10898

2013

AOPXXX10.1177/1060028013510898Annals of PharmacotherapyWillson et al

Research Report

Medication Regimen Complexity and Hospital Readmission for an Adverse Drug Event

Annals of Pharmacotherapy 2014, Vol. 48(1) 26­–32 © The Author(s) 2013 Reprints and permissions: sagepub.com/journalsPermissions.nav DOI: 10.1177/1060028013510898 aop.sagepub.com

Megan N. Willson, PharmD, BCPS1, Christopher L. Greer, BPharm2, and Douglas L. Weeks, PhD2

Abstract Background: Adverse drug events (ADEs) are costly, dangerous, and often preventable. Little is known about the link between medication regimen complexity and rehospitalization as a result of an ADE. Objective: The objective of this study was to compare admission and discharge medication regimen complexity in 2 cohorts: patients readmitted for an ADE within 30 days and patients not readmitted for an ADE. Methods: The study used a retrospective parallel-group case-control design. Participants from 4 urban acute care hospitals were included in the revisit cohort if they were rehospitalized within 30 days as a result of an adverse event coded as accidental poisoning. The no-revisit cohort was formed by randomly sampling patients with the same disease classification codes as the revisit group but without history of a readmission within 30 days. Complexity of medication regimens at the initial admission and discharge was quantified with the medication regimen complexity index (MRCI). Results: The revisit group comprised 92 individuals and the no-revisit group, 228. The revisit group had a significantly higher MRCI score at admission and discharge than the norevisit group (all P < .005). Receiver operating characteristic curves, used to determine a potential MRCI cutoff score for risk of an ADE, revealed MRCI scores of 8 or greater to optimally predict increased risk for readmission caused by an ADE. Conclusions: Complex medication regimens at hospital admission are predictive of rehospitalizations for ADEs. This finding suggests that medication regimen complexity be considered as a target for interventions to decrease the risk for readmission. Keywords medication regimen complexity, adverse drug event, acute care, hospital readmission

Introduction A medication error is “any preventable event that may cause or lead to inappropriate medication use or patient harm while the medication is in the control of the health care professional, patient, or consumer.”1 Medication errors are common occurrences that have increased in frequency over the past 25 years.2 They can be compounded when patients receive care through multiple practitioners and when transitioning between different health care settings, with most medication errors occurring within the first 7 days a patient self-manages his or her medications following hospital discharge.3 An adverse drug event (ADE) is defined as an “injury resulting from the use of a drug.”4 ADEs are very prevalent in the ambulatory care setting, with approximately a quarter believed to be preventable.5-10 The prevalence of ADEs differs between age groups, with elderly individuals being the most likely to experience an ADE.3,9-11 The seriousness of

ADEs can range from minimal and managed in the ambulatory care setting to severe, resulting in hospitalization or death.3,8,11,12 Repeat hospital admissions for ADEs are estimated to be in excess of 17%, thus indicating lingering medication-related problems in many patients.13 The economic burden from ADE-related morbidity and mortality has been estimated to be $30 to $130 billion annually in the United States.14 Because many ADEs are considered preventable, costs and complications associated with ADEs represent major quality and cost-efficiency concerns for the US health care system. 1

Washington State University, Spokane, WA, USA St Luke’s Rehabilitation Institute, Spokane, WA, USA

2

Corresponding Author: Megan N. Willson, PharmD, BCPS, Department of Pharmacotherapy, College of Pharmacy, Washington State University, PO Box 1495, Spokane, WA 99210, USA. Email: [email protected]

Downloaded from aop.sagepub.com at NORTHWESTERN UNIV LIBRARY on September 9, 2014

27

Willson et al The seriousness of ADEs has led researchers to focus on risk factors for ADEs. Among patient-related risk factors are advanced age, comorbid conditions, patient ethnicity, mental illness diagnosis, medication misuse, poor cognitive function, insurance status, and poor health literacy.9,13,15-18 Several medication-related risk factors are reported to be independently associated with out-of-hospital ADEs, such as class of medication and number of medications.4,7,11,12,19,20 With medication misuse being cited as a known risk factor for ADEs, understanding reasons for this misuse could help decrease the rate of medication-related problems. One factor that may contribute to misuse is the complexity of the medication regimen that the patient and/or caretaker must manage following hospital discharge. A recent study by Mansur et al21 demonstrated that an increase in medication regimen complexity increased the rate of nonadherence. Yet little is known about the link between medication regimen complexity and medication misuse leading to hospitalization precipitated by an ADE. Also, the link between the number of medications and the resultant increase in ADEs needs to be explored further to define a potential relationship with medication regimen complexity. Various definitions have been used for medication regimen complexity in the literature. Some studies have only accounted for number of medications or number of tablets consumed per day, whereas other studies have accounted for complexity by using a combination of the number of medications and doses per day.16,18,22 The inconsistent definitions prevent conclusions from being drawn regarding how regimen complexity is associated with ADEs. In addition, complexity of the medication regimen should also consider features such as frequency of administration, additional medication directions, and route of administration for each medication. The medication regimen complexity index (MRCI) is one such tool that accounts for each of these features in quantifying regimen complexity.23 Using a more valid method of capturing regimen complexity could enhance the ability to identify a regimen that is too complex for an individual to self-manage, leading to adjustments to the regimen, which could potentially avoid ADEs. This study used the MRCI to examine the association between discharge medication regimen complexity and subsequent readmission to the hospital for an ADE. The primary objective of this study was to compare the MRCI of individuals who had a rehospitilization for an ADE with those who did not. The secondary objective was to determine if the MRCI could predict a rehospitalization.

