and methodological issues relevant to measuring diabetes regimen adherence ... compliance with diabetes control. ... prevalence of adherence problems exists.
S E R I E S
Methodological Issues in Diabetes Research Measuring adherence SUZANNE BENNETT JOHNSON, PHD
The prevalence of nonadherence in IDDM and NIDDM populations and conceptual and methodological issues relevant to measuring diabetes regimen adherence are reviewed. The prevalence of nonadherence varies across the different components of the diabetes regimen, during the course of the disease, and across the patient's life span. Although prevalence rates might be expected to differ between IDDM and NIDDM populations, this rarely has been evaluated. Conceptual problems in defining and measuring adherence include: the absence of explicit adherence standards against which the patient's behavior can be compared; inadvertent noncompliance attributable to patient-provider miscommunication and patient knowledge/skill deficits; the behavioral complexity of the diabetes regimen; and the confounding of compliance with diabetes control. Methods for measuring adherence include: health status indicators, provider ratings, behavioral observations, permanent products, and patient self-reports, including behavior ratings, diaries, and 24-h recall interviews. A measurement method should be selected on the basis of reliability, validity, nonreactivity, sensitivity to the complexity of diabetes regimen behaviors, and measurement independence from the patient's health status. The timing of measurements should be based on the stability of adherence behaviors and temporal congruity with other measures of interest (e.g., indexes of metabolic control). Directions for future research and suggestions for clinical practice are provided.
s with many chronic diseases, diabetes places numerous behavioral demands on the patient. Most are required to take daily medication either orally or by injection. Timing and fre-
quency of meals is deemed important as is the type of foods consumed; foods high in concentrated sweets and fats are to be avoided. Exercise is thought to have a beneficial effect by improving in-
FROM THE DEPARTMENTS OF PSYCHIATRY, PEDIATRICS, AND CLINICAL AND HEALTH PSYCHOLOGY, UNIVERSITY OF FLORIDA HEALTH SCIENCES CENTER, GAINESVILLE, FLORIDA. ADDRESS CORRESPONDENCE AND REPRINT REQUESTS TO SUZANNE BENNETT JOHNSON, PHD, UNIVERSITY OF FLORIDA HEALTH SCIENCE CENTER, DEPARTMENT OF PSYCHIATRY, P.O. Box,
FL. RECEIVED FOR PUBLICATION 26 AUGUST 1991
AND ACCEPTED IN REVISED FORM 31 MARCH
I D D M , INSULIN-DEPENDENT DIABETES MELLITUS; N I D D M , NON-INSULIN-DEPENDENT DIABETES MELLITUS;
AMERICAN DIABETES ASSOCIATION.
THIS ARTICLE IS ONE OF A SERIES PRESENTED AT THE MEETING ON THE BEHAVIORAL ASPECTS OF DIABETES MELLITUS.
sulin action. However, exercise should be timed carefully in relationship to meals so as to avoid hypoglycemia. Home blood glucose testing, as often as 2 - 4 times/day, often is required, along with follow-up urine ketone tests when blood glucose test results are high. Although the physician serves as a consultant, prescribing medication and making dietary, exercise, and testing recommendations, the ultimate responsibility for daily care rests with the patient. The myriad of behaviors required for daily diabetes care represents a true behavioral challenge. Perhaps it should come as no surprise that many patients fail to meet this challenge successfully.
PREVALENCE OF NONADHERENCE— Watkins et al (1) published one of the first studies documenting the extent of noncompliance in adults with diabetes. Of the patients studied, > 50% were making insulin dosage errors, ~66% were testing incorrectly, and —75% were judged unacceptable in terms of quality, quantity, and timing of meals. Although it is difficult to make across-study comparisons because of differences in definitions of adherence as well as the quality and type of measures used, the consensus is that a high prevalence of adherence problems exists in both IDDM and NIDDM populations. Some evidence suggests that published prevalence rates may underestimate the extent of the problem, as some patients attempt to appear more compliant or in better metabolic control than is actually the case. Although patient reports of noncompliant behaviors are usually accurate, patient reports of highly compliant behavior are considered suspect (2). Examples of biased reporting have been documented within the glucose testing literature. Not only are patient glucose testing results frequently inaccurate, but errors are often in the direction of underestimating glucose levels (3-6). Biased reporting appears to be less common in adult populations, com-
DIABETES CARE, VOLUME 15, NUMBER 11, NOVEMBER 1992
ADA recommends that 55-60% of total calories should be comprised of carbohydrates (32). Studies consistently report carbohydrate consumption rates of 90% (14,18,19,21-24). consumption is common in NIDDM Some evidence indicates that patients are populations (17,39,40). more adherent with insulin administraExercise adherence levels have tion than oral medication (23), although not been well documented. Arey et al. this is not a consistent finding (22). Al- (17) noted that only 31% of IDDM and though many patients may take their 53% of NIDDM patients adhered to exmedication, they are less adherent with ercise prescriptions. When asked, adult the timing of its administration (19,20, patients report following their exercise 25-28). prescriptions only 50% of the time (14), Numerous studies have focused exercising an average of 2 - 4 times/wk on dietary behaviors, primarily within (19,31,41). Exercise may be more fre1DDM populations. Of adult IDDM pa- quent in child and adolescent samples tients, —60-75% fail to eat the appro- (19,26-28); rates as high as 1.8 times/ priate types or amount of foods at the day have been reported. correct time (1,12). When asked, paThe literature on urine or blood tients acknowledge following dietary glucose testing is more substantive. Once prescriptions only 60-70% of the time again, most studies have focused on (14,29). When unobtrusively observed, IDDM populations. As is the case with children with diabetes exhibit dietary de- the adherence literature in general, prevviation rates of —35% (30). The number alence rates for nonadherence vary of meals and snack consumed has been across different methods of measurethe focus of only a few studies. Children ment. When the focus is on the number and adolescents average 5 meals or of IDDM patients who fail to regularly snacks/day (26-28). NIDDM adults skip test, rates vary from 36 to 82% 1 prescribed meal or snack/day, usually (10,12,18); as many as 34% of NIDDM the noon meal (31). Studies of types of adults do not test at all (25). When the food consumed suggest that IDDM pa- focus is on the percentage of prescribed tients eat far too many fats and too few tests actually conducted, rates vary from carbohydrates. Although the ADA rec- 38 to 78% (9,14,17,19,24,42). Studies ommends that 45% in ports range from 0 to 3 tests/day children and adolescents (26-28,33,34). (6,8,9,26-28,29,31). Fat consumption rates (39-41%) are somewhat lower in adults (23,35) but are Developmental issues still well above recommended levels. Each stage in the human life span is While fat consumption is too high, car- associated with different cognitive and bohydrate consumption is too low. The social capabilities as well as different bi-
pared with childhood or adolescent populations, or when blood glucose testing meters with memory capacity (that is known to the patient) are used (5,7-11).
