Household Decision-Making and HIV Care Continuum Outcomes Among Women Living with HIV in Mozambique

Abstract

Gender inequality has been associated with HIV infection among women. Less is known about the relationship between gender inequality and HIV care continuum outcomes. This study assessed whether household decision-making power (DMP), one component of gender inequality, was associated with linkage to HIV care, antiretroviral therapy (ART) initiation, or retention in care among women living with HIV in Mozambique. The sample included 600 women ≥18 years of age newly diagnosed with HIV in Mozambique. Data were collected between April 2013 and June 2016. DMP was assessed with three questions about participation in decisions regarding: major household decisions, routine household decisions, and visits to family. Women who did not contribute to decision making in any domain were categorized as having low DMP. HIV care information was obtained from electronic medical records. Multivariable log-Poisson regression analyses were conducted to assess the relationship between DMP and HIV care continuum outcomes. Almost half (49%) of participants reported medium DMP, followed by high (39%) and low DMP (12%). In multivariable regression analyses, when compared with respondents with medium DMP, those with low DMP had a lower likelihood of retention in care 12 months after diagnosis [adjusted risk ratio (aRR) 0.69 (95% CI 0.49–0.98)]. There was no significant relationship between low DMP and linkage to HIV care or ART initiation. High DMP was not significantly associated with HIV outcomes assessed. Women who report low DMP may face barriers to long-term engagement in HIV care. Interventions to improve retention in care among women should consider the role of household decision making and gender inequality.

Introduction

Globally, 19 million women and girls are living with HIV, constituting approximately half of all people living with HIV (PLWH). In sub-Saharan Africa, a region which bears the greatest burden of the HIV epidemic, women constitute 58% of PLWH.¹ Gender disparities in HIV incidence and prevalence persist and are particularly acute among younger populations. In sub-Saharan Africa, 80% of new infections among adolescents ages 15–19 occur among girls.² Despite advances in HIV treatment, AIDS-related illnesses remain the leading cause of death of women of reproductive age globally.³ As the global community strives to achieve the UNAIDS 90-90-90 goals, it is critical to advance our understanding of factors that may influence outcomes across the HIV care cascade among women living with HIV.

Gender inequality has been associated with HIV sexual risk behaviors and HIV infection among women.^4–7 In recognition of this, the 2016 United Nations General Assembly Political Declaration on Ending AIDS identified eliminating gender inequalities as a key component of increasing the capacity of women to protect themselves from HIV.⁸ Similarly, the UNAIDS Gap Report included promoting and implementing laws and policies related to gender equality as a top strategy to reduce women's vulnerability to HIV.⁹ Gender inequality can take many forms, including limited household decision-making power (DMP), limited access to financial resources, restricted mobility, and intimate partner violence.

While the relationship between gender inequality and HIV vulnerability has been well established, less is known about the relationship between gender inequality and HIV care continuum outcomes. Existing studies examining gender equality and HIV treatment outcomes have largely focused on pregnant women, and findings have been equivocal. For example, decision-making autonomy was associated with participation in prevention of mother to children transmission services among women living with HIV in Malawi and Uganda, but not significantly associated with maternal initiation of antiretroviral therapy (ART) during pregnancy in Zambia.^10,11 Greater understanding of the relationship between gender inequality and HIV care continuum outcomes among women living with HIV is needed and can inform the development, implementation, and evaluation of strategies to improve the health of women living with HIV. This work examined the relationship between household decision making, an important component of gender equality, and HIV care linkage, ART initiation, and retention in HIV care among women recently diagnosed with HIV in Mozambique.

Methods

Data for this analysis were drawn from the Engage4Health study, which evaluated the effectiveness of a combination intervention strategy (CIS) to enhance linkage to and retention in HIV care following HIV diagnosis at HIV care clinics in Mozambique.¹² Ten primary health clinics in Maputo City and Inhambane Province were matched on study volume and urbanicity, then randomized to implement either standard of care (SOC) or a CIS consisting of point-of-care CD4 testing at diagnosis, accelerated ART initiation, and SMS health messages, and appointment reminders. A pre/post intervention two-sample design was nested within the intervention arm to assess the effectiveness of an enhanced version of the CIS (CIS+) providing noncash financial incentives for linkage and retention. Eligibility criteria included being ≥18 years of age, understanding Portuguese or Xitsua, not currently pregnant, not planning to leave the community in the next 12 months, agreeing to be referred to HIV care services at the diagnosing facility, and not having been enrolled in HIV care or initiated ART in the prior 6 months. Participants were followed for 12 months, and completed three interviewer-administered questionnaires. Baseline questionnaires gathered information on sociodemographic and psychosocial characteristics, including household decision making. For participants who linked to care at study facilities, dates of linkage to care and subsequent clinic visits were abstracted from electronic medical records. Participants were enrolled in the Engage4Health study between April 2013 and June 2015. Study data were collected between April 2013 and June 2016. Ethics approval was provided by Mozambique's National Committee for Bioethics for Health and Columbia University. Participants provided informed written consent. Further details on the study methodology and primary results are available elsewhere.^12,13

