Interactive effects of social support and social conflict on medication adherence in multimorbid older adults

Social Science & Medicine 87 (2013) 23e30

Contents lists available at SciVerse ScienceDirect

Social Science & Medicine journal homepage: www.elsevier.com/locate/socscimed

Interactive effects of social support and social conflict on medication adherence in multimorbid older adults Lisa M. Warner a, b, *, Benjamin Schüz c, Leona Aiken d, Jochen P. Ziegelmann b, Susanne Wurm b, Clemens Tesch-Römer b, Ralf Schwarzer a, e a

Department of Psychology, Freie Universität Berlin, Germany German Centre of Gerontology, Berlin, Germany University of Tasmania, Australia d Department of Psychology, Arizona State University, USA e University of Social Sciences and Humanities, Wroclaw, Poland b c

a r t i c l e i n f o

a b s t r a c t

Article history: Available online 18 March 2013

With increasing age and multimorbidity, medication regimens become demanding, potentially resulting in suboptimal adherence. Social support has been discussed as a predictor of adherence, but previous findings are inconsistent. The study examines general social support, medication-specific social support, and social conflict as predictors of adherence at two points in time (6 months apart) to test the mobilization and social conflict hypotheses. A total of 309 community-dwelling multimorbid adults (65e85 years, mean age 73.27, 41.7% women; most frequent illnesses: hypertension, osteoarthritis and hyperlipidemia) were recruited from the population-representative German Ageing Survey. Only medicationspecific support correlated with adherence. Controlling for baseline adherence, demographics, physical fitness, medication regimen, and attitude, Time 1 medication-specific support negatively predicted Time 2 adherence, and vice versa. The negative relation between earlier medication-specific support and later adherence was not due to mobilization (low adherence mobilizing support from others, which over time would support adherence). Social conflict moderated the medication-specific support to adherence relationship: the relationship became more negative, the more social conflict participants reported. Presence of social conflict should be considered when received social support is studied, because wellintended help might have the opposite effect, when it coincides with social conflict. Ó 2013 Elsevier Ltd. All rights reserved.

Keywords: Germany Medication adherence Multimorbidity Medication-specific social support Received social support Social conflict

Introduction Individuals with multiple chronic conditions (multimorbidity) often face the problem of complex treatment regimens, as the number of medications increases with the number of diseases (Tinetti, Bogardus, & Agostini, 2004). Whereas adherence rates are relatively high among patients with acute conditions, those with chronic conditions often fail to follow their prescribed treatment reasonably closely during the long-term course of their illness (e.g., Doggrell, 2010). Poor medication adherence, however, adds to the burden of multimorbidity, as it can worsen overall health status, lead to medication-related hospital admissions, and increase mortality (Simpson et al., 2006). A substantial body of research has examined predictors of adherence to medication as prescribed. Beyond demographic and * Corresponding author. Health Psychology, Freie Universität Berlin, Habelschwerdter Allee 45,14195 Berlin, Germany. Tel.: þ49 30 83855179; fax: þ49 30 83855634. E-mail address: [email protected] (L.M. Warner). 0277-9536/$ e see front matter Ó 2013 Elsevier Ltd. All rights reserved. http://dx.doi.org/10.1016/j.socscimed.2013.03.012

socio-economic factors (e.g., age, gender, education), characteristics of the disease and treatment (e.g., chronic versus acute diseases, number of medicines), and individual resources (e.g., health status, medication beliefs), in particular the effects of social resources (e.g., social network characteristics, perception of support, receipt of support) have been researched intensively (DiMatteo, 2004a, 2004b; Doggrell, 2010; Schüz, Marx, et al., 2011; Schüz, Wurm, et al., 2011). However, the effects of these social resources on health are not unequivocally positive. There is consistent evidence that social interactions such as social conflict or receiving social support can also negatively affect physical and mental health and specific health behaviours, particularly in older adults and those with chronic conditions (Hays, Saunders, Flint, Kaplan, & Blazer, 1997; Scholz et al., 2012; Uchino, 2009). These negative effects have been explained with regard to mobilization effects (low adherence mobilizing higher social support resulting in a negative relation; Uchino, 2009; Väänänen, Vahtera, Pentti, & Kivimäki, 2005), mismatches in the operationalization of outcome and social support (Aaronson, 1989; Tay et al., 2013), or the

24

L.M. Warner et al. / Social Science & Medicine 87 (2013) 23e30

effects of social conflict between provider and recipient, which might render support detrimental (Holt-Lunstad, Uchino, Smith, & Hicks, 2007). Together with the multidimensional nature of social resources, these inconsistent effects of social resources on adherence in older adults require a more nuanced approach, considering different types and levels of specificity of social influences. Social resources and risk factors for health and well-being in older adults Received (enacted) social support describes the experience of receiving help, whether or not this help was solicited (Uchino, 2009). There is evidence for positive effects of receiving social support, e.g., on effective coping and physical functioning in persons with chronic diseases (for an overview, see Schwarzer & Knoll, 2007). However, other studies find no relations (e.g., Brown, Nesse, Vinokur, & Smith, 2003) or even negative relations between received support and physical and mental health, especially in older and chronically ill adults (Hays et al., 1997; Seeman, Bruce, & McAvay, 1996). These negative effects were found for both receiving tangible social support that negatively affects functional health (Uchino, 2009) and emotional support that has been found to negatively affect quality of life (Warner, Schüz, Wurm, Ziegelmann, & Tesch-Römer, 2010). One explanation for negative effects of receiving support might be that it is beneficial only at times when the recipient needs help, requests support, and when the support needs are met, hence when adequate support is received (Bolger & Amarel, 2007). Although research has shown that social support measures in a non-behaviour-specific format can predict health behaviour to some extent (Uchino, 2009), behaviour-specific operationalizations of received social support are assumed to be more predictive of health behaviours (Aaronson, 1989). General measures of support should, therefore have lower associations with specific health measures due to a failure to capture the nature of support in response to specific health behaviour-related needs (Tay et al., 2013). A further potential explanation for negative relationships between received support and health is that those who are in greatest need (i.e., who are experiencing significant adverse events that threaten health) mobilize more social support from their environment in response to their need (mobilization hypothesis). Over time, these relations should reverse: whereas cross-sectionally received support coincides with adverse events, the longitudinal outcomes of received support should be positive (Schwarzer & Leppin, 1991; Uchino, 2009; Väänänen et al., 2005). Since negative effects of receiving support have been observed in both cross-sectional and longitudinal studies, an alternative explanation of the adverse effects of receiving social support might be the social conflict hypothesis. Social conflict describes the potentially negative sides of social interactions, including the expression of negative affect, disregard and disaffirmation. It has consistently been associated with adverse effects for health and well-being in older and chronically ill adults (Everson-Rose & Lewis, 2005; Krause & Rook, 2003). This indicates that individuals may experience high levels of support and social conflict simultaneously, which implies no or only small correlations between receiving support and social conflict (Argyle & Furnham, 1983). The presence of conflict in relationships does not only affect outcomes directly, but may be responsible for rendering the effects of support on wellbeing negative, thus acting as a moderator (Holt-Lunstad et al., 2007; Liang, Krause, & Bennett, 2001). This is of particular relevance for older adults, as they were found to experience high levels of social conflict, mostly with their children and family (Krause & Rook, 2003).

