Measuring Medication: Do Interviews Agree With Medical Record and Pharmacy Data?

Self-reported medication intake vs information from other data sources such as pharmacy records or medical records: Identification and description of existing publications, and comparison of agreement results for publications focusing on patients with cancer - a systematic review

PURPOSE

To identify and describe publications addressing the agreement between self-reported medication and other data sources among adults and, in a subgroup of studies dealing with cancer patients, seek to identify parameters which are associated with agreement.

METHODS

A systematic review including a systematic search within five biomedical databases up to February 28, 2019 was conducted as per the PRISMA Statement. Studies and agreement results were described. For a subgroup of studies dealing with cancer, we searched for associations between agreement and patients' characteristics, study design, comparison data source, and self-report modality.

RESULTS

The literature search retrieved 3392 publications. Included articles (n = 120) show heterogeneous agreement. Eighteen publications focused on cancer populations, with relatively good agreement identified in those which analyzed hormone therapy, estrogen, and chemotherapy (n = 11). Agreement was especially good for chemotherapy (proportion correct ≥93.6%, kappa ≥0.88). No distinct associations between agreement and age, education or marital status were identified in the results. There was little evaluation of associations between agreement and study design, self-report modality and comparison data source, thus not allowing for any conclusions to be drawn.

CONCLUSION

An overview of the evidence available from validation studies with a description of several characteristics is provided. Studies with experimental design which evaluate factors that might affect agreement between self-report and other data sources are lacking.

Entry into primary care-based buprenorphine treatment is associated with identification and treatment of other chronic medical problems

BACKGROUND

Buprenorphine is an effective treatment for opioid dependence that can be provided in a primary care setting. Offering this treatment may also facilitate the identification and treatment of other chronic medical conditions.

METHODS

We retrospectively reviewed the medical records of 168 patients who presented to a primary care clinic for treatment of opioid dependence and who received a prescription for sublingual buprenorphine within a month of their initial visit.

RESULTS

Of the 168 new patients, 122 (73%) did not report having an established primary care provider at the time of the initial visit. One hundred and twenty-five patients (74%) reported at least one established chronic condition at the initial visit. Of the 215 established diagnoses documented on the initial visit, 146 (68%) were not being actively treated; treatment was initiated for 70 (48%) of these within one year. At least one new chronic medical condition was identified in 47 patients (28%) during the first four months of their care. Treatment was initiated for 39 of the 54 new diagnoses (72%) within the first year.

CONCLUSIONS

Offering treatment for opioid dependence with buprenorphine in a primary care practice is associated with the identification and treatment of other chronic medical conditions.

Osteoporosis screening, prevention, and treatment in systemic lupus erythematosus: application of the systemic lupus erythematosus quality indicators

OBJECTIVE

Osteoporosis and fragility fractures are associated with significant morbidity for patients with systemic lupus erythematosus (SLE). New quality indicators (QIs) for SLE advise bone mineral density testing, calcium and vitamin D use, and antiresorptive or anabolic treatment for specific subgroups of patients receiving high-dose steroids.

METHODS

Subjects were participants in the University of California, San Francisco Lupus Outcomes Study, an ongoing longitudinal study of patients with physician-confirmed SLE, in 2007-2008. Patients responded to an annual telephone survey and were queried regarding demographic, clinical, and other health care-related variables. Multiple logistic regression was used to predict receipt of care per the QIs described above.

RESULTS

One hundred twenty-seven patients met the criteria for the formal definitions of the denominators for QI I (screening) and QI II (calcium and vitamin D); 91 met the formal criteria for QI III (treatment). The proportions of patients receiving care consistent with the QIs were 74%, 58%, and 56% for QIs I, II, and III, respectively. In a sensitivity analysis of all steroid users (n = 427 for QI I and II and n = 224 for QI III), rates were slightly lower. Predictors of receiving care varied by QI and by denominator; however, female sex, older age, white race, and longer disease duration were associated with higher-quality care.

CONCLUSION

Bone health-related care in this community-based cohort of SLE patients is suboptimal. Quality improvement efforts should address osteoporosis prevention and care among all SLE patients, especially those receiving high-dose, prolonged steroids.

Hydroxychloroquine treatment in a community-based cohort of patients with systemic lupus erythematosus

OBJECTIVE

In recent years hydroxychloroquine (HCQ) has emerged as a key therapy in systemic lupus erythematosus (SLE). We determined the rates of HCQ use in a diverse, community-based cohort of patients with SLE and identified predictors of current HCQ use.

