[Skip to Content]
[Skip to Content Landing]
Figure.
Diabetes Medication Adherence
Diabetes Medication Adherence

Inadequate adherence to newly prescribed oral hypoglycemic medications based on new prescription medication gap by race/ethnicity, English-language proficiency, and patient-physician language concordance. Following convention, patients are categorized as demonstrating “adequate” or “inadequate” adherence based on commonly used cut points for gaps in medication supplies (ie, new prescription medication gap <20% vs ≥20% of follow-up time, respectively).

Table 1.  
Definitions of Measures of Medication Adherence
Definitions of Measures of Medication Adherence
Table 2.  
Characteristics of 30 838 Insured Patients with Diabetes by Race/Ethnicity, English-Language Proficiency, and Patient-Physician Language Concordancea
Characteristics of 30 838 Insured Patients with Diabetes by Race/Ethnicity, English-Language Proficiency, and Patient-Physician Language Concordancea
Table 3.  
Nonadherence to Newly Prescribed Diabetes Medications by Race/Ethnicity, English-Language Proficiency, and Patient-Physician Language Concordancea
Nonadherence to Newly Prescribed Diabetes Medications by Race/Ethnicity, English-Language Proficiency, and Patient-Physician Language Concordancea
Table 4.  
Nonadherence to Newly Prescribed Diabetes Medications for Latino Groups Compared With White Patients and for Latino Patients by English-Language Proficiency and Physician Language Concordancea
Nonadherence to Newly Prescribed Diabetes Medications for Latino Groups Compared With White Patients and for Latino Patients by English-Language Proficiency and Physician Language Concordancea
1.
Centers for Disease Control and Prevention. Age-adjusted rates of diagnosed diabetes per 100 civilian, non-institutionalized population, by Hispanic origin, United States, 1997-2014. https://www.cdc.gov/diabetes/statistics/prev/national/figbyhispanic.htm. Accessed December 27, 2016.
2.
Haynes  RB, Ackloo  E, Sahota  N, McDonald  HP, Yao  X.  Interventions for enhancing medication adherence.  Cochrane Database Syst Rev. 2008;(2):CD000011.PubMedGoogle Scholar
3.
Osterberg  L, Blaschke  T.  Adherence to medication.  N Engl J Med. 2005;353(5):487-497.PubMedGoogle ScholarCrossref
4.
Rodriguez  F, Cannon  CP, Steg  PG,  et al; REACH Registry Investigators.  Predictors of long-term adherence to evidence-based cardiovascular disease medications in outpatients with stable atherothrombotic disease: findings from the REACH Registry.  Clin Cardiol. 2013;36(12):721-727.PubMedGoogle ScholarCrossref
5.
Palacio  AM, Uribe  C, Hazel-Fernandez  L,  et al.  Can phone-based motivational interviewing improve medication adherence to antiplatelet medications after a coronary stent among racial minorities? a randomized trial.  J Gen Intern Med. 2015;30(4):469-475.PubMedGoogle ScholarCrossref
6.
van Servellen  G, Lombardi  E.  Supportive relationships and medication adherence in HIV-infected, low-income Latinos.  West J Nurs Res. 2005;27(8):1023-1039.PubMedGoogle ScholarCrossref
7.
Ryan  C. Language Use in the United States, 2011: American Community Survey Reports. Washington, DC: US Census Bureau; 2013.
8.
Fernandez  A, Schillinger  D, Warton  EM,  et al.  Language barriers, physician-patient language concordance, and glycemic control among insured Latinos with diabetes: the Diabetes Study of Northern California (DISTANCE).  J Gen Intern Med. 2011;26(2):170-176.PubMedGoogle ScholarCrossref
9.
Clay  BJ, Halasyamani  L, Stucky  ER, Greenwald  JL, Williams  MV.  Results of a medication reconciliation survey from the 2006 Society of Hospital Medicine national meeting.  J Hosp Med. 2008;3(6):465-472.PubMedGoogle ScholarCrossref
10.
Karter  AJ, Subramanian  U, Saha  C,  et al.  Barriers to insulin initiation: the translating research into action for diabetes insulin starts project.  Diabetes Care. 2010;33(4):733-735.PubMedGoogle ScholarCrossref
11.
Rivadeneyra  R, Elderkin-Thompson  V, Silver  RC, Waitzkin  H.  Patient centeredness in medical encounters requiring an interpreter.  Am J Med. 2000;108(6):470-474.PubMedGoogle ScholarCrossref
12.
Schenker  Y, Karter  AJ, Schillinger  D,  et al.  The impact of limited English proficiency and physician language concordance on reports of clinical interactions among patients with diabetes: the DISTANCE Study.  Patient Educ Couns. 2010;81(2):222-228.PubMedGoogle ScholarCrossref
13.
Karliner  LS, Pérez-Stable  EJ, Gildengorin  G.  The language divide: the importance of training in the use of interpreters for outpatient practice.  J Gen Intern Med. 2004;19(2):175-183.PubMedGoogle ScholarCrossref
14.
Karter  AJ, Ferrara  A, Liu  JY, Moffet  HH, Ackerson  LM, Selby  JV.  Ethnic disparities in diabetic complications in an insured population.  JAMA. 2002;287(19):2519-2527.PubMedGoogle ScholarCrossref
15.
Karter  AJ, Schillinger  D, Adams  AS,  et al.  Elevated rates of diabetes in Pacific Islanders and Asian subgroups: the Diabetes Study of Northern California (DISTANCE).  Diabetes Care. 2013;36(3):574-579.PubMedGoogle ScholarCrossref
16.
Rosenthal  A, Wang  F, Schillinger  D, Pérez Stable  EJ, Fernandez  A.  Accuracy of physician self-report of Spanish language proficiency.  J Immigr Minor Health. 2011;13(2):239-243.PubMedGoogle ScholarCrossref
17.
Parker  MM, Moffet  HH, Adams  A, Karter  AJ.  An algorithm to identify medication nonpersistence using electronic pharmacy databases.  J Am Med Inform Assoc. 2015;22(5):957-961.PubMedGoogle ScholarCrossref
18.
