 |
|
 |
|
 |
 |
|
 |
 |
 |
 |
 |
 |
 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
,
* Department of Pediatrics, Mount Sinai Medical Center, New York, New York
Department of Psychiatry, Mount Sinai Medical Center, New York, New York
Recanati-Miller Transplant Institute, Mount Sinai Medical Center, New York, New York
|| Department of Pathology, Mount Sinai Medical Center, New York, New York
¶ Department of Psychiatry, University of California Los Angeles, Los Angeles, California
 |
ABSTRACT |
|---|
 TOP ABSTRACT METHODSRESULTSDISCUSSIONCONCLUSIONREFERENCES |
|---|
Methods. We present an analysis of data obtained in the first year of the implementation of an adherence assessment protocol at a pediatric liver transplant clinic in a tertiary medical care center. Data were obtained for children and adolescents who had a liver transplant at least 1 year before the assessments took place. We used 5 adherence detection methods. The 4 subjective methods were self-reported, scaled questionnaires answered by nurses, physicians, caregivers, and patients. For the objective method, a standard deviation (SD) was calculated for tacrolimus blood levels obtained from each patient over time. A higher SD suggests increased variation among patients’ blood levels and hence more erratic medication taking. We also asked the patients and caregivers who is responsible for taking the medications and what are the reasons for not taking them. The medical outcome measures were biopsy-proven rejection episodes, number of biopsies regardless of the results, number of hospital admissions, and number of in-patient days.
Results. An analysis of 81 cases (258 assessments) revealed that the only method that predicted the medical outcome variables (biopsy-proven rejection and number of biopsies) was the SD of medication blood levels. Patients’, clinicians’, and caregivers’ reports were not predictive. Clinicians’ ratings of adherence were not correlated with patients’ or caregivers’. The transition of responsibility for medication taking occurred approximately at the age of 12 years. Forgetfulness was cited as the most common reason for nonadherence by patients and caregivers; medication side effects were not frequently cited.
Conclusions. Our results indicate that clinical impression is not sufficient to determine whether children and adolescents are taking their medications after they have had a liver transplant. An objective assessment method should be used. Interventions targeting adherence should address the child’s increasing role beginning in early adolescence. A clinical protocol incorporating objective assessments of adherence could potentially be implemented in other settings. It could form the basis for the evaluation of efficacy of interventions seeking to improve adherence to medications.
Key Words: liver transplantation • adherence • nonadherence • compliance • noncompliance
Abbreviations: SD, standard deviation
Children who have had a liver transplant need to take immunosuppressant medications to avoid rejection of the transplanted organ. Some children and adolescents do not take their medications as prescribed, even though failure to take these medications is likely to cause the loss of the transplant or even death.1,2 Although adherence to the medication regimen is indisputably an important factor in posttransplantation morbidity1–4 and survival,2 adherence is rarely routinely assessed in a standardized manner in clinical practice. There are no established norms or procedures with regard to which measures should be used to determine whether children are taking their medications in this setting.
There is no widely accepted "gold standard" method for the assessment of adherence in transplant recipients.5,6 Therefore, it is not possible to validate a method that tries to assess whether a child is taking his or her medications against an accepted norm. However, the impact on the child’s health is ultimately the reason for assessing whether a child is adherent. It is possible to examine whether a particular adherence detection method predicts the occurrence of a measure of poor health outcome. In other words, if one determines that a child did not take the medication by using a specific method and this determination is associated with poor health outcome, then the method is clinically useful. However, health outcome is affected by many other factors (eg, disease type, comorbid conditions). Therefore, for examining the performance of a method that purports to detect whether a child is taking the medication, the health outcome measure should be as specifically related to adherence as possible. In the transplant setting, one such outcome measure is the likelihood of a biopsy-proven rejection in individuals who are taking immunosuppressant medications. Rejection of the transplant is, in many cases, a direct outcome in individuals who are not taking prescribed immunosuppressants.2,3 Other, nonspecific medical outcome measures (eg, number of inpatient days within a defined period of time, number of admissions) are expected to be less closely tied to nonadherence. Methods that purport to assess adherence but are correlated with all medical outcome measures or predict only measures that are not closely tied to the effect of the medication that is supposedly being taken are likely to be measures of medical outcome, not measures of adherence. An adherence detection method will be preferred if it is correlated with outcome measures that are closely related to nonadherence (eg, rejection episodes) but is less correlated with less specific ones (eg, rate of admission, length of inpatient stay).