METHODS

Patients were participants in the University of California San Francisco Lupus Outcomes Study, an ongoing longitudinal study of patients with confirmed SLE. We examined the prevalence of HCQ use per person-year and compared baseline characteristics of users and nonusers, including demographic, socioeconomic, clinical, and health system use variables. Multiple logistic regression with generalized estimating equations was used to evaluate predictors of HCQ use.

RESULTS

A total of 881 patients contributed 3,095 person-years of data over 4 interview cycles. The prevalence of HCQ use was 55 per 100 person-years and was constant throughout the observation period. In multivariate models, the odds of HCQ use were nearly doubled among patients receiving their SLE care from a rheumatologist compared with those identifying generalists or nephrologists as their primary sources of SLE care. In addition, patients with shorter disease duration were more likely to use HCQ, even after adjusting for age and other covariates.

CONCLUSION

In this community-based cohort of patients, HCQ use was suboptimal. Physician specialty and disease duration were the strongest predictors of HCQ use. Patients who are not using HCQ, those with longer disease duration, and those who see nonrheumatologists for their SLE care should be targeted for quality improvement.

Does self-report data on HIV primary care utilization agree with medical record data for socially marginalized populations in the United States?

To test whether self-report data agree with medical record data in marginalized, HIV-infected populations, we collected information about HIV primary care visits over a 6-month period from both sources. Patients were drawn from a large study of engagement and retention in care conducted between 2003 and 2005. Self-report data were collected in face-to-face interviews and medical records were extracted using a rigorous, standardized protocol with multiple quality checks. We found poor overall agreement (weighted kappa = 0.36, 95% confidence interval = 0.28, 0.43). Factors associated with disagreement included younger age (adjusted odds ratio for 20 versus 40 years = 1.25, 95% confidence interval = 0.98, 1.60), non-Hispanic black race/ethnicity (adjusted odds ratio for non-Hispanic blacks versus non-Hispanic whites = 1.48, 95% confidence interval = 1.03, 2.13), lower education (adjusted odds ratio for high school education, GED, or less versus some college or college graduate = 1.43, 95% confidence interval = 0.96, 2.13), and substance use (adjusted odds ratio for any illicit drug/heavy alcohol use in the past 6 months versus no use = 1.39, 95% confidence interval = 1.02, 1.90). These findings do not support a conclusion that unconfirmed self-report data of HIV primary care visits are a sufficient substitute for rigorously collected medical record data in studies focusing on marginalized populations. Use of other data sources (e.g., administrative data), use of other self-reported outcome measures that have better concordance with medical records/administrative data (e.g., CD4 counts), or incorporation of rigorous measures to increase reliability of self-report data may be needed. Limitations of this study include the lack of a true gold standard with which to compare self-report data.

Type and pattern of illicit drug use and access to health care services for HIV-infected people

Approximately 28% of HIV-infected people in treatment in the United States report using illicit drugs. Illicit drug users have poorer course of HIV disease than non-drug users, which is thought to be due to their irregular use of HIV medical services. We examined associations between type (cocaine versus opioids) and pattern of drug use (drug use at baseline, 6-month follow-up, both periods, and nonuse) and health care utilization for a large sample of HIV-infected individuals drawn from a multisite project that evaluated the impact of medical outreach interventions for populations at risk of poor retention in HIV care. Across all types and patterns of drug use, drug users were more likely to have suboptimal ambulatory care, miss scheduled appointments, use the emergency department, have unmet support services needs, and were less likely to take antiretroviral medications. Additionally, while people who started using drugs during the follow-up period and consistently used drugs across both periods differed from nonusers on missed appointments (odds ratio [OR] = 2.20 for starters versus nonusers, OR = 2.92 for consistent users versus nonusers), emergency department use (OR = 4.93 for starters versus nonusers, OR = 2.24 for consistent users versus nonusers), and antiretroviral medication use at follow-up (OR = 0.23 starters versus nonusers, OR = 0.19 for consistent users versus nonusers), those who stopped using drugs after the baseline period did not differ from nonusers. We conclude that health care utilization is poorer for people who use illicit drugs than those who do not, and stopping drug use may facilitate improvements in health care utilization and HIV outcomes for this population.

Prescribing in palliative care as death approaches

OBJECTIVES

To determine how prescribing for comorbid illnesses and symptom control changes during the palliative phase of a terminal illness.

DESIGN

This prospective cohort study explores the relative contribution to prescribing of symptom-specific medications (SSMs) and long-term medications for comorbid medical conditions.

SETTING

Regional consultative palliative care program, Adelaide, South Australia.

PARTICIPANTS

Two hundred sixty consecutive patients, 96% of whom had cancer, who enrolled and subsequently died in a larger randomized trial exploring palliative service delivery.

