The Network of Psychological Variables  in Patients With Diabetes and Their Importance for Quality of Life and Metabolic Control

It has repeatedly been proven that quality of life is influenced by the type and number of secondary illnesses (1,2,3) and the frequency of serious metabolic complications that arise (4,5). However, particular personality characteristics play a decisive, mediating role in the quality of life experienced (6,7,8). We find that depressed subjects generally have a poorer quality of life, independent of the physical illnesses from which they might suffer (9,10,11,12). This also applies to patients with diabetes whose personality dispositions appear to be more significant for the quality of life than, for instance, the presence of secondary illnesses (8,13,14). The patient’s psychological disposition is, in turn, closely connected to cognitive and emotional aspects of his/her illness-coping strategies, which indirectly affect their health-seeking behavior (15). The patient’s subjective assessment that a health-relevant behavior exists (“outcome expectancy”) and the belief that he or she is able to carry out this behavior (“self-efficacy expectancy”) (15) have been described as predictors for adherence to therapy (16,17) as well as for the present quality of life of patients with diabetes (18). Most studies found a positive connection between active, problem-oriented coping behavior and an improved quality of life (19,20,21,22). In the case of adolescents, however, it appears that it can be intermittently useful as well to adopt an attitude of emotion-oriented coping, even denial (23). Because the treatment of patients with diabetes primarily involves instruction in self-therapy, the central belief is that the attending physician can influence the way the patients deal with their illness. Indeed, it can be shown that the effective communication of treatment-relevant information (24,25) or the effective change of emotional illness-coping strategies (20,21) can be influenced by particular educational programs, with the change in illness-coping strategies likely exerting the greater influence on relevant changes in behavior (20,21). Relatively little research has investigated the significance of the type and quality of the doctor-patient relationship, the personality of the attending physician, or the type and/or particular specialization of the clinical institution for the successful treatment of patients with diabetes. The few available studies (26,27,28) and the clinical experience itself suggest that these components play essential roles in achieving the previously mentioned goals of therapy.

It is undisputed that the patient’s motivation and willingness to take part in therapy are crucial foundations for good self-therapy and the resulting optimal blood glucose regulation (29). It has been shown that a successful communication of illness-relevant knowledge in the context of specific, educational programs has a positive effect (30). An additional (24,25) benefit seems to be achievable if the education program also focuses on the development of adequate emotional coping skills (20,21,31)

Some authors assume that the present quality of life influences metabolic regulation and vice versa (32,33,34). Although a positive effect has appeared to be plausible in various, detailed studies (35), there are, in our opinion, still too many contradictory results to allow us to assume that this relationship is proven at this time (36,37,38,39,40).

The preceding overview demonstrates that many somatic, psychological, treatment-relevant factors are already known that either directly or indirectly influence the two central treatment goals, blood glucose regulation and quality of life. Glasgow (41) attempted to summarize some of these factors as “patient characteristics” and “clinic characteristics” that determine the “cycle of care” made up of “self-management behavior” and “doctor-patient relationship.” We would like to make use of his model and assume that the patients with diabetes are embedded in complex network of interacting agents and forces that can be illustrated in a simplified manner, as shown in Fig. 1.

The relationships between these elements are usually examined individually. What is problematic in depicting individual factors from this conceptual network is that the relative significance of the individual factors can only be accurately represented within the context of all factors (7). Consequently, in the present study, we have attempted to simultaneously include as many of the above-mentioned factors as possible in the analysis in order to present as realistic a picture of the treatment conditions of patients with diabetes. To keep the analysis within manageable boundaries, we decided to focus solely on the patient factors and to leave out those influencing factors coming from the physician or the social background.

A structural equation model was used for the analysis of the data. Compared with a regression analysis, a structural equation model has the advantage of being able to examine the significance of psychological factors for both target criteria at the same time, as well as being able to examine the interrelationship among the observed factors, according to the previously mentioned theoretical assumptions, without having to allow for covariation among all variables.

