Awareness, agreement, adoption and adherence to type 2 diabetes mellitus guidelines: a survey of Indonesian primary care physicians

Based on the evaluation of hypertension guidelines questionnaire by Heneghan et al. [19], we developed a similar questionnaire centered on items in the Consensus on the Management of Type 2 Diabetes Mellitus 2011 of the Indonesian Society of Endocrinology [7]. We included questions on the respondent characteristics: gender, age, specialization, practice duration, type of practice, location of practice, previous participation on type 2 diabetes management training and number and proportion of diabetes patients seen in their practice. According to the guideline recommendations we grouped the questionnaire content into screening, diagnosis, treatment, life-style modification, management of co-morbidities and diabetes complications (Table 1). For the nominal and ordinal response options we followed the Pathman ‘awareness-to-adherence’ model, notably awareness (Yes/no), agreement (Yes/Unsure/No), and adoption (i.e. the recommendation is being followed in general in the appropriate patients; always/more than half/less than half/never). Adherence was assessed with an open ended question about the system responders had in place to promote or monitor the guideline application.

The questionnaire was pilot tested to five GPs from the Community Medicine Department of the Faculty of Medicine Universitas Indonesia to determine whether the questions were clear, understandable, and in a logical order (face validity). Moreover the same GPs and three endocrinologists who are familiar with the diabetes guideline were asked to criticize the content of the questionnaire (content validity). Based on the results of this pilot, minor changes were made. Further psychometric evaluation of the reliability was not performed.

The final questionnaire was distributed to all physicians attending the Indonesian Association of Family Practitioners annual conference on November 2012 in Jakarta, Indonesia. The questionnaire was put in the delegates pack together with an information leaflet on consent for survey participation. Returning of the self-completed questionnaire by responders was seen as their token of consent. Before handing out their certificate of attendance conference participants were informed about the possibility of participation. The Health Research Ethics Committee of the Faculty of Medicine Universitas Indonesia reviewed and approved the study.

We explored the association between respondent characteristics and adherence to each of the seven guideline recommendations. For this we used multivariate logistic regression analysis. To prevent for a type 2 error, only GP characteristics with a univariate p-value of 0.20 or less were selected for such multivariate analyses. The outcome for these data analyses was the number of responders’ adherent to a recommendation. To prevent for spurious findings at least 10 participants adherent to the recommendation were needed for each respondent characteristic included in the multivariable analysis [20]. We used SPSS, version 20.0 for Windows (SPSS Inc., Chicago, IL, USA) for all data analysis.

Responders were classified as ‘unaware of a recommendation’ if they answered “no” on the question on familiarity with that recommendation. Agreement was classified according to whether they agreed or not with the guideline. They were considered to have adopted a guideline when they reported implementing it ‘more than half of the time’. They were considered to adhere to the recommendation when they ‘always’ or ‘more than half of the time’ applied it in clinical practice and specified the system they used to promote or monitor application. The proportions of doctors who were aware of, agreed with, adopted and adhered to each recommendation were calculated over the total number of responders.

Years of practice experience was grouped in 15 years or less and more than 15 years. The Perkeni guideline was firstly introduced in 1993. Therefore we assume that GPs who practiced less than 15 years should be aware of the guideline during their medical training.

We assumed the missing outcome data to represent no awareness, no agreement, no adoption and no adherence. For respondent characteristics with up to 15% missing data, we used conditional imputation, imputing the mean or median. We used mean and median values for imputation since we thought we had not the proper participant characteristic’s to do regression analyses for imputation.

There was high awareness among responders on the need for screening for type 2 diabetes risk factors among those without diabetes symptoms. Still, most of our responders waited until several cardiovascular risk factors emerged in the patient before they advised a screening test to the patient or they did not have a system to identify patients with risk factors. This was indicated by an extremely low (2%) adherence to the screening recommendation for identification of type 2 diabetes. This result is in contrast with findings among GPs in Switzerland about their adherence to the screening of diabetes which reach 83% [21].

We found the least agreement (41%) with the recommendation to use random blood (plasma) glucose in the diagnosis of diabetes in patients with classic diabetes symptoms. The majority of our responders believed it is more appropriate to examine patients with classic diabetes symptoms with a fasting plasma glucose test and 2-hour post-prandial plasma glucose test. Due to the fasting needed for these blood sugar tests, patients may delay or avoid the examination.

