The mean age of this study population was higher compared with published local data which evaluated the diabetes control of 1670 patients in Malaysia . Both local and global data revealed that in developing countries like Malaysia, the prevalence of diabetes mellitus was highest in the age group between 45 and 64 years  but in this study only 36% of the patients fell into this age group. The higher mean age reported in this study might be because the sample was not representative of the whole population. It also suggests that more older hypertensive diabetic patients were admitted to the ward than were younger patients.
In this study, there was an average of 1.9 ± 1.2 DRPs per patient. To date, there has been no comparable study done specifically on DRPs in T2DM patients with hypertension both locally and globally. The number of DRPs identified was only half the number detected by some other studies which were conducted in diabetes mellitus patients [10, 11, 25]. When compared with a recent study with an almost equivalent sample size (193 geriatric clinic patients in Taiwan), which also used the PCNE classification system, the average number of problems identified was 2.2 ± 1.6 per patient, slightly higher than in this study .
Although a similar PCNE classification of DRPs was used, the discrepancy with the study by Van Roozendaal et al. (2009)  could be due to the different methods and references used to identify DRPs. For example, the concurrent use of an ACE inhibitor and a sulfonylurea or insulin was considered as a potential DRP in that study and accounted for 46 cases out of the 682 DRPs detected. However, this combination of drugs was not considered as a potential drug interaction in this study because there is a lack of strong evidence of interaction . Also, Van Roozendaal et al. (2009)  could not identify any possible contraindications as no information on patients’ renal and hepatic functions were successfully retrieved but this is not the case in this study as several contraindications were identified based on patients’ medical records and laboratory results.
Apart from that, the discrepancy with other studies may be attributed to the differences in study method and setting, different classification systems of DRPs used, and different methods to assess DRPs. Both the studies by Haugbolle & Sorensen (2006)  and Eichenberger et al. (2011)  conducted home visits and interviews. Also, the Problem Intervention Documentation (PI-Doc) coding system was used in the study by Haugbolle & Sorensen (2006) whereas the classification system of DRPs used by Eichenberger et al. (2011) was unclear, as it was not mentioned in its Methods section. Also, the clinical knowledge of the investigator(s) might also influence the assessment and identification of DRPs.
This study revealed that 90.5% of the patients had at least one DRP, which was much greater than the 80.7% reported by Haugbolle & Sorensen (2006) . However, in a study conducted on ambulatory hemodialysis patients, 97.7% of the patients were found to have at least one DRP . This variation across the studies can be attributed to the different study populations and study methods.
In this study, almost one-third of the adverse drug reactions implicated antidiabetic drugs. Similar to the finding by Van Roozendaal & Krass (2008) , there was a potential risk of hypoglycemia in patients receiving oral antidiabetic drugs or insulin. Antihypertensive agents were also commonly associated with adverse reactions  and this finding was clearly demonstrated in this study. Calcium channel blockers caused a higher incidence of adverse reactions than diuretics, consistent with a study in an outpatient setting by Goncalves et al. (2007) . Therefore, all these potential adverse reactions should be taken into consideration, especially in the elderly who might suffer significant deleterious effects. However, since this study was retrospective in nature, only the ones that were important for the hospital admission were noted.
Drug choice problems
A drug choice problem was the second most common DRP in this study and this finding was comparable to other studies [10, 11]. In this study, most of the contraindications identified were related to the use of metformin. Approximately 24% of the patients who received metformin were found to have contraindications. This was much less compared with the study carried out by Sweileh (2007) , in which up to 60% of the patients receiving metformin had contraindications to it. This difference is probably due to the variations in defining metformin’s contraindications. For instance, the decompensated stage, but no other stage of congestive heart failure, was defined as a contraindication in this study.
Apart from metformin, drugs that were categorized as high risk in the modified Beers criteria were frequently prescribed to the elderly, placing them at higher risk of developing drug toxicity [4, 31]. The high frequency of drug choice problems may highlight a need for the health care providers to pay more attention when prescribing these drugs to older hypertensive diabetic patients.