Social support and medication adherence These considerations are particularly important in the domain of adherence, since a meta-analysis revealed that most studies of the relationship of social influences to medication adherence have employed general and non-behaviour-specific social support measures (DiMatteo, 2004a). DiMatteo found that unidimensional measures of social resources such as structural network characteristics or general received and anticipated social support relate positively to medication adherence in the aged and chronically ill (DiMatteo, 2004a). However, general social support has exhibited negative effects on adherence under certain circumstances as well (Hamilton, Razzano, & Martin, 2007). It has been suggested that characterizing social support in a behaviour-specific manner, such as asking recipients whether interaction partners reminded them to take, bought or organized their medication, should provide better prediction of medication adherence (Aaronson, 1989; Tay et al., 2013). Few studies have measured the receipt of medication-specific social support, and results of these studies are inconsistent. For example, Stirratt et al. (2006) found that receiving medication-specific support predicted adherence in HIV-positive patients. However, in a study of hypertensive older adults with a co-morbid cardiovascular disease, medication-specific support predicted worse medication adherence (Friedberg et al., 2009) and in a study on patients with epilepsy, medication-specific support predicted increased anxiety, which predicted decreased medication self-management (DiIorio et al., 1996). Studies of patients with diabetes further reported practical assistance with metabolic control to be unrelated to adherence (Burroughs, Pontious, & Santiago, 1993). A study of HIV patients found received support (consisting of both emotional support and medication-specific support) to be unrelated to adherence (Simoni, Frick, Lockhart, & Liebovitz, 2002), whereas patients with organ transplantation reported both negative and positive effects of receiving support on their medication adherence (Scholz et al., 2012). Although these studies operationalized social support in a behaviour-specific format, they generated inconsistent results. Aims of the study This study therefore aims to clarify the role of social resources and risk factors (support and conflict) on medication adherence over time in multimorbid older adults e a high-risk population with both complex medication regimens and a particular need for improvement in medication adherence. We examined associations of general and medication-specific support with medication adherence to find out whether received medication-specific social support relates more strongly to adherence than general received support. Previous studies have considered support as a driver behind adherence, and have not considered the potential mobilizing role of non-adherence in eliciting support. To disentangle this and explain possible negative associations between medication-specific support and adherence, we examined bidirectional relationships over a six-month period to explore the mobilization hypothesis (whether lower adherence evokes more support from the network). We also examine whether social conflict acts as a moderator in this relationship. The study hence aims to concurrently test the mobilization (Väänänen et al., 2005), and social conflict hypotheses (Holt-Lunstad et al., 2007). Method Participants and procedure Participants for this study were recruited from the third assessment wave of the German Ageing Survey (DEAS, Wurm, Tomasik,

L.M. Warner et al. / Social Science & Medicine 87 (2013) 23e30

& Tesch-Römer, 2010), a population-representative survey of community-dwelling adults aged 40 and over, with a total N of 8200 between July and September 2008. DEAS participants were considered eligible for the present study if they were a) 65 years or older, b) suffered from at least two chronic physical conditions mentioned either in the Charlson Comorbidity Index (Charlson, Szatrowski, Peterson, & Gold, 1994) or the Functional Comorbidity Index (Groll, To, Bombardier, & Wright, 2005) and c) had given consent to be contacted for further studies. Ethical consent was granted from the appropriate body (ethics commission of the German Psychological Society: DGPs-SW.012009). Of a total eligible n ¼ 443 participants, n ¼ 309 (69.7%) provided informed consent and made an appointment for the first measurement point (Time 1, March 2009). Participants were visited at their homes by trained interviewers, where they completed a 30min personal interview. Following interviewer departure, participants completed a self-report questionnaire and returned it in a prepaid envelope. In total, 305 of 309 interviewed participants returned the questionnaire. At Time 2 (September 2009), n ¼ 277 participants (88.03% of Time 1 sample) completed the interview, and n ¼ 272 of them returned the questionnaire. Participants at Time 1 were on average 73.27 years of age (SD ¼ 5.10), and 41.7% were women. Measures Independent variables General received social support and social conflict were assessed in the Time 1 questionnaire with 6 and 4 items, respectively, from the German version of the Illness-specific Social Support Scale (ISSS, Ramm & Hasenbring, 2003; Revenson, Schiaffino, Majerovitz, & Gibofsky, 1991). The ISSS was abbreviated to 10 items by excluding illness-specific items that were unsuited for a multimorbid target population and by selecting the most valid and reliable items based on Ramm and Hasenbring (2003). Examples for the six general received support items are: ‘If you think of the previous three months, what have persons close to you (partner, children, friends, acquaintances) done for you? Persons close to me. a) took care of many things for me, b) talked to me about important decisions, c) gave me the feeling that I can rely completely on them’. Item examples for the four social conflict items are: ’If you think of the previous three months, what have persons close to you (partner, children, friends, acquaintances) done for you? Persons close to me. a) found it hard to understand the way I felt, b) became annoyed when I didn’t accept their advice’. Participants rated the frequency with which they received support or experienced conflict on a four-point scale from (1) ‘(almost) never’ to (4) ‘(almost) always’. Mean scores were computed from responses on both scales. Higher scores reflect higher general received support and higher social conflict. Cronbach’s alpha for the six general received support items was 0.83, and 0.55 for the four social conflict items. As Cronbach’s alpha for social conflict was low, we conducted an additional confirmatory factor analysis in Mplus to examine the factorial structure underlying the instrument. We compared a one-factorial model (all 10 items load on one factor) with the assumed twofactorial model (four items indicating social conflict and six items indicating general received support) and found the two-factorial model to fit the data significantly better, Dc2 ¼ 86.95, df ¼ 1, p < .001. Medication-specific social support was assessed in the Time 1 questionnaire with two items adapted from the Medical Care Questionnaire (Sayers, White, Zubritsky, & Oslin, 2006). The Medical Care Questionnaire encompasses items on families’ and friends’ support for patients’ overall medical and self-care decisions. The contents of the Medical Care questions were integrated into two