MEASUREMENTS

Medication and performance data were collected monthly from referral until death (mean 107 days, median 93 days, standard deviation (SD) 103 days, range 11-752 days). Prespecified subgroup analyses of age, performance status, and the baseline use of medications for comorbid medical conditions were performed.

RESULTS

At baseline, the mean total number of medications+/-SD was 4.9+/-2.8 (range 0-16), SSMs was 2.3+/-1.5 (range 0-7), and medications for comorbid medical conditions was 2.6+/-2.4 (range 0-13). As death approached, the total number of medications increased because of SSM prescribing (2.5 more medications, 95% confidence interval (CI)=2.2-2.9; P<.001) with a decrease in medications for comorbid medical conditions (1.1 fewer medications, 95% CI=0.8-1.3; P<.001). There was an increase in the number of medications meeting Beers' criteria for high-risk inappropriate medication use for SSMs (29% to 48%). More SSMs were prescribed in people with better performance status, and older participants took more medications for comorbid medical conditions.

CONCLUSION

Prescribing changes as life-limiting illnesses progress, with older people taking more medications. Medications for comorbid medical conditions should be reviewed in the context of their original therapeutic goals.

Prospective cohort study into post-disaster benzodiazepine use demonstrated only short-term increase

OBJECTIVES

Benzodiazepines are typically prescribed for anxiety and insomnia, two complaints often reported after disasters. Benzodiazepines can cause mental or physical dependence, especially when taken for a long time. This study aims at evaluating benzodiazepine use in a disaster-stricken community with the help of computer-based records.

STUDY DESIGN AND SETTING

This prospective cohort study covers a period of 4 years. For every patient, predisaster baseline data are available. Multilevel regression is applied to study differences in benzodiazepine use in 496 patients whose children were involved in the Volendam café fire on January 1, 2001 compared with 1,709 community controls, and 4,530 patients from an unaffected cohort.

RESULTS

In community controls and patients from the unaffected cohort, benzodiazepine use remained stable in the course of the years. In the first year postfire, parents of disaster victims were 1.58 times more likely to use benzodiazepines than community controls (95% confidence interval 1.13-2.23). With regard to long-term use, differences between community controls and parents were statistically nonsignificant.

CONCLUSIONS

In the studied community, benzodiazepines were predominantly prescribed as a short-term intervention. Clinical guidelines that advocate a conservative prescription policy were well adhered to.

A Comparison of HIV Health Services Utilization Measures in a Marginalized Population

BACKGROUND

In studies examining the use of human immunodeficiency virus (HIV) health services, researchers often use subjects' self-reported measures. Agreement between a subject's self-reports and medical records in marginalized populations is uncertain, yet important to understand, as this population is disproportionately affected by HIV.

METHODS

We sought to examine agreement between self-report and medical record health care utilization measures. Using a cross-sectional study, we studied 428 unstably housed HIV-infected adults in New York City. Self-reported data were collected from Audio Computer-Assisted Self-Interviews, and medical record data from health care providers' and facilities' ambulatory medical records. Agreement for a 6-month period was compared for ambulatory visits (0, 1, >or=2), HIV medications (antiretroviral therapy, opportunistic infection prophylaxis), whether CD4 counts and viral loads were performed and their values (CD4: <200, 200-500, >500 cells/mm; Viral load: undetectable, detected).

RESULTS

Agreement between self-report and medical records was 55.2% (kappa=0.12) for visits, and 68.2-79.1% (kappa=0.27-0.48) for medications. Agreement on whether laboratory tests were performed was 62.3-65.7% (kappa=0.11-0.14), whereas agreement on laboratory values was 77.6-79.3% (kappa=0.52-0.70). Most disagreement resulted in greater number of self-reported visits, use of medications, and laboratory tests compared with medical record data.

CONCLUSIONS

Among HIV-infected marginalized individuals, agreement between self-report and medical records was poor for ambulatory visits, poor to fair for medication use, and poor for laboratory tests performed. However, agreement for CD4 count value was substantially better. These findings have implications on health services research in marginalized populations that relies only on self-report or medical record data. This study underscores the importance of understanding how self-reported and medical record data are correlated in marginalized populations.

