Factors associated with self-rated health in primary care in the South-Western health zone of Malawi

Data were collected from patients who sought health care services in 12 health government-owned facilities located in three districts of the south-western health zone of Malawi, namely: Blantyre, Neno, and Thyolo. Blantyre and Neno were purposively chosen due to their unique characteristics of having an urban population (Blantyre), and the highest per capita health spending in the south-western zone respectively (Neno). Thyolo was randomly selected from the remaining five districts in the zone, which in addition to Thyolo include; Nsanje, Chikhwawa, Mwanza, and Chiradzulu. All government-owned primary care clinics and hospitals in these districts are in the first and second levels of the three tier health system respectively, and they operate as walk-in clinics that offer a standard essential health package with additional secondary care services being offered at the hospitals to patients for whom such is necessary. All health services in government owned facilities are offered free for all without any formal consultation lists. Thus, there is no financial barrier to accessing care, and all conditions beyond the scope of primary care services need to be referred to hospitals that offer secondary or tertiary health care.

Neno is the smallest of the three districts with a total population of less than 138,000, and a population density of 89 people per square kilometre in contrast with Blantyre and Thyolo, which have populations and population densities over 700,000 and 400 people per square kilometres respectively [24]. Over three-quarters of residents over the age of five years in Thyolo (79%) and Neno (77%) have no formal educational qualifications in contrast to 48% in Blantyre, which also has the highest proportion of people with tertiary education qualifications (7%) in contrast to less than 1% in the other districts [24].

The study population comprised of all adults of at least 18 years of age from the catchment area of the respective facilities who primarily sought care at the facility for at least two years and had visited the facility at least three times. The participants also needed to be capable of independently giving informed consent and answering the questions in the study questionnaire. Potential participants were excluded from the study if they could not give informed consent or if they were in urgent need of medical care at the time of data collection. The participant selection procedure has been detailed elsewhere  [20]: briefly, each data collector daily interviewed 12 randomly selected patients recruited in outpatient primary care clinics throughout the study period. Participants were recruited at a constant pre-calculated rate determined by the expected duration of each interview, average primary care consultation time, and the number of patients already present at the clinic at the beginning of the day shift. This approach ensured that none of the study participants got inconvenienced by their participation in the study.

As shown in Table 1, 1001 eligible people were approached, and 962 (96.1%) were recruited into the study. Most of the participants were female (64.0%; n = 616), aged between 18 and 45 years (82.1%; n = 790), or residents of a rural area (68.7%; n = 661). Eighty-eight percent (n = 854) of participants had attained some formal education, with 52.8% (n = 508) having attained primary school education only, and 5.2% (n = 50) having some tertiary education. Thus, the overall educational status and age distribution of the participants approximated published findings from other community-based studies in Malawi and our estimated general distribution of the 2018 population census figures. [24, 26]. Disabilities were self-reported among 11.6% (n = 112) participants only. Most participants sought care at a primary care clinic (70.6%; n = 679), had a primary affiliation to the study facility (94.8%; n = 912), had been affiliated with the same facility for at least four years (72.8%; n = 701), had sought care at a health facility more than 5 times in the immediate past 24 months (549; 57.1%).

Neno district had the highest proportion of participants who, (I) had more than five primary care visits in the immediate past 24 months 64.0%; versus 61.5% & 44.9% in Thyolo & Blantyre respectively; χ² p-value < 0.001), (II) perceived the treatment they received to be of good quality (60.7% versus 52.2% & 41.5%; χ² p-value < 0.001), and (III) had to travel for more than an hour to access primary care (60.1% versus 46.7% & 16.6%; χ² p-value < 0.001).

Three hundred and forty-one participants (35.5%) reported having poor SRH. Statistically significant higher prevalence of poor SRH was present among participants who; were female (38.8% versus 29.5%; p = 0.004), disabled (50.9% versus 33.4%; p < 0.001), sought care at clinics as compared to hospitals that offer primary care (43.1% versus 32.3%; p = 0.001) or had more than five primary care visits in the immediate past 24 months (40.3% versus 29.1%; p < 0.001). Participants who perceived the treatment provided at the index facility to be of poor quality were also more likely to have poor SRH (40.1% versus 31.0%; p = 0.003).