25

items that reflect tangible received social support for medication adherence only (the original items relating to doctors’ appointments were not used in this study, because of their equivocal relation to medication adherence). Participants were asked: ‘If you think of the previous three months, what have persons close to you (partner, children, friends, acquaintances) done for you? Persons close to me a) have reminded me to take my medication regularly, b) have handled my medication (e.g., have fetched or organized it)’. Answers ranged from (1) ‘(almost) never’ to (5) ‘(almost) always’. These two items were correlated (r ¼ 0.58, p < .001) at both points in time. Primary outcome Medication adherence was assessed in the Time 2 questionnaire with the item assessing unintentional nonadherence from the Reported Adherence to Medication Scale (RAM; Horne, Weinman, & Hankins, 1999): ‘Some people forget to take their medicines. How often does this happen to you?’ with responses on a 5-point scale from (1) ‘(almost) never’ to (5) ‘(almost) always’. The RAM was found to have good reliability, criterion-related and discriminant validity (Horne et al., 1999). Unintentional nonadherence (forgetting to take medication) is a more relevant outcome for health in older adults with multiple illnesses, as intentional adjustments of medication might be indicated in this population due to adverse drug interactions and potentially inappropriate prescriptions (Klarin, Wimo, & Fastbom, 2005). Though continuous measures of adherence are preferable, the distribution of this measure was highly skewed, with 73.1% of the participants reporting complete adherence at Time 1 and 68.0% at Time 2. Thus we dichotomized the scale into 1 for full adherence and 0 for non-adherence (reflecting any answer from 1 to 4). By defining only the ‘(almost) always’ answer as adherence, we used the suggested and conservative cut-off point of 100% adherence (Pearson, Simoni, Hoff, Kurth, & Martin, 2007). Covariates Covariates were sex, age, education, number of medicines, physical fitness and attitude towards medication, because of their well-established relation to medication adherence (DiMatteo, 2004a, 2004b; Doggrell, 2010; Schüz, Marx, et al., 2011; Schüz, Wurm, et al., 2011). Number of medications was assessed with a computer-assisted full medication inventory in the Time 1 interview (Psaty et al., 1992). Because the interviews took place at participants’ houses, interviewers asked participants to bring all of the medicines they currently took. The national drug code for each medication was then entered in the electronic records. Education level was assessed in the German Ageing Survey one year before this study at Time 0 and classified according to the International Standard Classification of Education (ISCED, Unesco, 1997), with (1) indicating low education (at most 9 years school education), (2) indicating medium education (secondary school) and (3) indicating high education (qualifying for university admission). Physical fitness was measured in the Time 1 interview with a peak expiratory flow meter that assesses the maximum pulmonary expiratory flow. It has been shown that this is a reliable and sensitive indicator of general physical fitness in older and frail adults (Nunn & Gregg, 1989). Participants exhale into the instrument twice with maximum effort. Scores could range from 60 to 800 L per minute, and the better result of the two trials was taken as the measure of fitness. Higher scores indicate better fitness status. Attitudes towards medication were assessed in the Time 1 questionnaire with one item adapted from Horne et al. (1999),

10

0.43***

11

because patients’ beliefs about the efficacy of their medication were found to be strongly related to adherence and should generalize to adults with multiple medication prescriptions: ‘My medication or my treatment regimen is effective for my disease’(1) ‘totally disagree’ to (4) ‘totally agree’.

0.37***

L.M. Warner et al. / Social Science & Medicine 87 (2013) 23e30

0.22***

26

0.22*** 0.06 0.07 0.02 0.08 0.01 0.01 0.05

0.11 0.12 <0.01 0.04 0.24*** 0.27*** 0.02 0.19** 0.01 0.07 0.09 0.06 0.18** 0.15** 0.04 <0.01 0.03 0.05 0.18** 0.22 0.19** 0.14* 0.10 0.16* 0.03 0.09 0.36*** 0.08 0.12 0.09 0.02 0.04 0.06 0.07 0.10 0.24*** 0.11 0.08 0.03 0.14* <0.01 0.01 0.09 0.41*** 0.03 0.50*** 0.04 0.19** 0.01 0.09 0.11 0.06

0.03 Note. *p < .05; **p < .01; ***p < .001.

Table 1 Means, standard deviations, ranges, and correlations.

