Antipsychotic Medication Adherence Among Outpatients Attending the Centre Neuro Psycho Pathologique in Kinshasa: A Cross-Sectional Pilot Study of Treatment Attitudes, Insight, and Side-Effect Burden
Yves Tshangala Kavunga, Julienne Nzuzi Mananga, Jael Won Vuvu Ngimbi, Philippe Ntalaja Kabuayi
TL;DR
This study found that nearly half of patients in Kinshasa did not properly take their antipsychotic medications, with treatment attitude and insight being key factors.
Contribution
The study provides new insights into antipsychotic adherence in the DRC, identifying modifiable factors like treatment attitude and insight.
Findings
Poor adherence was reported in 45.2% of patients.
Treatment attitude and insight were independently associated with lower odds of poor adherence.
Adverse effects were strongly correlated with adherence but not independently associated after adjustment.
Abstract
Background and objective Incomplete adherence to antipsychotics commonly undermines relapse prevention and functional stabilization in severe mental disorders. Published data from the Democratic Republic of the Congo (DRC) are scarce. The objective of this pilot study was to estimate antipsychotic medication adherence among outpatients attending the Centre Neuro Psycho Pathologique (CNPP) in Kinshasa, DRC, and examine associations with treatment attitude, insight, and adverse effects using brief standardized instruments. Methods We conducted a cross-sectional pilot study from April to May 2025 at the outpatient service of the CNPP in Kinshasa, DRC. Adult patients (≥18 years) who had been prescribed at least one antipsychotic for a minimum duration of five weeks were included. Adherence was measured with the 10-item Medication Adherence Rating Scale (MARS). Adherence was dichotomized…
Genes, proteins, chemicals, diseases, species, mutations and cell lines named across the full text — each resolved to its canonical identifier and authoritative record.
| Variable | Total (n=62) | W (Shapiro-Wilk) | p (Shapiro-wilk) |
| Age, mean ± SD (years) | 37.5 ± 14.4 | 0.909618 | 0.0002374649 |
| Gender male, n(%) | 37 (59.7) | - | - |
| Married, n(%) | 19 (30.6) | - | - |
| Years of study, median (IQR) | 13.0 (12.0-15.0) | 0.903025 | 0.0001328778 |
| Income-generating activity (yes), n(%) | 13 (21.0) | - | - |
| Diagnosis group: schizophrenia-spectrum / primary psychotic, n(%) | 42 (67.7) | - | - |
| Diagnosis group: substance-induced psychotic disorder, n(%) | 5 (8.1) | - | - |
| Diagnosis group: bipolar disorder, n(%) | 6 (9.7) | - | - |
| Diagnosis group: depressive disorder, n(%) | 4 (6.5) | - | - |
| Diagnosis group: neurocognitive disorder, n(%) | 4 (6.5) | - | - |
| Diagnosis group: other, n(%) | 1 (1.6) | - | - |
| Accompaniment, n(%) | 40 (64.5) | - | - |
| Antipsychotic type: atypical only, n(%) | 23 (37.1) | - | - |
| Antipsychotic type: typical only, n(%) | 28 (45.2) | - | - |
| Antipsychotic type: combination, n(%) | 11 (17.7) | - | - |
| Antipsychotic polypharmacy (≥2), n(%) | 16 (25.8) | - | - |
| Number of antipsychotics, median (IQR) | 1.0 (1.0-1.8) | 0.580401 | 4.880485x10-12 |
| Total number of medications, median (IQR) | 3.0 (2.0-3.0) | 0.909678 | 0.0002387270 |
| Haloperidol decanoate, n (%) | 10 (16.1) | - | - |
| EqCPZ, median (IQR) mg/day | 233.5 (100.0-500.0) | 0.876400 | 0.0000151979 |
| Previous hospitalizations, median (IQR) | 1.0 (0.0-1.0) | 0.826248 | 4.648403x10-7 |
| Duration of disorder, median (IQR) years | 3.0 (1.0-8.0) | 0.795280 | 7.196232x10-8 |
| Measure | Total (n=62) | W (Shapiro-Wilk) | p (Shapiro-Wilk) |
| MARS, Mean ± SD | 5.89 ± 3.04 | ||
| MARS, median (IQR) | 7.00(3.25-8.00) | 0.899140 | 0.0000952111 |
| Poor adherence, n/n (%) CI95% | 28/62 (45.2) 33.4-57.5 | ||
| DAI, mean ± SD | 2.52 ± 5.37 | 0.934995 | 0.0026947357 |
| DAI positive attitude, n(%) | 39 (62.9) | ||
| DAI neutral attitude, n(%) | 8 (12.9) | ||
| DAI negative attitude, n(%) | 15 (24.2) | ||
| BIS, mean ± SD | 5.80 ± 2.86 | 0.964524 | 0.0701181902 |
| BIS good insight, n(%) | 9 (14.5) | ||
| GASS, mean ± SD | 20.63 ± 20.84 | 0.744034 | 4.661350x10-9 |
| GASS absent/mild, n(%) | 47 (75.8) | ||
| GASS moderate, n(%) | 3 (4.8) | ||
| GASS severe, n(%) | 12 (19.4) |
| Scale | MARS ρ (p) | DAI | BIS | GASS | Cronbach α |