Prevalence of poor SRH increased with age such that 50% (95% CI: 42.3—57.7) of participants over the age of 45 had poor SRH compared to 38.0% (95% CI: 32.8—43.4) and 27.9% (95% CI: 23.8—32.3) among participants aged 31–45 years and 18–30 years respectively. Poor SRH, however, exhibited an inverse relationship with the participants’ highest attained education such that 59.3% (95% CI: 49.4—68.6) of participants with no formal education had poor SRH compared to 28.6% among those who had attained secondary or tertiary education (Table 1).

The mean total primary care score was 16.1 (95% CI: 15.87—16.33), and scores suggestive of good quality care (scores ≥ 21) were computed from 10% (n = 96) of the participants only. There was no statistically significant difference in the overall scores computed across the SRH divide (p-value = 0.84).

Table 1 demonstrates how demographic characteristics and level of perceived quality of health care related to SRH in unadjusted bivariate analyses. Poor SRH varied significantly between the different districts, Sex, age of participants, level of highest attained education, and presence of disability. Higher odds of poor SRH were also recorded for the following categories of binary variables: (I) those who visited hospitals (OR: 1.59; p-value = 0.002), (II) those who received treatment they perceived to be of poor quality (OR: 1.49, p-value = 0.003), and (III) those who sought care more than 5 times within the immediate past 24 months (OR: 1.65; p-value < 0.001). Participants rating increased level of first contact access, and comprehensiveness of services available also had statistically significant positive associations with the odds of poor SRH.

In the fully adjusted model (Table 1 and Fig. 1), participants who were from Neno district (OR: 1.61; 95% CI: 1.08—2.38), female (OR: 1.64; 95% CI: 1.21—2.23), or with disabilities (OR: 2.14; 95% CI: 1.39—3.3) had statistically significant higher odds of poor SRH than their counterparts. The odds of poor SRH increased with age with the highest odds of poor SRH being recorded in the oldest group (OR: 2.20; 95% CI: 1.45—3.33). Poor SRH also had an inverse relationship with the education of participants, such that the odds of poor SRH were highest among those with no formal education (OR: 2.51; 95% CI: 1.5—4.2). Participants who sought care more than five times in the immediate past 24 months had 1.57 higher odds of poor SRH (95% CI: 1.17—2.11) whereas those who received treatment they perceived as poor had 1.68 higher odds of poor SRH (95% CI: 1.26—2.26). Relational continuity was the only PCAT-MW dimension that had a statistically significant inverse association with poor SRH in the adjusted model with the odds of poor SRH decreasing by a factor of 0.82 (95% CI: 0.68—0.98) for every unit increase in the domain score.

To the best of our knowledge, this is the first study on self-rated health in the south-western health zone of Malawi. Although the prevalence of poor SRH is lower than those of other Sub-Saharan African (SSA) countries with higher per capita expenditure on health, and comparable to findings from high-income settings it ought to be considered as suboptimal and a cause for concern [12‐14, 22, 23, 27]. The latter is due to Malawi’s good SRH prevalence that is below the continental mean [21]. Resource constraints, growing double burden of communicable and non-communicable diseases, increasing life expectancy, and declining mortality from conditions that once carried high mortality also portend further declines in prevalence of good SRH [26, 28]. Since SRH is a stable construct, increasing life expectancy and declining HIV-associated mortality also portend a higher burden of poor SRH [29, 30].

The prevalence of poor SRH also highlights the importance of looking beyond the provision of freely accessible care, as is the case in Malawi, but also good quality patient-centred care to improve clinical outcomes and patients’ confidence in the health system [31].

The pattern of disparity in SRH outcomes across the study sites was consistent with findings from other studies, which report better SRH among urban residents and participants with a better socioeconomic context [22, 23, 32, 33]. Contrary to our expectations, the odds of poor SRH were highest in Neno despite Neno having the highest per capita expenditure on health, the highest proportion of participants who considered the treatment they received to be of good quality and being anecdotally reported as a popular health tourism destination for residents from neighbouring districts. Neno also has integrated care programmes, which are unique to the district, and are credited to have resulted in increased case finding and uptake of various primary care services including chronic infectious & non-infectious diseases, with consequential improvements in survival rates [34‐36].