Karter  AJ, Parker  MM, Moffet  HH, Ahmed  AT, Schmittdiel  JA, Selby  JV.  New prescription medication gaps: a comprehensive measure of adherence to new prescriptions.  Health Serv Res. 2009;44(5, pt 1):1640-1661.PubMedGoogle ScholarCrossref
19.
Raebel  MA, Schmittdiel  J, Karter  AJ, Konieczny  JL, Steiner  JF.  Standardizing terminology and definitions of medication adherence and persistence in research employing electronic databases.  Med Care. 2013;51(8)(suppl 3):S11-S21.PubMedGoogle ScholarCrossref
20.
Greenland  S, Pearl  J, Robins  JM.  Causal diagrams for epidemiologic research.  Epidemiology. 1999;10(1):37-48.PubMedGoogle ScholarCrossref
21.
Robins  JM.  Data, design, and background knowledge in etiologic inference.  Epidemiology. 2001;12(3):313-320.PubMedGoogle ScholarCrossref
22.
Hernán  MA, Hernández-Díaz  S, Werler  MM, Mitchell  AA.  Causal knowledge as a prerequisite for confounding evaluation: an application to birth defects epidemiology.  Am J Epidemiol. 2002;155(2):176-184.PubMedGoogle ScholarCrossref
23.
Pearl  J.  Causality: Models, Reasoning, and Inference. 2nd ed. New York, NY: Cambridge University Press; 2009.
24.
Pearl  J.  Causal diagrams for empirical research.  Biometrika. 1995;82(4):669-688. doi:10.1093/biomet/82.4.669Google ScholarCrossref
25.
Laraia  BA, Karter  AJ, Warton  EM, Schillinger  D, Moffet  HH, Adler  N.  Place matters: neighborhood deprivation and cardiometabolic risk factors in the Diabetes Study of Northern California (DISTANCE).  Soc Sci Med. 2012;74(7):1082-1090.PubMedGoogle ScholarCrossref
26.
Alsabbagh  MH, Lemstra  M, Eurich  D,  et al.  Socioeconomic status and nonadherence to antihypertensive drugs: a systematic review and meta-analysis.  Value Health. 2014;17(2):288-296.PubMedGoogle ScholarCrossref
27.
Zou  G.  A modified poisson regression approach to prospective studies with binary data.  Am J Epidemiol. 2004;159(7):702-706.PubMedGoogle ScholarCrossref
28.
Heisler  M, Faul  JD, Hayward  RA, Langa  KM, Blaum  C, Weir  D.  Mechanisms for racial and ethnic disparities in glycemic control in middle-aged and older Americans in the health and retirement study.  Arch Intern Med. 2007;167(17):1853-1860.PubMedGoogle ScholarCrossref
29.
Traylor  AH, Schmittdiel  JA, Uratsu  CS, Mangione  CM, Subramanian  U.  Adherence to cardiovascular disease medications: does patient-provider race/ethnicity and language concordance matter?  J Gen Intern Med. 2010;25(11):1172-1177.PubMedGoogle ScholarCrossref
30.
Hu  D, Juarez  DT, Yeboah  M, Castillo  TP.  Interventions to increase medication adherence in African-American and Latino populations: a literature review.  Hawaii J Med Public Health. 2014;73(1):11-18.PubMedGoogle Scholar
31.
McQuaid  EL, Everhart  RS, Seifer  R,  et al.  Medication adherence among Latino and non-Latino white children with asthma.  Pediatrics. 2012;129(6):e1404-e1410.PubMedGoogle ScholarCrossref
32.
Compton  S, Haack  S, Phillips  CR.  Identification of barriers to medication adherence in a Latino population.  Res Social Adm Pharm. 2010;6(4):365-371.PubMedGoogle ScholarCrossref
33.
Lyles  CR, Seligman  HK, Parker  MM,  et al.  Financial strain and medication adherence among diabetes patients in an integrated health care delivery system: the Diabetes Study of Northern California (DISTANCE).  Health Serv Res. 2016;51(2):610-624.PubMedGoogle ScholarCrossref
34.
Gazmararian  JA, Kripalani  S, Miller  MJ, Echt  KV, Ren  J, Rask  K.  Factors associated with medication refill adherence in cardiovascular-related diseases: a focus on health literacy.  J Gen Intern Med. 2006;21(12):1215-1221.PubMedGoogle ScholarCrossref
35.
Conn  VS, Enriquez  M, Ruppar  TM, Chan  KC.  Cultural relevance in medication adherence interventions with underrepresented adults: systematic review and meta-analysis of outcomes.  Prev Med. 2014;69:239-247.PubMedGoogle ScholarCrossref
36.
Wilson  E, Chen  AH, Grumbach  K, Wang  F, Fernandez  A.  Effects of limited English proficiency and physician language on health care comprehension.  J Gen Intern Med. 2005;20(9):800-806.PubMedGoogle ScholarCrossref
37.
Wassertheil-Smoller  S, Arredondo  EM, Cai  J,  et al.  Depression, anxiety, antidepressant use, and cardiovascular disease among Hispanic men and women of different national backgrounds: results from the Hispanic Community Health Study/Study of Latinos.  Ann Epidemiol. 2014;24(11):822-830.PubMedGoogle ScholarCrossref
38.
Gany  F, Leng  J, Shapiro  E,  et al.  Patient satisfaction with different interpreting methods: a randomized controlled trial.  J Gen Intern Med. 2007;22(suppl 2):312-318.PubMedGoogle ScholarCrossref
39.
Slean  GR, Jacobs  EA, Lahiff  M, Fisher  L, Fernandez  A.  Aspects of culturally competent care are associated with less emotional burden among patients with diabetes.  Med Care. 2012;50(9)(suppl 2):S69-S73.PubMedGoogle ScholarCrossref
40.
Meyers  K, Tang  G, Fernandez  A.  Responding to the language challenge: Kaiser Permanente’s approach.  Perm J. 2009;13(3):77-83.PubMedGoogle ScholarCrossref
41.
Parker  MM, Fernandez  A, Moffet  HH, Grant  RW. Torreblanca  A, Karter  AJ.  Association of patient-physician language concordance and glycemic control for limited-English proficiency Latinos with type 2 diabetes [published online January 23, 2017].  JAMA Intern Med. doi:10.1001/jamainternmed.2016.8648Google Scholar
42.
Gee  GC, Walsemann  KM, Takeuchi  DT.  English proficiency and language preference: testing the equivalence of two measures.  Am J Public Health. 2010;100(3):563-569.PubMedGoogle ScholarCrossref
Original Investigation
March 2017