| MARS | - | 0.787 (p=0.000000000000033) | 0.616 (p=0.000000098260369) | -0.660 (p=0.000000005291995) | 0.830 |
| DAI | 0.787 (p=0.000000000000033) | - | 0.481 (p=0.000076313358278) | -0.570 (p=0.000001319544528) | 0.758 |
| BIS | 0.616 (p=0.000000098260369) | 0.481 (p=0.000076313358278) | - | -0.353 (p=0.004884424173122) | 0.730 |
| GASS | -0.660 (p=0.000000005291995) | -0.570 (p=0.000001319544528) | -0.353 (p=0.004884424173122) | - | 0.963 |
| Factor | Good adherence (n=34) | Poor adherence (n=28) | Test (df) | p-value |
| Gender male, n (%) | 15 (44.1) | 22 (78.6) | χ²(1)=7.57 | 0.006 |
| Married, n (%) | 12 (35.3) | 7 (25.0) | χ²(1)=0.77 | 0.382 |
| Years of study, median (IQR) | 15.0 (12.0-15.0) | 12.0 (12.0-15.2) | U=521 | 0.517 |
| Income-generating activity (yes), n (%) | 6 (17.6) | 7 (25.0) | χ²(1)=0.50 | 0.479 |
| Diagnosis group, n (%) | χ²(5)=7.05 | 0.217 | ||
| - Schizophrenia-spectrum/primary psychotic | 22 (64.7) | 20 (71.4) | ||
| - Substance-induced psychotic disorder | 4 (11.8) | 1 (3.6) | ||
| - Bipolar disorder | 2 (5.9) | 4 (14.3) | ||
| - Depressive disorder | 2 (5.9) | 2 (7.1) | ||
| - Neurocognitive disorder | 4 (11.8) | 0 (0.0) | ||
| - Other | 0 (0.0) | 1 (3.6) | ||
| Accompaniment (yes), n (%) | 18 (52.9) | 22 (78.6) | χ²(1)=4.41 | 0.036 |
| Antipsychotic type, n (%) | χ²(2)=3.64 | 0.162 | ||
| - Atypical only | 16 (47.1) | 7 (25.0) | ||
| - Typical only | 12 (35.3) | 16 (57.1) | ||
| - Combination (typical + atypical) | 6 (17.6) | 5 (17.9) | ||
| Antipsychotic polypharmacy (≥2), n(%) | 7 (20.6) | 9 (32.1) | χ²(1)=1.07 | 0.301 |
| Number of antipsychotics, median (IQR) | 1.0 (1.0-1.0) | 1.0 (1.0-2.0) | U=422 | 0.320 |
| Total number of medications, median (IQR) | 2.5 (2.0-3.0) | 3.0 (2.0-4.0) | U=347 | 0.059 |
| Haloperidol decanoate, n (%) | 8 (23.5) | 2 (7.1) | Fisher's exact | 0.097 |
| EqCPZ, median (IQR) mg/day | 175.0 (100.0-342.0) | 500.0 (175.0-525.0) | U=279 | 0.005 |
| Previous hospitalizations, median (IQR) | 1.0 (0.0-1.0) | 1.0 (1.0-2.0) | U=407 | 0.297 |
| Duration of disorder (years), median (IQR) | 3.0 (1.0-8.0) | 3.5 (1.0-9.0) | U=467 | 0.904 |
| DAI, mean ± SD | 5.88 ± 3.07 | -1.57 ± 4.69 | U=859.5 | 0.000000048263269 |
| BIS, mean ± SD | 7.29 ± 2.19 | 3.98 ± 2.53 | U=777 | 0.000020274546116 |
| GASS, mean ± SD | 9.68 ± 6.03 | 33.93 ± 24.53 | U=165.5 | 0.000011293272583 |
| DAI negative attitude, n (%) | 0 (0.0) | 15 (53.6) | Fisher's exact | 0.000000402375642 |
| GASS moderate/severe, n (%) | 2 (5.9) | 13 (46.4) | Fisher's exact | 0.000260738407633 |
| Factor | OR (95% CI) | Wald χ² (df = 1) | p-value | aOR (clinical/treatment) | Wald χ²(df = 1) | p-value | aOR (extended) | Wald χ²(df) = 1 | p-value |
| Gender, male | 4.64 (1.50-14.35) | 7.12 | 0.008 | 4.85 (1.20-19.56) | 4.93 | 0.026 | 3.97 (0.39-40.16) | 1.36 | 0.243 |
| Accompaniment (yes) | 3.26 (1.06-10.05) | 4.23 | 0.040 | 2.79 (0.64-12.09) | 1.89 | 0.170 | 0.24 (0.01-3.94) | 1.00 | 0.318 |
| Total medications (per +1) | 1.53 (0.97-2.42) | 3.42 | 0.064 | - | - | - | - | - | - |
| Total antipsychotics (per +1) | 1.57 (0.59-4.16) | 0.83 | 0.364 | - | - | - | - | - | - |
| Antipsychotic polypharmacy (>=2) | 1.83 (0.58-5.76) | 1.06 | 0.304 | 0.30 (0.03-2.55) | 1.22 | 0.269 | 0.28 (<0.01-17.87) | 0.36 | 0.548 |
| Antipsychotic type: Typical only (ref atypical) | 3.05 (0.95-9.74) | 3.54 | 0.060 | 2.34 (0.39-14.06) | 0.87 | 0.351 | 3.77 (0.22-65.00) | 0.84 | 0.361 |
| Antipsychotic type: Combination (ref atypical) | 1.90 (0.43-8.39) | 0.73 | 0.394 | 6.53 (0.38-112.93) | 1.67 | 0.197 | 0.07 (<0.01-65.50) | 0.59 | 0.444 |
| EqCPZ (per +100 mg) | 1.34 (1.06-1.70) | 5.89 | 0.015 | 1.42 (1.00-2.02) | 3.85 | 0.050 | 0.87 (0.50-1.51) | 0.25 | 0.614 |
| Haloperidol decanoate | 0.25 (0.05-1.29) | 2.74 | 0.098 | 0.03 (0.01-0.47) | 6.25 | 0.012 | 0.68 (0.01-39.45) | 0.03 | 0.854 |
| DAI (per 1-point increase) | 0.63 (0.51-0.79) | 16.86 | 4.0276e-05 | - | - | - | 0.54 (0.32-0.93) | 5.00 | 0.025 |
| BIS (per 1-point increase) | 0.57 (0.43-0.76) | 15.36 | 8.8914e-05 | - | - | - | 0.39 (0.17-0.87) | 5.29 | 0.022 |
| GASS (per 1-point increase) | 1.12 (1.03-1.22) | 7.29 | 0.006943 | - | - | - | 1.30 (0.96-1.78) | 2.84 | 0.092 |
Peer Reviews
No public reviews on file for this paper yet. If you reviewed it on a platform where reviews are public (OpenReview, ICLR, NeurIPS, ICML), you can paste yours below so the community can read it here.