The present study examines a sample of 625 diabetic patients (224 type 1 and 401 type 2 diabetic patients). The patients come from three sources: the diabetes outpatient center of the Charité (n = 174), 3 specialist practices (n = 264), and 28 general practitioner practices (n = 187). The aim of this investigation was to acquire a relatively realistic picture of patients with diabetes who are undergoing routine treatment in various outpatient facilities.

A total of 194 general practitioners were contacted randomly (Medical Handbook in Berlin [42]) until 30 general practitioners were recruited. Two general practitioners later refused to take part in the study. All general practitioners were expected to include eight patients consecutively within a time period of 3 months. All of them made the patients’ files freely accessible, enabling us to secure data quality and control for the fulfillment of the recruitment conditions. All three randomly chosen specialist practices were willing to take part in the study and to include 80 of their patients. The random selection of patients was assured by the participation of one of our researchers on particular days on location, where all patients present on this day were recruited. All patients being treated at the university policlinic within a 9-month period were included in the study, resulting in a total of 174 patients. Eight patients refused to participate.

The sociodemographic characteristics and characteristics with respect to illness and therapy status are listed in Table 1.

Health-related quality of life (HRQL) was measured using the multidimensional World Health Organization Quality of Life questionnaire (WHOQOL-Bref) (43) in its short version (26 items). It was developed in 17 countries on >17,000 patients worldwide. The German representative sample consisted of 2,055 subjects. The scales showed a good internal consistency for the sample under investigation (Table 2).

The Freiburger Illness-Coping Strategies questionnaire (FKV-lis) (44) covers a broad spectrum of illness-coping strategies on cognitive, emotional, and behavioral levels. The FKV-lis is made up of 35 items and five scales. The validation process was carried out on random samples of various acute and chronically ill patients (n = 350). In this study, we used the scale “active problem-oriented coping strategy” (with items such as “I am actively seeking information about the illness and its treatment”). It showed a good internal consistency for the sample under investigation (Table 2).

The Medical Interview Satisfaction Scale (MISS) (45) examines the patients’ assessment of their relationship with their physician, taking into account the level of trust in the relationship, the assessment of the physician’s reliability, and the time that the physician takes for the patient. Therefore, the questionnaire examines the patients’ subjective assessment of the quality of the treatment situation. The German version is made up of 16 items (46). The scales “empathy,” “competence,” and “information” were used. The scales showed a good internal consistency for the sample under investigation (Table 2).

The Giessen Test (GTS) (47) is a self-assessment test, based on theoretical assumptions from depth psychology, that offers subjects the opportunity to assess themselves according to their self-image. The test consists of 40 bipolar questions (e.g., “I blame myself very seldom… very often”). A representative German sample (n = 1,575) as well as various clinical samples served as a validation of the test. The test has been used and cited in >2,000 German studies. In this study, we used one scale that consisted of six items asking for different trait markers of depression (“depressive disposition”). It showed a sufficient internal consistency for the sample under investigation (Table 2).

The Assessment of Beliefs in Self-Efficacy and Optimism (SWOP) questionnaire (48) was created by restructuring two older questionnaires (49,50). The SWOP consists of a scale assessing self-efficacy (five items; for instance: “I face difficulties with relative ease because I can count on my abilities”) and four items assessing optimism/pessimism (for instance: “Things never go the way I want”). The validation of this instrument was carried out on five clinical samples with a total of 726 patients. The scales showed a good internal consistency for the sample under investigation (Table 2).

