- Email: [email protected]
Associations between Joblessness and Oral Anti-diabetic Medication Adherence in US Diabetic Working-age Adults
ORIGINAL RESEARCH ARTICLE
Associations between Joblessness and Oral Anti-diabetic Medication Adherence in US Diabetic Working-age Adults Mary L. Davis-Ajami, PhDa, Milap C. Nahata, PharmDb, Gregory Reardon, RPh, PhDc, Eric E. Seiber, PhDd, Rajesh Balkrishnan, PhDe a
DEPARTMENT OF ORGANIZATIONAL SYSTEMS AND ADULT HEALTH, UNIVERSITY OF MARYLAND SCHOOL OF NURSING, BALTIMORE; bCOLLEGE OF PHARMACY AND DEPARTMENTS OF INTERNAL MEDICINE AND PEDIATRICS, COLLEGE OF MEDICINE, THE OHIO STATE UNIVERSITY, COLUMBUS; cINFORMAGENICS, LLC, WORTHINGTON, OHIO; dDIVISION OF HEALTH SERVICES MANAGEMENT AND POLICY, THE OHIO STATE UNIVERSITY COLLEGE OF PUBLIC HEALTH, COLUMBUS; AND eDEPARTMENT OF RESEARCH AND EDUCATION, UNIVERSITY OF MICHIGAN CENTER FOR GLOBAL HEALTH, DEPARTMENT OF CLINICAL, SOCIAL & ADMINISTRATIVE SCIENCES, DEPARTMENT OF HEALTH MANAGEMENT AND POLICY, UNIVERSITY OF MICHIGAN COLLEGE OF PHARMACY, ANN ARBOR
-
ABSTRACT
O B J E C T I V E : To assess potential associations between joblessness and oral anti-diabetic (OAD) medication adherence in US diabetic working-age adults. S T U D Y D E S I G N : A retrospective longitudinal panel design used pooled 2001-2007 Medical Expenditure Panel Survey (MEPS) data forming a nationally representative sample of diabetic individuals, ages 24-59 years. Pregnancy, seasonal job status, retired persons, a student designation, and those prescribed insulin were excluded. Adherence was measured using the proportion of days covered (PDC). A PDC $0.80 was classified as adherent. Descriptive statistics and multivariate regression analysis accounting for the MEPS’ complex survey design were conducted. R E S U L T S : There were 2256 individuals (means: age 48.3 years [SD 8.15], body mass index 31.1 [SD 0.30], Charlson Comorbidity Index 0.37 [SD 0.79]) who met study criteria. Thirty-four percent were jobless at the first interview round and 29% remained jobless all 5 interview rounds during the 2-year panel period. Reasons cited for joblessness included: waiting to start a new job (73%) and unable to work due to illness or disability (20%). Negligible proportions cited staying home to care for family members or maternity leave as reasons for joblessness. Proportionately, more individuals were nonadherent (55%, SE 0.006). Joblessness was associated with a 16% significant reduction in the PDC (b 15.9, P < 0.001), and a 25% less likelihood of OAD medication adherence compared with those employed (odds ratio 0.75; 95% confidence interval, 0.64-0.90, P ¼ 0.002), while holding all other variables constant.
See funding, conflict of interest, and authorship disclosures at the end of this article. Ó 2012 Elsevier Inc. All rights reserved Health Outcomes Research in Medicine (2012) 3, e139-e151
e140
Oral Anti-diabetic Medication Adherence and Joblessness in the US
C O N C L U S I O N S : The results indicate that jobless working-age individuals with diabetes were significantly less likely to adhere to OAD medication than employed individuals. K E Y W O R D S : Diabetes; Employment; Job status; Joblessness; Medication adherence; Medication use
Employment
offers individuals the means to generate income, and often provides benefits such as employer-sponsored insurance. As such, employment status may give additional context for health care affordability. Jobless, nondisabled, work-eligible, working-age individuals vulnerable to noninclusion in Medicare and Medicaid, those facing Consolidated Omnibus Budget Reconciliation Act (COBRA) premiums, or those uninsured, face challenging health care choices. Separate from health insurance coverage, joblessness, with its associated income loss, income poverty, and material deprivation often constrains resource availability, potentially impacting personal expenditures.1 Personal resource reduction may influence spending, purchasing, or consumption patterns for chronic disease management, including pharmacologic intervention. However, few studies examine associations between joblessness and chronic disease in working-age individuals. Joblessness, in individuals with chronic disease, may provide additional context explaining suboptimal medication adherence.
The World Health Organization suggests that chronic disease represents 46% of the global disease burden.2 Individuals with chronic disease experience health expenditures twice those without a chronic condition.3 United Nations Resolution 61/225 describes diabetes as “a chronic, debilitating, and costly disease” with devastating human, social, and economic consequences. Globally, diabetes prevalence continues to escalate, and US data report diabetes as the seventh leading cause of death, and the eighth most costly disease to treat, with 2007 total health care costs estimated at $174 billion.4 Type 2 diabetes shows increased out-of-pocket medical and medication costs compared with nondiabetics.5 Increased medication costs, including out-of-pocket expenses, influence medication nonadherence.6 Diabetes medication adherence remains suboptimal, with reported overall medication nonadherence estimated as 24.8%. Average diabetes medication adherence across 23 studies was reported as 67.5% (95% confidence interval [CI], 58.5%-75.8%).6 Pharmacologic nonadherence shows associations with disease progression, complications, hospitalization, premature disability, and mortality.6,7 Jobless individuals with diabetes may face financial constraints affecting relationships between medication underuse and poor diabetic health outcomes. Theoretical constructs suggest that contextual factors, medication regime complexity, clinical factors, and the health system influence cost-related medication underuse.8-11 Moreover, studies report significant associations between increased medication cost, as well as lower net worth and medication nonuse.11 Although lower net worth is associated with medication nonuse, few studies assess joblessness or unemployment in relation to medication adherence among US working-age diabetics. For working-age individuals, employment generally establishes a main source of income. The literature about diabetes and employment spanned topics related to worker productivity, eligibility for employment, and job suitability,12-14 as well as workplace discrimination, accommodations, and safety.15 Many studies used a payer perspective to assess workplace disease management, strategic cost minimization, absenteeism, presenteeism, work disability, or health insurance premiums.16 Research assessing diabetes and unemployment examined indirect costs associated with workforce nonparticipation, or the effects of disease severity or diabetes-related complications on retirement and disability.17,18 This literature reported significant associations between socioeconomic status and medication adherence, with incomes #US $46,000, food insufficiency, hunger, poverty, low literacy, substandard housing, and homelessness as factors in nonadherence.19 Associations between US labor force nonparticipation and medication adherence in diabetes remained unexplored.
Health Outcomes Research in Medicine - Vol. 3 / No. 3 / August 2012
This study focused on one contextual factordjob statusdspecifically, joblessness and its relationship to oral antidiabetic medication use in adult working-age individuals diagnosed with diabetes. We hypothesized that joblessness in working-age individuals with diabetes would show significantly lower oral antidiabetic (OAD) prescription medication adherence levels as measured by the proportion of days covered (PDC) per interview round, as well as a lower likelihood of medication adherence below a PDC 0.80 threshold compared with those employed.
RESEARCH DESIGN AND METHODS This study used a population-based retrospective longitudinal design with public domain deidentified pooled years 2001-2007 data from complete panels from the Household Component of the Medical Expenditure Panel Survey (MEPS). As a national probability survey sponsored by the Agency for Healthcare Research and Quality, MEPS provided comprehensive, nationally representative, valid, and accurate individual-level health care data with detailed employment and prescription medication information.20 The survey uses an overlapping panel design, surveying sample households during 5 separate interviews over 2 1/2 years. This study used MEPS over other publicly accessible health care databases for its demonstrated use in prior diabetes research, accessible employment information, and validated pharmacy prescription medication information. MEPS data have been used to assess diabetes pharmacologic treatment trends,21 diabetes disease management,22 direct medical costs,23 and diabetes utilization and expenditures,24 among other uses. Moreover, the survey elicits information about ambulatory prescription medication purchases or otherwise procured medicines such as samples during each interview round. Most self-reported prescription medication use gets linked to, for example, pharmacy claims data giving information about date filled, National Drug Code, medication name, strength, and quantity.25,26 Because the study used publicly accessible, de-identified secondary preexisting data, it was granted an institutional review board exemption from The Ohio State University. Full-year consolidated files provided administrative variables, employment status, and demographic information. Medication condition files were accessed to assess comorbidity and diabetes-related complications. Prescribed medicines files provided medication adherence data. Eligible respondents aged 24-59 years with any International Classification of Diseases, 9th Revision, Clinical Modification (ICD-9-CM) 3-digit code “250” for diabetes, from complete 2-year MEPS panels, were included. Publicly downloadable MEPS data provided only 3-digit ICD-9-CM codes, limiting finer diagnostic gradations. Pregnant females, those with a seasonal job status, those who were retired, or those with a student designation, as well as any individuals prescribed insulin, were excluded. Additionally, persons traditionally excluded from US Bureau of Labor Force employment data: persons under age 16 years, all persons confined to nursing homes or prisons, and active duty Armed Forces members, were excluded. The age upper bound was set at 59 years to help control for early retirement, and 24 years lower bound to control for workforce nonparticipation due to education. Pregnancy was excluded to control for gestational diabetes, and individuals prescribed insulin were excluded to control for type I diabetes and disease severity, as well as the difficulty measuring complex pharmaceutical regimens with injectable medications. The outcome variable, OAD pharmacotherapeutics adherence, was calculated for complete 2-year panel periods using the PDC with a predetermined $0.80 cut-point. The PDC represented an objective, indirect method for estimating medication adherence appropriate for claims data linked to administrative data.27 The PDC showed reasonable predictive value for subsequent hospitalization in diabetes research.28 A 0.80 cut-point showed a fair balance between sensitivity and specificity for research assessing medication adherence using administrative data,29 as well as for research comparing adherence rates for individuals with chronic conditions, including diabetes.30 Moreover, MEPS data were better suited to calculating medication adherence by the PDC than by the medication possession ratio because the MEPS complex survey overlapping panel design precluded using an index date, among other reasons. MEPS prescribed medicines file does not
e141
e142
Oral Anti-diabetic Medication Adherence and Joblessness in the US
provide a days supply variable. Days supply was calculated using the OAD unique identifier, strength, dosing unit, beginning and ending date for each prescription, and number of pills supplied. A pharmacotherapy product master list was created using the unique drug code (National Drug Code) for OAD prescriptions for the following drug classes: sulfonylurea, meglitinide, biguanide, thiazolidinedione, alpha-glycosidase inhibitors, and DPP-4 inhibitors. The primary variable of interest, joblessness, was assessed at each interview round using the MEPS EMPST[round] variable. Individuals reporting “they were not employed with no job to return to” were classified as jobless.31 Individuals were classified as employed when respondents reported having a job, being employed, or having a job to return to. The EMPST[round] variable was collapsed into a zero or one (0 ¼ yes, employed; 1 ¼ no, jobless) categorical variable, with employed persons serving as the reference group. Demographic covariates included: age, sex, educational attainment, ethnicity, race, marital status, health insurance status, total income, total prescription, and total prescription medication expense. Geographical region was used only in descriptive analysis. Clinical covariates were classified as comorbidities or complications. The enhanced Charlson Comorbidity Index (CCI) was used as a proxy for disease-burden risk adjustment. Descriptive analysis used a body mass index (BMI) continuous variable, and multivariate analysis used categorical variables created from standard US BMI designations for underweight, normal weight, overweight, obese, and moribund obese. Medical conditions file ICD-9-CM codes identified first interview panel round complication status. Categorical variables were created for specific eye, neurological, macrovascular, microvascular, and renal complications commonly associated with diabetes and further collapsed into a zero or one (0 ¼ none, 1 ¼ one or more complication) categorical variable.32 Descriptive analyses were conducted using a pooled person-level analytical dataset to report means for continuous variables, independent sample t-tests for differences between continuous variables, and chisquared tests for differences between categorical variables. Multicollinearity was assessed using procedures appropriate for complex survey data.33 A series of regressions were estimated using the variables representing individual employment status, insurance status, and spouse employment status.33 All variance inflation factor calculations were well below 10, suggesting no problem with multicollinearity. Separately, interactions between individual employment status and insurance status, as well as individual and spousal employment status, were nonsignificant and not reported here. Multivariate analyses were conducted on a pooled analytical panel dataset. Regression diagnostics showed a right skew for the PDC outcome variable. Estimates were robust to generalized linear models (GLM) with a log link versus a natural log transformation.34 Multivariate GLM with a log link assessed overall PDC level, and logistic regression assessed medication adherence defined by the 0.80 threshold. The GLM results were reported as log-normal.34,35 Statistical analyses were conducted using Stata, version 11 (StataCorp LP, College Station, TX) with appropriate complex survey estimation commands. The linearized Taylor series expansion method estimated variance.