Measures

Because uncertainty persists in how best to measure household decision making, we examined two measures of household decision making as described below.¹⁴

Household DMP

Household DMP examines the extent to which participants participated in household decision making, and was assessed with the Household Decision-Making Scale.¹⁵ This scale comprises three questions about who makes decisions about: (1) major household decisions, (2) routine household decisions, and (3) visits to family or relatives. Response options included: participant alone; spouse or partner alone; both participant and spouse or partner; parent alone; or other. Informed by previous research, a three-level measure of DMP was constructed (low, medium, high).^16–18 Women who reported that they did not contribute to decision making (alone or with their partners) in any of the three domains were categorized as having low DMP; women who reported that they contributed to decision making (alone or with their partners) in at least one, but not all three domains, were categorized as having medium DMP; women who reported that they contributed to decision making (alone or with their partners) in all three domains were categorized as having high DMP. Similar to previous research focused on household DMP, this construction of household DMP assesses whether or not the respondent participates in household decision making, but does not distinguish between respondents who report making decisions by themselves and those who report making decisions jointly with their spouses or partners.^16,17

Household decision making by domain

Because it remains unclear if sole and joint decision making are meaningfully distinct expressions of autonomy, we further examined whether each type of decision (i.e., major household decisions, routine decisions, and visits to family) was made without input from the respondent, jointly by the respondent and her partner, or solely by the respondent.¹⁴ For each of the three variables that comprise the Household Decision-Making Scale, we categorized each type of decision as made: without the involvement of the respondent, jointly by the respondent and her partner, or solely by the respondent.

HIV care continuum outcomes

This analysis examined three HIV care continuum outcomes: linkage to HIV care, ART initiation, and retention in HIV care. Linkage to care was defined by at least one clinical consultation for HIV that included assessment of the patient's medical history or physical exam. Retention in care was defined by a clinic visit in the 90 days before the end of the 12-month study follow-up period, with no documentation that the patient had transferred to another facility following linkage to care or had died. ART initiation was defined as having initiated ART during the study follow-up period. HIV care continuum outcomes were assessed from the perspective of the diagnosing facility using data from the electronic medical records maintained by the study sites as part of routine clinical care. All study participants were included in these analyses, including those who did not complete follow-up interviews. Participants whose electronic medical records were not located were considered not to have linked to care at the diagnosing facility.

Sociodemographic and household characteristics

Sociodemographic covariates included age, education, number of children, living environment, food insufficiency, and employment status. Informed by previous DMP literature, household characteristics included head of household and number of adults in household.¹⁸

Analyses

Univariate analyses were conducted with baseline data to assess the prevalence of household DMP (i.e., low, medium, high) and decision making for each of the three decision-making domains. Analyses between key participant characteristics and DMP were conducted using Pearson chi-squared tests and Fisher's Exact tests, as appropriate. Bivariate and multivariable random-intercept multilevel log-Poisson models assessed the relationship between household DMP and HIV care continuum outcomes. All models accounted for clustering within health facilities with an empirical variance adjustment for small numbers of sampling units described by Morel.¹⁹ Bivariate and multivariable models assessed the relationship between each of the three spheres of household decision making assessed and HIV care continuum outcomes. Because individual decision-making variables were correlated with each other, multivariable regression models were run separately to examine the relationship between each type of decision making and each HIV care continuum outcome. Independent variables considered for inclusion were those deemed a priori as theoretically important such as age, level of education, employment, food insufficiency, and region (Fig. 1). Level of education and employment were not included in adjusted models because these variables were correlated with age.

FIG. 1.

Conceptual model of the relationship between household decision-making power and HIV care continuum outcomes.

