Factors associated with fluoxetine adherence among outpatients with common mental disorders in Western Kenya

Abstract

Objective

Non-adherence to antidepressants has been linked to increased symptom severity, relapse and hospitalisation from common mental disorders. However, there is limited knowledge of factors associated with antidepressant adherence in low-income and middle-income countries, especially in public sector, primary care settings.

Methods

We quantified fluoxetine adherence using the medication possession ratio. A limitation of this measure is that it does not always reflect the ingestion of medication. We constructed a generalised estimating equations linear regression with robust SEs, clustered by the participant, to identify independent predictors of fluoxetine adherence.

Results

Participants randomised to fluoxetine were dispensed an average of 126 daily doses, or 70% of the 180 possible doses. Adherence was higher in the first half of the treatment period at 86.3%, 95% CI (83.5% to 89.2%) compared with 46.5% in the second half (44.3% to 48.8%) (p<0.001). Participants who opted for community-delivered fluoxetine demonstrated adherence at 79.7% (77.0% to 82.4%) compared with 58.6% (55.7% to 61.5%) of those who only picked up medication at the facility (p<0.001). Use of mHealth for at least one but less than half of the visits had the highest level of adherence at 84.6% (82.4% to 86.9%) compared with 49.6% (46.1% to 53.0%) among those who did not use mHealth and 67.2% (62.5% to 72.0%) for those who used mHealth at least half their visits (p<0.001).

Conclusions

Adherence to fluoxetine was high relative to existing selective serotonin reuptake inhibitors adherence data, the majority of which is from high-income countries. Adherence was higher during the first half of treatment. People who were older, living with HIV and opted to use community delivery of medication and/or mHealth had higher adherence.

Trial registration number

NCT03466346.

WHAT IS ALREADY KNOWN ON THIS TOPIC

• We searched using PubMed for ‘factors influencing fluoxetine adherence in Africa’. While a significant body of research on antidepressant adherence exists for high-income countries, we found very few studies of populations in low-income and middle-income countries (LMICs).

WHAT THIS STUDY ADDS

• This study provides evidence on adherence to fluoxetine in a public-sector outpatient setting in Kenya for patients with major depression and/or post-traumatic stress disorder.

HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY

• Incorporating both in-person and mHealth appointments with fluoxetine delivery may improve adherence to medication in high-burden regions.

Background

The prevalence of mental disorders in Africa is one of the highest in the world, led by common, treatable illnesses such as major depression and trauma-related disorders, including post-traumatic stress disorder (PTSD). A key driver is the scarcity of evidence-based mental health services—only 3.7% of those with major depression and 2.3% with anxiety disorders, including trauma disorders, receive minimally adequate care.^{1 2} Psychosocial and pharmacological treatments are identified by the WHO as the two pillars of essential mental health care in non-specialised health settings (mental health treatment gap-intervention guide (mh-GAP)).³ However, many mental health treatment studies in low-income and middle-income countries (LMICs) test psychological or psychotherapy interventions only and exclude pharmacological strategies.

Fluoxetine: efficacy, availability and adherence studies

Premiered for clinical treatment in the 1980s, fluoxetine was the first widely used ‘modern’ antidepressant (selective serotonin reuptake inhibitors (SSRIs)) and is effective for treating symptoms of major depressive disorder (MDD).^{4 5} Despite the subsequent development of other SSRIs, none have shown greater clinical efficacy for depression than fluoxetine, and it remains a first-line treatment for depression.⁶ Fluoxetine is also in the class of first-line treatments for PTSD.⁷

Access to psychotropic medication, such as fluoxetine, is known to be very poor and disproportionately expensive in LMICs,8,10 including Africa. Fluoxetine is by far the most widely available modern SSRI antidepressant available in Africa.¹¹ Fluoxetine is typically the only modern SSRI antidepressant on the essential drugs list for some countries in Africa. Medications listed as ‘essential’ are theoretically available at most levels of public sector healthcare, where the vast majority of citizens access healthcare. However, despite the designation as ‘essential’, fluoxetine is often not prescribed, accessible or affordable.^{11 12} Improving our understanding of the effectiveness of fluoxetine for African populations is an impactful component of advocating for better availability.

