Attachment relationships of preschool-aged children of mothers with HIV and HIV-related psychosis

ABSTRACT

Children from mothers with HIV-related psychosis are frequently raised in challenging contexts, yet the extent to which these children grow up in insecure or disordered attachment relationships is unknown. Using the Strange Situation Procedure the distribution of attachment relationships of children from mothers with HIV and psychosis (n = 45) was compared with children from mothers with HIV without psychosis (n = 41). No significant differences in the distributions were found between the two groups and attachment was not associated with specific psychotic symptomatology. Security of attachment was associated with more people providing the mother with emotional support, but only in the psychosis group. Disordered attachment (24%) was more often found in the total sample than in studies with other normal and high risk populations. Recommendations were made for future research about factors facilitating resilience in the children and on interventions increasing emotional support for affected mothers.

KEYWORDS:

• HIV/AIDS
• psychosis
• attachment
• mothers
• high risk

Introduction

Parenting while having a diagnosis of both HIV and a psychotic disorder has never been empirically studied. In particular, studies are lacking on the attachment relationships of mother–child dyads living in this situation (Spies, Sterkenburg, Schuengel, & Van Rensburg, 2013). Literature reporting on the attachment relationship where the mother has either HIV or psychosis alone is also limited. Two independent systematic literature reviews on child attachment and maternal psychotic disorders by Spies and colleagues (2013) and Wan and Green (2009) found only two studies, both from the 1980s. D’Angelo (1986) and Näslund, Persson-Blennow, McNeil, Kaij, and Malmquist-Larsson (1984) compared the attachment relationships of infants of mothers with non-organic psychosis to demographically similar unaffected dyads using the Strange Situation Procedure (SSP). A significant elevation in insecure attachment in the psychosis groups was reported in both studies. D’Angelo (1986) reported more specifically that anxious-avoidant attachment was significantly higher in the psychosis group. As Main and Solomon (1986) proposed the category of disorganized attachment in 1986, the disorganized pattern was not taken into consideration. Following the work of Main and Solomon, operationalizations of quality of attachment relationships in older children also have identified maladaptive attachment patterns beyond the organized secure, avoidant, and resistant patterns, including so-called disordered attachment relationships among children up to and including the age of five (Cassidy & Marvin, 1992) and disorganized-controlling (punitive or role-reversing) patterns among children at age six (Main & Cassidy, 1988). Disorganized attachment in infancy has been found moderately predictive of this disorder and disorganized-controlling relationships in longitudinal studies (Van IJzendoorn, Schuengel, & Bakermans-Kranenburg, 1999). It would thus be important to study the possible association between attachment and psychotic symptomatology with the current knowledge of attachment theory, especially for families that have to deal with HIV infection as well.

Studies focusing on HIV and attachment are mostly focused on attachment styles of adults with HIV (Feeney & Raphael, 1992; Koopman et al., 2000) or the attachment relationships of HIV-positive children (Dobrova-krol, Bakermans-Kranenburg, Van IJzendoorn, & Juffer, 2010). Only one study could be found on mother–infant attachment relationships where the mothers were HIV positive. Using the Waters Attachment Q-Set, Peterson (1994) found no significant difference between the attachment quality of infants with mothers who were HIV-positive versus HIV-negative. A distinction was made however, between HIV-positive healthy mothers and those who were diagnosed with AIDS. Significantly fewer infants had secure attachment relationships with their mothers in the AIDS group compared to the HIV-positive healthy group. These findings underscore the importance of distinguishing between the complex risk factors and challenges that HIV-positive mothers and their children have to deal with.

A meta-analytic study by Cyr, Euser, Bakermans-Kranenburg, and Van IJzendoorn (2010) identified seven socio-economic risk factors that significantly predict insecure and disorganized attachment. These are substance abuse, being part of a minority group, being an adolescent mother, low education, low income, single parenthood, and maltreatment. An accumulation of at least five of the first six risk factors is as predictive of disorganized attachment as maltreatment. It is therefore important to investigate the possible links between attachment and the parenting contexts of children of mothers with HIV and psychosis when possible confounding variables are accounted for such as socio-economic stressors.