Although pill counts and electronic monitoring devices (that register every time a patient takes a medication out of the bottle) are sometimes used as a way to assess whether a child took the medication, these methods are not free of bias. A patient may open the bottle but discard the medication (rather than take it). These methods would not be able to differentiate between patients who actually took the medication and those who were discarding them. For this reason, methods that use pill counts or their variants are sometimes called indirect measures of adherence.7 An examination of a blood level of a medication, however, seems to be a reasonable and direct way to determine whether a patient took the medication. However, examining a single blood level may not be adequate for clinical purposes, because adherence is a dynamic phenomenon: patients may be taking the medications at times and not taking them at other times.8 Therefore, an adherence detection method should make use of repeated measures. One example is to examine the degree of fluctuation between medication blood levels that are taken over time. We previously evaluated the feasibility of comparing standard deviations (SDs) of consecutive blood levels of tacrolimus in children who had a liver transplant.5 A higher SD and, therefore, more fluctuation over time were deemed to be indicative of erratic adherence. It was shown to be consistent with a panel assessment of adherence in the same subjects. This method assumes that medication blood levels are related to intake. This was shown to be the case for tacrolimus but is not true for cyclosporine.9 Others10,11 have used similar methods that try to capture the degree of fluctuation among medication blood levels in transplant recipients. Nonetheless, it seems as though many practitioners prefer to rely on their clinical impression and on patients’ and parents’ reports when assessing adherence. Indeed, a recent literature review cited self-reports as a leading method for determination of adherence in organ transplant recipients.12 This approach may or may not be adequate, and it has not been evaluated critically in children who have had a transplant. No study to date has tried to evaluate the correlation between these different assessment methods with measures of outcome in children who have had a transplant.
In children, the examination of nonadherence is especially complex because responsibility for taking the medications changes as the child grows older. In very young children, the parents are responsible for the child’s medications. Eventually, the children become primarily responsible. Transition is assumed to occur during adolescence. However, we are not aware of any data-driven reports that sought to identify when children who have had a transplant assume this responsibility. Knowledge about the age of transition could be used to provide children with guidance and support at that age. Also, this knowledge could guide the shift of target in interventions aimed at improving adherence to medications from parent to child.
In January 2002, the Mount Sinai Medical Center’s pediatric liver transplant program started a clinical protocol that aimed to assess in a standard and focused manner whether children were taking their medications. We present an analysis of data that were culled from the first year of this protocol. We examine the degree of correlation between 5 different methods of assessment of adherence and medical outcome in the cohort. We assess the degree of agreement between different informants, including nurses, physicians, patients, and caregivers. We present data that suggest when a child who had a transplant assumes responsibility for taking his or her medications. Finally, we describe patient responses when asked about the reasons for nonadherence. These data provide an indication about which methods could be used by practitioners who care for children who had a solid organ transplant to determine whether these patients are actually taking their medications. The method that we describe could be applied to other populations of chronically ill children whose adherence to medications needs to be monitored.
|
METHODS |
|---|
|
TOPABSTRACTMETHODSRESULTSDISCUSSIONCONCLUSIONREFERENCES |
|---|
Clinicians rated the patient’s adherence at each visit on a similar scale. Clinicians were not aware of the patients’ or caregivers’ ratings when they filled their own ratings but were made aware of the responses later. Each patient’s adherence was rated by 2 clinicians: the nurse who was assigned to examine the patient and the attending physician who supervised the visit and examined the patient.