Long-term adherence is crucial to effectiveness, given that standard first-line treatment with SSRIs is 6 months of daily treatment.^{13 14} Studies in high-income countries (HICs) find that approximately 30% of patients stop taking prescribed SSRIs in the first month and 60% within the first 3 months—with most patients receiving less than half of the recommended 6 months of initial SSRI treatment.^{15 16} Non-adherence has been linked to increased symptom severity, relapse and hospitalisation, exacerbating the prevalence and burden of mental disorders at a population level.^{13 14 17 18} Limited SSRI adherence data are available from public, primary care settings in Africa and other LMICs. A study in Ethiopia (n=217) measured adverse drug reactions and self-reported adherence among inpatients and outpatients with MDD.¹⁹ This study showed that the long duration of the disease (>2 years) and transport time (>2 hours) to the clinic were associated with self-reported non-adherence.¹⁹ Other studies of SSRI adherence in Africa are small and focus on narrowly defined populations. A feasibility study among pregnant women with postpartum depression in Zambia found a high level of adherence to SSRI (0.85, n=40), using a pill count measure of adherence.²⁰ A study (n=127) in a private healthcare cohort in South Africa showed slightly lower adherence to antidepressants among patients diagnosed with MDD and HIV compared with those with MDD alone, using medication possession ratio (MPR) to assess adherence.²¹ A study in India (n=626) found that 47% of patients adhered to antidepressants for 1 month.²²

The goal of this study is to address gaps in knowledge regarding SSRI adherence in Africa using a subset of data from the first-line fluoxetine treatment arm of Sequential, Multiple Assignment Randomised Trial (SMART) DAPPER. We conducted a substudy using data from the SMART DAPPER trial (n>2000) among public sector primary care patients with major depressive episode (MDE) and/or PTSD in Kisumu, Kenya. We evaluated fluoxetine adherence using the first-line treatment data from SMART DAPPER. Specifically, we aimed to identify key predictors of adherence to fluoxetine. The potential impact of this study is clear—given that adherence is a crucial underpinning of treatment effectiveness, advancing our understanding of fluoxetine adherence in Africa heavily informs advocacy for improved access.

Methods

Study design and population

As detailed elsewhere,²³ SMART DAPPER study participants were aged 18 years or older, primary care patients at Kisumu County Referral Hospital who met the threshold for MDE and/or PTSD on the Mini International Neuropsychiatric Interview (7.0.2). We excluded individuals with moderate or high risk of suicide, current/previous hypomania or mania, moderate or severe drug or alcohol use disorders, severe cognitive dysfunction, and those who were pregnant or breastfeeding, unable to attend weekly treatment visits, or were already receiving outside mental health treatment. Excluded participants were referred to a higher level of specialised care, as indicated. All participants provided written informed consent. Participants were enrolled between September 2020 and October 2021 and followed up for 30 months. Participants were randomised using a SMART design. Public sector primary care patients were randomised to receive fluoxetine or interpersonal psychotherapy (IPT) delivered by non-specialists.²³ Participants not in remission at the end of treatment were randomised to the opposite or combination treatment.²³ This study focuses on first-line fluoxetine treatment.

Treatment format: dispensing and duration

Fluoxetine treatment was initiated and monitored by trained clinical officers and nurses via eight treatment appointments occurring at baseline, 2 weeks, 4 weeks and then monthly until month 6. At the first treatment visit, participants were prescribed a daily dose of 20 mg capsule of fluoxetine. At each visit, the Patient Health Questionnaire-2 (PHQ-2) was administered to assess the severity of depressed mood and anhedonia over the past 2 weeks. After 1 month, if there were no prohibitive side effects and the PHQ-2 had not improved from the prior visit and was not ‘0’, then the dose was increased from 20 to 40 mg. If not, the participant continued with 20 mg of fluoxetine unless they had prohibitive side effects. At subsequent visits, if they did not have any prohibitive side effects and the PHQ-2 had not improved from the prior visit and was not ‘0’, then the dose was increased by 20 mg from their previous visit up to a maximal dose of 60 mg. 99.4% (n=5126) of delivered fluoxetine doses were 20 mg (the starting dose), 0.4% were 40 mg (n=20) and 0.2% were 60 mg (n=10). If participants missed or delayed treatment visits and subsequently returned to the study, they were dispensed medication as indicated and scheduled for return, but the duration of study treatment was not lengthened, so it was not possible for them to receive all 180 daily dosages of medication (6 months).