RESULTS Of the 240,409 individuals from MEPS complete panels for pooled years 2001-2007, 2256 subjects (means: age 48.3 years [SD 8.15], BMI 31.5 [SD 0.30], and CCI 0.37 [SD 0.79]) were included in the study. Mean total income approximated $11,967 [SD 9632] and mean total prescription medication expense approximated $2556 [SD 3177] (Table 1: summary characteristics). Approximately 34% were jobless during the first
Health Outcomes Research in Medicine - Vol. 3 / No. 3 / August 2012
e143
- T A B L E 1 : Summary Characteristics in a Study Assessing Associations between Joblessness and OAD Medication Adherence (Using the PDC) in a Cohort of Working-age Adults All Individuals n ¼ 2256
Age, years Mean 48.29, SE 0.09 24-40 41-59 Sex Male Female Marital status Married Unmarried Educational degree No degree High school College Graduate school Ethnicity Hispanic Other Health insurance status Private health insurance Public health insurance Uninsured Poverty category Poor, near poor, or low Middle income High income Race White Black Other Region Northeast Midwest South West Total income, mean $11,797, SE 98.91 $0-$19,000 $20,000-$29,999 $$30,000 Total prescription med expense, mean $2556, SE 28.11 $0-$1499 $1500-$2999 $$3000 Body mass index category, mean 31, SE 0.30 Underweight Normal Overweight Obese/moribund obese Charlson Comorbidity Index, mean 0.37, SE 0.79 0 1 $2
Employed Individuals n ¼ 1488
Jobless Individuals n ¼ 767
n (%)
SE
n (%)
SE
n (%)
SE
384 (17%) 1872 (83%)
0.003 0.003
268 (18%) 1220 (82%)
0.005 0.005
107 (14%) 660 (86%)
0.006† 0.006
1105 (49%) 1151 (51%)
0.004 0.004
833 (56%) 655 (44%)
0.006 0.006
268 (35%) 767 (65%)
0.008‡ 0.008
1421 (63%) 835 (37%)
0.004 0.004
1011 (68%) 477 (32%)
0.006 0.006
414 (54%) 353 (46%)
0.008‡ 0.008
474 (21%) 1218 (54%) 271 (12%) 293 (13%)
0.004 0.004 0.003 0.003
193 (13%) 803 (54%) 238 (16%) 254 (17%)
0.004 0.007 0.005 0.005
261 (34%) 407 (53%) 46 (06%) 53 (07%)
0.007‡ 0.008 0.004 0.004
360 (16%) 1896 (84%)
0.003 0.003
208 (14%) 1280 (86%)
0.004 0.004
138 (18%) 629 (82%)
0.005‡ 0.005
1557 (69%) 474 (21%) 225 (10%)
0.003 0.003 0.003
1280 (86%) 60 (04%) 148 (09%)
0.002 0.004 0.004
276 (36%) 407 (53%) 84 (11%)
0.004‡ 0.008 0.008
767 (34%) 699 (31%) 790 (35%)
0.004 0.005 0.005
283 (19%) 520 (35%) 685 (46%)
0.005 0.007 0.007
499 (65%) 169 (22%) 99 (13%)
0.008‡ 0.007 0.006
1669 (74%) 406 (18%) 181 (08%)
0.004 0.004 0.002
1131 (76%) 238 (16%) 119 (08%)
0.005 0.004 0.004
537 (70%) 169 (22%) 61 (08%)
0.007† 0.006 0.004
338 (15%) 384 (17%) 970 (43%) 564 (25%)
0.003 0.003 0.004 0.004
253 (17%) 327 (22%) 565 (38%) 343 (23%)
0.006 0.006 0.006 0.006
153 (20%) 138 (18%) 314 (41%) 162 (21%)
0.007* 0.006 0.008 0.007
1715 (76%) 429 (19%) 499 (05%)
0.004 0.004 0.002
714 (48%) 565 (38%) 209 (14%)
0.005 0.005 0.007
705 (92%) 54 (07%) 08 (01%)
0.008‡ 0.007 0.008
1308 (58%) 406 (18%) 542 (24%)
0.005 0.004 0.004
982 (66%) 268 (18%) 238 (16%)
0.006 0.005 0.005
330 (43%) 153 (20%) 284 (37%)
0.008‡ 0.007 0.008
90 (04%) 248 (11%) 587 (26%) 1331 (59%)
0.002 0.003 0.005 0.005
45 (03%) 179 (12%) 417 (28%) 847 (57%)
0.002 0.004 0.006 0.005
31 (04%) 84 (11%) 184 (24%) 486 (61%)
0.003 0.005 0.007 0.007
1692 (75%) 406 (18%) 158 (07%)
0.004 0.004 0.002
1235 (83%) 208 (14%) 45 (03%)
0.005 0.004 0.002
434 (57%) 207 (27%) 126 (16%)
0.005‡ 0.004 0.002
e144
Oral Anti-diabetic Medication Adherence and Joblessness in the US
- T A B L E 1 (continued ): Summary Characteristics in a Study Assessing Associations between Joblessness and OAD Medication Adherence (Using the PDC) in a Cohort of Working-age Adults All Individuals n ¼ 2256
Diabetes-related complications Yes No
Employed Individuals n ¼ 1488
Jobless Individuals n ¼ 767
n (%)
SE
n (%)
SE
n (%)
SE
948 (42%) 1308 (58%)
0.003 0.003
1220 (82%) 268 (18%)
0.005 0.005
522 (68%) 245 (32%)
0.007‡ 0.007
OAD ¼ oral anti-diabetic; PDC ¼ proportion of days covered. Data from Medical Expenditure Panel Survey (MEPS) pooled panel years 2001-2007. Significantly different from employed individuals. *P < 0.05. † P < 0.01. ‡ P < 0.001.
interview round, with 29% remaining jobless through all 5 interview rounds. For those giving reasons for being jobless, the majority (74%) were waiting to start a new job, followed by those unable to work due to illness or disability (20%). Compared with those employed, jobless individuals were proportionately older, female, less
- T A B L E 2 : Employment Status Summary Characteristics for a Cohort of Working-age Adults from MEPS Pooled Panel Years 2001-2007 Assessing Associations between Joblessness and OAD Medication Adherence as Measured by the PDC
Employment status at first interview round Employed Jobless Job status by interview round 1 Employed No job for 2 interview rounds No job for 3-5 interview rounds Job status by individual interview rounds Employed No job for 1 round No job for 2 interview rounds No job for 3 interview rounds No job for 4 interview rounds No job for 5 interview rounds Reasons cited for no job Could not find work Retired Unable to work (ill/disabled) On temporary layoff Maternity/paternity leave Going to school Taking care of home or family Wanted some time off Waiting to start a new job Other
Proportion
SE
0.66 0.34
0.004 0.004
0.6500-0.6693 0.3289-0.3480
0.604 0.057 0.339
0.005 0.002 0.004
0.5939-0.6142 0.0524-0.0620 0.3289-0.3480
0.604 0.036 0.022 0.034 0.012 0.292
0.005 0.002 0.001 0.002 0.001 0.004
0.5939-0.6142 0.0317-0.0397 0.0188-0.0244 0.0303-0.0381 0.0097-0.0143 0.2833-0.3014
0.015 <0.000 0.201 0.008 <0.000 <0.000 0.034 0.002 0.736 0.004
0.001 >0.001 0.004 >0.001 >0.001 >0.001 >0.001 >0.001 0.004 0.0005
MEPS ¼ Medical Expenditure Panel Survey; OAD ¼ oral anti-diabetic; PDC ¼ proportion of days covered. *Confidence intervals <0.001.
95% Confidence Interval
0.1029-0.0182 * 0.1930-0.2089 0.0003-0.0014 0.000003-0.0001 * 0.0306-0.0376 0.0009-0.0025 0.7265-0.7520 0.0024-0.0044
Health Outcomes Research in Medicine - Vol. 3 / No. 3 / August 2012
e145
- T A B L E 3 : OAD Medication Adherence (Adherent ¼ PDC $0.80) in a Cohort of Working-age Adults with Diabetes Using Data from MEPS Pooled Panel Years 2001-2007 All Individuals
PDC Mean (SD) 0.58 (0.49) Job status Employed Jobless
Age, years 24-40 41-59 Sex Male Female Marital status Married Unmarried Educational degree No degree High school College Graduate school Ethnicity Hispanic Other Health insurance status Private health insurance Public health insurance Uninsured Race White Black Other Region Northeast Midwest South West Total income $0-$19,000 $20,000-$29,999 $$30,000 Total prescription med expense $0-$1499 $1500-$2999 $$3000 Body mass index category Underweight Normal Overweight Obese/moribund obese Charlson Comorbidity Index 0 1 $2
PDC $.8-1
PDC $.8-1
PDC <0.8
n ¼ 2256
OAD Adherent
OAD Nonadherent
n (%) 1489 (66%) 767 (34%)
SE 0.009 0.009
n (%) 676 (30%) 340 (15%) PDC $.8-1 n ¼ 1015
n (%) (835) 37% (406) 18% PDC <0.8 n ¼ 1241
384 (17%) 1872 (83%)
0.003 0.003
n, % 102 (10%) 913 (90%)
SE 0.008 0.005
n, % 161 (13%) 1080 (87%)
SE 0.008 0.006
1105 (49%) 1151 (51%)
0.004 0.004
507 (50%) 507 (50%)
0.010 0.010
645 (52%) 596 (48%)
0.012 0.012
1421 (63%) 835 (37%)
0.004 0.004
640 (63%) 375 (37%)
0.010 0.009
807 (65%) 434 (35%)
0.011 0.011
474 (21%) 1218 (54%) 271 (12%) 293 (13%)
0.004 0.004 0.003 0.003
234 (23%) 528 (52%) 112 (11%) 141 (14%)
0.007 0.010 0.007 0.008
248 (20%) 695 (56%) 137 (11%) 161 (13%)
0.008 0.012 0.008 0.009
360 (16%) 1896 (84%)
0.003 0.003
152 (15%) 863 (85%)
0.006 0.006
211 (17%) 1030 (83%)
0.007 0.008
1557 (69%) 474 (21%) 225 (10%)
0.003 0.003 0.003
721 (71%) 213 (21%) 81 (08%)
0.008 0.008 0.005
881 (71%) 235 (19%) 124 (10%)
0.009 0.008 0.006
1669 (74%) 406 (18%) 181 (08%)
0.004 0.004 0.002
761 (75%) 173 (17%) 82 (08%)
0.008 0.006 0.007
931 (75%) 199 (16%) 111 (09%)
0.010 0.008 0.007
338 (15%) 384 (17%) 970 (43%) 564 (25%)
0.003 0.003 0.004 0.004
182 (18%) 182 (18%) 396 (39%) 255 (25%)
0.008 0.007 0.009 0.008
236 (19%) 261 (21%) 496 (40%) 248 (20%)
0.009* 0.009 0.012 0.008
1715 (76%) 429 (19%) 499 (05%)
0.004 0.004 0.002
711 (70%) 213 (21%) 91 (09%)
0.011 0.009 0.007
819 (66%) 348 (28%) 74 (06%)
0.014* 0.013 0.007
1308 (58%) 406 (18%) 542 (24%)
0.005 0.004 0.004
355 (35%) 244 (24%) 416 (41%)
0.009 0.009 0.010
236 (19%) 223 (18%) 782 (63%)
0.009* 0.009 0.009
90 (04%) 248 (11%) 587 (26%) 1331 (59%)
0.002 0.003 0.005 0.005
20 (02%) 71 (07%) 244 (24%) 681 (66%)
0.003 0.005 0.009 0.009
50 (04%) 112 (09%) 310 (25%) 769 (62%)
0.004 0.007 0.010 0.011
1692 (75%) 406 (18%) 158 (07%)
0.004 0.004 0.002
741 (73%) 193 (19%) 81 (08%)
0.008 0.008 0.008
943 (76%) 223 (18%) 75 (06%)
0.010 0.009 0.001
e146
Oral Anti-diabetic Medication Adherence and Joblessness in the US
- T A B L E 3 (continued ): OAD Medication Adherence (Adherent ¼ PDC $0.80) in a Cohort of Working-age Adults with Diabetes Using Data from MEPS Pooled Panel Years 2001-2007 All Individuals
PDC Mean (SD) 0.58 (0.49) Diabetes-related complications Yes No
PDC $.8-1
PDC $.8-1
PDC <0.8
n ¼ 2256
OAD Adherent
OAD Nonadherent
948 (42%) 1308 (58%)
0.003 0.003
792 (78%) 223 (22%)
0.008 0.008
1005 (81%) 236 (19%)
MEPS ¼ Medical Expenditure Panel Survey; OAD ¼ oral anti-diabetic; PDC ¼ proportion of days covered. Significantly different from adherent individuals. *P < 0.05.