Participation in household decision making

Most research to date has modeled household DMP as a dichotomous variable (i.e., low DMP vs. medium/high DMP). Such a construction precludes one's ability to assess to what extent medium and high levels of household DMP function as meaningfully distinct manifestations of DMP in their relation to HIV care continuum outcomes. To advance understanding the extent to which both low and high DMP differ as compared with medium DMP, high and low DMP were modeled as unordered categorical exposure variables. In this way, analysis of household DMP assesses to what extent participation in household decision making (i.e., participation in no, some, or all decisions) is associated with HIV care continuum outcomes.

Joint versus sole decision making

We also sought to understand to what extent sole versus joint decision making function as meaningfully different expressions of autonomy. As such, we also examined to what extent the relationship between household decision making and HIV care outcomes differed based on whether the respondent reported sole, joint, or no decision making within each decision-making domain. Because the proportion of sole and joint decision making varied substantially by domain, the relationship between household decision making (no, joint, or sole decision making) and HIV care continuum outcomes was assessed separately by decision-making domain.

Results

Figure 2 illustrates the screening and enrollment of participants into the parent study. Details on study enrollment and exclusion has been previously published.¹² Briefly, 5327 adults 18 years of age or older were diagnosed with HIV at the study clinics during the study period. Of these, 265 individuals were not referred to study staff because they indicated that they were not interested in the study, were already receiving HIV services, or were not willing to be referred to the diagnosing health facility. Among the 5062 individuals referred to study staff, 3058 did not meet study eligibility criteria. The most common reasons for ineligibility were inability to provide informed consent due to distress of HIV diagnosis (19%), inability to understand Portuguese or Xitsua (12%), and refusal to be referred to the diagnosing health facility for HIV services (10%). Among the 2004 individuals enrolled in the parent study, 1292 were female (490 in the CIS group, 319 in the CIS+ group, and 483 in the SOC group). Of these 1292 women, 601 were married or cohabitating (223 in the CIS group, 141 in the CIS+ group, and 237 in the SOC group). This analysis includes 600 married or cohabitating women living with HIV for whom data regarding decision making were available (223 in the CIS group, 140 in the CIS+ group, and 237 in the SOC group). Most (65%) women included in this analysis were between 25 and 39 years of age (Table 1). Participation in formal education systems was common with 60% of women reporting having completed primary school and 28% having completed secondary school or above. Most women were employed (57%) and had at least one living child (85%). Approximately one-third of participants (37%) reported recent household food insufficiency.

FIG. 2.

Flow chart for participation in Engage4Health parent study. CIS, combination intervention strategy; SOC, standard of care; VCT, voluntary counseling and testing. Reprinted with permission from Elul et al.¹²

Table 1.

Participant Characteristics and Household Decision-Making Power Among Married or Cohabitating Women Living with HIV in Mozambique (n = 600)

	Total	Low (n = 74)	Medium (n = 294)	High (n-232)	p-Value
	n (%)	n (%)	n (%)	n (%)
Age					<0.0001
18–24	94 (16)	22 (30)	53 (18)	19 (8)
25–39	388 (65)	43 (58)	193 (66)	152 (66)
40+	118 (20)	9 (12)	48 (16)	61 (26)
Education					0.60
None	73 (12)	10 (14)	32 (11)	31 (13)
Primary	358 (60)	39 (53)	183 (62)	136 (59)
Secondary or above	169 (28)	25 (34)	79 (27)	65 (28)
Employment					0.09
Employed	342 (57)	38 (51)	159 (54)	145 (63)
Unemployed	258 (43)	36 (49)	135 (46)	87 (38)
No. of living children
0	85 (14)	20 (27)	43 (15)	22 (10)	0.001
1–2	268 (45)	30 (41)	140 (48)	98 (42)
3 or more	247 (41)	24 (32)	111 (38)	112 (48)
Another household member has HIV					0.47
No	388 (65)	45 (61)	197 (67)	146 (63)
Yes	212 (35)	29 (39)	97 (33)	86 (37)
No. of adults in household (besides respondent)					<0.0001
0–1	110 (18)	7 (10)	36 (12)	67 (29)
2–3	355 (59)	31 (42)	191 (65)	133 (57)
4+	135 (23)	36 (49)	67 (23)	32 (14)
Frequency of household food insufficiency in prior 12 monthsa					0.02
Never	375 (63)	49 (67)	168 (57)	158 (68)
Sometimes/always	222 (37)	24 (33)	125 (43)	73 (32)
Head of householda					<0.0001
Participant alone	81 (14)	2 (3)	12 (4)	67 (29)
Spouse or partner alone	359 (60)	25 (34)	227 (77)	107 (46)
Participant and spouse or partner together	71 (12)	0	17 (6)	54 (23)
Parent/other	88 (15)	47 (64)	38 (13)	3 (1)
Region					0.02
Inhambane	278 (46)	27 (37)	128 (44)	123 (53)
Maputo	322 (54)	47 (64)	166 (57)	109 (47)
Intervention arm					0.21
SOC	237 (40)	34 (46)	123 (42)	80 (35)
CIS	223 (37)	26 (35)	109 (37)	88 (38)
CIS+	140 (23)	14 (19)	62 (21)	64 (28)

^aMissing by variable: household food insufficiency n = 3; head of household n = 1.