Treatment delivery model: visits and medication

SMART DAPPER enrolled participants from September 2020 to October 2021 and was therefore affected by the COVID-19 pandemic. To maintain participant and staff safety and adhere to local government travel restrictions, we revised our delivery model to include optional (not randomised) mHealth (audio-only mobile phone) treatment visits and/or community delivery of medication. We also allowed them to switch between in-person and mobile phone treatment as needed (eg, COVID infection). These strategies not only allowed us to safely continue study interventions during the COVID-19 pandemic, but they also afforded an opportunity to evaluate fluoxetine adherence when participants were provided with flexible treatment modalities. The first appointment was conducted in person. Participants could choose mHealth or face to face for subsequent appointments (and could alternate as desired). Participants could choose to pick up medication at the large outpatient primary care facility where the study was based or to have medications delivered to their home or community health facility. These strategies were adapted to meet the needs of the population (ie, audio-only mobile phones). Lessons were also adapted from HIV care programmes.

Measures

We quantified fluoxetine adherence using MPR. For this project, we defined MPR as a percentage—the number of capsules dispensed out of the total 180 possible daily doses (multiplied by 100 to achieve per cent). The MPR has been used in other studies because it is acceptable, acceptably accurate, convenient, objective, non-invasive and inexpensive to obtain in a large study population.¹⁷ Although the MPR does not reflect actual ingestion of oral medication, it is superior to subjective measures (eg, self-report), which are prone to recall and social desirability bias.¹⁶ The number of capsules dispensed was based on the prescribed dose and interval for the next visit plus two extra tablets (to reduce the chance of missed doses) minus the number of remaining pills. Fluoxetine prescribers instructed participants to return all their medication packets, both empty and partially used at each visit. A pill count was conducted at every visit to capture the number of capsules remaining. If the visit was conducted in person, the provider conducted the pill count, and if done over the phone, the participants were asked to count how many capsules were remaining. For example, if a participant attends the visits as scheduled and they are prescribed 20 mg, they should have two pills remaining from the prior visit, and the provider will prescribe 14 additional capsules at visits one and two (2-week intervals), and 30 additional capsules at visit three through seven (1-month interval). However, if the participant returns with no medication because their visit was delayed or they lost their medication, the provider would prescribe 16 or 32 capsules depending on the next scheduled visit date.

Potential factors affecting fluoxetine adherence

A conceptual model was used to determine adjustments and a sequence of models. We adapted Ickovics’ and Meisler’s conceptual framework²⁴ of factors affecting HIV treatment adherence to assess factors associated with fluoxetine adherence. Adherence to fluoxetine is influenced by multiple factors including: (1) demographic, (2) family and social support, (3) disease, (4) treatment and (5) health system factors (table 1). We began testing demographic factors and treatment factors, including the treatment phase, and then we added family and social support and health system factors.

Table 1. Potential factors affecting fluoxetine adherence.

Demographic factors	Family and social support factors	Disease factors	Treatment factors	Health system factors
Age	Marital status	Depression diagnostic threshold (BDI-2)*	Treatment Phase (1–3 months), (4–6 months)	Time to treatment initiation
Gender	Intimate partner violence (IPV) (CTS2)†	PTSD diagnostic threshold (PTSD Checklist for DSM-5 (PCL-5))‡	Side effects	Transport time to facility
Highest level of education		HIV and other co-morbidities§		Use of optional mHealth treatment
				Use of optional community delivery of fluoxetine

^*BDI-2 ≥ 19.36

^†IPV among partnered participants—CTS2, physical violence subscale.

^‡PTSD Checklist for DSM-5 (PCL-5) ≥23.37

^§High-prevalence medical comorbidities: HIV, hypertension, diabetes, tuberculosis, syphilis, hypothyroidism, hyperthyroidism.

BDI-2, Beck Depression Inventory; CTS2, Conflict Tactics Scale; PTSD, post-traumatic stress disorder.

