Health Policy and Planning, 34, 2019, ii45–ii55
doi: 10.1093/heapol/czz109
Supplement Article
Diabetes epidemics: inequalities increase the
burden on the healthcare system
Carolina Santamarı́a-Ulloa 1,*, Melina Montero-López1 and
Luis Rosero-Bixby2
1Instituto de Investigaciones en Salud, Universidad de Costa Rica, San José, Costa Rica, 2Centro Centroamericano
de Población, Universidad de Costa Rica, San José, Costa Rica
*Corresponding author. Instituto de Investigaciones en Salud, Universidad de Costa Rica, De la UNED carretera a
Sabanilla, 200 m este y 125 m sur, San José, Costa Rica. E-mail: carolina.santamaria@ucr.ac.cr
Accepted on 5 September 2019
Abstract
Diabetes is a major cause of morbidity and mortality and represents a source of demands on
already constrained healthcare systems in Latin America and the Caribbean. We estimate inequal-
ities in diabetes incidence, prevalence and mortality and assess the economic burden on the
healthcare system in Costa Rica. The main source of data is the Costa Rican Longevity and Healthy
Aging Study, a longitudinal nationally representative survey of the elderly population (n¼ 2827).
Data analyses include descriptive statistics, multiple regression models and survival analysis
models. More than a fifth of Costa Rican elderly experience diabetes. Incidence is estimated at
5 per 1000 person-years in the population 30þ. Gender and geographical inequalities were found.
Men have a significantly lower prevalence (16.51% vs 24.02%, P< 0.05) and incidence (4.3 vs 6.0
per 1000 person-years, P< 0.05), but higher mortality (hazard ratio¼ 1.31, P< 0.01). Longer time to
the closest facility translates into a lower probability of having the condition diagnosed [odds ratio
(OR)¼0.77, P< 0.05]. The diabetic as compared to the non-diabetic population imposes a larger
economic burden on the healthcare system with a higher probability of using outpatient care
(OR¼ 3.08, P<0.01), medications (OR¼ 3.44, P< 0.01) and hospitalizations (OR¼ 1.24, P>0.05).
Individuals living in the Metro Area have a significantly lower probability of being hospitalized
(OR¼ 0.72, P< 0.05), which may be evidence of better access to primary care that prevents
hospitalization. Along the same line, women have higher utilization rates of outpatient care
(OR¼ 2.02, P<0.01) and medications (OR¼ 1.73, P< 0.01), which may contribute to lower odds of
hospitalization (OR¼ 0.61, P<0.01). Aligned with the aim of attaining Sustainable Development
Goals, this study highlights the importance of generating health policies focused on prevention of
diabetes that take into consideration gender and geographical inequalities. Strategies should
booster preventive healthcare utilization by men and aim to make healthcare services accessible
to all, regardless of geographical location.
Keywords: Developing countries, diabetes, healthcare system, inequalities
Introduction role. To some extent, urbanization is a proxy for lifestyle changes,
In virtually all populations, hyper-caloric diets and decreased including a more sedentary lifestyle, and increased obesity
physical activity have accompanied the benefits of modernization. (Shaw et al., 2010). There is evidence that the principal, albeit not
These changes, which have led to an increasing prevalence of obes- exclusive, driver of the DM2 epidemic is overweight and obesity,
ity, combined with increasing longevity have formed the basis for especially abdominal fat deposition (Hu et al., 2001).
dramatic increases in the prevalence of type 2 diabetes (DM2) Diabetes is a major cause of both morbidity and mortality in the
worldwide. Urbanization processes in developing countries play a elderly. DM2 is a well-established risk factor for coronary heart
VC The Author(s) 2019. Published by Oxford University Press in association with The London School of Hygiene and Tropical Medicine. All rights reserved.
For permissions, please e-mail: journals.permissions@oup.com ii45
Downloaded from https://academic.oup.com/heapol/article-abstract/34/Supplement_2/ii45/5625033 by UCR user on 13 November 2019
ii46 Health Policy and Planning, 2019, Vol. 34, Suppl. 2
Key Messages
• Gender inequality was found in the context of the diabetes epidemic in the Costa Rican elderly. Although men have
significantly lower prevalence and incidence of diabetes, they face a higher mortality than women.
• Geographical inequality was also described. The longer the time to the closest healthcare facility the lower the probabil-
ity of having the condition diagnosed.
• The diabetic as compared to the non-diabetic population imposes a larger economic burden on the healthcare system
with a higher probability of using outpatient care, medications and hospitalizations.
• Aligned with the aim of attaining Sustainable Development Goals, this study highlights the importance of generating
health policies focused on prevention of diabetes that take into consideration gender and geographical inequalities.
disease, cardiovascular and cerebrovascular diseases. Furthermore, Costa Rican medicine is highly socialized (Rosero-Bixby, 1996).
hypertension is more prevalent in the diabetic population (Grundy The health system includes a public and a private sector. The public
et al., 1999; Barceló, 2000). This condition is therefore increasingly sector is led by the Costa Rican Social Security Fund (CCSS, for its
constraining the healthcare systems and imposing a high economic Spanish acronym) whose main functions are those of financing, pur-
burden (Barceló et al., 2003; Palloni et al., 2006). Higher propor- chasing and delivering most of the healthcare services. CCSS delivers
tions of elderly combined with an increasing number of diabetic healthcare in three levels with differential resolution capacity.
individuals—who have higher risk of premature mortality—have The first level corresponds to Basic Teams for Comprehensive
made diabetes a challenge for the medical care systems in Latin Healthcare (EBAIS, for its Spanish acronym) along with outpatient
America (Barceló, 2000; Barceló et al., 2006) and in developed and peripheral clinics. The second level includes peripheral and
regions (Sloan et al., 2008; Solli et al., 2010). regional hospitals. The third level includes national and specialized
The population composition will play an important role in the hospitals (Sáenz et al., 2011). The private sector offers outpatient
impact the diabetes epidemic will have in the years to come. In care and specialty services, and it is mainly financed by out-of-
Costa Rica, like in other Latin American countries, the population pocket payments.
aging process is clearly on its way. In 1950, only 48 000 people A series of reforms have been under way since the 1980s
(6%) were aged 60 or older. Currently, in 2019, there are already (Villalobos and Piedra, 1998). This resulted in a better allocation of
650 000 people (13%) aged 60 or older. It is projected that in 2050 resources to the CCSS and has been associated with a reduction of
a total of 1.7 million will belong to this age group, which will inequalities and a strengthening of access to primary healthcare
represent nearly 30% of the total population (Figure 1). (Sáenz et al., 2011). Currently, the Costa Rican healthcare system
As a consequence of demographic aging and the prevalence of is focused on non-communicable diseases (Organización para la
diabetes risk factors, such as obesity, the elderly population with Cooperación y el Desarrollo Económicos (OCDE), 2017), which are
DM2 is expected to continue growing. Actions to reduce diabetes complex and multi-causal (Sáenz et al., 2011). Bolstering access to
will therefore have noticeable results only in the medium or in the primary healthcare will be crucial to effectively reduce inequalities.
long term. Nonetheless, if the diabetes epidemic is left unaddressed, Little is known about diabetes epidemics among elderly popula-
the burden on both the population and the healthcare system will tions in Latin America because most epidemiologic studies have
increase to unsustainable proportions. focused on the general population or on younger population
Figure 1 Population distribution in three large age groups. Costa Rica: 1950–2050.