CIS, combination intervention strategy; SOC, standard of care.

As shown in Table 2, most participants reported some participation in household decision making. About half (49%) noted that they regularly participated in decisions in at least one, but not all three spheres included in the decision-making scale (i.e., medium DMP) and 39% reported participating in decisions in all three spheres (i.e., high DMP). Decision-making participation varied by the sphere of decision making assessed. The majority of participants noted involvement in decisions related to routine household decisions, with 65% reporting making such decisions alone and 20% reporting making such decisions with a partner. Only 13% of participants reported that they made major household decisions alone while 30% reported making such decisions with a partner. Approximately half (48%) of participants reported making decisions to visit family jointly with their partner.

Table 2.

Household Decision Making Among Married or Cohabitating Women Living with HIV in Mozambique (n = 600)

	Household DMP n (%)	Household decision making by scenario
			Big household decisions	Routine household decisions	Visiting family or friends
			n (%)	n (%)	n (%)
Lowa	74 (12)	Woman does not contribute to decision making	343 (57)	90 (15)	207 (35)
Mediumb	294 (49)	Woman makes decision together with her spouse or partner	180 (30)	119 (20)	286 (48)
Highc	232 (39)	Woman makes decision alone	77 (13)	391 (65)	107 (18)

^aLow DMP: woman does not contribute to decision making across three scenarios.

^bMedium DMP: woman contributes to decision making in at least one, but not all three scenarios.

^cHigh DMP: woman contributes to decision making across all three scenarios.

DMP, decision-making power.

Household DMP and decision making by scenario were significantly associated with sociodemographic and household characteristics. As shown in Table 1, women who reported low DMP were more likely to be younger (i.e., ages 18–24), to not have children, to live in larger households (i.e., households with four or more adults), to report a parent or other individual (excluding a partner) as the head of the household, and to live in Maputo compared with women with medium or high DMP. Women who reported high DMP were more likely to be older (i.e., age 40 or older), to have three or more children, to have no or one other adult in the household, to report that they were the head of their household, and to live in Inhambane compared with women with lower levels of DMP. Women with medium DMP were more likely to report that their partner was head of the household and to report recent household food insufficiency compared with women with low or high DMP.

Table 3 shows the relationships between participant characteristics and decision making regarding big household decisions among study participants. When examined separately by scenario, all three spheres of decision making were significantly associated with participant age, number of living children, number of adults in the household, identified head of household, and region (routine decisions and visits to family: data not shown). Across all three decision-making scenarios, women who reported not participating in decision making were more likely to be younger, to not have children, to live in larger households, and to report that a parent or other individual (excluding a partner) was head of the household. Women who reported not participating in big household decisions were significantly more likely to be unemployed compared with those who participated in such decisions. Similarly, women who reported not participating in decisions to visit family were significantly more likely to be unemployed compared with women who reported participating in such decisions. Employment was not significantly associated with decision making related to routine household decisions.

Table 3.

Participant Characteristics and Decision Making Regarding Big Household Decisions Among Married or Cohabitating Women Living with HIV in Mozambique (n = 600)

