“Damm Sokkor”: Factors Associated With the Quality of Care of Patients With Diabetes: A study in primary care in Tunisia

Tunisia is a country of 10 million inhabitants, situated on the North African coast. There are ∼2,000 public sector, primary care health centers situated throughout the 24 regions of the country. The majority of these centers are small, nurse-ran health posts that do not manage patients with chronic diseases; we therefore chose to include only health centers that hold medical consultations four or more times a week (n = 567). Two health centers were randomly selected from each region using data obtained from the Ministry of Public Health.

Patient data were extracted from manual medical records. A maximum of 50 patients with diabetes were randomly selected per health center. Patient details included demographic data, clinical background, processes of care (i.e., whether a measurement had been recorded in a 12-month period), outcomes of care (i.e., the result of the measurement), and prescriptions of blood glucose–lowering, antihypertensive, and lipid-lowering medication. Physician and organizational characteristics were collected from national and health center reports, a structured questionnaire administered at each center and interviews with the staff at the health centers. Explanatory factors were selected on the basis of research findings elsewhere and exploratory qualitative work in Tunisia (23) (Table 1).

The quality-of-care indicators were based on two process-of-care scores and two outcome-of-care scores. The process-of-care scores were calculated based on recommendations from the Tunisian national program (21), namely assessments of fasting glucose, blood pressure, weight, total cholesterol, creatinine, foot examination, cardiovascular examination, electrocardiogram, eye examination, and A1C. The latter four tests usually require referral to a local hospital, as they cannot be performed on site at the health center. Following a model proposed by Gulliford et al. (4) in Trinidad and Tobago, we combined the process-of-care results to create two quality-of-care scores.

Nonweighted process-of-care scores were obtained by assigning to each patient a score of 1 for each measurement undertaken in the previous 12-month period (maximum score: 10).

To take into account the importance of glycemic and blood pressure control, a score was calculated in which glucose and blood pressure measurement were given a weighted score of 4 rather than 1; the other measurements remained with a score of 1 (maximum score: 16). The outcome-of-care scores were based on levels of fasting glucose, blood pressure, total cholesterol, and BMI. The assessment was based on an average of all the results collected per patient.

An outcome-of-care score was calculated based on how many of the following targets a patient achieved: blood pressure <140/80 mmHg, fasting glucose ≤7.8 mmol/l, total cholesterol ≤5 mmol/l, and BMI ≤25 kg/m² (24). Missing data were excluded. A score was assigned to each patient based on the proportion of targets achieved.

A second outcome-of-care score was calculated using fasting glucose and blood pressure levels only. The scoring system used a range from good control (using definitions above), borderline control, and poor control (defined as blood pressure ≥160/95 mmHg and fasting glucose ≥11.1 mmol/l). Each patient was assigned a score of 2 for good control, 1 for borderline control, and 0 for poor control for both fasting glucose and blood pressure using a denominator of 2 (if only one variable recorded) or 4 (if both variables recorded). A mean of each of the four scores was calculated for each health center. The scores were assessed for normality, and the value of Cronbach's α was calculated to measure the internal consistency of each score.

The health center was used as the unit of randomization in order to cluster patients into practices, as recommended in primary care studies (25). All explanatory variables were first tested against each of the outcome variables (quality-of-care scores) using ANOVA (categorical variables) or linear regression (continuous variables). Logarithmic transformations were made for variables not normally distributed; if the variable remained not normally distributed, the variable was converted into a categorical variable. Analyses were weighted for number of patients per center and date of data collection. Potentially significant variables (P < 0.15) were entered into three separate multilinear regression models, grouping variables into patient, health professional, or organizational with each of the outcome variables as the dependent variable. Potentially significant variables (P < 0.15) from each of the three separate models were then entered into a final regression model against each outcome variable. The data were analyzed using SPSS software package (version 12.0.1). Approval for the study was granted by the Tunisian Ministry of Public Health.

A total of 2,160 patients with diabetes were selected for medical record review from 48 health centers; the mean age of patients in the study was 62.4 years, mean duration of diabetes was 8.4 years, 62% were female, and 94% had type 2 diabetes.

A mean of 45 patients were selected per health center. The ratio of urban to rural health centers was 2:1. Health centers had a mean of 2.1 primary care doctors and 5.6 nurses, and 20% had a nutritionist available for patients with diabetes. On average, each health center served a population of 15,986 and managed 162 patients with diabetes, and 26 patients attended per clinic per doctor. Among the 48 health centers, 85% had the new disease-specific medical records available, 70% had a chronic disease register, 63% used patient-held records, 79% had a weekly chronic disease clinic, 39% had an electrocardiogram machine on site, 93% had a glucometer on site, and 57% ran regular patient education sessions. Table 2 depicts selected patient characteristics, and Table 3 depicts the results of the process and intermediate outcomes of care of the study population. All four quality-of-care scores were normally distributed. Internal consistency was high for the process-of-care scores (0.84 and 0.81) but lower for the outcome-of-care scores (0.58 and 0.29) due to the lower number of variables incorporated.

Univariate analysis demonstrated a potential association among 16, 18, 13, and 11 of 53 explanatory factors with the four quality-of-care indicators (nonweighted process-of-care, weighted process-of-care, four-variable outcome-of-care, and two-variable outcome-of-care scores, respectively; online appendix [available at http://dx.doi.org/10.2337/dc07-0520). All factors potentially related to each quality-of-care indicator were entered into the three separate multilinear regression models, grouping factors into patient, health professional, or organizational. Factors that remained potentially significant were entered into a final regression model for each indicator, and these are demonstrated in Table 4. The final models explained 71.3% (nonweighted process-of-care score), 62.7% (weighted process-of-care score), 64.4% (four-variable outcome-of-care score), and 55.9% (two-variable outcome-of-care score) of the variations in scores.

