Blackwell Publishing LtdOxford, UKADDAddiction0965-2140© 2006 The Authors. Journal compilation © 2006 Society for the Study of Addiction 2006 101••••

Original Article Network and risk behavior dynamics Elizabeth C. Costenbader et al.

RESEARCH REPORT

doi:10.1111/j.1360-0443.2006.01431.x

The dynamics of injection drug users’ personal networks and HIV risk behaviors Elizabeth C. Costenbader1, Nan M. Astone2 & Carl A. Latkin3 Substance Abuse Treatment Evaluations and Interventions Program, Research Triangle Institute, International, NC, USA,1 Department of Population and Family Health Sciences, Johns Hopkins Bloomberg School of Public Health, Baltimore, MD, USA2 and Department of Health Behavior and Society, Johns Hopkins Bloomberg School of Public Health, Baltimore, MD, USA3

ABSTRACT Aims While studies of the social networks of injection drug users (IDUs) have provided insight into how the structures of interpersonal relationships among IDUs affect HIV risk behaviors, the majority of these studies have been crosssectional. The present study examined the dynamics of IDUs’ social networks and HIV risk behaviors over time. Design Using data from a longitudinal HIV-intervention study conducted in Baltimore, MD, this study assessed changes in the composition of the personal networks of 409 IDUs. We used a multi-nomial logistic regression analysis to assess the association between changes in network composition and simultaneous changes in levels of injection HIV risk behaviors. Using the regression parameters generated by the multi-nomial model, we estimated the predicted probability of being in each of four HIV risk behavior change groups. Findings Compared to the base case, individuals who reported an entirely new set of drug-using network contacts at follow-up were more than three times as likely to be in the increasing risk group. In contrast, reporting all new non-drug-using contacts at follow-up increased the likelihood of being in the stable low-risk group by almost 50% and decreased the probability of being in the consistently high-risk group by more than 70%. Conclusions The findings from this study show that, over and above IDUs’ baseline characteristics, changes in their personal networks are associated with changes in individuals’ risky injection behaviors. They also suggest that interventions aimed at reducing HIV risk among IDUs might benefit from increasing RESEARCH REPORT IDUs’ social contacts with individuals who are not drug users. Keywords

HIV, injection drug users (IDUs), longitudinal, risk behaviors, social networks.

Correspondence to: Elizabeth Costenbader, Substance Abuse Treatment Evaluations and Interventions Program, Research Triangle Institute, International, 3040 Cornwallis Road, Research Triangle Park, NC 27709–2194, USA. E-mail: [email protected] Submitted 23 March 2005; initial review completed 24 May 2005; final version accepted 5 December 2005

INTRODUCTION In the United States, injection drug use is the third leading risk factor for HIV infection. In the state of Maryland, however, injection drug use has been the primary mode of transmission of HIV since 1991. According to the Maryland AIDS Administration, 59% of prevalent HIV cases reported in Baltimore City to December of 2003 were attributable to injection drug use (AIDS Administration Maryland Department of Health and Mental Hygiene 2004). Sharing syringes for injection drugs is a wellknown route of HIV transmission. In addition, several ‘syringe mediated’ or indirect sharing behaviors have been identified that place injection drug users (IDUs) at risk of HIV transmission (Jose et al. 1993; Grund et al. 1996; Koester 1996; Needle et al. 1998). These behaviors include using one individual’s syringe to measure and dis-

tribute drug solution shares to other individuals (referred to as backloading or frontloading), rinsing a used syringe in water in which others have previously placed used syringes for mixing and rinsing, drawing drug shares from a common cooker where the drug is prepared and shared use of a filter used to draw up the drug solution. The identification of sharing behavior as central to HIV transmission among IDUs brings the social context of intravenous drug use to the forefront. Studies that have examined IDUs’ social networks have been helpful in this regard. Social networks have been defined as ‘those people with whom there are social interactions in which members are, at least potentially, mutually oriented to one another and may influence each others’ behavior’ (Jose et al. 1993). Recently there has been interest in characterizing quantitatively the effects of social networks on behaviors.

© 2006 The Authors. Journal compilation © 2006 Society for the Study of Addiction

Addiction, 101, 1003–1013

1004

Elizabeth C. Costenbader et al.

Quantitative social network studies among IDU populations have found associations between network structural characteristics such as size and density (the number of network members divided by the number of potential ties among members) and reported injection sharing behaviors. For instance, in a study of IDUs recruited from the streets of Baltimore, Maryland, Latkin and colleagues found that the size of the drug subnetwork (number of individuals the study participant reported ‘doing drugs with’ in the 6 months prior to being interviewed) and personal network density were associated significantly with greater frequency of injecting drugs (Latkin et al. 1995a). In addition, IDUs with denser and larger personal networks were more likely to share syringes and those with smaller material aid networks (relationships with individuals who would provide them with economic assistance) and larger positive feedback networks (more and closer relationships with other IDUs) were more likely to inject at shooting galleries (Latkin et al. 1995b). One limitation of the majority of quantitative social network studies to date is that they are cross-sectional and characterize networks as static or non-changing entities. Evidence from the epidemiological and ethnographic research, however, indicates that social relationships among many IDUs are likely to be transient and short term (Curtis et al. 1995; Neaigus et al. 1995; Hoffmann, Su & Pach 1997; Rothenberg et al. 1998; Valente & Vlahov 2001). One study of a cohort of individuals at high risk for HIV found that the names of individuals identified as drug-using partners changed significantly more over time than the names of individuals identified as sexual partners or friends (Rothenberg et al. 1998). Unstable or changing relationships in IDU networks have several implications for both the spread and prevention of HIV (Kretzschmar & Morris 1996; Laumann & Youm 1999). A study comparing HIV-related risk behaviors across 22 communities of IDUs with low, moderate and high seroprevalence rates found that risky injection and sexrelated behaviors were reported at significantly higher rates in the low seroprevalence communities (Deren et al. 2001). The researchers concluded that IDUs were adjusting their level of risk behavior in response to the level of HIV seroprevalence in the community. An alternative explanation may be that the transmission of HIV in these communities is affected largely not only by the frequency with which IDUs are sharing injection equipment but also by the selectivity and stability of the social relationships within which sharing occurs. IDUs in the high seroprevalence communities may be sharing injection equipment on only an infrequent basis but may not be being selective about with whom they share. If they are sharing over time with a succession of individuals whose HIV status is unknown to them, this may be contributing to the more rapid spread of HIV in those communities.

