Journal of Diabetes and Its Complications xxx (2015) xxxxxx

Contents lists available at ScienceDirect

Journal of Diabetes and Its Complications

journal homepage: WWW.JDCJOURNAL.COM

Adherence to diabetes guidelines for screening, physical activity and

medication and onset of complications and death
Yiqun Chen 1, 2, Frank A. Sloan , Arseniy P. Yashkin 3
Department of Economics, Duke University, 213 Social Sciences Building, Box 90097, Durham, NC, 27708

a r t i c l e i n f o a b s t r a c t

Article history: Aims: Analyze relationships between adherence to guidelines for diabetes care regular screening; physical
Received 29 May 2015 activity; and medication and diabetes complications and mortality.
Received in revised form 30 June 2015 Methods: Outcomes were onset of congestive heart failure (CHF), stroke, renal failure, moderate complications
Accepted 1 July 2015 of lower extremities, lower-limb amputation, proliferative diabetic retinopathy (PDR), and mortality during
Available online xxxx
follow-up. Participants were persons aged 65 + in the Health and Retirement Study (HRS) 2003 Diabetes
Study and had Medicare claims in follow-up period (20048).
Keywords:
Guidelines
Results: Adherence to screening recommendations decreased risks of developing CHF (odds ratio (OR) = 0.83;
Physical activity 95% condence interval (CI): 0.720.96), stroke (OR = 0.80; 95% CI: 0.680.94); renal failure (OR = 0. 82; 95%
Diabetes complications CI: 0.710.95); and death (OR = 0.86; 95% CI: 0.740.99). Adherence to physical activity recommendation
Mortality reduced risks of stroke (OR = 0.64; 95% CI: 0.450.90), renal failure (OR = 0.71; 95% CI: 0.520.97), moderate
Health and Retirement Study lower-extremity complications (OR = 0.71; 95% CI: 0.510.99), having a lower limb amputation (OR = 0.31,
95% CI: 0.110.85), and death (OR = 0.56, 95% CI: 0.410.77). Medication adherence was associated with lower
risks of PDR (OR = 0.35, 95% CI: 0.130.93).
Conclusions: Adherence to screening, physical activity and medication guidelines was associated with lower risks
of diabetes complications and death. Relative importance of adherence differed among outcome measures.
2015 Elsevier Inc. All rights reserved.

1. Introduction diabetes by 2050 if recent increases in diabetes incidence continue

(Boyle, Thompson, Gregg, Barker, & Williamson, 2010), suggesting an
Diabetes mellitus imposes a substantial societal burden. The even larger burden of diabetes on society in subsequent years.
American Diabetes Association (ADA) estimated that diabetes led to In an effort to improve diabetes outcomes, the ADA has provided a
a total cost of $245 billion in 2012, including $176 billion in direct comprehensive set of guidelines for diabetes care. Although benets
medical costs and $69 billion in reduced productivity (American of adherence to guidelines have been documented, most research to
Diabetes Association, 2013a). As the leading cause of blindness and date has focused on the association between adherence and glycemic
lower-limb amputations among U.S. adults (Centers for Disease control (Krapek, King, Warren, et al., 2004; Raum, Krmer, Rter, et al.,
Control and Prevention, 2011), diabetes also has increased the cost 2012), followed in order of frequency by health care utilization and
to such public programs as Social Security Disability Insurance and costs (Balkrishnan et al., 2003; Wong, Bryson, Hebert, & Liu, 2014),
Medicaid. One study projected that 1 in 3 U.S. adults could have and quality of life (Martnez, Prado-Aguilar, Rascn-Pacheco, &
Valdivia-Martnez, 2008; Saleh, Mumu, Ara, Hafez, & Ali, 2014).
These studies have generally concluded that there are links between
improved adherence and better glycemic control, lower health care
utilization and costs, and better quality of life.
Conict of Interest Statement: This research was supported in part by the National
Much less attention has been devoted to impacts of adherence to
Institute on Aging (grant R01-AG017473). The sponsors had no such involvement in
design and conduct of the study, collection, management, analysis, interpretation of the
guidelines on onset of diabetes complications and mortality. Persons
data, preparation, review, approval of the manuscript, writing the report/article, nor the with diabetes are more likely to experience neuropathic, cardiovas-
decision to submit the manuscript for publication. Conicts of interest: none. cular, nephropathic, and ophthalmic complications that seriously
Corresponding author. Tel.: +1 919 613 9358; fax: +1 919 681 7984. erode quality of life and lead to lost productivity and premature
E-mail addresses: yiqun.chen@duke.edu (Y. Chen), fsloan@duke.edu (F.A. Sloan),
mortality (Bethel, Sloan, Belsky, & Feinglos, 2007; Boyle et al., 2010).
arseniy.yashkin@duke.edu (A.P. Yashkin).
1
Tel.: +1 919 660 1825; fax: +1 919 681 7984. About half of the diabetes treatment costs are attributable to
2
Alternative Proof Reader. complications (Zhuo, Zhang, & Hoerger, 2013). The estimated total
3
Tel.: +1 919 660 1800; fax: +1 919 681 7984. cost of lost productivity due to diabetes-attributable premature death

http://dx.doi.org/10.1016/j.jdiacomp.2015.07.005
1056-8727/$ 2015 Elsevier Inc. All rights reserved.