Adherence to Newly Prescribed Diabetes Medications Among Insured Latino and White Patients With Diabetes

Author Affiliations
  • 1Division of General Internal Medicine, San Francisco General Hospital, San Francisco, California
  • 2Department of Medicine, University of California–San Francisco, San Francisco
  • 3Center for Vulnerable Populations, San Francisco General Hospital, San Francisco, California
  • 4Division of Research, Kaiser Permanente, Oakland, California
JAMA Intern Med. 2017;177(3):371-379. doi:10.1001/jamainternmed.2016.8653
Key Points

Question  What is the role of patient race/ethnicity, preferred language, and physician language concordance on adherence to newly prescribed diabetes medications among Latino and white patients?

Findings  In this 2-year study, overall nonadherence to newly prescribed diabetes medications was observed in 60.2% of Spanish-speaking Latino patients, 51.7% of English-speaking Latino patients, and 37.5% of white patients, indicating significant differences among groups. Nonadherence among Spanish-speaking Latino patients did not vary with the Spanish-language fluency of their physicians.

Meaning  Interventions beyond ensuring access to interpreters or patient-physician language concordance will be required to improve medication adherence among Latino Spanish- and English-speaking patients with diabetes.

Abstract

Importance  Medication adherence is essential to diabetes care. Patient-physician language barriers may affect medication adherence among Latino individuals.

Objective  To determine the association of patient race/ethnicity, preferred language, and physician language concordance with patient adherence to newly prescribed diabetes medications.

Design, Setting, and Participants  This observational study was conducted from January 1, 2006, to December 31, 2012, at a large integrated health care delivery system with professional interpreter services. Insured patients with type 2 diabetes, including English-speaking white, English-speaking Latino, or limited English proficiency (LEP) Latino patients with newly prescribed diabetes medication.

Exposures  Patient race/ethnicity, preferred language, and physician self-reported Spanish-language fluency.

Main Outcomes and Measures  Primary nonadherence (never dispensed), early-stage nonpersistence (dispensed only once), late-stage nonpersistence (received ≥2 dispensings, but discontinued within 24 months), and inadequate overall medication adherence (>20% time without sufficient medication supply during 24 months after initial prescription).

Results  Participants included 21 878 white patients, 5755 English-speaking Latino patients, and 3205 LEP Latino patients with a total of 46 131 prescriptions for new diabetes medications. Among LEP Latino patients, 50.2% (n = 1610) had a primary care physician reporting high Spanish fluency. For oral medications, early adherence varied substantially: 1032 LEP Latino patients (32.2%), 1565 English-speaking Latino patients (27.2%), and 4004 white patients (18.3%) were either primary nonadherent or early nonpersistent. Inadequate overall adherence was observed in 1929 LEP Latino patients (60.2%), 2975 English-speaking Latino patients (51.7%), and 8204 white patients (37.5%). For insulin, early-stage nonpersistence was 42.8% among LEP Latino patients (n = 1372), 34.4% among English-speaking Latino patients (n = 1980), and 28.5% among white patients (n = 6235). After adjustment for patient and physician characteristics, LEP Latino patients were more likely to be nonadherent to oral medications and insulin than English-speaking Latino patients (relative risks from 1.11 [95% CI, 1.06-1.15] to 1.17 [95% CI, 1.02-1.34]; P < .05) or white patients (relative risks from 1.36 [95% CI, 1.31-1.41] to 1.49 [95% CI, 1.32-1.69]; P < .05). English-speaking Latino patients were more likely to be nonadherent compared with white patients (relative risks from 1.23 [95% CI, 1.19-1.27] to 1.30 [95% CI, 1.23-1.39]; P < .05). Patient-physician language concordance was not associated with rates of nonadherence among LEP Latinos (relative risks from 0.92 [95% CI, 0.71-1.19] to 1.04 [95% CI, 0.97-1.1]; P > .28).

Conclusions and Relevance  Nonadherence to newly prescribed diabetes medications is substantially greater among Latino than white patients, even among English-speaking Latino patients. Limited English proficiency Latino patients are more likely to be nonadherent than English-speaking Latino patients independent of the Spanish-language fluency of their physicians. Interventions beyond access to interpreters or patient-physician language concordance will be required to improve medication adherence among Latino patients with diabetes.

Introduction

More than 3.1 million Latino individuals in the United States have a diagnosis of diabetes and require daily medication use.1 Adherence to long-term medications is poor in the general population,2,3 and perhaps particularly poor among Latino individuals.4-6 Poor patient-physician communication has been posited as a key barrier to medication adherence.3 Language barriers in health care are common for the Latino population, as more than 44% have limited English proficiency (LEP), defined as speaking English less than “very well.”7 Language barriers between LEP Latino patients and their non–Spanish-speaking physicians have been associated with higher risk for poor glycemic control. In a prior study, we found that LEP Latino patients whose physicians did not speak Spanish were twice as likely as those with Spanish-speaking physicians to have poor glycemic control (28% vs 16%).8

Language barriers might lead to poorer glycemic control via several pathways. Medication reconciliation can be difficult across language barriers,9 and physicians may be reluctant to initiate or intensify medications when they are uncertain of a patient’s current medication use. Patients are less likely to initiate insulin if they feel their physician did not adequately explain the risks and benefits.10 Encounters carried out across language barriers are often less patient centered,11 and LEP patients with language-discordant physicians are less likely to report trust in their physicians.12 Physicians who lack fluency in the patient’s language may feel stymied when attempting to elicit or address a patient’s medication concerns,13 and patients’ lingering questions and lack of comprehension of their medication regimen may lead to poorer medication adherence. However, the extent to which language barriers affect medication adherence for newly prescribed medications, and how these potential barriers play out in health care settings with uniform access to professional interpreter services, is not well understood.