In addition to offering protection through continuation with the same risk partners whose HIV status is known (i.e. making it less likely that an IDU shares with an infected person), stability of networks also may promote protective behavior, possibly through the establishment of behavioral norms. In one study, for example (Hoffmann et al. 1997), a higher proportion of new members entering a personal network in the 3-month interval was associated with a significant increase in the odds of sharing needles in the absence of bleach cleaning or sharing cookers, cottons, rinse water or front/backloading. To date relatively little is known about how IDUs’ social networks change over time, and consequently how changes in these relationships affect risk-taking behaviors and the establishment of behavioral norms. Using data from three waves of a longitudinal study of IDUs recruited in Baltimore City, the goal of the present study was to examine the association between changes in network characteristics and changes in IDUs’ injection HIV risk behaviors over time.

DATA AND METHODS Survey respondents The data used in this analysis come from three waves of the SHIELD (Self-Help In Eliminating Life-threatening Diseases) HIV intervention study (Latkin, Sherman & Knowlton 2003). The aim of the SHIELD study was to examine empirically social processes and their relationship to HIV risk behaviors. Recruitment areas in Baltimore City were identified through ethnographic observations, focus groups and geocodes (i.e. geographic coordinates corresponding to a street address) of the locations of all the drug-related arrests in Baltimore in a 3-year period. Criteria for inclusion in the study were: being 18 years of age or older; having weekly contact with drug users; and a willingness to conduct network education and to bring in members of one’s social network for a baseline interview. The demographics of Baltimore’s AIDS cases and previous research among Baltimore’s illicit drug users indicate that the SHIELD study is representative of the population of illicit drug users in Baltimore. The sample was primarily low-income, unemployed and African American, consistent with previous research among Baltimore IDUs (Mandell et al. 1994; Latkin 1995). Baseline data collection, referred to as the T1 interviews, occurred from June 1997 to February 1999; 1637 individuals were interviewed. At the respondent’s baseline interview, study staff gave each participant information cards to give to as many as five members of her or his social network in order to recruit additional individuals into the study. Study staff asked respondents to bring in

© 2006 The Authors. Journal compilation © 2006 Society for the Study of Addiction

Addiction, 101, 1003–1013

Network and risk behavior dynamics

individuals with whom they had used drugs or had sex within the 6 months prior to the interview. The payment for the baseline interview was $20 with additional payments for bringing in network members who completed a baseline interview. Trained interviewers administered the approximately 90-minute survey to study participants in a private setting. Insufficient funding required reduction of the followup sample. For this reason the investigators prioritized participants based on individuals’ intervention status and whether they had brought in a network member to be interviewed. Seventy-eight per cent of the individuals selected for the first follow-up, the T2 interview, were located and interviewed again. At the second follow-up interview, the T3 interview, 72% of those interviewed at T2 were relocated and re-interviewed. Given the time-consuming nature of locating and enrolling a hidden population such as IDUs, the baseline and follow-up surveys occurred over variable periods of time. The T1 interviews continued for 29 months, the T2 interviews continued for 37 months and the T3 interviews continued for 24 months. The average interval between the T1 and T2 interviews was 10 months [standard deviation (SD) = 6 months]. The average interval between the T1 and T3 interviews was 3.5 years (SD = 9 months) and the average interval between the T2 and the T3 interviews was 2.5 years (SD = 8 months). Because the present analysis focused on injection risk behaviors over time, we restricted the population to 409 individuals who injected any drug in the 6 months prior to their T1 interview or at any point during the course of the study, who came in for at least one follow-up interview and who provided information about their network contacts at both interviews. Injection drug use characteristics of the sample The 409 individuals who comprised the current sample are a group of relatively established injectors. Although there was a broad range of experience with injecting drugs from less than 1 year to as many as 44 years, more than three-quarters of participants had been injecting for 10 years or more at the T1 interview. The majority of respondents injected cocaine or speedball (heroin mixed with cocaine or amphetamine) in addition to heroin (82%) and more than three-quarters injected drugs at least once a day (76%). Nearly twice as many respondents admitted to having shared drug equipment over the past 6 months (80%) than to having shared injection syringes over the past 6 months (40%). Network contacts Study staff delineated the personal networks of survey respondents at all three times using the same method. The study instrument used 17 different name generator

1005

questions to elicit names of people with whom the survey respondent socializes, has sex, does drugs and/or calls upon for material and emotional support (Latkin et al. 2003). Using this method, the maximum number of network members that a respondent named at baseline was 24, the minimum was 2 and the average was 9. For each network contact mentioned at T1 and T2, participants provided first names and the first letter of last names. The T3 interview required that respondents also specify each network contact’s full last name and any pseudonyms or nicknames. Once the list of names was completed, participants provided information about each network member such as age, gender, employment status and relationship type. Participants also described each network member’s drug-using practices. The final question in the network inventory was a request that survey respondents indicate which of their network members were friends with one another (density). At each follow-up survey a coder examined network names to determine if they corresponded to a person that had been mentioned previously, referred to as a ‘match’. Matches were made using name, age, gender and type of relationship (e.g. friend, spouse, etc.). One individual coded all the matches but a second reviewer re-coded a random 5% sample of the matches independently. The exact agreement between the two coders was 93% and the kappa coefficient of agreement between their matches was 85%. The coder gave each match a score from 1 to 3 to indicate the level of certainty with which the match was made. For this analysis, we used only matches designated with certainty codes of 1, ‘almost definitely the same person’, and 2, ‘highly likely to be the same person. Measurement of individual characteristics and network properties Because network characteristics are confounded with individual characteristics we controlled for indicators of socio-economic and mental health status, participation in the intervention, history of incarceration, drug use and drug treatment (Kang & De Leon 1993; Latkin et al. 1994; Mandell et al. 1999; Reynolds et al. 2000; Celentano et al. 2001; Stein et al. 2003). We also looked at indicators of survey respondents’ gender and age, as these characteristics have been found to have significant associations with changes in network membership over time (Fischer & Oliker 1983; Hoffmann et al. 1997; Ruan et al. 1997). Study participants reported the highest grade of school that they had completed. We created a dichotomous indicator of educational level (did not finish high school versus completed a high school degree or beyond). We considered study participants to have been homeless in the past 6 months if they reported being homeless or they described their current living situation as a ‘shelter’ or ‘on the streets, homeless’.

© 2006 The Authors. Journal compilation © 2006 Society for the Study of Addiction

Addiction, 101, 1003–1013

1006

Elizabeth C. Costenbader et al.