Please cite this article as: Chen, Y., et al., Adherence to diabetes guidelines for screening, physical activity and medication and onset of
complications and death, Journal of Diabetes and Its Complications (2015), http://dx.doi.org/10.1016/j.jdiacomp.2015.07.005
2 Y. Chen et al. / Journal of Diabetes and Its Complications xxx (2015) xxxxxx

is $18.5 billion per year in 2012 (American Diabetes Association, We rst identied patients screening, physical activity and
2013a). Given this substantial nancial burden, it is important to medication information from HRS-DM and HRS-DRUG, and then
investigate potential factors that could prevent diabetes complica- merged the interview responses to Medicare claims les to obtain
tions and early mortality. dates of diagnosis of diabetes complications and death during the
Another limitation of earlier studies is the limited scope of 5-year follow-up period of 20048. The 2002 wave of HRS provided
adherence measuresmostly only medication adherence (An & data on respondents demographic characteristics and income. Since
Nichol, 2013; Gibson, Song, Alemayehu, et al., 2010; Krapek et al., the goal of the analysis was to measure the relationship between
2004; Raum et al., 2012). Effects of adherence to guidelines for other adherence to guidelines and onset of complications and death during
aspects of diabetes care on diabetes outcomes, and the comparative follow-up, we excluded individuals diagnosed with the complication
importance of different adherence, e.g., physical activity adherence corresponding to the dependent variable prior to follow-up. Thus,
and dietary adherence, have not been extensively analyzed. numbers of observations varied by complication.
To overcome these limitations, this study examined 3 measures of
adherence in the same study screening, physical activity, and
medication adherence and their associations with mortality and 2.2. Dependent variables
onset of 5 frequently-occurring complications of diabetes: congestive
heart failure (CHF), renal failure, stroke, moderate severity low- Our primary outcomes were physician-diagnosed onset of: (1)
er-extremity complications, lower-limb amputation, and proliferative congestive heart failure (CHF); (2) stroke; (3) renal failurea general
diabetic retinopathy (PDR). We limited our analysis sample to category encompassing end-stage renal disease, acute renal failure,
individuals aged 65 +, an age group that incurs approximately 60% unspecied renal failure, kidney transplantation, or dialysis; (4)
of total medical expenditures attributable to diabetes (American moderate severity complications of the lower extremitiescellulitis,
Diabetes Association, 2013a). Limiting the study sample to elderly Charcot foot, osteomyelitis, gangrene, diabetic amyotrophy, diabetic
persons allowed us to obtain physician diagnoses of diabetes neuropathy; (5) serious complications of the lower extremities,
complications from the Medicare claims les, which should be more measured by having a lower-limb amputation; and (6) proliferative
reliable than patient self-reports. diabetic retinopathy; and (7) death during the 5-year follow up
period. The 2 lower-extremity categories were not dened to be
mutually exclusive, but few persons fell into both categories. The
2. Material and methods complication information was identied from the Medicare Part A and
Part B claims data using codes from the International Classication of
2.1. Data and subjects Disease, 9th Revision, Clinical Modication (ICD-9-CM) and Current
Procedural Terminology, version 4 (CPT-4) (Table 1). Mortality
Data came from the Health and Retirement Study (HRS), the HRS information was obtained from the Medicare denominator le.
2003 Diabetes Study (HRS-DM) conducted as a supplement to HRS, 2
other HRS supplemental surveysthe 2005 and 2007 Prescription
Drug Study (HRS-DRUG), and Medicare Part A (facility), Part B
(professional) and denominator les that contained respondents Table 1
dates of death and participation in Medicare Advantage (MA) plans Diagnosis codes.
when applicable. Medicare beneciaries in MA during follow-up were Conditions Administrative Codes
excluded from this study since claims for MA enrollees are not
Congestive heart failurea ICD-9 428.xx 398.91 402.01 402.11
provided on a public use basis. Medicare claims data were merged 402.91 404.11 404.91
with interview data from HRS. Strokea ICD-9 431.xx 436.xx 997.02
The HRS is a biannual longitudinal survey of persons aged 5161 Renal failure
at study initiation in 1992 and their spouses who could be of any Chronic Renal Failureb ICD-9 585.xx
Acute Renal Failure ICD-9 584.xx
age. The HRS is a nationally representative survey of community-
Unspecied Renal Failure ICD-9 586.xx
dwelling residents in the U.S. of all race/ethnicities. Individuals in Transplant ICD-9 V420
the HRS are followed until death; new replenishment cohorts are ICD-9(P) 55.69
added periodically. In 2003, the HRS-DM surveyed HRS respondents CPT 50360 50365
Dialysis ICD-9 V45.11 V56.xx
who reported a diagnosis of diabetes at the 2002 wave of HRS.
ICD-9(P) 39.95 54.98
Among eligible participants, 1901 returned questionnaires (response CPT 90921 90925 90960 90961
rate, 79.7%). This survey collected detailed data on patients treatment 90962 90966 90970 90935
and self-management of diabetes not reported in regular HRS 90937 90945 90947
interviews which covered a wide range of topics. In 2005 and 2007, Moderate severity lower extremity complications
Diabetic neuropathy ICD-9 250.6x 357.2 355.xx
another supplemental survey, the HRS-DRUG, was conducted to
Diabetic amyotrophy ICD-9 358.1
collect Medicare eligible HRS participants use of medications. Cellulitis ICD-9 681.1x 682.6 682.7
The HRS-DRUG allowed us to supplement respondents drug use Charcot foot ICD-9 707.11
information provided by the HRS-DM. The Medicare claims les Osteomyelitis ICD-9 730.06 730.07 730.16 730.17
730.26 730.27
contained information on Medicare beneciaries medical diagnoses
Gangrene ICD-9 250.7x 785.4
(using the International Classication of Diseases, 9th Revision Amputation of lower extremity ICD-9(P) 84.1x
Clinical Modication (ICD-9-CM)), use of laboratory tests (using Proliferative diabetic retinopathy ICD-9 362.02
Current Procedure Terminology (CPT) codes), and physician visits Low cognition
(as dened by U.S. Centers for Medicare and Medicaid medical Alzheimer's disease ICD-9 331.0x 331.1x 331.2x 331.9x
Senility ICD-9 797.xx
specialty codes).
Dementia ICD-9 290.xx 294.xx
Limiting our analysis sample to persons aged 65 + in 2003 yielded
ICD-9: International Classication of Disease, 9th Revision, Clinical Modication;
a sample of 1320 of the total 1901 respondents to the HRS-DM.
ICD-9(P): ICD-9 for procedure; CPT: Current Procedural Terminology.
Individuals not identiable in Medicare claims les (n = 140) or with a
Includes Part A claims only.
missing information on any explanatory variable (n = 38) were b
Must be accompanied by dialysis or transplant code, except End Stage Renal
excluded. The nal analysis sample consisted of 1142 individuals. Disease (ICD-9: 585.6) introduced in 2005.