Information was transferred to and recorded in patients’ charts on a specific "adherence monitoring sheet." Clinicians were able to refer to previous ratings on these sheets at each visit.
Because this was a clinical protocol, all patients were urged to participate and there were no specific exclusion criteria. Our Institutional Review Board approved the analysis of de-identified data, which were collected as a part of our standard practice.
Adherence Detection Methods
Adherence detection methods consisted of 5 different assessments (subjective [all recorded on a 1–4 scale, see above] and objective):
The objective method is the SD of medication blood levels that were taken during 2002 for each patient. Trough blood levels of the immunosuppressant medication (tacrolimus) are obtained in our clinic between 9 and 11 AM on clinic days. The SD method was described elsewhere.5 It is a calculation of the degree of variation among medication blood levels obtained for each patient. A higher SD means a higher degree of difference between individual levels, which suggests less consistent medication taking and, therefore, less adherence.
Medication blood levels may vary as a result of acute illness or in cases in which a more aggressive treatment is implemented. Therefore, we analyzed only medication blood levels that were obtained in the outpatient clinic during routine visits. We did not use levels obtained in the emergency department or in inpatient units, and we also did not use levels obtained immediately after hospital discharge. Hence, levels were obtained in medically stable patients.
Outcome Data
The survival rate in 2002 was 100%, and therefore mortality was not an outcome measure in the present study. We examined 4 different outcome measures, presented in a decreasing order of relatedness to potential nonadherence:
The occurrence of biopsy-proven rejection episodes was considered to be the outcome measure that is most closely related to nonadherence. Although nonadherence is not the only possible reason for rejection, it certainly is a leading cause for it.1–3,12 The number of biopsies performed for any reason and with any finding is less correlated with nonadherence but is still related to it. This is because a biopsy result is inconclusive at times, and it could be hard to distinguish between an autoimmune process and rejection. These cases are still potentially attributable to nonadherence but may not be read by a pathologist as "definite" or "likely" cases of rejection. Number of hospital admissions is less related to nonadherence because there are many potential reasons for admission, only some of which are conceivably related to not taking the medications. Finally, we did not expect the number of inpatient days to be linearly correlated with nonadherence. This is because a rejection episode could resolve relatively quickly. Hence, patients whose main reason for admission was nonadherence may, at times, be less likely to be hospitalized for a longer period compared with patients who had another reason for admission (eg, infection).
We decided not to include measures of liver function (eg, transaminases) among the medical outcome measures. These tests may considerably vary under situations that are not clinically significant and in response to many factors other than nonadherence. Hence, these measures would not have added any specificity to those that we already used. We sought to identify the adherence detection methods that correlate well with the more specific variables (1 and, potentially, 2) but are less or not correlated with the less specific outcome variables (3 and 4).
Other Variables
Age of Transition (Age at Which a Child Is Becoming Responsible for Taking the Medications)
Simply asking at which age the child in a given family has assumed responsibility for taking his or her medications would have introduced a potential recall bias. This method would have relied on patients’ or caregivers’ recollections of an event that might have occurred many years before the assessment. Therefore, we asked each patient and caregiver to write who is presently responsible for taking/administering the medications. In the entire cohort, we examined the reported age ranges at which the child or the parent was said to be responsible for taking the medications.
Reasons for Not Taking the Medications
Because this question was asked by a clinician when needed and hence did not necessarily require reading, we included slightly younger patients—7 years old and above, rather than 8 and above—for the other questions in the questionnaires. Caregivers and children were asked about potential reasons for not taking the medications as prescribed. The answer was recorded on the questionnaires. Options were given as a list of reasons, which were based on results obtained by asking 19 patients and caregivers in a preliminary pilot assessment. We included an open option to tell us about any other reason that the patient or the parent wanted to add. The options were "no reason," "the child forgets," "the caregiver forgets to give it to the child," "side effects of the medication," "it reminds the child that the child is sick," "it takes too much time," "it tastes bad," "doesn’t want to be different," "other (please specify)."