We designed a study to evaluate the role of ethnicity and language barriers, specifically patient-physician language concordance, on Latino patients’ nonadherence to newly prescribed oral hypoglycemic medications or insulin.

We compared medication nonadherence across 4 groups of insured patients with diabetes: (1) English-speaking non-Latino white patients; (2) English-speaking Latino patients; (3) LEP Latino patients whose physicians were fluent in Spanish (LEP concordant); and (4) LEP Latino patients whose physicians were not fluent in Spanish (LEP discordant). Our hypotheses were that LEP-discordant patients would have greater medication nonadherence than LEP-concordant patients, and that nonadherence among both English-speaking and LEP Latino patients would be higher than for white patients.

Methods
Study Setting and Population

The study setting was Kaiser Permanente Northern California (KPNC), an integrated health care delivery system that provides comprehensive medical services to approximately 3.9 million members. The study population was drawn from the KPNC Diabetes Registry, a well-characterized, ethnically diverse diabetic population.14,15

For this analysis, participants were eligible if they were active health plan members with type 2 diabetes, self-identified as Latino or white, whose preferred language was English or Spanish, and were aged 18 years or older on the date they were prescribed a new oral hypoglycemic medication or insulin during 2006-2010 (n = 116 802). Patients were excluded if they had end-stage renal disease (n = 1744), gaps in KPNC coverage and/or pharmacy benefits for more than 2 months during the 2 years prior through the 2 years after the date of the new medication prescription (n = 35 124), or were not empaneled at the time of the new medication prescription with a primary care physician for whom Spanish-language proficiency data existed (n = 49 096), leaving 30 838 patients for the study.

This study was approved by the institutional review boards of KPNC and the University of California, San Francisco; the requirement that informed consent be obtained from study participants was waived.

Measures

Latino patients whose preferred language was Spanish in the electronic health record were considered to have LEP. To determine language concordance, we identified the Spanish fluency of each LEP Latino patient’s physician using data from 2 surveys: (1) an administrative survey given to physicians on employment and (2) a web-based survey we developed specifically to ask KPNC physicians about their Spanish-language ability conducted in 2012. Both surveys asked physicians to indicate their level of Spanish-speaking proficiency as “not at all,” “low,” “moderate,” or “high.” If data from both surveys were available, we used answers from the more recent web-based survey. Physicians whose self-reported Spanish-speaking proficiency was “high” were considered fluent in Spanish16; their LEP Latino patients were designated as being in language-concordant relationships (LEP concordant) and LEP Latino patients whose physicians were not fluent in Spanish were designated as being language discordant (LEP discordant).

Nonadherence

Our outcomes of interest were measures of patient nonadherence to each newly prescribed hypoglycemic medication. Diabetes medications were classified as oral or insulin. See the eAppendix in the Supplement for details. Medication adherence was estimated using medication dispensing data from KPNC pharmacies; patients’ pharmacy benefits were limited to those pharmacies. To be considered a new medication, we required that there was no dispensing of the medication in the prior 24 months; for each new medication (“index therapy”), we measured adherence subsequent to “index date,” defined as the date of first dispensing (or prescribing date if there were zero dispensings). We used 4 validated measures to assess medication nonadherence to each newly prescribed diabetes medication.17-19 (See definitions in Table 1.) Three of these measures evaluated nonadherence at specific milestones, while the fourth, a continuous measure of overall adherence (new prescription medication gap [NPMG]), estimated gaps in medication supplies beginning with the index date until the end of follow-up (earlier of 24 months or censorship [date the primary care physician stopped the medication or switched the patient to an alternative medication]). Following convention, we then categorized patients as demonstrating ‘‘adequate’’ or ‘‘inadequate’’ adherence based on commonly used cut points for gaps in medication supplies (ie, NPMG <20% vs ≥20% of follow-up time, respectively).18 We calculated all 4 measures of medication nonadherence for oral medications. For insulin, we calculated only primary nonadherence and early-stage nonpersistence. We did not estimate NPMG as insulin use can vary daily with patient requirements, making estimates of this and similar continuous measures of adherence unreliable when based on pharmacy dispensing data alone.

Data Analysis

To guide covariate selection, we constructed a causal directed acyclic graph (DAG) based on our team’s interpretation of a broad review of the literature and hypothesized causal pathways that could link language barriers to medication adherence (eFigure in the Supplement).20-22 Directed acyclic graphs are particularly useful in analyses where many factors might influence an outcome, as they clarify the relationships between the factors and assist in defining potential confounders (which should be included as covariates) and mediators (which need not be included in analytic models). From this DAG analysis, we specified the set of covariates required in adjusted regression models to generate causal estimates of the effect of language barriers on adherence.20,23,24

Based on our DAG, the necessary covariates included patient age, sex, race/ethnicity, and socioeconomic status and the physician’s race/ethnicity, age, and sex. The Neighborhood Deprivation Index25 was included as a contextual measure of patients’ socioeconomic status as socioeconomic status has been associated with adherence.26 Mediators in our DAG model included comorbidity index, number of nondiabetes medications, out-of-pocket cost, and depression. See the eAppendix in the Supplement for variable definitions.

Bivariate analyses comparing covariates and other patient characteristics between specific LEP language groups were performed using χ2 tests for categorical variables, t tests for interval variables if normally distributed, and Wilcoxon rank sum tests for interval variables not normally distributed.

Repeated measures, multivariate, and modified Poisson regression models, with robust error variance and logarithm link,27 were generated to calculate adjusted relative risk of the binary outcomes (ie, nonadherence to newly prescribed oral hypoglycemic medications or insulin). We chose this method over logistic regression or binomial regression because (1) the odds ratio is a biased estimate of relative risk for common outcomes and (2) convergence problems are common in binomial regression models. We used robust variance estimates because log Poisson models tend to give conservative results.27 Our models also adjusted the residual covariance structure for within-patient clustering, as some patients were prescribed more than 1 new medication during the study. We compared English-speaking and LEP Latino vs white patients, LEP Latino vs English-speaking Latino patients, and LEP-discordant vs LEP-concordant Latino patients.