The German Test of Diabetes-Specific Knowledge (DWT-I-K) (51) is based on comparable Anglo-American questionnaires (52,53) and covers knowledge in the following areas: causes/pathophysiology, insulin/effects of insulin, insulin injection/storage, nutrition and physical activity, personal control over metabolic functions, hyperglycemia, hypoglycemia, illnesses, insulin adaptation, and consequential damages. The DWT-I-K consists of 30 questions with three possible dichotomized answers (yes/no), resulting in a total of 90 items. A question is considered to be answered correctly when all three possible answers have been answered correctly. The questionnaire has been validated on 912 persons in specialized treatment facilities and is widely used in the German-speaking world. It was given to all patients undergoing insulin therapy. Because there was no corresponding validated instrument in German for patients with type 2 diabetes who are not undergoing insulin treatment, these patients were only asked those DWT-I-K questions that did not concern insulin. Four ad hoc questions concerning diet were added. Therefore, this instrument consisted of 15 questions with the same multiple-answer options that were previously mentioned. The assessment of the results was carried out solely by means of sum scores (in percent) with respect to the maximum number of correct answers possible (54).

The attending physician was asked to fill out an ad hoc questionnaire (13 items), which yielded information concerning the patient’s present secondary illnesses.

The patients were asked to fill out an ad hoc questionnaire (30 items), which yielded information concerning their social status, present therapy, and the frequency of hypo- and hyperglycemic episodes that required professional treatment (metabolic complications).

The HbA_1c values were determined at the laboratories of the respective treatment facilities, all of which used the high-performance liquid chromatography method. Due to the slightly differing normal ranges, the HbA_1c values are presented as absolute deviations from the upper normal cutoff of the respective laboratory.

Based on the assumptions outlined in the introduction, a structural equations model was created to assess the factors as follows:

1) The construct “doctor-patient relationship” was understood as a subjective assessment of the doctor-patient relationship and was measured by means of the MISS. The scales measuring “empathy,” “information,” and subjectively experienced level of medical “competency” showed very high loadings (>0.90) on the one proposed factor, making it possible to operationalize this construct as a latent variable (Table 2).

2) “Patient characteristics” were assessed using the GTS and the SWOP. The scale “depressive disposition” showed a comparably low factor loading (−0.73) and communality (0.54), indicating that this factor is less homogeneous than the first one. Nevertheless, it seems to be justified to include this theoretically important construct in the latent variable, made up of the three scales “depressive disposition,” “self-efficacy,” and “optimism” (Table 2).

3) The construct “coping” was covered by two approaches: by a self-assessment asking the patients whether they use a problem-oriented active-coping behavior and by testing the present treatment-specific knowledge as an indicator for effective information-seeking behavior.

4) The number of secondary illnesses and the frequency of hypo- or hyperglycemias needing professional treatment were used to assess the “characteristics of the illness.” The type of secondary illness did not play a role in the statistical calculations.

5) The HbA_1c values were used to measure the targeted variable “metabolic control.”

6) Quality of life was measured using the WHOQOL-Bref. In accordance with the generally accepted understanding of HRQL as a multidimensional construct mirroring biopsychosocial aspects of illness, the three scales “physical domain,” “psychological domain,” and “social domain” were measured. All three scales showed loadings >0.70 in a one-factor solution, indicating that there is a fair amount of shared variance between them. We focused on this shared variance, which allowed us to form the latent variable, quality of life.

All observed variables were controlled for the age of the patient (Fig. 2). The statistical calculations were carried out with the AMOS 4.0 program (available online from http://www.smallwaters.com). The adequacy of fit of the model can be assumed if a P value >0.10 is reached. The standard error and the significance of the β-coefficients were calculated using a bootstrap procedure (maximum likelihood method, 95% CIs, 500 bootstrap samples).

If we construct the structural equation model as illustrated in Fig. 1, we come up with a valid model (χ² = 74.3, P = 0.16) and a good index of fit (adjusted goodness of fit [AGFI] = 0.94, root mean square error of approximation [RMSEA] = 0.03, C_min/df = 1.18). If we further restrict the model by using only the significant β-weights (P < 0.05) for recalculation, the new constructed model is also valid (χ² = 88.5, P = 0.16) (Fig. 2), showing very similar interrelationships and fit indexes (AGFI = 0.94, RMSEA = 0.02, C_min/df = 1.16). As there is a common interpretation of both models, the reported data in the following paragraphs will refer to the second model, which is more restricted.