Videos
No videos yet. Explain this paper in a talk, walkthrough, or lecture? Add one.
Taxonomy
TopicsMedication Adherence and Compliance · Schizophrenia research and treatment · Mental Health Treatment and Access
Introduction
Mental disorders rank among the most disabling health conditions globally due to their substantial and enduring contribution to years lived with disability across regions and income settings [1]. Antipsychotics are commonly used to treat mental disorders, on or off-label [2]. Antipsychotic maintenance treatment is an important factor to maintain stability in conditions like schizophrenia or bipolar disorders and to prevent relapse [3]. However, the clinical benefits noted in controlled settings are often lowered in standard care due to incomplete adherence, which is frequent and frequently hidden [4,5].
Poor adherence to antipsychotics is associated with worse outcomes like symptom exacerbations, relapse, emergency presentations, and rehospitalization [4]. It also contributes to more use of health care and higher costs [4-7]. Studies have shown that poor adherence is associated with factors such as poor insight, negative beliefs or attitudes toward medication, adverse effects, substance use, and service-related barriers [6,8]. These associated factors are important in low- and middle-income contexts, where limited medication options, out-of-pocket costs, and limited follow-up resources can lead to treatment discontinuation [9,10]. In sub-Saharan Africa, a meta-analysis estimated that approximately half of patients with schizophrenia were nonadherent to antipsychotic medication, with treatment-related adverse effects and negative attitudes toward medication among the repeatedly implicated correlates [10].
Three modifiable, patient-centered constructs warrant focused attention. First is treatment attitude, which summarizes the subjective appraisal of benefits, perceived necessity, and acceptability of antipsychotic treatment and has long been linked to medication-taking behavior [11]. Second is insight; it relates to illness awareness and attribution of symptoms and is repeatedly associated with adherence patterns in schizophrenia [6,12]. Third is the adverse effects that constitute a proximal and actionable barrier. Systematic elicitation of side effects improves clinical detection and may inform dose optimization and shared decision-making [13].
In the Democratic Republic of the Congo (DRC), available evidence highlights a large treatment gap and limited formal supply of mental health services in several settings, emphasizing the need for locally generated, pragmatic data to inform service-level interventions [10]. Against this background, the objective of this pilot study was to estimate antipsychotic medication adherence among outpatients attending the Centre Neuro Psycho Pathologique (CNPP) in Kinshasa and examine associations with treatment attitude, insight, and adverse effects using brief standardized instruments.
Materials and methods
Study design and setting
This cross-sectional pilot study was conducted at the outpatient service of the CNPP at the University of Kinshasa (Kinshasa, DRC). The protocol, including the data from this study, was submitted and approved by the Ethics Committee of the School of Public Health, University of Kinshasa (approval number: ESP/CE/21/25). All participants provided informed consent prior to inclusion. The study was designed and reported in accordance with the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) guidelines (see Appendix A) for cross-sectional studies [14].
Data were collected during routine ambulatory follow-up from April 1st to May 1st, 2025. The study was designed as a pilot to generate the first local estimates of antipsychotic adherence in this setting, quantify the distribution of key patient-reported determinants (treatment attitudes, insight, and side effect burden), and assess the feasibility of standardized data collection in a tertiary hospital. We used a pragmatic sample size constrained by the recruitment period. No formal sample-size calculation was made a priori because the primary objective was to assess the feasibility of standardized data collection and to generate preliminary estimates for future hypothesis-driven research.