	Total n (%)	Woman does not contribute to decision making	Woman makes decision together with her spouse or partner	Woman makes decision alone	p-Value
		n = 343	n = 180	n = 77
		n (%)	n (%)	n (%)
Age					<0.0001
18–24	94 (16)	72 (21)	13 (7)	9 (12)
25–39	388 (65)	221 (64)	129 (72)	38 (49)
40+	118 (20)	50 (15)	38 (21)	30 (39)
Education					0.78
None	73 (12)	38 (11)	24 (13)	11 (14)
Primary	358 (60)	209 (61)	102 (57)	47 (61)
Secondary or above	169 (28)	96 (28)	54 (30)	19 (25)
Employment					0.05
Employed	342 (57)	181 (53)	111 (62)	50 (65)
Unemployed	258 (43)	162 (47)	69 (38)	27 (35)
No. of living children					0.009
0	85 (14)	61 (18)	20 (11)	4 (5)
1–2	268 (45)	157 (46)	77 (43)	34 (44)
3 or more	247 (41)	125 (36)	83 (46)	39 (51)
Another household member has HIV					0.08
No	388 (65)	227 (66)	120 (67)	41 (53)
Yes	212 (35)	116 (34)	60 (33)	36 (47)
No. of adults in household (besides respondent)					<0.0001
0–1	110 (18)	40 (12)	34 (19)	36 (47)
2–3	355 (59)	208 (61)	120 (67)	27 (35)
4+	135 (23)	95 (28)	26 (14)	14 (18)
Frequency of household food insufficiency in prior 12 monthsa					0.06
Never	375 (63)	206 (60)	125 (70)	44 (57)
Sometimes/always	222 (37)	135 (40)	54 (30)	33 (43)
Head of householda					<0.0001
Participant alone	81 (14)	10 (3)	14 (8)	57 (75)
Spouse or partner alone	359 (60)	232 (68)	113 (63)	14 (18)
Participant and spouse or partner together	71 (12)	16 (5)	52 (29)	3 (4)
Parent/other	88 (15)	85 (25)	1 (1)	2 (3)
Region					0.006
Inhambane	278 (46)	143 (42)	88 (49)	47 (61)
Maputo	322 (54)	200 (58)	92 (51)	30 (39)
Intervention arm					0.41
SOC	237 (40)	146 (43)	66 (37)	25 (33)
CIS	223 (37)	124 (36)	68 (38)	31 (40)
CIS+	140 (23)	73 (21)	46 (26)	21 (27)

^aMissing by variable: Household food insufficiency n = 3; Head of household n = 1.

CIS, combination intervention strategy; SOC, standard of care.

In bivariate analyses, DMP was not significantly associated with linkage to care, ART initiation, or retention in care 6 months after HIV diagnosis (Table 4). However, DMP was significantly associated with retention in care 12 months after diagnosis. Indeed, 39%, 60%, and 55% of those with low, medium, and high DMP, respectively, were retained in care 12 months after diagnosis.

Table 4.

Decision-Making and HIV Care Continuum Outcomes Among Married or Cohabitating Women Living with HIV in Mozambique

		DMP				Routine household decisions			p-Value
	Total	Low	Medium	High	p-Value	Woman does not contribute to decision making	Woman makes decision together with her spouse or partner	Woman makes decision alone
	n = 600	n = 74	n = 294	n = 232
						n = 90	n = 119	n = 391
	n (%)	n (%)	n (%)	n (%)		n (%)	n (%)	n (%)
Linked to care at diagnosing facility
Same day as HIV test	371 (62)	39 (53)	180 (61)	152 (66)	0.14	53 (59)	79 (66)	239 (61)	0.48
Within 1 week of HIV test	442 (74)	49 (66)	214 (73)	179 (77)	0.16	64 (71)	97 (82)	281 (72)	0.09
Within 1 month of HIV test	494 (82)	59 (80)	241 (82)	194 (84)	0.73	74 (82)	100 (84)	320 (82)	0.86
Ever linked to care	534 (89)	65 (88)	260 (88)	209 (90)	0.79	81 (90)	105 (88)	348 (89)	0.92
Initiated ART	370 (62)	40 (54)	190 (65)	140 (60)	0.21	54 (60)	66 (55)	250 (64)	0.23
Retained at diagnosing facility
6 months after diagnosis	366 (61)	37 (50)	185 (63)	144 (62)	0.11	48 (53)	67 (56)	251 (64)	0.08
12 months after diagnosis	332 (55)	29 (39)	176 (60)	127 (55)	0.006	43 (48)	60 (50)	229 (59)	0.09

ART, antiretroviral therapy; DMP, decision-making power.

When examined separately by decision-making scenario, household decision making was not associated with ART initiation or retention in care across the three decision-making scenarios (Tables 4 and 5). However, decision making regarding visits to family was significantly associated with same-day linkage to care as well as linkage to care within 1 week and 1 month of diagnosis. More specifically, linkage to care was less common among those who did not participate in decision making regarding family visits. Decision making regarding big decisions or routine decisions was not significantly associated with linkage to care at any time points assessed.

Table 5.