In the study by Spies, Sterkenburg, Schuengel, and Van Rensburg (2016) the mothers emphasized the value of support figures in their lives. They reported that these support figures at times completely took over their caregiving responsibilities when they were incapable of looking after their children. They emphasized the importance of family support in times of their illness. Dolman, Jones, and Howard (2013) also reported on the importance of the support network for women with schizophrenia where the women reported that support alleviated some of the demands during times of illness. Social support and in particular emotional support were found to be a predictor of quality of life for people diagnosed with psychosis in a study by Caron, Lecomte, Stip, and Renaud (2005). With increased emotional support and quality of life, women with psychosis may be more empowered in being mothers as well. Emotional support is valued in these studies and may need further investigation. This can be done by measuring the possible moderating effects between psychosis and parenting and in particular the attachment relationship.

The first aim of this study was to contrast the distribution of attachment patterns of preschool and early school-aged children from a group of mothers with HIV with a group of mothers with HIV and psychosis. Secondly, possible associations between attachment status and socio-economic risks were investigated. Thirdly, the possible role of emotional support as a moderator for the attachment relationships of this sample was investigated. Fourthly, the distribution of attachment classifications from the sample in this study was compared to other studies of children with comparable ages from community as well as at risk samples.

Firstly, on the basis of previous studies showing elevated rates of insecure attachment relationships of children with mothers with HIV (Peterson, 1994) or psychosis (D’Angelo, 1986; Näslund et al., 1984) and on the basis of the associations between anomalous parental behaviour with disorganized attachment (Madigan et al., 2006; Main & Hesse, 1990), more disordered attachment relationships was hypothesized between children and mothers with HIV and psychosis than between children and mothers with HIV only, while rates on insecure attachment may be similar. Secondly, based on the meta-analytic links between accumulation of socio-economic risk and disorganized and insecure attachment (Cyr et al., 2010), a positive association was hypothesized between the sum of socio-economic risk factors and disorganized and insecure attachment. Thirdly, based on studies suggesting a protective effect of social support (Caron et al., 2005; Dolman et al., 2013) as well as based on qualitative reports of the importance of support figures in the lives of mothers with psychosis and/or HIV (Spies et al. 2016), the hypothesis was tested that access to emotional support would compensate and buffer the risk of insecure and disorganized attachment. Fourthly, the hypothesis was tested that insecure and disordered attachment in this high risk population that is confronted with psychosis and/or HIV was elevated in comparison to general population samples as well as other high risk samples studied in the literature (Cyr et al., 2010; Van IJzendoorn et al., 1999).

Methods

Participants

Researchers identified three hospitals and seven clinics from the North-West-, Gauteng-, and Kwa-Zulu Natal provinces of South Africa who provided specialized treatment of people with HIV and psychiatric illnesses. Participants were included if they were mothers of a child between the ages of two and nine years and had a diagnosis of HIV or HIV and a psychotic illness. Only mothers who were treated as outpatients, thus not staying in hospital, were included. To prevent risk to the children, mothers with acute psychotic symptoms were not included in the study. The length of time that the children were exposed to mothers who were experiencing psychotic symptoms was unknown. The case group included 45 mothers with HIV and psychosis, ranging between 20 and 49 years old (M = 33.4, SD = 6.4) and their children aged between two and nine years old (M = 5.1, SD = 2). The comparison group included 41 mothers with HIV and no psychosis, ranging between 22 and 46 years old (M = 33.5, SD = 6.4) and their children aged between two and nine years old (M = 4.9, SD = 2).

All the participants lived in under-resourced, peri-urban settings (townships) and reported a high number of socio-economic and maltreatment risk indicators for attachment insecurity and disorganisation as identified by Cyr and colleagues (2010). These were low incomes (94% indicated financial difficulties and at times not having food in their homes), low education (23% completed school and 19% only had primary level education), all were part of a minority group, and maltreatment of their children (75.6%) (see Table 1). Maltreatment, as assessed in the study by Cyr and colleagues (2010), refers to children being physically (80%) or sexually abused (11%) or physically neglected (74%). Ten (12%) of the participating children’s fathers were deceased at the time of data collection.

Table 1. Demographic characteristics of mothers in the sample (N = 86).