We conducted a sensitivity analysis incorporating physicians who reported “moderate” fluency into the language-concordant group.

Results

The study population of 30 838 included 21 878 white, 5755 English-speaking Latino, and 3205 LEP Latino patients. Among LEP Latinos, 1610 (50.2%) had a physician who was language concordant) and 1595 patients (49.8%) had a physician who was language discordant) (Table 2).

Compared with white patients, Latino patients were somewhat younger and lived in poorer neighborhoods. They also had fewer comorbid illnesses, fewer long-term nondiabetes medications, lower out-of-pocket costs, and were less likely to be diagnosed as having depression. However, Latino patients had higher rates of poor glycemic control (hemoglobin A1c >9%) than white patients, with LEP Latino patients at 37%, English-speaking Latino patients at 36%, and white patients at 24%. Limited English proficiency–concordant and LEP-discordant patients were similar, although depression was more commonly diagnosed among the LEP concordant (15.1% vs 11.0%; P < .05) (Table 2).

A total of 696 physicians cared for the 30 838 patients, with each physician caring for 45 patients on average (SD, 33; range, 1-156). Physicians differed slightly across groups by age, sex, and race/ethnicity, with physicians caring for LEP Latino patients more likely to identify as Latino than physicians caring for white or English-speaking Latino patients (42.3% vs 7.2% and 13.3%). There were 46 131 new prescriptions for diabetes medications (39 157 for oral medications and 6974 for insulin) over the 5-year study period. At each adherence stage, LEP Latino patients had greater nonadherence to oral medication prescriptions than English-speaking Latino patients, who in turn had greater nonadherence than white patients (Table 3). Initiation of treatment to oral medications was poor: 32.2% of LEP Latino patients, 27.2% of English-speaking Latino patients, and 18.3% of white patients were either primary nonadherent or early nonpersistent. More than half (54.2%) of LEP Latino patients were nonadherent to treatment with oral diabetes medications before the end of follow-up.

The prevalence of inadequate adherence for oral medications based on NPMG was substantially greater for Latino than white patients, and greater still for LEP Latino patients (60.2% for LEP Latino patients, 51.7% for English-speaking Latino patients, and 37.5% for white patients) (P < .05 for each comparison) (Figure).

The ethnic-language patterns for primary nonadherence for the far fewer new insulin prescriptions were similar to those of oral medications, although the differences were not statistically significant. Early-stage nonpersistence with insulin varied substantially: 42.8% among LEP Latino patients, 34.4% among English-speaking Latino patients, and 28.5% among white patients (P < .05 for each comparison). There were no significant differences between the LEP-concordant and LEP-discordant groups for any of the measures.

The patterns described here persisted after adjustment for variables guided by our DAG and in fuller models incorporating potential mediators (comorbidity index, number of nondiabetes medications, out-of-pocket cost, and depression) (Table 4 and the eTable in the Supplement). Relative to white patients, English-speaking Latino patients had adjusted relative risks ranging from 1.23 (95% CI, 1.19-1.27) to 1.30 (95% CI, 1.23-1.39) for nonadherence to oral medications and insulin, while LEP Latino patients had somewhat larger adjusted relative risks ranging from 1.36 (95% CI, 1.31-1.41) to 1.49 (95% CI, 1.32-1.69). Limited English proficient–concordant and LEP-discordant patients had similar risks of nonadherence relative to white patients, ranging from a low of 1.33 (95% CI, 1.26-1.4) to a high of 1.52 (95% CI, 1.3-1.76). The differences in primary nonadherence of insulin were smaller and not significant.

Differences in nonadherence between LEP Latino and English-speaking Latino patients, while smaller than those noted between Latino groups and white patients, also persisted after adjustment for patient and physician factors (Table 4). Compared with English-speaking Latino patients, LEP Latino patients had relative risks ranging from 1.11 (95% CI, 1.06-1.15) to 1.17 (95% CI, 1.02-1.34). As in the unadjusted analyses, no differences in nonadherence were noted between LEP-concordant vs LEP-discordant patients at any stage of medication use. Relative to LEP-concordant patients, LEP-discordant patients had relative risks ranging from 0.92 (95% CI, 0.71-1.19) to 1.04 (95% CI, 0.97-1.1).

A sensitivity analysis incorporating physicians reporting “moderate” Spanish into the language-concordant group resulted in a slight change in patient nonadherence by NPMG (LEP concordant, 52.6% vs LEP discordant, 55.3%; P = .13).

Discussion

In this study of nonadherence to newly prescribed diabetes medications among insured patients in an integrated health care delivery system with uniform access to interpreter services, we found substantive, graded differences in nonadherence among white, English-speaking Latino, and LEP Latino patients. While the greatest nonadherence was among LEP Latino patients, large differences in adherence between white and Latino groups existed irrespective of English-language proficiency. Contrary to our hypothesis that language discordance in the patient-physician encounter would be associated with greater medication nonadherence for LEP patients, we found no difference in nonadherence measures among the LEP Latino patients when examined by the Spanish fluency of their physician.

The relationship between Latino ethnicity, English-language proficiency, and medication nonadherence has been difficult to study. Although a population-level study found no differences between elderly Latino and white patients in self-report of diabetes medication nonadherence,28 the association between Latino ethnicity and medication nonadherence has been observed in various clinical settings and with diverse illnesses.28-31 Using this same Kaiser diabetes registry, a previous study of ongoing medication adherence (using other measures of adherence and models with different covariates) found that a combined lipid, blood pressure, and diabetes medication adherence outcome varied by ethnicity of the patient and by the language capability of the physicians, although diabetes medications alone did not.29 Multiple explanations have been posited for greater nonadherence among Latino patients, including issues of pharmacy logistics,32 financial barriers,33 numeracy and literacy barriers,34 cultural attitudes toward medications,35 and language barriers.36 Latino individuals, particularly immigrants, also have high rates of depression, which may contribute to medication nonadherence.37

Our study adds to this literature in several ways. First, it offers a clinically relevant measurement of the substantial difference in newly prescribed medication adherence between Latino and white patients with the same disease, in a setting conducive to medication adherence, and irrespective of Latino language preference. Despite favorable clinical characteristics (ie, fewer comorbidities and less medication burden), approximately a third of LEP Latino patients never started their prescribed treatment to oral medications and half never started prescribed insulin. Between 50% and 60% of Latino patients (English-speaking or LEP) did not have adequate supply of a newly prescribed diabetes medication during follow-up (eg, were without medication at least 4.8 of 24 months) while the same was true for about 37% of white patients.