Please cite this article as: Chen, Y., et al., Adherence to diabetes guidelines for screening, physical activity and medication and onset of
complications and death, Journal of Diabetes and Its Complications (2015), http://dx.doi.org/10.1016/j.jdiacomp.2015.07.005
 Y. Chen et al. / Journal of Diabetes and Its Complications xxx (2015) xxxxxx 3

2.3. Explanatory variables Table 2

Summary statistics (N = 1142).

Our choice of screening adherence measures was based on ADA Variable Mean Std.
Guidelines (American Diabetes Association, 2013b). Binary variables Dev.
specied whether an individual obtained the recommended tests in HbA1c test 1 if had 2 HbA1c tests within the last year, 0.789 0.406
the last year, i.e., lipid test, eye exam, urine analysis and HbA1c test. To 0 otherwise
generate a single comprehensive measure of screening adherence, we Lipid test 1 if had at least 1 cholesterol test within 0.595 0.487
the last year, 0 otherwise
performed a factor analysis on the 4 binary variables and used the rst
Urinalysis 1 if had at least 1 urinalysis within the 0.757 0.427
extracted factor (eigenvalue 1.50) that accounted for most of the last year, 0 otherwise
variation in the 4 variables as a covariate (Table A.1). Higher factor Eye exam 1 if had at least 1 eye exam within the last 0.700 0.459
scores indicated it was more likely that the respondent received the year, 0 otherwise
tests at the recommended frequency. Screening adherence Extracted screening adherence score 0.034 0.963
from factor analysis
For physical activity adherence, following earlier papers (Ains-
Physical activity 1 if exercising regularly, 0 otherwise 0.532 0.499
worth, Haskell, Whitt, et al., 2000; Plotnikoff, Taylor, Wilson, et al., Medication adherence 1 if never/rarely miss oral diabetes 0.693 0.462
2006), we considered individuals to be adherent to the physical medication, 0 otherwise
activity guideline if the person engaged in 600 + metabolic equivalent Age 6574 1 if aged 6574 0.545 0.498
Age 7584 1 if aged 7584 0.368 0.482
tasks/week of moderate or vigorous physical activity (total minutes/
Age 85+ 1 if aged 85+ 0.088 0.283
week of moderate activity*4+total minutes/week of vigorous Female 1 if female, 0 if male 0.504 0.500
activity*7.5)600). Medication adherence was a binary variable, Black 1 if black, 0 otherwise 0.164 0.370
dened as whether the respondents never or rarely (fewer than 1 out Married 1 if married, 0 otherwise 0.604 0.489
of 10 scheduled doses) missed a prescribed dose of oral diabetes Education Years of schooling completed 11.320 3.571
Household income Unit: $10,000 (2002 dollars) 3.638 5.606
medication. We also controlled for whether the respondent was
TIBI Total Illness Burden Index score 35.138 18.359
taking oral diabetes medications to account for the fact that some (scale: 0100)
patients were not prescribed such medications. Insulin-use was not BMI Body Mass Index 29.061 5.568
accounted in the medication adherence as insulin-users are likely to Low cognition 1 if had Alzheimer's disease, senility, or 0.055 0.228
dementia
have more severe diabetes and thus would be more likely to have
Insulin 1 if use insulin, 0 otherwise 0.238 0.426
diabetes complications and premature death in the follow-up period. Oral diabetes 1 if use oral diabetes medications, 0 0.748 0.434
This could downward-bias our estimations on benecial effects or medications otherwise
even make us nd counter-intuitive adverse effects of medication Baseline conditions (2003)
adherence on diabetes outcomes. This concern is supported by our Congestive heart 1 if had CHF in baseline, 0 otherwise 0.194 0.395
failure(CHF)
data: we found that insulin-using respondents were more likely to
Stroke 1 if had stroke in baseline, 0 otherwise 0.088 0.283
have complications and die in the follow-up period. However, no Renal failure 1 if had ESRD/acute/unspecied renal 0.053 0.223
association between oral diabetes medication use and diabetes failure/kidney transplantation/dialysis in