DIABETES CARE, VOLUME 15,
NUMBER 11, NOVEMBER
ological changes, personal goals, and social demands. Adherence is likely to be affected as the patient enters and moves through each stage of human development. Young children with diabetes often show significant skill deficits when attempting to administer insulin or test glucose, as they have not yet developed the necessary cognitive and motor skills to perform these tasks accurately (43,44). Adolescents are more knowledgeable and skillful than their younger counterparts; yet, during this developmental period, their strivings for independence, reduced parental supervision, and peer influences may have detrimental effects on patient adherence. Indeed, several studies have documented greater adherence problems among adolescents compared with younger children (26,27, 30,36,45-47). This issue rarely has been examined in adult populations (15). Several studies suggest that older adults may be more compliant with medication administration and glucose testing (2,17), whereas younger adults may be more compliant with exercise (19).
Course of disease Few studies have examined change in adherence over the course of the disease. In children and adolescents with IDDM, adherence appears to be highest at diagnosis and to deteriorate thereafter (18,46,47). We do not know whether a similar pattern occurs in adult NIDDM populations. Some evidence indicates that introducing a new management technique (such as a meter for blood glucose monitoring) or sending a child to diabetes camp results in a temporary increase in adherence. However, unless the home environment is changed to support the increased adherence, patient behavior often gradually declines to previous levels (28,42). Researchers have not examined change in adherence associated with change in health status, such as the onset of complications.
Prevalence rates in IDDM versus NIDDM populations Although researchers often include both IDDM and NIDDM patients in their study population (1,15,24,41), direct comparison of adherence rates in IDDM and NIDDM has been rare. Arey et al. (17) found few differences between the IDDM and NIDDM patients they studied; the NIDDM adults who were prescribed insulin were more adherent with the timing of their insulin administration than their IDDM counterparts. Pozzilli et al. (48) found IDDM adults kept more complete glucose testing records than NIDDM patients. IDDM and NIDDM samples differ in several important respects: regimen prescriptions, family environment, life stage, time since diagnosis, age of diagnosis, and onset of complications. One or more of these factors could readily influence adherence. Consequently, one might expect greater differences in adherence behaviors between IDDM and NIDDM populations than actually has been documented in the literature.
DEFINING ADHERENCE: CONCEPTUAL ISSUES— Haynes (49), probably the most cited author on the subject, defines compliance as "the extent to which a person's behavior (in terms of taking medications, following diets, or executing lifestyle changes) coincides with medical or health advice." The definition requires that the patient's behavior be compared with a standard determined by the medical community. Some have taken issue with the term compliance, suggesting it places too much emphasis on the physician's role as an authority, determining what the patient should or should not do. Adherence is the preferred term since it connotes a willingness on the patient's part to follow the physician's recommendations. Either term, however, requires a comparison of the patient's behavior with medical or health advice.
Medical/health advice: the illusive adherence standard One of the first problems encountered in defining adherence is the illusiveness of medical or health advice. In diabetes, prescriptions for oral agents or insulin may be documented in the patient's chart. However, written recommendations concerning timing of medications, diet, exercise, and glucose testing rarely are found. The physician may discuss these components of everyday diabetes care with the patient, often using vague and nonspecific language. Patients may be told to "get some exercise" or "avoid high-fat foods." Such suggestions are too general to serve as a standard against which the patient's behavior can be compared. Sometimes standards can be ascertained in other ways—from medical textbooks, statements from relevant medical organizations, or agreed upon practice standards within a medical community. For example, the ADA uses panels of experts to develop consensus standards with regard to patient care. Standards are published relevant to insulin and oral medication, diet, glucose testing, exercise, and so on (50). As an alternative approach, Glasgow et al. (51) suggested that levels of diabetes self-care behaviors be measured without regard to standards when no clear provider prescription is available. Researchers also may elect to intervene in the patient care process and insure that specific, daily care recommendations are provided. Inadvertent noncompliance Although we often think of noncompliance as willful disregard of provider prescriptions, patients often are noncompliant inadvertently. Through patientprovider miscommunication, failure to recall information accurately, or knowledge and skill deficits, a patient may believe he or she is compliant while behaving in ways that are contrary to provider recommendations. Researchers need to be aware that patient and provider reports of provider prescriptions are often inconsistent. Similarly, patients may ap-
pear to be compliant, taking their insulin the correct time each day and testing their glucose at specified intervals. However, closer inspection of their actual behavior may reveal numerous technical errors, each potentially leading to incorrect insulin administration and incorrect glucose test results. Page et al. (52), for example, compared recommendations given by health-care providers in a childhood diabetes clinic with patients' and parents' recall of those recommendations. On the average, providers gave seven recommendations per patient. Patients (and parents of younger children) recalled an average of only two recommendations. Furthermore, 40% of the patient- and parent-recalled recommendations were not made by the provider. In such cases, the patient may be inadvertently noncompliant, because the patient fails to understand or recall the regimen prescriptions. As children grow and develop, they exhibit an increased ability to abstract information, and more sophisticated vocabulary and disease knowledge. Patient- provider miscommunications may be particularly common among children whenever the provider fails to modify vocabulary and communication style depending on the cognitive capacity of the child. Research by Perrin and Perrin (53) suggests that providers often are unable to accurately discriminate between the conceptual abilities of children of different ages. Whenever this occurs, miscommunication between provider and patient, leading to inadvertent patient noncompliance, is a likely result. Provider-patient miscommunications with regard to diabetes regimen prescriptions are not limited to children. In a NIDDM adult sample, Hulka et al. (54) reported that only 66% of physician instructions actually were understood by the patients. In a follow-up study, adult patients were found to be frequently misinformed as to the function of their medication and the appropriate timing of its administration (55). Patients are unlikely
DIABETES CARE, VOLUME 15,
NUMBER 11, NOVEMBER
to be adherent to physician recommendations that they do not understand or recall. We need to clarify how adherence standards can be more effectively communicated to patients, and we need to educate providers to behave accordingly. Inadvertent noncompliance may occur as the result of knowledge or technical skill deficits as well. When Watkins et al. (1) actually observed adult patients self-injecting insulin, they found that 80% used unacceptable techniques, and 52% made errors in dosage. Similar problems have been reported in juveniles with this disease; as many as 40% of youngsters make insulin injection errors (43,44). Three studies reported that >66% of adult patients test glucose incorrectly (1,39,56). Error rates in children are as high as 80% (4,43,44). Misconceptions about dosage and timing of medication are relatively common (1,54, 55). Similar misconceptions exist with regard to dietary composition of foods (30). To assure adequate adherence, the provider must assess whether the patient knows what to do and how to do it.