On statin prescription, we found that the level of awareness and adoption were the lowest (72% and 52%). With 32% adherence among our responders to the recommendation to prescribe statins, it is slightly higher than in a Seoul tertiary hospital (29%) [14] but much lower than the 68% adherence among Australian GPs [11]. Factors related to the adherence to the recommendation to prescribe statins in our study were more years of practice experience and larger proportion of diabetes patients the practice.

Adoption to guideline recommendations has been shown to be facilitated by the acquisition of the necessary knowledge and skills. High health costs for patients and practice, patient’s knowledge, expectations, compliance, motivation and support for recommendation, lack of materials, logistic support and time of health professionals, and high proportions of patients without insurance have been reported as barriers to guidelines adherence [16, 17]. Adherence to actions recommended in guidelines may require practice organization. Besides, adherence is facilitated when tools are in place to put the recommendations into practice [22]. A systematic review [17] reported a median adherence of 36% (Inter Quartile Range 30 to 56%) to recommendations from various guidelines. Our data reveal that adherence to all the recommendations was quite low since less than 50% responders did not implement a system to promote and monitor recommendations in their practice.

Our findings on the low adherence of the recommendation on screening and diagnosis may explain why two thirds of the diabetes population in Indonesian remain un-diagnosed [2]. The generally low adherence to the recommendations of the Indonesian guideline may partly explain the finding by Soewondo et al. that 68% of patients in Indonesia diagnosed with diabetes were in poor glycemic control [6].

Based on a systematic review on the utilization of clinical guidelines [17] we expected to find a consistent pattern of ‘leakage’ , i.e. lower number of positive responders, over the four subsequent steps of the Pathman’s awareness-to-adherence model. In our study, adoption and adherence rates are generally progressively lower, except for the recommendations on diagnosis of diabetes and sulfonylurea treatment.

Our findings on the non-progressiveness of the recommendation on diagnosis of diabetes and sulfonylurea treatment deserve further consideration. As was stated previously, a majority of our responders believed that fasting and 2-hour post-prandial blood (plasma) glucose test are more appropriate than random blood (plasma) glucose in the diagnosis of diabetes in patients with classic diabetes symptoms. However, random blood glucose is more practical to be implemented in the practice so it is more adopted.

The Indonesian guideline recommends sulfonylurea as the drug of choice to manage hyperglycemia in normal and underweight type 2 diabetes patients. In all, 58% of our responders agree with this recommendation. In contrast to the Indonesian guidelines, the consensus statement of the American Diabetes Association and the European Association for the Study of Diabetes recommends simultaneous initiation of metformin and lifestyle intervention at diagnosis [10]. Familiarity with this ADA/EASD recommendation may cause some uncertainty, but metformin and other classes of blood glucose–lowering medications are not generally available in primary health centers [23]. This may explain why agreement with sulfonylurea as recommended treatment is lower than its adoption.

Our study supports the usefulness of the awareness-to-adherence model and provides valuable information on the utilization of an important guideline in Indonesia. Still, some aspects of our study, notably the sampling method, response rate of the survey, missing data and the use of self-reporting questionnaire, need further consideration.

Obtaining representative samples from the large number of Indonesian GPs, exceeding 70,000, who are distributed over the archipelago possess challenges for our type of research.

Recruitment of responders among those attending a conference in Jakarta (capital city of Indonesia) was seen as more practical although it might not represent physicians who do not have opportunity to attend such meeting. However, our responders represent GPs from all parts of Indonesia. Our response rate is within the range of that of similar studies as reported by the systematic review of Mickan et al. [17].

Self-reporting is the most simple and inexpensive method of measuring adherence. However, it has several limitations including over-estimation due to recall bias and social desirability bias [24]. These drawbacks have been addressed through determination of specific time period in the questionnaire, assessment of specific behavior related to the recommendations, non-judgmental statements and confidentiality.

The awareness-to adherence model may help to identify GPs’ specific concerns with recommended practice changes. If uptake of a specific recommendation is low, qualitative approach to the concerns and barriers might be useful. Implementation of a quality assurance system which could further illustrate the care being received by diabetes patients in relation to the clinical outcomes is believed to be beneficial to promote adherence to guideline recommendations and increase the quality of diabetes care [25].