Quality of primary care from patients’ perspective: a cross sectional study of outpatients’ experience in public health facilities in rural Malawi

Within primary health care research, the US Primary Care Assessment Tool (PCAT) has been widely adapted and used in patient surveys in many countries [21‐26]. Based on the 1994 American Institute of Medicine’s definition of primary care [27] the PCAT aims at a global assessment of primary care organization and its performance in the core dimensions of accessibility, comprehensiveness, coordination and continuity, and accountability. In addition, it also assesses derivative dimensions of family orientation, community orientation, and cultural competence.

The development and validation of the Malawian version of the primary care assessment tool (PCAT-Mw) has been documented in another paper [28]. The tool has 29 items (Table 1) measuring primary care performance in seven dimensions: first contact access (3 items), communication continuity of care (4 items), relational continuity of care (4 items), coordination (3 items), comprehensiveness of services available (6 items), comprehensiveness of services provided (6 items) and community orientation (3 items). First contact access is here defined as the manner in which services are organized to accommodate access whenever needed and ensure patient satisfaction. Continuity of care entails the existence of a regular source of care and the longitudinal relationship between primary care providers and patients, in terms of accommodation of patient’s needs and preferences, such as communication and respect for patients. Coordination of care reflects the ability of primary care providers to facilitate and support patients to navigate use of other levels of health care when needed. Comprehensiveness of primary care services represents the range of services available in primary care to meet patients’ health care needs. A distinction is made between services that are available and those that are actually provided. Community orientation defines the extent to which the primary care providers understand and address priority health problems in a particular community with evidence of community participation.

Items are scored on a 4-point Likert scale, with 1 indicating “definitely not,” 2 indicating “probably not,” 3 representing “probably,” and 4 representing “definitely.” For consistency with methods used in PCAT studies in other countries, a mid-scale value of 2.5 is assigned to “not sure” answers while the mean item score is used for missing data [22‐24]. Additionally, there are questions to identify the usual primary care facility the patient uses and the patient’s sociodemographic data. This paper excludes the 3 coordination items because insufficient number of patients had been referred for secondary level care.

A face to face administered cross sectional study was carried out in August –September, 2016 in outpatient clinics of ten facilities – the two hospitals and eight health centers in Neno district. Facilities were selected purposefully to include all the public health facilities in the district. One of the faith-based health centers was included as it had signed a memorandum of understanding with the authorities to remove the user fees and run as a public facility. Patients were at least 18 years of age, must have used the facility for at least six months and must have visited the facility for at least 3 times. Acutely ill, frail looking or severe mental health patients were excluded in order to allow them to receive urgent medical attention. As this study’s data collection was part of the validation of the PCAT-Mw through metric analyses, sample size was calculated based on similar studies using at least 5:1 subject to item ratio [22‐26]. Sample size of 600 was targeted, 60 from each facility.

Six interviewers were trained to conduct the PCAT-Mw survey. A pilot study showed that the questionnaire would take about 45 min to administer. Each interviewer was therefore expected to interview seven patients per day. The sampling frame was 50–60 patients waiting to be seen on each working day. Sampling interval (n) was calculated by dividing the number of waiting patients by seven. A random starting point was obtained using a smart phone random number generator. Each ‘n^th’ patient was then asked for consent to participate in the study.

Independent variables were sex, age, education, geographical location, duration of contact with facility, reason for attending: chronic or acute condition, distance to facility measured through time taken to walk to the facility, cost of travel to the facility, waiting time, individual health facility affiliation and self- rated health status.

Data were entered into and analyzed using the IBM SPSS Statistics 24.0.0 (2016) package. Dimension mean scores were derived by dividing the sum of the item means by the number of items in the dimension. A score ≥ 3 was considered ‘acceptable to good performance’ and < 3 as ‘poor performance’. [28, 29] Total primary care was calculated as the sum of all dimension means. Sociodemographic, health care and health characteristics of the patients were compared between sexes by performing cross table analyses with chi squared significance testing to highlight differences between male and female patients.