Behavioral complexity Patients often are described as compliant or noncompliant, as if compliance were a traitlike characteristic of the patient. Traitlike conceptualizations of adherence presume that if the patient is adherent (or nonadherent) with one component of the diabetes regimen, he or she will be adherent (or nonadherent) with all aspects of the regimen. The available literature flatly rejects this assumption. Studies of adult, childhood, IDDM, and NIDDM populations all report that the various components of the diabetes regimen are not related strongly to one another (17,19,23,24,26,27,31,57). A patient's behavior with regard to one aspect of the diabetes regimen is not predictive of the same patient's behavior relevant to other regimen components. Even within a regimen component, such as diet, patient behavior often is inconsistent from one dietary measure to another. For example, among children and adolescents,
DIABETES CARE, VOLUME 15,
total calories consumed, the frequency of eating, the percentage of calories consisting of fats, and concentrated sweets consumption all are unrelated to one another (26,57). Diabetes regimen adherence is clearly a multivariate construct requiring a multicomponent measurement strategy.
Adherence and metabolic control Within the diabetes literature, the tendency has been to treat adherence and metabolic control as interchangeable constructs. It is common, for example, for diabetologists to assess patient adherence by using measures of metabolic control (58). Unfortunately, the literature has failed to document a clear 1:1 relationship between patient adherence and diabetes control (19,23,27,28,59). Perhaps this is not so surprising, because metabolic control is a function of various factors other than adherence: appropriateness of the prescribed regimen, duration of disease, presence of other illness conditions, hormonal changes associated with growth and development, and heredity, to name a few. Adherence is one factor, but not the only factor, that may influence a patient's metabolic status. In fact, adherence should impact the patient's metabolic status only when an effective treatment regimen has been prescribed by the physician. Even perfect adherence will not make an ineffective treatment prescription effective. At the same time, the behavioral and education intervention literature might be criticized for conducting adherence assessments in the absence of biological assessments relevant to the patient's health status. If the purpose of the intervention is to improve patient adherence to positively impact the patient's medical condition, then changes in patient behavior should be examined in the context of changes in biological outcomes. Patient adherence and metabolic control need to be assessed both independently and concomitantly.
NUMBER 11, NOVEMBER
MEASURING ADHERENCE: METHODOLOGICAL ISSUES— Numerous adherence measurement strategies are available from which to choose. Measurement selection should be based on the conceptual issues discussed previously and on the demonstrated reliability and validity of the method. Available reliability data should demonstrate that the method measures the adherence behavior of interest in a consistent fashion. Relevant validity studies should suggest that the method is sensitive to differences between patients in levels of adherence behaviors and any behavior change within a patient. The method selected should be nonreactive, i.e., the process of measuring adherence should not induce change in patient adherence behaviors. Once a measurement method is selected, issues relevant to the timing of measurements must be considered.
