Multi-morbidity and its associations with healthcare service utilisation and glycaemic control in diabetic patients in a primary care clinic in Hong Kong

Cheuk-chung Sung 宋卓聰，Tsun-kit Chu 朱晉傑，Jun Liang 梁峻

HK Pract 2020;42:77-86

Summary

Objective: Multi-morbidity is common among diabetic patients and often complicates its treatment. The objectives of this study were: (1) to investigate the magnitude of multi-morbidity among patients with Type 2 diabetes, and (2) to examine its relationship with glycaemic control and public healthcare service utilisation.
Design: A cross-sectional study, retrospectively analysing electronic medical records
Subjects: A random sample of 386 adults, aged 18 or over, with Type 2 diabetes who attended a General Out-patient Clinic (GOPD) in Hong Kong, for their diabetic follow-up between 1 January 2014 and 31 December 2015.
Main outcome measures: The outcome measures were the number of co-existing chronic conditions, glycaemic control (as measured by HbA1c) and healthcare service utilisation (as measured by the number of attendances for medical consultations in the General Out-patient Clinics, Specialist Out-patient Clinics, Accident and Emergency Departments, and number of unplanned hospital admission episodes) within the reported period. Other data collected were the number of chronic medications (polypharmacy), and patient demographics including age, gender, body mass index (BMI) and smoking status. Regression analyses were used to analyse the associations between multi-morbidity and HbA1c level and public healthcare service utilisation.
Results: Multi-morbidity was observed in almost all (99.2%) patients in our sample, with 91% of them having two or more co-existing conditions. The four most common co-existing conditions were hypertension, hyperlipidaemia, visual conditions and chronic kidney disease. After adjusting for age, sex, BMI, smoking status and the presence of polypharmacy, an increase in number of co-morbid conditions significantly increased the healthcare service utilisation in secondary care. The number of co-existing conditions had no statistically significant association with the glycaemic control or the number of attendances to General Out-patient Clinics.
Conclusions: The magnitude of multi-morbidity is associated with healthcare service utilisation, independent of the presence of polypharmacy. This study highlighted the importance of adopting a multidisciplinary and holistic patient-centred approach when managing patients with diabetes.

Keywords: Diabetes, multi-morbidity, polypharmacy, disease control, healthcare utilisation

摘要

目 標 : 糖尿病人經常伴有其他病患，而使治療變得複雜。本研究旨在：(1) 探研二型糖尿病人伴有其他病患的普遍性，和(2) 檢視其他病患與血糖控制及對公共醫療服務需求的關係。
設計 ： 按病人的電子醫療紀錄作橫切面式，回顧性分析。
對象 ： 隨機挑選386名18歲或以上，於2014至2015年間在香港一所普通科門診因二型糖尿病而覆診的病人進行研究。
主要結果測量 ： 結果測量項目為長期並存病患的數目，血糖控制(按HbA1C量度)，在研究期內對公共醫療服務的需要(包括前往普通科門診、專科門診、急症室求診和非預約入院的次數)。其他收集的數據有長期服用藥物的數量(多元藥物治療)和病人的統計學個人資料：包括年齡、性別、體重指數(BMI)和吸煙習慣。以還原分析法分析共存病患、血糖控制和對公共醫療服務需求的關係。
結果 ： 在研究組別中，接近所有(99.2%)病人伴有其他病患。其中91%有兩種或以上不同病患。最常見的四種病症是高血壓、高血脂，視力障礙和慢性腎病。經調整年齡、性別、體重指數、吸煙狀況和使用多元藥物等因素後，發現並存病患的多少和對公共醫療服務需求有密切的直接關係。而它與血糖控制和往普通科門診次數則無明顯聯繫。
結論 ： 共存病患的多少和對公共醫療服務的需求有相聯關係，但與服用多種藥物無關。本研究結果突顯在醫治糖尿病時，採納以不同專科匯診方式，並以病人為本作全人治療的重要。

主要詞彙 ： 糖尿病，多種病患，多元藥物治療，疾病控制，醫療服務需求。

Introduction

The prevalence of diabetes in Hong Kong has increased significantly over the past decade and in 2014, the overall prevalence of diabetes in Hong Kong was 10.29%.¹ Multi-morbidity, defined as the presence of two or more chronic conditions² , is increasingly recognised as a major challenge in healthcare service globally. A recent study undertaken in Hong Kong showed multi-morbidity was commonly encountered in the public General Out-patient Clinics (GOPDs), with diabetes being the chronic condition which had the strongest association with multi-morbidity.³ With an ageing population, it is expected that the prevalence of multi-morbidity will increase, along with the rising burden and risks associated with polypharmacy.