Downloaded from https://academic.oup.com/heapol/article-abstract/34/Supplement_2/ii45/5625033 by UCR user on 13 November 2019
Health Policy and Planning, 2019, Vol. 34, Suppl. 2 ii47
segments (Barceló et al., 2007). An estimation of the burden of System (Death Index). For the purpose of this study, mortality was
diabetes in terms of healthcare will be an important input for the tracked up to 31 October 2017.
establishment of public policy that is relevant not only to Costa Rica
but also to other developing countries facing similar scenarios. Classification of individuals’ diabetes status
In the search to meet the Sustainable Development Goals Diabetes status was defined based on self-report of a diagnosis by a
(SDGs), attempts to reduce inequalities are important. By providing medical doctor. Self-reports have high specificity (ability to identify
evidence of inequalities that need to be addressed, this research con- correctly those who ‘do not have the diagnosis’), but lower sensitiv-
tributes to SDG goal number 3: ‘Ensure healthy lives and promote ity (ability to identify correctly those who ‘have the disease’).
well-being for all at all ages’. As stated by Chaparro-Dı́az (2016), Nonetheless, self-reports are still the easiest and most widely used
efforts towards reducing inequalities are also related to SDG goal way to measure health conditions in population studies.
number 1: ‘End poverty in all its forms everywhere’. Diabetes as well
as many chronic diseases is related to poverty and increases costs to
Sociodemographic characteristics
the healthcare system due to a higher demand of public healthcare
services. Finally, by generating scientific knowledge that may be Age and sex were used as sociodemographic characteristics in the
applicable to other developing countries, this research contributes to prevalence, incidence and mortality models, and they were used as
SDG goal 10: ‘Reduce inequality within and among countries’. predisposing characteristics in the economic burden models. Age
The objective of this study is to estimate inequalities in diabetes was used as a continuous variable. Sex was a dichotomous variable,
incidence, prevalence and mortality, and to provide estimates of the with female as the reference category.
elderly diabetes burden on the healthcare system. This research is an Education was used as a sociodemographic variable in the preva-
input for decision makers in terms of allocation and efficient use of lence and mortality models and as a personal enabling resource in the
resources. Although information alone does not translate into economic burden models. It was a dichotomous variable that refers
policymaking to reduce inequalities, this study provides evidence of to incomplete or complete primary school (reference category).
the need for health promotion and prevention programmes aimed at Income was used as a sociodemographic variable in the
reducing specific diabetes inequalities. prevalence model. It was a dichotomous variable that refers to low
or high income. The cut-off point is 100 US dollars (USD) 2011 per
elderly individual per month. That is the elder’s own income if not
married, or the average of the couple’s monthly income if married.
Methods USD100 was considered the minimum income for an elderly person
to cover his or her expenses over a 1-month period during the time
Data analyses and estimations were conducted with STATA com-
period of the baseline survey. This cut-off point for income has also
puter software (StataCorp, 2013). The analyses include descriptive
been used in similar studies with CRELES (Méndez-Chacón et al.,
statistics, multiple regression models and survival analysis models
2008; Brenes-Camacho and Rosero-Bixby, 2008b).
depending on the nature of the phenomena to be described.
The main source of data is the Costa Rican Longevity and
Healthy Aging Study (CRELES, for its Spanish acronym), a nation- Diabetes risk factors
ally representative longitudinal survey of health and life-course Two variables were used as diabetes risk factors in the prevalence
experiences of Costa Ricans ages 60 and over in 2005. As with any model: family history of diabetes and a combined measure of waist
elderly survey, data are subject to selection bias since they refer circumference (WC) and body mass index (BMI).
solely to the population that had survived at least to the age of 60 in Family history of diabetes was a dichotomous variable. It
2005 and was therefore eligible for the study. A baseline (n¼2827) referred to whether or not (reference category) any of the individu-
and two subsequent 2-year follow-up interviews were conducted. al’s parents, siblings or grandparents had ever had the condition.
Data collection occurred between 2004 and 2006 for the baseline, Obesity is known to be the main risk factor for diabetes. There
between 2006 and 2007 for the second wave and between 2008 are different indicators of obesity. BMI is an indicator of general
and 2009 for the third wave. Loss to follow-up between baseline obesity and WC of central obesity. According to BMI, individuals
and wave 2 was 7% and between waves 2 and 3, it was 9% of the were classified as (1) underweight: <18.5, (2) normal: 18.5–24.9,
baseline sample. (3) overweight: 25.0–29.9 or (4) obese: 30.0 (WHO, 2000).
The baseline sample was randomly drawn from Costa Rican Participants were also classified in the following WC categories.
residents in the 2000 population census who were born in 1945 or For men: (1) normal: <94, (2) increased: 94–101 or (3) substantially
earlier, regardless of their nationality. It was stratified by 5-year age increased: 102 cm. For women: (1) normal: <80, (2) increased:
groups in order to have similar sample sizes in each age group. This 80–87 or (3) substantially increased: 88 (WHO, 2000).
strategy, that implied an over sampling of older people, assured a WC and BMI combined categories were combined into a more
sufficiently large number of observations for advanced ages. Because comprehensive measure of obesity and were used as follows: (1) nor-
of the use of sampling weights in the analyses conducted, oversam- mal WC, normal BMI; (2) normal WC, overweight or obese; (3)
pling of the oldest adults does not have an impact on the findings increased or substantially increased WC, normal BMI; (4) increased
reported in this study. Details on the estimation of sampling weights WC, overweight or obese; (5) substantially increased WC, over-
for the CRELES survey have been previously published (Rosero- weight; and (6) substantially increased WC, obese.
Bixby et al., 2013).
This survey collected information on different health conditions, Behavioural health risks
living conditions, health behaviours, healthcare utilization and Three variables were used as behavioural health risks in the preva-
socioeconomic status, among other variables. Baseline data, wave 2 lence model: smoking, alcohol consumption and hyper-caloric diet.
and wave 3 were all used to follow individuals regarding diabetes Questions regarding active smoking behaviour in this study were
diagnosis and mortality. Additionally, mortality was tracked by asked only to those who had smoked 100 or more cigarettes or
linking the CRELES dataset with the National Vital Registration cigars during their lives. Those who were not current active smokers
Downloaded from https://academic.oup.com/heapol/article-abstract/34/Supplement_2/ii45/5625033 by UCR user on 13 November 2019
ii48 Health Policy and Planning, 2019, Vol. 34, Suppl. 2
but lived with a smoking partner were classified as passive smokers. Need is the most proximate determinant of utilization, and
Individuals were categorized as: (1) never smoked (reference cat- it varies as a function of the predisposing and enabling factors.
egory), (2) former active or passive smoker, (3) current passive Cultural factors can also influence need, but no measurement of
smoker and (4) current active smoker. Information on alcohol refers such variables was available in the CRELES questionnaire. The fol-
to alcoholic drinks ever consumed along individual’s lives. lowing variables were included: having poor self-perceived health,
Individuals were categorized into: (1) never (reference category), (2) having at least one limitation in activities of daily living (ADL),
former and (3) current alcohol drinker. having at least one limitation in instrumental activities of daily living
The estimation of calorie daily consumption was made from a (IADL) and having been diagnosed with diabetes, cancer, lung dis-
tracer food consumption questionnaire that was part of the baseline ease, cardiovascular disease (myocardial infarction, ischaemic heart
interview. A cut-off point of 3000 kcal/day was used. This value is a disease or stroke), hypertension, dyslipidaemia (hypercholesterol-
standard cut point associated with differential risk of cardiovascular aemia or hypertriglyceridaemia), arthritis or osteoporosis. They
disease (Brown, 2008) and has been used in similar population were all treated as dichotomous variables.