Second, we also observed that language concordance between clinician and patient is insufficient to eliminate disparities in adherence for LEP Latino patients with diabetes. Prior work has consistently demonstrated that LEP patients with language-discordant physicians report less comprehension of diagnosis and treatment, including medication instructions.36,38 Limited English proficient patients report more adverse medication events, less trust in physicians, and less satisfaction with the medical encounter.12,36,39 Physicians caring for LEP patients report that language barriers make it difficult to elicit symptoms, reconcile medications, and establish rapport.13 Why then, did we find no differences in medication adherence by patient-physician language concordance and only small differences in adherence between LEP and English-speaking Latino patients? Several facts are worth considering. Medical homes that harness the skills of ethnically and linguistically diverse medical staff (eg, nurse care managers) may play a large role, as support personnel may supplant the central role of the primary care physician in medication coaching. Uniform access to professional interpreter systems, testing and certification of bilingual staff,40 and greater physician awareness of language-associated disparities may also help mitigate language barriers in the clinical encounter. Yet, the high nonadherence rate observed suggests that even these programs are insufficient to overcome barriers to medication adherence among LEP and English-speaking Latino patients; more research, using qualitative and quantitative methods, is clearly needed to determine key factors and enable successful interventions.

Our study of medication nonadherence should not be taken to imply that language concordance is not associated with better glycemic control among Latino individuals. Glycemic control is dictated by a range of lifestyle factors (eg, diet, exercise, weight control, and stress) in addition to use of pharmacotherapy. In a recent study, we found that LEP Latino patients who switched from a language-discordant physician to a language-concordant physician improved glycemic control more than those who switched to another language-discordant physician.41 Even in the absence of adherence differences, improved attention to lifestyle, or intensification of insulin, which we could not capture, could account for the improvement in glycemic control observed when patients switched to language-concordant physicians.

Strengths and Limitations

Our study has additional strengths. To our knowledge, this is the first study to determine ethnic-language medication adherence differences by examining all new diabetes medication prescriptions in the clinical record. Most studies have relied on estimates of ongoing medication adherence. The use of detailed electronic records allowed us to use multiple, validated measures of medication nonadherence to examine oral diabetes medication and insulin adherence overall and at different stages of medication use, including primary nonadherence, which is typically not captured. We have shown previously that because Kaiser Permanente maintains a closed pharmacy system, ascertainment of pharmacy use in a sample with pharmacy benefits is quite complete.18 We were able to include a direct measure of physician Spanish fluency via a physician self-report survey in addition to administrative data. Sensitivity analysis showed little variation when we altered our cut-off for physician fluency. Finally, the integrated health care system setting in this study affords all patients relatively uniform access and quality of care, relatively small financial and logistical barriers to medication use, and access to bilingual staff and interpreter services, thereby increasing the internal validity of our findings.

Our study also has several limitations. First, we were unable to measure individual-level socioeconomic indicators, such as education and income, health literacy, or acculturation level of patients, which are all potentially important factors in medication adherence. Instead, we relied on the Neighborhood Deprivation Index, a contextual measure of socioeconomic status. Second, we may have misclassification errors in adjudication of patient English-proficiency status using administrative data. Data from our own work and others indicate that about 20% to 30% of patients indicate a non–English-language preference despite speaking English well.42 If patients classified as LEP with language-discordant clinicians were actually English speaking, a small difference between the LEP groups may have been obscured. Adoption of current recommendations to include an English-language proficiency question when capturing race, ethnicity, and language data would strengthen studies of this type. Third, we measured adherence after each new prescription in the electronic medical record. If a patient refused a medication during the clinical encounter, the physician would most likely not write a prescription. In our study, lower rates of insulin use among Latino than white patients, along with higher rates of poor glycemic control among Latino patients, suggest that Latino patients might be refusing insulin within the clinical encounter, resulting in no prescription and no opportunity to measure adherence. Thus, our findings likely underestimate the adherence gap between Latino and white patients. Fourth, our findings may not generalize to other settings with less access to interpreters or with different Latino groups. Fifth, we were unable to systematically measure interpreter use. Sixth, we had no objective measure of physician Spanish proficiency. Finally, we had no direct measure of medication adherence beyond dispensing.

Conclusions

Our study among insured patients suggests that more needs to be done to improve adherence to newly prescribed medications among Latino patients at all levels of English proficiency. Regardless of their language ability, Latino patients had much lower rates of initiating newly prescribed diabetes medications, suggesting that early interventions should be evaluated. Given the lack of evidence of substantive differences in adherence between language-concordant vs language-discordant LEP patients, these interventions need to go beyond simply addressing language barriers. More research on nonadherence in this vulnerable subgroup is needed. Finally, interventions both within and outside the patient-physician interaction will likely be required to adequately support English- and Spanish-speaking Latino patients with the challenges of diabetes medication adherence.

Back to top
Article Information

Corresponding Author: Alicia Fernández, MD, Division of General Internal Medicine, San Francisco General Hospital, 1001 Potrero Ave, UCSF Box 1364, Bldg 10, Ward 13, San Francisco, CA 94143 (alicia.fernandez@ucsf.edu).

Published Online: January 23, 2017. doi:10.1001/jamainternmed.2016.8653

Author Contributions: Dr Fernández had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Concept and design: Fernández, Moffet, Karter.

Acquisition, analysis, or interpretation of data: All authors.

Drafting of the manuscript: Fernández, Quan.

Critical revision of the manuscript for important intellectual content: Quan, Moffet, Parker, Schillinger, Karter.

Statistical analysis: Quan, Parker, Karter.