complications or premature death was found, thus estimates of baseline, 0 otherwise
Lower extremity 1 if had moderate severity lower 0.406 0.491
effects of medication adherence based on oral medication use would
complications extremity complications in baseline,
be less subject to bias. To account for the fact that a few 0 otherwise
oral-medication-taking patients also used insulin, we added a binary Amputation 1 if had amputation in baseline, 0 0.008 0.088
variable for whether the patient was using insulin in the regressions. otherwise
Information on medication adherence came from 2 sources. The Proliferative diabetic 1 if had PDR in baseline, 0 otherwise 0.025 0.157
retinopathy(PDR)
primary source was the HRS-DM. When such information was
Follow up period (20042008)
unavailable in HRS-DM, we imputed the value from the HRS-DRUG. CHF 1 if diagnosed with CHF in 20042008, 0.417 0.493
Participants in the 2005 HRS-DRUG reporting a 2 + year length of use 0 otherwise
and participants in the 2007 HRS-DRUG reporting a 4+ year length of Stroke 1 if diagnosed with stroke in 20042008, 0.215 0.411
0 otherwise
use were considered to have taken diabetes drugs at baseline (2003).
Renal failure 1 if diagnosed with ESRD/acute/ 0.252 0.434
Other explanatory variables were age, gender, race, marital status, unspecied renal failure/kidney
years of schooling completed, household income (in 2002 dollars), transplantation/dialysis in 20042008,
whether obese based on the persons body mass index (BMI), total illness 0 otherwise
burden index (TIBI) score, and low cognition dened as having a Lower extremity 1 if diagnosed with moderate severity 0.615 0.487
complications lower extremity complications in 2004
diagnosis of Alzheimer's disease, senility, or dementia. The TIBI is a
2008, 0 otherwise
widely used comorbidity measure specically developed for persons Amputation 1 if had amputation in 20042008, 0 0.019 0.138
with diabetes (Greeneld, Billimek, & Kaplan, 2010; Greeneld et al., otherwise
1995). The TIBI assesses comorbid illness using self-reported presence Proliferative diabetic 1 if diagnosed with PDR in 20042008, 0 0.053 0.223
and severity of symptoms and conditions in each of 16 body system retinopathy(PDR) otherwise
Death 1 if died in 20042008, 0 otherwise 0.279 0.449
domains on a scale from 0 to 100. A higher TIBI score indicates worse
health. We included the persons TIBI score as a covariate to address the Note: ESRD = End Stage Renal Disease.

potential concern that persons in worse health and thus at higher risk of
future illness and death are more (less) likely to follow care
recommendations, which could bias our estimates upward (downward). Over half (53.2%) exercised regularly according to our study criterion;
We used logistic regression (STATA 11 SE). Duke Universitys most (69.3%) never or rarely missed oral diabetes drugs. These rates
Institutional Review Board approved our study protocol and use of the are slightly higher than those reported in earlier studies that focused
restricted data prospectively. on younger samples (Bayer et al., 2014; Feil, Zhu, & Sultzer, 2012).
Of the total study sample, 50.4% were female, 16.4% were black,
3. Results and 60.4% were married. Most were aged 6584. Mean educational
attainment, household income, BMI and TIBI score were 11.3 years,
Most respondents had HbA1c tests (78.9%), lipid tests (59.5%), $36,380, 29.1, and 35.1, respectively. Most sample persons (74.8%)
urinalysis (75.7%) and eye exams (70.0%) in the last year (Table 2). used oral diabetes medications; many fewer used insulin (23.8%).

Please cite this article as: Chen, Y., et al., Adherence to diabetes guidelines for screening, physical activity and medication and onset of
complications and death, Journal of Diabetes and Its Complications (2015), http://dx.doi.org/10.1016/j.jdiacomp.2015.07.005
4 Y. Chen et al. / Journal of Diabetes and Its Complications xxx (2015) xxxxxx