studies (Méndez-Chacón et al., 2008; Rosero-Bixby and Dow, Point prevalence rates of diabetes and their 95% confidence
2009; Rehkopf et al., 2010). intervals were estimated by sex. Logistic regression models were
used to analyse the relationship among individuals’ diabetes status
Chronic morbidity and sociodemographic characteristics (age, sex, education, income),
Three comorbidities were included in the prevalence model: diabetes risk factors (family history of diabetes, WC—BMI com-
hypertension, elevated high-density lipoprotein (HDL)/low-density bined measure), behavioural health risks (smoking, alcohol drink-
lipoprotein (LDL) cholesterol, and elevated triglycerides. Six chronic ing, hyper-caloric diet), access to healthcare (health insurance, living
conditions were included in the mortality analysis: diabetes, cancer, in the GMA, time to closest facility) and comorbidities (hyperten-
lung disease, myocardial infarction, ischaemic heart disease (not sion, HDL/LDL cholesterol, triglycerides).
infarction), and stroke. The estimation of incidence relies on self-reporting on how old
Chronic morbidities were self-reported, they refer to whether the subject was or what the date was when diabetes diagnosis
or not the individual has ever been told by a medical doctor to have occurred. When reporting a previous medical diagnosis, people will
the condition and were defined as dichotomous variables, with not likely offer approximate rather than exact dates of the diagnosis of
having the condition as the reference category. their condition. However, there is no way to attenuate this bias since
no external sources of data, such as medical records, were used.
To avoid potential confounding with type 1 diabetes, only indi-
Access to healthcare viduals reporting a diagnosis of diabetes at 30 years or older were
Three variables related to healthcare access were included in the included in incidence models. This criterion has been used in other
prevalence model: having a health insurance, living in the Great population studies (Hu et al., 2001, 2007; Suh et al., 2008).
Metropolitan Area (GMA), and mean time to the nearest healthcare This estimation of adult population incidence is therefore a re-
facility. construction that relies on both retrospective and prospective data
Following Gulliford et al. (2002), the choice of the aforemen- on the timing of diabetes diagnosis. Prospective information comes
tioned variables responds to access being measured in terms of from the three waves of the longitudinal study, and retrospective
utilization of available services, which depends on affordability, data come from participants’ recall back to their age of 30. Because
physical accessibility and acceptability of services. No measurements CRELES sampling weights were estimated to reproduce the struc-
regarding acceptability were available. Access to healthcare was ture of the Costa Rican elderly population in 2005 (Rosero-Bixby
operationalized in terms of affordability (having a health insurance) et al., 2013), and this is rather a reconstruction of adult population
and physical accessibility (living in the GMA and mean time to the rates, CRELES weights were not used for incidence estimates.
nearest healthcare facility). Living in the GMA was used as a Incidence rates and their 95% confidence intervals were esti-
measure of physical accessibility because that is the geographical mated for total population and by sex, no control variables were
area where most important healthcare facilities are clustered in used for these estimates. The data weres set as survival time. Follow-
Costa Rica. up time starts at the date each individual was aged 30. Censoring
occurs when individuals are lost to follow-up—either because of
Determinants of healthcare services utilization death or because of other reasons—or at the time of interview in the
Variables used in the models of economic burden on the healthcare third wave. Starting at the age of 30, incidence rates were computed
system are based on Andersen’s theoretical model of healthcare use as the ratio of new diabetes diagnoses to the exact count of person-
(Andersen, 1995), which is widely used and adjusts to the Costa years.
Rican context (Brenes-Camacho and Rosero-Bixby, 2009; Llanos Parametric survival models with a log-logistic distribution for
et al., 2009). This model proposes that the use of healthcare services the baseline hazard were used to model incidence. This distribution
is mediated by the interaction of predisposing characteristics, ena- has a fairly flexible functional form (Hosmer and Lemeshow, 1999).
bling resources, and need (Andersen, 1995). Reason to select this distribution is that the incidence process does
Predisposing characteristics included in the model are: age, sex, not grow monotonically. DM2 incidence increases from the age of
living in the GMA, married or in union, and retired. Age has been 30 up to a certain point around the age of 60, and then starts going
included as a continuous variable, the rest of predisposing character- down at older ages. The log-logistic function can effectively repre-
istics were dichotomous variables. sent a pattern of increasing incidence, followed by a decrease and
Enabling factors refer to conditions that allow a greater avail- was therefore selected for this analysis.
ability and access to the services. They include individual (education Because of the longitudinal nature of the data, parametric
and income) as well as contextual (mean time to the nearest health- regression models were used to estimate the association between
care facility) characteristics. Time has been included as a continuous mortality and sociodemographic characteristics (sex and educa-
variable, the rest of enabling factors were dichotomous variables. tion) and chronic morbidity (diabetes, cancer, lung disease,
Downloaded from https://academic.oup.com/heapol/article-abstract/34/Supplement_2/ii45/5625033 by UCR user on 13 November 2019
Health Policy and Planning, 2019, Vol. 34, Suppl. 2 ii49
myocardial infarction, ischaemic heart disease and stroke). The to urban areas (39 vs 27min, t¼2.86, P¼0.004) and outside the
data were set as survival time. Follow-up time starts at the date GMA (38 vs 26 min, t¼2.85, P¼0.004).
each individual was 60. Respondent’s vital status was assessed Women have higher prevalence of general and central obesity as
during the three waves of CRELES and it was also tracked by link- measured by BMI, WC and the combined measure of both variables.
ing the CRELES dataset with the National Vital Registration Obesity prevalence decreases with age. Smoking and especially alco-
System (the Death Index) up to 31 October 2017. Mortality hol drinking is higher in males, and both risk behaviours decrease
rates were computed as the ratio of deaths to the exact count of with age (Table 1).
person-years. Chronic comorbidities are common in the elderly. Hypertension
Parametric survival models with a Gompertz distribution for the and hypertriglyceridaemia are highly prevalent in Costa Rican
baseline hazard were used to model mortality (Hosmer and population. Hypertension is the most common cardiovascular dis-
Lemeshow, 1999). Costa Rican mortality rates have been shown to ease and the most common comorbid condition for diabetic elderly:
follow a Gompertz function, especially after the age of 45 (Rosero- 82% of diabetic are hypertensive, as compared to 59% of non-
Bixby and Antich, 2010). diabetic who are hypertensive.
Two-part models were used to analyse the factors that affect the Diabetes incidence estimation according to CRELES is at least
propensity to use hospitalizations, outpatient consultations and 5.2 per 1000 people aged 30 and above. As well as in prevalence,
medications and those that affect the volume of utilization once the gender inequalities exist in incidence, with a significantly higher rate
person makes use of them. This is a common tool used in health eco- in the female adult population (Table 2).
nomics applications in which the outcomes are measures of health- Results of a longitudinal model of general mortality are shown
care utilization (Diehr et al., 1999). It basically assumes that the in Table 3. When controlling for sociodemographic characteristics
probability of the outcome is >0 given a set of covariates is gov- and other chronic morbidity as the most proximate determinants of
erned by a binary probability model. That is part 1, and is usually mortality, diabetes is significantly associated with higher mortality.
modelled as a logistic regression, as is the case in this study. It also Men have higher mortality rates than women (hazard ratio¼1.31,
assumes that the expected logarithm of the outcome given that the P<0.01).
outcome is >0, and given the same covariates, is a linear function of Geographical inequalities translate into a lower probability of
those covariates. That is part 2, and has been modelled as a general- having the condition diagnosed. As time to the closest facility
ized linear model (GLM) for the three healthcare services analysed increases, the odds of having been diagnosed significantly decreases
in this study. [odds ratio (OR)¼0.77, P<0.05] (Table 4).