In this study, excessive dosage was the most frequent dosing problem. Also, there were cases of subtherapeutic dosages and inappropriate durations of treatment. The percentage of dosing problems in this study was higher than that reported by Van Roozendaal and Krass (2009)  and this was probably due to the lack of assessment of patients’ renal and hepatic functions in the latter study.
Ranitidine was the most implicated drug for dosing problems. In clinical practice, dosage adjustments of ranitidine are not frequently applied although recommended by manufacturers and this is probably because its potential adverse effects are underestimated [36, 37]. In a study by Manlucu et al. (2005) , H2-receptor antagonists were demonstrated to significantly increase the area under the curve (AUC) and the elimination half-life (t
1/2) of serum drug concentrations when the glomerular filtration rate (GFR) was decreasing. Dosage reduction of drugs in patients with impaired renal function and low GFR may prevent adverse effects and decrease unnecessary drug expenditures [36, 37]. Therefore, efforts should be made to minimize these dosing errors such as the involvement of a pharmacist in deciding the dosing of drugs or a computerized dosing program .
Drug use problem
The most frequent drug use problem encountered was “drugs not taken by patients prior to admission”, which were mostly antihypertensive and antidiabetic drugs. In this study, forgetfulness might be one of the reasons for non-adherence. Also, complicated regimens for the treatment of diabetic complications may contribute to non-adherence. A systematic review confirmed the poor compliance in diabetic patients who were prescribed diabetic medications, whether oral agents or insulin . Non-adherence has proven to be associated with negative outcomes such as higher A1c levels and blood lipid levels in diabetes patients [12, 38].
In this study, the drugs that were most implicated in drug interactions were aspirin, clopidogrel, simvastatin and amlodipine. By contrast, beta-blockers, non-steroidal anti-inflammatory agents (NSAIDs) and ACE inhibitors were the drug most involved in drug interactions in a study conducted in Singapore . The differences in prescribing patterns and practice in different hospitals may explain this discrepancy. The drug interactions identified in this study were mostly based on established literature and evidence. In clinical practice, several drugs can still be used together, yet close monitoring is fundamental and any toxicity should be identified and immediately followed by corrective actions.
Many of the patients in this study did not engage in regular physical activity, did not adhere to diabetic diets, did not perform any routine blood glucose monitoring, and defaulted on follow-up or medications, and all these problems were clearly stated by the health care providers in the medical records. These problems would lead to poor glycemic control and accelerate the development or worsening of diabetic complications .
A local study conducted in an endocrine clinic in a teaching hospital in Kuala Lumpur identified barriers to optimal control of Malay type 2 diabetic patients by interviewing 18 patients and health care providers. It was not surprising to find that most of the patients had a lack of understanding of the disease itself and of its management, which would eventually contribute to non-adherence . This is particularly true for the elderly as they tend to have decreased memory, health beliefs not in line with drug therapy, and often neglect the importance of adherence to medications and dietary control. Hence, counseling may be important to increase the awareness and knowledge of this patient population since they frequently encounter these problems.
Causes of DRPs
The results from our study revealed that among all the causes, “drug or dose selection” was the most frequently identified cause for DRPs such as drug choice problems, dosing problems or drug interactions. According to the PCNE classification of DRPs, this domain of causes is directly related to the drug or dosage selection while the other domains are concerned with patient-related causes .
On the other hand, patients usually had “drug use problems” caused by “drug use process, lack of information, and physiological or patient factor”. Generally, the number of causes identified was lower than the causes identified in other studies such as Chan et al. (2011) . This is because most of the problems identified were matched with the one most relevant cause rather than several causes, which might be seen in other studies.