Study participants reported the age at which they first injected drugs, their attendance at any drug treatment program in the last 30 days and whether they had spent any time in prison during the past 6 months. Interviewers assessed participants’ depressive symptoms at the time of interview using the Center for Epidemiologic Studies Depression Scale (CES-D) (Radloff 1977). We regarded respondents with a score of 20 or higher on the CES-D to be depressed. We calculated two network measures that have been shown to have an effect on risk behaviors in other network studies of IDUs (Latkin et al. 1995a,b): network density and the percentage of drug users in the personal network. In addition to these cross-sectional network measures, which are measured as of the baseline interview, we were also interested in assessing the association between changes in network composition and changes in risk behavior. We adapted the measures of network turnover developed by Hoffman et al. (1997). They developed two separate indicators for movement into and out of the network over time, turnover-in and turnover-out, reasoning that ‘members moving into a network might have different implications than members moving out of a network’ (p. 45). We further divided the counts of turnover-in and turnover-out into turnover-in and turnover-out of the drug and non-drug networks, respectively. We were interested in capturing the proportion of movement relative to the total size of the network summed between baseline and follow-up. Therefore, the resulting turnover indices ranged between 0 (no turnover) and 1 (complete turnover). Injection risk scale The measure of injection behavioral risk used in this analysis was a four-level scale with each level denoting increasing probability of exposure to HIV as a result of risky injection practices. The scale was adapted from behavioral risk scales used in other cohort studies of HIV seroprevalence among injection drug users (Celentano et al. 2001; Vanichseni et al. 2001). Study participants who reported no instances of injecting drugs in the 6 months prior to the interview were assigned the lowest risk level categorization, level 1 or no risk. Study participants who reported injecting drugs in the past 6 months but reported no sharing needles or indirect sharing behaviors during this time were assigned a risk level of 2. Study participants who reported indirect sharing behaviors in the past 6 months but no instances of sharing syringes were assigned a risk level of 3. Indirect sharing behaviors that we assessed included sharing cookers, cotton and rinse water and using a dirty or used syringe to divide or distribute drugs. The highest risk level designation, level 4, was given to study partici-

pants who reported any instances of using needles that they were not sure were clean or sharing needles, defined as using a needle immediately after someone else. We classified study participants into the following four mutually exclusive outcome groups based on their pattern of injection risk behavior across subsequent interviews. 1 Stable low risk. Study participants with baseline values of 1 or 2 on the risk scale who remained at levels 1 or 2 at their follow-up interview. 2 Decreasing risk. Study participants whose risk level categorization decreased between their baseline and follow-up interviews. 3 Increasing risk. Study participants whose risk level categorization increased between their baseline and follow-up interviews. 4 Consistently high-risk. Survey respondents with baseline values of 3 or 4 on the risk scale who remained at levels 3 or 4 at their follow-up interview. A cross-tabulation of injection risk level categories at baseline and follow-up for each observation revealed that in this study sample the largest numbers of observations were individuals who were in the same risk-level category at both baseline and follow-up. Of those who changed injection behavior classifications over time there was a predominance of risk reduction over risk increase in this population. Analyses Because our HIV injection risk change variable had four possible levels, we used a multinomial logistic regression model to assess the associations between our predictors and change in risk. In our multivariate model, we included all covariates for which the bivariate χ2 test of association with risk change was significantly different from zero at the P < 0.20 level. Our multivariate models included only individual-level covariates and baseline network characteristics. The second included individual and baseline network characteristics as well as dynamic network measures. We compared the different models by examining the difference in their deviance. This difference is distributed as χ2 with degrees of freedom equal to the difference in the number of predictors in each model (Powers & Xie 2000). Using the regression parameters generated by this final model, we estimated the predicted probability of being in each of the four injection HIV risk behavior change groups. Although there were only 409 individuals in our analysis subsample, 53% of the 409 individuals in the sample were re-interviewed at both T2 and T3 and were therefore allowed to contribute more than one observation in the analysis. For these individuals we analyzed their injection risk behavior and concurrent network

© 2006 The Authors. Journal compilation © 2006 Society for the Study of Addiction

Addiction, 101, 1003–1013

Network and risk behavior dynamics

changes between T1 and T2 and between T2 and T3, treating these as two separate observations. We used the robust cluster estimator to account for the correlation within individuals with more than one observation (White 1982; Royall 1986). All analyses were performed using Stata statistical software (StatCorp 2003).

1007

RESULTS Bivariate relationships Table 1 displays the associations between network and individual-level characteristics and injection HIV risk behavior groups. Based on the tabulations in Table 1, the multivariate model included: percentage of drug

Table 1 Percentage of observations within each change group by network and individual characteristics ( n = 624). Change group Stable low n

Decreasing risk %

Increasing risk %

Percentage of drug users in the network Less than 30% 30–42% 43–56% More than 56%

165 149 154 156

27 15 19 16

30 30 27 32

Network density Less than 25% 25–32% 33–43% More than 43%

153 145 177 149

17 18 19 24

Percentage turnover-in to the drug network Zero 1–17% 18–39% More than 39%

230 79 158 157

Percentage turnover-out of the drug network Less than 12% 12–33% 34–60% More than 60%

Consistently high %

%

Univariate P-valuea

7 6 6 6

36 50 47 46

0.167

31 26 30 31

7 6 8 5

45 50 44 40

0.715

30 11 12 16

38 34 20 24

3 8 4 12

29 47 63 48

0.000

163 162 126 173

25 10 17 24

65 23 28 40

6 7 10 3

43 59 45 32

0.000

Percentage turnover-in to the non-drug network Less than 14% 14–26% 27–41% More than 41%

147 141 169 167

16 22 16 24

25 25 32 36

8 6 4 8

52 47 49 32

0.014

Percentage turnover-out of the non-drug network Less than 11% 11–22% 23–38% More than 38%

145 161 163 155

21 17 18 21

28 33 26 32

7 7 7 5

45 43 48 42

0.923

Unemployed at time of interview Yes No

513 111

19 23

31 24

6 9

45 44

0.334

Depressed at time of interview Yes No

293 331

15 23

29 31

5 8

51 39

0.006

Incarcerated past 6 months Yes No

103 521

20 19

32 29

6 7

42 45

0.900

Homeless past 6 months Yes No

87 537

21 18

26 33

4 9

48 41

0.884

© 2006 The Authors. Journal compilation © 2006 Society for the Study of Addiction

Addiction, 101, 1003–1013

1008

Elizabeth C. Costenbader et al.