Methods of measurement Methods of measuring adherence include health-status indicators, health-provider ratings, behavioral observations, permanent products, and patient self-reports. As mentioned previously, indexes of metabolic control or other indicators of health status offer a poor substitute for a more direct assessment of adherence. The literature has failed to document a 1:1 correspondence between adherence and diabetes control (19,23, 27,28,59). Yet, providers continue to use GHb levels or other indexes of health status to assess patient adherence (58). Patients in good metabolic control are presumed to be adherent; those in poor diabetes control are assumed to be noncompliant. Unfortunately, such an approach may subtly encourage the provider to blame the patient whenever the patient is in less than satisfactory metabolic control. Furthermore, using healthstatus indicators to measure adherence provides no information about what the patient is or is not doing relevant to the multiple components of diabetes care. An elevated GHb level indicates that
something is wrong, but it does not tell us what specifically is wrong with either the provider's prescriptions or the patient's daily management of the disease. Assessment of adherence requires a more direct, multibehavior measurement strategyProvider ratings of patient adherence often are unreliable (60-62) and suffer from the same conceptual and methodological problems as healthstatus indicators. Because the provider usually is aware of the patient's current level of metabolic control, knowledge of health status is likely to influence provider ratings of patient adherence. Consequently, provider adherence ratings rarely are made independently of the patient's current metabolic control. Furthermore, provider ratings are typically global in nature; the patient is rated as adherent or nonadherent based on a traitlike conceptualization of compliance. Now, convincing evidence is available indicating that diabetes adherence is a multivariate construct; multiple behaviors are involved that are not highly related to one another (17,19,23,24,27,31, 57). Global ratings of patient adherence will never capture this complexity accurately. In contrast, behavioral observation offers an assessment strategy that is highly specific and potentially unconfounded by concurrent patient health status. Observational methods have been particularly useful for reliably detecting technical skill deficits in the administration of insulin or glucose testing that could lead to inadvertent noncompliance (4,43,44). Observational methods also have been used successfully in home (63,64) and summer camp settings (30,34,37) to assess a variety of diabetes adherence behaviors in IDDM childhood populations. Although clearly useful, observational methods are usually labor intensive, requiring solicitation and training of observers to reliably code the behaviors of interest. Issues of measurement reactivity also need to be considered. In some cases, if the patient knows
that highly personal, diabetes-related behaviors are being observed by members of the family or others, the patient may change the behaviors selected for observation. The patient may become more adherent than usual. Or, the patient may seek ways to deceive the observer by making observation exceedingly difficult or by appearing more adherent than is actually the case. Counting permanent products is an interesting, but rarely used, assessment strategy. If a behavior is associated consistently with a quantifiable product, the product can be counted as an indirect measure of the associated adherence behavior. Counting the number of pills left in a bottle of prescribed medication is probably the most well-known use of a permanent product as a measure of adherence. Other examples include weighing bottles of insulin (2) and counting glucose test tablets or test strips (65). The advent of meters for home blood glucose monitoring with large memory capacities offers numerous permanent products, including the frequency, date, time, and result of glucose tests. Counting permanent products cannot be used as a general adherence assessment strategy, because many diabetes-relevant behaviors are not associated with a permanent product. Furthermore, permanent products, when available, are not always valid indicators of patient behavior. For example, a patient could engage in a variety of behaviors that would invalidate a pill count as a measure of medication adherence. These include removing but not taking pills from the bottle, giving pills from the bottle to others for their use, taking prescribed pills from a different medication dispenser, or taking the correct number of pills but at the wrong time. Nevertheless, permanent products sometimes can be used creatively to corroborate other types of adherence data. Patient self-reports concerning regimen adherence have been considered suspect; what patients say they do may bear little resemblance to actual behavior, because patients may be influenced
markedly by what they believe the doctor wants to hear. However, patient reports of noncompliance appear to be more valid than reports of compliance (2). Furthermore, when patients are asked to report about specific behaviors, better quality adherence data may be obtained. One approach is to ask patients to rate their adherence for specific behaviors relevant to medication, diet, exercise, testing, and foot care; from these ratings, a total compliance score is obtained (13,14,16,66-71). Only a few studies have examined the reliability of such reports (66,70). The use of a total compliance score also may be problematic, because it is insensitive to the complexity of diabetes management behaviors. Certain compliance behaviors (e.g., medication taking) might be expected to be related to metabolic control, whereas others (e.g., foot care) may not. By constructing a global adherence score, researchers may fail to detect relationships between specific adherence behaviors and health-outcome variables. Other approaches, also based on patient self-reports, have been developed to assess the multiple behavioral components of daily diabetes management. Good quality data can be obtained when patients are asked to report about highly specific behaviors over a specific time interval. For example, Glasgow et al. (19) have successfully used written diaries to measure adherence. Johnson et al. (26) have adapted the 24-h recall interview, a standard dietary assessment method, for use as a general adherence assessment strategy with IDDM patients. In a series of studies, the authors have demonstrated both the reliability and validity of this technique (26-28,34,57,72). The method also appears to be nonreactive— that is, it does not induce changes in adherence behaviors as a consequence of the interview process (72). Multiple interviews are recommended to assure a representative sample of patient adherence behaviors. Because memory errors can and do occur with all recall strategies, where possible, both the patient
DIABETES CARE, VOLUME 15,
NUMBER 11, NOVEMBER
and a significant other should be interviewed about the patient's behavior. What the patient may forget, the significant other may remember. Special consideration should be given when interviewing young children about behaviors involving time (e.g., time of insulin administration and meals). However, it appears that with brief practice, children as young as 6 yr can provide reliable adherence data about all aspects of their daily care (72). The method has numerous advantages: reliable information can be obtained readily about the multiple components of diabetes care; interviews can be conducted by telephone; little is demanded from the patient other than the time (—20 min) required to conduct each interview. However, trained interviewers are required; and when multiple interviews with multiple informants are conducted, the process becomes labor intensive.