educated, in the lowest income category, and had greater BMI. Although proportionately more females were jobless than males, for women citing reasons for joblessness, only 3% reported not working due to taking care of home or family, and none (0%) reported begin on maternity leave (Table 2: job status summary characteristics). Of the approximately 34% categorized as poor, near poor, or low income, approximately 22% were jobless. Ten percent of the study cohort reported no health insurance. Jobless individuals showed slightly greater proportions uninsured when compared with employed respondents. Jobless individuals with health insurance were predominately covered by public insurance. Summary descriptive comorbidity and complication characteristics showed significant differences between employed and jobless individuals for the CCI and diabetes-related complications. Overall, approximately 55% were nonadherent (SE 0.006). Truncated mean PDC was 0.58 (SD 0.49). Of those found nonadherent, 18% were jobless, versus 37% employed. Proportionately, the younger age group, males, those with a high school education, those in the middle total income category, and those with the highest total prescription medication expense were proportionately more nonadherent than those aged 41-59 years. Individuals living in the western region of the US, as well as individuals who were obese/ moribund obese, were proportionately more adherent (Table 3, medication adherence). Multivariate GLM with a log link regression evaluated associations between joblessness and a PDC continuous measure. The transformed log-normal GLM regression showed jobless individuals with a 16% lower PDC level (b ¼ 0.157, P < 0.01) compared with employed individuals, when holding all other variables constant. The covariates age, educational attainment, race, total income, total prescription medication expense, CCI, and the presence of one or more diabetic complication(s) also showed statistically significant differences (Table 4: GLM regression). Logistic regression evaluated associations between joblessness and adherence (PDC $0.80) using a dichotomous measure. Logistic regression showed that the odds of OAD medication adherence was 25% significantly lower for jobless individuals compared with those employed (OR 0.75; 95% CI, 0.64-0.90, P ¼ 0.002) while holding all other variables constant. Other parameters significantly affecting the likelihood of OAD medication adherence included: age, female sex, educational attainment, ethnicity, total income, total prescription medication expenses, and the CCI. For individuals aged 41-59 years, every 1-year increase in age increased the likelihood of being adherent 45% (OR 1.45; 95% CI, 1.17-1.74, P < 0.001). Females were 17% less likely than males to be adherent (OR 0.83; 95% CI, 0.72-0.95, P ¼ 0.008), and those with high school or college degrees were less likely than those without degrees to be adherent. Those whose total incomes ranged from $20,000-$29,999 were 35% less likely to be adherent than those with lower incomes (OR 0.65; 95% CI, 0.54-0.78, P # 0.001). Individuals with lower total prescription medication expenses were less likely to be adherent than those with the highest total prescription medication expenses. Every one-unit increase in the CCI resulted in a 15% reduction in the likelihood of adherence (OR 0.85; 95% CI, 0.78-0.99, P < 0.001) (Table 4: logistic regression).
0.009 0.009
Health Outcomes Research in Medicine - Vol. 3 / No. 3 / August 2012
e147
- T A B L E 4 : Estimates of Associations between Joblessness and OAD Medication Adherence (Measured by PDC) in a Cohort of US Working-age Adults from MEPS Pooled Panel Years 2001-2007 GLM Log-normal b Primary variable of interest Jobless individual1 Demographic characteristics Age2 Sex3 Marital status4 Education: degree5 High school College Graduate school Ethnicity6 Hispanic Race7 Black Other Economic-related factors Health insurance coverage8 Private health insurance Public health insurance Total income9 $20,000-$29,999 $$30,000 Total prescription medication expenses10 $1500-$2999 $$3000 Comorbidity/complications of diabetes Charlson comorbidity index Body mass index category11 Normal Overweight Obese/moribund obese Diabetes-related complication12 Constant
SE
Logistic Regression Model P-Value
Odds Ratio
SE
95% CI
0.16
0.054
0.003*
0.75
0.066
0.64-0.90*
0.15 0.04 0.04
0.068 0.041 0.045
0.039* 0.295 0.359
1.45 0.83 0.92
0.147 0.060 0.068
1.17-1.74* 0.72-0.95* 0.77-1.03
0.06 0.32 0.01
0.049 0.107 0.075
0.224 0.005* 0.907
0.78 0.56 0.94
0.061 0.100 0.130
0.69-0.93* 0.41-0.83* 0.74-1.27
0.02
0.056
0.714
0.85
0.41
0.74-1.03
0.02 0.15
0.050 0.074
0.633 0.035*
0.92 1.12
0.82 0.139
0.76-1.07 0.89-1.45
0.02 0.04
0.072 0.077
0.876 0.712
1.12 0.95
0.120 0.107
0.92-1.40 0.76-1.18
0.18 0.04
0.057 0.078
0.001* 0.602
0.65 1.06
0.061 0.158
0.54-0.78* 0.77-1.38
0.24 0.72
0.054 0.052
<0.001* <0.001*
0.65 0.31
0.063 0.027
0.50-0.78* 0.23-0.33*
0.10
0.030
0.001*
0.85
0.037
0.78-0.99*
0.14 0.25 0.17 0.11
0.128 0.117 0.111 0.050
0.242 0.051 0.110 0.017*
1.20 1.28 1.25 0.94
0.270 0.263 0.245 0.078
0.74-1.79 0.81-1.82 0.79-1.72 0.80-1.08
0.155
0.000
0.490
CI ¼ confidence interval; GLM ¼ generalized linear model; MEPS ¼ Medical Expenditure Panel Survey; OAD ¼ oral anti-diabetic; PDC ¼ proportion of days covered. Reference groups: 1Jobless: employed, 2Age: 24-40 years, 3Sex: male, 4Marital status: not married, 5Degree: no degree, 6Ethnicity: not-Hispanic, 7Race: white, 8Health insurance: no insurance coverage, 9Total income: $0-$19,999, 10Total prescription medication expenses: #$1499, 11BMI: underweight, 12Diabetes-related complication: none. *Confidence intervals <0.001.
CONCLUSIONS This research assessing associations between joblessness and OAD medication adherence in adult working-age individuals showed significant proportions reporting no job, striking proportions waiting to start another job, and a troubling chronic dimension to workforce nonparticipation, a finding not fully attributable to disability.
e148
Oral Anti-diabetic Medication Adherence and Joblessness in the US
Individual regression models found significant associations between joblessness and lower PDC levels, as well as an increased likelihood of suboptimal OAD medication adherence as measured by a 0.80 PDC adherence threshold. These findings suggest that joblessness may influence OAD pharmacotherapeutics nonadherence. Joblessness may decrease available personal resources for necessary and discretionary spending, placing financial constraints on personal health care expenditures. Reduced monetary resources, coupled with uncertainty for future economic viability, may influence decisions to reduce medication refills and collaterally affect behavioral dimensions, including changes in consumption patterns.36 The results suggest differing medication refill patterns for jobless individuals compared with those employed. Fully 45% of the study cohort was found adherent using a 0.80 OAD pharmacotherapy threshold. These results were lower than those reported in a 2001-2004 longitudinal study reporting mean medication possession ratio for type 2 diabetes using claims data,37 as well as previously reported mean (percent) overall medication adherence from a meta-analysis of all medication adherence studies for years 1948-1998.6 The divergence may stem from differences in calculating adherence, among other reasons. Our results suggest that a suboptimal percentage of individuals met the 0.80 adherence threshold. This has implications for overall diabetic disease management, such as the known correlation between medication adherence and the ability to reach target glucose control.38 Interestingly, insurance status was not a significant covariate. Although uninsured individuals were proportionately more nonadherent than those covered by private or public health insurance, the regression models showed no significant associations between insurance status and OAD adherence as measured by a 0.80 PDC threshold. This suggests that other factors may influence OAD medication adherence. Moreover, our data came from the MEPS household component of the survey. Due to the characteristics of this database, insurance and employment may not interact like data directly linked to medical claims. In MEPS, although an individual respondent with diabetes may be jobless, others in the household may serve as the main health insurance policy holder. However, there were no significant interactions between jobless individuals and spouses. Higher proportions of employed persons were adherent. Respondents reporting a jobless work status showed a 16% significant reduction in the PDC compared with those employed, and were 24% less likely to meet or exceed the 0.80 PDC adherence threshold. Poor medication adherence has known associations with poor health outcomes and increased health care costs.38 Over time, poor health outcomes may become more apparent for those out of work, particularly for individuals experiencing chronic joblessness.39 Jobless individuals with diabetes may represent a vulnerable segment for lagged long-term suboptimal health outcomes. The World Health Organization suggests that social and economic factors, among others, influence medication adherence.40 Economic factors associated with chronic diabetes disease management may interface with financial constraints during periods of joblessness. This is concerning in light of recent research reporting increased deferrals for health care services to enhance personal financial savings, as well as reports of changing medication consumption patterns as evidenced by skipped dosages and pill splitting stemming from the 2007 economic downturn.36 Moreover, earlier research conducted after the recession in the early 1980s found financial strain stemming from unemployment significantly associated with ill health.41 Our study, with its focus on joblessness and medication adherence, may shed some light on scant previous research suggesting countercyclical relationships in diabetes.42 Loosely coupled linkages between joblessness leading to reduced personal resources, scaled back OAD medication purchases and consumption, or outright nonadherence to OAD medication regimens, and known poor health outcomes for medication nonadherence may, over time, show lagged increases in diabetes mellitus complications or other serious health consequences. This conjecture may be difficult to empirically assess, but is within the realm of possibility. Significant proportions of the study cohort were out of work the entire 2-year MEPS panel time frame, suggesting a troubling longevity to their joblessness. Workforce participation for adults with diabetes and other chronic conditions command the attention of public policy-makers, particularly when prioritizing resource allocation. As a starting position, health care providers and systems need standard processes to identify individuals facing financial pressure and their
Health Outcomes Research in Medicine - Vol. 3 / No. 3 / August 2012
vulnerability to lower medication adherence. Future research could focus on assessing links between financial strain stemming from joblessness and changes in personal choices resulting in differing patterns of medication use, as well as other treatment deferrals. Another area of interest for future research suggested by findings from this study includes a need to further understand the magnitude of chronic joblessness in the diabetic population and possible correlates with macroeconomic indicators. There were several limitations to this research. We could not assess long-term joblessness beyond the 2-year MEPS panel window. The study minimally assessed length of joblessness. The research did not control for the economic cycle or nuances in employment patterns using Bureau of Labor Statistics unemployment data or other macroeconomic indicators, and made no distinction between full-time versus part-time work or job churning. No provision was made to examine lagged effects. The study did not control for changes in diabetic treatment patterns over time, and was subject to limitations common to retrospective database analysis such as selection bias, missing or incomplete information, and recall bias. Adherence was measured indirectly without information about consumption patterns or sample medications. Finally, observational studies may not infer causation. Joblessness in working-age adults with diabetes deserves policy considerations. As a starting position, health care providers and systems need standard processes to identify individuals facing financial pressure and employment vulnerability. Job loss, as well as length of joblessness, could be assessed in tandem with other financial measures assessing income, and affordability and willingness to pay for medications, medical supplies, and services. For individuals identified as needing long-term pharmacotherapy for chronic disease management, future research could assess how the presence or lack of safety-net provisions among workeligible but nonworking individuals with chronic disease impacts disease management cost drivers, complication rates, and mortality. More information about how joblessness affects personal decisions regarding disease management and personal health care spending is needed. Future research could assess differences in health care consumption patterns, particularly medication use patterns, between employed and jobless individuals. Secondly, further research could further identify and evaluate factors associated with nonadherence. Pharmacologic nonadherence risk-assessment tools geared for health professionals need further refinement. Researchers should continue assessing optimal medication adherence thresholds with the view to identifying areas resulting in suboptimal disease management. This research identified significant associations between joblessness and OAD medication adherence as measured by the PDC level and a 0.80 adherence threshold in working-age adults with diabetes. Jobless individuals showed significant reductions in the PDC, as well as a decreased likelihood of OAD medication adherence when compared with those employed. Overall, 55% of the study cohort had OAD medication adherence levels less than a predetermined 0.80 threshold, suggesting suboptimal OAD medication adherence in this nationally representative sample. This study identified joblessness as a significant parameter for diabetes oral medication adherence for working-aged adults with diabetes.