68% adherent 32% non-adherent

Characteristics of n ¼ 309 participants are reported in Table 1. Around 12.6% indicated low, 52.1% medium, and 35.3% high education. Participants came from all regions of Germany, with n ¼ 108 (35%) living in the eastern federal states (former German Democratic Republic). Participants had on average 5.49 chronic conditions (SD ¼ 2.86) at Time 1, with hypertension (67.6%), osteoarthritis (63.1%), hyperlipidemia (49.2%), arthritis (31.1%) and peripheral vascular disease (30.7%) being the five most prominent conditions. Participants took 4.26 medicines (SD ¼ 2.96) on average. Only 19% of participants took only one medication daily, whereas 28% took five or more. Table 1 shows correlations between the raw scores of all variables. The significant correlation between gender and physical fitness, as indicated by peak expiratory flow, is due to genderspecific differences in un-weighted peak expiratory flow scores (Nunn & Gregg, 1989). Physical fitness (peak expiratory flow) correlated with education even with gender controlled (rpartial(300) ¼ 0.19, p < .01), indicating better fitness in better educated participants. Age was positively related to receipt of medication-specific support, as was the number of medications.

M

Participant characteristics

65e85 1e3 1e17 70e780 1e4 1e4 1e3.25 1e5 1e5

9 8 7 6 5 4 3 2 1 Range SD

Results

42% 5.10 0.66 2.96 134.48 1.08 0.68 0.42 0.88 0.82 78% adherent

Those 37 participants who dropped out between Time 1 and Time 2 were examined for significant differences in the study variables at Time 1. The main reasons for drop-out were hospital admissions, serious health problems and general reasons (e.g., no time or interest in another interview and questionnaire). There were no differences between participants, who dropped out and those who stayed in the study in terms of medication adherence c2(1) ¼ 0.21, p > .05 (drop-outs 30.30% non-adherent, completers 26.51% nonadherent), gender c2(1) ¼ 0.34, p > .05 (drop-outs 37.14% women, completers 42.36% women), age t(307) ¼ 0.41, p > .05 (Mdropouts ¼ 73.60, SEdrop-outs ¼ 5.66; Mcompleters ¼ 73.23, SEcompleters ¼ 5.03), education t(307) ¼ 0.29, p > .05 (Mdrop-outs ¼ 2.26, SEdropouts ¼ 0.61; Mcompleters ¼ 2.22, SEcompleters ¼ 0.66), physical fitness t(301) ¼ 0.46, p > .05 (Mdrop-outs ¼ 335.16, SEdrop-outs ¼ 146.22; Mcompleters ¼ 346.73, SEcompleters ¼ 133.26), number of medication t(307) ¼ 0.18, p > .05 (Mdrop-outs ¼ 4.34, SEdrop-outs ¼ 2.93; Mcompleters ¼ 4.25, SEcompleters ¼ 2.97), social conflict t(300) ¼ 0.0005, p > .05 (Mdrop-outs ¼ 1.42, SEdrop-outs ¼ 0.41; Mcompleters ¼ 1.42, SEcompleters ¼ 0.42), general support t(300) ¼ 0.64, p > .05 (Mdrop-outs ¼ 2.57, SEdrop-outs ¼ 0.62; Mcompleters ¼ 2.65, SEcompleters ¼ 0.69) or medication-specific support t(295) ¼ 0.15, p > .05 (Mdrop-outs ¼ 1.45, SEdrop-outs ¼ 1.06; Mcompleters ¼ 1.43, SEcompleters ¼ 0.85).

58% 73.27 2.23 4.26 345.51 3.05 2.65 1.42 1.43 1.43 27% non-adherent

Attrition analysis

1. Sex (0 ¼ women, 1 ¼ men) 2. Age 3. Education 4. Number of medicines 5. Physical fitness 6. Attitude towards medication 7. General received Support 8. Social conflict 9. Medication-specific social support T1 10. Medication-specific social support T2 11. Medication adherence T1 (0 ¼ non-adherent, 1 ¼ adherent) 12. Medication adherence T2 (0 ¼ non-adherent, 1 ¼ adherent)

Descriptive data analysis, Pearson correlations and logistic regressions were carried out with SPSS 18.0. The cross-lagged panel analysis was performed using the WLSMV estimator in Mplus 5.21. Simple slopes analyses were used to detect the direction of the interaction term (Aiken & West, 1991) and transformed to probabilities for dichotomous outcomes (Cohen, Cohen, West, & Aiken, 2003).

0.54*** 0.18**

Analytic procedure

L.M. Warner et al. / Social Science & Medicine 87 (2013) 23e30

The number of medications was positively associated of general support and also with social conflict. Only specific support at both time points was associated ence both cross-sectionally and over time; all these were negative.

with receipt medicationwith adherassociations

Prediction of adherence and medication support in cross-lagged panel analysis Table 1 shows that adherence T1 and medication-specific support T2 had a significant negative relationship of precisely the same magnitude as medication-specific support at Time 1 and adherence at Time 2 (r ¼ 0.22, p < .001 in both cases). Medication-specific support was somewhat more stable over time (r ¼ 0.54, p < .001) than was adherence (r ¼ 0.37, p < .001). To examine the mobilization hypothesis, we investigated the bidirectional relationships between medication-specific support and adherence over time. We performed a cross-lagged panel analysis in which Time 1 medication-specific support predicted Time 2 adherence with Time 1 adherence controlled, and Time 1 adherence predicted Time 2 medication-specific support with Time 1 support controlled. Sex, age, education, number of medicines, physical fitness and attitude towards medication were included as covariates predicting Time 2 adherence and Time 2 medicationspecific support. Fig. 1 shows that higher initial adherence was negatively related to later receipt of medication-specific social support (b ¼ 0.12, p < .001). However, high medication support at Time 1 was also associated with lower adherence at Time 2 (b ¼ 0.19, p < .05). Apart from a positive relation between general social support and medication-specific support (b ¼ 0.11, p < .001), none of the other covariates had a significant relation to the outcomes. This model explained 38% of the variance in medication-specific social support and 27% of the variance in medication adherence. To test whether the path between medication support Time 1 and adherence Time 2 and the path between adherence Time 1 and medication support Time 2 were significantly different, a second cross-lagged panel analysis was performed with these paths constrained to be equal. Chi square difference tests for these two models showed that the more restrictive model (in which the paths were restrained to be equal) fit the data as well as the model with freely estimated parameters (Dc2(1) ¼ 0.002, p > .05). This means that the path from previous adherence to later support did not significantly differ from the path from previous support to later adherence. In sum, participants reported receiving more support when they had been less adherent previously. However, they also reported that receiving medication-specific support had a negative relation to their later medication adherence.