This SRH pattern suggests that the aforementioned qualities of primary care in Neno, though necessary, might be insufficient in efforts to improve SRH trajectories of communities such as in Neno. The pattern is, however, consistent with the “paradox of health” observed by Barsky, where excellent health, in the presence of (I) advanced medical care, (II) heightened consciousness of health, (III) commercialization of health, and (IV) medicalization of daily life, is associated with poor subjective health [37]. The relative importance of each of the elements in this complex web of factors in Malawi is not yet known, but these elements highlight the overall importance of patients’ social context in their overall health and SRH trajectories. However, Barsky points at a paradoxical relationship that may be universal: health care succeeding at combating disease and mortality will not necessarily improve the perceptions of health of patients and populations.

The finding of female sex as an independent factor associated with poor SRH, though consistent with literature from other settings, is counterintuitive because (I) Malawi provides free health care for all, and (II) men in Malawi have a greater burden of disease and lower life expectancy [32, 38, 39]. On the other hand, SRH is often associated with social, contextual factors and subjective wellbeing [30]. This disparity, which has also been demonstrated among adolescents in other studies, can be attributed, at least in part to the male–female health survival paradox that may result from community views on health and masculinity [40‐43]. These result in women having more contact with the health system, having more diagnoses of conditions that are often non-lethal, and potentially having more knowledge about their health [39, 43]. Consequently, women may be more responsive to changes in their bodies and may, more often than men, factor these when asked to rate their health. Other studies have also attributed Sex disparities in SRH, at least in part, to differences in stress and social determinants of health [38, 40].

Relational continuity is independently associated with a reduced cost of care, improvements in uptake of preventive care, adherence to treatment, patient satisfaction with care, and health outcomes [44]. It is a surrogate marker of patients’ trust and satisfaction, and its importance is expected to increase as the burden of chronic diseases grows [44, 45]. Relational continuity is not an innate aspect of Malawi’s approach to primary care. Based on the personal experience of the authors (SK and LD), relational continuity in Malawian public primary care facilities is often a consequence of patient preference, healthcare worker shortages, and the presence of chronic diseases (e.g. HIV and diabetes). The presence of statistically significant associations between relational continuity and better SRH, albeit without systemic efforts to institute the same, suggests that relational continuity is a crucial and cost-effective ingredient for improving the quality of primary care and SRH in the study communities.

Participants’ age, presence of disabilities, and highest attained education are markers of social determinants of health. Marginalised populations (e.g. the disabled and people of low socioeconomic status) and the elderly are likely to be the greatest victims of these factors since they tend to experience socio-economic exclusion, reduced access to care, and numerous unique healthcare needs. Thus, the pillars of universal health coverage seem indispensable in the quest to improve SRH. The highest attained education of a participant is a valid measure of socioeconomic status and is probably associated with health knowledge and perception of self-efficacy [46].

The absence of a statistically significant association between comprehensiveness of services and SRH was another unexpected finding, which is probably a consequence of the uniformity in the range of services provided as part of the essential health package in Malawian primary care facilities.

Our study has several limitations. The most important is the cross-sectional design where we cannot establish causal relationships between any of the factors identified and SRH. Another limitation is in the exploration of patient’s subjective reports. However, procedural strategies, namely: (I) explaining the purpose of the study to participants, (II) use of unambiguous Likert scale labels, and (III) using a mixture of response approaches, were used to minimise recall and desirability tendencies on the part of participants. All questions were asked in the vernacular language with clear wording that is consistent with similar international tools.

Although the tool is validated for use in our setting, the use of the tool in a clinical setting may have made some of the PCAT domains, especially “first-contact access”, prone to selection bias. Data on the objective health status of the participants, except self-reported disability, was not collected. Thus, the effect of various pathological processes, especially mental health conditions, which may affect one's perception of the quality of care received, were not factored in the analysis. We also recognise the existence of the possibility that the findings we have attributed to the health paradox may be due to confounders not adjusted for in this study.