A total of 649 patients were approached and 18 (2.8%) declined to participate in the study. This paper presents results from 631 completed questionnaires. Missing data accounted for approximately 1.9% of all data. Table 2 compares sociodemographic, health care and health characteristics of study participants between sexes. Overall, 65.0% of primary care visits were from female patients. (Table 2: Sociodemographic, health care and health characteristics among 631 patients attending outpatient clinics in Neno district, Malawi in August and September, 2016 compared between sexes).

Table 3 shows poor performance in relational continuity (2.3), comprehensiveness of services available (2.4) and first contact access (2.8). The highest score was in communication continuity of care (3.6). Community orientation and comprehensiveness of services provided also achieved acceptable performance at 3.1 and 3.2 respectively. Female patients scored lower than male patients in all dimensions but the difference was significant only in total primary care (p = 0.01), first contact access (p = 0.021), relational continuity (p = 0.044) and comprehensiveness of services available (p = 0.017).

(Table 4: Linear regression models assessing association between sociodemographic and health care factors and total primary care scores with unstandardized beta values among 631 patients attending outpatient clinics in Neno district, Malawi (August–September, 2016). Table 4 presents the linear regression models assessing association between patient characteristics and total primary care scores. Male patients scored 0.7 points higher than females (95% CI = 0.2, 1.2; p = 0.01). After adjusting for sex and age, patients in upper Neno scored total primary care 0.5 points higher than lower Neno patients (95% CI = 0.04, 1.0; p = 0.033). Increasing self-rated health status (rated on a 5 point Likert scale from very poor to excellent) was associated 0.8 points higher scores at good (95% CI = 0.1, 1.5; p = 0.034) and 0.9 points for very good to excellent (95% CI = 0.3, 1.4; p = 0.002), duration of contact with facility of more than 4 years was associated with scores 1.1 points higher (95%CI = 0.4, 1.2; p = 0.003) while acute presentation was associated with 0.6 points lower (95% CI = − 1.0, − 0.1; p = 0.03). At the individual facility level, patients from the health centers scored significantly below the reference outpatient clinic at the district hospital by points ranging from 0.6 to 2.0. Level of education, distance to the facility, cost of travel to the facility and waiting time were not associated with total primary care scores.

(Table 5: Association between predictors and total primary care scores, access and comprehensiveness of services available mean scores with unstandardized beta values among 631 patients attending outpatient clinics in Neno, Malawi (August – September, 2016)). The investigated factors explained 10.9% of the noted variance in total primary care scores. Looking at each dimension, these sociodemographic and health care characteristics explained 29.4% of variance in first contact access and 25.2% in comprehensiveness of services available (Table 5). These factors also explained 3% of variance in comprehensiveness of services provided, 3.7% in community orientation, 4.4% in relational continuity of care and 5.2% in communication continuity of care (data not shown in the table).

Strengths of the study include use of a globally accepted tool that had been culturally adapted and validated for use in Malawi [28]. The PCAT-Mw has advantages compared to other tools that measure patient perspectives in that it assesses patient experience with care. Since this is the first time the PCAT-Mw has been applied in a clinical setting in Malawi, the results of the paper provide a measure of the performance of primary care in Malawi. This also adds to the construct validation of the questionnaire.

The study had a number of limitations. First, because this was a cross-sectional study, causal inferences to findings are not possible. Second, liability to several types of bias is noted: recall, response and selection. The face to face interview partly minimized recall bias through clarifying questions whenever that was necessary. Potential for response bias was possible because data collection was done during clinic visit. Selection bias might have resulted from excluding those who were acutely ill, frail or had severe mental illness and interviewing only patients who attended clinics and might have had better experience than the patients excluded. The study was also carried out in one district only. In another subsequent study, we have included multiple sites to improve generalizability of results. Third, the factors identified accounted for 10.9% of total primary care score variances, 25.2% in the comprehensiveness of services available and 29.4% in the first contact access. Potential unmeasured factors such as the actual quality of services provided and health care workers’ skills, attitude and behaviors might confound the results. Fourth, this was a study of patient experiences of primary care and not of health outcomes. Further studies could assess correlations between clinical outcomes and patient experiences of care and the extent to which patient experiences predict later health outcomes.