Language, literacy, and communication regarding medication in an anticoagulation clinic: a comparison of verbal vs. visual assessment

Despite the importance of clinician-patient communication, little is known about rates and predictors of medication miscommunication. Measuring rates of miscommunication, as well as differences between verbal and visual modes of assessment, can inform efforts to more effectively communicate about medications. We studied 220 diverse patients in an anticoagulation clinic to assess concordance between patient and clinician reports of warfarin regimens. Bilingual research assistants asked patients to (1) verbalize their prescribed weekly warfarin regimen and (2) identify this regimen from a digitized color menu of warfarin pills. We obtained clinician reports of patient regimens from chart review. Patients were categorized as having regimen concordance if there were no patient-clinician discrepancies in total weekly dosage. We then examined whether verbal and visual concordance rates varied with patient's language and health literacy. Fifty percent of patients achieved verbal concordance and 66% achieved visual concordance with clinicians regarding the weekly warfarin regimen (P < .001). Being a Cantonese speaker and having inadequate health literacy were associated with a lower odds of verbal concordance compared with English speakers and subjects with adequate health literacy (AOR 0.44, 0.21-0.93, AOR 0.50, 0.26-0.99, respectively). Neither language nor health literacy was associated with visual discordance. Shifting from verbal to visual modes was associated with greater patient-provider concordance across all patient subgroups, but especially for those with communication barriers.Clinician-patient discordance regarding patients' warfarin regimen was common but occurred less frequently when patients used a visual aid. Visual aids may improve the accuracy of medication assessment, especially for patients with communication barriers.

The importance of establishing regimen concordance in preventing medication errors in anticoagulant care

Miscommunication between patients and providers can have serious consequences, especially where medications are concerned. We examined whether regimen discordance between patient and provider, a possible result of miscommunication, contributes to unsafe medication management. We studied 220 patients taking warfarin in an anticoagulation clinic to characterize two medication assessment methods. We measured (1) adherence by asking patients to report any missed doses and (2) concordance between patients' and providers' reports of warfarin regimens. We categorized patients as having regimen adherence if they missed no doses, and concordance if there was patient-provider agreement in weekly dosage. We characterized anticoagulant outcomes as unsafe if international normalized ratio (INR) values were <2.0 (at risk for thrombosis) or >4.0 (at risk for hemorrhage), and explored relationships among adherence, concordance, and anticoagulant outcomes. One hundred fifty-five patients (71%) reported no missed doses during the prior 30 days. Poor adherence was associated with underanticoagulation (AOR 2.33, 1.56-3.45), but not overanticoagulation (AOR 1.36, 0.69-2.66). One hundred ten patients (50%) reported regimens discordant with clinicians' report. There was no relationship between patients' reports of adherence and concordance. Among adherent patients, discordance was associated with underanticoagulation (AOR 1.67, 1.00-2.78) and overanticoagulation (AOR 3.44, 1.32-9.09). Discordance regarding warfarin regimens is common and places patients at risk for adverse events. To promote safe and effective care, clinicians should separately determine adherence and regimen concordance during routine medication assessments. Systems need to be developed to ensure concordance in medication regimens.

Lessons learned about implementing research evidence into clinical practice. Experiences from VA QUERI

The mission of the Veterans Health Administration's (VHA) quality enhancement research initiative (QUERI) is to enhance the quality of VHA health care by implementing clinical research findings into routine care. This paper presents lessons that QUERI investigators have learned through their initial attempts to pursue the QUERI mission. The lessons in this paper represent those that were common across multiple QUERI projects and were mutually agreed on as having substantial impact on the success of implementation. While the lessons are consistent with commonly recognized ingredients of successful implementation efforts, the examples highlight the fact that, even with a thorough knowledge of the literature and thoughtful planning, unexpected circumstances arise during implementation efforts that require flexibility and adaptability. The findings stress the importance of utilizing formative evaluation techniques to identify barriers to successful implementation and strategies to address these barriers.

Invoices and farmer interviews for vaccine-exposure measurement in turkey-broiler production

Exposure measurement in pharmaco-epidemiological studies can be based on various sources that do not always concur. However, reliability is an important criterion when selecting the method used to assess exposure and interpreting the results obtained. An analysis based on invoices might be more informative and more accurate to assess vaccines exposure (yes/no) in turkey broiler production than a questionnaire administered to farmers, which is nevertheless more feasible and less time-consuming. We compared the two methods to assess vaccination exposure in 239 turkey broiler flocks reared in 129 farms in 2000-2001. The agreement (crude agreement and kappa) was calculated, and association between discrepancy and farm and flock characteristics was investigated. Marek's-disease vaccine, Newcastle-disease vaccine, turkey haemorrhagic-enteritis vaccine and turkey-rhinotracheitis vaccine exposures were reported on the questionnaire for 2.1, 27.6, 93.0, and 98.3% of flocks, respectively, and for 2.1, 29.3, 89.4, and 86.6%, of invoices for the flocks studied, respectively. A discrepancy was observed in 24.9% of flocks. A discrepancy was observed more frequently in specialised farms without any other animal production (OR = 3.6; CI = 1.5, 8.9) and when the farmer did not know whether vaccination had been performed in the hatchery (OR = 7.1; CI = 2.6, 19.7).