Decision-Making and HIV Care Continuum Outcomes Among Married or Cohabitating Women Living with HIV in Mozambique

		Big household decisions				Decisions to visit family
	Total	Woman does not contribute to decision making	Woman makes decision together with her spouse or partner	Woman makes decision alone	p-Value	Woman does not contribute to decision-making	Woman makes decision together with her spouse or partner	Woman makes decision alone	p-Value
	n = 600	n = 343	n = 180	n = 77
						n = 207	n = 286	n = 107
	n (%)	n (%)	n (%)	n (%)		n (%)	n (%)	n (%)
Linked to care at diagnosing facility
Same day as HIV test	371 (62)	206 (60)	114 (63)	51 (66)	0.53	109 (53)	192 (67)	70 (65)	0.003
Within 1 week of HIV test	442 (74)	244 (71)	136 (76)	62 (81)	0.19	136 (66)	224 (78)	82 (77)	0.005
Within 1 month of HIV test	494 (82)	280 (82)	149 (83)	65 (84)	0.83	159 (77)	245 (86)	90 (84)	0.03
Ever linked to care	534 (89)	303 (88)	162 (90)	69 (90)	0.83	180 (87)	255 (89)	99 (93)	0.33
Initiated ART	370 (62)	212 (62)	110 (61)	48 (62)	0.98	122 (59)	177 (62)	71 (66)	0.44
Retained at diagnosing facility
6 months after diagnosis	366 (61)	207 (60)	113 (63)	46 (60)	0.83	117 (57)	184 (64)	65 (61)	0.21
12 months after diagnosis	332 (55)	187 (55)	105 (58)	40 (52)	0.58	103 (50)	168 (59)	61 (57)	0.13

ART, antiretroviral therapy.

In multivariable regression analyses, when compared with respondents with medium DMP, those with low DMP had a significantly lower likelihood of retention in care 12 months after diagnosis [adjusted risk ratio (aRR) _{low vs. medium DMP} 0.69 (95% CI 0.49–0.98)] (Table 6). There was no significant relationship between the level of DMP and linkage to care or ART initiation. When examined separately by decision-making sphere, household decision making was not significantly related to HIV care outcomes assessed (Table 7).

Table 6.

Bivariate and Multivariable Analyses of Level of Decision-Making Power and HIV Care Continuum Outcomes Among Married or Cohabitating Women Living with HIV in Mozambique

	Linked 1 month after diagnosis		Initiated ART		Retained 12 months after diagnosis
	Bivariate RR (95% CI)	Multivariable aRR (95% CI)a	Bivariate RR (95% CI)	Multivariable aRR (95% CI)a	Bivariate RR (95% CI)	Multivariable aRR (95% CI)a
Decision-making power
Low	0.99 (0.79–1.25)	1.00 (0.78–1.29)	0.85 (0.63–1.13)	0.87 (0.64–1.18)	0.66 (0.47–0.94)	0.69 (0.49–0.98)
Medium	1.00	1.00	1.00	1.00	1.00	1.00
High	0.97 (0.83–1.14)	0.95 (0.80–1.13)	0.91 (0.74–1.12)	0.87 (0.70–1.08)	0.89 (0.72–1.09)	0.85 (0.70–1.03)

All models adjusted for intervention assignment and accounted for clustering by site.

^aMultivariable models also adjusted for age, region, and household food insufficiency.

aRR, adjusted relative risk; ART, antiretroviral therapy; CI, confidence interval; RR, relative risk.

Table 7.

Bivariate and Multivariable Analyses of Domains of Household Decision Making and HIV Care Continuum Outcomes Among Married or Cohabitating Women Living with HIV in Mozambique

	Linked 1 month after diagnosis		Initiated ART		Retained 12 months after diagnosis
	Bivariate RR (95% CI)	Multivariable aRR (95% CI)a	Bivariate RR (95% CI)	Multivariable aRR (95% CI) a	Bivariate RR (95% CI)	Multivariable aRR (95% CI)a
Routine household decisionsb
Sole decision making	1.00	1.00			1.00	1.00
Joint decision making	0.97 (0.81–1.16)	0.98 (0.81–1.18)	0.86 (0.65–1.15)	0.85 (0.63–1.14)	0.86 (0.68–1.08)	0.84 (0.66–1.06)
No input in decision making	1.00 (0.83–1.20)	1.01 (0.83–1.24)	0.93 (0.72–1.21)	0.98 (0.76–1.28)	0.81 (0.63–1.05)	0.87 (0.67–1.12)
Decisions to visit familyc
Sole decision making	1.00	1.00	1.00	1.00	1.00	1.00
Joint decision making	1.00 (0.84–1.21)	1.02 (0.85–1.23)	0.93 (0.74–1.17)	0.93 (0.73–1.17)	1.03 (0.79–1.34)	1.02 (0.79–1.33)
No input in decision making	0.97 (0.80–1.18)	0.99 (0.80–1.23)	0.91 (0.70–1.18)	0.97 (0.74–1.27)	0.90 (0.69–1.18)	0.98 (0.75–1.29)
Big household decisionsd
Sole decision making	1.00	1.00	1.00	1.00	1.00	1.00
Joint decision making	1.01 (0.81–1.25)	1.03 (0.82–1.28)	0.99 (0.74–1.33)	0.99 (0.73–1.35)	1.14 (0.81–1.61)	1.16 (0.83–1.62)
No input in decision making	1.04 (0.85–1.27)	1.07 (0.86–1.32)	1.02 (0.76–1.37)	1.07 (0.78–1.47)	1.09 (0.78–1.51)	1.18 (0.85–1.64)