Fig. 1. Cross-lagged panel analysis. Note. *p < .05, **p < .01, ***p < .001. Medication adherence and medication-specific support at Time 2 were statistically controlled for sex, age, education, number of medicines, physical fitness, attitude towards medication, general received social support, social conflict and baseline of both measurements; T1 ¼ Time 1, T2 ¼ Time 2.

27

Prediction of adherence in logistic regression To test the social conflict hypothesis as an explanation for the negative effects between medication-specific support and adherence, we conducted a logistic regression analysis predicting Time 2 adherence, while statistically controlling for adherence at Time 1. The results for the logistic regression are reported in Table 2. Medication adherence was regressed on baseline adherence, all covariates, general received social support and social conflict in Step 1 of the logistic regression analysis. Apart from baseline adherence (OR ¼ 1.79, 95% CI 3.18e11.29, p < .001) neither the covariates nor general social support or social conflict predicted medication adherence at Time 2. Medication-specific social support was added to the prediction equation at Step 2. Medication-specific support increased the amount of predicted variance in medication adherence at Time 2 (Dc2 (1, N ¼ 236) ¼ 4.25, p < .05) over and above all other predictors, including prediction from baseline adherence (OR ¼ 5.37, 95% CI 2.82e10.24, p < .001). Medication-specific social support was negatively related to later medication adherence (OR ¼ 0.69, 95% CI 0.49e0.98, p < .05). In Step 3, we added the interaction between social conflict and medication-specific social support. Model fit improved (Dc2 (1, N ¼ 236) ¼ 4.40, p < .05), with OR ¼ 0.42 (95% CI 0.17e0.99, p < .05) for the interaction term. To characterize the nature of the interaction effect, we used simple slopes analyses (Aiken & West, 1991). Fig. 2 depicts the association of received medication-specific support at Time 1 with the predicted probability of medication adherence at Time 2 at three levels of social conflict at Time 1: at the value 1 on the social conflict scale, indicating no social conflict, at the value of 2, indicating social conflict that occurred sometimes and at the value of 3.25, which was the highest observed value in the data, indicating social conflict that occurred slightly more than often. The simple slopes analyses revealed that the negative relation between medication-specific support and medication adherence was significant only for individuals with social conflict that occurred sometimes (B ¼ 0.86, p < .01) or often (referring to the highest observed value of 3.25; B ¼ 1.73, p < .05). In individuals with no social conflict, medication-specific social support was not significantly related to later medication adherence (B ¼ 0.02, p > .05). Discussion In this study, we examined the role of general received social support, medication-specific social support and social conflict on medication adherence. Medication-specific social support emerged as the only significant predictor of medication adherence (apart from baseline medication adherence). In line with previous research on other health behaviours, this study found a negative association between received social support for medication adherence and medication adherence (Martire, Stephens, Druley, & Wojno, 2002; Silverman, Hecht, & McMillin, 2002). This negative relation may be explained by the mobilization hypothesis, which poses that older adults with deficits in their health-related behaviours might mobilize the support of network members. According to this hypothesis, being worse off in the first place activates coping processes such as disclosure or the search for social support. These coping processes mobilize social support from a network and might explain the negative relations between receiving social support and positive outcomes (Uchino, 2009; Väänänen et al., 2005). In the long run however, the mobilization hypothesis assumes a positive effect of support. To investigate the mobilization hypothesis, we tested the mutual relation of medication-specific support and medication

28

L.M. Warner et al. / Social Science & Medicine 87 (2013) 23e30

Table 2 Logistic regression to predict medication adherence. Predictor

Step 1

Step 2

B (SE) Constant Adherence Time 1 Sex Age Education Number of Medicines Physical fitness Attitude towards medication General received Support Social Conflict Medication-specific social support Interaction med.-support * conflict

0.26 1.79*** 0.12 0.01 0.15 0.02 0.002 0.09 0.44 0.44

(2.66) (0.32) (0.42) (0.03) (0.26) (0.05) (0.002) (0.15) (0.25) (0.35)

Odds ratio (Lower, Upper 95% CI)

B (SE)

0.77 5.99 1.14 1.01 1.16 1.02 9.998 1.09 0.64 0.65

0.79 1.68*** 0.25 0.02 0.11 0.01 0.002 0.11 0.31 0.40 0.37*

(3.18, 11.29) (0.50, 2.57) (0.95, 1.08) (0.69, 1.94) (0.92, 1.13) (0.995, 1.002) (0.81, 1.47) (0.40, 1.05) (0.32, 1.29)

Step 3

(2.71) (0.33) (0.42) (0.03) (0.27) (0.05) (0.002) (0.15) (0.26) (0.36) (0.18)

Odds ratio (Lower, Upper 95% CI)

B (SE)

0.45 5.37 1.28 1.02 1.11 1.01 0.998 1.11 0.73 0.67 0.69

1.15 1.73*** 0.15 0.02 0.18 0.01 0.001 0.12 0.31 0.35 0.35 0.88*

(2.82, 10.24) (0.56, 2.94) (0.95, 1.08) (0.66, 1.89) (0.91, 1.12) (0.995, 1.002) (0.83, 1.50) (0.44, 1.21) (0.33, 1.35) (0.49, 0.98)

Odds ratio (Lower, Upper 95% CI) (2.74) (0.34) (0.43) (0.03) (0.27) (0.06) (0.002) (0.16) (0.26) (0.38) (0.19) (0.45)

0.32 5.62 1.17 1.02 1.20 1.01 0.999 1.13 0.73 0.71 0.70 0.42

(2.91, 10.85) (0.50, 2.70) (0.95, 1.09) (0.70, 2.05) (0.91, 1.12) (0.996, 1.002) (0.83, 1.53) (0.44, 1.22) (0.35, 1.49) (0.49, 1.02) (0.17, 0.99)