The main characteristics of these data are that the outcomes are Following the Andersen’s model of access to medical care,
positive numeric values, there is an important fraction of zeros, and the economic burden of diabetes on the public healthcare system
the non-zero outcomes are positively skewed (Manning and Mullahy, was estimated by modelling the probability and volume of utiliza-
2001). A gamma stochastic distribution with a log link has been used tion (Table 5) as well as the costs (Table 6) of outpatient care,
to estimate the parameters associated to each covariate in the part 2 hospitalizations and medications. CRELES data on healthcare
of the models. Using the gamma distribution is common in models to utilization had high completeness rates, which allowed models to be
explain healthcare services utilization and costs (Diehr et al., 1999). estimated in the base of 2189 cases (77% of baseline sample). Because
The expected levels of use of each of these three services were of the nature of data collected by the survey, both outpatient care and
estimated by multiplying the estimates of part 1 and part 2 of the hospitalizations costs were estimated over one calendar year.
two-part models. Each individual’s estimated utilization of health- Medications costs were estimated for those drugs that were currently
care services is his probability of having any use multiplied by the being taken at baseline, rather than for one calendar year.
expected volume of utilization conditional on being a user. Outpatient consultation probability (OR¼3.08, P<0.01) and
Economic cost is the dependent variable in each of the three its volume of use (OR¼1.11, P<0.01) are significantly higher for
models. Costs are inputted for each individual based on their volume diabetic individuals. Once diabetic individuals make use of out-
of utilization of each of the following services: hospitalizations over patient care, they have an 11% higher volume of utilization than
a calendar year, outpatient visits over a calendar year and medica- their non-diabetic counterparts (Table 5). Mean cost of outpatient
tions currently taken. Mean costs for these services as reported by care is 24% higher for the diabetic as compared to non-diabetic
the CCSS were used. Costs are reported in United States Dollars elderly (Table 6).
from the year 2011 (2011 USD). Hospitalization probability (OR¼1.24, P>0.05) and its vol-
ume of use (OR¼1.09, P>0.05) are higher for diabetic individuals,
although not statistically significant (Table 5). Mean hospitalization
Results
costs are 50% higher for diabetic elderly (Table 6).
At least one-fifth (20.5%) of Costa Rican elderly is diabetic Medications probability of utilization (OR¼3.44, P<0.01) and
(Table 1). Gender inequalities in prevalence exist in this population. its volume of use (OR¼1.28, P<0.01) are significantly higher for
Diabetes prevalence in Costa Rican elderly is significantly higher elderly, diabetic individuals. Once diabetic individuals make use of
among women [24.02%, 95% confidence interval (CI) 21.88– medications, they have a 28% higher volume of utilization than the
26.16] than men (16.51%, 95% CI 14.48–18.53). Diabetes preva- non-diabetic elderly (Table 5). Mean cost of medications for the dia-
lence is lowest in the oldest old (Table 1). betic elderly is almost twice the cost for non-diabetic (Table 6).
Out of 10 individuals, 7 have general or abdominal fat deposition Costs presented in Table 6 are particularly low for medications
that puts them at higher risk of diabetes. This elderly population because their estimation is based on the mean cost of a drug pre-
seems to have good access to healthcare. The great majority of them scription, which is the way the CCSS registers and provides mean
have health insurance; more than half of them live in the GMA, where costs of drugs.
the urban conditions make healthcare facilities geographically more Once controlling for diabetes and other comorbidities, individu-
accessible (Table 1). Mean time to the nearest healthcare facility is als living in the Metro Area have a significantly lower probability of
half an hour across the country, but it is longer in rural as compared being hospitalized (OR¼0.72, P<0.05), which may be evidence of
Downloaded from https://academic.oup.com/heapol/article-abstract/34/Supplement_2/ii45/5625033 by UCR user on 13 November 2019
ii50 Health Policy and Planning, 2019, Vol. 34, Suppl. 2
Table 1 Descriptive information of the CRELES Costa Rican elderly at baseline: 2004–2006
Characteristics, n¼ 2827 unless otherwise noted Total population Sex Age
Male Female 60–69 70–79 80þ
Sociodemographics
Education: % with complete primary 49.0 48.8 52.4 58.6 40.0 33.5
Low income, 2799 40.6 38.1 42.8 36.8 42.9 49.4
Risk factors
Waist circumference, 2632
Normal 31.6 51.5 13.5 30.5 30.9 37.6
Increased 21.3 24.7 18.2 22.2 18.9 23.7
Substantially increased 47.0 23.8 68.3 47.4 50.2 38.7
Body mass index, 2698
Underweight 13.3 12.3 14.1 9.4 13.0 28.9
Normal 39.6 41.4 37.9 38.2 40.1 43.9
Overweight 34.6 36.4 32.8 37.1 36.3 20.7
Obese 12.6 9.8 15.2 15.4 10.5 6.4
Waist circumference and BMI, 2627
Normal WC and BMI 21.4 31.9 11.8 17.6 22.9 32.8
Normal WC, overweight or obese 10.3 19.7 1.8 12.7 8.8 4.6
Increased or substantially increased WC, normal weight 10.0 2.9 16.5 7.5 9.9 19.8
Increased WC, overweight or obese 17.0 23.8 10.8 19.0 15.3 13.3
Substantially increased WC, overweight 18.0 7.3 27.9 17.0 20.3 16.8
Substantially increased WC, obese 23.2 14.4 31.3 26.1 22.9 12.7
Behavioural health risks
Smoking, 2810
Never 10.0 29.42 41.09 36.68 34.12 34.5
Former active or passive smoker 33.1 51.0 17.0 31.5 33.5 37.8
Current passive smoker 21.4 2.9 38.1 19.5 24.2 22.2
Current active smoker 35.5 16.8 3.8 12.3 8.2 5.4
Alcohol
Never 35.8 7.3 61.6 34.2 36.0 41.2
Former alcohol drinker 29.7 46.3 14.7 27.4 31.0 35.2
Current alcohol drinker 34.5 46.5 23.8 38.3 33.1 23.6
Calorie daily consumption 3000, 2819 12.3 16.0 9.0 13.9 11.6 7.9
Access to healthcare
Having health insurance 94.6 92.9 96.1 92.9 96.5 96.5
Living in the Great Metropolitan Area 53 50.7 55.0 51.1 55.1 55.0
Mean time to the nearest health facility, 2401 31.0 32.3 29.8 29.9 30.4 37.3
Health condition
Chronic morbidity
Diabetesa 20.5 17.0 24.2 21.1 23.2 14.2
Hypertension, 2823 64.5 59.8 68.8 61.3 68.8 67.2
Dyslipidaemia,b 2656 51.2 50.8 51.6 54.8 47.6 45.9
Elevated total/HDL cholesterol ratio, 2654 28.5 32.7 24.6 31.8 24.7 24.0
Elevated triglycerides, 2573 44.9 42.0 47.4 47.6 42.5 39.7
Cardiovascular disease
Myocardial infarction 4.6 5.6 3.7 3.0 6.3 6.8
Ischaemic heart attack (no infarction) 12 11.7 12.4 10.1 13.5 16.1
Stroke 3.8 3.5 4.1 2.1 4.7 8.2
Cancer 5.8 4.9 6.6 4.8 6.4 8.4
Lung disease 16.6 4.9 6.6 15.6 17.3 18.7
aDiabetes refers to self-report of MD diagnosis.
bDyslipidaemia refers to any or both: hypercholesterolemia (total/HDL ratio) and hypertriglyceridaemia.
better access to primary care that prevents hospitalization. Along the
estimates from this study, based on self-reporting of diagnosis, can be
same line, women have higher utilization rates of outpatient
¼ taken as a conservative estimate of an even greater magnitude epidem-care (OR 2.02, P<0.01) and medications (OR¼1.73, P<0.01), with
¼ ic. Self-reports are known to underestimate true prevalence since therea consequential lower odds of hospitalization (OR 0.61, P<0.01).
is a percentage of late-onset cases that go undiagnosed for a period of
time. High prevalence also occurs in other Latin American and
Caribbean (LAC) countries. In seven LAC cities, it is estimated to
Discussion range from 12.4% in Buenos Aires, Argentina to 21.7% in
Diabetes is highly prevalent. At least one-fifth of the Costa Rican eld- Bridgetown, Barbados (Andrade, 2006). Compared to the LAC re-
erly population has this condition. This is despite the fact that gion, the prevalence of diabetes in Costa Rica is among the highest.