Categories	HIV mothers without psychosis	HIV mothers with psychosis	Total sample	Local population demographic comparison (Ramjee et al., 2012) n = 5753
Mother’s age (M, SD)	33.49 (6.38)	33.47 (6.36)	33.48 (6.33)
Child’s age (M, SD)	4.93 (1.85)	5.11 (1.96)	5.02 (1.9)
Number of siblings (M, SD)	1.34 (1.26)	1.48 (1.02)	1.4 (1.1)
Highest level of education (n, %)
≤ Primary completed	9 (22)	7 (16)	16 (19)
Some secondary completed	24 (59)	26 (58)	50 (58)	66%
≥Secondary completed	8 (20)	12 (27)	20 (24)	34%
Employed mothers (n, %)	12 (30)	4 (15)	16 (18)	21%
Town (n, %)
Potchefstroom	39 (95)	9 (21)	48 (57)
Klerksdorp	0 (0)	2 (5)	2 (2)
Stilfontein	1 (2)	5 (11)	6 (7)
Newcastle	0 (0)	1 (2)	1 (1)
Johannesburg	0 (0)	26 (59)	26 (31)
other	1 (2)	1 (2)	2 (2)
Single parent (n, %)	24 (59)	39 (87)	63 (73)	78%
Financial problems (n, %)	39 (95)	42 (93)	81 (94)
Financial support from father (n, %)	18 (44)	10 (23)	28 (33)
Attachment risk factors (Cyr et al., 2010) (n, %)
Low income	41 (100)	44 (98)	85 (99)
Single mother	24 (59)	40 (89)	64 (74)	78%
Low education	33 (81)	34 (76)	67 (78)	66%
Adolescent mother	1 (2)	2 (4)	3 (4)
Minority	41 (100)	45 (100)	86 (100)	100%
Substance abuse	20 (49)	19 (42)	39 (45)
Maltreatment	30 (73)	35 (78)	65 (76)
Father passed away	4 (10)	6 (13)	10 (12)

M = mean, SD = standard deviation, n = number and % = percentage

The North-West University Ethics Committee granted ethical approval for this study, registered under NWU-00046–12-A1. Additionally, the departments of health of Gauteng-, North-West-, and KwaZulu Natal provinces approved the collection of data from participating health care facilities. Written informed consent was received from the mothers for participation for themselves as well as their children. Researchers followed the rules of ethical conduct of the Health Professions Council of South Africa.

Measures

Attachment assessment

Quality of attachment relationships between mothers and children was assessed with the SSP (Ainsworth, Blehar, Waters, & Wall, 1978) and coded with the MacArthur Preschool Attachment Classification System (Cassidy & Marvin, 1992). The SSP entails a series of episodes where children are separated and reunited from their attachment figure and a stranger. The procedure is designed to encourage natural interaction between attachment figures and children with the occasional introduction of mildly anxiety provoking situations (i.e. separations) that will enable the trained observer to witness a range of attachment- and exploration-related child behaviours. These patterns of behaviour are classified in one of five possible attachment categories. These are: secure, insecure-ambivalent, insecure-avoidant, controlling-disorganized, and insecure other. Children with secure attachment relationships will actively seek proximity and they are easily comforted upon reunion with the caregiver. Insecure-ambivalent children will seek proximity to the caregiver upon reunion, but also resist it and they experience little security in the caregiver’s presence. Children with insecure-avoidant attachment classifications may actively avoid or ignore the caregiver or maintain emotional neutrality during reunions. The controlling disorganized group is characterised by a strategy of controlling the caregiver during reunions in either a caregiving or punitive manner. Children classified as insecure other form part of a disordered group together with children classified as controlling-disorganized (Cassidy & Marvin, 1992).

The mean age of the children from this sample was five years and therefore the Preschool Attachment Classification System for children between the ages of two and five years (Cassidy & Marvin, 1992) was valid for the large majority of the children. Conceptually there is similarity with the Main and Cassidy (1988) system and thus for consistency the Cassidy and Marvin model of assessment was also applied to children of six and seven years old (cf. Oosterman, De Schipper, Fisher, Dozier, & Schuengel, 2010). The separation episodes were extended to five minutes instead of the usual three minutes for the age group six to nine years as has been suggested by Marvin when used in a clinical setting (R.S. Marvin, personal communication, 29 August 2011).

Procedures

Nursing staff from the health care facilities invited a total of 227 mothers who met the inclusion criteria for participation. The first author telephoned the mothers who agreed to be contacted and after research procedures were explained, 86 consenting mothers were included in the study. Mothers were not asked to give a reason for declining participation. Table 1 provides a comparison between the sample demographics and the population demographics of HIV positive women who lived in the same peri-urban setting (Ramjee et al., 2012). Mothers in the sample were similar to the population with respect to employment and marital status, but were slightly less educated as only 24% completed school compared to 34% in the wider population.