Obtained funding: Fernández, Moffet, Schillinger, Karter.

Administrative, technical, or material support: Fernández, Moffet, Karter.

Supervision: Fernández, Karter.

Conflict of Interest Disclosures: None reported.

Funding/Support: Funding for this study was provided by grant R01 DK090272 from the National Institute of Diabetes and Digestive and Kidney Diseases. This study was an ancillary study of DISTANCE, which provided additional support for data collection and analysis via grants DK081796, DK080726, and DK065664 from the National Institute of Diabetes and Digestive and Kidney Diseases and grant R01 HD046113 from the Eunice Kennedy Shriver National Institute of Child Health and Human Development. Drs Karter and Schillinger’s work was also supported by grant P30DK092924 from the Center for Diabetes Translational Research. Dr Fernández was additionally partially supported by grant K24 DK102057 from the National Institute of Diabetes and Digestive and Kidney Diseases.

Role of the Funder/Sponsor: The funders had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.

Additional Contributions: We thank Eric Vittinghoff, PhD, of the University of California, San Francisco, School of Medicine, for his input on the development of the study design. He received compensation.

References
1.
Centers for Disease Control and Prevention. Age-adjusted rates of diagnosed diabetes per 100 civilian, non-institutionalized population, by Hispanic origin, United States, 1997-2014. https://www.cdc.gov/diabetes/statistics/prev/national/figbyhispanic.htm. Accessed December 27, 2016.
2.
Haynes  RB, Ackloo  E, Sahota  N, McDonald  HP, Yao  X.  Interventions for enhancing medication adherence.  Cochrane Database Syst Rev. 2008;(2):CD000011.PubMedGoogle Scholar
3.
Osterberg  L, Blaschke  T.  Adherence to medication.  N Engl J Med. 2005;353(5):487-497.PubMedGoogle ScholarCrossref
4.
Rodriguez  F, Cannon  CP, Steg  PG,  et al; REACH Registry Investigators.  Predictors of long-term adherence to evidence-based cardiovascular disease medications in outpatients with stable atherothrombotic disease: findings from the REACH Registry.  Clin Cardiol. 2013;36(12):721-727.PubMedGoogle ScholarCrossref
5.
Palacio  AM, Uribe  C, Hazel-Fernandez  L,  et al.  Can phone-based motivational interviewing improve medication adherence to antiplatelet medications after a coronary stent among racial minorities? a randomized trial.  J Gen Intern Med. 2015;30(4):469-475.PubMedGoogle ScholarCrossref
6.
van Servellen  G, Lombardi  E.  Supportive relationships and medication adherence in HIV-infected, low-income Latinos.  West J Nurs Res. 2005;27(8):1023-1039.PubMedGoogle ScholarCrossref
7.
Ryan  C. Language Use in the United States, 2011: American Community Survey Reports. Washington, DC: US Census Bureau; 2013.
8.
Fernandez  A, Schillinger  D, Warton  EM,  et al.  Language barriers, physician-patient language concordance, and glycemic control among insured Latinos with diabetes: the Diabetes Study of Northern California (DISTANCE).  J Gen Intern Med. 2011;26(2):170-176.PubMedGoogle ScholarCrossref
9.
Clay  BJ, Halasyamani  L, Stucky  ER, Greenwald  JL, Williams  MV.  Results of a medication reconciliation survey from the 2006 Society of Hospital Medicine national meeting.  J Hosp Med. 2008;3(6):465-472.PubMedGoogle ScholarCrossref
10.
Karter  AJ, Subramanian  U, Saha  C,  et al.  Barriers to insulin initiation: the translating research into action for diabetes insulin starts project.  Diabetes Care. 2010;33(4):733-735.PubMedGoogle ScholarCrossref
11.
Rivadeneyra  R, Elderkin-Thompson  V, Silver  RC, Waitzkin  H.  Patient centeredness in medical encounters requiring an interpreter.  Am J Med. 2000;108(6):470-474.PubMedGoogle ScholarCrossref
12.
Schenker  Y, Karter  AJ, Schillinger  D,  et al.  The impact of limited English proficiency and physician language concordance on reports of clinical interactions among patients with diabetes: the DISTANCE Study.  Patient Educ Couns. 2010;81(2):222-228.PubMedGoogle ScholarCrossref
13.
Karliner  LS, Pérez-Stable  EJ, Gildengorin  G.  The language divide: the importance of training in the use of interpreters for outpatient practice.  J Gen Intern Med. 2004;19(2):175-183.PubMedGoogle ScholarCrossref
14.
Karter  AJ, Ferrara  A, Liu  JY, Moffet  HH, Ackerson  LM, Selby  JV.  Ethnic disparities in diabetic complications in an insured population.  JAMA. 2002;287(19):2519-2527.PubMedGoogle ScholarCrossref
15.
Karter  AJ, Schillinger  D, Adams  AS,  et al.  Elevated rates of diabetes in Pacific Islanders and Asian subgroups: the Diabetes Study of Northern California (DISTANCE).  Diabetes Care. 2013;36(3):574-579.PubMedGoogle ScholarCrossref
16.
Rosenthal  A, Wang  F, Schillinger  D, Pérez Stable  EJ, Fernandez  A.  Accuracy of physician self-report of Spanish language proficiency.  J Immigr Minor Health. 2011;13(2):239-243.PubMedGoogle ScholarCrossref
17.
Parker  MM, Moffet  HH, Adams  A, Karter  AJ.  An algorithm to identify medication nonpersistence using electronic pharmacy databases.  J Am Med Inform Assoc. 2015;22(5):957-961.PubMedGoogle ScholarCrossref
18.
Karter  AJ, Parker  MM, Moffet  HH, Ahmed  AT, Schmittdiel  JA, Selby  JV.  New prescription medication gaps: a comprehensive measure of adherence to new prescriptions.  Health Serv Res. 2009;44(5, pt 1):1640-1661.PubMedGoogle ScholarCrossref