Table 1 Cont. Change group

Education completed Less than high school High school/GED or more

Stable low n

Decreasing risk %

Increasing risk %

325 299

21 18

26 33

Consistently high %

%

Univariate P-valuea

4 9

48 41

0.019

Intervention participation Yes No Sex Male Female

381 243

18 21

28 32

6 7

48 40

0.319

402 222

19 20

29 30

6 8

46 42

0.734

Age (years) 18–28 29–35 36–44 More than 44

154 171 168 131

29 16 18 13

29 30 29 31

6 6 7 8

36 48 46 48

0.086

Tried but unable to get into a drug treatment program during past year Yes No

255 369

16 21

30 30

7 6

47 43

0.455

In a drug treatment program past 30 days Yes No

235 389

17 21

31 29

6 6

46 44

0.815

Years of injection drug use prior to enrollment Less than 10 10–18 19–27 More than 27

140 169 165 150

23 24 13 17

32 27 29 31

5 7 9 4

40 41 48 48

0.170

Univariate P-value obtained from χ2 test.

a

users in the network, turnover-in to the drug network, turnover-out of the drug network, turnover-in to the non-drug network, respondent age and educational level, depression and number of years injecting drugs. Multinomial logistic regression coefficients Table 2 presents the multinomial logistic regression coefficients for selected contrasts from the final model. Our baseline model included only individual and baseline network characteristics. We used the deviance of this baseline model to assess improvements in fit due to the inclusion of new variables. Our final model significantly improved the fit relative to the baseline (improvement χ2(9) = 79, P = 0.001). Table 2 presents four of six possible contrasts, those that we believed were the most informative from the perspective of HIV prevention as they demonstrated movement out of a high- or increasing-risk group into a low- or

decreasing-risk group. All three of our dynamic network variables were significant predictors of injection risk behavior outcome in more than one model. Calculated predicted probabilities Because multinomial logistic regression estimates are difficult to interpret directly, we estimated a corresponding series of predicted probabilities which are presented in Table 3. Specifically, we present here the probabilities associated with changes occurring only in the networklevel variables and for illustrative purposes the most extreme examples of change (i.e. changing from all drugusers in the network to no drug users in the network), as this provides us with the upper and lower bounds of change. Our base case was an individual with modal values for each of the categorical variables and mean values for each of the continuous covariates. In this sample, the base case was an individual who was aged 41 years, had

© 2006 The Authors. Journal compilation © 2006 Society for the Study of Addiction

Addiction, 101, 1003–1013

Network and risk behavior dynamics

1009

Table 2 Multinomial logistic coefficients for the multivariate model of injection HIV risk behavior change group over time.

Constant Network characteristics Turnover-in drug network Turnover-out drug network Turnover-in non-drug network Percentage drug users Individual characteristics Age Completed less than high school education Depressed at time of interview Number of years injecting

High to low

Increasing to low

High to decreasing

−1.46

−4.15**

−0.05

−2.75†

1.61** −1.21** −2.56*** 3.61***

3.01** −0.37 −1.53 3.29**

1.45** −1.53** −2.35** 2.43***

2.85** −0.69* −1.31 2.11*

0.02 0.01 0.80** 0.02

0.04 −0.77† 0.14 −0.01

–0.01 0.39† 0.45* 0.02

0.01 −0.39 −0.21 −0.01

Total model characteristics Log-likelihood Model χ2 df Number of observations †

Increasing to decreasing

−699.82 90.09 24 624

P < 0.1, *P < 0.05, **P < 0.01, ***P < 0.001.

Table 3 Predicted probabilities of classification into each HIV risk behavior change group by network characteristics. a Change group

1 Base caseb Changes in dynamic network characteristics 2 Zero turnover-in drug network 3 100% turnover-in drug network 4 Zero turnover-out drug network 5 100% turnover-out drug network 6 Zero turnover-in non-drug network 7 100% turnover-in non-drug network 8 Zero turnover-in drug and non-drug network and zero turnover-out drug network 9 100% turnover-in drug and non-drug network and 100% turnover-out drug network Changes in baseline network characteristics 10 Zero drug users in the network 11 All drug users in the network

Stable low

Decreasing risk

Increasing risk

Consistently high

0.29

0.25

0.05

0.42

0.33 0.13 0.20 0.42 0.20 0.43 0.17

0.30 0.10 0.20 0.29 0.17 0.43 0.17

0.03 0.19 0.05 0.05 0.05 0.03 0.03

0.34 0.58 0.55 0.24 0.58 0.12 0.63

0.41

0.30

0.16

0.13

0.33 0.13

0.48 0.06

0.02 0.07

0.17 0.74

a When comparing differences in these numbers some discrepancies might arise as a result of the fact that these numbers are rounded to two decimal places. bEvaluated by using the modal values for all categorical variables and mean values for all continuous covariates in the model.

completed less than a high school education, was not depressed at the time of interview had been injecting drugs for 19 years and had mean values for each of the other continuous network covariates. The probabilities of being in each of the four risk change categories for the base case are in row 1 of Table 3. The base case in this sample population was most likely to be in the consistently high-risk group (0.42) and least likely to be in the increasing-risk group (0.05).

Comparison of row 1 to rows 2–7 indicates the impact of network turnover or changes in network membership over time on an individual’s likelihood of being in each of the risk injection categories. Rows 2 and 3 show that the probability of decreasing one’s injection risk behavior or remaining at a stable low level of injection risk increased when there was less turnover-in to the drug network. By contrast, the probability of increasing one’s injection risk behavior or remaining at a consistently high level of

© 2006 The Authors. Journal compilation © 2006 Society for the Study of Addiction

Addiction, 101, 1003–1013

1010

Elizabeth C. Costenbader et al.