Participants knew their HIV status prior to data collection and received the necessary pre- and post-test counselling at the health care facilities. The diagnosis was made at the hospitals when a reactive rapid test and an Enzyme Linked Immunosorbent Assay (ELISA) or confirmatory rapid tested positive. All participants had also been seen by a multidisciplinary team that included at least a clinical psychologist, psychiatrist and nursing staff, who made a diagnosis of psychosis based on observed criteria of the DSM-IV-TR (American Psychiatric Association [APA], 2000). The National Consolidated Guidelines for the treatment of persons with a diagnosis of HIV state that pregnant or breastfeeding women, children under the age of five years, and people with a comorbid diagnosis of TB or hepatitis B will be provided with lifelong anti-retroviral treatment (ART) regardless of CD4 Cell count. For all other adults and children above the age of five years, ART is initiated when CD4 cell count drops to 500 cells/μl or below (Department of Health, 2015). Long-term treatment adherence studies indicate a loss to follow-up (not returning to hospitals or clinics for medication) of 24% after a 24 month period (Fox & Rosen, 2010). Patients who are retained in the health care system and who returned monthly for medication showed adherence rates over 90% (Dewing, Mathews, Fatti, Grimwood, & Boulle, 2014).

In cases where the mother was not fluent in English, interpreters facilitated the communication. Interpreters were students who studied African languages at the North-West University in South Africa (NWU) and whose first language was Zulu, Tswana, or Sotho. Participants received a R100 gift card (€7) and they were provided with transportation and a meal.

The settings for the data collection were observation rooms with video recording facilities at the NWU in Potchefstroom and Chris Hani Baragwanath Hospital (CHBH) in Johannesburg that were converted into playrooms for the SSP. Upon arrival at the venue for data collection, participants were again informed about confidentiality, the voluntary nature of participation and given the opportunity to ask questions. Demographic information was retrieved in interview format.

Following completion of the questionnaires, the SSP was performed. Female research assistants who were students from the NWU and VU University Amsterdam or colleagues of the first author were the strangers as they were unfamiliar to the child. All strangers received training and signed confidentiality contracts with the first author to protect the identities of participants. The procedure that was followed for the SSP as well as the instructions given to the strangers came from the training manual of Cassidy and Marvin (1992).

A professional videographer recorded videos at NWU and the first author did the recordings at CHBH. Video recordings were translated to English and subtitles were generated for the videos. Two independent, certified coders who were blind to the samples and objectives of the study coded the attachment classifications. Professor R.S. Marvin from The Mary D. Ainsworth Child–Parent Attachment Clinic trained the coders. Both coders had additional experience in the infant-classification system of the SSP. Ten of the 86 videos were coded by both coders. Inter-rater reliability for the coding of the attachment classifications was Cohen’s kappa = .71. Discrepancies were found in two of the videos and in this instance a third certified independent coder coded the videos. In both cases there was a match, so where two of the three coders agreed upon the classification, this classification was used. The coders’ notes of the SSPs were also examined to verify that they were observing similar behaviour that related to the specific attachment classification.

Analysis

Analysis of the data was performed using SPSS version 22. To compare the distribution of attachment of the two groups (HIV with or without psychosis), Chi-square test for independence were conducted. Binary logistic regression analysis was used to determine whether group (case versus comparison) predicted security or insecurity and organization or disordering of attachment. A sequential logistic regression analysis was used to investigate the possible effect of covariates and moderating variables on group and attachment classification. For a comparison of the distribution of attachment of our sample with those of other studies, the Multinom programme (Kroonenberg, 1998) was used to test the deviation of the distribution from the standard distribution. Standardized residuals of >±2 were considered significantly larger or smaller than expected marginals. This comparison closely corresponds that of Archer and colleagues (2015).

Results

Preliminary analysis

Demographic variables were compared between the two groups, HIV with and without psychosis in order to identify potential confounding variables. There were no significant differences on the demographic variables between the two groups with regards to age of the women and children, reported financial problems, alcohol and drug use of mother and father, education level of mother and father, domestic conflict, occupational status, and mortality of fathers.

