators
of access difficulties like use of the emergency room, unmet needs for
care, and problem aggravation due to delayed care, and the associations
are stronger among low- than among high-income persons.
Other studies in Canada have pointed to differential health service
use by income category, (3-7) but to our knowledge, this is the first
valid and reliable direct measure of cost-related barriers to necessary
medical services. Although problems with health care affordability were
more prevalent in low- than in high-income respondents in our sample,
higher-income respondents also reported problems with affordability. In
our sample, almost one third of respondents, regardless of income, did
not follow through with physician-recommended services due to cost
issues, particularly for prescribed services not covered by public
insurance. This observed avoidance of health care due to cost across all
income groups suggests that the measure reflects both willingness-to-pay
and affordability, which may be seen as a limitation.
Suggested application
This measure is for use in health care surveys. Both population and
individual overall scores can be derived. The population score measures
health system coverage (24) and can be represented two ways: 1) the
average percent of respondents reporting cost-related access
difficulties across the five dimensions; and 2) the percent reporting at
least one access difficulty due to direct and indirect cost barriers. It
can also be interpreted as the probability that an individual will not
obtain the needed services, conditional on need. This measure reflects
the extent to which health system structures, resources, procedures, and
regulations render services available to the population on the basis of
need. It can be used to monitor the accessibility clause in the Canada
Health Act.
The overall score for an individual is the sum of reported cost
barriers and it reflects the cumulative affordability burden borne by
individuals. This score will be particularly relevant for comparisons
between population subgroups and for researchers to track the
consequences of cost-related barriers at an individual level. Both
cost-related lack of coverage and cumulative problems with health care
affordability to medically necessary services are more than twice as
high among low-income compared to high-income respondents.
Though the factor analyses support use of an overall meaningful
score of health care affordability, the scores on specific items may be
more relevant for decision-makers and for comparisons over time and
between provinces. The cost-related barriers are expected to differ from
one Canadian province to another, depending on the extent of public
insurance coverage. For instance, problems affording medication will be
higher in provinces that, unlike Quebec, do not provide publicly funded
drug coverage.
We recommend that a 12-month reference period be used. In this
study we did not specify a time-frame, even though the remainder of the
questionnaire referred consistently to the past 12 months. This was an
oversight, because in applying the measure to an independent sample,
(25) we found that scores are sensitive to the reference period. The
scores rate dropped appropriately when a 12-month period was specified
compared to no reference period.
We also recommend measuring the occurrence or magnitude of direct
costs billed to patients by physicians at different points of care,
another issue that needs to be monitored. A recent Quebec survey of
randomly-selected health service users found an increase in the
occurrence and frequency of direct payment for medications,
examinations, and other services at their physicians' offices
between 2005 and 2010, with 71.8% paying for at least one service in
2010.26
In conclusion, our study supports the reliability and validity of
this measure of cost barriers to medically necessary services. It is
important that this subscale be applied in other jurisdictions to allow
benchmarking of the economic accessibility of Canada's publicly
funded system.
REFERENCES
(1.) Canadian Government. Canada Health Act, Bill C-3. Statutes of
Canada, 3233 Elizabeth II (RSC 1985, c 6; RSC 1989, c C-6). 1984.
(2.) Mendelsohn M. Canadians' thoughts on their health care
system: Preserving the Canadian model through innovation. Saskatoon, SK:
Commission on the Future of Health Care in Canada, 2002.
(3.) Allin S. Does equity in healthcare use vary across Canadian
provinces?. Healthc Policy 2008;3(4):83-99. PMID: 19377331.
(4.) Asada Y, Kephart G. Equity in health services use and
intensity of use in Canada. BMC Health Serv Res 2007;7(1):41. PMID:
17349059. doi: 10.1186/14726963-7-41.
(5.) Jimenez-Rubio D, Smith PC, Van Doorslaer E. Equity in health
and health care in a decentralised context: Evidence from Canada. Health
Econ 2008; 17(3):377-92. PMID: 17721900. doi: 10.1002/hec.1272.