Multi-morbidity often complicates diabetic treatment and its outcomes, as evidenced by studies undertaken with Caucasian subjects, which analysed the relationship between multi-morbidity, diabetic care goal achievement and healthcare utilisation.^4-10 However, these associations are less studied in the Asian population. Given the phenotypic variabilities, it is important to study these associations in the local population. Polypharmacy can increase the risk of drug-drug interactions, and hence, the adverse drug reactions, particularly in the older population and those with multi-morbidity.^11-12 Therefore, both multimorbidity and polypharmacy create significant concerns for the individual as well as for the healthcare system.

The public General Out-patient Clinics in Hong Kong manage a significant proportion of diabetic patients in the locality.¹³ With diabetes shown to have the strongest association with multi-morbidity in the public GOPDs³ , further analysis on the multi-morbidity pattern of diabetic patients is warranted. In addition, there is no study published, in the literatures thus far, on the association of multi-morbidity and glycaemic control and healthcare service utilisation in the local Chinese diabetic patients.

We hypothesised that the number of co-existing conditions would be associated with poor glycaemic control (as measured by HbA1c), and a higher degree of healthcare service utilisations (as measured by the number of medical consultation visits to the GOPDs, Specialist Out-patient Clinics and Accident and Emergency Departments, and number of unplanned hospital admission episodes). We envisaged polypharmacy as a potentially significant confounding factor which would influence glycaemic control and the degree of healthcare service utilisation. Therefore, in addition to age, sex, BMI and smoking status, we adjusted for the presence of polypharmacy in our regression analysis.

Methods
Study Design and Population

This was a retrospective cross-sectional study using electronic health record data of a computer generated random sample of adult patients aged 18 or over with Type 2 diabetes, who were followed-up in a GOPD in Hong Kong within the period of 1 January 2014 (Time 0) – 31 December 2015 (Time 2 years). Sample size calculation was determined by using the following formula, using confidence level and margin of error as 95% and 5% respectively¹⁴:-

n = Z² x P (1 – P) / d²

where:

Z = Z value, which is the standard normal variate. For 95% confidence level, Z value is 1.96.

P = expected prevalence of a condition in a population in decimal. We set the P figure as 0.5, as among our clinic patients requiring regular follow-up appointments for chronic conditions, we estimated the actual proportion with diabetes would not be more than 50%.

d = absolute error or precision in decimal. As we set the margin of error as 5%, the d figure is 0.05.

The sample size thus was calculated to be 384.

The Clinical Data Analysis and Reporting System (CDARS) is one of electronic medical systems used in the Hong Kong’s public healthcare service. A list of Type 2 diabetic patients, i.e. those with International Classification of Primary Care – 2^nd Edition [(ICPC – 2) codes T89 and T90], who attended follow-up in our clinic within the period between 1 September 2013 and 31 December 2013 was drawn from CDARS. We believed a 4-month-pool of patient attendances was a representative cohort because follow-up within 4-month-interval was our clinic policy. 430 patients (over 10% greater than the calculated sample size) were randomly selected from the list by computer generated random numbers, as we anticipated that some cases had to be excluded from our study. We excluded patients under 18 years of age, and patients with gestational or Type 1 diabetes, as well as the ones who died before Time 0, those who were actually followed up in Secondary Care or in Private Sector, and the ones who were misdiagnosed with diabetes. For the cases included, the patients had to have at least 12-month-history of Type 2 diabetes before Time 0.

Definitions

The World Health Organisation (WHO) defined multi-morbidity as the co-existence of two or more chronic conditions in the same individual. ² In this study, a list of 28 chronic conditions was used to identify patients with multi-morbidity (Table 1). These 28 conditions had included 14 chronic conditions shown to be associated with multi-morbidity among patients attending the Hong Kong General Out-patient Clinics³ , as well as 15 chronic conditions, which were considered prevalent in Singapore.¹⁵ Papers from Chu et al³ and Subramaniam et al¹⁵ were chosen to formulate the list of the chronic conditions since both Hong Kong and Singapore have similar demographics and social structure with the majority of the population being of Chinese ethnicity. Both cities also are noted to have a highly developed medical and economic infrastructure.

There is no consensus on the definition of polypharmacy, and diverse definitions have been used by researchers to date. The most commonly reported definition of polypharmacy is a numerical definition of five or more medications daily.¹⁶ In this study, we adopted the definition of polypharmacy as 5 or more medications daily as this had the advantage of being simple and easily classified in clinical practice. As-required medications (e.g. analgesics or symptomatic treatment) as well as topical medications (e.g. emollient for eczema) were excluded while counting a patient’s total medication usage.