In a multigroup approach, we checked whether significant differences occur between the regression patterns of type 1 and type 2 diabetic patients. If we force all regression coefficients of type 1 and type 2 diabetic patients to be the same and compare this model with one in which the regression coefficients are allowed to differ, the model with the forced restrictions is favored (df = 38, C_min = 37.62, P = 0.487). We can conclude that there are no significant differences between the types with regard to the models of the adult type 1 and type 2 diabetic patients.

In Fig. 2, we can see that 5% of the variance in the observed HbA_1c values can be explained by the factors included. Patients suffering from a severe somatic illness condition who report acute metabolic complications needing professional treatment more often had the poorest HbA_1c values (β = 0.15). In contrast, patients who reported an active coping style were found to have better HbA_1c values (β = 0.19). The diabetes-specific knowledge and the patient’s subjective assessment of the doctor-patient relationship played a subordinate role in connection with the HbA_1c values (P > 0.05).

Sixty-two percent of the variance of the HRQL can be explained by the model. The most significant determinants for the HRQL were the “patient characteristics” (β = 0.60). Patients with a more optimistic attitude, showing stronger beliefs in self-efficacy and a generally more positive disposition, had the higher HRQL values.

Patients with a more severe somatic condition had, as expected, the lower HRQL values (β= −0.39). Independent of this, with a standardized regression value of β = 0.20, the subjectively experienced doctor-patient relationship also played an important role for the HRQL values. Patients who were found to be better informed (“knowledge”) (Fig. 2) also reported a higher quality of life (β = 0.15).

The more optimistic patients, those with stronger beliefs in self-efficacy, and patients with better doctor-patient relationships reported a more active coping style (β = 0.21/0.15). The relevance of these values for the treatment-specific knowledge is minimal. Only the age of the patient played a particularly important role for this variable (β = −0.42). The younger patients were, on average, better informed than the older ones.

The goals in treating patients with diabetes are, most importantly, a sufficient blood glucose regulation and a quality of life with as few restrictions as possible. There are innumerable studies that can identify various factors related to these two treatment goals (1). In the daily treatment context of patients, however, these individual factors are part of a complex network of interacting forces. For these reasons, we wanted to construct and assess a model that can present the various individual factors simultaneously, making it possible to examine the relative significance of the single factors for realizing the treatment goals. It was indeed possible to develop a statistically valid structural equation model with a sound goodness of fit, allowing us to assume that most of the conceptual assumptions regarding the interactive character of the various influencing factors are in fact correct.

With a variance explanation of only 5%, all included factors have little explanatory value for the observed HbA_1c variance, and it is obvious that other factors that were not studied are more significant. Contrary to our assumptions but similar to results reported by Roth et al. (51), we did not find a connection between actual treatment-specific knowledge and HbA_1c values, at least when measured under routine treatment conditions. Obviously, good glycemic control can be achieved by means of routine or intuition, without having detailed knowledge about the disease, just as having such knowledge does not necessarily mean being able to implement it. The only significant psychological factor for the metabolic control was how active the patients describe their coping behavior. In accordance with other studies (22), we found that, in adults, a more active coping style covaries with better blood glucose regulation.

A large number of methodologically sophisticated studies have focused on the identification of individual factors that influence the self-reported quality of life of patients. Rubin and Peyrot (1) have recently summarized these studies very clearly.

When including these known factors in one structural equation model, a total of 62% of the variance in the reported quality of life could be correctly predicted. This is much more than the variance that could be explained for the HbA_1c values. When interpreting these results, however, we must take into account that predictors in the model include patient self-assessments. Because the present quality of life measure is also a self-assessment, the methodological overlap with this criterion is larger than with the chemical lab results of HbA_1c values.