Methods

medical records databases from 4 separate hospitals. For the period September 1, 2009 to July 31, 2010, electronic medical record data were obtained for all patients with any discharge diagnosis who were subsequently rehospitalized within 30 days (up to August 31, 2010) with an admission diagnosis for an adverse event caused by accidental poisioning by medications. International Classification of Diseases, 9th revision (ICD-9) diagnostic codes 960.x through 977.x and external cause codes (E-codes) E850.x through E858.x were used to identify a patient record for inclusion in the data set. The control cohort was selected from the same time period through stratified random sampling. The control cohort patients were proportionately sampled from within strata formed to match the rehospitalized controls by sex, age group (5-year intervals), and initial hospitalization discharge diagnosis. For each readmission, 2 to 3 control patients were identified. The study protocol was reviewed and approved by the local institutional review board.

Measurements Data elements obtained were patient age, sex, length of stay (LOS) for each hospitalization, race, admission source, admission diagnostic code, ICD-9 codes for primary etiological diagnosis and secondary diagnoses, discharge disposition, and the admission and discharge medication list. Complexity of the medication lists was measured by the MRCI, a tool shown to have good interrater and test-retest reliability for the total derived score and scores on each of 3 individual compontents: dosage form (section A), dosing frequency (section B), and additional directions required for administration (section C).23 The MRCI provides a weighted complexity score in which each medication in the regimen contributes to the score on each component. Section A (dosage form) quantifies complexity based on route of administration. All drugs are reviewed, and a score is given only once for each different route of administration. Section B (dosing frequency) quantifies complexity by number of medications and frequency of administration. For example, if medication A is given once a day, it would be scored as 1, but if medication B is given twice daily, it would receive a score of 2. Section C (additional directions) contributes to the complexity score for additional directions the patient is expected to follow, such as multiple doses at a given time, specific timing (ie, with meals or at bedtime), or range dosing. The total MRCI score is calculated from summing scores for each section.

Analytical Plan

Design and Sample A retrospective parallel-group case-control design was used in which patient-level information was extracted from

Continuous outcomes were characterized by means and standard deviations, with nominal outcomes characterized by frequencies and proportions. Differences among groups

Downloaded from aop.sagepub.com at NORTHWESTERN UNIV LIBRARY on September 9, 2014

28

Annals of Pharmacotherapy 48(1)

for continuous measures were examined with independentgroup t tests. For continuous measures, the standardized mean difference (difference between revisit and no-revisit group means divided by the no-revisit group standard deviation) was used to express the magnitude of differences among groups on a relative scale. For the group experiencing a revisit within 30 days, we also analyzed differences between MRCI scores and total number of medications from the index hospitalization and those from the rehospitalization with paired group t tests. Differences among groups for nominal data were examined with Fisher’s exact test for doubly dichotomous relationships and χ2 when nominal data included at least 3 levels. Receiver operating characteristic (ROC) curves were developed for index hospitalization number of medications, MRCI scores, and separate components of the MRCI scores at admission and discharge to determine which measure resulted in the greatest area under the curve (AUC) for detecting revisit risk (sensitivity) versus detecting when a revisit would not occur (specificity). An AUC value of 1.0 would indicate perfect ability to detect when a revisit would occur, whereas an AUC score of 0.5 would indicate detection ability that was no greater than chance. For the measure with the greatest AUC, ROC analyses were used to obtain a cutoff score that optimized classification accuracy while adequately balancing sensitivity and specificity. Whereas there are no established thresholds for sensitivity and specificity, we considered values at 80% or above to be acceptable for establishing positive and negative predictive adequacy of a measure. The value of 80% was chosen as the threshold for sensitivity because of the complexity of the predictors for hospital readmission. However, in the tradeoff between true positive and true negative classification ability, we considered the ability to truly detect when a revisit would occur (positive classification relative to the cutoff score) to be of greater importance to the health of the patient and to costs for a revisit for an ADE than accuracy in detecting when a revisit would not occur (true negative classifications). We chose the cutoffs so that they exceeded our criterion of 80% for accurate detection of a revisit (sensitivity) while maximizing specificity. The identified cutoff score was used to create a binary variable that classified MRCI scores as at or above the cutoff score versus below the cutoff score. The binary score was further used as a predictor variable in logistic regression to obtain the odds ratio (OR) for a revisit resulting from an ADE based on the cutoff score. The logistic regression results we report included the binary cutoff score as the sole predictor, but we conducted additional logistic regression analyses that included sex, age, and LOS as covariates with the binary cutoff score. Addition of the covariates did not significantly moderate the odds of a readmission (and none of the covariates had significant ORs themselves) over inclusion of the binary cutoff score variable as the sole predictor.