Note. *p < .05, **p < .01, ***p < .001.

adherence over time and found that the constructs are reciprocally associated. On the one hand, participants reported receiving more medication-specific support when they were previously less adherent. On the other hand, more medication-specific social support was associated with later non-adherence, as well. Hence, the mobilization hypothesis did not suffice to explain the on-going process in this sample. Another explanation for negative associations between received social support, health, and health behaviour outcomes is that support might not be perceived to be helpful if it coincides with social conflict (Liang et al., 2001). In fact, the closest network members can be major sources of social support and social conflict at the same time (Argyle & Furnham, 1983). The interaction between social conflict and medication-specific social support in our study supports the assumption that the negative relation between medication-specific social support and medication adherence can

be traced to those individuals who report experiencing social conflict at least sometimes. This is in line with previous research showing that the impact of social support can be rendered negative, if too much social conflict is present (Liang et al., 2001). For individuals who report that they experience no social conflict, however, the expected positive relation between medication support and medication adherence did not emerge. To further investigate why participants without conflict report lower adherence when they receive higher levels of medicationspecific support, future studies should assess further personal resources such as self-efficacy for medication adherence (Simoni et al., 2002). The inclusion of measures of both overprotection and social control in future research may provide a more complete understanding of the mechanisms that underlie the negative link between support and adherence (e.g., Everson-Rose & Lewis, 2005).

Limitations

Fig. 2. Simple slopes analysis depicting the interaction between medication-specific support and social conflict. Note. Reported are probabilities to adhere to medication regimens and centred medication-specific social support values. Medication adherence Time 2 was statistically controlled for Time 1 medication adherence, sex, age, education, number of medicines, physical fitness, attitude towards medication, general received social support and social conflict. The three social conflict levels are the raw score values of 1 ¼ “no conflict”, 2 ¼ “sometimes conflict” and 3.25 ¼ the highest observed value, which refers to “often conflict”.

With regard to the measures in this study, future research would benefit from a more complete characterization of the multiple sources of medication-specific social support. We asked participants to rate medication-specific support from family and friends only, as support from significant others seems to be more relevant in most domains of life (Hogan, Linden, & Najarian, 2002). However, it would be of great importance to assess medication-specific social support from health care providers in future research as well, because they were found to provide considerable support to chronically ill adults (Hamilton et al., 2007). Medication adherence was assessed via one self-report item only. This might have led to over-reporting, as adherence rates ranged between 68 and 70% in this study and could be due to a social desirability bias. Even though previous comparative research showed that self-reported measures of adherence can be considered valid, future research, especially on older adults with multiple chronic conditions, should consider assessing adherence for each medication and use more objective methods such as electronic adherence measurement or pharmacy refill (Pearson et al., 2007). Because memory biases or medication that is not stored at home might have affected the assessment of the amount of medication in this sample, an ideal measurement would have been a comparison with medical records. Although adherence ratings are often dichotomized, the use of continuous measurements should be preferred in future studies, if the distribution of adherence is more widely distributed (Pearson et al., 2007). Social conflict was measured with four very broad items in this study, which makes it impossible to determine whether

L.M. Warner et al. / Social Science & Medicine 87 (2013) 23e30

medication-specific social support was indeed a source of conflict or whether reported social conflict derived from different life domains or interaction partners. Qualitative studies might shed light on these mechanisms and should assess the identities of the major providers of medication-specific support in detail and whether the supporting behaviour per se leads to social conflict or whether social conflict is present for different reasons and obscures the positively intended acts of the medication support provider. Our measure of social conflict had no satisfactory reliability, suggesting that this construct is quite heterogeneous and should be assessed with different measures in future studies. Our usage of raw, un-weighed peak expiratory flow data could be considered a limitation of the study, in particular because this created artificial correlations between gender and physical fitness (Table 1). However, as the transformation formula between genders (Nunn & Gregg, 1989) is both dated and based on non-multimorbid individuals from the US, a non-equivalent transformation such as weighing might have biased the distribution of peak expiratory flow scores. As peak expiratory flow was used as a control for spurious relations between physical fitness and medication adherence, raw data was deemed sufficient. Further, sex of participants was included as a control variable along with physical fitness, thereby partialling sex from the relationship of our measure for physical fitness to outcomes. Future studies should not rely on a single measure of physical fitness, but incorporate measures such as (instrumental) activities of daily living to give consideration to the disabilities that older adults with multiple chronic conditions encounter, and which make them needier of social support in everyday live (Katz, Down, Cash, & Grotz, 1970). Even though we tried to exclude alternative explanations and showed in the cross-lagged panel analysis that medication-specific social support and medication adherence relate to one another reciprocally over time, causal assumptions cannot be drawn from this longitudinal study and more adequate study designs (such as randomized controlled trials) are needed. Implications Notwithstanding these limitations, the results of our study bear some important implications for older adults with multimorbidity e a group at high risk for medication non-adherence due to highly prevalent and complicated treatment regimens (e.g., Tinetti et al., 2004). Older adults with multimorbidity are challenged in several ways: on the one hand, the need for social support grows with health deterioration, but on the other hand, older adults often disapprove support in order to sustain their sense of self-esteem and autonomy (Fisher, Nadler, & Whitcher-Alagna, 1982; Warner et al., 2010, 2011). They have to cope with a number of different diseases and experience higher social conflict than younger adults at the same time (Krause & Rook, 2003). Because of this double burden, harmful effects of receiving support have been found in several previous studies (e.g., Brown et al., 2003). If negative effects of receiving support are not explainable by alternative explanations e e.g., the mobilization hypothesis e it is reasonable to examine the role of negative social experiences within the social network, including social conflict, overprotection, or social control. Such negative interplay between individuals may be a major source of the perception that support provided by others is not, in fact, supportive from the perspective of the recipient (e.g., Everson-Rose & Lewis, 2005). This suggests that the impact of social resources on health and health behaviour in older adults must be investigated carefully to find the fine line between those social resources that are helpful and those that might act as a “double-edged sword” (Revenson et al., 1991). It is important to equip providers and especially older and multimorbid recipients of social support with good