The CNPP is the main public neuropsychiatric facility affiliated with the University of Kinshasa and is a tertiary referral center for the city-province of Kinshasa (estimated population ≥ 17 million) [15] and referred patients from other provinces of the DRC [16]. The outpatient services provide follow-up for patients discharged from inpatient services and those referred from primary and secondary facilities [17]. Because CNPP is a tertiary hospital, the case mix may be more clinically complex than in community or primary care settings. Therefore, the generalizability of the prevalence estimate and associated factors should be interpreted with caution.
Participants
Eligible participants were adults (≥ 18 years) followed in ambulatory care at CNPP and had been on an unchanged antipsychotic regimen for at least five weeks at the time of the survey. This threshold was chosen to ensure that participants had attained pharmacokinetic steady state for the majority of oral antipsychotics and that early-onset extrapyramidal and metabolic side effects would have manifested themselves. This lessens confounding from recent dose changes or transient adverse effects. Patients who could not complete the questionnaires (e.g., severe cognitive impairment, marked behavioral disorganization, or acute agitation at the visit) were excluded from the study. The sample was a pragmatic convenience series of consecutive eligible patients encountered during the study period. Data was collected by a senior psychiatrist assistant through a face-to-face interview. Recruitment took place on all consultation days during the study period. Each consecutive eligible participant was systematically approached. No formal screening log was maintained, but reasons for exclusion were documented (see Appendix B for participant exclusion criteria). The total number of patients attending the outpatient service during the recruitment period was not recorded.
Data sources and variables
Clinical and treatment-related information was obtained from the patient interview and medical records, including age, gender, marital status, education, income, activity, accompaniment to the visit, duration of the trouble, and DSM-5 diagnosis recorded in the file by the psychiatrist in charge as part of routine clinical assessment. No structured diagnostic interview was administered. Diagnostic inter-rater reliability was not formally evaluated for this pilot survey. This is a methodological limitation. Diagnoses were grouped for analytic purposes into the following categories: schizophrenia-spectrum and primary psychotic disorders (including schizophrenia, schizoaffective disorder, schizophreniform disorder, brief psychotic disorder, and chronic hallucinatory psychosis), substance-induced psychotic disorder, bipolar disorder, depressive disorder, neurocognitive disorder, and other.
Current treatment was recorded (molecule and dose). Antipsychotic polypharmacy was defined as the concurrent use of ≥2 antipsychotics. Exposure to long-acting injectable (LAI) haloperidol decanoate (the only LAI used) was coded present if listed among current medications. The total antipsychotic daily dose was expressed as chlorpromazine equivalent (EqCPZ) based on published conversion tables and summed when more than one was prescribed [18-21]. The prescribed daily dose was converted using the molecule-specific EqCPZ ratio (e.g., chlorpromazine 100 mg=100 mg EqCPZ; haloperidol 2 mg=100 mg EqCPZ; risperidone 2 mg=100 mg EqCPZ; olanzapine 5 mg=100 mg EqCPZ). For haloperidol decanoate, the administered dose and injection interval were converted to an average daily dose before applying the EqCPZ ratio [20]. The EqCPZ was analyzed per 100 mg increase (EqCPZ/100) to improve interpretability. These conversion ratios covered all antipsychotic molecules prescribed in the study sample (see Appendix C).
Measures
All questionnaires were administered in French. For MARS, a previously published French version was used. For the DAI-10, BIS-8, and GASS, French-language versions/translations of the original instruments were used for study administration in our setting. Each participant read and completed questionnaires independently. The interviewer was available to clarify any ambiguous items at the participant's request. The average completion time for the full survey was approximately 15 to 20 minutes. Prior to the study, the survey package was pre-tested in five outpatients to ensure comprehension and feasibility within routine visits. This pre-test served as a feasibility check for item comprehension and administration logistics rather than a formal cultural adaptation procedure. No wording changes were required.
Medication adherence
Medication adherence was measured with the 10-item Medication Adherence Rating Scale (MARS). The total score ranges from 0 to 10. Higher scores indicate better adherence [22,23]. Adherence was dichotomized as good adherence (MARS total ≥ 7) versus poor adherence (MARS < 7) [22,24]. We used the French translation of the MARS validation [25].
Treatment attitude
We assessed treatment attitude toward antipsychotics with the 10-item Drug Attitude Inventory (DAI). Items were scored +1/-1 and summed to a total score ranging from -10 to +10. Higher scores indicate a more favorable attitude toward medication [11]. The DAI was categorized as negative (<0), neutral (=0), or positive (>0). The questionnaire was administered using the French translation. No local validation study has been conducted in the DRC.
Insight
Insight was assessed with the 8-item Birchwood Insight Scale (BIS-8). Total score ranges from 0 to 12. Higher scores indicate better insight [12]. We used a ≥ 9 threshold to describe 'good insight' vs. 'poor insight' [26,27]. The BIS-8 was administered using the French translation. The DRC-specific validation data are unavailable.
Adverse effects were assessed using the Glasgow Antipsychotic Side-Effect Scale (GASS). It is a patient-rated checklist scored 0-3 per item and summed to a total score [13]. Side-effect severity was categorized as absent/mild (0-21), moderate (22-42), and severe (≥ 43) [28]. The GASS was administered using the French-translated version of the original English-language scale. Interpretation of absolute scores should consider the absence of local cultural adaptation data.