Exclusions From Analysis
We excluded patients who were using cyclosporine (because cyclosporine intake is not linearly related to its blood level after liver transplantation9), patients who had <2 levels of medications recorded in 2002 (because it is impossible to measure the degree of fluctuation between individual levels in these patients), and patients who had a transplant during 2002 (because their initial medical outcome data, before they were stable, would have skewed the results.)
Statistical Analyses
All analyses were performed using SPSS v-10 statistical package and supervised by J.S. Partial correlations were computed for each pair of the 5 adherence detection methods and the 4 outcome variables. This analysis was done while controlling for age, because diagnoses and outcome measures differ for different ages in our cohort. Because we performed 20 related analyses, we applied Holm’s method,13 a variant of the Bonferroni inequality, to correct for the effect of multiple tests of significance. Holm’s method improves the Bonferroni inequality still with no assumptions about the relationships among the tests. The most significant statistic in the set of k comparisons has the same 0.05/k level of significance as the Bonferroni inequality procedure. If it is not significant, then testing stops, because none of the other tests are as good. However, if it is significant, then the next most significant statistic is tested using the 0.05/(k – 1) level of significance, etc (rather than using 0.05/k for all tests).
Pearson correlations for the relationship among pairs of different methods of assessment of adherence were also computed. Again, we applied Holm’s method to correct for multiple tests.
|
RESULTS |
|---|
|
TOPABSTRACTMETHODSRESULTSDISCUSSIONCONCLUSIONREFERENCES |
|---|
Age distribution is presented in Fig 1, and disease categories are presented in Table 1. There were 41 girls and 40 boys in the final cohort. There were 258 questionnaire assessment points for these patients, with an average of 3.2 assessments/patient (range: 1–10 assessments).
|
|
Assessment of Adherence
"Ideally adherent" ratings (ratings that were always "1" without any other rating across all measurements for the same patient) were given to 60.5% (n = 49) of the patients by attending physicians, 51.9% (n = 42) by nurses, and 70.4% (n = 57) by caregivers. Forty-three of 46 eligible patients who were 8 years old or older answered the adherence questionnaires; of these, 69.8% (n = 30) self-reported ideal adherence.
Patients had 2 to 10 blood levels of tacrolimus taken during 2002 (mean: 3.5 assessments/patient). The mean SD of tacrolimus levels in this cohort was 2.48 (range: 0.44–5.60). Figure 2 presents the distribution of SD levels in the cohort.
|
|
|
|
|
|
|
DISCUSSION |
|---|
|
TOPABSTRACTMETHODSRESULTSDISCUSSIONCONCLUSIONREFERENCES |
|---|
We did not have a "gold standard" measure of adherence in the present study. Such "gold standard" does not exist for this population.5,6,12 Hence, the present report should not be interpreted as a validation study of adherence assessment methods. Rather, we aimed to determine which of the methods is the most clinically relevant in that it predicts fairly specific outcome measures.
The SDs of medication levels predicted the medical outcome variable that is likely to be closely related to nonadherence (biopsy-proven rejection episodes) and not the more general outcomes (eg, length of inpatient stay). Hence, SDs are fairly specific predictors of clinically meaningful nonadherence rather than nonspecific predictors of poor outcome. In contrast, subjective methods of assessment of adherence have proved inaccurate in our setting, regardless of who the informer was. Our study reaffirms the previously reported finding that clinicians are frequently not able to detect nonadherence reliably.14
That nurses’ and physicians’ assessments were correlated may be an artifact of the procedure that we used, in which clinicians were not blind to each other’s assessments. Because clinicians had access to the information in patients’ charts, they may have looked at past ratings, and this may have influenced their determinations at each point. This kind of "bias" is usual in clinical practice—clinicians are supposed to look at patients’ histories and charts when making their determinations. Hence, our procedure is a representation of clinical practice, and our results pertain to the usual way in which clinicians perform their assessments.