We report the first nationwide study from primary care of the factors that influence the care of patients with diabetes from a low-/middle-income country. Use of chronic disease clinics, availability of medication, and doctor motivation appear to be the most strongly related modifiable factors influencing diabetes care in our context. The other factors that were independently and significantly associated with improved processes or outcomes of care were regional affluence and younger age.

The process-of-care results show that the majority of patients are having their blood pressure and fasting glucose recorded annually. These results compare favorably with studies from similar countries (4–6,26,27). Around half of the patients have most of the other measures performed annually. Fewer patients are recorded as having an electrocardiogarm, eye examination, and A1C measurement. The latter is almost certainly due to the fact that this test is not generally available within primary care. The low recording of eye and electrocardiogram examinations may be due to the fact that these tests are usually performed in secondary care; primary care physicians report difficulties in persuading patients to attend and in receiving the results from secondary care. Since the time of the study, training of primary care doctors in the use of ophthalmoscopes has been introduced, and it is hoped that this will improve the uptake of eye examinations. Particularly striking is the variation in results between health centers as has been demonstrated in other countries (3,4). The percentage of patients achieving targets of blood pressure, fasting glucose, and cholesterol is variable and suboptimal but again compares favorably with results from other countries (4,5,26).

Assessing the relative influence of specific factors that influence diabetes care is essential for the development of targeted interventions to improve the quality of care. Our study showed five factors to be clearly associated with improved processes or outcomes of care: regional affluence, doctor motivation, use of chronic disease clinics (processes), younger age, and increased availability of medication (outcomes).

An association between affluence and quality of care has been demonstrated previously in studies from the developed world (11,28), and it appears that this influence is equally important in less affluent countries. Financial aspects strongly influence the care of patients, especially those with chronic diseases, from developing nations (29).

The significant influence of doctor motivation is perhaps unexpected. Historically, more emphasis has been placed on the training and education of clinicians rather than their attitudes and beliefs, but motivation of the health professionals is increasingly being recognized as having as central role in diabetes care (12). However, this finding must be approached with caution given the subjective nature of the term “motivation,” even within the context of a theoretical model (30), and the subjective method of data collection (interviews and observations). Further investigation is required using more formal methods, such as validated questionnaires or surveys, to confirm this potential discovery. The introduction of weekly chronic disease clinics at most of the health centers studied seems to have been a major success in improving the quality of diabetes care in Tunisia. Structured care in the primary care setting has been shown in systematic reviews from developed nations to be associated with improved quality (19), and our findings suggest that these results can be generalized to less affluent nations. The association of younger age with improved outcomes of care seems to be related to the inclusion of BMI and cholesterol in the four-variable outcome-of-care score. A national nutrition survey in Tunisia 10 years ago demonstrated the association of age with BMI and cholesterol in Tunisia, as in other countries (31).

Finally, the association of improved outcomes of care at health centers with increased availability of medication suggests a direct link between intermediate patient outcomes and medication availability. In the Tunisian public sector, medications are free with the payment of a small consultation fee. If the medications are unavailable, patients are required to buy them privately from pharmacists, and many cannot afford to do so. Other authors from developing nations have stressed the essential role of the provision of medication (29,32). One of the aims of the Tunisian national program has been to prioritize the supply of medicines for chronic diseases, and our evidence supports this initiative.

Quality of health care is a multidimensional concept that has been identified as including a combination of access (assessed in our study by processes of care) and effectiveness (assessed by outcomes of care) (33). Much debate has centered on the use of processes or outcomes to assess quality of care (34). We chose to use a combination of process and outcome measures in an attempt to give a more accurate overall picture of the factors influencing both the recording of care and the achievement of clinical outcomes. We recognize that our outcome variables are intermediate and not long term; it is not possible at present to identify long-term outcomes, such as complication and mortality rates, in our setting.

Our study is the first nationwide study from primary care on the factors that influence diabetes care from a low-/middle-income country to be reported. In addition, it is one of the first to incorporate an extensive number and range of potential variables, including patient, health professional, and organizational factors. Selection of the variables was based on exploratory, qualitative work from Tunisia (23) in addition to reported findings from elsewhere. Our inclusion of >50 potential factors, though larger than previous studies, is not exhaustive, and other unexplored factors may be playing a role. However, it is reassuring to note that our final models did explain most of the variations in quality scores observed. Certain explanatory variables could be subject to bias; for example, availability of medication was based on reports from staff rather than an objective measure.

Based on a two-stage randomized procedure, our study is nationally representative of the public sector primary care management of patients with diabetes, covering >150,000 patients throughout the country. It is possible that some of the factors discovered may be contextual and not transferable to other settings. Nonetheless, being one of the first and largest studies to be reported from a low-/middle-income country, we would suggest that our findings are more likely to be relevant to other similar countries than previous work from developed nations.

In summary, we found the use of chronic disease clinics, the availability of medication, and possibly doctor motivation to be the most strongly related modifiable factors influencing the quality of diabetes care in the Tunisian primary care setting. We suggest that our findings be evaluated in other settings. However, it is unlikely that such a large, encompassing study can be undertaken in every context, particularly in less affluent nations. We would therefore recommend that clinicians, managers, and health policymakers take our results into consideration in order to develop and implement culturally appropriate quality improvement interventions in other low-/middle-income countries.

We thank all the staff at the primary care health centers for their hospitality and assistance. We are also grateful for helpful advice from Rudy Bilous, Nigel Oswald, Janine Gray, Tom Chadwick, George Alberti, Robbie Foy, and colleagues at the DSSB.

A summary of this article was presented at the International Diabetes Federation Conference, South Africa, 2006.