Factors found to be associated with DRPs
Generally, this study did not identify any factors with a statistically significant association with DRP. This was in agreement with a study by Koh et al. (2005)  which found no statistical correlation between DRPs and age or gender. Research by Samoy et al. (2006)  also concluded that there were no risk factors for drug-related hospitalization in a tertiary care hospital in Canada. One of the possible explanations is the nature of the PCNE classification system (6 problem domains with 22 categories) which could possibly affect the results. Also, it remains unclear whether the result was affected by the sample size (Samoy et al., 2006) .
When examining each of the problem domains with several possible factors, statistically significant associations were observed. These associations should receive the attention of the health care providers in order to minimize preventable DRPs.
In our study, the non-elderly were found to be associated with drug choice problems. From the literature reviewed, the findings on the association between age and DRPs are conflicting. In one study on the elderly in an ambulatory setting, age of 80 and above was found to be an independent risk factor for adverse drug events . While a study by Chan et al. (2011)  on geriatrics also reported a significant association between age and DRPs, a few studies did not agree with this finding. A study on hospitalized patients from several internal medicine departments found that age was not a risk factor of DRPs . Similarly, Koh et al. (2005)  did not report any statistically significant correlation between these two.
It is a well-known fact that polypharmacy is strongly associated with DRPs and this has been shown by numerous studies [12, 14, 15, 43]. It has been reported that a one unit increase in the number of drugs can lead to an increase of 8.6% in the number of DRPs .
The results of our study revealed a significant statistical association between polypharmacy and drug interactions, which was consistent with the result from Moura et al. (2009) , a retrospective study on drug interactions in a public hospital in Brazil. The increasing number of drugs used can lead to an increased risk of potential drug interactions [23, 45]. Since polypharmacy is an inherent problem in T2DM patients with hypertension, the clinically important and significant drug interactions should be routinely checked and monitored .
Renal impairment was associated with both the drug choice problem (p = 0.029) and the dosing problem (p = 0.027). Drugs with doses that were higher than required were often prescribed to T2DM patients with hypertension and renal impairment in our study. Also, dosage adjustment was commonly ignored by physicians, suggesting that the severity of inappropriate drug and dosing selection might be underestimated [36, 37]. DRPs were common among patients with renal impairment due to co-existing medical conditions, as most of them were receiving multiple medications which require dosage adjustment and routine monitoring .
The study by Manley et al. (2003)  revealed that in ambulatory hemodialysis patients, the presence of diabetes mellitus is one of the factors associated with DRPs. In other words, diabetes patients on hemodialysis were more likely to experience DRPs. Similarly, Leendertse el al. (2008)  also found that impaired renal function was a risk factor for potentially preventable DRPs. Therefore, in patients with renal impairment, dosage adjustment and close monitoring of renal function are fundamental in order to minimize drug toxicity or subtherapeutic effect .
Patients with cardiovascular events had more potential drug interactions than patients without cardiovascular disease in our study. This can be explained by the wide use of cardiovascular drugs such as antihypertensive drugs, antiplatelet drugs, anticoagulants, and lipid lowering drugs in T2DM patients with hypertension. Many studies concluded that the most common drug category involved in DRPs was cardiovascular agents [12, 47]. Also, cardiovascular events often add an additional burden to patient conditions and complicate their therapies.
Duration of stay
Patients who stayed for not more than one week in the hospital tended to experience the drug choice problem as compared with those who stayed for more than one week. This finding of our study was not in line with that reported by Moura et al. (2009) , which revealed an association between duration of hospital stay and potential drug interactions.
Another study on hospitalized cancer patients also found a correlation between duration of hospital stay (≥ 6 days) and potential interactions . The relationship between length of hospital stay and DRPs needs to be investigated in future studies since there is a lack of published literature investigating this association.
Because of the retrospective nature of our study, the identification and assessment of the DRPs were based on the data available from the medical records with reference to established literature and guidelines. The number of studied patients was less than 50% of potentially eligible patients.