All analyses were conducted with SPSS v19.0, and type I error rate was established at P < .05.

Results Comparison of the Revisit Group and No-Revisit Group for the Index Stay The distribution of male to female patients among groups was 41% to 59% for the revisit group and 42% to 58% for the no-revisit group; these proportions did not differ (P = .47). Additional descriptive statistics for the revisit and norevisit groups for the index hospitalization are given in Table 1. Group differences were not significant for age, LOS for the index hospitalization, or change in MRCI from admission to discharge for the index stay (all P > .18). The frequency of patients per group in each major ICD-9 classification is listed in the Appendix. The number of medications listed at admission and the MRCI for those medications were significantly greater for the index stay for the group that would eventually experience a revisit than for the no-revisit group (both P < .001). The standardized mean difference among groups was 0.71 for number of admission medications and 0.75 for admission MRCI. Likewise, the number of medications on the discharge list and the MRCI for those medications were significantly greater for the index stay for the revisit group than the no-revisit group (both P < .001). The standardized mean difference among groups was 0.65 for number of discharge medications and 0.67 for discharge MRCI. To examine whether groups differed significantly on specific aspects of complexity, we used independent group t tests to compare group means for the 3 separate MRCI components for admission and discharge MRCI scores; means per MRCI section are displayed in Table 1. The revisit group had significantly larger scores, indicating greater complexity, for each component of the MRCI at admission and discharge (all P < .005). The largest standardized mean difference among groups was for section B of the admission medication MRCI, at 0.75.

Accuracy of Index MRCI for Predicting Likelihood of a Revisit The AUC for each index hospitalization measure is displayed in Table 2. All AUC statistics were significant at P < .001 indicating that the ability to detect a revisit from the given cutoff score exceeded chance. However, the admission medications MRCI and the score for section B of the admission medications MRCI were equally the most effective measures at distinguishing among those who would or would not have a revisit for an ADE (both AUC = 0.694). Cutoff scores for each of these measures were derived from the ROC analyses, which displayed all combinations of

Downloaded from aop.sagepub.com at NORTHWESTERN UNIV LIBRARY on September 9, 2014

29

Willson et al

Table 1.  Descriptive Statistics for the Index Hospitalization for the Group Experiencing a Hospital Encounter for an ADE Within 30 Days of Index Discharge (Revisit Group, n = 92) and the Group Without a Hospital Encounter for an ADE Within 30 Days of Index Discharge (No-Revisit Group, n = 228).a Measure

Group

Age

Revisit No revisit Revisit No revisit Revisit No revisit Revisit No revisit Revisit No revisit Revisit No revisit Revisit No revisit Revisit No revisit Revisit No revisit Revisit No revisit Revisit No revisit Revisit No revisit Revisit No revisit

Length of stay Number of medications reported at admission Admission medications MRCI Number of medications on discharge list Discharge medications MRCI Admission-to-discharge MRCI change Section A of the MRCI for admission medications Section B of the MRCI for admission medications Section C of the MRCI for admission medications Section A of the MRCI for discharge medications Section B of the MRCI for discharge medications Section C of the MRCI for discharge medications

Mean

Standard Deviation

50.29 49.39 4.08 4.06 10.64 6.78 27.38 16.21 11.21 7.78 30.11 20.42 2.74 4.21 5.41 3.20 16.78 9.78 4.03 2.33 5.65 3.86 18.95 12.59 3.99 2.76

P Value for Group Comparison

17.21 17.46 4.69 3.79 6.43 5.44 17.78 14.84 6.65 5.31 18.82 14.49 9.07 8.74 4.87 3.72 11.27 9.32 3.46 2.75 5.12 3.82 11.98 9.06 3.22 2.74

.673   .979  

Medication regimen complexity and hospital readmission for an adverse drug event.

Adverse drug events (ADEs) are costly, dangerous, and often preventable. Little is known about the link between medication regimen complexity and reho...
277KB Sizes 0 Downloads 0 Views