Statistical analysis
Analyses were performed using SPSS Statistics version 31 (IBM Corp., Armonk, NY, USA). All tests were two-sided with α = 0.05. We used the Shapiro-Wilk test [29] to check distributional assumptions. Continuous variables were summarized as mean ± standard deviation (SD) when normally distributed and as median (interquartile range, IQR) when skewed. Means were additionally provided for key psychometric totals to facilitate comparison with the literature. The prevalence of poor adherence was reported with 95% confidence intervals (CI).
Internal consistency of psychometric scales (MARS, DAI, BIS, and GASS) was evaluated using Cronbach’s alpha, acknowledging the tau-equivalence assumption. Corrected item-total correlations and alpha-if-item-deleted statistics were examined. We evaluated associations between continuous scale scores using Spearman’s rank correlation rho (ρ). For bivariable comparisons (good vs. poor adherence), categorical variables were compared using Pearson’s chi-square test or Fisher’s exact test when expected cell counts were small. Continuous variables were compared using Student’s t-test or the Mann-Whitney U test as appropriate.
We used univariable logistic regression to estimate crude odds ratios (ORs) for poor adherence. Multivariable logistic regression was conducted with parsimonious models to limit sparse-data bias, given the sample size. We specified a priori a 'clinical/treatment' model and included gender, accompaniment, antipsychotic type, polypharmacy, EqCPZ (per +100 mg), and haloperidol decanoate exposure (events per variable (EPV) = 4.0). The extended model added the psychometric predictors (DAI, BIS, and GASS total scores) with EPV = 2.8. Multicollinearity was assessed using variance inflation factors (VIF) (see Appendix D). Model fit and performance were summarized using the likelihood-ratio chi-square test, the Hosmer-Lemeshow goodness-of-fit test [30], and the area under the receiver operating characteristic curve (AUC). The AUC is reported as the apparent (in-sample) value and as an optimism-corrected estimate based on 500 bootstrap resamples. Missing values were handled using complete-case analysis for each model. Key psychometric measures and the adherence outcome had no missing data. No missing data were observed for sociodemographic, clinical, or treatment variables. Sensitivity analysis using Firth’s penalized (bias-reduced) logistic regression for the clinical/treatment and extended models was performed, given sparse cells and quasi-complete separation for some predictors.
Results
Participants and baseline characteristics
Table 1 shows the participants’ sociodemographic, clinical, and antipsychotic treatment characteristics. A total of 62 outpatients receiving antipsychotic treatment were included. The mean age was 37.5 ± 14.4 years; 37 participants were male (59.7%), and 40 patients (64.5%) who attended the visit were accompanied. Participant diagnoses were schizophrenia-spectrum/psychotic disorders in 42/62 (67.7%), mood disorders in 10/62 (16.1%), neurocognitive disorders in 4/62 (6.5%), and other diagnoses in 6/62 (9.7%). Of the included patients, 23/62 (37.1%) received atypical antipsychotics only, 28/62 (45.2%) received typical antipsychotics only, and 11/62 (17.7%) received a combination of typical and atypical antipsychotics; 16 patients (25.8%) had polypharmacy. Haloperidol decanoate was prescribed to 10 patients (16.1%). The median total antipsychotic exposure was 233.5 mg EqCPZ per day (IQR 100.0-500.0). The median duration of the disorder was 3.0 years (IQR 1.0-8.0), and the median of prior hospitalizations was 1.0 (IQR 0.0-1.0).
Adherence and psychometric measures
The mean MARS score was 5.89 ± 3.04 (median 7.0; IQR 3.25-8.0). Poor adherence was observed in 28/62 patients (45.2%; 95% CI 33.4-57.5). Overall mean scores were 2.52 ± 5.37 for DAI, 5.80 ± 2.86 for BIS, and 20.63 ± 20.84 for GASS. A positive treatment attitude was present in 39/62 patients (62.9%), neutral in eight/62 (12.9%), and negative in 15/62 (24.2%). Good insight was identified in nine/62 patients (14.5 %). Side-effect burden was classified as absent/mild in 47/62 (75.8%), moderate in three/62 (4.8%), and severe in 12/62 (19.4%). Table 2 presents adherence outcomes and psychometric scores. Internal consistency was acceptable to excellent; Cronbach’s α = 0.830 (MARS), 0.758 (DAI), 0.730 (BIS), and 0.963 (GASS). See Table 3 for correlations across scales and reliability coefficients.
**Table 3: Correlations between adherence, treatment attitude, insight, and side effects, and internal consistency of scalesOff-diagonal cells are Spearman’s correlations and the exact two-sided p-value. Cronbach’s alpha values are reported in the sixth column. All correlations are significant at p<0.05. The complete
We examined the item-total statistics for each of the four scales. For MARS (α = 0.830), the corrected item-total correlations ranged from 0.345 (item 6) to 0.707 (item 4), while the alpha-if-item-deleted values ranged from 0.796 to 0.832. Removing any one item would not appreciably raise the overall alpha. The corrected item-total correlations for DAI (α = 0.758) ranged from 0.245 (item 10) to 0.579 (item 9), and the alpha-if-item-deleted values ranged from 0.715 to 0.763. The corrected item-total correlations for BIS (α = 0.730) were between 0.136 (item 2) and 0.663 (item 6), and the alpha-if-item-deleted values were between 0.648 and 0.752. If item 2 ('my doctor is right in prescribing medication for me') were taken out, alpha would increase up to 0.752. The item was maintained, nevertheless, to ensure the content was valid and could be compared to past BIS investigations.