ACKNOWLEDGMENTS Mary L. Davis-Ajami, PhD, affiliated with the University of Maryland School of Nursing, researched data and wrote the manuscript and serves as corresponding author. Milap C. Nahata, PhD, of The Ohio State University College of Pharmacy, reviewed/edited the manuscript. Rajesh Balkrishnan, PhD, of the University of Michigan College of Pharmacy, contributed to the research design, researched data, and reviewed/edited the manuscript. Gregory Reardon, PhD, of Informagenics, researched data and reviewed/edited the manuscript. Eric E. Seiber, PhD, of The Ohio State University College of Public Health, reviewed/edited the manuscript. We thank Helen Leonard, PharmD for assistance in identifying each OAD agent listed in the
e149
e150
Oral Anti-diabetic Medication Adherence and Joblessness in the US
Prescribed Medicines files. We also wish to acknowledge James W. McAuley, PhD, of The Ohio State University College of Pharmacy for his helpful review of the study.
SUPPLEMENTARY DATA Supplementary data associated with this article can be found, in the online version, at doi:10.1016/ j.ehrm.2012.06.001. Corresponding Author: Mary L. Davis-Ajami, PhD, Department of Organizational Systems and Adult Health, University of Maryland School of Nursing, 655 Lombard Street, Room 465D, Baltimore, MD 21201. E-mail address: [email protected] The authors have no relevant conflicts of interest to disclose. There were no funding sources. Mary Lynn Davis-Ajami wrote the first draft of the manuscript. No grants or honoraria were received.
R E F E R E N CE S 1. Färber H. What do we know about job loss in the United States? Evidence from the Displaced Workers Survey, 1984-2004. Federal Reserve Bank of Chicago. Econ Perspect. 2005;29(2):13-28. 2. World Health Organization (WHO). WHO technical report series: Diet, nutrition and the prevention of chronic diseases. Joint WHO/FAO Expert Consultation, WHO. Geneva, Switzerland: World Health Organization. Available at: http://whqlibdoc.who.int/trs/who_trs916.pdf; 2003. Accessed July 6, 2011. 3. Stanton MW, Rutherford MK. The High Concentration of US Health Care Expenditures (Research in Action issue 19 AHRQ Pub. No. 06-0060). Rockville, MD: Agency for Healthcare Research and Quality; 2006. 4. Centers for Disease Control and Prevention (CDC). National Diabetes Fact Sheet: National Estimates and General Information on Diabetes and Prediabetes in the United States, 2011. Atlanta, GA: U.S. Department of Health and Human Services, Centers for Disease Control and Prevention; 2011. 5. Robard HW, Green AJ, Fox KM, Grandy S, SHIELD Study Group. Impact of type 2 diabetes mellitus on prescription medication burden and out-ofpocket healthcare expenses. Diabetes Res Clin Pract. 2010;87(3):360-365. 6. DiMatteo MR. Variations in patients’ adherence to medical recommendations: a quantitative review of 50 years of research. Med Care. 2004;42(3): 200-209. 7. Cramer JA. A systematic review of adherence with medications for diabetes. Diabetes Care. 2004;27(5):1218. 8. Asche C, LaFleur J, Conner C. A review of diabetes treatment adherence and the association with clinical and economic outcomes. Clin Ther. 2011; 33(1):74-109. 9. Piette JD, Heisler M, Horne R, et al. A conceptually based approach to understanding chronically ill patients’ responses to medication cost pressures. Soc Sci Med. 2006;62:846-857. 10. Piette JD. Cost-related medication underuse: a window into patient’s medication-related concerns. Diabetes Spectr. 2009;22(2):77-80. 11. Zivin K, Ratliff S, Heisler MM, Langa K, Piette JD. Factors influencing cost-related nonadherence to medication in older adults: a conceptually based approach. Value Health. 2010;13(4):338-345. 12. Lavigne J, Phelps C, Mushlin A, Lednar W. Reductions in individual work productivity associated with type 2 diabetes mellitus. Pharmacoeconomics. 2003;21(15):1123-1134. 13. Ramsey S, Summers KH, Leong SA, et al. Productivity and medical costs of diabetes in a large employer population. Diabetes Care. 2002;25:23.
14. Tunceli K, Bradley C, Nerenz D, et al. The impact of diabetes on employment and work productivity. Diabetes Care. 2005;11(28):2662-2667. 15. American Diabetes Association. Position statement: diabetes and employment. Diabetes Care. 2011;34:S82-S86. 16. Crawley J, Rizzo J, Haas K. The association of diabetes with job absenteeism costs among obese and morbidly obese workers. J Occup Environ Med. 2008; 50:527-534. 17. Herquelot E, Guéguen A, Bonenfant S, Dray-Spira R. Impact of diabetes on work cessation: data from the GAZEL cohort study. Diabetes Care. 2011;34(6):1344-1349. 18. Vijan S, Hayward RA, Langa KM. The impact of diabetes on workforce participation: Results from a national household sample. Health Serv Res. 2004;39(6 Pt. 1):1653-1669. 19. Peeters B, Van Tongelen I, Boussery K, et al. Factors associated with medication adherence to oral hypoglycemic agents in different ethnic groups suffering from Type 2 diabetes: a systematic literature review and suggestions for future research. Diabet Med. 2011;28:262-275. 20. Olin G, Zuvekas S, Kumar V, Ward P, Williams K Wobus D. 2008. Medicare-MEPS Validation Study: A Comparison of Hospital and Physician Expenditures. Agency for Healthcare Research and Quality Working paper No. 08003, March 2008. Available at: http://gold.ahrq.gov. Accessed April 2, 2012. 21. Sarpong E, Miller GE. Trends in the Pharmaceutical Treatment of Diabetes, 1997 to 2007. Research Findings #30. Rockville, MD: Agency for Healthcare Research and Quality. Available at: http://meps.ahrq.gov/mepsweb/ data_files/publications/rf30/rf30.pdf; 2010. Accessed April 2, 2012. 22. Soni A. Trends in Use and Expenditures for Diabetes among Adults 18 and Older, U.S. Civilian Noninstitutionalized Population, 1996 and 2007. Statistical Brief #304. Rockville, MD: Agency for Healthcare Research and Quality; 2010. 23. Olin G, Machlin SR, Rhoades J. Estimating the Cost of Illness: The Case of Diabetes. Agency for Healthcare Research and Quality Working Paper No. 08001, January 2008. Available at: http://gold.ahrq.gov. Accessed April 2, 2012. 24. Olin G. Proportion and Medical Expenditures of Adults Being Treated for Diabetes, 1996 and 2003. Statistical Brief #146. Rockville, MD: Agency for Healthcare Research and Quality. Available at: http://www.meps.ahrq. gov/mepsweb/data_files/publications/st146/stat146.pdf; 2006. Accessed April 2, 2012.
Health Outcomes Research in Medicine - Vol. 3 / No. 3 / August 2012
e151
R E F E R E N C E S Continued 25. Moeller JF, Stagnitti MN, Horan E, Ward P, Kieffer N, Hock E. Outpatient Prescription Drugs: Data Collection and Editing in the 1996 Medical Expenditure Panel Survey (HC-010A) (MEPS Methodology Report No. 12, AHRQ Pub. No01-0002). Rockville, MD: Agency for Healthcare Research and Quality; 2001. 26. Zodet MW, Hill SC, Miller E. Comparison of Retail Drug Prices in the MEPS and MarketScan: Implications for MEPS Editing Rules. AHRQ Working Paper No. 10001, February 2010. Available at: www.ahrq.gov. Accessed April 17, 2012. 27. Rasmussen JN, Chong A, Alter DA. Relationship between adherence to evidence-based pharmacotherapy and long-term mortality after acute myocardial infarction. JAMA. 2007;297(2):177-186. 28. Karve S, Cleves MA, Helm M, et al. Good and poor adherence: optimal cut-point for adherence measures using administrative claims data among diabetic patients. Med Care. 2008;46(11):112. 29. Hansen RA, Kim MM, Song L, Tu W, Wu J, Murray MD. Comparison of methods to assess medication adherence and classify nonadherence. Ann Pharmacother. 2009;43(3):413-422. 30. Briesacher BA, Andrade SE, Fouayzi H, et al. Comparison of drug adherence rates among patients with seven different medical conditions. Pharmacotherapy. 2008;28(4):437-443. 31. Agency for Healthcare Research and Quality. Medical Expenditure Panel Survey Downloadable Data Files e Codebook. Available at: http://www. meps.ahrq.gov. Accessed January 6, 2010. 32. Caro JJ, Ward AJ, O’Brien JA. Lifetime costs of complications resulting from type 2 diabetes in the U.S. Diabetes Care. 2002;25(3): 476-481.
33. UCLA Academic Technology Services, Statistical Consulting Group. Stata FAQ: How can I check for collinearity in survey regression? Available at: http://www.ats.ucla.edu/stat/stata/faq/svycollin.htm. Accessed July 2011. 34. Manning WG, Mullahy J. Estimating log models: to transform or not to transform? J Health Econ. 2001;20(4):461-494. 35. Hardin JW, Hilbe JM. The negative binomial family: interpreting coefficients. In: Generalized Linear Models and Extensions. 2nd ed. College Station, TX: Stata Press; 2007:218-219. 36. Bernstein J. Changes in Health Care Financing and Organization: Issue Brief: Impact of the Economy on Health Care. Princeton, NJ: Robert Wood Johnson Foundation; 2009. 37. Briesacher BA, Andrade SE, Fouayzi H, Chan KA. Comparison of drug adherence rates among patients with seven different medical conditions. Pharmacotherapy. 2008;28(4):437-443. 38. Salas M, Dyfrig H, Zuluaga A, et al. Costs of medication nonadherence in patients with diabetes mellitus: a systematic review and critical analysis of the literature. Value Health. 2009;12(6):915-922. 39. Mathers CD, Schofield DJ. The health consequences of unemployment: the evidence. Med J Aust. 1998;168(4):178-182. 40. World Health Organization. Adherence to Long Term Therapies: Evidence for Action. Geneva, Switzerland: World Health Organization; 2003. 41. Kessler RC, Turner JB, House JS. Effects of unemployment on health in a community survey: main, modifying, and mediating effects. J Soc Issues. 1988;44(4):59-85. 42. Granados JA. Macroeconomic fluctuations and mortality in postwar Japan. Demography. 2008;45(2):323-343.