Relative to the baseline, substantial proportions of respondents Table A.1

were diagnosed with complications during follow-up: CHF (22.3%); Distribution of scores for factor 1 from factor analysis and percentages of persons with
tests.
moderate lower extremity complications (20.9%); renal failure
(19.9%); and stroke (12.7%). Fewer persons were diagnosed with Percentiles Score HbA1c (%) Lipid test (%) Urinalysis (%) Eye exam (%)
PDR (2.8%) or had lower-limb amputations (1.1%). 10 1.423 100 0 0 0
Adherence to ADA screening recommendations for screening tests 25 0.623 100 100 0 0
at baseline year decreased the probability of being diagnosed with 50 0.262 100 0 100 100
75 1.062 100 100 100 100
CHF (odds ratio (OR) = 0.83; 95% condence interval (CI): 0.72
90 1.062 100 100 100 100
0.96), stroke (OR = 0.80; 95% CI: 0.680.94); renal failure (OR = 0.
82; 95% CI: 0.710.95); and death (OR = 0.86; 95% CI: 0.740.99) Note: numbers in column 3 to 6 mean percentages of patients had HbA1c tests, lipid
tests, eye exams and urinalyses in the last year, respectively, for all patients with each of
(Table 3). As the covariate for screening adherence was measured by a the factor score reported in column 2.
factor score from the factor analysis, to interpret the ndings, we
compared the 10th percentile of the factor score for adherence
screening to the score at the 90th percentile of scores. Comparing In our analysis sample, individuals who died during follow-up
screening adherence at the 10th percentile of the sample, at which were more likely to have been diagnosed with the study complica-
point persons only received HbA1c tests (Table A.1), to the 90th tions beforehand. In the subsample of persons who died during
percentile of the sample, where individuals received all 4 recom- follow-up, 61.8% were diagnosed with CHF, and 32.6% and 43.9% were
mended tests-HbA1c tests, eye exams, urinalysis, and lipid tests, the diagnosed with stroke and renal failure prior to death, respectively.
probability of developing CHF during follow-up fell by 0.42. The The corresponding percentages of diagnoses among persons who did
corresponding reductions for stroke were 0.50, for renal failure, 0.68, not die during follow-up were about half as large: 33.9%, 17.1%, and
and death, 0.61. 18.0%, respectively. When we added death during follow-up as a
Relationships between adherence to the physical activity recom- covariate to our analysis of complications diagnosis, the parameter
mendations and outcomes during follow-up were similarly strong, estimate on this covariate generally indicated that diagnosis was more
although there were differences in outcomes for which adherence likely. Results on adherence were not materially affected (results not
was statistically signicant and in the magnitudes of the reductions in shown). Thus, if anything, the results presented above underestimate
the odds of adverse outcomes. Adherence to such recommendations the importance of adherence to guidelines.
reduced the probability of being diagnosed with stroke (OR = 0.64;
95% CI: 0.450.90), renal failure (OR = 0.71; 95% CI: 0.520.97), 4. Discussion
moderate lower-extremity complications (OR = 0.71; 95% CI: 0.51
0.99), having a lower limb amputation (OR = 0.31, 95% CI: 0.11 Our results show that adherence to ADA recommendations often
0.85), and death (OR = 0.56, 95% CI: 0.410.77) during follow-up. was followed by reduced risks of diabetes complications and death,
Fewer of the results for medication adherence were statistically but there were differences in the relationship between adherence and
signicant. Such adherence was negatively associated with the specic outcomes according to the type of adherence. Better
probability of developing PDR (OR = 0.35, 95% CI: 0.130.93). adherence to recommended screenings was associated with lower
Overall, older age, a higher TIBI score, lower cognition, and being risks of being diagnosed with CHF, renal failure, and stroke and death
obese were associated with increased odds of complications and during follow up among persons diagnosed with diabetes at baseline.
mortality (not shown). The sample sizes varied among regressions Physical activity adherence was related to reduced risks of stroke,
since we excluded persons who had the complication at baseline. The renal failure, moderate and serious severity lower-extremity compli-
most common complication at baseline was for moderate low- cations, and death. For medication adherence, statistically signicant
er-extremity complications. association between medication adherence and lower rates of PDR
Persons who died during follow-up were included in the analysis of was also observed. While most attention in studying adherence has
complications onset. A potential concern is that death is a competing been devoted to drugs designed to treat diabetes, the evidence
risk to complications onset. If death occurs during follow-up, persons presented in this study demonstrates that adherence to screening and
may die before complications onset, thus biasing the estimated physical activity guidelines is at least as important in delaying a range
benecial effects of adherence downward if decedents were relatively of diabetes complications and death.
less likely to be adherent to guidelines and conversely if the odds of Our estimates are comparable to a few existing studies evaluating
complications onset is increased immediately before death. the link between diabetes medication adherence and complication

Table 3
Logistic regression results.

Variables CHF Stroke Renal Lower extremity Amputation PDR Death

Screening adherence 0.827 0.801 0.822 0.926 1.178 1.150 0.855

[0.715, 0.957] [0.680, 0.942] [0.709, 0.952] [0.785, 1.094] [0.748, 1.853] [0.829, 1.594] [0.741, 0.987]
Physical activity 0.783 0.637 0.706 0.711 0.311 0.781 0.564
[0.576, 1.065] [0.451, 0.898] [0.517, 0.965] [0.509, 0.994] [0.114, 0.850] [0.391, 1.561] [0.414, 0.769]
Medication 0.778 1.265 0.869 0.895 0.382 0.348 0.809
Adherence [0.411, 1.473] [0.602, 2.658] [0.460, 1.643] [0.476, 1.682] [0.074, 1.959] [0.130, 0.934] [0.441, 1.486]
Observations 921 1,042 1,082 678 1,133 1,113 1,142

CHF = congestive heart failure; Renal = renal diseases that encompass end stage renal disease, acute renal failure, unspecied renal failure, dialysis and kidney transplantation;
Lower extremity = moderate severity lower extremity complications (cellulitis, Charcot foot, osteomyelitis, gangrene, diabetic amyotrophy, diabetic neuropathy); amputation =
lower-limb amputation; PDR = proliferative diabetic retinopathy.
Other covariates included but not shown are aged 7584, aged 85+ (omitted group: aged 6574), gender, race, marital status, years of schooling completed, household income, body
mass index, total illness burden index score, low cognition (had Alzheimer's disease, senility, or dementia), whether use insulin, and whether use oral diabetes medications.
95% condence intervals in brackets.
Boldfaced indicates statistical signicance.
* pb0.05.
pb0.01.