Downloaded from https://academic.oup.com/heapol/article-abstract/34/Supplement_2/ii45/5625033 by UCR user on 13 November 2019
Health Policy and Planning, 2019, Vol. 34, Suppl. 2 ii51
Table 2 Incidencea rate of diabetes by sex (rates per 1000 person- Table 4 Odds ratios and confidence intervals from logistic regres-
years) sion models of diabetes prevalence
Population Incidence rate 95% CI Variables OR 95% CI
Total population 5.2 4.9–5.6 Sociodemographics
Female 6.0 5.5–6.6 Age 0.99 0.98–1.01
Male 4.3 3.8–4.9 Male 0.77* 0.56–1.05
Incidence rate ratiob 1.4*** 1.2–1.7 Complete primary school 0.77** 0.60–0.98
Low income 1.06 0.84–1.35
aIncidence estimated for adult ages 30 and above as reported by the Risk factors
subject. Family history of diabetes 2.57*** 2.05–3.21
bMantel-Haenszel estimates of the rate ratio. Normal WC and BMI 1.00
Significance level: ***P< 0.01. Normal WC, overweight or obese 1.41 0.86–2.33
Increased or substantially increased 1.64* 0.98–2.75
WC, normal weight
Table 3 Hazard ratios and confidence intervals from Gompertz lon- Increased WC, overweight or obese 2.10*** 1.38–3.21
gitudinal regression models of general mortality at age 60 and Substantially increased WC, overweight 2.11*** 1.36–3.26
above Substantially increased WC, obese 3.67*** 2.45–5.48
Behavioural health risks
Variables Self-report Never smoker 1.00
Former active or passive smoker 1.02 0.78–1.33
Hazard ratio 95% CI Current passive smoker 1.18 0.84–1.65
Current active smoker 2.24*** 1.54–3.26
Sociodemographics
Never drinker 1.00
Male 1.35*** 1.18–1.53
Former drinker 0.82 0.59–1.13
Complete primary school 1.14 0.99–1.31
Current drinker 1.06 0.78–1.43
Chronic morbidity
Calorie daily consumption 3000 0.85 0.59–1.21
Diabetes 1.57*** 1.35–1.83
Access to healthcare
Cancer 1.03*** 0.81–1.31
Has health insurance 1.05 0.62–1.79
Lung disease 1.16*** 0.99–1.36
Living in the Great Metropolitan Area 0.89 0.71–1.12
Myocardial infarction 1.40** 1.09–1.82
Time to the closest facility (min) 0.77** 0.62–0.96
Ischaemic heart 1.09*** 0.94–1.31
Comorbidities
disease (no infarction)
Hypertension 2.61*** 1.99–3.43
Stroke 1.31** 1.03–1.67
 Elevated HDL/LDL cholesterol 1.28* 0.99–1.67Log pseudolikelihood 1477.04
Elevated triglycerides 0.76** 0.60–0.97
Prob > Chi2 0.0000
Pseudo R2 0.1225
Prob > Chi2 0.0000
Significance levels: ***P< 0.01. **P< 0.05. *P< 0.10.
Significance levels: ***P< 0.01. **P< 0.05. *P< 0.10.
Metabolic conditions, hypertension and diabetes included, have
common risk factors. Similar to what has been reported in other studies show significant gender differences in prevalence, complica-
countries such as Mexico (Velázquez-Monroy et al., 2003; Instituto tions and mortality, those differences are not sufficiently explained
Nacional de Salud Pública (INSP), 2012), in Costa Rica the odds of (Sandı́n et al., 2011). In the USA and some cities in LAC, diabetes
hypertension are higher in the diabetic elderly, which increases the prevalence has been reported to be higher in elderly men (Barceló
burden both on individuals and on the healthcare system. et al., 2007). In Costa Rica, however, it is lower for men in all
Smoking (Willi et al., 2007; Kowall et al., 2010; Zhang et al., 10-year age groups. After controlling for other sociodemographic
2011) and alcohol consumption (Nakanishi et al., 2003; Beulens characteristics, risk factors, behavioural health risks and access to
et al., 2005; Baliunas et al., 2009; Pietraszek et al., 2010) have been healthcare, men are less likely to have a diabetes diagnosis in Costa
shown to be associated with diabetes. In this Costa Rican cohort, as Rica. Although this male advantage in terms of incidence may come
it has also been reported by Velázquez-Monroy et al. (2003), the in part from their lower prevalence of obesity; male disadvantage in
prevalence of diabetes increases as obesity and smoking increase. mortality may come from late diagnoses that can be related to a
Public policies should take into account diabetes risk factors, lower utilization of outpatient care. The difficulty in understanding
such as overweight, tobacco and alcohol consumption. Efforts to the sex differences that exist in the prevalence of diabetes has to do
reduce risks while promoting health can contribute to reductions in with the complexity of this disease. Obesity prevention, especially
inequalities surrounding the prevalence of diabetes, and promote a in women, and early detection, especially in men, are modifiable
healthy life and universal well-being, as described in the SDGs. factors on which health policy should focus.
The prevalence of diabetes among the elderly is higher at younger Having health insurance was not found to have a significant
ages and decreases with age. This holds not only for Costa Rica but association with diabetes prevalence. This may be explained by the
also for the Latin American population. The same age patterning of fact that Costa Rica has a universal healthcare system with no co-
diabetes prevalence has been observed in many Latin American cities payments associated to the use of services, which makes healthcare
(Barceló et al., 2006). Diabetes itself is associated with premature affordable to the entire population. Access to healthcare is a com-
mortality, which contributes to lower prevalence at older ages. plex concept. A population may have access if an adequate supply of
Gender differences in the diabetes epidemic have not been con- services is available, but the extent to which a population actually
sistent across the world (Ávila-Curiel et al., 2007). While some gains access to healthcare depends on financial, organizational and
Downloaded from https://academic.oup.com/heapol/article-abstract/34/Supplement_2/ii45/5625033 by UCR user on 13 November 2019
ii52 Health Policy and Planning, 2019, Vol. 34, Suppl. 2
Table 5 Results from two-part regression models of cost of outpatient care, hospitalizations and medications at baseline: 2004–2006
Determinants Outpatient care Hospitalizations Medications
Part 1—logistic Part 2—GLM Part 1—logistic Part 2—GLM Part 1—logistic Part 2—GLM
Odds ratios Exp(b) Odds ratios Exp(b) Odds ratios Exp(b)
Predisposing characteristics
Age 1.01 1.00 1.00 1.00 1.03 1.00
Female 2.02*** 1.06** 0.61*** 0.85 1.73*** 1.10
Live in the Metro Area 1.29 1.00 0.72** 1.32 1.97 1.10
Married or in union 1.30 1.13*** 0.84 0.48*** 1.42 1.07***
Retired 2.13*** 1.04 1.10 1.21 1.80*** 1.05
Enabling resources
Personal
Complete primary school 1.18 1.07** 1.08 0.64** 1.39 1.06**
Low income 1.13 0.91*** 0.79 1.40* 0.74 1.02***
Contextual
Mean time to nearest health facility 1.06 1.01 1.05 1.06 1.00 1.01
Need
Poor self-perceived health 1.37* 1.10*** 1.90*** 0.97 1.68* 1.13***
At least 1 ADL limitation 1.48** 1.05 1.51** 1.18 1.26** 1.09
At least 1 IADL limitation 0.89 1.08** 2.06*** 1.17 1.47 1.12**
Diabetes 3.08*** 1.11*** 1.24 1.09 3.44*** 1.28***
Hypertension 1.18 0.97 1.35* 1.39 3.42 1.42
Dyslipidaemia 0.64*** 0.98 0.70** 1.03 0.85*** 0.97
Cardiovascular disease 7.91*** 1.16*** 1.58*** 0.96 3.52*** 1.36***
Lung disease 2.99*** 1.06* 0.84 1.01 1.96*** 1.05*
Cancer 1.23 1.03 1.75** 1.01 0.70 1.01
Arthritis 1.46 1.09** 1.48** 0.85 1.75 1.15**
Osteoporosis 1.54 1.04 0.92 1.34 2.88 1.06
Pseudo R2 0.1461 0.0841 0.2559
Prob > Chi2 0.0000 0.0000 0.0000
AIC 27.43 29.69 21.31
BIC 14 345.83 1043.17 12 381.65
*P< 0.10, **P< 0.05, ***P< 0.001.