Associations between Joblessness and Oral Anti-diabetic Medication Adherence in US Diabetic Working-age Adults Mary L. Davis-Ajami, PhDa, Milap C. Nahata, PharmDb, Gregory Reardon, RPh, PhDc, Eric E. Seiber, PhDd, Rajesh Balkrishnan, PhDe a
DEPARTMENT OF ORGANIZATIONAL SYSTEMS AND ADULT HEALTH, UNIVERSITY OF MARYLAND SCHOOL OF NURSING, BALTIMORE; bCOLLEGE OF PHARMACY AND DEPARTMENTS OF INTERNAL MEDICINE AND PEDIATRICS, COLLEGE OF MEDICINE, THE OHIO STATE UNIVERSITY, COLUMBUS; cINFORMAGENICS, LLC, WORTHINGTON, OHIO; dDIVISION OF HEALTH SERVICES MANAGEMENT AND POLICY, THE OHIO STATE UNIVERSITY COLLEGE OF PUBLIC HEALTH, COLUMBUS; AND eDEPARTMENT OF RESEARCH AND EDUCATION, UNIVERSITY OF MICHIGAN CENTER FOR GLOBAL HEALTH, DEPARTMENT OF CLINICAL, SOCIAL & ADMINISTRATIVE SCIENCES, DEPARTMENT OF HEALTH MANAGEMENT AND POLICY, UNIVERSITY OF MICHIGAN COLLEGE OF PHARMACY, ANN ARBOR
-
ABSTRACT
O B J E C T I V E : To assess potential associations between joblessness and oral anti-diabetic (OAD) medication adherence in US diabetic working-age adults. S T U D Y D E S I G N : A retrospective longitudinal panel design used pooled 2001-2007 Medical Expenditure Panel Survey (MEPS) data forming a nationally representative sample of diabetic individuals, ages 24-59 years. Pregnancy, seasonal job status, retired persons, a student designation, and those prescribed insulin were excluded. Adherence was measured using the proportion of days covered (PDC). A PDC $0.80 was classified as adherent. Descriptive statistics and multivariate regression analysis accounting for the MEPS’ complex survey design were conducted. R E S U L T S : There were 2256 individuals (means: age 48.3 years [SD 8.15], body mass index 31.1 [SD 0.30], Charlson Comorbidity Index 0.37 [SD 0.79]) who met study criteria. Thirty-four percent were jobless at the first interview round and 29% remained jobless all 5 interview rounds during the 2-year panel period. Reasons cited for joblessness included: waiting to start a new job (73%) and unable to work due to illness or disability (20%). Negligible proportions cited staying home to care for family members or maternity leave as reasons for joblessness. Proportionately, more individuals were nonadherent (55%, SE 0.006). Joblessness was associated with a 16% significant reduction in the PDC (b 15.9, P < 0.001), and a 25% less likelihood of OAD medication adherence compared with those employed (odds ratio 0.75; 95% confidence interval, 0.64-0.90, P ¼ 0.002), while holding all other variables constant.
See funding, conflict of interest, and authorship disclosures at the end of this article. Ó 2012 Elsevier Inc. All rights reserved Health Outcomes Research in Medicine (2012) 3, e139-e151
e140
Oral Anti-diabetic Medication Adherence and Joblessness in the US
C O N C L U S I O N S : The results indicate that jobless working-age individuals with diabetes were significantly less likely to adhere to OAD medication than employed individuals. K E Y W O R D S : Diabetes; Employment; Job status; Joblessness; Medication adherence; Medication use
Employment
offers individuals the means to generate income, and often provides benefits such as employer-sponsored insurance. As such, employment status may give additional context for health care affordability. Jobless, nondisabled, work-eligible, working-age individuals vulnerable to noninclusion in Medicare and Medicaid, those facing Consolidated Omnibus Budget Reconciliation Act (COBRA) premiums, or those uninsured, face challenging health care choices. Separate from health insurance coverage, joblessness, with its associated income loss, income poverty, and material deprivation often constrains resource availability, potentially impacting personal expenditures.1 Personal resource reduction may influence spending, purchasing, or consumption patterns for chronic disease management, including pharmacologic intervention. However, few studies examine associations between joblessness and chronic disease in working-age individuals. Joblessness, in individuals with chronic disease, may provide additional context explaining suboptimal medication adherence.
The World Health Organization suggests that chronic disease represents 46% of the global disease burden.2 Individuals with chronic disease experience health expenditures twice those without a chronic condition.3 United Nations Resolution 61/225 describes diabetes as “a chronic, debilitating, and costly disease” with devastating human, social, and economic consequences. Globally, diabetes prevalence continues to escalate, and US data report diabetes as the seventh leading cause of death, and the eighth most costly disease to treat, with 2007 total health care costs estimated at $174 billion.4 Type 2 diabetes shows increased out-of-pocket medical and medication costs compared with nondiabetics.5 Increased medication costs, including out-of-pocket expenses, influence medication nonadherence.6 Diabetes medication adherence remains suboptimal, with reported overall medication nonadherence estimated as 24.8%. Average diabetes medication adherence across 23 studies was reported as 67.5% (95% confidence interval [CI], 58.5%-75.8%).6 Pharmacologic nonadherence shows associations with disease progression, complications, hospitalization, premature disability, and mortality.6,7 Jobless individuals with diabetes may face financial constraints affecting relationships between medication underuse and poor diabetic health outcomes. Theoretical constructs suggest that contextual factors, medication regime complexity, clinical factors, and the health system influence cost-related medication underuse.8-11 Moreover, studies report significant associations between increased medication cost, as well as lower net worth and medication nonuse.11 Although lower net worth is associated with medication nonuse, few studies assess joblessness or unemployment in relation to medication adherence among US working-age diabetics. For working-age individuals, employment generally establishes a main source of income. The literature about diabetes and employment spanned topics related to worker productivity, eligibility for employment, and job suitability,12-14 as well as workplace discrimination, accommodations, and safety.15 Many studies used a payer perspective to assess workplace disease management, strategic cost minimization, absenteeism, presenteeism, work disability, or health insurance premiums.16 Research assessing diabetes and unemployment examined indirect costs associated with workforce nonparticipation, or the effects of disease severity or diabetes-related complications on retirement and disability.17,18 This literature reported significant associations between socioeconomic status and medication adherence, with incomes #US $46,000, food insufficiency, hunger, poverty, low literacy, substandard housing, and homelessness as factors in nonadherence.19 Associations between US labor force nonparticipation and medication adherence in diabetes remained unexplored.
Health Outcomes Research in Medicine - Vol. 3 / No. 3 / August 2012
This study focused on one contextual factordjob statusdspecifically, joblessness and its relationship to oral antidiabetic medication use in adult working-age individuals diagnosed with diabetes. We hypothesized that joblessness in working-age individuals with diabetes would show significantly lower oral antidiabetic (OAD) prescription medication adherence levels as measured by the proportion of days covered (PDC) per interview round, as well as a lower likelihood of medication adherence below a PDC 0.80 threshold compared with those employed.
RESEARCH DESIGN AND METHODS This study used a population-based retrospective longitudinal design with public domain deidentified pooled years 2001-2007 data from complete panels from the Household Component of the Medical Expenditure Panel Survey (MEPS). As a national probability survey sponsored by the Agency for Healthcare Research and Quality, MEPS provided comprehensive, nationally representative, valid, and accurate individual-level health care data with detailed employment and prescription medication information.20 The survey uses an overlapping panel design, surveying sample households during 5 separate interviews over 2 1/2 years. This study used MEPS over other publicly accessible health care databases for its demonstrated use in prior diabetes research, accessible employment information, and validated pharmacy prescription medication information. MEPS data have been used to assess diabetes pharmacologic treatment trends,21 diabetes disease management,22 direct medical costs,23 and diabetes utilization and expenditures,24 among other uses. Moreover, the survey elicits information about ambulatory prescription medication purchases or otherwise procured medicines such as samples during each interview round. Most self-reported prescription medication use gets linked to, for example, pharmacy claims data giving information about date filled, National Drug Code, medication name, strength, and quantity.25,26 Because the study used publicly accessible, de-identified secondary preexisting data, it was granted an institutional review board exemption from The Ohio State University. Full-year consolidated files provided administrative variables, employment status, and demographic information. Medication condition files were accessed to assess comorbidity and diabetes-related complications. Prescribed medicines files provided medication adherence data. Eligible respondents aged 24-59 years with any International Classification of Diseases, 9th Revision, Clinical Modification (ICD-9-CM) 3-digit code “250” for diabetes, from complete 2-year MEPS panels, were included. Publicly downloadable MEPS data provided only 3-digit ICD-9-CM codes, limiting finer diagnostic gradations. Pregnant females, those with a seasonal job status, those who were retired, or those with a student designation, as well as any individuals prescribed insulin, were excluded. Additionally, persons traditionally excluded from US Bureau of Labor Force employment data: persons under age 16 years, all persons confined to nursing homes or prisons, and active duty Armed Forces members, were excluded. The age upper bound was set at 59 years to help control for early retirement, and 24 years lower bound to control for workforce nonparticipation due to education. Pregnancy was excluded to control for gestational diabetes, and individuals prescribed insulin were excluded to control for type I diabetes and disease severity, as well as the difficulty measuring complex pharmaceutical regimens with injectable medications. The outcome variable, OAD pharmacotherapeutics adherence, was calculated for complete 2-year panel periods using the PDC with a predetermined $0.80 cut-point. The PDC represented an objective, indirect method for estimating medication adherence appropriate for claims data linked to administrative data.27 The PDC showed reasonable predictive value for subsequent hospitalization in diabetes research.28 A 0.80 cut-point showed a fair balance between sensitivity and specificity for research assessing medication adherence using administrative data,29 as well as for research comparing adherence rates for individuals with chronic conditions, including diabetes.30 Moreover, MEPS data were better suited to calculating medication adherence by the PDC than by the medication possession ratio because the MEPS complex survey overlapping panel design precluded using an index date, among other reasons. MEPS prescribed medicines file does not
e141
e142
Oral Anti-diabetic Medication Adherence and Joblessness in the US
provide a days supply variable. Days supply was calculated using the OAD unique identifier, strength, dosing unit, beginning and ending date for each prescription, and number of pills supplied. A pharmacotherapy product master list was created using the unique drug code (National Drug Code) for OAD prescriptions for the following drug classes: sulfonylurea, meglitinide, biguanide, thiazolidinedione, alpha-glycosidase inhibitors, and DPP-4 inhibitors. The primary variable of interest, joblessness, was assessed at each interview round using the MEPS EMPST[round] variable. Individuals reporting “they were not employed with no job to return to” were classified as jobless.31 Individuals were classified as employed when respondents reported having a job, being employed, or having a job to return to. The EMPST[round] variable was collapsed into a zero or one (0 ¼ yes, employed; 1 ¼ no, jobless) categorical variable, with employed persons serving as the reference group. Demographic covariates included: age, sex, educational attainment, ethnicity, race, marital status, health insurance status, total income, total prescription, and total prescription medication expense. Geographical region was used only in descriptive analysis. Clinical covariates were classified as comorbidities or complications. The enhanced Charlson Comorbidity Index (CCI) was used as a proxy for disease-burden risk adjustment. Descriptive analysis used a body mass index (BMI) continuous variable, and multivariate analysis used categorical variables created from standard US BMI designations for underweight, normal weight, overweight, obese, and moribund obese. Medical conditions file ICD-9-CM codes identified first interview panel round complication status. Categorical variables were created for specific eye, neurological, macrovascular, microvascular, and renal complications commonly associated with diabetes and further collapsed into a zero or one (0 ¼ none, 1 ¼ one or more complication) categorical variable.32 Descriptive analyses were conducted using a pooled person-level analytical dataset to report means for continuous variables, independent sample t-tests for differences between continuous variables, and chisquared tests for differences between categorical variables. Multicollinearity was assessed using procedures appropriate for complex survey data.33 A series of regressions were estimated using the variables representing individual employment status, insurance status, and spouse employment status.33 All variance inflation factor calculations were well below 10, suggesting no problem with multicollinearity. Separately, interactions between individual employment status and insurance status, as well as individual and spousal employment status, were nonsignificant and not reported here. Multivariate analyses were conducted on a pooled analytical panel dataset. Regression diagnostics showed a right skew for the PDC outcome variable. Estimates were robust to generalized linear models (GLM) with a log link versus a natural log transformation.34 Multivariate GLM with a log link assessed overall PDC level, and logistic regression assessed medication adherence defined by the 0.80 threshold. The GLM results were reported as log-normal.34,35 Statistical analyses were conducted using Stata, version 11 (StataCorp LP, College Station, TX) with appropriate complex survey estimation commands. The linearized Taylor series expansion method estimated variance.