Timing of measurement When researchers think about timing of measurements they most often consider the time it takes to measure adherence. Although such practical concerns are warranted, they should not take precedence over issues of reliability and validity. If a new laboratory assay was developed to assess metabolic status in diabetic patients, acceptance would be determined by demonstrated reliability and validity. Issues of cost and time to conduct the assay might influence the frequency of assay use, but would never be the primary determinant in its initial acceptability. Yet, when it comes to measuring human behavior, we sometimes apply different standards. Issues of brevity and expense may become paramount, whereas issues of reliability and validity become secondary. Human behavior is not easier to measure than human biology. To measure it well often takes patient time, provider time, or both. Because our ultimate goal is to develop reliable and valid measures of patient adherence behavior, the importance of
DIABETES CARE, VOLUME 15,
brevity per se must be kept in perspective. Once an appropriate measurement strategy has been selected, issues relevant to how often adherence behaviors should be measured need to be addressed. Because diabetes is a life-long disease, adherence behaviors, as well as metabolic status, will need to be assessed repeatedly over time. How often adherence behaviors need to be assessed is a function of the stability of those behaviors. Only recently have researchers begun to study the consistency of diabetes regimen behaviors over rime. In children, we know that adherence deteriorates during the first year after diagnosis (18,46,47), suggesting that repeated adherence assessments should be conducted during this time. In patients with established diabetes, diabetes regimen behaviors have been monitored over intervals as long as 6 mo. Adherence behaviors appear to be moderately stable over 2-3-mo intervals but deteriorate thereafter. Furthermore, certain behaviors (glucose testing) appear to be more stable than others (dietary intake, exercise, timing of insulin administration) (19,72). These data suggest that patients are not highly consistent in their diabetes care from month to month and from year to year. Just as patients' metabolic status can change, their behavioral management of the disease also may change. Often adherence behaviors are assessed because of their presumed relationship to patient health status. In such cases, it is important that adherence behaviors be monitored during a time interval that is temporally congruent with the health-status measure. For example, GHb is an index of blood glucose levels during the preceding 2 - 3 mo (73). If we are interested in the relationship between patient adherence behavior and glycemic control, we would need to assess patient adherence during the same time period indexed by the GHb measure, i.e., during the 2 - 3 mo before the blood draw for the GHb assay. Failure to assure temporal congruence between behavioral
NUMBER 11, NOVEMBER
and health status measures may be one reason behavior/health status linkages have been difficult to document in diabetes. MEASURING ADHERENCE: DIRECTIONS FOR FUTURE RESEARCH— We have begun to identify methods of reliably measuring adherence independent of the patient's health status. However, much more psychometric work needs to be done. Studies need to determine the reliability, validity, and usefulness of specific methods with different populations (e.g., 1DDM vs. NIDDM, children vs. adults, newly diagnosed vs. long-time patients). Longitudinal studies of the stability of adherence behaviors over time will help us understand the static versus dynamic nature of adherence as it relates to changes in disease course and human growth and development. Such studies will help us determine the optimal temporal frequency of adherence assessments. Application of research findings to the clinical situation is paramount. Once a psychometrically adequate method of adherence is developed, it must be tested within the clinical setting. Issues of staff training, efficiency, usefulness of data collection and analysis, and patient acceptance must be addressed. Our conceptualization of adherence may need to be reconsidered. In the past, patient adherence has been conceptualized as the relationship between patient behavior and some standard determined by the health provider. In this model, the standard is static; we examine how consistent the patient's behavior is from day to day in relationship to this standard. Recently, interest has increased in viewing adherence as a dynamic event. Instead of engaging in the same behaviors from day to day, the adherent patient would correct discrepancies between actual and normal blood glucose by appropriately changing behavior in response to blood glucose test results (74). Developing measures of adherence based on this dynamic model, in which
the appropriateness of behavioral change determines adherence, represents an interesting methodological challenge. Developing adequate measures of adherence is critical to our understanding of the link between adherence and health status in this population. Unless we can establish that adherence really makes a difference to patients' longevity and/or health-related quality of life, studies that examine predictors of adherence or methods of changing adherence may seem premature. The available literature's failure to document a clear and significant relationship between patient adherence and glycemic control is disturbing. However, it would be inappropriate to conclude that adherence has no effect on patient health; rather, the relationship is far more complex than initially surmised. Our task is to successfully address this complexity. Adequate measures of adherence are critical to this endeavor because measurement error will only reduce our ability to detect behavioral and health-status relationships. We also must reconsider the measures we select as health outcome variables. GHb is currently the most widely accepted measure of glycemic control. It may be an adequate measure of mean blood glucose but a poor measure of blood glucose variability. If adherence affects blood glucose variability more than mean blood glucose, the sole use of GHb as an index of health status would lead to a failure to detect this relationship. Lipid metabolism also is disrupted in diabetes. Yet, measures of lipid metabolism rarely have been used in studies of adherence and health-status relationships. Results from one study suggest that lipid metabolism may be associated more strongly with adherence behaviors than glucose metabolism (27). Finally, adherence behaviors have not been examined in relationship to healthrelated daily functioning. More adherent patients may be hospitalized less, miss school or work less, and experience fewer debilitating or distressing symptoms. Certainly, the quality of a patient's