(6.) Schoen C, Doty MM. Inequities in access to medical care in
five countries: Findings from the 2001 Commonwealth Fund International
Health Policy Survey. Health Policy 2004;67(3):309-22. PMID: 15036818.
doi: 10.1016/j. healthpol.2003.09.006.
(7.) Levesque JF, Pineault R, Hamel M, Roberge D, Kapetanakis C,
Simard B., et al. Emerging organisational models of primary healthcare
and unmet needs for care: Insights from a population-based survey in
Quebec province. BMC Fam Pract 2012;13(1):66. PMID: 22748060. doi:
10.2105/AJPH.2004.059402.
(8.) Penchansky R, Thomas JW. The concept of access: Definition and
relationship to consumer satisfaction. Med Care 1981;19(2):127-40. PMID:
7206846. doi: 10.1097/00005650-198102000-00001.
(9.) Donabedian A. Capacity to produce services in relation to need
and demand. In: Donabedian A (Ed.), Aspects of Medical Care
Administration: Specifying Requirements for Health Care. Cambridge, MA:
Harvard University Press, 1973:418-85.
(10.) Bashshur RL, Shannon GW, Metzner CA. Some ecological
differentials in the use of medical services. Health Serv Res
1971;6(1):61-75. PMID: 5569227.
(11.) Starfield B. Primary Care: Balancing Health Needs, Services,
and Technology. New York, NY: Oxford University Press, 1998.
(12.) Frenk J. The concept and measurement of accessibility. In:
White KL, Ordonez C, Paganini JM, Starfield B (Eds.), Health Services
Research: An Anthology. Washington, DC: Pan American Health
Organization, 1992; 858-64.
(13.) Levesque JF, Harris MF, Russell G. Patient-centred access to
health care: Conceptualising access at the interface of health systems
and populations. Int J Equity Health 2013;12(1):18. PMID: 23496984. doi:
10.1136/jech.2003. 017731.
(14.) Glazier RH, Agha MM, Moineddin R, Sibley LM. Universal health
insurance and equity in primary care and specialist office visits: A
population-based study. Ann Family Med 2009;7(5):396-405. PMID:
19752467. doi: 10.1370/ afm.994.
(15.) Schoen C, Osborn R, Huynh PT, Doty M, Davis K, Zapert K,
Peugh J. Primary care and health system performance: Adults'
experiences in five countries. Health Affair 2004;1037(W4):487-503.
PMID: 15513956.
(16.) Haggerty JL, Bouharaoui F, Santor DA. Differential item
functioning in primary healthcare evaluation instruments by
French/English version, educational level and urban/rural location.
Healthc Policy 2011;7(Spec Issue):47-65. PMID: 23205035.
(17.) Haggerty JL, Roberge D, Levesque J-F, Gauthier J, Loignon C.
An exploration of rural-urban differences in healthcare-seeking
trajectories: Implications for measures of accessibility. Health Place
2014;28:92-98. PMID: 24793139.
(18.) Borowsky SJ, Nelson DB, Fortney JC, Hedeen AN, Bradley JL,
Chapko MK. VA community-based outpatient clinics: Performance measures
based on patient perceptions of care. Med Care 2002;40(7):578-86. PMID:
12142773. doi: 10.1097/00005650-200207000-00004.
(19.) Schoen C, Osborn R. The Commonwealth Fund 2004 International
Health Policy Survey of primary care in five countries. 2004. Available
at: http:// www.commonwealthfund.org/-
/media/Files/Surveys/2004/2004%20Commonwealth%20Fund%20International%20Health%20Policy%20Survey%
20of%20Adults%20Experiences%20with%20Primary%20Care/2004_survey_
charts%20pdf.pdf (Accessed April 23, 2014).
(20.) Schoen C, Osborn R, Bishop M, How S. The Commonwealth Fund
2007 International Health Policy Survey in Seven Countries. 2007.
Available at: http://www.commonwealthfund.org/-/media/Files/Surveys/
2007/2007%20International%20Health%20Policy%20Survey%20in%20
Seven%20Countries/Schoen_intlhltpolicysurvey2007_chartpack%20pdf. pdf
(Accessed April 23, 2014).