injection risk behavior increased when there was more turnover-in to the drug network. In row 2 we see that having no new drug contacts at follow-up (i.e. zero turnover-in to the drug network) increased the likelihood of being in the stable low-risk group and the decreasing-risk group, approximately 14% (0.33/0.29) and 20% (0.30/0.25), respectively, and reduced the likelihood of being in the increasing-risk and consistently high-risk groups, approximately 40% [1 − (0.03/0.05)] and 19% [1 − (0.34/0.42)], respectively. In row 3 we see the opposite. If all an individual’s drug network contacts were new contacts at follow-up, the likelihood of being in the increasing injection-risk behavior group over time increased more than threefold (0.19/0.05) and the likelihood of being in the consistently high-risk group was more than 1.3 times higher (0.58/0.42), while the likelihood of being in the stable low- or decreasing-risk groups was reduced by 55% [1 − (0.13/0.29)] and 60% [1 − (0.10/0.25)], respectively. Comparison of rows 4 and 5 to rows 2 and 3 shows that the movement of drug-using contacts out of the network over time (i.e. drug network turnover-out) had an opposite effect on the probabilities of being in each of the injection risk behavior groups than the movement of new drug-using contacts in to the network over time (i.e. drug network turnover-in). For instance, in row 5 movement of all drug associates out of the network (i.e. 100% turnover-out) increased the probability of being in the decreasing-risk group by approximately 16% (0.29/ 0.25) and in row 2 zero turnover-in to the drug network increased the probability of being in the decreasing-risk group by approximately 20% (0.30/0.25). In rows 6 and 7 the contrasting probabilities demonstrate the impact on injection risk behavior of gaining new network members who are not drug users. The complete absence of non-drug-using contacts entering the network over time decreased the probability of being in the stable low-risk group and of being in the decreasingrisk group by about 31% [1 − (0.20/0.29)] and 32% [1 − (0.17/0.25)], respectively, and increased the probability of being in the consistently high-risk group by as much as 38% (0.58/0.42), although it did not have an impact on the probability of being in the decreasing-risk group. In contrast, reporting all new non-drug-using contacts at follow-up increased the likelihood of being in the stable low-risk group and the decreasing-risk group by about 48% (0.43/0.29) and 72% (0.43/0.25), respectively, and decreased the probability of being in the increasing-risk group and the consistently high-risk group by about 40% [1 − (0.03/0.05)] and 71% [1 − (0.12/0.42)], respectively. Comparison of rows 4, 5, 6 and 7 reveal that turnoverin to the non-drug network (i.e. the addition of new network contacts who were not identified as drug-using

contacts) had an effect on the predicted probabilities of being in each of the four risk-level groups similar to turnover-out of the drug network, although differences in magnitude were measured. Because average overall network size in this study population changed relatively little over time (from nine to eight individuals), as turnover-out of the network of drug-using contacts increased, turnover-in to the network of non-drug-using contacts also increased. In rows 8 and 9 we provide contrasting probabilities to demonstrate the impact of staying with the all the same network contacts over time versus changing all one’s network contacts over time. The complete absence of new contacts entering the network over time increased the probability of being in the consistently high-risk group by 50% (0.63/0.42) and decreased the probability of being in all other risk groups by between 30 and 40% [(1 − (0.17/0.29)] [− (0.17/0.25)] and [1 − (0.03/0.05)]. In contrast, complete turnover of an individual’s network contacts decreased the probability of being in the consistently high-risk group by almost 70% [1 − (0.13/0.42)] and increased the probability of being in all the other risk groups. In particular, the likelihood of being in the increasing-risk group was more than three times higher (0.16/0.05).

DISCUSSION The findings from this analysis demonstrate that the likelihood of behavior change due to network factors is not captured fully in a snapshot of the network at one timepoint. We have shown that changes occurring over time in personal network membership have an important impact on whether individuals increase, decrease or maintain their level of risky injection behaviors, net of individual characteristics. We have also shown that network turnover cannot be categorized as either a uniformly negative or positive influence on injectors’ behaviors. In our study sample, network turnover-in made it more likely that an injector engaged in the highest levels of injection risk behavior (i.e. sharing injection equipment and needles), if the new network contacts were also drug users. Yet network turnover-in decreased an injector’s likelihood of engaging in high-level injection risk behaviors if the new network contacts were not drug users. Network turnover-out also contributed to a decrease in the likelihood of engaging in high-level injection risk behaviors if the contacts who left the network were drug users. The results of this study are consistent with previous studies that show that changes in networks are associated with changes in the likelihood of risky injection behavior (Curtis et al. 1995; Latkin et al. 1995b; Hoffmann et al. 1997). This study, however, expands upon the

© 2006 The Authors. Journal compilation © 2006 Society for the Study of Addiction

Addiction, 101, 1003–1013

Network and risk behavior dynamics

findings of these previous studies in several significant ways. First, in the only other study that assessed network turnover in a manner analogous to our study (Hoffmann et al. 1997), the period of follow-up was a 3-month period. In contrast, the average interval between baseline and follow-up interviews in our analysis is 1.5 years—a substantially larger window of time over which to assess behavior change. Another strength of our analysis is the large population size that was followed. We were able to calculate network turnover for 624 observations in comparison to only 55 observations in the analysis completed by Hoffman et al. (1997). Finally, while Hoffman and colleagues found that network turnover-in was associated with a significant increase in the odds of injection risk behavior we were able to show, by distinguishing between contacts in the drug and the non-drug network, that network turnover-in can have either a positive or a negative influence on individual risk behavior depending upon whether or not the new network contacts are drug users. Several important caveats accompany our findings. First, there was undoubtedly some amount of misclassification of relationships that would affect our measures of network turnover. For instance, it is entirely possible that respondents forgot to include a network member with whom they were still in contact at follow-up or conversely (i.e. network members included at follow-up had a relationship with the respondent at baseline). In both cases we may be overestimating network turnover. On the other hand, the study design is likely to have exerted downward pressure on our estimate of network turnover in two respects. A willingness to bring in two members of one’s social network for interview was one criterion for study participation, thereby selecting into the study individuals who are more likely to have potentially more stable networks of relationships (i.e. networks experiencing less turnover of membership). In addition, in order for individuals to be included in our analysis, they had to have returned for a follow-up interview. It is likely that individuals who were located and recruited easily for follow-up interviews had more stable networks of relationships (i.e. networks with less turnover) than those who were not interviewed. Our injection risk measures are also likely to be underestimates of the true prevalence of these behaviors. Respondents were not asked about renting needles or about using discarded previously used syringes. Asking individuals about their use of ‘clean’ needles has been shown to be an unreliable indicator of whether HIV is actually present in a syringe (Abdala et al. 2001). In addition, several studies have shown that survey respondents generally tend to under-report behaviors that are considered socially undesirable (Parry, Balter & Cisin 1970; Hindelang, Hirschi & Weis 1981; Brewer et al. 2000). Although the fact that SHIELD was an intervention study