Physicians’ assessments but not nurses’ correlated with the SD, perhaps because physicians were more likely to base their adherence assessments on the medication blood levels that they reviewed. Because physicians’ assessments did not predict outcome, the correlation between their report and the SD method should not be interpreted as though physicians’ assessments were more "accurate" than nurses’ in this cohort.
Patients’ and caregivers’ reports about taking the medications were not predictive of any of the outcome measures. This is consistent with previously published findings that patient self-reports are not reliable measures of adherence.15–19 Caregivers’ reports did predict the SD levels. The same cautionary note presented above applies here. That caregivers rated adherence in a manner that correlated with the SD of blood levels should be understood in the context of the lack of prediction of outcome measures. Hence, the interpretation is not that caregivers’ assessments are more accurate. Rather, caregivers were probably more likely than patients (in the same way that physicians were more likely than nurses) to use the medication blood levels as an indicator of adherence.
We have found that early adolescence (ages 9–16 years, with an average of approximately 12 years) is typically the age range when responsibility for taking the medications is shifted to the child. This finding is important in the assessment and management of adherence, because the transition period should probably be accompanied by a clinician’s shift in the focus of the education provided to the family. Because early adolescence seems like an early age for this shift to occur, support and guidance should probably be offered to patients who reach this stage.
The reasons that caregivers and children reported for not taking the medications were consistent. Forgetfulness was the most commonly reported reason. However, simple forgetfulness is hardly a likely explanation for severe nonadherence except in the rare case of actual cognitive damage (eg, dementia, encephalopathy). Hence, the reason for such forgetfulness should probably be explored further in practice. Medication side effects were not a commonly cited reason. This is consistent with our previous finding2 in a similar group of patients.
Several limitations of our findings must be acknowledged. Our results with regard to the self-reported adherence are applicable only to a self-report that is obtained through the use of a questionnaire. It is possible that a dedicated, extensive "adherence interview" with each patient/caregiver would have been a better predictor of outcome. We did not investigate this possibility. Such interviews would have taken much more time and resources and therefore would have been of limited relevance to clinical practice. We did not include patients who were receiving cyclosporine, because trough levels are not predictive of intake. However, transplant programs that are extensively using a newer preparation—Neoral—may consider obtaining postintake (not trough) levels to apply the same kind of monitoring that we have done. Postintake levels of Neoral may be good measures of intake.20 Our clinicians do not routinely obtain Neoral postintake levels because we rarely use this medication. Finally, because of the relatively small number of participants (81), a replication of our findings is desirable.
|
CONCLUSION |
|---|
|
TOPABSTRACTMETHODSRESULTSDISCUSSIONCONCLUSIONREFERENCES |
|---|
|
ACKNOWLEDGMENTS |
|---|
|
FOOTNOTES |
|---|
Reprint requests to (E.S.) Department of Psychiatry, Box 1230, Mount Sinai Medical Center, 1 Gustave L Levy Pl, New York, NY 10029. E-mail: eyal.shemesh{at}mssm.edu
|
REFERENCES |
|---|
|
TOPABSTRACTMETHODSRESULTSDISCUSSIONCONCLUSIONREFERENCES |
|---|
This article has been cited by other articles:
 |
|
 |
|
  L. E. Simons and R. L. Blount Identifying Barriers to Medication Adherence in Adolescent Transplant Recipients J. Pediatr. Psychol., August 1, 2007; 32(7): 831 - 844. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
 |
||
|
||
Home | My Pediatrics | Journal Information | Current Issue | Past Issues | Subscriptions & Services | Contact Us Click here for faster international access © 2004 American Academy of Pediatrics. All rights reserved. |
||
|
||