Please cite this article as: Chen, Y., et al., Adherence to diabetes guidelines for screening, physical activity and medication and onset of
complications and death, Journal of Diabetes and Its Complications (2015), http://dx.doi.org/10.1016/j.jdiacomp.2015.07.005
 Y. Chen et al. / Journal of Diabetes and Its Complications xxx (2015) xxxxxx 5

outcomes and death. Using data on elderly individuals, Kuo et al. data or adverse side effects of these drugs experienced by patients.
(2003) concluded that persons with diabetes with poor consistency in Third, we did not include measures of adherence to other lifestyle
use of diabetes medications (inconsistent use or discontinuation of activities such as diet, smoking and drinking. Diet information,
drugs in follow-up surveys) were 43% more likely to die over a 7-year although provided by the HRS-DM, are not sufciently detailed to
period compared with patients who used medications consistently, an permit construction of a valid measure. Smoking and excess alcohol
effect size somewhat larger than the one we reported. No signicant consumption, 2 other behaviors associated with recommended
association between diabetes medication adherence and risk of diabetes self-care, were too low in the sample to permit meaningful
circulation problems was observed, a result consistent with our inference. Fourth, since we excluded persons with diabetes who had
study. Another study documented that adherence to diabetic the study complications at baseline except for the mortality analysis,
medications is associated with lower rates of retinopathy (Gibson et the remaining persons in our analysis could have been healthier than
al., 2010), which is consistent with our results. Based on a sample of elderly persons with a diabetes mellitus diagnosis as a whole.
employees diagnosed with diabetes, Bayer et al. (2014) concluded However, we controlled for TIBI, a widely used health measure for
that patients who received regular screenings for lipids, microalbu- diabetes patients, which could mitigate such concern to some extent.
minuria tests and HbA1c measurements experienced decreased risks Fifth, we do not account for patients use of routine diabetes care, a
of complications compared to those who received less complete potential determinant of diabetes outcomes, but our measure of
testing, with a hazard rate (HR) of 0.39 for congestive heart failure and screening test adherence should reect use of routine diabetes care
0.49 for renal disease, both results indicating higher effect sizes than since patients typically obtain tests at physician visits or immediately
the ones we obtained. prior to such visits. Also, controlling for patients socio-demographic
The effects of physical activity in reducing diabetes-complication characteristics, which are proxies to individuals access to and use
risk factors such as dyslipidemia, dysglycemia, hypertension and of medical care, in the estimation could help mitigate concerns on
excess weight in persons diagnosed with diabetes have been well this limitation.
documented in earlier literature (Agosti, Graziano, Artiaco, & These limitations notwithstanding, our results underscore the
Sorrentino, 2009; Herbst, Kordonouri, Schwab, Schmidt, & Holl, importance of improving adherence to screening tests which in turn
2007; Kirk, Mutrie, MacIntyre, & Fisher, 2003). For example, Boul, permits timely administration of effective therapeutic interventions,
Haddad, Kenny, Wells, and Sigal (2001) found that an 8-week as well as self-care measured in terms of physical activity and
duration of exercise training on patients with type 2 diabetes reduced medication use. At least half of cost of diabetes treatment is
HbA1c levels by 0.66% on average, an amount equivalent to a attributable to its complications (American Diabetes Association,
reduction of the risk of microvascular complication of 24.4% (Stratton 2013a; Brown, Pedula, & Bakst, 1999). The lost productivity due to
et al., 2000) and cerebrovascular diseases by 11.9% (Selvin et al., diabetes-attributed premature mortality had led to a total cost of
2004). Blomster et al. (2013) found that diabetes patients who $18.5 billion per year in 2012 (American Diabetes Association, 2013a).