(21.) Schoen C, Osborn R. The Commonwealth Fund 2010 International
Health Policy Survey in eleven countries. 2010. Available at:
http://www.commonwealthfund.org/-
/media/Files/Publications/In%20the%20Literature/2010/
Nov/Int%20Survey/PDF_2010_IHP_Survey_Chartpack_FINAL_white_
bkgd_111610_ds.pdf (Accessed April 23, 2014).
(22.) Joreskog KG, Sorbom D. LISREL 8: User's Reference Guide,
1996.
(23.) S. A. S. Institute. SAS User's Guide: Statistics. Cary,
NC: SAS Institute Inc., 2003.
(24.) Shengelia B, Murray CJL, Adams OB. Beyond access and
utilization: Defining and measuring health system coverage. In: Murray
CJL, Evans DB (Eds.), Health Systems Performance Assessment: Debates,
Methods, and Empiricism. Geneva, Switzerland: WHO, 2003;221-34.
(25.) Haggerty J, Fortin M, Beaulieu MD, Hudon C, Loignon C,
Preville M., et al. At the interface of community and healthcare
systems: A longitudinal cohort study on evolving health and the impact
of primary healthcare from the patient's perspective. BMC Health
Serv Res 2010;10(1):258. PMID: 20815880. doi: 10.1186/1472-6963-10-258.
(26.) Lemoine O, Simard B, Provost S, Levesque J-F, Pineault R,
Tousignant P. Rapport descriptif global de l'enquete
populationnelle sur l'experience de soins a Montreal et en
Monteregie. 2011. Available at: http://www.mspq.
qc.ca/pdf/publications/1372_RappDescripGlobalEnquetePopuExperienceSoinsMtlMonteregie.pdf (Accessed April 23, 2014).
Received: May 20, 2014
Accepted: December 26, 2014
Jeannie L. Haggerty, PhD, [1] Jean-Frederic Levesque, MD, PhD [2]
Author Affiliations
[1.] Department of Family Medicine, McGill University; St.
Mary's Hospital Research Centre, Montreal, QC
[2.] Centre for Primary Health Care and Equity, University of New
South Wales, Sydney, NSW
Correspondence: Jeannie L. Haggerty, St. Mary's Hospital
Research Centre, Hayes Pavilion--Suite 3734, 3830 Lacombe Avenue,
Montreal, QC H3T 1M5, Tel: [??] 514-345-3511, ext. 6334, E-mail:
jeannie.haggerty@mcgill.ca
Acknowledgements: This study was funded by the Canadian Health
Services Research Foundation (now Canadian Foundation for Health
Improvement) and by the Centre de recherche de l'Hopital Charles
LeMoyne. We gratefully acknowledge Prof. Daniele Roberge, Universite de
Sherbrooke, who contributed to the conception and analysis of the
qualitative study. Questionnaire preparation and field work were
overseen by Christine Beaulieu and the statistical analyses were
conducted by Fatima Bouharaoui.
Conflict of Interest: None to declare.
Table 1. Socio-demographic, health and health care
utilization characteristics of validation samples
Characteristic Initial sample Repeat sample
(n = 750) Respondents
(n = 316)
Socio-demographics
Mean age (SD) 50.9 (15.5) 51.4 (14.4)
Percent female 69.6% (522) 71.8% (227)
Geographic context
Urban 33.3% (250) 31.7% (100)
Rural 33.3% (250) 34.8% (110)
Remote 33.3% (250) 33.5% (106)
Percent with high school 55.0% (402) 51.0% (161)
education or less
Self-perceived financial
status
Comfortable 68.2% (502) 66.1% (208)
Tight 22.6% (166) 23.5% (73)
Very tight or poor 9.2% (68) 9.6% (30)
Health characteristics
Percent rating health as very 53% (397) 53.9% (170)
good or excellent
Percent with a chronic 36.7% (274) 39.8% (125)
physical condition
Health care utilization
Have a personal physician 81.2% (609) 84.5% (267)
Used health care services in 87.3% (655) 89.2% (282)
last year
Characteristic Repeat sample
Non-respondents Test for
(n = 434) difference (repeat)
Socio-demographics
Mean age (SD) 50.6 (16.2) T = -0.72; p = 0.47
Percent female 68.0% (295) [chi square] = 1.29;
1df; p = 0.26
Geographic context
Urban 34.6% (150) [chi square] = 0.83;
2df; p = 0.66
Rural 32.3% (140)
Remote 33.2% (144) [chi square] = 7.87;
3df; p = 0.049
Percent with high school 58.0% (251)
education or less
Self-perceived financial
status
Comfortable 69.2% (294) [chi square] = 0.97;
4df; p = 0.91
Tight 21.9% (93)
Very tight or poor 8.9% (38)
Health characteristics
Percent rating health as 52.3% (227) [chi square] = 3.47;
very good or excellent 4 df; p = 0.48
Percent with a chronic 34.5% (149) [chi square] = 2.21;
physical condition 1df; p = 0.14
Health care utilization
Have a personal physician 78.8% (342) [chi square] = 3.88;
1df; p = 0.048
Used health care services in 85.9% (373) [chi square] = 1.80;
last year 1df; p = 0.18
Table 2. Health care affordability subscale items and scoring
with descriptive statistics and comparison by income status,
second administration (n = 316)
Item Not applicable True missing
% (n) * % (n)
Score for response options
1. Are there times when you don't 10.4 (33) 1.3(4)
take drugs prescribed by a doctor
because of their costs?