The Epidemiologic Approach to Pharmacogenomics

The epidemiologic approach enables the systematic evaluation of potential improvements in the safety and efficacy of drug treatment which might result from targeting treatment on the basis of genomic information. The main epidemiologic designs are the randomized control trial, the cohort study, and the case-control study, and derivatives of these proposed for investigating gene-environment interactions. However, no one design is ideal for every situation, and methodological issues, notably selection bias, information bias, confounding and chance, all play a part in determining which study design is best for a given situation. There is also a need to employ a range of different designs to establish a portfolio of evidence about specific gene-drug interactions. In view of the complexity of gene-drug interactions, pooling of data across studies is likely to be needed in order to have adequate statistical power to test hypotheses. We suggest that there may be opportunities (i) to exploit samples from trials already completed to investigate possible gene-drug interactions; (ii) to consider the use of the case-only design nested within randomized controlled trials as a possible means of reducing genotyping costs when dichotomous outcomes are being investigated; and (iii) to make use of population-based disease registries that can be linked with tissue samples, treatment information and death records, to investigate gene-treatment interactions in survival.

Does racial concordance between HIV-positive patients and their physicians affect the time to receipt of protease inhibitors?

BACKGROUND

Compared to whites, African Americans have been found to have greater morbidity and mortality from HIV, partly due to their lower use of effective antiretroviral therapy. Why racial disparities in antiretroviral use exist is not completely understood. We examined whether racial concordance (patients and providers having the same race) affects the time of receipt of protease inhibitors.

METHODS

We analyzed data from a prospective, cohort study of a national probability sample of 1,241 adults receiving HIV care with linked data from 287 providers. We examined the association between patient-provider racial concordance and time from when the Food and Drug Administration approved the first protease inhibitor to the time when patients first received a protease inhibitor.

RESULTS

In our unadjusted model, white patients received protease inhibitors much earlier than African-American patients (median 277 days compared to 439 days; P < .0001). Adjusting for patient characteristics only, African-American patients with white providers received protease inhibitors significantly later than African-American patients with African-American providers (median 461 days vs. 342 days respectively; P < .001) and white patients with white providers (median 461 vs. 353 days respectively; P= .002). In this model, no difference was found between African-American patients with African-American providers and white patients with white providers (342 vs. 353 days respectively; P > .20). Adjusting for patients' trust in providers, as well as other patient and provider characteristics in subsequent models, did not account for these differences.

CONCLUSION

Patient-provider racial concordance was associated with time to receipt of protease inhibitor therapy for persons with HIV. Racial concordance should be addressed in programs, policies, and future racial and ethnic health disparity research.

Direct measurement of health care costs

Cost identification is fundamental to many economic analyses of health care. Health care costs are often derived from administrative databases. Unit costs may also be obtained from published studies. When these sources will not suffice (e.g., in evaluating interventions or programs), data may be gathered directly through observation and surveys. This article describes how to use direct measurement to estimate the cost of an intervention. The authors review the elements of cost determination, including study perspective, the range of elements to measure, and short-run versus long-run costs. They then discuss the advantages and drawbacks of alternative direct measurement methods such as time-and-motion studies, activity logs, and surveys of patients and managers. A parsimonious data collection effort is desirable, although study hypotheses and perspective should guide the endeavor. Special reference is made to data sources within the Department of Veterans Affairs (VA) health care system.

Longitudinal study of mental health and psychosocial predictors of medical treatment adherence in mothers living with HIV disease

Cross-sectional studies to date that examine psychosocial correlates of antiretroviral adherence have insufficiently addressed the challenges of long-term adherence. This longitudinal study examined mental health, substance abuse, and psychosocial predictors of long-term adherence to antiretroviral medications and medical appointments among HIV-seropositive mothers recruited from an infectious disease clinic of a large urban medical center. Individual interviews were conducted at baseline and two follow-up points, 8 to 18 months after enrollment. Based on a model of health behavior, we examined psychiatric and psychosocial predictors of adherence to antiretroviral medications and medical appointments over time. Presence of a psychiatric disorder, negative stressful life events, more household members, and parenting stress were significantly associated with both missed pills and missed medical appointments at follow-up. Baseline substance abuse was associated with missed pills at follow-up and lack of disclosure to family members at baseline was associated with missed medical appointments at follow-up. These findings suggest that interventions that integrate mental health, substance abuse and medical care may be important to improving the medical adherence and health of HIV-seropositive women, particularly in multistressed populations with substantial caregiving and other life demands.