1011

makes it unlikely that respondents will under-report drug use per se, they may under-report sharing needles and sharing behaviors (Edwards 1957; Latkin, Vlahov & Anthony 1993). Although there was a broad range of years of experience with injecting drugs in this sample, it is likely that the longer-term injectors who are in this sample are simply a proportion of those who began injecting when they did. Presumably the survivors may be the less risky injectors, as studies have shown that individuals who exhibit greater drug dependence and more risky injection practices are more likely to experience an overdose (Darke, Ross & Hall 1996; Ochoa et al. 2001). The lower reported levels of injection risk behaviors at followup in our sample may represent true behavior change as a result of exposure to the SHIELD HIV prevention intervention. We are not able to make statements about causality with this analysis. Network contacts and risk behaviors were changing over the same period of time, therefore it is also possible that behavior change could be determining network change or that some third factor could be causing both network and behavior changes (e.g. police harassment). A final limitation to this study design is that the only network characteristic we were able to examine was changes in membership. There are other ways in which a relationship and a network of relationships could change over time (i.e. changes in relationship strength and structural network changes). These other facets of network change may have implications for social influence and disease spread and are of interest for future research. Several studies indicate that IDUs are aware of the risk of HIV through the sharing of injection equipment and that many IDUs have altered their behaviors in response to this threat (Selwyn et al. 1987; Bloor 1995; Des Jarlais et al. 1995). In particular, it appears that most contemporary syringe-sharing occurs within limited social circles (e.g. close friendships, sexual relationships and family ties) (McKeganey & Barnard 1992; Neaigus et al. 1994; Bloor 1995; Valente & Vlahov 2001). Our findings support the practice of limiting syringe-sharing behaviors to a small social circle of contacts not only because the number of individuals with whom these IDUs share needles and other injection equipment will be reduced, but also because less turnover-in to the drug-using network over time is associated with an increased likelihood of either staying at a low level of injection risk behavior or decreasing one’s injection risk behavior. Although we were unable to assess whether changes in injecting practices were occurring at the network-level, the association between turnover-in and injection risk may be an indication that fewer members entering a network over time allows network members to establish certain norms of injection practice.

© 2006 The Authors. Journal compilation © 2006 Society for the Study of Addiction

Addiction, 101, 1003–1013

1012

Elizabeth C. Costenbader et al.

Our findings also indicate that encouraging injectors to stay with the same drug-using partners over time is not likely to be sufficient to eliminate high-risk injection behaviors. In our study sample, turnover-out of the drug network and turnover-in to the non-drug-using network were also critical elements to behavioral risk reduction. These findings suggest that interventions which work with injectors to supplant their drug-using friends with non-drug-using contacts may be the most successful in achieving sustained behavioral risk reductions. Nonetheless, intervention approaches that aim to increase IDUs’ social contacts with non-drug-using individuals will need to be designed and monitored carefully for several reasons. Realistically, the ability of IDUs to replace drug-using contacts with non-drug-using contacts will vary. Among those who are least likely to be able to replace their drug contacts with non-drug-using contacts are those who are most dependent on the access to drugs and other resources provided by those contacts (i.e. highly addicted individuals and individuals with the least economic resources). Similarly, individuals residing in communities where drug use is endemic will be at a disadvantage to establish non-drug-using contacts. There is also the potential that an IDU could influence a noninjector to transition to injecting. Given that this study is the first study to assess specifically the impact of changes in risk and non-risk (i.e. drug-using versus non-drugusing) network membership on intra-individual injection HIV risk behavior change, additional research is needed to determine whether these findings are replicable in other study settings, sample populations and for other types of risk behaviors, as well as to characterize the correlates of network turnover. Acknowledgements This research was supported by grant number DA016005 from the National Institutes of Drug Abuse (NIDA). We thank Steve Muth for his assistance with the matching process, Susan Sherman for her comments on a previous draft and our anonymous reviewers for their thoughtful and helpful reviews. References Abdala, N., Gleghorn, A., Carney, J. & Heimer, R. (2001) Can HIV-1 contaminated syringes be disinfected? Implications for transmission among injection drug users. Journal of Acquired Immune Deficiency Syndrome, 28, 487–494. AIDS Administration, Maryland Department of Health and Mental Hygiene (2004) The Maryland HIV/AIDS Annual Report, pp, 1–96. MD: AIDS Administration, Maryland Department of Health and Mental Hygiene. Bloor, M. (1995) The Sociology of HIV Transmission. Thousand Oaks, CA: Sage Publications. Brewer, D. D., Potterat, J. J., Garrett, S. B., Muth, S. Q., Roberts, J. M., Kasprzyk, D. et al. (2000) Prostitution and the sex dis-

crepancy in reported number of sexual partners. Proceedings of the National Academy of Sciences USA, 97, 12385–12388. Celentano, D. D., Munoz, A., Cohn, S. & Vlahov, D. (2001) Dynamics of behavioral risk factors for HIV/AIDS: a 6-year prospective study of injection drug users. Drug and Alcohol Dependence, 61, 315–322. Curtis, R., Friedman, S. R., Neaigus, A., Jose, B., Goldstein, M. & Ildefonso, G. (1995) Street-level drug markets: network structure and HIV risk. Social Networks, 17, 229–249. Darke, S., Ross, J. & Hall, W. (1996) Overdose among heroin users in Sydney, Australia. I. Prevalence and correlates of non-fatal overdose. Addiction, 91, 405–411. Deren, S., Beardsley, M., Coyle, S., Singer, M. & Kang, S. (2001) HIV risk behaviors among injection drug users in low, medium, and high seroprevalence communities. AIDS and Behavior, 5, 45–50. Des Jarlais, D. C., Friedman, S. R., Friedmann, P., Wenston, J., Sotheran, J. L., Choopanya, K. et al. (1995) HIV/AIDS-related behavior change among injecting drug users in different national settings. AIDS, 9, 611–617. Edwards, A. L. (1957) The Social Desirability Variable in Personality Assessment and Research. New York: Dryden Press. Fischer, C. S. & Oliker, S. J. (1983) A research note on friendship, gender, and the life-cycle. Social Forces, 62, 124–133. Grund, J. P., Friedman, S. R., Stern, L. S., Jose, B., Neaigus, A., Curtis, R. et al. (1996) Syringe-mediated drug sharing among injecting drug users: patterns, social context and implications for transmission of blood-borne pathogens. Social Science and Medicine, 42, 691–703. Hindelang, M. J., Hirschi, T. & Weis, J. G. (1981) Measuring Delinquency. Beverly Hills: Sage Publications. Hoffmann, J. P., Su, S. S. & Pach, A. (1997) Changes in network characteristics and HIV risk behavior among injection drug users. Drug and Alcohol Dependence, 46, 41–51. Jose, B., Friedman, S. R., Neaigus, A., Curtis, R., Grund, J. P., Goldstein, M. F. et al. (1993) Syringe-mediated drug-sharing (backloading): a new risk factor for HIV among injecting drug users. AIDS, 7, 1653–1660. Kang, S. Y. & De Leon, G. (1993) Correlates of drug injection behaviors among methadone outpatients. American Journal of Drug and Alcohol Abuse, 19, 107–118. Koester, S. (1996) The process of drug injection: applying ethnography to the study of HIV risk among IDUs. In: Rhodes, T. & Hartnoll, R., eds. AIDS, Drugs and Prevention: Perspectives on Individual and Community Action, pp. 133–148. London: Routledge. Kretzschmar, M. & Morris, M. (1996) Measures of concurrency in networks and the spread of infectious disease. Mathematical Biosciences, 133, 165–195. Latkin, C. A. (1995) A personal network approach to AIDS prevention: an experimental peer group intervention for streetinjecting drug users: the SAFE study. NIDA Research Monograph, 151, 181–195. Latkin, C. A., Mandell, W., Oziemkowska, M., Celentano, D., Vlahov, D., Ensminger, M. et al. (1995a) Using social network analysis to study patterns of drug use among urban drug users at high risk for HIV/AIDS. Drug and Alcohol Dependence, 38, 1–9. Latkin, C., Mandell, W., Vlahov, D., Knowlton, A., Oziemkowska, M. & Celentano, D. (1995b) Personal network characteristics as antecedents to needle-sharing and shooting gallery attendance. Social Networks, 17, 219–228. Latkin, C. A., Mandell, W., Vlahov, D., Oziemkowska, M., Knowlton, A. & Celentano, D. (1994) My place, your place, and no