Attachment: HIV with and without psychosis

Distributions of attachment patterns across the case and comparison groups are displayed in Table 2. In the total sample, attachment security was observed in only 41% of the mother–child dyads, with 59% being insecurely attached. Amongst the insecure classifications, avoidant strategies were most prevalent (26%). When the role-reversed controlling and insecure-other classifications were combined as a larger disordered group as proposed by Hoffman, Marvin, Cooper, and Powell (2006), 24% of the sample was classified as disordered.

Table 2. Distribution of attachment patterns across the case and comparison groups (n, %).

	HIV without psychosis	HIV with psychosis	Total sample
A: Insecure-avoidant	10 (24)	12 (27)	22 (26)
B: Secure	16 (39)	19 (42)	35 (44)
C: Insecure-resistant	4 (10)	4 (9)	8 (9)
D: Role-reversed/controlling	4 (10)	2 (4)	6 (7)
IO: Insecure other	7 (17)	8 (18)	15 (17)
Disordered (D + IO)	11 (27)	10 (22)	21 (24)

A chi-square test for independence was performed to determine if psychosis was related to attachment classification (secure, insecure ambivalent, insecure avoidant, and disordered). No significant association was found. In addition, the classifications of attachment were divided into participants who were classified disordered and participants who were ordered (i.e. secure, insecure avoidant, or insecure ambivalent). Chi-square test was performed to determine if group (mothers with HIV and mothers with HIV-related psychosis) predicted disordered attachment relationships. No significant differences were found.

Similarly two new groups were created based on security of attachment classification that included participants with secure attachment and participants with insecure attachments (i.e. insecure-avoidant, insecure-ambivalent, and disordered). The chi-square test revealed no significant effect of having a diagnosis of psychosis as a predictor for attachment security.

Socio-economic risks and attachment

The role of the seven socio-economic risk factors identified by Cyr and colleagues (2010) in disorganized/disordered as well as insecure attachment was investigated in this sample. Chi-square tests indicated no significant associations between presence of the individual aforementioned risks and disordered attachment. Chi-square tests also indicated no significant associations between having a father who passed away and disordered attachment. Having a diagnosis of psychosis with HIV, however, was associated with single motherhood. Significantly more mothers with HIV-related psychosis reported being single parents (χ2 (1, N = 86) = 10,381, p = .001). When a hierarchical multiple logistic regression was performed, none of these risk factors, nor the age and gender of the child moderated the effect of having a psychotic disorder on attachment security or disordered attachment.

Emotional support and attachment

A sequential logistic regression was performed using attachment security vs. insecurity as a dichotomous dependent variable. The number of family members forming part of the support network, the diagnosis of psychosis, and the number of identified support figures providing emotional support, were each entered as predictor variables. This step of the analysis yielded no significant effects and the model did not have a good fit when either of the predictor variables (family members, emotional support, and psychosis) was entered alone. In the next step an interaction factor consisting of having a diagnosis of psychosis in combination with the number of people providing emotional support was added. This interaction variable significantly predicted attachment security (p = .031) with an odds ratio [Exp (β)] of 3.14. This indicated that the odds for secure attachment were three times more likely when a mother with psychosis has a high number of people providing her with emotional support compared to less people providing emotional support. Figure 1 illustrates the two-way interaction effect for the logistic regression analysis. The proportion of dyads classified with secure attachment was higher in mothers with a diagnosis of HIV and psychosis if a higher number of people provided emotional support to the mothers.

Figure 1. Proportion of attachment security by psychosis diagnosis and number of emotional supporters.

Comparison with other populations

The results from this study were compared to other studies (Table 3). Standardized residuals were computed to assess deviations with summaries of results from other populations. The meta-analysis of Van IJzendoorn and his colleagues (1999) grouped the results from normal USA samples of children older than 24 months. Compared to the normal samples, the distribution of attachment did not deviate significantly for secure (B), avoidant (A), or resistant (C) classifications, but significantly more participants (Z = 2.31) were classified as disordered (D) in the current study. A comparison was also made with studies identified in the meta-analysis of Cyr and colleagues (2010) with high risk populations and where the Cassidy and Marvin (1992) system for the SSP was used on preschool-aged children. The frequencies for each classification were calculated using the sum of the frequencies of the four identified studies (Table 3). Disordered attachment was significantly more frequent in the current study (Z = 2.52) and insecure resistant attachment was significantly less frequent (Z = −1.96). Secure and insecure avoidant patterns did not show significant differences.