report moderate to vigorous activity were 15% less likely to Given this high cost, even only a modest reduction in rates of diabetes
experience microvascular events and 17% less likely to die when complications and premature death is causally associated with better
compared to persons with diabetes who undertook no or only mild adherence to care recommendations; there would be a wide margin
physical activity. All these studies suggest potential benets of for benets by driving improvement in this realm.
adherence to physical activity requirements on reducing risks of
diabetes-associated complications and mortality, which are con- Financial disclosure
rmed by our study results.
We found no signicant link between physical activity and a No nancial disclosures were reported by the authors of this paper.
diagnosis of PDR during follow-up. Regular physical activity may
reduce the risk of PDR through its benecial effects on blood pressure
Acknowledgments
and high-density lipoprotein cholesterol, but earlier studies have also
found that physical activity has no effect on PDR (Cruickshanks, Moss,
This research was supported in part by the National Institute on
Klein, & Klein, 1992; Kriska, LaPorte, Patrick, Kuller, & Orchard, 1991),
Aging (grant R01-AG017473). The sponsors had no role in design and
possibly because physical activity is a relatively unimportant factor in
conduct of the study, collection, management, analysis, interpretation
the etiology of PDR (Cruickshanks et al., 1992).
of the data, preparation, review, approval of the manuscript, or
Our study has several important strengths. First, we focused on the
decision to submit the manuscript for publication. There is no conict
understudied link between adherence and onset of diabetes compli-
of interest to be reported. No nancial disclosures were reported by
cations and death. Many studies have reported a signicant
the authors of this paper. All three authors were involved in the
association between adherence to guidelines for care of diabetes
conception and design of the study, data analysis and interpretation,
and patient outcomes including glycemic control, health care
wrote and revised the paper, and approved the nal version of the
utilization, expenditures on personal health care services, and quality
submitted manuscript.
of life (Balkrishnan et al., 2003; Krapek et al., 2004; Martnez et al.,
2008; Raum et al., 2012; Saleh et al., 2014; Wong et al., 2014).
However, much less attention has been paid to effects of adherence on References
diabetes complications and mortality. Second, in contrast to most Agosti, V., Graziano, S., Artiaco, L., & Sorrentino, G. (2009). Biological mechanisms of
earlier studies that have focused on a limited scope of adherence stroke prevention by physical activity in type 2 diabetes. Acta Neurologica
measures, mainly only medication adherence, we used a more Scandinavica, 119(4), 213223.
Ainsworth, B. E., Haskell, W. L., Whitt, M. C., Irwin, M. L., Swartz, A. M., Strath, S. J., et al.
comprehensive set of adherence measures. This allowed us to (2000). Compendium of physical activities: An update of activity codes and MET
compare the relative importance of different types of adherence. intensities. Medicine and Science in Sports and Exercise, 32(9; SUPP/1), S498S504.
We acknowledge these study limitations. First, the HRS-DM survey American Diabetes Association (2013a). Economic costs of diabetes in the US in 2012.
Diabetes Care, 36(4), 10331046.
from which most of our data were obtained only contains 1901
American Diabetes Association (2013b). Standards of medical care in diabetes2013.
observations. After exclusions, the analysis sample fell to 1142 for Diabetes Care, 36(Suppl. 1), S11S66.
deaths and even less for analysis of outcomes conditional on not An, J., & Nichol, M. B. (2013). Multiple medication adherence and its effect on clinical
having the study complication at baseline. Second, we were unable to outcomes among patients with comorbid type 2 diabetes and hypertension.
Medical Care, 51(10), 879887.
document the reasons for non-use of diabetic drugs, which could Balkrishnan, R., Rajagopalan, R., Camacho, F. T., Huston, S. A., Murray, F. T., & Anderson,
reect either a lack of clinical indicators for use not observed in our R. T. (2003). Predictors of medication adherence and associated health care costs in