19.
Raebel  MA, Schmittdiel  J, Karter  AJ, Konieczny  JL, Steiner  JF.  Standardizing terminology and definitions of medication adherence and persistence in research employing electronic databases.  Med Care. 2013;51(8)(suppl 3):S11-S21.PubMedGoogle ScholarCrossref
20.
Greenland  S, Pearl  J, Robins  JM.  Causal diagrams for epidemiologic research.  Epidemiology. 1999;10(1):37-48.PubMedGoogle ScholarCrossref
21.
Robins  JM.  Data, design, and background knowledge in etiologic inference.  Epidemiology. 2001;12(3):313-320.PubMedGoogle ScholarCrossref
22.
Hernán  MA, Hernández-Díaz  S, Werler  MM, Mitchell  AA.  Causal knowledge as a prerequisite for confounding evaluation: an application to birth defects epidemiology.  Am J Epidemiol. 2002;155(2):176-184.PubMedGoogle ScholarCrossref
23.
Pearl  J.  Causality: Models, Reasoning, and Inference. 2nd ed. New York, NY: Cambridge University Press; 2009.
24.
Pearl  J.  Causal diagrams for empirical research.  Biometrika. 1995;82(4):669-688. doi:10.1093/biomet/82.4.669Google ScholarCrossref
25.
Laraia  BA, Karter  AJ, Warton  EM, Schillinger  D, Moffet  HH, Adler  N.  Place matters: neighborhood deprivation and cardiometabolic risk factors in the Diabetes Study of Northern California (DISTANCE).  Soc Sci Med. 2012;74(7):1082-1090.PubMedGoogle ScholarCrossref
26.
Alsabbagh  MH, Lemstra  M, Eurich  D,  et al.  Socioeconomic status and nonadherence to antihypertensive drugs: a systematic review and meta-analysis.  Value Health. 2014;17(2):288-296.PubMedGoogle ScholarCrossref
27.
Zou  G.  A modified poisson regression approach to prospective studies with binary data.  Am J Epidemiol. 2004;159(7):702-706.PubMedGoogle ScholarCrossref
28.
Heisler  M, Faul  JD, Hayward  RA, Langa  KM, Blaum  C, Weir  D.  Mechanisms for racial and ethnic disparities in glycemic control in middle-aged and older Americans in the health and retirement study.  Arch Intern Med. 2007;167(17):1853-1860.PubMedGoogle ScholarCrossref
29.
Traylor  AH, Schmittdiel  JA, Uratsu  CS, Mangione  CM, Subramanian  U.  Adherence to cardiovascular disease medications: does patient-provider race/ethnicity and language concordance matter?  J Gen Intern Med. 2010;25(11):1172-1177.PubMedGoogle ScholarCrossref
30.
Hu  D, Juarez  DT, Yeboah  M, Castillo  TP.  Interventions to increase medication adherence in African-American and Latino populations: a literature review.  Hawaii J Med Public Health. 2014;73(1):11-18.PubMedGoogle Scholar
31.
McQuaid  EL, Everhart  RS, Seifer  R,  et al.  Medication adherence among Latino and non-Latino white children with asthma.  Pediatrics. 2012;129(6):e1404-e1410.PubMedGoogle ScholarCrossref
32.
Compton  S, Haack  S, Phillips  CR.  Identification of barriers to medication adherence in a Latino population.  Res Social Adm Pharm. 2010;6(4):365-371.PubMedGoogle ScholarCrossref
33.
Lyles  CR, Seligman  HK, Parker  MM,  et al.  Financial strain and medication adherence among diabetes patients in an integrated health care delivery system: the Diabetes Study of Northern California (DISTANCE).  Health Serv Res. 2016;51(2):610-624.PubMedGoogle ScholarCrossref
34.
Gazmararian  JA, Kripalani  S, Miller  MJ, Echt  KV, Ren  J, Rask  K.  Factors associated with medication refill adherence in cardiovascular-related diseases: a focus on health literacy.  J Gen Intern Med. 2006;21(12):1215-1221.PubMedGoogle ScholarCrossref
35.
Conn  VS, Enriquez  M, Ruppar  TM, Chan  KC.  Cultural relevance in medication adherence interventions with underrepresented adults: systematic review and meta-analysis of outcomes.  Prev Med. 2014;69:239-247.PubMedGoogle ScholarCrossref
36.
Wilson  E, Chen  AH, Grumbach  K, Wang  F, Fernandez  A.  Effects of limited English proficiency and physician language on health care comprehension.  J Gen Intern Med. 2005;20(9):800-806.PubMedGoogle ScholarCrossref
37.
Wassertheil-Smoller  S, Arredondo  EM, Cai  J,  et al.  Depression, anxiety, antidepressant use, and cardiovascular disease among Hispanic men and women of different national backgrounds: results from the Hispanic Community Health Study/Study of Latinos.  Ann Epidemiol. 2014;24(11):822-830.PubMedGoogle ScholarCrossref
38.
Gany  F, Leng  J, Shapiro  E,  et al.  Patient satisfaction with different interpreting methods: a randomized controlled trial.  J Gen Intern Med. 2007;22(suppl 2):312-318.PubMedGoogle ScholarCrossref
39.
Slean  GR, Jacobs  EA, Lahiff  M, Fisher  L, Fernandez  A.  Aspects of culturally competent care are associated with less emotional burden among patients with diabetes.  Med Care. 2012;50(9)(suppl 2):S69-S73.PubMedGoogle ScholarCrossref
40.
Meyers  K, Tang  G, Fernandez  A.  Responding to the language challenge: Kaiser Permanente’s approach.  Perm J. 2009;13(3):77-83.PubMedGoogle ScholarCrossref
41.
Parker  MM, Fernandez  A, Moffet  HH, Grant  RW. Torreblanca  A, Karter  AJ.  Association of patient-physician language concordance and glycemic control for limited-English proficiency Latinos with type 2 diabetes [published online January 23, 2017].  JAMA Intern Med. doi:10.1001/jamainternmed.2016.8648Google Scholar
42.
Gee  GC, Walsemann  KM, Takeuchi  DT.  English proficiency and language preference: testing the equivalence of two measures.  Am J Public Health. 2010;100(3):563-569.PubMedGoogle ScholarCrossref
×