Nevertheless, the relatively high β-weights for those variables that are not self-assessments confirm that we already know a great deal about the determinants of quality of life: solid diabetes-specific knowledge was found to be associated with a higher HRQL (β = 0.15), and of course, the well-being of the patients is determined to a large extent by the presence and number of secondary illnesses and acute metabolic complications (β = −0.39).

Much more significant, however, is the personal attitude of the patient (β = 0.60). As we have similarly reported earlier (14), we were now able to confirm that if a patient reports a good, optimistic outlook on life and strong beliefs in self-efficacy, he/she is more likely to report a higher quality of life, even in the presence of secondary illnesses.

Therefore, the personality characteristics “self-efficacy,” “optimism,” and “depressive disposition” may appear to be related to the quality of life construct, particularly in terms of its psychological dimensions, so that a high covariation could appear tautological. To address this potential problem, we made sure that the personality characteristics assessed and the questionnaires used measured stable, illness-independent traits of the individual, whereas the quality of life measurements only assessed the patient’s present situation. In fact, there are no overlapping questions in the questionnaires used. Of course, even when it is possible to establish a sufficient conceptual and methodological separation of the two, based on present research standards, the principally close relationship between these constructs naturally remains. These data thereby illustrate the peculiar situation to be found in the present understanding of the concept of HRQL. If we want to include and influence psychological dimensions of health or illness, we must admit that in many situations, a large proportion of HRQL is more fundamentally determined by factors that are to be found in the personality development of the patients than in the illness or its treatment.

The key variable among the characteristics of the illness, the patient, and the physician often appears to be the coping behavior of the patient. Here, we might be able to name a construct that is crucially significant for the self-treatment of chronic illnesses and one that also lies within the physician’s realm of influence. In fact, on one hand, the coping behavior covaried with the metabolic control and, on the other hand, with a positively attributed doctor-patient relationship and a personal attitude of self-efficacy. The cognitive “knowledge” was exclusively related to the age of the patient (54).

The key psychological variable for both therapy goals—namely, blood glucose regulation and HRQL—proved to be the patient’s belief in his/her self-efficacy and an optimistic outlook on life. These variables were directly connected with a more positively experienced doctor-patient relationship, more active coping behavior, and better HRQL. If the patient is able to maintain this attitude for as long as possible despite his/her somatic condition, then the context appears to be favorable for the realization of both therapeutic goals. It is therefore worthwhile to consider whether the patient training programs, for instance, should routinely include the goal of enhancing beliefs in self-efficacy in their interventions, particularly in light of the fact that positive experiences in this area have already been reported (21) and the fact that pure, cognitive knowledge of treatment appears to have limited significance under routine conditions.

This study was supported by Humboldt University Grant no. UFF 98-513.

We would like to thank Dr. Livia Cotta and Dr. Bärbel Seidlitz for their talent to organize and manage the data collection and Chie Hee Cho, Manuel Franco, Tobias Hofmann, and Sacha Wohlfahrt for their work collecting the data. For their kind cooperation, special thanks goes to the following colleagues: Dr. Volker Boest, Karin Dienz-Graening, Dr. Helga Flatow, Brigitte Galwitz, Sonja Gericke, Alexander Grafe, Dr. Jutta Haiduga, Dr. Jordan, Anita Kral, Dr. Ingo Kutscher, Dr. Rüdiger Langer, Gisela Löser Dr. Angelika Lutze, Dr. Evelyn Mahlke, Dr. Meier, Prof. Dr. Hans Peter Meiβner, Renate Mollenhauer-Prell, Dr. Larissa Niederstadt, Eduard Obermeier, Dr. Rita Patzwald, Dipl.-Med. Christiane Prescher, Dr. Ronald Raabe, Dr. Rabenow, PD Dr. Ruhnau, Christine Schley, Dr. Holgen Strohscheer, Dr. Manfred Stündel, Dr. Albrecht Sturma, Dr. Inge Thyen, Dr. Thomas Verstege, Dr. Stefan Weiβach, and Dr. Eberhard Zühlsdorff.