0 = Never, rarely
1 = Sometimes, Often, Very often
2. Are there times when you don't 18.7 (59) 1.3(4)
take laboratory tests or
exams because of their costs?
0 = Never,rarely
1 = Sometimes, Often, Very often
3. Are there times when you decide 34.5 (109) 1.6 (5)
not to get services prescribed by a
doctor but not covered by health
insurance (such as physiotherapy,
psychotherapy or nutrition)
because of their costs?
0 = Never, rarely
1 = Sometimes, Often, Very often
4. Are there times when you find it (0) 0.6 (2)
difficult to get health care
services because of the loss of
income it involves?
0 = Never, rarely
1 = Sometimes, Often, Very often
5. Are there times when you find it (0) 1.0 (3)
difficult to get health care services
because of the additional costs it
involves (babysitting, parking,
etc.)?
0 = Never, rarely
1 = Sometimes, Often, Very often
Overall scores
Population: average percent
reporting cost-related lack of
coverage of medically needed
services (mean of item percentage)
Population: percent reporting at
least one problem with health care
affordability
Individual: sum of cost barriers to
medically needed services (sum of
item scores)
Item Overall occurrence
% (n/N) ([dagger])
Score for response options
1. Are there times when you don't 9.7 (27/279)
take drugs prescribed by a doctor
because of their costs?
0 = Never, rarely
1 = Sometimes, Often, Very often
2. Are there times when you don't 6.7 (17/253)
take laboratory tests or
exams because of their costs?
0 = Never,rarely
1 = Sometimes, Often, Very often
3. Are there times when you decide 27.2 (55/202)
not to get services prescribed by a
doctor but not covered by health
insurance (such as physiotherapy,
psychotherapy or nutrition)
because of their costs?
0 = Never, rarely
1 = Sometimes, Often, Very often
4. Are there times when you find it 10.5 (33/314)
difficult to get health care
services because of the loss of
income it involves?
0 = Never, rarely
1 = Sometimes, Often, Very often
5. Are there times when you find it 6.7 (21/313)
difficult to get health care services
because of the additional costs it
involves (babysitting, parking,
etc.)?
0 = Never, rarely
1 = Sometimes, Often, Very often
Overall scores
Population: average percent 12.2
reporting cost-related lack of
coverage of medically needed
services (mean of item percentage)
Population: percent reporting at 27.9
least one problem with health care
affordability
Individual: sum of cost barriers to 0.49 (SD = 0.96)
medically needed services (sum of
item scores)
Item Occurrence by
income level
Low n =174
Score for response options
1. Are there times when you don't 12.3 (19/154)
take drugs prescribed by a doctor
because of their costs?
0 = Never, rarely
1 = Sometimes, Often, Very often
2. Are there times when you don't 10.4 (14/135)
take laboratory tests or
exams because of their costs?