Table 6 Individual mean cost of outpatient carea, hospitalizationsa to disentangle what portion of inequality is driven by cultural
and medicationsb forces. This piece of information should be explored in future
researches.
Characteristics Mean cost 95% confidence interval
Longer mean times to the nearest healthcare facility were associ-
Outpatient carea ated with a decreased probability of having a medical diagnose of
Total population 337 334–341 diabetes. This is evidence of geographical barriers to healthcare that
Non-diabetic 319 316–322 translate into a lower probability of diagnosis. A previous study by
Diabetic 404 398–410 Brenes-Camacho and Rosero-Bixby (2008a) reported differential
Hospitalizationsa access to care in this same elderly population since individuals not
Total population 827 763–891 living in the GMA had a lower probability of having their diabetes
Non-diabetic 745 683–807
controlled. Inequalities in access to diabetes care can result from
Diabetic 1.124 930–1318
b various factors including the geographical distribution of healthMedications
Total population 20 20–20 services and therefore the distance needed to travel to have access to
Non-diabetic 17 16–17 them (Whiting et al., 2010).
Diabetic 31 31–32 Although most diabetes diagnoses occur during adulthood, the
lack of an official population registry of incident cases prevents the
Predicted mean costs from a two-part regression model (2011 USD). direct estimation of incidence rates. An indirect estimation was
aEstimated along one calendar year. therefore conducted based on the reconstruction of this elderly
bEstimated for prescribed drugs currently taken at baseline.
cohort back to age 30. Although subject to selection bias, this esti-
mation can be used to project the impact of this condition on the
social or cultural barriers. Access measured in terms of utilization of healthcare system. Diabetes, as it has also been shown in other stud-
available services is therefore dependent on affordability, physical ies, is associated with increased risks of all-cause mortality (Hu
accessibility and acceptability of services (Gulliford et al., 2002). et al., 2007). The most highly effective interventions to reduce mor-
Acceptability refers to the social and cultural influences that mediate bidity, premature mortality and the incidence of diabetes-related
access to healthcare. Leaving acceptability of healthcare services complications are both education for lifestyle change, and the cre-
out of this study because of a lack of data results in a limitation ation of environments in which individual behavioural initiatives
Downloaded from https://academic.oup.com/heapol/article-abstract/34/Supplement_2/ii45/5625033 by UCR user on 13 November 2019
Health Policy and Planning, 2019, Vol. 34, Suppl. 2 ii53
can succeed. As stated by Yach et al. (2006), overweight and obesity Besides the impact diabetes epidemic has on individuals lives,
have become to diabetes what tobacco is to lung cancer. Acting this condition clearly puts the public healthcare system under
on preventable risk factors for diabetes is therefore mandatory. pressure. Caring for diabetic elderly is more expensive in terms
Diabetes costs are on the rise around the world (Zhang et al., of hospitalizations, outpatient care and medications. The costs of
2010). In Latin America, the increase in the prevalence of DM2 has healthcare will increase because of the population aging process it-
already impacted healthcare systems. A 33% increase in economic self. But the impact of diabetes on these costs can be reduced if risk
spending for diabetes care between 2009 and 2011 has been factors are attenuated in the population and earlier diagnoses of the
reported in Mexico, and this figure is expected to rise (Arredondo condition are attained. Policies to reduce risk factors will not affect
and De Icaza, 2011). Countries such as Argentina have reported diabetes incidence in the short term but will do so in the medium
that deficiencies in programmes aimed at preventing complications and long terms. Programmes to promote health, to improve
in diabetic patients have increased the expected costs of diabetes detection and management of diabetes would reduce the burden as a
care by a 23% (Gagliardino et al., 2000). result of a slower progression to complications.
In Costa Rica, diabetes is among the three highest-cost condi- Directing public funds at treating diabetes and its complications
tions for the healthcare system (Jiménez, 2018). Developing policies is important. Nonetheless, the rapid escalation of expected numbers
that prevent chronic non-communicable diseases prevention is an of elderly with diabetes in the near future demands urgent action on
alternative to reducing costs in the healthcare system. Physiological health promotion and prevention. Not doing so would have the ad-
changes related to DM2 begin in childhood or adolescence verse effect of increasing economic costs due to premature morbidity
(Guerrero and Rodrı́guez, 2015) therefore creating policies to pro- and mortality from diabetes that would absorb much of the health-
mote health as early as during childhood is critical. care budgets.
A study to find out what prevention-focused actions can reduce Strategies to tackle obesity might be incorporated into other exist-
costs in diabetes care has been carried out in Brazil, to analyse the ing health promotion programmes. But strategies should be framed in
relationship between physical activity and the expenditures in public contexts that reduce obesogenic environments. Educational strategies
healthcare on DM2 treatment. It was found that physical activity may lead to a better diet in individuals, but sustainable changes occur
in diabetic populations was consistently associated with lower in the population when supporting environments for these behaviour-
healthcare expenditures for the public healthcare system (Codogno al shifts are also part of the equation.
et al., 2011).
A challenge for policymakers is to develop policy and pro-
grammes aimed at reaching SDGs, as they respond to context Acknowledgements
needs and target inequity reduction. According to the Pan American The CRELES project is a longitudinal study of the Universidad de Costa Rica,
Health Organization (2017) actions aimed at reaching SDGs and carried on by the Centro Centroamericano de Población in collaboration with
alleviating inequity must focus on policy that prioritizes actions to the Instituto de Investigaciones en Salud, with the support of the Wellcome
target structural determinants. Producing evidence of inequality Trust Foundation (Grant No. 072406). Principal Investigator: Luis Rosero-
related to diabetes will allow policy makers to identify policy Bixby. Co-principal investigators: Xinia Fernández and William H. Dow.
targets. Collaborating investigators: Ericka Méndez, Guido Pinto, Hannia Campos,
To face the diabetes burden, health promotion aspects must be Kenia Barrantes, Floribeth Fallas, Gilbert Brenes and Fernando Morales.