Statistical analysis

Descriptive statistics with 95% CIs were used to quantify the percentage of pills dispensed. A comparison of overall adherence rates between groups was conducted using the Kruskal-Wallis rank test to accommodate a skewed distribution for adherence. A comparison of the change in adherence rates over time (visits 1–4 compared with visits 5–8) was conducted using a signed-rank test. We used a modified intention-to-treat analysis, requiring attendance of one or more treatment sessions. We split the course of treatment into two phases, months 1–3 and 4–6. We also constructed a generalised estimating equations linear regression with robust standard errors, clustered by the participant, to identify independent predictors of fluoxetine adherence. All p values are two-sided and we regard p<0.05 as statistically significant. All statistical analyses were done with SAS (V.9.4).

Results

Participants

2162 participants were enrolled in the study (figure 1). 1080 participants were randomised to fluoxetine for first-line treatment. 1041 fluoxetine participants were included in the modified intention-to-treat analysis.

Figure 1. CONSORT diagram. Tx, treatment; reasons for discontinuation. *Deaths, •Recommendation by investigator, ǂParticipant requests withdrawal or relocates. CONSORT, Consolidated Standards of Reporting Trials.

At baseline (table 2), the average age of participants was 35 years old, and the majority were married women with some primary or secondary education. At baseline, over 90% of participants had a diagnosis of major depression; approximately 52% had PTSD. Nearly half had both major depression and PTSD. Participants had high levels of lifetime trauma, including nearly 65% having experienced two or more types of trauma. Only 1% of participants had any previous mental healthcare.

Table 2. Baseline characteristics of subjects randomised to fluoxetine.

	(N=1080)
Age (years)
Mean±SD (N)	35.3±10.8
Median (min-max)	33 (18–77)
Gender
Male	99 (9.2%)
Female	981 (90.8%)
Education
None	19 (1.8%)
Some primary/primary	541 (50.1%)
Some secondary/secondary	418 (38.7%)
Post secondary	102 (9.4%)
Marital status
Currently married	518 (48.0%)
Never married	146 (13.5%)
Separated	238 (22.1%)
Divorced	7 (0.6%)
Widowed	161 (14.9%)
Cohabiting	9 (0.8%)
Time to the health facility (minutes)
Mean±SD (N)	37.7±24.8 (N=1079)
Median (Q1–Q3)	30 (30–45)
Transportation cost to the health facility (Ksh)
No cost	69 (6.4%)
1–99	576 (53.4%)
100–299	394 (36.5%)
300–499	33 (3.1%)
500 or more	7 (0.6%)
Baseline diagnosis(es) (MINI)
Major depression	1006 (93.1%)
PTSD	563 (52.1%)
Major depression and PTSD	508 (47.0%)
Depression symptoms (BDI II)
Mean±SD (N)	29.1±10.5 (n=1079)
Median (min-max)	28 (1–60)
PTSD symptoms (PCL-5)
Mean±SD (N)	43.3±17.2 (n=1079)
Median (min-max)	42 (0.0–80)
Comorbidities
HIV+	413 (38.2%)
Other comorbidities*	96 (8.9%)
Trauma
Physical intimate partner violence among partnered participants, ever (CTS)†	343 (60.4%)
No lifetime trauma history (THQ)	80 (7.4%)
1 type of lifetime trauma (THQ)	300 (27.8%)
2 types of lifetime trauma (THQ)	479 (44.4%)
3 or more types of lifetime trauma (THQ)	220 (20.4%)

^*Diabetes, high blood pressure, tuberculosis, hypothyroidism, hyperthyroidism, syphilis.

^†Among partnered partnered participants, N=568 (52.6%).

BDI II, Beck Depression Inventory; MINI, Mini International Neuropsychiatric Interview; PTSD, post-traumatic stress disorder; THQ, Trauma History Questionaire.

Adherence outcomes

Participants randomised to fluoxetine were dispensed an average of 126 daily doses, or 70% of the 180 possible doses (SD 60.1, n=1034). The main reason why participants did not receive 180 daily dosages was due to missed visits or not returning to the clinic on time. The mean treatment visit adherence was 59.1% (n=1080).