It's important to note that the GASS has a very high alpha (α = 0.963, 23 items). The corrected item-total correlations ranged from -0.094 (item 21, which questions about sexual function) to 0.924 (item 18). Twenty-two out of twenty-three items had adjusted item-total correlations of more than 0.40, indicating a strong association among the items. Item 21 had a negative adjusted item-total correlation (r = -0.094), and removing it would raise alpha to 0.969.
Bivariable comparisons by adherence status
Table 4 compares clinical and treatment factors by adherence status. Compared with adherent patients, those with poor adherence were more frequently men (22/28 vs. 15/34; p = 0.006) and more often accompanied to the visit (22/28 vs. 18/34; p = 0.036). They had higher antipsychotic dose exposure (EqCPZ median 500.0 vs 175.0; p = 0.005). Haloperidol decanoate was less common among poorly adherent patients, but not significant (Fisher’s p = 0.097). Polypharmacy was not associated with adherence status (p = 0.301).
Table 4: Bivariable comparisons by adherence status (good vs. poor adherence)Data Percentages are column percentages and may not sum to 100 due to rounding. Good adherence was defined as a MARS total score ≥7; poor adherence as MARS <7.EqCPZ: Chlorpromazine equivalents (mg/day), DAI: Drug Attitude Inventory, BIS: Birchwood Insight Scale, GASS: Glasgow Antipsychotic Side-effect Scale, MARS: Medication Adherence Rating Scale, df: Degrees of freedom, SD: Standard deviation, IQR: Interquartile range
Psychometric differences were marked. Poor adherent patients had lower DAI scores (-1.57 ± 4.69 vs. 5.88 ± 3.07; p < 0.001), lower BIS scores (3.98 ± 2.53 vs. 7.29 ± 2.19; p < 0.001), and higher GASS scores (33.93 ± 24.53 vs. 9.68 ± 6.03; p < 0.001). All participants with a negative treatment attitude were in the poor adherence group (15/15). Severe adverse effects occurred exclusively in the poor adherence group (12/12). When moderate and severe categories were combined, poor adherence was present in 13/15 vs. 15/47 in the absent/mild group (Fisher’s p < 0.001).
Correlations between adherence, attitude, insight, and adverse effects
The MARS total score correlated strongly with DAI (Spearman ρ = 0.787; p < 0.001) and moderately with BIS (ρ = 0.616; p < 0.001). The MARS correlated inversely with GASS (ρ = -0.660; p < 0.001). The DAI correlated positively with BIS (ρ = 0.481; p < 0.001) and negatively with GASS (ρ = -0.570; p < 0.001). The BIS also correlated negatively with GASS (ρ = -0.353; p = 0.005).
Logistic regression analyses for poor adherence (MARS < 7)
Table 5 presents univariable and multivariable logistic regression analyses of factors associated with poor adherence. In univariable logistic regression, poor adherence was associated with male gender (OR 4.64; 95% CI 1.50-14.35; p = 0.008), being accompanied (OR 3.26; 95% CI 1.06-10.05; p = 0.040), higher EqCPZ (per 100 mg: OR 1.34; 95% CI 1.06-1.70; p = 0.015), lower BIS (per 1-point increase: OR 0.57; 95% CI 0.43-0.76; p < 0.001), and higher GASS (per 1-point increase: OR 1.12; 95% CI 1.03-1.22; p = 0.007).
**Table 5: Logistic regression analyses for poor adherence (MARS < 7)The clinical/treatment model included gender, accompaniment, antipsychotic type (reference: atypical only), polypharmacy, EqCPZ (per +100 mg/day), and haloperidol decanoate EPV ≈4.0). Model fit: likelihood-ratio χ²(7)=25.07 (p=0.00074); Hosmer-Lemeshow χ²(8)=7.72 (p=0.461). Discrimination: apparent AUC=0.842 (95% bootstrap CI 0.733-0.935); optimism-corrected AUC=0.770 (bootstrap, 500 resamples). Multicollinearity was low (max VIF=2.75).The extended model added DAI, BIS, and GASS totals (EPV ≈2.8). Model fit: likelihood-ratio χ²(10)=56.20 (p<0.001); Hosmer-Lemeshow χ²(8)=5.65 (p=0.687). Discrimination: apparent AUC=0.962 (95% bootstrap CI 0.910-0.998); optimism-corrected AUC=0.911 (bootstrap, 500 resamples). Multicollinearity remained accep
In the multivariable clinical/treatment model, male gender (adjusted odds ratio (aOR) 4.85 (1.20-19.56); p = 0.026) remained independently associated with poor adherence. Haloperidol decanoate (aOR 0.03 (<0.01-0.47); p = 0.012) became significant, while EqCPZ showed a borderline association (aOR 1.42 (1.00-2.02); p = 0.050). In the extended model adding DAI, BIS, and GASS, both DAI (aOR 0.54 (0.32-0.93); p = 0.025) and BIS (aOR 0.39 (0.17-0.87); p = 0.022) remained independently associated with poor adherence. GASS showed a non-significant association (aOR 1.30 (0.96-1.78); p = 0.092).