taken into consideration, specifically the action lines proposed in the Informatics and support staff: Daniel Antich, Aaron Ramı́rez, Jeisson
Hidalgo, Juanita Araya, and Yamileth Hernández. Field workers: José
Ottawa Charter for Health Promotion (1986), dealing with the im-
Solano, Julio Palma, Jenny Méndez, Maritza Aráuz, Mabelyn Gómez,
plementation of public policy and healthy legislation, the creation
Marcela Rodrı́guez, Geovanni Salas, Jorge Vindas and Roberto Patin~o. The
and protection of healthy environments, the strengthening of com- following sources of funding were awarded to the corresponding author:
munitarian action, individual and collective potentiality and the re- Traineeship from the National Institutes of Health’s (NIH) Fogarty
orientation of healthcare services. International Center (FIC) training programme (5D43TW001586) at the
Primary attention actions focused on the obesity and diabetes Center for Demography and Ecology. Scholarships from the Costa Rican
problem should be taken into account by decision makers. Araúz Ministry of Health and the Universidad de Costa Rica. The authors express
et al. (2001) propose to search for methodologies that deal with their gratitude to Hazel Quesada-Leitón for her support in data management.
knowledge, perceptions, attitudes, fears and practices of patients They also thank the manuscript reviewers for their thoughtful comments.
both in the family and the community contexts. Health education Conflict of interest statement. None declared.
may therefore prove to be a useful tool. Ethical approval: The CRELES study was approved by the Ethical
The universal access to primary healthcare can be crucial to Science Committee of the University of Costa Rica (reference: VI-763-CEC-
effectively reduce inequalities, by preventing the onset of the disease. 23 -04). All the databases of the study have been made anonymous
As stated by Aguilar et al. (2015), healthcare services should be (the name or identifier has been removed) to avoid risks to the privacy of the
re-oriented based on a Social Determinants of Health approach participants.
that considers equal access to preventive programmes, strengthening
of primary healthcare and the development of human resources with References
adequate skills.
Aguilar C, Hernández S, Garcı́a E, Barquera S, Reyes H. 2015. Chapter 15.
Los sistemas de Salud en la Prevención y Control de la Diabetes. In:
Conclusion Acciones para enfrentar a la diabetes. México: Academia Nacional de
Medicina. https://www.anmm.org.mx/publicaciones/CAnivANM150/L15-
This study adds to the literature an estimation of the burden that
Acciones-para-enfrentar-a-la-diabetes.pdf, accessed 20 October 2018.
DM2 has on a developing country that has already started the popu- Andersen RM. 1995. Revisiting the behavioral model and access to medical
lation aging process. Although the data are specific to Costa Rica, care: does it matter? Journal of Health and Social Behavior 36: 1–10.
results may be applicable to other aging developing countries in the Andrade FCD. 2006. Functional disability and health inequality among elderly
LAC region. Latinos: the diabetes mellitus burden. Doctoral dissertation, University of
Downloaded from https://academic.oup.com/heapol/article-abstract/34/Supplement_2/ii45/5625033 by UCR user on 13 November 2019
ii54 Health Policy and Planning, 2019, Vol. 34, Suppl. 2
Wisconsin-Madison. Available from ProQuest Dissertations and Thesis Hu FB, Stampfer MJ, Solomon CG et al. 2007. The impact of diabetes mellitus
database (UMI No. 3245628). on mortality from all causes and coronary heart disease in women. 20 years
Araúz A, Sánchez G, Padilla G, Fernández M, Roselló M, Guzmán S. of follow-up. Archives of Internal Medicine 161: 1717–23.
2001. Intervención educativa comunitaria sobre la diabetes en el ámbito Instituto Nacional de Salud Pública (INSP). 2012. Encuesta Nacional de Salud
de la atención primaria. Revista Panamericana de Salud Pública 9: y Nutrición 2012. México: ENSANUT. http://ensanut.insp.mx/doctos/semi
145–53. nario/M0302.pdf36, accessed 7 November 2018.
Arredondo A, De Icaza E. 2011. Costos de la Diabetes en América Latina: Jiménez J. 2018. Envejecimiento de la población y alto gasto en medicamentos
evidencias del Caso Mexicano. Value in Health 14: S85. retan la sostenibilidad de la Caja. Noticias Universidad de Costa Rica.
Ávila-Curiel A, Shamah-Levy T, Galindo-Gómez C, Cuevas-Nasu L, Moreno- https://www.ucr.ac.cr/noticias/2018/12/21/envejecimiento-de-la-poblacion-
Macı́as H, Chávez-Villasana A. 2007. La diabetes mellitus en estratos y-alto-gasto-en-medicamentos-retan-la-sostenibilidad-de-la-caja.html, accessed
socioeconómicos bajos de la ciudad de México. Un problema relevante. 3 November 2018.
Revista de Investigación Clı́nica 59: 246–55. Kowall B, Rathmann W, Strassburger K et al. 2010. Association of passive and
Baliunas DO, Taylor BJ, Irving H et al. 2009. Alcohol as a risk factor for type active smoking with incident type 2 diabetes mellitus in the elderly popula-
2 diabetes. A systematic review and meta-analysis. Diabetes Care 32: tion: the KORA S4/F4 cohort study. European Journal of Epidemiology 25:
2123–32. 393–402.
Barceló A. 2000. Diabetes and hypertension in the Americas. The West Indian Llanos AA, Morera-Salas M, Barber-Pérez P, Hernández K, Xirinach-Salazar
Medical Journal 49: 262–5. Y, Varga JR. 2009. Factores relacionados con la utilización de servicios de
Barceló A, Aedo C, Rajpathak S, Robles S. 2003. The cost of diabetes in Latin salud en Costa Rica. Revista de Salud Pública 11: 323–35.
America and the Caribbean. Bulletin of the World Health Organization 81: Manning W, Mullahy J. 2001. Estimating log models: to transform or not to
19–27. transform? Journal of Health Economics 20: 461–94.
Barceló A, Peláez M, Rodrı́guez-Wong L, Pastor-Valero M. 2006. The preva- Méndez-Chacón E, Santamarı́a-Ulloa C, Rosero-Bixby L. 2008. Factors asso-
lence of diagnosed diabetes among the elderly of seven cities in Latin ciated with hypertension prevalence, unawareness and treatment among
America and the Caribbean. Journal of Aging and Health 18: 224–39. Costa Rican elderly. BMC Public Health 8: 275.
Barceló A, Gregg E, Pastor-Valero M, Robles S. 2007. Waist circumference, Nakanishi N, Suzuki K, Tatara K. 2003. Alcohol consumption and risk for de-
BMI and the prevalence of self-reported diabetes among the elderly of the velopment of impaired fasting glucose or type 2 diabetes in middle-aged
United States and six cities of Latin America and the Caribbean. Diabetes Japanese men. Diabetes Care 26: 48–54.
Research and Clinical Practice 78: 418–27. Organización para la Cooperación y el Desarrollo Económicos (OCDE).
Beulens JWJ, Stolk RP, van der Schouw YT, Grobbee DE, Hendriks HFJ, Bots 2017. OECD Reviews of Health Systems: Costa Rica. Evaluación y
ML. 2005. Alcohol consumption and risk of type 2 diabetes among older Reocmedaciones. Paris: OECD Publishing.
women. Diabetes Care 28: 2933–8. Palloni A, Wong R, Peláez M. 2006. The tide to come. Elderly health in Latin
Brenes-Camacho G, Rosero-Bixby L. 2008a. Diabetes mellitus en adultos America and the Caribbean. Journal of Aging and Health 18: 180–206.
mayores costarricenses. Población y Salud En Mesoamérica 5: 1–7. Pan American Health Organization. 2017. Hacia una Salud Sostenible. https://
Brenes-Camacho G, Rosero-Bixby L. 2008b. Metabolic control in a nationally www.paho.org/salud-en-las-americas-2017/?post_type¼post_t_es&p¼308&
representative diabetic elderly sample in Costa Rica: patients at community lang¼es, accessed 28 October 2018.
health centers vs. patients at other health care settings. BMC International Pietraszek A, Gregersen S, Hermansen K. 2010. Alcohol and type 2 diabetes: a
Health and Human Rights 8: 5. review. Nutrition, Metabolism, and Cardiovascular Diseases 20: 366–75.