Potential factors affecting adherence

Demographics

Older age was associated with higher adherence. Among the oldest participants (>45 years), adherence was 76.7% (95% CI) (71.6, 35.1), compared with middle age (26–45 years) at 66.9% (64.3, 69.6) and the youngest age (18–25 years) at 55.4% (50.5, 60.3), (p<0.001). Gender and education were not statistically significantly associated with adherence (table 3).

Table 3. Factors associated with fluoxetine adherence.

	Adherence (%)	P value*
Overall	66.4±36.09 (N=1080)	N/A
Demographics
Gender
Male	68.20±38.68 (N=99)	0.28
Female	66.23±35.83 (N=981)
Age
18–25	55.41±35.35 (N=203)	<0.0001
26–45	66.93±35.71 (N=695)
>45	76.73±35.15 (N=182)
Formal education
None	65.26±35.69 (N=19)	0.2
Some primary/primary	68.32±36.02 (N=541)
Some secondary/secondary	65.64±35.15 (N=418)
Post secondary	59.68±39.76 (N=102)
Family and social support
Marital status
Currently married and cohabitating	67.38±34.81 (N=527)	0.56
Never married, separated, divorced and widowed	65.61±37.20 (N=552)
Baseline intimate partner violence (IPV)†
IPV	67.04±34.89 (N=453)	0.55
No IPV	69.19±38.32 (N=232)
Disease factors
Depression symptoms
Baseline - BDI-2≥19	66.87±35.87 (N=935)	0.35
Baseline - BDI-2<19	63.94±37.19 (N=144)
PTSD symptoms - PTSD Checklist for DSM-5 (PCL-5)
Baseline—PCL-5≥23	66.54±35.82 (N=956)	0.91
Baseline—PCL-5<23	65.95±37.91 (N=123)
Baseline comorbidities reported
HIV negative	63.63±34.61 (N=667)	0.0002
HIV positive	70.91±37.96 (N=413)
Has no other comorbidities‡	66.06±36.17 (N=984)	0.3
Has other comorbidities‡	70.06±35.17 (N=96)
Treatment factors
Treatment phase
Months 1–3	86.30±47.82 (N=1080)	<0.0001§
Months 4–6	46.53±38.22 (N=1080)
Side effects
Side effects	79.34±32.67 (N=27)	0.046
No side effects	66.08±36.12 (N=1053)
Health system factors
Treatment initiation
On randomisation day	69.53±33.75 (N=899)	0.17
After randomisation day	65.08±38.14 (N=139)
Treatment location
None	49.58±38.76 (N=488)	<0.0001
1–49.9% visits by phone	84.62±24.37 (N=445)
≥50% visits by phone	67.20±29.11 (N=147)
Community delivery of fluoxetine
≥1 delivery	79.68±27.69 (N=400)	<0.0001
No delivery	58.61±38.13 (N=680)
Time to clinic
0–30 min	66.21±35.60 (N=595)	0.97
31–45 min	66.73±34.66 (N=278)
>45 min	66.89±39.20 (N=206)

^*Krusal-Wallis test (unadjusted).

^†Among partnered participants (n=685).

^‡Diabetes, high blood pressure, tuberculosis, hypothyroidism, hyperthyroidism, syphilis.

^§Signed rank test.

BDI-2, Beck Depression Inventory; N/A, not available; PTSD, post-traumatic stress disorder.

Family and social support factors

Marital status and intimate partner violence (IPV) among partnered participants were not associated with fluoxetine adherence.

Disease factors

We measured the association of disease factors (ie, meeting the threshold for depression and PTSD) and other comorbidities. Meeting the threshold for depression and PTSD did not significantly impact adherence. People living with HIV (PLWH) had 70.9% adherence (67.2%, 74.6) compared with 63.6% (61.0, 66.3) among those without HIV (p=0.0002).

Treatment factors

During the first half of the treatment, adherence was higher at 86.3% (83.5%, 89.2%) compared with 46.5% (44.3%, 48.8%) during the second half of treatment (p<0.001). There was a higher adherence among participants who reported side effects at 79.3% (66.4%, 92.3%) compared with 66.1% (63.9%, 68.3%) who did not have side effects (p=0.046). We evaluated the relationship between fluoxetine adherence and reduction of depression and PTSD symptoms between baseline and treatment mid-point (month 3) and found no significant association.