Table 3. Distribution of attachment classifications compared with other studies.

Study	Frequency				Standardized residuals
	A	B	C	D	A	B	C	D
This study	22	35	8	21	-	-	-	-
Normal US population ≥24 months	94	275	50	73	1.37	1.89	−0.25	2.31*
Studies from meta-analysis on high risk samples of preschoolers tested with SSP
Moore et al. (1997)^1	21	59	56	18
Moss, Cyr & Dubois-Comtois (2005)^2	37	139	29	37
Cicchetti & Barnett (1991)^2	11	11	4	10
Barnett, Kidwell & Leung (1998)^2	18	42	3	6
Total	87	251	92	71	1.83	1.23	1.96*	2.52*

* p < .05

Note. Comparison study sample with normal US > 24 months χ2 = 10.82, p < .05; Comparison study sample with at risk > 24 months χ2 = 15.08, p < .01

¹Van IJzendoorn et al. (1999)

²Cyr et al. (2010)

Discussion

Contrary to previous studies from D’Angelo (1986) and Näslund and colleagues (1984) on maternal psychosis, the classifications of attachment quality were comparably distributed between children with HIV-positive mothers diagnosed with a psychotic disorder versus children with HIV-positive mothers without psychosis. Psychosis on its own was therefore not a significant risk factor for insecure or disordered attachment in this sample. This ran counter to the expectation that the psychosis would increase the risk for disordered attachment, as Madigan and colleagues (2006) found moderate effect sizes for the associations between anomalous parental behaviour and disorganized attachment relationships. Similarly, Cyr and colleagues (2010) reported large effect sizes for the association between maltreatment and disorganized attachment relationships. Both anomalous behaviour and maltreatment were characteristic of the behaviour previously reported in our sample of mothers with HIV-related psychosis (Spies et al., 2016). For example, the mothers reported being aggressive towards their children and behaving in a frightening manner when they were experiencing active symptoms of psychosis.

A possible explanation for the similarity in the distribution of attachment between the two groups (HIV with and without psychosis) was that the psychotic symptoms were under effective control. The group of mothers with HIV-related psychotic disorders had confirmed diagnoses during recruitment, but received antipsychotic treatment. Consequently their symptoms were in remission when data collection took place. Recruitment was done among treated mothers, as potential exposure of the child during the SPP to the mothers’ acute psychotic symptoms would be unethical. The length of time and severity of the children’s exposure to the mother’s psychotic behaviour before the problems were detected and treated were unknown, making it impossible to establish whether a longer history of exposure to maternal psychotic symptoms may have increased the risk of insecure or disordered attachment. Another limitation concerns the use of the adapted preschool SSP for 11 children who were older than seven years. While in consultation with the coding system developer, the length of the separation episodes was increased, little is known about the validity for this age group. Removing these 11 children from the data did not substantially change the results, however.

The effects of exposure to maternal psychotic symptoms may also have been overshadowed by the large number of other risk factors that these families experience. Focusing on the risk factors linked by Cyr and colleagues (2010) to disorganized attachment, both groups of mothers reported high numbers of serious risk factors, which was consistent with the elevated percentage of disordered attachment classifications, even when compared to other at risk populations (Cyr et al., 2010). If maternal psychosis had any effects, it can apparently not be seen as distinct from the other risk factors in the lives of HIV-positive mothers in South Africa. Shisana and colleagues (2014) found that a disproportionate number of HIV positive people in South Africa are from lower socio-economic backgrounds that are under-resourced, thus making it difficult to study HIV associated psychosis in isolation from other risk factors. When comparing socio-economic risk factors from a different study conducted in a similar setting as our study, there was similarity in terms of employment and marital status, and a slight difference in education levels (Ramjee et al., 2012). This indicates the representativeness of the sample, yet the low level of participation may still be a confounder for attachment quality.