29

communication skills to ensure that well-intended social support fosters disease management rather than adding to the illness burden. A promising approach for future studies on older adults with high disease burden and their support providers is, therefore, a social skills training (Krause & Shaw, 2000). Acknowledgements The German Aging Survey was funded under Grant 301-1720-2/ 2 by the German Federal Ministry for Family, Senior Citizens, Women, and Youth. For the present study, the first and the second author are funded by the German Federal Ministry of Education and Research (Grant No. 01ET0702); the fourth author is funded by Grant No. 01ET0801 by the same funding body. The content is the sole responsibility of the authors. References Aaronson, L. S. (1989). Perceived and received support: effects on health behavior during pregnancy. Nursing Research, 38, 4e9. Aiken, L. S., & West, S. G. (1991). Multiple regression: Testing and interpreting interactions. Newbury Park, CA: Sage. Argyle, M., & Furnham, A. (1983). Sources of satisfaction and conflict in long-term relationships. Journal of Marriage & Family, 45, 481e493. Bolger, N., & Amarel, D. (2007). Effects of social support visibility on adjustment to stress: experimental evidence. Journal of Personality and Social Psychology, 92, 458e475. Brown, S. L., Nesse, R. M., Vinokur, A. D., & Smith, D. M. (2003). Providing social support may be more beneficial than receiving it: results from a prospective study of mortality. Psychological Science, 14, 320e327. Burroughs, T. E., Pontious, S. L., & Santiago, J. V. (1993). The relationship among six psychosocial domains, age, health care adherence, and metabolic control in adolescents with IDDM. The Diabetes Educator, 19, 396e402. Charlson, M. E., Szatrowski, T. P., Peterson, J., & Gold, J. (1994). Validation of a combined comorbidity index. Journal of Clinical Epidemiology, 47, 1245e1251. Cohen, J., Cohen, P., West, S. G., & Aiken, L. S. (2003). Applied multiple regression/ correlation analysis for the behavioral sciences (3rd ed.). Mahwah: NJ US: Lawrence Erlbaum Associates Publishers. DiIorio, C., Hennessy, M., & Manteuffel, B. (1996). Epilepsy self-management: a test of a theoretical model. Nursing Research, 45(4), 211e217. http://dx.doi.org/ 10.1097/00006199-199607000-00004. DiMatteo, M. R. (2004a). Social support and patient adherence to medical treatment: a meta-analysis. Health Psychology, 23, 207e218. DiMatteo, M. R. (2004b). Variations in patients’ adherence to medical recommendations: a quantitative review of 50 years of research. Medical Care, 42, 200e209. Doggrell, S. A. (2010). Adherence to medicines in the older-aged with chronic conditions: does intervention by an allied health professional help? Drugs and Aging, 27, 239e254. Everson-Rose, S. A., & Lewis, T. T. (2005). Psychosocial factors and cardiovascular diseases. Annual Review of Public Health, 26, 469e500. Fisher, J. D., Nadler, A., & Whitcher-Alagna, S. (1982). Recipient reactions to aid. Psychological Bulletin, 91, 27e54. Friedberg, J. P., Robinaugh, D., Ulmer, M. E., Antonopoulos, M. S., Sathe, N., & Natarajan, S. (2009). Does comorbid cardiovascular disease affect antihypertensive medication adherence? Baltimore: Health Services Research & Development Service. Groll, D. L., To, T., Bombardier, C., & Wright, J. G. (2005). The development of a comorbidity index with physical function as the outcome. Journal of Clinical Epidemiology, 58, 595e602. Hamilton, M. M., Razzano, L. A., & Martin, N. B. (2007). The relationship between type and quality of social support and HIV medication adherence. Journal of HIV/AIDS & Social Services, 6, 39e63. Hays, J., Saunders, W., Flint, E., Kaplan, B., & Blazer, D. (1997). Social support and depression as risk factors for loss of physical function in late life. Aging & Mental Health, 1, 209e220. Hogan, B. E., Linden, W., & Najarian, B. (2002). Social support interventions: do they work? Clinical Psychology Review, 22, 381e440. Holt-Lunstad, J., Uchino, B. N., Smith, T. W., & Hicks, A. (2007). On the importance of relationship quality: the impact of ambivalence in friendships on cardiovascular functioning. Annals of Behavioral Medicine, 33, 278e290. Horne, R., Weinman, J., & Hankins, M. (1999). The beliefs about medicines questionnaire: the development and evaluation of a new method for assessing the cognitive representation of medication. Psychology & Health, 14, 1. Katz, S., Down, T. D., Cash, H. R., & Grotz, R. C. (1970). Progress in the development of the index of ADL. The Gerontologist, 10, 20e30. Klarin, I., Wimo, A., & Fastbom, J. (2005). The association of inappropriate drug use with hospitalisation and mortality: a population-based study of the very old. Drugs and Aging, 22, 69e82. http://dx.doi.org/10.2165/00002512-20052201000005.