Data Collection

Data were collected by reviewing the medical records and consultation notes in the period between 1 January 2014 (Time 0) and 31 December 2015 (Time 2 years). The review was performed by the principal investigator. Whenever there were uncertainties with the data, the records would be reviewed by the second investigator, and any discrepancies were resolved via discussion and consensus.

Our main explanatory variable was the number of co-existing chronic conditions (ever diagnosed and actively under treatment) collected at the time closest to Time 0. The actual diagnoses of the co-existing conditions were also recorded.

Our primary outcome variables were glycaemic control (as measured by HbA1c) collected at the time closest to Time 0, and the record of public healthcare service utilisation between Time 0 and Time 2 years. The parameters of healthcare service utilisation included the number of attendances for medical consultation in the public General Out-patient Clinics, Specialist Out-patient Clinics, Accident and Emergency Departments, as well as the number of unplanned hospital admission episodes.

The covariates we collected were patient characteristics, namely age, sex, BMI, smoking status and the number of regular medications prescribed at the time closest to Time 0.

Other data we collected but not used for our analysis in this study were the new conditions developed between Time 0 and Time 2 years, the actual diagnoses and number of co-existing chronic conditions and the number of regular medications at the time closest to Time 2 years.

Statistical analysis

All analyses were conducted using Statistical Package for the Social Sciences (SPSS). Regression analyses were performed to analyse the association between the number of co-morbid conditions and HbA1c level and the degree of public healthcare service utilisation, after adjusting for patient demographic characteristics and the presence of polypharmacy.

Results
Study population

In total, we reviewed the electronic medical records of 386 patients. At Time 0, the mean age was 63.4 years and 56.5% were female. Multi-morbidity was observed in almost all subjects (n=383, 99.2%), with a mean number of co-existing conditions of 3 per person. Polypharmacy was observed in 144 (37.3%) of patients, with a mean of 4.2 drugs per patient in the study population. The mean HbA1c was 7.2% and the mean BMI was 26.0 kg/m² . 34 patients (8.8%) were current smokers (Table 2).

Nature of multi-morbidity

At Time 0 of our study period, of the 386 records reviewed, the majority (n = 242, 62.7%) had 2 or 3 co-existing conditions. The four most common co-existing chronic conditions were hypertension (85.2%), hyperlipidaemia (83.4%), visual condition (24.4%) and chronic kidney disease (24.1%). Other co-morbidities were observed in less than 10% of the study population (Table 3).

Association between number of co-existing conditions
and diabetes control, as measured by HbA1c

The mean HbA1c levels stratified by different categories of number of co-existing chronic conditions among the diabetic patients are shown in Table 4. Using multiple linear regression adjusted for age, sex, BMI, smoking status and presence of polypharmacy, there was no statistically significant association between the number of co-existing conditions and the mean HbA1c (p = 0.138).

Association between the number of co - existing
conditions and healthcare utilisation

Figure 1 compared the patterns of public healthcare service utilisation among the subjects with different numbers of co-existing chronic conditions. The mean number of annual attendances in the General Out-patient Clinics, Specialist Out-patient Clinics, Accident and Emergency Departments and the mean of annual unplanned hospital admission episodes were 4.18, 2.18, 0.44 and 0.16 per person respectively.

A poisson regression was used to predict the annual attendances in the public healthcare services based on the number of co-existing conditions. After adjusting for age, sex, BMI, smoking status and the presence of polypharmacy, there were statistically significant associations between the number of co-existing conditions and the number of visits to Specialist Out-patient Clinics, Accident and Emergency Departments, as well as the number of unplanned hospital admission episodes. As presented in Table 5, for every extra number of chronic co-existing condition, 1.41 (95% CI, 1.312-1.516; p ＜ 0.001) times more Specialist Out-patient Clinic attendances, 1.57 (95% CI, 1.292-1.907; p ＜ 0.001) times more Accident and Emergency Department attendances and 1.66 (95% CI, 1.208-2.281, p ＜ 0.05) times more unplanned hospital admission episodes were observed in the following year.

5 patients (1.3%) died within the 2-year reporting period. The mean age of death was 75 years (range: 65 – 90 years), and all the deceased were non-smoker. They had a mean number of co-existing chronic conditions of 4.2 per person and a mean BMI of 25.9 kg/m².

Discussion

This is the first-ever study evaluating the magnitude of multi-morbidity and its relationship with glycaemic control and healthcare service utilisation in diabetic patients in a Hong Kong Primary Care Outpatient Clinic. We managed to achieve our primary objectives by collecting and analysing the data we set out to achieve at the start of the project.