Health system factors

Participants who selected community-delivered fluoxetine at least once demonstrated adherence at 79.7% (77.0%, 82.4%) compared with 58.6% (55.7, 61.5) among those who only picked up their medication at the health facility (p<0.001). Use of mHealth for at least one but less than half of the visits had the highest level of adherence (84.6% (82.4%, 86.9%) compared with 49.6% (46.1%, 53.0%) who did not use mHealth and 67.2% (62.5%, 72.0%) among those who used mHealth for more than half their visits (p<0.001). The timing of treatment initiation after randomisation or the time to the facility did not have a significant impact on adherence.

Multivariate analysis

We constructed a generalised estimating equations linear regression to identify independent predictors of fluoxetine adherence (table 4). Adherence to fluoxetine was higher in the first half of treatment than in the second half. Adherence in months 4–6 was 11% points lower compared with months 1–3 (−11% (95% CI −13.1% to –9.0%), p<0.001). Participants older than 45 years (7.7% (3.3% to 12.0%) p=0.015) and 26–45 years (3.9% (0.8% to 7.1%), p=0.006) had higher adherence compared with the youngest age category, 18–25 years. Community delivery was independently associated with better adherence (11.0% (8.2% to 13.8%), p<0.001) compared with those who did not utilise community delivery. Participants who had at least one but less than half of their visits by audio-only cell phone (mHealth) had greater per cent adherence (+9.5% (6.4% to 12.5%) p<0.001) compared with those who did not use telehealth (reference) and those who used telehealth more than half of the time (−6.7% (−9.6% to –3.8%), p<0.001). Gender, HIV status, other comorbidities, time to clinic and history of IPV were not statistically significantly associated with adherence to treatment.

Table 4. Multivariate generalised estimating equations linear regression for adherence.

	Per cent point difference (95% CI) compared with reference	P value
Treatment phase
Months 1–3	Reference
Months 4–6	−11.1% (−13.1% to –9.0%)	<0.0001
Gender
Male	Reference
Female	−2.8% (−6.1% to 0.5%)	0.097
Age
18–25	Reference
26–45	3.9% (0.8% to 7.1%)	0.015
>45	7.7% (3.3% to 12.0%)	0.0006
Community delivery of FLX
No delivery	Reference
≥1 delivery	11.0% (8.2% to 13.8%)	<0.0001
Baseline intimate partner violence (IPV)*
No history of IPV	Reference
History IPV	−2.2% (−4.7% to 0.4%)	0.10
Baseline comorbidities
HIV negative	Reference
HIV positive	1.6% (−1.2% to 4.1%)	0.22
No other comorbidities†	Reference
Other comorbidities†	−1.6% (−5.8% to 2.6%)	0.45
Time to clinic
0–30 min	Reference
31–45 min	−0.1% (−2.6% to 2.5%)	0.96
>45 min	2.1% (−1.3% to 5.5%)	0.23
Treatment location
None	Reference
1–49.9% visits by phone	9.5% (6.4% to 12.5%)	<0.0001
≥50% visits by phone	−6.7% (−9.6% to –3.8%)	<0.0001

^*Among partnered participants (n=685).

^†Diabetes, high blood pressure, tuberculosis, hypothyroidism, hyperthyroidism and syphilis.

Discussion

This study adds to the scarce data on adherence to SSRIs in LMICs, despite the designation of pharmacological interventions as essential aspects of mh-GAP to reduce the population’s mental health burden. Given the relationship between adherence and treatment success,^{13 14} SSRIs may hold promise for reducing regional gaps in mental healthcare.

Global context: We found that average fluoxetine adherence in this setting was more than double that observed in HICs.^{15 16} The differences seen in SSRI adherence between this study and HIC studies may reflect the types of people enrolled. For example, many people who enrol in clinical trials in the USA have tried and failed other treatments, and this may inadvertently target participants who may have lower treatment adherence. In contrast, only 1% of the participants in this study ever had mental health treatment. This underscores the importance of global mental health research to capture the full picture of treatment adherence.