RESULTS Of the 240,409 individuals from MEPS complete panels for pooled years 2001-2007, 2256 subjects (means: age 48.3 years [SD 8.15], BMI 31.5 [SD 0.30], and CCI 0.37 [SD 0.79]) were included in the study. Mean total income approximated $11,967 [SD 9632] and mean total prescription medication expense approximated $2556 [SD 3177] (Table 1: summary characteristics). Approximately 34% were jobless during the first
Health Outcomes Research in Medicine - Vol. 3 / No. 3 / August 2012
e143
- T A B L E 1 : Summary Characteristics in a Study Assessing Associations between Joblessness and OAD Medication Adherence (Using the PDC) in a Cohort of Working-age Adults All Individuals n ¼ 2256
Age, years Mean 48.29, SE 0.09 24-40 41-59 Sex Male Female Marital status Married Unmarried Educational degree No degree High school College Graduate school Ethnicity Hispanic Other Health insurance status Private health insurance Public health insurance Uninsured Poverty category Poor, near poor, or low Middle income High income Race White Black Other Region Northeast Midwest South West Total income, mean $11,797, SE 98.91 $0-$19,000 $20,000-$29,999 $$30,000 Total prescription med expense, mean $2556, SE 28.11 $0-$1499 $1500-$2999 $$3000 Body mass index category, mean 31, SE 0.30 Underweight Normal Overweight Obese/moribund obese Charlson Comorbidity Index, mean 0.37, SE 0.79 0 1 $2
Employed Individuals n ¼ 1488
Jobless Individuals n ¼ 767
n (%)
SE
n (%)
SE
n (%)
SE
384 (17%) 1872 (83%)
0.003 0.003
268 (18%) 1220 (82%)
0.005 0.005
107 (14%) 660 (86%)
0.006† 0.006
1105 (49%) 1151 (51%)
0.004 0.004
833 (56%) 655 (44%)
0.006 0.006
268 (35%) 767 (65%)
0.008‡ 0.008
1421 (63%) 835 (37%)
0.004 0.004
1011 (68%) 477 (32%)
0.006 0.006
414 (54%) 353 (46%)
0.008‡ 0.008
474 (21%) 1218 (54%) 271 (12%) 293 (13%)
0.004 0.004 0.003 0.003
193 (13%) 803 (54%) 238 (16%) 254 (17%)
0.004 0.007 0.005 0.005
261 (34%) 407 (53%) 46 (06%) 53 (07%)
0.007‡ 0.008 0.004 0.004
360 (16%) 1896 (84%)
0.003 0.003
208 (14%) 1280 (86%)
0.004 0.004
138 (18%) 629 (82%)
0.005‡ 0.005
1557 (69%) 474 (21%) 225 (10%)
0.003 0.003 0.003
1280 (86%) 60 (04%) 148 (09%)
0.002 0.004 0.004
276 (36%) 407 (53%) 84 (11%)
0.004‡ 0.008 0.008
767 (34%) 699 (31%) 790 (35%)
0.004 0.005 0.005
283 (19%) 520 (35%) 685 (46%)
0.005 0.007 0.007
499 (65%) 169 (22%) 99 (13%)
0.008‡ 0.007 0.006
1669 (74%) 406 (18%) 181 (08%)
0.004 0.004 0.002
1131 (76%) 238 (16%) 119 (08%)
0.005 0.004 0.004
537 (70%) 169 (22%) 61 (08%)
0.007† 0.006 0.004
338 (15%) 384 (17%) 970 (43%) 564 (25%)
0.003 0.003 0.004 0.004
253 (17%) 327 (22%) 565 (38%) 343 (23%)
0.006 0.006 0.006 0.006
153 (20%) 138 (18%) 314 (41%) 162 (21%)
0.007* 0.006 0.008 0.007
1715 (76%) 429 (19%) 499 (05%)
0.004 0.004 0.002
714 (48%) 565 (38%) 209 (14%)
0.005 0.005 0.007
705 (92%) 54 (07%) 08 (01%)
0.008‡ 0.007 0.008
1308 (58%) 406 (18%) 542 (24%)
0.005 0.004 0.004
982 (66%) 268 (18%) 238 (16%)
0.006 0.005 0.005
330 (43%) 153 (20%) 284 (37%)
0.008‡ 0.007 0.008
90 (04%) 248 (11%) 587 (26%) 1331 (59%)
0.002 0.003 0.005 0.005
45 (03%) 179 (12%) 417 (28%) 847 (57%)
0.002 0.004 0.006 0.005
31 (04%) 84 (11%) 184 (24%) 486 (61%)
0.003 0.005 0.007 0.007
1692 (75%) 406 (18%) 158 (07%)
0.004 0.004 0.002
1235 (83%) 208 (14%) 45 (03%)
0.005 0.004 0.002
434 (57%) 207 (27%) 126 (16%)
0.005‡ 0.004 0.002
e144
Oral Anti-diabetic Medication Adherence and Joblessness in the US
- T A B L E 1 (continued ): Summary Characteristics in a Study Assessing Associations between Joblessness and OAD Medication Adherence (Using the PDC) in a Cohort of Working-age Adults All Individuals n ¼ 2256
Diabetes-related complications Yes No
Employed Individuals n ¼ 1488
Jobless Individuals n ¼ 767
n (%)
SE
n (%)
SE
n (%)
SE
948 (42%) 1308 (58%)
0.003 0.003
1220 (82%) 268 (18%)
0.005 0.005
522 (68%) 245 (32%)
0.007‡ 0.007
OAD ¼ oral anti-diabetic; PDC ¼ proportion of days covered. Data from Medical Expenditure Panel Survey (MEPS) pooled panel years 2001-2007. Significantly different from employed individuals. *P < 0.05. † P < 0.01. ‡ P < 0.001.
interview round, with 29% remaining jobless through all 5 interview rounds. For those giving reasons for being jobless, the majority (74%) were waiting to start a new job, followed by those unable to work due to illness or disability (20%). Compared with those employed, jobless individuals were proportionately older, female, less
- T A B L E 2 : Employment Status Summary Characteristics for a Cohort of Working-age Adults from MEPS Pooled Panel Years 2001-2007 Assessing Associations between Joblessness and OAD Medication Adherence as Measured by the PDC
Employment status at first interview round Employed Jobless Job status by interview round 1 Employed No job for 2 interview rounds No job for 3-5 interview rounds Job status by individual interview rounds Employed No job for 1 round No job for 2 interview rounds No job for 3 interview rounds No job for 4 interview rounds No job for 5 interview rounds Reasons cited for no job Could not find work Retired Unable to work (ill/disabled) On temporary layoff Maternity/paternity leave Going to school Taking care of home or family Wanted some time off Waiting to start a new job Other
Proportion
SE
0.66 0.34
0.004 0.004
0.6500-0.6693 0.3289-0.3480
0.604 0.057 0.339
0.005 0.002 0.004
0.5939-0.6142 0.0524-0.0620 0.3289-0.3480
0.604 0.036 0.022 0.034 0.012 0.292
0.005 0.002 0.001 0.002 0.001 0.004
0.5939-0.6142 0.0317-0.0397 0.0188-0.0244 0.0303-0.0381 0.0097-0.0143 0.2833-0.3014
0.015 <0.000 0.201 0.008 <0.000 <0.000 0.034 0.002 0.736 0.004
0.001 >0.001 0.004 >0.001 >0.001 >0.001 >0.001 >0.001 0.004 0.0005
MEPS ¼ Medical Expenditure Panel Survey; OAD ¼ oral anti-diabetic; PDC ¼ proportion of days covered. *Confidence intervals <0.001.
95% Confidence Interval
0.1029-0.0182 * 0.1930-0.2089 0.0003-0.0014 0.000003-0.0001 * 0.0306-0.0376 0.0009-0.0025 0.7265-0.7520 0.0024-0.0044
Health Outcomes Research in Medicine - Vol. 3 / No. 3 / August 2012
e145
- T A B L E 3 : OAD Medication Adherence (Adherent ¼ PDC $0.80) in a Cohort of Working-age Adults with Diabetes Using Data from MEPS Pooled Panel Years 2001-2007 All Individuals
PDC Mean (SD) 0.58 (0.49) Job status Employed Jobless
Age, years 24-40 41-59 Sex Male Female Marital status Married Unmarried Educational degree No degree High school College Graduate school Ethnicity Hispanic Other Health insurance status Private health insurance Public health insurance Uninsured Race White Black Other Region Northeast Midwest South West Total income $0-$19,000 $20,000-$29,999 $$30,000 Total prescription med expense $0-$1499 $1500-$2999 $$3000 Body mass index category Underweight Normal Overweight Obese/moribund obese Charlson Comorbidity Index 0 1 $2
PDC $.8-1
PDC $.8-1
PDC <0.8
n ¼ 2256
OAD Adherent
OAD Nonadherent
n (%) 1489 (66%) 767 (34%)
SE 0.009 0.009
n (%) 676 (30%) 340 (15%) PDC $.8-1 n ¼ 1015
n (%) (835) 37% (406) 18% PDC <0.8 n ¼ 1241
384 (17%) 1872 (83%)
0.003 0.003
n, % 102 (10%) 913 (90%)
SE 0.008 0.005
n, % 161 (13%) 1080 (87%)
SE 0.008 0.006
1105 (49%) 1151 (51%)
0.004 0.004
507 (50%) 507 (50%)
0.010 0.010
645 (52%) 596 (48%)
0.012 0.012
1421 (63%) 835 (37%)
0.004 0.004
640 (63%) 375 (37%)
0.010 0.009
807 (65%) 434 (35%)
0.011 0.011
474 (21%) 1218 (54%) 271 (12%) 293 (13%)
0.004 0.004 0.003 0.003
234 (23%) 528 (52%) 112 (11%) 141 (14%)
0.007 0.010 0.007 0.008
248 (20%) 695 (56%) 137 (11%) 161 (13%)
0.008 0.012 0.008 0.009
360 (16%) 1896 (84%)
0.003 0.003
152 (15%) 863 (85%)
0.006 0.006
211 (17%) 1030 (83%)
0.007 0.008
1557 (69%) 474 (21%) 225 (10%)
0.003 0.003 0.003
721 (71%) 213 (21%) 81 (08%)
0.008 0.008 0.005
881 (71%) 235 (19%) 124 (10%)
0.009 0.008 0.006
1669 (74%) 406 (18%) 181 (08%)
0.004 0.004 0.002
761 (75%) 173 (17%) 82 (08%)
0.008 0.006 0.007
931 (75%) 199 (16%) 111 (09%)
0.010 0.008 0.007
338 (15%) 384 (17%) 970 (43%) 564 (25%)
0.003 0.003 0.004 0.004
182 (18%) 182 (18%) 396 (39%) 255 (25%)
0.008 0.007 0.009 0.008
236 (19%) 261 (21%) 496 (40%) 248 (20%)
0.009* 0.009 0.012 0.008
1715 (76%) 429 (19%) 499 (05%)
0.004 0.004 0.002
711 (70%) 213 (21%) 91 (09%)
0.011 0.009 0.007
819 (66%) 348 (28%) 74 (06%)
0.014* 0.013 0.007
1308 (58%) 406 (18%) 542 (24%)
0.005 0.004 0.004
355 (35%) 244 (24%) 416 (41%)
0.009 0.009 0.010
236 (19%) 223 (18%) 782 (63%)
0.009* 0.009 0.009
90 (04%) 248 (11%) 587 (26%) 1331 (59%)
0.002 0.003 0.005 0.005
20 (02%) 71 (07%) 244 (24%) 681 (66%)
0.003 0.005 0.009 0.009
50 (04%) 112 (09%) 310 (25%) 769 (62%)
0.004 0.007 0.010 0.011
1692 (75%) 406 (18%) 158 (07%)
0.004 0.004 0.002
741 (73%) 193 (19%) 81 (08%)
0.008 0.008 0.008
943 (76%) 223 (18%) 75 (06%)
0.010 0.009 0.001
e146
Oral Anti-diabetic Medication Adherence and Joblessness in the US
- T A B L E 3 (continued ): OAD Medication Adherence (Adherent ¼ PDC $0.80) in a Cohort of Working-age Adults with Diabetes Using Data from MEPS Pooled Panel Years 2001-2007 All Individuals
PDC Mean (SD) 0.58 (0.49) Diabetes-related complications Yes No
PDC $.8-1
PDC $.8-1
PDC <0.8
n ¼ 2256
OAD Adherent
OAD Nonadherent
948 (42%) 1308 (58%)
0.003 0.003
792 (78%) 223 (22%)
0.008 0.008
1005 (81%) 236 (19%)
MEPS ¼ Medical Expenditure Panel Survey; OAD ¼ oral anti-diabetic; PDC ¼ proportion of days covered. Significantly different from adherent individuals. *P < 0.05.