Brenes-Camacho G, Rosero-Bixby L. 2009. Differential by socioeconomic sta- Rehkopf D, Dow H, Rosero-Bixby L. 2010. Differences in the association of
tus and institutional characteristics in preventive service utilization by older cardiovascular risk factors with education: a comparison of Costa Rica
persons in Costa Rica. Journal of Aging and Health 21: 730–58. (CRELES) and the USA (NHANES). Journal of Epidemiology &
Brown JE. 2008. Nutrición en las diferentes etapas de la vida. 3rd edn. Community Health 64: 821–8.
México: McGraw Hill. Rosero-Bixby L. 1996. Adult mortality decline in Costa Rica. In: Timaeus IM,
Chaparro-Dı́az L. 2016. Objetivos de Desarrollo Sostenible y Enfermedad Chackiel J, Rusicka L (eds). Adult Mortality in Latin America. Chapter 9.
Crónica no Transmisible. Revista Latino-Americana de Enfermagem 24: 1–2. Oxford: Clarendon Press, 33–42.
Charter for Health Promotion. 1986. World Health Organization. https:// Rosero-Bixby L, Dow WH. 2009. Surprising SES gradients in mortality,
www.paho.org/hq/dmdocuments/2013/Carta-de-ottawa-para-la-apromocion- health, and biomarkers in a Latin American population of adults. The
de-la-salud-1986-SP.pdf. Journals of Gerontology Series B: Psychological Sciences and Social
Codogno J, Fernandes R, Sarti F, Freitas I, Monteiro H. 2011. The burden of Sciences 64B: 105–17.
physical activity on type 2 diabetes public healthcare expenditures Rosero-Bixby L, Antich D. 2010. Estudio longitudinal de mortalidad de
among adults: a retrospective study. BMC Public Health 11: 1–7. adultos costarricenses 1984-2007. Población y Salud En Mesoamérica 7:
Diehr P, Yanez D, Ash A, Hornbrook M, Lin DY. 1999. Methods for analyzing 1–24.
health care utilization and costs. Annual Review of Public Health 20: 125–44. Rosero-Bixby L, Dow WH, Fernández X. 2013. CRELES: Costa Rican
Gagliardino J, Olivera E, Etchegoyen G, Gonzalez C, Guidi M. 2000. Longevity and Healthy Aging Study. Methods, Wave 3. Berkeley, CA:
Evaluación y costos del proceso de atención a pacietnes diabéticos. Department of Demography, University of California, Berkeley. Electronic
Medicina (Buenos Aires) 60: 880–8. Document http://www.creles.berkeley.edu, accessed 5 November 2018.
Grundy SM, Benjamin IJ, Burke GL, Chait A, Eckel RH, Howard BV. 1999. Sandı́n M, Espelt E, Escolar-Pujolar A, Arriola L, Larran~a I. 2011.
Diabetes and cardiovascular disease: a statement for healthcare professio- Desigualdades de género y diabetes mellitus tipo 2: la importancia de la
nals from the. Circulation 100: 1134–46. diferencia. Avances En Diabetologı́a 27: 78.
Guerrero F, Rodrı́guez M. 2015. Chapter 10. Prevención de la Diabetes tipo 2: Sáenz M, Acosta M, Muiser J, Bermúdez J. 2011. Sistema de Salud en Costa
estrategias aplicables en México. In: Acciones para enfrentar a la diabetes. Rica. Revista de Salud Pública de México 53: 156–67.
México: Academia Nacional de Medicina. https://www.anmm.org.mx/pub Shaw J, Sicree R, Zimmet P. 2010. Global estimates of the prevalence of dia-
licaciones/CAnivANM150/L15-Acciones-para-enfrentar-a-la-diabetes.pdf, betes for 2010 and 2030. Diabetes Research and Clinical Practice 87: 4–14.
accessed 3 November 2018. Sloan FA, Bethel MA, Ruiz D, Shea AH, Feinglos MN. 2008. The growing
Gulliford M, Figueroa-Munoz J, Morgan M et al. 2002. What does ‘access to burden of diabetes mellitus in the US elderly population. Archives of
health care’ mean? Journal of Health Services Research & Policy 7: 186–8. Internal Medicine 168: 192–9.
Hosmer DW, Lemeshow S. 1999. Applied Survival Analysis. New York: John Solli O, Jenssen T, Kristiansen IS. 2010. Diabetes: cost of illness in Norway.
Wiley & Sons. BMC Endocrine Disorders 10: 15.
Hu FB, Manson JE, Stampfer MJ et al. 2001. Diet, lifestyle and the risk of type StataCorp. 2013. Stata Statistical Software: Release 13.0. College Station, TX:
2 diabetes in women. The New England Journal of Medicine 345: 790–7. Stata Corporation.
Downloaded from https://academic.oup.com/heapol/article-abstract/34/Supplement_2/ii45/5625033 by UCR user on 13 November 2019
Health Policy and Planning, 2019, Vol. 34, Suppl. 2 ii55
Suh DC, Kim CM, Choi IS, Plauschinat CA. 2008. Comorbid conditions and gly- World Health Organization. 2000. Obesity: preventing and managing the glo-
cemic control in elderly patients with type2 diabetes mellitus, 1988 to 1994 bal epidemic: Report of a WHO consultation. Geneva, Switzerland. 2000.
to 1999 to 2004. Journal of the American Geriatrics Society 56: 484–92. Technical Report Series no. 894.
Velázquez-Monroy Ó, Rosas-Peralta M, Lara-Esqueda A, Pastelı́n-Hernández Willi C, Bodenmann P, Ghali WA, Faris PD, Cornuz J. 2007. Active smoking and
G;Grupo ENSA 2000, Sánchez-Castillo C. 2003. Prevalencia e interrelación the risk of type 2 diabetes. A systematic review and meta-analysis. JAMA 298:
de enfermedades crónicas no transmisibles y factores de riesgo cardiovascu- 2654–64.
lar en México: resultados finales de la Encuesta Nacional de Salud (ENSA) Yach D, Stuckler D, Brownell KD. 2006. Epidemiologic and economic conse-
2000. Archivos de Cardiologı́a de México 73: 62–77. quences of the global epidemics of obesity and diabetes. Nature Medicine
Villalobos L, Piedra M. 1998. Reforma en la salud en Costa Rica: implica- 12: 62–6.
ciones en la gestión municipal. Revistas de Ciencias Administrativas Zhang L, Curhan GC, Hu FB, Rimm EB, Forman JP. 2011. Association
y Financieras de la Seguridad Social 6: 43–51. between passive and active smoking and incident type 2 diabetes in women.
Whiting D, Unwin N, Roglic G. 2010. Diabetes: equity and social Diabetes Care 34: 892–7.
determinants. In: Blas E, Sivasankara Kurup A (eds). Equity, Social Zhang P, Zhang X, Brown J, Vistisen D, Sicree R, Shaw J, Nuchols G. 2010.
Determinants and Public Health Programmes. Switzerland: World Global healthcare expenditure on diabetes for 2010 and 2030. Diabetes
Health Organization, 77–94. Research and Clinical Practice 87: 293–301.
Downloaded from https://academic.oup.com/heapol/article-abstract/34/Supplement_2/ii45/5625033 by UCR user on 13 November 2019