All models adjusted for intervention assignment and accounted for clustering by site.

^aMultivariable models also adjusted for age and region.

^bModel includes routine household decisions, but not decisions to visit family or big household decisions.

^cModel includes decisions to visit family, but not routine or big household decisions.

^dModel includes big household decisions, but not routine household decisions or decisions to visit family.

aRR, adjusted relative risk; ART, antiretroviral therapy; CI, confidence interval; RR, relative risk.

Discussion

We examined patterns of household DMP and its relationship to HIV care linkage, ART initiation, and retention in HIV care among 600 women newly diagnosed with HIV at 10 primary health facilities in Mozambique. Participation in some aspects of household decision making was high with over 85% of respondents reporting that they participated in decision making in at least one of the scenarios assessed. A study of women enrolled in HIV care in Ethiopia that used a similar approach to operationalizing DMP, found some participation in household decision making among 90% of participants.¹⁸ A study among pregnant women living with HIV in Zambia found that 77% reported participation in at least one aspect of household decision making.¹⁰ However, differences in study populations and measurement of DMP make direct comparisons difficult. Participation in routine household decisions was most common among study participants, with approximately two-thirds of women reporting making these types of decisions alone. Participation in big household decisions was least commonly reported with more than half of participants reporting not participating in these types of decisions.

When compared with participants reporting medium levels of DMP, participants reporting low levels of DMP were less likely to be retained in care 12 months after diagnosis, but not less likely to have linked to care or to have initiated ART. However, overall linkage to care was high among participants and may have limited our ability to identify a relationship between these variables. Our findings suggest that women with low DMP may find long-term engagement in health care particularly challenging. Women with low DMP may face restrictions on mobility, limited access to transportation, or financial or social resources that facilitate long-term retention in HIV care.²⁰ In addition, women with low household DMP may also have limited autonomy to make decisions regarding their own health care. Indeed, surveys of married women across multiple settings have found that many married women do not participate in decisions regarding their own health care.^21–23 Participation in health care-related decisions has been associated with health care use among women.^24,25 The Household Decision-Making Scale, the measure used in this study, does not ask about health care decision making. Future research should examine the relationship between health care decision making and HIV care outcomes. The Demographic and Health Surveys commonly ask about health care decision making and may serve as a useful resource for such research.^22,24,25

When compared with medium DMP, high DMP was not significantly associated with HIV outcomes assessed. This is among the first research to investigate the extent to which participating in all (i.e., high DMP) versus some (i.e., medium DMP) household decisions is associated with HIV care outcomes. Most extant research related to household DMP and health has used a dichotomous construction of DMP (i.e., low vs. high DMP), precluding such a comparison. Our findings suggest that not participating in household decision making (i.e., low DMP) may serve as a barrier to long-term engagement in HIV care. However, participation in all (i.e., high DMP) versus some (i.e., medium DMP) household decisions may not serve meaningfully different functions in relation to HIV engagement and retention in care. Additional research into the relationship between medium and high levels of DMP is warranted, given the limited extant research in this area. Qualitative research to explore access to and engagement with health care among individuals with low, medium, and high DMP may yield important insights.

Across decision-making domains, joint decision making was not associated with HIV care outcomes as compared with sole decision making. To the authors' knowledge, this is the first research to examine the role of joint versus sole household decision making in relation to HIV outcomes. Our findings suggest that joint and sole decision making may not function as meaningfully different expressions of autonomy in relation to engagement and retention in HIV care. Additional research is needed to advance understanding of the relationship between household decision making and HIV care. Such research should include various cultural and geographic settings as household DMP norms likely vary across settings and cultures. Greater understanding of interactions between decision-making norms at the community and individual levels is also warranted. Ecological analyses, which capture individual deviations from community norms, may provide insight into decision-making dynamics and relationships with health care access.