educated, in the lowest income category, and had greater BMI. Although proportionately more females were jobless than males, for women citing reasons for joblessness, only 3% reported not working due to taking care of home or family, and none (0%) reported begin on maternity leave (Table 2: job status summary characteristics). Of the approximately 34% categorized as poor, near poor, or low income, approximately 22% were jobless. Ten percent of the study cohort reported no health insurance. Jobless individuals showed slightly greater proportions uninsured when compared with employed respondents. Jobless individuals with health insurance were predominately covered by public insurance. Summary descriptive comorbidity and complication characteristics showed significant differences between employed and jobless individuals for the CCI and diabetes-related complications. Overall, approximately 55% were nonadherent (SE 0.006). Truncated mean PDC was 0.58 (SD 0.49). Of those found nonadherent, 18% were jobless, versus 37% employed. Proportionately, the younger age group, males, those with a high school education, those in the middle total income category, and those with the highest total prescription medication expense were proportionately more nonadherent than those aged 41-59 years. Individuals living in the western region of the US, as well as individuals who were obese/ moribund obese, were proportionately more adherent (Table 3, medication adherence). Multivariate GLM with a log link regression evaluated associations between joblessness and a PDC continuous measure. The transformed log-normal GLM regression showed jobless individuals with a 16% lower PDC level (b ¼ 0.157, P < 0.01) compared with employed individuals, when holding all other variables constant. The covariates age, educational attainment, race, total income, total prescription medication expense, CCI, and the presence of one or more diabetic complication(s) also showed statistically significant differences (Table 4: GLM regression). Logistic regression evaluated associations between joblessness and adherence (PDC $0.80) using a dichotomous measure. Logistic regression showed that the odds of OAD medication adherence was 25% significantly lower for jobless individuals compared with those employed (OR 0.75; 95% CI, 0.64-0.90, P ¼ 0.002) while holding all other variables constant. Other parameters significantly affecting the likelihood of OAD medication adherence included: age, female sex, educational attainment, ethnicity, total income, total prescription medication expenses, and the CCI. For individuals aged 41-59 years, every 1-year increase in age increased the likelihood of being adherent 45% (OR 1.45; 95% CI, 1.17-1.74, P < 0.001). Females were 17% less likely than males to be adherent (OR 0.83; 95% CI, 0.72-0.95, P ¼ 0.008), and those with high school or college degrees were less likely than those without degrees to be adherent. Those whose total incomes ranged from $20,000-$29,999 were 35% less likely to be adherent than those with lower incomes (OR 0.65; 95% CI, 0.54-0.78, P # 0.001). Individuals with lower total prescription medication expenses were less likely to be adherent than those with the highest total prescription medication expenses. Every one-unit increase in the CCI resulted in a 15% reduction in the likelihood of adherence (OR 0.85; 95% CI, 0.78-0.99, P < 0.001) (Table 4: logistic regression).
0.009 0.009
Health Outcomes Research in Medicine - Vol. 3 / No. 3 / August 2012
e147
- T A B L E 4 : Estimates of Associations between Joblessness and OAD Medication Adherence (Measured by PDC) in a Cohort of US Working-age Adults from MEPS Pooled Panel Years 2001-2007 GLM Log-normal b Primary variable of interest Jobless individual1 Demographic characteristics Age2 Sex3 Marital status4 Education: degree5 High school College Graduate school Ethnicity6 Hispanic Race7 Black Other Economic-related factors Health insurance coverage8 Private health insurance Public health insurance Total income9 $20,000-$29,999 $$30,000 Total prescription medication expenses10 $1500-$2999 $$3000 Comorbidity/complications of diabetes Charlson comorbidity index Body mass index category11 Normal Overweight Obese/moribund obese Diabetes-related complication12 Constant
SE
Logistic Regression Model P-Value
Odds Ratio
SE
95% CI
0.16
0.054
0.003*
0.75
0.066
0.64-0.90*
0.15 0.04 0.04
0.068 0.041 0.045
0.039* 0.295 0.359
1.45 0.83 0.92
0.147 0.060 0.068
1.17-1.74* 0.72-0.95* 0.77-1.03
0.06 0.32 0.01
0.049 0.107 0.075
0.224 0.005* 0.907
0.78 0.56 0.94
0.061 0.100 0.130
0.69-0.93* 0.41-0.83* 0.74-1.27
0.02
0.056
0.714
0.85
0.41
0.74-1.03
0.02 0.15
0.050 0.074
0.633 0.035*
0.92 1.12
0.82 0.139
0.76-1.07 0.89-1.45
0.02 0.04
0.072 0.077
0.876 0.712
1.12 0.95
0.120 0.107
0.92-1.40 0.76-1.18
0.18 0.04
0.057 0.078
0.001* 0.602
0.65 1.06
0.061 0.158
0.54-0.78* 0.77-1.38
0.24 0.72
0.054 0.052
<0.001* <0.001*
0.65 0.31
0.063 0.027
0.50-0.78* 0.23-0.33*
0.10
0.030
0.001*
0.85
0.037
0.78-0.99*
0.14 0.25 0.17 0.11
0.128 0.117 0.111 0.050
0.242 0.051 0.110 0.017*
1.20 1.28 1.25 0.94
0.270 0.263 0.245 0.078
0.74-1.79 0.81-1.82 0.79-1.72 0.80-1.08
0.155
0.000
0.490
CI ¼ confidence interval; GLM ¼ generalized linear model; MEPS ¼ Medical Expenditure Panel Survey; OAD ¼ oral anti-diabetic; PDC ¼ proportion of days covered. Reference groups: 1Jobless: employed, 2Age: 24-40 years, 3Sex: male, 4Marital status: not married, 5Degree: no degree, 6Ethnicity: not-Hispanic, 7Race: white, 8Health insurance: no insurance coverage, 9Total income: $0-$19,999, 10Total prescription medication expenses: #$1499, 11BMI: underweight, 12Diabetes-related complication: none. *Confidence intervals <0.001.
CONCLUSIONS This research assessing associations between joblessness and OAD medication adherence in adult working-age individuals showed significant proportions reporting no job, striking proportions waiting to start another job, and a troubling chronic dimension to workforce nonparticipation, a finding not fully attributable to disability.
e148
Oral Anti-diabetic Medication Adherence and Joblessness in the US
Individual regression models found significant associations between joblessness and lower PDC levels, as well as an increased likelihood of suboptimal OAD medication adherence as measured by a 0.80 PDC adherence threshold. These findings suggest that joblessness may influence OAD pharmacotherapeutics nonadherence. Joblessness may decrease available personal resources for necessary and discretionary spending, placing financial constraints on personal health care expenditures. Reduced monetary resources, coupled with uncertainty for future economic viability, may influence decisions to reduce medication refills and collaterally affect behavioral dimensions, including changes in consumption patterns.36 The results suggest differing medication refill patterns for jobless individuals compared with those employed. Fully 45% of the study cohort was found adherent using a 0.80 OAD pharmacotherapy threshold. These results were lower than those reported in a 2001-2004 longitudinal study reporting mean medication possession ratio for type 2 diabetes using claims data,37 as well as previously reported mean (percent) overall medication adherence from a meta-analysis of all medication adherence studies for years 1948-1998.6 The divergence may stem from differences in calculating adherence, among other reasons. Our results suggest that a suboptimal percentage of individuals met the 0.80 adherence threshold. This has implications for overall diabetic disease management, such as the known correlation between medication adherence and the ability to reach target glucose control.38 Interestingly, insurance status was not a significant covariate. Although uninsured individuals were proportionately more nonadherent than those covered by private or public health insurance, the regression models showed no significant associations between insurance status and OAD adherence as measured by a 0.80 PDC threshold. This suggests that other factors may influence OAD medication adherence. Moreover, our data came from the MEPS household component of the survey. Due to the characteristics of this database, insurance and employment may not interact like data directly linked to medical claims. In MEPS, although an individual respondent with diabetes may be jobless, others in the household may serve as the main health insurance policy holder. However, there were no significant interactions between jobless individuals and spouses. Higher proportions of employed persons were adherent. Respondents reporting a jobless work status showed a 16% significant reduction in the PDC compared with those employed, and were 24% less likely to meet or exceed the 0.80 PDC adherence threshold. Poor medication adherence has known associations with poor health outcomes and increased health care costs.38 Over time, poor health outcomes may become more apparent for those out of work, particularly for individuals experiencing chronic joblessness.39 Jobless individuals with diabetes may represent a vulnerable segment for lagged long-term suboptimal health outcomes. The World Health Organization suggests that social and economic factors, among others, influence medication adherence.40 Economic factors associated with chronic diabetes disease management may interface with financial constraints during periods of joblessness. This is concerning in light of recent research reporting increased deferrals for health care services to enhance personal financial savings, as well as reports of changing medication consumption patterns as evidenced by skipped dosages and pill splitting stemming from the 2007 economic downturn.36 Moreover, earlier research conducted after the recession in the early 1980s found financial strain stemming from unemployment significantly associated with ill health.41 Our study, with its focus on joblessness and medication adherence, may shed some light on scant previous research suggesting countercyclical relationships in diabetes.42 Loosely coupled linkages between joblessness leading to reduced personal resources, scaled back OAD medication purchases and consumption, or outright nonadherence to OAD medication regimens, and known poor health outcomes for medication nonadherence may, over time, show lagged increases in diabetes mellitus complications or other serious health consequences. This conjecture may be difficult to empirically assess, but is within the realm of possibility. Significant proportions of the study cohort were out of work the entire 2-year MEPS panel time frame, suggesting a troubling longevity to their joblessness. Workforce participation for adults with diabetes and other chronic conditions command the attention of public policy-makers, particularly when prioritizing resource allocation. As a starting position, health care providers and systems need standard processes to identify individuals facing financial pressure and their
Health Outcomes Research in Medicine - Vol. 3 / No. 3 / August 2012
vulnerability to lower medication adherence. Future research could focus on assessing links between financial strain stemming from joblessness and changes in personal choices resulting in differing patterns of medication use, as well as other treatment deferrals. Another area of interest for future research suggested by findings from this study includes a need to further understand the magnitude of chronic joblessness in the diabetic population and possible correlates with macroeconomic indicators. There were several limitations to this research. We could not assess long-term joblessness beyond the 2-year MEPS panel window. The study minimally assessed length of joblessness. The research did not control for the economic cycle or nuances in employment patterns using Bureau of Labor Statistics unemployment data or other macroeconomic indicators, and made no distinction between full-time versus part-time work or job churning. No provision was made to examine lagged effects. The study did not control for changes in diabetic treatment patterns over time, and was subject to limitations common to retrospective database analysis such as selection bias, missing or incomplete information, and recall bias. Adherence was measured indirectly without information about consumption patterns or sample medications. Finally, observational studies may not infer causation. Joblessness in working-age adults with diabetes deserves policy considerations. As a starting position, health care providers and systems need standard processes to identify individuals facing financial pressure and employment vulnerability. Job loss, as well as length of joblessness, could be assessed in tandem with other financial measures assessing income, and affordability and willingness to pay for medications, medical supplies, and services. For individuals identified as needing long-term pharmacotherapy for chronic disease management, future research could assess how the presence or lack of safety-net provisions among workeligible but nonworking individuals with chronic disease impacts disease management cost drivers, complication rates, and mortality. More information about how joblessness affects personal decisions regarding disease management and personal health care spending is needed. Future research could assess differences in health care consumption patterns, particularly medication use patterns, between employed and jobless individuals. Secondly, further research could further identify and evaluate factors associated with nonadherence. Pharmacologic nonadherence risk-assessment tools geared for health professionals need further refinement. Researchers should continue assessing optimal medication adherence thresholds with the view to identifying areas resulting in suboptimal disease management. This research identified significant associations between joblessness and OAD medication adherence as measured by the PDC level and a 0.80 adherence threshold in working-age adults with diabetes. Jobless individuals showed significant reductions in the PDC, as well as a decreased likelihood of OAD medication adherence when compared with those employed. Overall, 55% of the study cohort had OAD medication adherence levels less than a predetermined 0.80 threshold, suggesting suboptimal OAD medication adherence in this nationally representative sample. This study identified joblessness as a significant parameter for diabetes oral medication adherence for working-aged adults with diabetes.