© 2006 The Authors. Journal compilation © 2006 Society for the Study of Addiction

Addiction, 101, 1003–1013

Network and risk behavior dynamics

place: behavior settings as a risk factor for HIV-related injection practices of drug users in Baltimore, Maryland. American Journal of Community Psychology, 22, 415–430. Latkin, C. A., Sherman, S. & Knowlton, A. (2003) HIV prevention among drug users: outcome of a network-oriented peer outreach intervention. Health Psychology, 22, 332–339. Latkin, C. A., Vlahov, D. & Anthony, J. C. (1993) Socially desirable responding and self-reported HIV infection risk behaviors among intravenous drug users. Addiction, 88, 517–526. Laumann, E. O. & Youm, Y. (1999) Racial/ethnic group differences in the prevalence of sexually transmitted diseases in the United States: a network explanation. Sexually Transmitted Diseases, 26, 250–261. Mandell, W., Kim, J., Latkin, C. & Suh, T. (1999) Depressive symptoms, drug network, and their synergistic effect on needle-sharing behavior among street injection drug users. American Journal of Drug and Alcohol Abuse, 25, 117–127. Mandell, W., Vlahov, D., Latkin, C., Oziemkowska, M. & Cohn, S. (1994) Correlates of needle sharing among injection drug users. American Journal of Public Health, 84, 920–923. McKeganey, N. P. & Barnard, M. (1992) AIDS, Drugs, and Sexual Risk: Lives in the Balance. Buckingham: Open University Press. Neaigus, A., Friedman, S. R., Curtis, R., Des Jarlais, D. C., Furst, R. T., Jose, B. et al. (1994) The relevance of drug injectors’ social and risk networks for understanding and preventing HIV infection. Social Science and Medicine, 38, 67–78. Neaigus, A., Friedman, S. R., Goldstein, M., Ildefonso, G., Curtis, R. & Jose, B. (1995) Using dyadic data for a network analysis of HIV infection and risk behaviors among injecting drug users. NIDA Research Monograph, 151, 20–37. Needle, R. H., Coyle, S., Cesari, H., Trotter, R., Clatts, M., Koester, S. et al. (1998) HIV risk behaviors associated with the injection process: multiperson use of drug injection equipment and paraphernalia in injection drug user networks. Substance Use and Misuse, 33, 2403–2423. Ochoa, K. C., Hahn, J. A., Seal, K. H. & Moss, A. R. (2001) Overdosing among young injection drug users in San Francisco. Addictive Behaviors, 26, 453–460. Parry, H. J., Balter, M. B. & Cisin, I. H. (1970) Primary levels of

1013

underreporting psychotropic drug use. Public Opinion Quarterly, 34, 584–&. Powers, D. A. & Xie, Y. (2000) Statistical Methods for Categorical Data Analysis. San Diego: Academic Press. Radloff, L. (1977) The CES-D scale: a self-report depression scale for research in the general population. Applied Psychological Measurement, 1, 385–401. Reynolds, G. L., Fisher, D. G., Estrada, A. L. & Trotter, R. (2000) Unemployment, drug use, and HIV risk among American Indian and Alaska Native drug users. American Indian and Alaska Native Mental Health Research, 9, 17–32. Rothenberg, R. B., Potterat, J. J., Woodhouse, D. E., Muth, S. Q., Darrow, W. W. & Klovdahl, A. S. (1998) Social network dynamics and HIV transmission. AIDS, 12, 1529–1536. Royall, R. M. (1986) Model robust confidence-intervals using maximum-likelihood estimators. International Statistical Review, 54, 221–226. Ruan, D. C., Freeman, L. C., Dai, X. Y., Pan, Y. K. & Zhang, W. H. (1997) On the changing structure of social networks in urban China. Social Networks, 19, 75–89. Selwyn, P. A., Feiner, C., Cox, C. P., Lipshutz, C. & Cohen, R. L. (1987) Knowledge about AIDS and high-risk behavior among intravenous drug users in New York City. AIDS, 1, 247–254. StatCorp (2003) Stata Statistical Software. Special edition, release 8.2. College Station, TX: Stata Corporation. Stein, M. D., Solomon, D. A., Herman, D. S., Anderson, B. J. & Miller, I. (2003) Depression severity and drug injection HIV risk behaviors. American Journal of Psychiatry, 160, 1659– 1662. Valente, T. W. & Vlahov, D. (2001) Selective risk taking among needle exchange participants: implications for supplemental interventions. American Journal of Public Health, 91, 406– 411. Vanichseni, S., Kitayaporn, D., Mastro, T. D., Mock, P. A., Raktham, S., Des Jarlais, D. C. et al. (2001) Continued high HIV-1 incidence in a vaccine trial preparatory cohort of injection drug users in Bangkok, Thailand. AIDS, 15, 397–405. White, H. (1982) Maximum-likelihood estimation of misspecified models. Econometrica, 50, 1–25.