30

L.M. Warner et al. / Social Science & Medicine 87 (2013) 23e30

Krause, N., & Rook, K. S. (2003). Negative interaction in late life: issues in the stability and generalizability of conflict across relationships. The Journals of Gerontology: Series B, Psychological Sciences and Social Sciences, 58, 88e99. Krause, N., & Shaw, B. A. (2000). Giving social support to others, socioeconomic status, and changes in self-esteem in late life. Journals of Gerontology Series B: Psychological Sciences & Social Sciences, 55, 323e333. http://dx.doi.org/10.1093/ geronb/55.6.S323. Liang, J., Krause, N. M., & Bennett, J. M. (2001). Social exchange and well-being: is giving better than receiving? Psychology and Aging, 16, 511e523. Martire, L. M., Stephens, M. A. P., Druley, J. A., & Wojno, W. C. (2002). Negative reactions to received spousal care: predictors and consequences of miscarried support. Health Psychology, 21, 167e176. Nunn, A. J., & Gregg, I. (1989). New regression equations for predicting peak expiratory flow in adults. BMJ, 298, 1068e1070. Pearson, C., Simoni, J., Hoff, P., Kurth, A., & Martin, D. (2007). Assessing antiretroviral adherence via electronic drug monitoring and self-report: an examination of key methodological issues. AIDS and Behavior, 11, 161e173. Psaty, B., Lee, M., Savage, P., Rutan, G., German, P., & Lyles, M. (1992). Assessing the use of medications in the elderly: methods and initial experience in the Cardiovascular Health Study. Journal of Clinical Epidemiology, 646, 683e692. Ramm, G. C., & Hasenbring, M. (2003). Die deutsche Adaptation der Illness-specific Social Support Scale und ihre teststatistische Überprüfung beim Einsatz an Patienten vor und nach Knochenmarktransplantation [The German adaptation of the Illness-specific Social Support Scale and the test statistical evaluation on the basis of patients undergoing bone marrow transplantation]. Zeitschrift für Medizinische Psychologie, 12, 29e38. Revenson, T. A., Schiaffino, K. M., Majerovitz, S. D., & Gibofsky, A. (1991). Social support as a double-edged sword: the relations of positive and problematic support to depression among rheumatoid arthritis patients. Social Science & Medicine, 33, 807e813. Sayers, S. L., White, T., Zubritsky, C., & Oslin, D. W. (2006). Family involvement in the care of healthy medical outpatients. Family Practice, 23, 317e324. Scholz, U., Klaghofer, R., Dux, R., Roellin, M., Boehler, A., Muellhaupt, B., et al. (2012). Predicting intentions and adherence behavior in the context of organ transplantation: gender differences of provided social support. Journal of Psychosomatic Research, 72, 214e219. http://dx.doi.org/10.1016/j.jpsychores.2011.10.008. Schüz, B., Marx, C., Wurm, S., Warner, L. M., Ziegelmann, J. P., Schwarzer, R., et al. (2011). Medication beliefs predict medication adherence in older adults with multiple illnesses. Journal of Psychosomatic Research, 70, 179e187. http:// dx.doi.org/10.1016/j.jpsychores.2010.07.014. Schüz, B., Wurm, S., Ziegelmann, J. P., Warner, L. M., Tesch-Römer, C., & Schwarzer, R. (2011). Changes in functional health, changes in medication beliefs and medication adherence. Health Psychology, 30, 31e39. http://dx.doi.org/10.1037/a0021881. Schwarzer, R., & Knoll, N. (2007). Functional roles of social support within the stress and coping process: a theoretical and empirical overview. International Journal of Psychology, 42, 243e252.

Schwarzer, R., & Leppin, A. (1991). Social support and health: a theoretical and empirical overview. Journal of Social and Personal Relationships, 8, 99e127. Seeman, T. E., Bruce, M. L., & McAvay, G. J. (1996). Social network characteristics and onset of ADL disability: MacArthur studies of successful aging. Journals of Gerontology Series B: Psychological Sciences and Social Sciences, 51, 191e200. Silverman, P., Hecht, L., & McMillin, J. D. (2002). Social support and dietary change among older adults. Ageing & Society, 22, 29. Simoni, J. M., Frick, P. A., Lockhart, D., & Liebovitz, D. (2002). Mediators of social support and antiretroviral adherence among an indigent population in New York City. AIDS Patient Care and STDs, 16, 431e439. http://dx.doi.org/10.1089/ 108729102760330272. Simpson, S. H., Eurich, D. T., Majumdar, S. R., Padwal, R. S., Tsuyuki, R. T., Varney, J., et al. (2006). A meta-analysis of the association between adherence to drug therapy and mortality. BMJ, 333, 1e6. Stirratt, M. J., Remien, R. H., Smith, A., Copeland, O. Q., Dolezal, C., & Krieger, D. (2006). The role of HIV serostatus disclosure in antiretroviral medication adherence. AIDS and Behavior, 10, 483e493. Tay, L., Tan, K., Diener, E., & Gonzalez, E. (2013). Social relations, health behaviors, and health outcomes: a survey and synthesis. Applied Psychology: Health and Well-Being, 5(1), 28e78. http://dx.doi.org/10.1111/aphw.12000. Tinetti, M. E., Bogardus, S. T., & Agostini, J. V. (2004). Potential pitfalls of diseasespecific guidelines for patients with multiple conditions. New England Journal of Medicine, 351, 2870e2874. Uchino, B. N. (2009). Understanding the links between social support and physical health: a life-span perspective with emphasis on the separability of perceived and received support. Perspectives on Psychological Science, 4, 236e255. Unesco. (1997). International standard classification of education. www.uis.unesco. org/Library/Documents/isced97-en.pdf. Väänänen, A., Vahtera, J., Pentti, J., & Kivimäki, M. (2005). Sources of social support as determinants of psychiatric morbidity after severe life events: prospective cohort study of female employees. Journal of Psychosomatic Research, 58, 459e467. Warner, L. M., Schüz, B., Wurm, S., Ziegelmann, J. P., & Tesch-Römer, C. (2010). Giving and taking e differential effects of providing, receiving and anticipating emotional support on quality of life in adults with multiple illnesses. Journal of Health Psychology, 15, 660e670. http://dx.doi.org/10.1177/ 1359105310368186. Warner, L. M., Ziegelmann, J. P., Schüz, B., Wurm, S., Tesch-Römer, C., & Schwarzer, R. (2011). Maintaining autonomy despite multimorbidity: self-efficacy and the two faces of social support. European Journal of Aging, 8, 3e12. http://dx.doi.org/ 10.1007/s10433-011-0176-6. Wurm, S., Tomasik, M. J., & Tesch-Römer, C. (2010). On the importance of a positive view on ageing for physical exercise among middle-aged and older adults: cross-sectional and longitudinal findings. Psychology & Health, 25, 25e42. http://dx.doi.org/10.1080/08870440802311314.