0 = Never,rarely
1 = Sometimes, Often, Very often
3. Are there times when you decide 29.7 (30/101)
not to get services prescribed by a
doctor but not covered by health
insurance (such as physiotherapy,
psychotherapy or nutrition)
because of their costs?
0 = Never, rarely
1 = Sometimes, Often, Very often
4. Are there times when you find it 13.8 (24/174)
difficult to get health care
services because of the loss of
income it involves?
0 = Never, rarely
1 = Sometimes, Often, Very often
5. Are there times when you find it 9.2 (16/174)
difficult to get health care services
because of the additional costs it
involves (babysitting, parking,
etc.)?
0 = Never, rarely
1 = Sometimes, Often, Very often
Overall scores
Population: average percent 15.1
reporting cost-related lack of
coverage of medically needed
services (mean of item percentage)
Population: percent reporting at 31.6
least one problem with health care
affordability
Individual: sum of cost barriers to 0.59 (SD = 1.08)
medically needed services (sum of
item scores)
Item Occurrence by
income level
High n = 118
Score for response options
1. Are there times when you don't 5.7 (6/105)
take drugs prescribed by a doctor
because of their costs?
0 = Never, rarely
1 = Sometimes, Often, Very often
2. Are there times when you don't 3.0 (3/100)
take laboratory tests or
exams because of their costs?
0 = Never,rarely
1 = Sometimes, Often, Very often
3. Are there times when you decide 25.0 (22/88)
not to get services prescribed by a
doctor but not covered by health
insurance (such as physiotherapy,
psychotherapy or nutrition)
because of their costs?
0 = Never, rarely
1 = Sometimes, Often, Very often
4. Are there times when you find it 5.9 (7/118)
difficult to get health care
services because of the loss of
income it involves?
0 = Never, rarely
1 = Sometimes, Often, Very often
5. Are there times when you find it 3.4 (4/118)
difficult to get health care services
because of the additional costs it
involves (babysitting, parking,
etc.)?
0 = Never, rarely
1 = Sometimes, Often, Very often
Overall scores
Population: average percent 8.6
reporting cost-related lack of
coverage of medically needed
services (mean of item percentage)
Population: percent reporting at 23.7
least one problem with health care
affordability
Individual: sum of cost barriers to 0.36 (SD = 0.78)
medically needed services (sum of
item scores)
* Those indicating no need of this service; e.g., don't need
medications, laboratory tests, complementary health services.
([dagger]) Among those for whom it is applicable.
Table 3. Association between health care affordability subscale
score and indicators of access difficulty, overall and by income
Consequence Prevalence Overall
Odds ratio &
predictive capacity
Odds ratio *
(95% CI)
Nuisance * 24.1% (74) 1.52 * (1.17, 1.98)
Unmet needs 21.9% (67) 1.54 * (1.17, 2.03)
Emergency 12.3% (38) 1.41 ** (1.02, 1.95)
room use
([section])
Problem 8.4% (26) 1.80 *** (1.29, 2.51)
aggravation
Consequence Overall
Odds ratio &
predictive capacity
Pseudo Somer's D
[R.sup.2]
Nuisance * 0.13 0.39
Unmet needs 0.13 0.41
Emergency 0.05 0.26
room use
([section])
Problem 0.23 0.57
aggravation
Consequence By income grouping
Low income High income
Odds ratio * Odds ratio *
(95% CI) (95% CI)
Nuisance * 1.74 *** (1.24, 2.43) 1.12 (0.62, 2.03)
Unmet needs 1.81 *** (1.28, 2.56) 1.34 (0.73, 2.47)
Emergency 1.40 (0.98, 2.01) 1.21 (0.55, 2.67)
room use
([section])
Problem 2.44 *** (1.56, 3.82) 0.42 (0.06, 3.11)
aggravation
Note: Odds ratios indicate likelihood of consequence with each
unit increase in score, i.e., each affordability problem
regardless of dimension.
* p <0.01; ** p <0.05; *** p <0.001.
([dagger]) Odds ratio for every unit increase in subscale score
from 0 to 5, controlling for physical limitations and chronic
illness.
* Patients reporting frequency as sometimes or often.
([section]) Emergency room use for health system reasons only
(e.g., doctor not available, clinic not open, confused about what
to do).