Sensitivity analysis using Firth’s penalized logistic regression yielded similar conclusions for the psychometric measures (see Appendix F). For DAI (per +1 point), the Firth's OR was 0.65 (95% CI 0.53-0.80; p<0.001). For BIS (per +1 point), the Firth's OR was 0.59 (95% CI 0.45-0.77; p<0.001). For GASS (per +1 point), the Firth's OR was 1.10 (95% CI 1.03-1.17; p=0.004). These estimates were slightly less than the null, but the direction and inference were unchanged. In the Firth clinical/treatment model, the haloperidol decanoate estimate was attenuated (Firth's aOR 0.07; 95% CI 0.001-0.72; p = 0.026) in comparison to the standard maximum-likelihood estimate (aOR 0.03). This supports a cautious interpretation of this association.
Discussion
In this study, nearly one in two participants met criteria for poor antipsychotic adherence using the MARS cut-off (MARS < 7). This magnitude is broadly consistent with the persistent adherence gap described in schizophrenia-spectrum disorders across settings, where nonadherence is common and clinically consequential despite the established efficacy of antipsychotics for relapse prevention [3-6,8]. Given that adherence was measured by self-report, the observed prevalence should be interpreted as a conservative estimate rather than an upper bound due to social desirability and recall effects of self-reporting [22,23].
Interpretation of the main associations
The most informative correlates clustered around modifiable, patient-centered dimensions: treatment attitude (DAI), insight (BIS), and adverse-effect burden (GASS). In the extended multivariable model, both DAI and BIS remained independently associated with adherence status. This pattern aligns with the conceptual view that adherence behaviors are strongly driven by subjective appraisal of medication (perceived benefits, acceptability, and necessity) and by illness-related awareness and attribution processes [6,8,11,12]. The DAI was developed to identify the experiential and attitudinal factors that influence daily medication-taking decisions. The results of this study support its clinical usefulness as a brief indicator for adherence vulnerability [11].
Insight showed that adherent patients reported elevated BIS scores. The BIS maintained an independent association after adjustment for clinical and treatment covariates. Previous research showed that impaired insight was associated with discontinuation and inconsistent use of antipsychotics, partially due to reduced perceived need for treatment and reduced engagement in follow-up care [6,8,12]. This supports focusing on interventions that improve shared decision-making and collaborative illness understanding, instead of just changing medications or controlling symptoms.
Adverse-effect burden (GASS) was higher in the poor-adherence group (see Appendix G). It showed a negative correlation with MARS. Side effects are frequently cited as a reason for skipping doses or discontinuing treatment [5,8]. They lower the quality of life, functioning, and the willingness to continue with treatment [5,6,8,13]. In the fully adjusted model, the GASS association became not significant. This may be consistent with a mediation model in the sense that adverse effects may indirectly affect adherence by influencing treatment attitude and acceptability, rather than functioning as an entirely independent variable. Cross-sectional data cannot resolve directionality. But the observed configuration is hypothetically compatible with a pathway in which adverse effects may be associated with worse medication attitudes and, thereby, lower adherence [5,11,13].
The association between being accompanied to the visit and poor adherence should be interpreted carefully. In routine ambulatory care, accompaniment may serve as an indicator of increased clinical severity, cognitive or functional impairment, or family concern arising from recent decompensation [31]. These factors could be upstream determinants of both accompaniment and nonadherence. Reverse causality is also possible: patients who are nonadherent and unstable may be more likely to be accompanied by relatives to facilitate access to care [32]. Additionally, a selection bias may operate: some patients who would otherwise have missed their appointment may only attend because they are accompanied, leading to an over-representation of poorly adherent patients among accompanied attendees. Because symptom severity and functional status were not measured, residual confounding cannot be excluded.
Dose, antipsychotic class, and long-acting formulations
Higher antipsychotic dose exposure (EqCPZ) was associated with poor adherence in univariable analyses and remained borderline in the clinical/treatment model, suggesting that dose burden may contribute to nonadherence. Antipsychotic type and polypharmacy showed no associations once clinical covariates and psychometric measures were considered. It is important to recognize that the EqCPZ has limitations because it compresses different pharmacodynamic profiles into a single metric [18,19]. Still, the association between higher EqCPZ and higher adverse-effect burden in poor-adherent patients supports a pragmatic implication: adherence-oriented care should focus on regular reviews of the minimum effective dose, simplification of treatment plans, and structured adverse-effect monitoring [13,19].
Exposure to haloperidol decanoate showed a protective association in the clinical model, with an extreme adjusted OR and a very wide CI, which may indicate instability. This pattern aligns with sparse-cell bias and potential quasi-separation in a small sample. In Firth penalized sensitivity analyses (Appendix F), the association’s direction was similar. The magnitude was less extreme. This supports cautious interpretation. The LAI antipsychotics can reduce the burden of daily pill-taking, provide clearer signals of treatment interruption, and improve continuity of taking medication. However, the effect varies depending on the design and context [3,5,6]. In this study, interpretation must be cautious. Depot use may be preferentially prescribed to patients perceived as high risk for poor adherence or, on the contrary, to those who are open to continuing treatment. The attenuation of the depot association in the extended model hypothesizes that attitude and insight may partially explain who receives and benefits from a long-acting formulation [11,12]. This finding is hypothesis-generating and should be interpreted cautiously, given the small number of participants receiving LAI haloperidol.