life is as important, if not more important, than the results of a particular laboratory assay. Kaplan (75) has argued eloquently that behavior (longevity, disability, discomfort, health-related quality of life) is the central outcome in health care; an exclusive focus on laboratory assay results may prove too myopic an approach. Because the relationship between adherence and health outcomes is presumably complex, studies need to identify and include other variables that may influence metabolic control in addition to adherence. Such variables might include biological predictors: available insulin production in newly diagnosed IDDM patients, insulin resistance in IDDM and NIDDM patients, hormonal changes associated with puberty, illness, or psychological stress, to name a few. Environmental predictors might include physician and patient decision making related to medication prescriptions and the appropriateness of those prescriptions for good metabolic control. Finally, the importance of individual differences to the study of behavior and health status linkages needs to be given further consideration. Previous studies have been conducted on the premise that behavior influences health status in similar ways for all patients. If different behaviors are more or less critical to metabolic control in different patients, group studies will fail to detect these relationships. Only within-subject data analytical strategies would detect such an association. Some precedent has been set for this approach within the IDDM literature (76-78). Studies that have examined symptom-blood glucose relationships have failed to find any significant association between patient symptomotology and episodes of hypoglycemia or hyperglycemia when a group design was used. When the same data were reanalyzed using a within-subjects approach, multiple significant associations emerged. In other words, most patients exhibited one or more symptoms that were associated consistently with
hypoglycemia or hyperglycemia. But the pattern of these symptom-glycemic relationships differed across patients. A similar within-subjects approach to the study of adherence and health-status relationships might prove to be particularly enlightening. MEASURING ADHERENCE: CLINICAL APPLICATIONS — The conceptual and methodological issues relevant to the measurement of adherence are important for research but also have widespread clinical application. How the health provider conceptualizes adherence will impact both diabetes management prescriptions and providerpatient relationships. The provider who understands the problems associated with effectively communicating adherence standards will provide more specific and detailed management prescriptions. Attention will be given to how the information is communicated to the patient; language and terms will be selected carefully to assure patient understanding. Time will be taken during the clinical visit to ensure that the patient accurately recalls the provider's recommendations and has the prerequisite knowledge and skills to follow those recommendations. The provider who appreciates the complexity of diabetes regimen behaviors will avoid global impressions of patient adherence and attempt to gather information about the specific behaviors involved in diabetes care. Metabolic status will be monitored but never used to assess patient adherence. Instead of automatically blaming the patient whenever GHb levels are high, the provider will examine the adequacy of the management prescriptions provided the patient, and will assess what the patient is or is not doing with regard to diabetes care. Adherence assessment strategies will be selected based on the reliability and validity of the information provided rather than brevity of the instrument. Aware that human behavior may change, the provider will reassess patient adherence behaviors on a regular basis. To
DIABETES CARE, VOLUME 15,
NUMBER 11, NOVEMBER
help clarify the relationship between patient adherence behaviors and metabolic control, the provider will assess adherence behaviors during the same temporal interval indexed by laboratory assays. Such an approach requires monitoring of patient adherence behaviors to become as much a part of standard care as an assay of glycemic control. In the absence of accurate information about what the patient is or is not doing, the provider can never be sure the management prescriptions offered are understood by the patient or are the most appropriate recommendations for the patient's care. Measures of diabetes knowledge, technical skill at insulin administration and glucose testing, and methods of assessing the patient's daily diabetes management are currently available. Developed by behavioral researchers, these methods can be adapted readily for use by clinic staff. Although some patient and staff time is required, the information obtained is well worth the time investment. The procedures are not associated with pain or discomfort (as are the blood draws needed for laboratory assays), thereby increasing the likelihood of patient acceptance. Furthermore, it is possible that increased patient-provider interaction may improve communication and increase patient satisfaction which, in turn, may increase patient compliance (70,79,80).
Acknowledgments—This work was supported by National Institutes of Health Grants RO1-HD-13820 and PO1-DK-39079.
References 1. Watkins JD, Williams R, Martin DA, Hogan MD, Anderson E: A study of diabetic patients at home. Am J Pub Health 57:452-59, 1967 2. Diehl AK, Bauer RL, Sugarek N: Correlates of medication compliance in noninsulin-dependent diabetes mellitus. South Med] 80:332-35, 1987 3. Malone JI, Hellrung JM, Malphus EW, Rosenbloom AL, Grgic A, Weber FT:
Good diabetic control: a study in mass delusion. J Ped 88:943-47, 1976 Epstein LH, Cobum PC, Becker D, Drash A, Siminero L: Measurement and modification of the accuracy of determinations of urine glucose concentrations. Diabetes Care 3:535-36, 1980 Mazze RS, Shamoon H, Pasmanner R, Lucindo D, Murphy J, Hartmann K, Kuykendall V, Lopatin W: Reliability of blood glucose monitoring by patients with diabetes mellitus. Am J Med 77: 211-17, 1984 WingRR, Lamparski DM, ZaslowS, Betschart J, Siminerio L, Becker D: Frequency and accuracy of self-monitoring of blood glucose in children: relationship to glycemic control. Diabetes Care 8:214-18, 1985 Pohl SL, Gonder-Frederick L, Cox DJ, Evans WS: Self-measurement of blood glucose concentration: clinical significance of patient-generated measurements (Letter). Diabetes Care 8:617-19, 1985 Most RS, Gross AM, Davidson PC, Richardson R: The accuracy of glucose monitoring by diabetic individuals in their home setting. Diabetes Ed 12:24-27, 1986 Wilson DP, Endres RK: Compliance with blood glucose monitoring in children with type 1 diabetes mellitus. J Ped 108: 1022-24, 1986 Gonder-Frederick LA, Julian DM, Cox DJ, Clarke WL, Carter WR: Self-measurement of blood glucose: accuracy of self-reported data and adherence to recommended regimen. Diabetes Care 11: 579-85, 1988 Hoskins PL, Alford JB, Handelsman DJ, Yue DK, Turtle JR: Comparison of different models of diabetes care on compliance with self-monitoring of blood glucose by memory glucometer. Diabetes Care 11:719-24, 1988 Wysocki M, Czyzk A, Slonska Z, Krolewski A, Janeczko D: Health behaviour and its determinants among insulindependent diabetics: results of the diabetes Warsaw study. Diab Metab 4:117-22, 1978 Cerkoney KA, Hart LK: The relationship between the health belief model and
DIABETES CARE, VOLUME 15, NUMBER 11, NOVEMBER 1992
compliance of persons with diabetes mellitus. Diabetes Care 3:594-98, 1980 14. Cox DJ, Taylor AG, Nowacek G, HolleyWilcox P, Pohl SL: The relationship between psychological stress and insulindependent diabetic blood glucose control: Preliminary investigations. Health Psychol 3:63-75, 1984 15. Jenny J: A comparison of four age groups' adaptation to diabetes. Can ] Pub Health 75:237-44, 1984 16. Shlenk EA, Hart LK: Relationship between health locus of control, health value, and social support and compliance of persons with diabetes mellitus. Diabetes Care 7:566-74, 1984 17. Ary DV, Toobert D, Wilson W, Glasgow RE: Patient perspective on factors contributing to nonadherence to diabetes regimen. Diabetes Care 9:168-72, 1986 18. Kovacs M, Brent D, Steinberg TF, Paulauskas S, Reid J: Children's self-reports of psychologic adjustment and coping strategies during the first year of insulindependent diabetes mellitus. Diabetes Care 9:472-79, 1986 19. Glasgow RE, McCaul KD, Schafer LC: Self-care behaviors and glycemic control in type 1 diabetes. J Chron Dis 40:399412, 1987 20. Pendleton L, House WC, Parker LE: Physicians' and patients' views of problems of compliance with diabetes regimens. Pub Health Rep 102:21-26, 1987 21. Glasgow RE, McCaul KD, Shafer LC: Barriers to regimen adherence among persons with insulin-dependent diabetes. J Behav Med 9:65-77, 1984 22. Diel AK, Sugarek NJ, Baurer RL: Medication compliance in non-insulin dependent diabetes: A randomized comparison of chlorpropamide and insulin. Diabetes Care 8:219-23, 1985 23. Glasgow RE, Toobert DJ, Riddle M, Donnelly J, Mitchell DJ, Calder D: Diabetesspecific social learning variables and selfcare behaviors among persons with type II diabetes. Health Psychol 8:285-303, 1989 24. Orme CM, Binik YM: Consistency of adherence across regimen demands. Health Psychol 8:27-43, 1989 25. Bernal H: Self-management of diabetes in a Puerto Rican population. Pub Health
Nursing 3:38-47, 1986 26. Johnson SB, Silverstein J, Rosenbloom A, Carter R, Cunningham W: Assessing daily management of childhood diabetes. Health Psychol 5:545-64, 1986 27. Johnson SB, Freund A, Silverstein J, Hansen CA, Malone J: Adherence-health status relationships in childhood diabetes. Health Psychol 9:606-31, 1990 28. Spevack M, Johnson SB, Riley R, Silverstein J: The effect of diabetes summer camp on adherence behaviors and glyce-
tus. Diabetes Care 6:245-50, 1983 37. Hanson SL, Pichert JW: Perceived stress and diabetes control in adolescents. Health Psychol 5:439-52, 1986 38. Marcus MD, Wing RR: Eating disorders
cents with insulin-dependent diabetes mellitus over a four-year longitudinal follow-up: I. The influence of patient coping and adjustment. J Ped Psychol 15: 511-26
48. Pozzilli P, Occhigrossi F, Buzzetti R, Fiore V: Record books and patients with CS, Ed. New York, Springer-Verlag, NIDDM: waste of time (Letter)? Diabetes 1990, p. 102-21 Care 11:299-300, 1988 39. Mazzuca SA, Moorman NH, Wheeler 49. Haynes RB: Introduction. In Compliance in Health Care. Haynes RB, Taylor DW, ML, Norton JA, Fineberg NS, Vinicor F, Sackett DL, Eds. Baltimore, Maryland, Cohen SJ, Clark CM: The diabetes edumic control. In Advances in Child Health Johns Hopkins Press, 1979, p. 2 - 3 cation study: a controlled trial of the efPsychology. Johnson JH, Johnson SB, Eds. fects of diabetes patient education. Dia- 50. American Diabetes Association: Clinical Gainesville, Florida, University of Florida betes Care 9:1-11, 1986 practice recommendations, 1990-1991. Press, 1991, p. 285-92 Diabetes Care 14 (Suppl. 2), 1991 40. Wing RR, Epstein LH, Marcus MD: Behavioral strategies for improving weight 51. Glasgow R, Wilson W, McCaul K: RegiDelamater AM, Bubb J, Davis SG, Smith men adherence: a problematic concept in loss in obese type II diabetic patients. In JA, Schmidt L, White NH, Santiago JV: diabetes research. Diabetes Care 8:300Neuropsychological and Behavioral Aspects Randomized prospective study of self301, 1985 of Diabetes. Holmes CS, Ed. New York, management training with newly diagSpringer-Verlag, 1990, p. 198-221 nosed diabetic children. Diabetes Care 52. Page P, Verstraete DG, Robb JR, Etzwiler 13:492-98, 1990 DD: Patient recall of self-care recommen41. Rubin RR, Peyrot M, Saudek CD: Effect dations in diabetes. Diabetes Care 4:96of diabetes education on self-care, metaLorenz RA, Christensen NK, Pichert JW: 98, 1981 bolic control, and emotional well-being. Diet-related knowledge, skill, and adherDiabetes Care 12:673-79, 1989 ence among children with insulin-de53. Pen-in EC, PerrinJM: Clinician's assesspendent diabetes mellitus. Pediatrics 75: 42. Wysocki T, Green L, Huxtable K. Blood ments of children's understanding of ill872-76, 1985 ness. Am] Dis Child 137:874-78, 1983 glucose monitoring in diabetic adolesRost RM, Flavin KS, Schmidt LE, McGill cents: compliance and metabolic control. 