ACKNOWLEDGMENTS Mary L. Davis-Ajami, PhD, affiliated with the University of Maryland School of Nursing, researched data and wrote the manuscript and serves as corresponding author. Milap C. Nahata, PhD, of The Ohio State University College of Pharmacy, reviewed/edited the manuscript. Rajesh Balkrishnan, PhD, of the University of Michigan College of Pharmacy, contributed to the research design, researched data, and reviewed/edited the manuscript. Gregory Reardon, PhD, of Informagenics, researched data and reviewed/edited the manuscript. Eric E. Seiber, PhD, of The Ohio State University College of Public Health, reviewed/edited the manuscript. We thank Helen Leonard, PharmD for assistance in identifying each OAD agent listed in the
e149
e150
Oral Anti-diabetic Medication Adherence and Joblessness in the US
Prescribed Medicines files. We also wish to acknowledge James W. McAuley, PhD, of The Ohio State University College of Pharmacy for his helpful review of the study.
SUPPLEMENTARY DATA Supplementary data associated with this article can be found, in the online version, at doi:10.1016/ j.ehrm.2012.06.001. Corresponding Author: Mary L. Davis-Ajami, PhD, Department of Organizational Systems and Adult Health, University of Maryland School of Nursing, 655 Lombard Street, Room 465D, Baltimore, MD 21201. E-mail address: [email protected] The authors have no relevant conflicts of interest to disclose. There were no funding sources. Mary Lynn Davis-Ajami wrote the first draft of the manuscript. No grants or honoraria were received.
R E F E R E N CE S 1. Färber H. What do we know about job loss in the United States? Evidence from the Displaced Workers Survey, 1984-2004. Federal Reserve Bank of Chicago. Econ Perspect. 2005;29(2):13-28. 2. World Health Organization (WHO). WHO technical report series: Diet, nutrition and the prevention of chronic diseases. Joint WHO/FAO Expert Consultation, WHO. Geneva, Switzerland: World Health Organization. Available at: http://whqlibdoc.who.int/trs/who_trs916.pdf; 2003. Accessed July 6, 2011. 3. Stanton MW, Rutherford MK. The High Concentration of US Health Care Expenditures (Research in Action issue 19 AHRQ Pub. No. 06-0060). Rockville, MD: Agency for Healthcare Research and Quality; 2006. 4. Centers for Disease Control and Prevention (CDC). National Diabetes Fact Sheet: National Estimates and General Information on Diabetes and Prediabetes in the United States, 2011. Atlanta, GA: U.S. Department of Health and Human Services, Centers for Disease Control and Prevention; 2011. 5. Robard HW, Green AJ, Fox KM, Grandy S, SHIELD Study Group. Impact of type 2 diabetes mellitus on prescription medication burden and out-ofpocket healthcare expenses. Diabetes Res Clin Pract. 2010;87(3):360-365. 6. DiMatteo MR. Variations in patients’ adherence to medical recommendations: a quantitative review of 50 years of research. Med Care. 2004;42(3): 200-209. 7. Cramer JA. A systematic review of adherence with medications for diabetes. Diabetes Care. 2004;27(5):1218. 8. Asche C, LaFleur J, Conner C. A review of diabetes treatment adherence and the association with clinical and economic outcomes. Clin Ther. 2011; 33(1):74-109. 9. Piette JD, Heisler M, Horne R, et al. A conceptually based approach to understanding chronically ill patients’ responses to medication cost pressures. Soc Sci Med. 2006;62:846-857. 10. Piette JD. Cost-related medication underuse: a window into patient’s medication-related concerns. Diabetes Spectr. 2009;22(2):77-80. 11. Zivin K, Ratliff S, Heisler MM, Langa K, Piette JD. Factors influencing cost-related nonadherence to medication in older adults: a conceptually based approach. Value Health. 2010;13(4):338-345. 12. Lavigne J, Phelps C, Mushlin A, Lednar W. Reductions in individual work productivity associated with type 2 diabetes mellitus. Pharmacoeconomics. 2003;21(15):1123-1134. 13. Ramsey S, Summers KH, Leong SA, et al. Productivity and medical costs of diabetes in a large employer population. Diabetes Care. 2002;25:23.
14. Tunceli K, Bradley C, Nerenz D, et al. The impact of diabetes on employment and work productivity. Diabetes Care. 2005;11(28):2662-2667. 15. American Diabetes Association. Position statement: diabetes and employment. Diabetes Care. 2011;34:S82-S86. 16. Crawley J, Rizzo J, Haas K. The association of diabetes with job absenteeism costs among obese and morbidly obese workers. J Occup Environ Med. 2008; 50:527-534. 17. Herquelot E, Guéguen A, Bonenfant S, Dray-Spira R. Impact of diabetes on work cessation: data from the GAZEL cohort study. Diabetes Care. 2011;34(6):1344-1349. 18. Vijan S, Hayward RA, Langa KM. The impact of diabetes on workforce participation: Results from a national household sample. Health Serv Res. 2004;39(6 Pt. 1):1653-1669. 19. Peeters B, Van Tongelen I, Boussery K, et al. Factors associated with medication adherence to oral hypoglycemic agents in different ethnic groups suffering from Type 2 diabetes: a systematic literature review and suggestions for future research. Diabet Med. 2011;28:262-275. 20. Olin G, Zuvekas S, Kumar V, Ward P, Williams K Wobus D. 2008. Medicare-MEPS Validation Study: A Comparison of Hospital and Physician Expenditures. Agency for Healthcare Research and Quality Working paper No. 08003, March 2008. Available at: http://gold.ahrq.gov. Accessed April 2, 2012. 21. Sarpong E, Miller GE. Trends in the Pharmaceutical Treatment of Diabetes, 1997 to 2007. Research Findings #30. Rockville, MD: Agency for Healthcare Research and Quality. Available at: http://meps.ahrq.gov/mepsweb/ data_files/publications/rf30/rf30.pdf; 2010. Accessed April 2, 2012. 22. Soni A. Trends in Use and Expenditures for Diabetes among Adults 18 and Older, U.S. Civilian Noninstitutionalized Population, 1996 and 2007. Statistical Brief #304. Rockville, MD: Agency for Healthcare Research and Quality; 2010. 23. Olin G, Machlin SR, Rhoades J. Estimating the Cost of Illness: The Case of Diabetes. Agency for Healthcare Research and Quality Working Paper No. 08001, January 2008. Available at: http://gold.ahrq.gov. Accessed April 2, 2012. 24. Olin G. Proportion and Medical Expenditures of Adults Being Treated for Diabetes, 1996 and 2003. Statistical Brief #146. Rockville, MD: Agency for Healthcare Research and Quality. Available at: http://www.meps.ahrq. gov/mepsweb/data_files/publications/st146/stat146.pdf; 2006. Accessed April 2, 2012.
Health Outcomes Research in Medicine - Vol. 3 / No. 3 / August 2012
e151
R E F E R E N C E S Continued 25. Moeller JF, Stagnitti MN, Horan E, Ward P, Kieffer N, Hock E. Outpatient Prescription Drugs: Data Collection and Editing in the 1996 Medical Expenditure Panel Survey (HC-010A) (MEPS Methodology Report No. 12, AHRQ Pub. No01-0002). Rockville, MD: Agency for Healthcare Research and Quality; 2001. 26. Zodet MW, Hill SC, Miller E. Comparison of Retail Drug Prices in the MEPS and MarketScan: Implications for MEPS Editing Rules. AHRQ Working Paper No. 10001, February 2010. Available at: www.ahrq.gov. Accessed April 17, 2012. 27. Rasmussen JN, Chong A, Alter DA. Relationship between adherence to evidence-based pharmacotherapy and long-term mortality after acute myocardial infarction. JAMA. 2007;297(2):177-186. 28. Karve S, Cleves MA, Helm M, et al. Good and poor adherence: optimal cut-point for adherence measures using administrative claims data among diabetic patients. Med Care. 2008;46(11):112. 29. Hansen RA, Kim MM, Song L, Tu W, Wu J, Murray MD. Comparison of methods to assess medication adherence and classify nonadherence. Ann Pharmacother. 2009;43(3):413-422. 30. Briesacher BA, Andrade SE, Fouayzi H, et al. Comparison of drug adherence rates among patients with seven different medical conditions. Pharmacotherapy. 2008;28(4):437-443. 31. Agency for Healthcare Research and Quality. Medical Expenditure Panel Survey Downloadable Data Files e Codebook. Available at: http://www. meps.ahrq.gov. Accessed January 6, 2010. 32. Caro JJ, Ward AJ, O’Brien JA. Lifetime costs of complications resulting from type 2 diabetes in the U.S. Diabetes Care. 2002;25(3): 476-481.
33. UCLA Academic Technology Services, Statistical Consulting Group. Stata FAQ: How can I check for collinearity in survey regression? Available at: http://www.ats.ucla.edu/stat/stata/faq/svycollin.htm. Accessed July 2011. 34. Manning WG, Mullahy J. Estimating log models: to transform or not to transform? J Health Econ. 2001;20(4):461-494. 35. Hardin JW, Hilbe JM. The negative binomial family: interpreting coefficients. In: Generalized Linear Models and Extensions. 2nd ed. College Station, TX: Stata Press; 2007:218-219. 36. Bernstein J. Changes in Health Care Financing and Organization: Issue Brief: Impact of the Economy on Health Care. Princeton, NJ: Robert Wood Johnson Foundation; 2009. 37. Briesacher BA, Andrade SE, Fouayzi H, Chan KA. Comparison of drug adherence rates among patients with seven different medical conditions. Pharmacotherapy. 2008;28(4):437-443. 38. Salas M, Dyfrig H, Zuluaga A, et al. Costs of medication nonadherence in patients with diabetes mellitus: a systematic review and critical analysis of the literature. Value Health. 2009;12(6):915-922. 39. Mathers CD, Schofield DJ. The health consequences of unemployment: the evidence. Med J Aust. 1998;168(4):178-182. 40. World Health Organization. Adherence to Long Term Therapies: Evidence for Action. Geneva, Switzerland: World Health Organization; 2003. 41. Kessler RC, Turner JB, House JS. Effects of unemployment on health in a community survey: main, modifying, and mediating effects. J Soc Issues. 1988;44(4):59-85. 42. Granados JA. Macroeconomic fluctuations and mortality in postwar Japan. Demography. 2008;45(2):323-343.