Please cite this article as: Chen, Y., et al., Adherence to diabetes guidelines for screening, physical activity and medication and onset of
complications and death, Journal of Diabetes and Its Complications (2015), http://dx.doi.org/10.1016/j.jdiacomp.2015.07.005
6 Y. Chen et al. / Journal of Diabetes and Its Complications xxx (2015) xxxxxx

an older population with type 2 diabetes mellitus: A longitudinal cohort study. Herbst, A., Kordonouri, O., Schwab, K. O., Schmidt, F., & Holl, R. W. (2007). Impact of
Clinical Therapeutics, 25(11), 29582971. physical activity on cardiovascular risk factors in children with type 1 diabetes: A
Bayer, F. J., Galusha, D., Slade, M., Chu, I. M., Taiwo, O., & Cullen, M. R. (2014). Process of multicenter study of 23,251 patients. Diabetes Care, 30(8), 20982100.
care compliance is associated with fewer diabetes complications. The American Kirk, A., Mutrie, N., MacIntyre, P., & Fisher, M. (2003). Increasing physical activity in
Journal of Managed Care, 20(1), 4152. people with type 2 diabetes. Diabetes Care, 26(4), 11861192.
Bethel, M. A., Sloan, F. A., Belsky, D., & Feinglos, M. N. (2007). Longitudinal incidence and Krapek, K., King, K., Warren, S. S., George, K. G., Caputo, D. A., Mihelich, K., et al. (2004).
prevalence of adverse outcomes of diabetes mellitus in elderly patients. Archives of Medication adherence and associated hemoglobin A1c in type 2 diabetes. Annals of
Internal Medicine, 167(9), 921927. Pharmacotherapy, 38(9), 13571362.
Blomster, J. I., Chow, C. K., Zoungas, S., Woodard, M., Patel, A., Poulter, N. R., et al. (2013). Kriska, A. M., LaPorte, R. E., Patrick, S. L., Kuller, L. H., & Orchard, T. J. (1991). The
The inuence of physical activity on vascular complications and mortality in association of physical activity and diabetic complications in individuals with
patients with type 2 diabetes mellitus. Diabetes, Obesity and Metabolism, 15(11), insulin-dependent diabetes mellitus: The Epidemiology of Diabetes Complications
10081012. StudyVII. Journal of Clinical Epidemiology, 44(11), 12071214.
Boul, N. G., Haddad, E., Kenny, G. P., Wells, G. A., & Sigal, R. J. (2001). Effects of exercise Kuo, Y. -F., Raji, M. A., Markides, K. S., Ray, L. A., Espino, D. V., & Goodwin, J. S. (2003).
on glycemic control and body mass in type 2 diabetes mellitus: A meta-analysis of Inconsistent use of diabetes medications, diabetes complications, and mortality in older
controlled clinical trials. The Journal of the American Medical Assocaition, 286(10), Mexican Americans over a 7-year period data from the Hispanic established population
12181227. for the epidemiologic study of the elderly. Diabetes Care, 26(11), 30543060.
Boyle, J. P., Thompson, T. J., Gregg, E. W., Barker, L. E., & Williamson, D. F. (2010). Martnez, Y. V., Prado-Aguilar, C. A., Rascn-Pacheco, R. A., & Valdivia-Martnez, J. J.
Projection of the year 2050 burden of diabetes in the US adult population: Dynamic (2008). Quality of life associated with treatment adherence in patients with type 2
modeling of incidence, mortality, and prediabetes prevalence. Population Health diabetes: A cross-sectional study. BMC Health Services Research, 8(1), 164173.
Metrics, 8(1), 2940. Plotnikoff, R. C., Taylor, L. M., Wilson, P. M., Courneya, K. S., Sigal, R. J., Birkett, N., et al.
Brown, J. B., Pedula, K. L., & Bakst, A. W. (1999). The progressive cost of complications in (2006). Factors associated with physical activity in Canadian adults with diabetes.
type 2 diabetes mellitus. Archives of Internal Medicine, 159(16), 18731880. Medicine and Science in Sports and Exercise, 38(8), 15261534.
Centers for Disease Control and Prevention (2011). National diabetes fact sheet: National Raum, E., Krmer, H. U., Rter, G., Rothenbacher, D., Rosemann, T., Szecsenyi, J., et al.
estimates and general information on diabetes and prediabetes in the United States, (2012). Medication non-adherence and poor glycaemic control in patients with
2011. Atlanta, GA: US Department of Health and Human Services, Centers for type 2 diabetes mellitus. Diabetes Research and Clinical Practice, 97(3), 377384.
Disease Control and Prevention. Saleh, F., Mumu, S. J., Ara, F., Hafez, M. A., & Ali, L. (2014). Non-adherence to self-care
Cruickshanks, K. J., Moss, S. E., Klein, R., & Klein, B. E. (1992). Physical activity and practices & medication and health related quality of life among patients with type 2
proliferative retinopathy in people diagnosed with diabetes before age 30 yr. diabetes: A cross-sectional study. BMC Public Health, 14(1), 431438.
Diabetes Care, 15(10), 12671272. Selvin, E., Marinopoulos, S., Berkenblit, G., Rami, T., Brancati, F. L., Powe, N. R., et al.
Feil, D. G., Zhu, C. W., & Sultzer, D. L. (2012). The relationship between cognitive (2004). Meta-analysis: Glycosylated hemoglobin and cardiovascular disease in
impairment and diabetes self-management in a population-based community sample diabetes mellitus. Annals of Internal Medicine, 141(6), 421431.
of older adults with type 2 diabetes. Journal of Behavioral Medicine, 35(2), 190199. Stratton, I. M., Adler, A. I., Neil, H. A. W., Matthews, D. R., Manley, S. E., Cull, C. A., et al.
Gibson, T. B., Song, X., Alemayehu, B., Wang, S. S., Waddell, J. L., Bouchard, J. R., et al. (2000). Association of glycaemia with macrovascular and microvascular compli-
(2010). Cost sharing, adherence, and health outcomes in patients with diabetes. cations of type 2 diabetes (UKPDS 35): Prospective observational study. BMJ Clinical
The American Journal of Managed Care, 16(8), 589600. Research, 321(7258), 405412.
Greeneld, S., Billimek, J., & Kaplan, S. (2010). The total illness burden index. In V. Wong, E. S., Bryson, C. L., Hebert, P. L., & Liu, C. -F. (2014). Estimating the impact of oral
Preedy, & R. Watson (Eds.), Handbook of disease burdens and quality of life measures diabetes medication adherence on medical costs in VA. Annals of Pharmacotherapy,
(pp. 7385). New York: Springer. 48(8), 978985.
Greeneld, S., Sullivan, L., Dukes, K. A., Silliman, R., D'Agostino, R., & Kaplan, S. H. (1995). Zhuo, X., Zhang, P., & Hoerger, T. J. (2013). Lifetime direct medical costs of treating type
Development and testing of a new measure of case mix for use in ofce practice. 2 diabetes and diabetic complications. American Journal of Preventive Medicine,
Medical Care, 33(4), AS47AS55. 45(3), 253261.

Please cite this article as: Chen, Y., et al., Adherence to diabetes guidelines for screening, physical activity and medication and onset of
complications and death, Journal of Diabetes and Its Complications (2015), http://dx.doi.org/10.1016/j.jdiacomp.2015.07.005