The strongest predictor of fluoxetine adherence was the treatment phase, with adherence significantly higher during the first half of the treatment period, compared with the second half. In HICs, factors for non-adherence to antidepressants have included patient characteristics,²³ side effects,²⁵ medication and health system factors.26,28 In this population, older age was associated with higher fluoxetine adherence. This finding is consistent with prior studies of antidepressant adherence.2228,31 Possible explanations include greater experience with the detrimental effects of depressive episodes across the lifespan, which may motivate treatment adherence.²⁸ Stigma and common side effects like weight gain or impaired sexual function may be more troublesome for the younger populations.²⁸ In this study population, younger adults also move frequently for work opportunities which may impact their adherence. Mobility leads to interruptions in care or early termination, similar to barriers in HIV care and treatment.³² On univariate analyses, PLWH had higher adherence to fluoxetine than people without HIV. Potential explanations include greater health literacy among PLWH versus HIV-negative populations, including the overall importance of medication adherence, which is repeatedly emphasised to patients in local HIV programmes.

Other reasons for non-adherence to fluoxetine include both patient factors (eg, concerns about side effects, fears of addiction, belief that these medications will not address personal problems) and clinician factors (eg, lack of sufficient patient education, poor follow-up).³³ Other research suggests that the lack of education on the chronicity of depression and the need to keep taking medication, fear of dependency and, in some cases, side effects, may outweigh the benefits. Hence, some patients may discontinue treatment, especially when they feel better.^{33 34}

African context: Adherence was higher in this study than one conducted in Ethiopia where the average self-reported adherence was 4.74±2.19 out of eight.¹⁰ Participants who selected (non-randomised) community delivery of fluoxetine at least once averaged 80% adherence, higher than those who opted to pick up their medication from the study facility pharmacy. A modified version of community delivery has been used in HIV care programmes in Zimbabwe and Kenya. In Zimbabwe, clients formed groups and one client collected antiretroviral therapy (ART) for all members. Clients and providers found community delivery beneficial as they reduced barriers to ART adherence including travel time and cost to the facility.³⁵ One US-based study showed that people using only retail pharmacies had lower SSRI adherence than those who used mail-order or retail pharmacies.²⁶ We are not aware of any other studies of SSRI adherence with community delivery in LMICs.

Using a hybrid treatment model, allowing for both in-person and mHealth appointments with medication delivery, may improve adherence to SSRIs in some of the world’s most heavily burdened regions. In particular, individuals who used mHealth, but met in person the majority of the time, had higher adherence compared with those who did not use mHealth. In Ethiopia, patients with MDD who had to travel long distances for treatment were five times less adherent to their medications compared with those travelling shorter distances.¹⁹ However, the mHealth treatment offered in this study was associated with improved adherence. Policy-makers should consider incorporating mHealth and community delivery to further improve fluoxetine adherence. Other chronic disease management programmes in Kenya should also explore if these approaches improve adherence to other chronic diseases.

Limitations

This study was conducted as part of a treatment study comparing a scalable model of non-specialist delivery of fluoxetine or IPT psychotherapy for major depression and PTSD. As such, the features of fluoxetine treatment designed to encourage adherence were not explicitly tested by the study, and the data are observational. Information on many clinical and demographic characteristics was lacking. Community delivery of fluoxetine to a nearby health facility and mHealth was an optional service driven by the COVID-19 pandemic and not randomised. This non-randomised design could have accounted for potential biases. The COVID-19 pandemic may have impacted fluoxetine adherence as local public health regulations restricted movement. Although mHealth and community delivery of medication were offered, this may not have completely mitigated the restrictions and lockdowns. Furthermore, work and school opportunities were limited, and many people faced increased financial hardships which led to additional stress. Pills dispensed data do not prove that medication was swallowed, and they do not reflect daily intake variation to adherence or the timing of missed doses. Furthermore, pill counts conducted during mHealth visits were self-reported.

Conclusions

This study provides data on adherence to fluoxetine in a public-sector primary care outpatient setting in Kenya for patients with major depression and/or PTSD and adds to the scarce literature on the use of SSRIs in LMICs. As measured by MPR, adherence to fluoxetine was high relative to existing SSRI adherence data, the majority of which is from HICs. Adherence was higher during the first half of treatment. People who were older, living with HIV and opted to use community delivery of medication and mHealth had higher adherence.

The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.

Data availability statement

Data are available in a public, open access repository.