54. Hulka BS, Kupper LL, Cassel JC, Mayo F: JB: Self-care predictors of metabolic conDoctor-patient communication and outHealth Psychol 8:267-84, 1989 trol in NIDDM patients. Diabetes Care 43. Johnson SB, Pollak T, Silverstein JH, comes among diabetic patients. J Com13:1111-13, 1990 munity Health 1:15-27, 1975 Rosenbloom AL, Spillar R, McCallum M, American Diabetes Association: Position Harkavy J: Cognitive and behavioral 55. Hulka BS, Cassel JC, Kupper LL, Burstatement: Nutritional recommendations dette JA: Communication, compliance knowledge about insulin dependent diaand principles for individuals with diaand concordance between physicians betes among children and parents. Pedibetes mellitus. Diabetes Care 14 (Suppl. and patients with prescribed medicaatrics 69:708-13, 1982 2):20-27, 1991 tions. Am] Pub Health 66:847-53, 1976 44. Harkavy J, Johnson SB, Silverstein J, Wharton CL, Wickinjg J, Wills HR, MarSpillar R, McCallum M, Rosenbloom: 56. Shenfield GM, Steel JM. Urinary glucose tin FIR: The diet of insulin-dependent testing by diabetic patients. The PractiWho leams what at diabetes summer diabetics. Med J Australia 2:707-13, tioner 218:147-50, 1977 camp. J Ped Psychol 8:143-53, 1983 1982 45. Ingersoll GM, Orr DP, Herrold AJ, 57. Johnson SB, Tomer A, Cunningham WR, Reynolds LA, Johnson SB, Silverstein J: HenrettaJC: Adherence in childhood diGolden MP: Cognitive maturity and selfAssessing daily diabetes management by abetes: results of a confirmatory factor management among adolescents with in24-hour recall interview: the validity of analysis. Health Psychol 9:493-501, sulin-requiring diabetes mellitus. J Ped children's reports. J Ped Psychology 15: 1990 108:620-23, 1986 493-509, 1990 46. Jacobson AM, Hauser ST, Wolfsdorf JI, 58. Clarke WL, Snyder AL, Nowacek G: Webb KL, Dobson AJ, Tuplin HE, Harris Outpatient pediatric diabetes: I. Current Houlihan J, Milley JE, Herskowitz RD, GW, O'Connell DL, Atkinson J, Sulway practices. J Chronic Dis 38:85-90, 1985 Wertlieb D, Watt E: Psychological preMJ, Leeder SR: Evaluation of a diabetes dictors of compliance in children with 59. Johnson SB: Adherence behaviors and education programme. Australian New health status in childhood diabetes. In recent onset of diabetes mellitus. J Ped Zealand] Med 12:153-60, 1982 Neuropsychological and Behavioral Aspects 110:805-11, 1987 Christensen NK, Terry RD, Wyatt S, 47. Jacobson AM, Hauser ST, Lavori P, of Diabetes. Holmes CS, Ed. New York, Pichert JW, Lorenz RA: Quantitative asSpringer-Verlag, 1990, 30-57 Wolfsdorf JI, Herskowitz RD, Milley JE, sessment of dietary adherence in patients Bliss R, Gelfand E, Wertlieb D, Stein J: 60. Chamey E, Bynum R, Eldredge D, Frank with insulin-dependent diabetes melliD, MacWhinneyJB, McNabb N, Scheiner Adherence among children and adoles-
and diabetes. In Neuropsychological and Behavioral Aspects of Diabetes. Holmes
DIABETES CARE, VOLUME
NUMBER 1 1 ,
A, Sumpter EA, Iker H: How well do patients take oral penicillin? A collaborative study in private practice. Pediatrics 40:188-95, 1967 Caron HS, Roth HP. Patient's cooperation with a medical regimen: difficulties in identifying the noncooperator. J Am MedAssoc 203:120-24, 1968 Davis MS: Physiologic, psychological, and demographic factors in patient compliance with doctors' orders. Med Care 6:115-22, 1968 Epstein LH, Beck S, FigueroaJ, Farkas G, Kazdin AD, Daneman D, Becker D: The effects of targeting improvements in urine glucose on metabolic control in children with insulin dependent diabetes. JApp BehavAnal 14:365-75, 1981 Lowe K, Lutzker JR: Increasing compliance to a medical regimen with a juvenile diabetic. Behav Ther 10:57-64, 1979 Gross AM: Self-management training and medication compliance in children with diabetes. Child Family Behav Ther 4:47-55, 1983 Brownlee-Duffeck M, Peterson L, Simonds J, Goldstein D, Kilo C, Hoette S: The role of health beliefs in the regimen adherence and metabolic control of adolescents and adults with diabetes mellitus. J Clin Consult Psychol 55:139-44, 1987 Hanson C, Henggler S, Burghen G: Race and sex differences in metabolic control
DIABETES CARE, VOLUME 15,
of adolescents with IDDM: a function of ing diabetic outpatients. J Clin Endocrinol psychosocial variables? Diabetes Care 10: Metab 64:269-73, 1987 313-18, 1987 74. Wing R, Epstein L, Nowalk M, Lamparski: Behavioral self-regulation in the 68. Hanson C, Henggler S, Burghen G: treatment of patients with diabetes melModel of associations between psychosolitus. Psych Bui! 99:78-89, 1986 cial variables and health-outcome measures in adolescents with IDDM. Diabetes 75. Kaplan R: Behavior as the central outCare 10:752-58, 1987 come in health care. Am Psychol 45: 1211-20, 1990 69. Hanson C, Henggler S, Burghen G: Social competence and parental support as me- 76. PennebakerJ, Cox D, Gonder-Frederick diators of the link between stress and L, Wunsch M, Evans W, Pohl S: Physical metabolic control in adolescents with insymptoms related to blood glucose in sulin-dependent diabetes mellitus. J Clin insulin-dependent diabetics. Psychosom Consult Psychol 55:529-33, 1987 Med 43:489-500, 1981 70. Hanson C, Henggeler S, Harris M, Mitch- 77. Cox D, Gonder-Frederick L, Pohl S, Penell K, Carle D, Burghen G: Associations nebaker J: Reliability of symptom-blood between family members' perceptions of glucose relationships among insulinthe health care system and the health of dependent adult diabetics. Psychosom youths with insulin-dependent diabetes Med 45:357-60, 1983 mellitus. J Ped Psychol 13:543-54, 1988 78. Freund A, Johnson S, Rosenbloom A, Alexander B, Hansen C: Subjective 71. Hanson C, Rodrigue J, Burghen G, symptoms, blood glucose estimation, Henggeler S, Onikul-Ross S: Peptic ulcer and blood glucose concentrations in addisease in youths with insulin-depenolescents with diabetes. Diabetes Care dent diabetes mellitus: a prospective 9:236-43, 1986 study. J Ped Psychol 15:595-604, 1990 72. Freund A, Johnson SB, Silverstein J, Th- 79. Litt 1, Cusky W: Satisfaction with health omas J: Assessing daily management of care: a predictor of adolescents' appointchildhood diabetes using 24-h recall inment keeping. J Adolesc Health Care terviews: reliability and stability. Health 5:196-200, 1984 Psychol 10:200-208, 1991 80. Hazzard A, Hutchinson S, Krawiecki N: 73. Ziel RH, Davidson MB: The role of glyFactors related to adherence to medicacosylated serum albumin in monitoring tion regimens in pediatric seizure paglycemic control in stable insulin-requirtients. J Ped Psychol 15:543-55, 1990
NUMBER 11, NOVEMBER