A limitation of this study was that other HIV-related illnesses and AIDS were not taken into consideration. As found by Peterson (1994), there may have been a difference in the attachment classifications in dyads where the mother was diagnosed with AIDS versus being HIV positive and being healthy. This information was not available to the researchers as it is standard practice for the doctors from participating hospitals not to write the word “AIDS” in patient files. A diagnosis of AIDS is assumed based on a combination of other diagnoses and/or test results. The mothers’ illness burden may have played a role in their psychological wellbeing as the qualitative study focussing on mothers’ illness experiences of HIV-related psychosis, indicated that some of the mothers suffered functional decline together with physical pain and fatigue (Spies et al., 2016). Myezwa, Stewart, Musenge, and Nesara (2009) assessed people being treated for HIV at one of the hospitals where our participants were recruited, using the International Classification of Functioning, Disability and Health (ICF). They found a number of impairments related to physical impairments (e.g. digestive, metabolic, and endocrine systems, sensory, mental functioning, etc.), activity limitations, difficulties performing everyday tasks, and interpersonal relationships, all add to the illness burden. Future studies may also include a third group with a similar demography without a diagnosis of HIV. This will allow for a comparison between dyads where the mother is healthy versus HIV positive.

The importance of emotional support for mothers with HIV and psychosis is highlighted in our study. Mothers with a diagnosis of a psychotic disorder who reported a high number of people providing them with emotional support had a three times higher chance of having a secure attachment relationship with their children, compared to mothers with fewer emotional supporters. This was not the case for mothers from the HIV only group. Tempier and colleagues (2013) found that emotional support for people with psychotic disorders predicted shorter times before remission of the symptoms, thus shortening the time of psychotic exposure to the child and possibly preventing negative consequences for the attachment relationship. Support in the South African peri-urban and rural areas is a normal occurrence and especially older women (e.g. grandmothers) play an active role in the caregiving of the children. They support the younger adults in the household with financial, physical, and emotional means (Schatz, 2007). The mothers from the psychosis group were, however, less likely to have a spouse as part of their support network. Spousal support did not predict attachment relationships in our study, but poor spousal support was associated with insecure attachment in another South African study by Tomlinson, Cooper, and Murray (2005). Their study was about the attachment relationships of infants of mothers living in poverty in a peri-urban setting similar to that of our sample.

The high percentage of disordered attachment classifications in this sample are put in relief by comparing with other normative and high risk samples. The dyads from this sample had significantly more frequently disordered attachment classifications even when compared to other high risk studies (Cyr et al., 2010). Although the children from this study are still of preschool age, their disordered attachment places them in a vulnerable position, as Sroufe (2005) reported that the consequences of disorganized attachment in infancy were associated with the development of psychopathology by age 17½ years and older. Children with a disordered attachment relationship were at high risk for dissociation, impulse control, and attention problems as well as conduct disorder. The high rate of disordered attachment relationships from our sample is thus concerning as these children have a prolonged risk for developing psychopathology.

An important finding from this study was that the number of securely classified dyads did not differ significantly from normal populations despite the high number of risk factors imposing on the mother–child relationship. This was similar to what Tomlinson and colleagues (2005) found in another South African peri-urban sample, testifying to the resilience of these populations. They attributed this to the collectivistic cultural and social organization of indigenous South Africans that may play a protective role in the upbringing of children. When a secure parent–child attachment relationship exists, it may buffer the negative effects of a high risk context. Houston and Grych (2015) found that children with secure attachments who were exposed to violence were less likely to be aggressive. The children from our sample were also exposed to contexts of violence and maltreatment and therefore secure attachment may safeguard them from becoming aggressive themselves. Tharner and colleagues (2012) also reported that secure attachment moderated the association between parenting stress and children’s behavioural and emotional problems.

Conclusion

Children and mothers who are HIV-infected are at heightened risk for disordered forms of attachment. Living in high risk socio-economic circumstances, these mothers have to raise their children while trying to cope with a diagnosis of HIV and in some cases also the burdens of a mental illness like psychosis. The high prevalence of insecure and disordered attachment relationships that develop from this context gives testimony to the need for intervention in this community as the attachment relationship forms the basis of children’s development in the long run. From the data, emotional support for the mothers with psychosis has been identified as a factor that future interventions may build on to promote attachment security. However, protective factors for the development of disordered attachment were not identified. Future efforts should thus focus on building the mothers’ support networks with a focus on emotional support. Resiliency studies investigating the factors that promote secure, ordered attachment relationships in this population will also be important as secure attachment may provide a buffer against in a high risk context.

Disclosure statement

No potential conflict of interest was reported by the author(s).