© 2006 The Authors. Journal compilation © 2006 Society for the Study of Addiction

Addiction, 101, 1003–1013

The dynamics of injection drug users' personal ... - Wiley Online Library

Substance Abuse Treatment Evaluations and Interventions Program, Research Triangle Institute, International, NC, USA,1 Department of Population and Family. Health Sciences, Johns Hopkins Bloomberg School of Public Health, Baltimore, MD, USA2 and Department of Health Behavior and Society, Johns Hopkins.

114KB Sizes 0 Downloads 187 Views

Recommend Documents

Ideas, Inheritances, and the Dynamics of ... - Wiley Online Library
budgeting over the past half-century, exploring such developments as the massive growth of entitlements, congressional budget reform, and the pro-.

Dynamics of fossil fuel CO2 neutralization by ... - Wiley Online Library
cm deep) with a bulk mixing coefficient of 150 cm 2 kyr 'l. The solid organic ...... Bundesstrasse 7, D-2000 Hamburg 54, Germany (email maier- [email protected]).

Dynamics of fossil fuel CO2 neutralization by ... - Wiley Online Library
Exxon Research and Engineering Company, Annandale, New Jersey. Ernst Maier- ..... A number of geoengineering techniques have been proposed to control .... been estimated to be on the order of 5000 Gt C [Sundquist,. 2000. 1800. 1600.

Seasonal dynamics and age of stemwood ... - Wiley Online Library
determined with a glucose hexokinase kit (Pointe Scientific,. Canton, MI, USA) at 340 nm with a microplate reader (ELx800. UV; Bio-Tek Instruments, Winooski, VT, USA). Starch concen- trations were calculated from standard curves and are expressed her

Personal Efficacy, the Information Environment ... - Wiley Online Library
2 The website for. An. Inconvenient Truth, http://www. climatecrisis.net, contains specific information and printable tips that are meant to foster individual efficacy, including infor- mation about how each individual can reduce his or her impact on

Finding Diachronic Likeâ•'Minded Users - Wiley Online Library
FINDING DIACHRONIC LIKE-MINDED USERS. 141. FIGURE 8. Sample topics from the graph-based topic detection method (GbT) that are relevant to the communities in Figure 9. [Color figure can be viewed at wileyonlinelibrary.com]. FIGURE 9. Sample temporal t

The Metaphysics of Emergence - Wiley Online Library
University College London and Budapest University of. Technology and Economics. I. Mental Causation: The Current State of Play. The following framework of ...

Dosing information in a standard drug reference - Wiley Online Library
dosing information were: neonates (13.1%), infants (23.3%), and children (32.2%). Therapeutic classes of drugs differed significantly in terms of dosing information adequacy for all pediatric age groups (P < 0.05). Most PI reviewed under all legal ca

Drug survival is not significantly different ... - Wiley Online Library
E-mail: [email protected]. Accepted for publication. 24 March 2014. Funding sources. None. Conflicts of interest. S.P.M. has carried out clinical trials for ...

international trade and industrial dynamics - Wiley Online Library
In this article, industrial evolution is driven by endogenous technology choices ... sity of California-San Diego, Deakin University, Georgetown University, Latrobe ...

Import dynamics and demands for protection - Wiley Online Library
XXX 2016. Printed in Canada / XXX 2016. Imprimé au Canada. 0008–4085 / 16 / 1–28 / © Canadian Economics Association. Import dynamics and demands for.

Volatility and commodity price dynamics - Wiley Online Library
value of storage, and by affecting a component of the total marginal cost of production .... futures market data to directly measure the marginal value of storage.3.

ELTGOL - Wiley Online Library
ABSTRACT. Background and objective: Exacerbations of COPD are often characterized by increased mucus production that is difficult to treat and worsens patients' outcome. This study evaluated the efficacy of a chest physio- therapy technique (expirati

Competing paradigms of Amazonian ... - Wiley Online Library
September 2014, immediately after the accepted version of this manuscript was sent to the authors on 18 September. 2014. doi:10.1111/jbi.12448. Competing ..... species are there on earth and in the ocean? PLoS Biology, 9, e1001127. Moritz, C., Patton

Principles of periodontology - Wiley Online Library
genetic make-up, and modulated by the presence or ab- sence of ... can sense and destroy intruders. ..... observation that apparently healthy Japanese sub-.

poly(styrene - Wiley Online Library
Dec 27, 2007 - (4VP) but immiscible with PS4VP-30 (where the number following the hyphen refers to the percentage 4VP in the polymer) and PSMA-20 (where the number following the hyphen refers to the percentage methacrylic acid in the polymer) over th

Recurvirostra avosetta - Wiley Online Library
broodrearing capacity. Proceedings of the Royal Society B: Biological. Sciences, 263, 1719–1724. Hills, S. (1983) Incubation capacity as a limiting factor of shorebird clutch size. MS thesis, University of Washington, Seattle, Washington. Hötker,

Kitaev Transformation - Wiley Online Library
Jul 1, 2015 - Quantum chemistry is an important area of application for quantum computation. In particular, quantum algorithms applied to the electronic ...

The knowledge economy: emerging ... - Wiley Online Library
explain the microfoundations and market mechanisms that underpin organizational disaggregation and the communal gover- nance forms observed in the knowledge economy. Because of the increasingly cen- tral role of HR professionals in knowledge manageme

XIIntention and the Self - Wiley Online Library
May 9, 2011 - The former result is a potential basis for a Butlerian circularity objection to. Lockean theories of personal identity. The latter result undercuts a prom- inent Lockean reply to 'the thinking animal' objection which has recently suppla

Aspects of the parametrization of organized ... - Wiley Online Library
uration is colder than the environment (the level of free sinking), with the initial mass ux set to 30% ...... XU95 criteria (see text) for the convective area. Panels (c) ...

Hormonal regulation of appetite - Wiley Online Library
E-mail: [email protected]. Hormonal regulation of appetite. S. Bloom. Department of Metabolic Medicine, Imperial. College London, London, UK. Keywords: ...

PDF(3102K) - Wiley Online Library
Rutgers University. 1. Perceptual Knowledge. Imagine yourself sitting on your front porch, sipping your morning coffee and admiring the scene before you.

Standard PDF - Wiley Online Library
This article is protected by copyright. All rights reserved. Received Date : 05-Apr-2016. Revised Date : 03-Aug-2016. Accepted Date : 29-Aug-2016. Article type ...