To our knowledge, this is among the first research to examine the relationship between DMP and HIV care cascade outcomes. Most research on the relationship between DMP and women's health has focused on reproductive or maternal health care access or outcomes. Given that HIV remains the leading cause of death of women of reproductive age globally, more research is needed into the relationship between autonomy and HIV care cascade outcomes among women living with HIV.³ Potential mediators or moderators of the relationship between low DMP and retention in HIV care, including mental health symptoms or disorders, social support, intimate partner violence, and coping mechanisms, should be investigated.

Greater attention is needed to the ways in which gender inequality affects women's ability to achieve optimal outcomes across the HIV care continuum. A systematic review of interventions to address gender inequality and improve self-efficacy and empowerment for sexual and reproductive health decision making among women living with HIV found that most interventions focused on reducing sexual risk behaviors, and few aimed to improve HIV care cascade outcomes.²⁶ Among intervention studies that examined HIV treatment outcomes, most focused on self-reported ART adherence, with equivocal findings. A gender-specific group-based intervention using motivational interviewing was associated with significantly higher levels of ART adherence among women living with HIV in Nigeria.²⁷ A gender-specific group-based intervention using cognitive/behavioral stress management with women living with HIV in the United States found no main intervention effects on self-reported ART adherence.²⁸ However, post-hoc analyses found that the intervention was significantly associated with increased ART adherence among those who reported low ART adherence at baseline.²⁸ More research is needed into the mechanisms through which gender inequality influences a broad range of outcomes across the HIV care continuum. Given the relationship between low DMP and retention in HIV care, couple-based interventions with women with low DMP and their partners may offer a promising path forward, particularly as couple-based interventions have shown promise in relation to reducing HIV risk and increasing engagement with health care.^29,30 For example, a home-based couple intervention with pregnant women living with HIV and their partners in Kenya was associated with increased uptake of perinatal health behaviors, including having participated in couples' HIV testing and counseling as well as having a postpartum checkup for the mother.²⁹

Our study has several important strengths. It is among the first to examine the relationship between household DMP and HIV care continuum outcomes, including linkage to HIV care, ART initiation, and retention in HIV care among women living with HIV. In addition, this study used objective markers of linkage and retention in care derived from participants' electronic medical records, rather than self-report. By using longitudinal data, our study moved beyond correlations to examine how gender inequality may shape HIV care outcomes. Finally, our analysis examined two measures of DMP, the first a more commonly used scale, and the second, an examination of decision making by scenario, allowing exploration of more nuanced aspects of DMP.

This study has limitations worth noting. Data were obtained from individuals receiving care at HIV clinics in two regions in Mozambique and may not be representative of women living with HIV in other regions of Mozambique. Outcome data were limited to retention, ART initiation, and linkage at the diagnosing facility. Some participants may have linked to care at other facilities. As such, these data may underestimate overall linkage, ART initiation, and retention rates. Linkage to care was high among participants and may have limited our ability to examine the role that decision making plays in relationship to linkage to care. Data were limited to women who consented to enroll in the parent study. It is possible that some women with low DMP did not feel comfortable enrolling in the parent study. If so, women with the lowest levels of DMP may be underrepresented in this study. On the other hand, it is possible that women with low levels of DMP are overrepresented in this sample if, for example, study participants were likely to have contracted HIV from their sexual partner due to limited power to negotiate condom use or other safer sex practices. In addition, differences in DMP may exist between married and unmarried cohabitating women. The current data do not allow for disaggregation of these two groups. Future research into the influence of marital and cohabitating status on household DMP is warranted.

As HIV treatment has largely transitioned to a chronic care treatment model, more nuanced understanding of barriers and facilitators to women's long-term retention in HIV care is needed. Qualitative research may be particularly helpful in gaining greater insight into the ways in which household decision making and gender inequality affect HIV care continuum outcomes. Longitudinal research which examines mediating and moderating factors of the relationship between household decision making and retention in HIV care is needed. Interventions to improve long-term retention in HIV care among women should consider the role of household decision making and gender inequality in long-term engagement in HIV care.

Author Disclosure Statement

No competing financial interests exist.

Funding Information

The parent study on which this analysis is based was supported by the United States Agency for International Development (USAID), USAID Award Number: AID-OAA-A-12-00027). This research was also supported by NIMH grant K01 MH114721 and NICHD grant P2C HD050924 (Carolina Population Center). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the article.