Limitations
Several limitations temper inference. First, the cross-sectional design precludes causal interpretation and does not establish temporal ordering between adverse effects, attitudes, insight, and adherence. Observed associations may reflect reverse causality or bidirectional relationships. Second, the sample was small, monocentric, and recruited by convenience in ambulatory care. This limits external validity. Third, adherence, attitude, insight, and adverse effects were self-reported and may have led to common-method bias and social desirability effects. A Harman single-factor exploration did not suggest the predominance of a single factor. But bias cannot be excluded [24]. Fourth, multivariable analyses were constrained by sparse data (28 outcome events), resulting in low EPV and unstable estimates for some predictors, including extensive CI and possible quasi-separation (particularly for haloperidol decanoate). The low EPV ratios (4.0 for the clinical/treatment model and 2.8 for the extended model) imply that parameter estimations should be seen as exploratory. This is shown by the very wide confidence intervals for several factors (e.g., haloperidol decanoate), which reflect instability. Moreover, some potentially significant confounders (such as symptom severity, substance use, therapeutic relationship, and treatment affordability) were not assessed and therefore could not be integrated into the current analysis. Finally, questionnaires were administered in French, and none have been formally validated in the DRC. Although the pre-test in five patients confirmed item comprehension, this does not substitute for a formal cross-cultural validation process. This should be considered when interpreting absolute scale values.
Conclusions
In a care setting facing major structural constraints, adherence interventions need to be feasible, low-cost, and closely integrated into routine follow-up. Although the cross-sectional design of this pilot study precludes causal conclusions, the observed associations suggest a clinical approach that focuses on quick standardized screening and rapid corrective action. This includes systematically asking about side effects using a structured tool, talking to the patient about their medication beliefs to guide counseling, and using psychoeducation methods and engagement that are adapted to the patient’s explanatory model and family context. This method fits with the idea that a combination of strategies is better than just one-on-one counseling for dealing with the many factors that affect adherence. In addition, if possible, using LAI for patients who have frequent breaks may help keep exposure stable over time. Larger longitudinal studies that use objective adherence measures and take into account other important contextual factors like substance use, symptom severity, affordability and continuity of supply, transportation barriers, and therapeutic alliance are needed to examine these associations and then guide targeted interventions in the DRC.
The reference list from the paper itself. Each links out to its DOI / PubMed record.
- 1Global, regional, and national burden of 12 mental disorders in 204 countries and territories, 1990-2019: a systematic analysis for the Global Burden of Disease Study 2019 Lancet Psychiatry GBD 2019 Mental Disorders Collaborators 137150920223502613910.1016/S 2215-0366(21)00395-3PMC 8776563 · doi ↗ · pubmed ↗
- 2Antipsychotic drugs: a concise review of history, classification, indications, mechanism, efficacy, side effects, dosing, and clinical application Am J Psychiatry Leucht S Priller J Davis JM 86587818120243935061410.1176/appi.ajp.20240738 · doi ↗ · pubmed ↗
- 3Oral and long-acting antipsychotics for relapse prevention in schizophrenia-spectrum disorders: a network meta-analysis of 92 randomized trials including 22,645 participants World Psychiatry Ostuzzi G Bertolini F Tedeschi F 2953072120223552462010.1002/wps.20972 PMC 9077618 · doi ↗ · pubmed ↗
- 4Prevalence of and risk factors for medication nonadherence in patients with schizophrenia: a comprehensive review of recent literature J Clin Psychiatry Lacro JP Dunn LB Dolder CR Leckband SG Jeste DV 8929096320021241659910.4088/jcp.v 63n 1007 · doi ↗ · pubmed ↗
- 5Nonadherence with antipsychotic medication in schizophrenia: challenges and management strategies Patient Relat Outcome Meas Haddad PM Brain C Scott J 4362520142506134210.2147/PROM.S 42735 PMC 4085309 · doi ↗ · pubmed ↗
- 6Medication adherence in schizophrenia: factors influencing adherence and consequences of nonadherence, a systematic literature review Ther Adv Psychopharmacol Higashi K Medic G Littlewood KJ Diez T Granström O De Hert M 200218320132416769310.1177/2045125312474019 PMC 3805432 · doi ↗ · pubmed ↗
- 7Systematic review of the economic aspects of nonadherence to antipsychotic medication in patients with schizophrenia Patient Prefer Adherence Dilla T Ciudad A Alvarez M 275284720132358968110.2147/PPA.S 41609 PMC 3622425 · doi ↗ · pubmed ↗
- 8A systematic review of factors influencing adherence to antipsychotic medication in schizophrenia-spectrum disorders Psychiatry Res Sendt KV Tracy DK Bhattacharyya S 143022520152546622710.1016/j.psychres.2014.11.002 · doi ↗ · pubmed ↗