Similar to the findings reported in studies conducted on Caucasian subjects^4,6-9, the majority of our study population had multi-morbidities. Hypertension was observed to be the commonest co-morbid condition. Almost all (99.2%) of our study cohort had at least one co-morbid condition. This incidence rate was the highest among all published in the literatures, and even higher than that in a similar study which only included older adults (66 years of age and older) with diabetes by Gruneir et al.⁶

Contrary to the first part of our hypothesis, there was no statistically significant association between the number of co-morbidities and the glycaemic control (as measured by HbA1c levels). This is in line with the finding shown in the study of a cohort of patients with Type 2 diabetes attending general practice in Ireland by Teljeur et al.⁴ This finding could be explained by the presumption that increasing number of co-morbidities is associated with increased patient awareness of their health condition. This might lead to better adherence to lifestyle measures and drug taking, resulting in better glycaemic control. As presented in Table 4, patients with 1 co-existing condition had the highest mean HbA1c of 7.28%, and other groups had insignificantly lower HbA1c. A possible explanation is that, with the comprehensive management programme and the extensive coverage of diabetes medications in the drug formulary of General Out-patient Clinics, a mean HbA1c of less than 7.3% was achieved, regardless of how many co-existing chronic conditions the patients have.

Other studies classified co-morbid conditions as diabetes concordant or discordant, depending on whether they shared the same pathophysiologic risk profile of diabetes and therefore more likely to be part of the same management plan.^8,10 The study by Magnan et al¹⁰ on 7 health systems in Midwestern US concluded having more concordant conditions made diabetic care goal achievement more likely.

We did show a significant positive relationship between the number of co-existing conditions and all parameters of healthcare service utilisation in secondary care. The increasing number of co-morbid conditions was associated with more attendances to Specialist Out-patient Clinics and Accident and Emergency Departments, as well as more unplanned hospital admission episodes. There was, however, no association between the number of co-morbidities and the number of attendances to the General Out-patient Clinics, and this is in contrast to the result found in Teljeur et al₄, which showed that the general practitioner visits increased significantly with the increasing numbers of chronic conditions. Our study findings could be explained by the distinctive scopes of public services provided by primary and secondary care in Hong Kong. While the drug formulary of General Outpatient Clinics of our cluster does cover the drugs for most medical conditions, drugs for certain conditions are only available in secondary care. For example, anti-psychotic medications for schizophrenia and disease-modifying anti-rheumatic drugs for rheumatoid arthritis can only be prescribed in psychiatry clinics and rheumatology clinics respectively. For a patient to be transferred from secondary care to primary care, certain criteria have to be met. In our opinion, this is beneficial for both the public health care provider and the population, as this can ensure the patients are being followed up by the Specialist Out-patient Clinics if indicated. Furthermore, primary health care will not be overwhelmed, and its resources can be allocated to a broad range of services, including health promotion, prevention of acute and chronic diseases, health risk assessment, self-management support and adjusting its services in unforeseen circumstances such as pandemics.

Other studies had shown that utilisation across all health services increased with the number of comorbid conditions.^6,7 This could be explained by the different structures of healthcare systems in different countries.

It is of interest to note that 3 out of the 5 deceased subjects had cancers of the gastro-intestinal system diagnosed within the reporting period. The causes of the 3 deaths were metastatic rectal cancer, metastatic pancreatic cancer and metastatic sigmoid cancer. In recent years, there had been emerging evidences showing the link between Type 2 diabetes and increased risk for cancer of gastro-intestinal tract.^17-19

Strengths and limitations

Firstly, the major strength was that we did not rely solely on the preliminary data drawn from the computerised system of electronic record (CDARS). Instead, the investigators reviewed each record individually. This significantly minimised the potential for misclassification and / or underestimation of chronic conditions. Although reporting of discrepancies between the actual health conditions of the patients and the information in the records was outside the scope of our objectives, it is worth pointing out that the investigators frequently identified chronic conditions which were not coded appropriately in the patient records, and these discrepancies would hence not have been identified via CDARS. Secondly, compared to similar researches conducted with Caucasian subjects, our study was unique as we considered polypharmacy as a potentially important confounding factor. In addition to the patient demographic characteristics, we adjusted the analyses for the presence of polypharmacy. Thirdly, although our list of chronic conditions was not exhaustive, we included a comprehensive set of 28 conditions.

However, several limitations should be mentioned. Firstly, this was a cross-sectional study, hence, we could only assess the associations of multi-morbidity with glycaemic control and healthcare service utilisation. We could not draw definite cause and effect relationships between our outcome measure parameters. Secondly, although the Electronic Health Record Sharing System (eHRSS) launched by Hong Kong Government in 2016 allows sharing of patient medical records between the public and private healthcare sectors, participation in the eHRSS is not mandatory for both healthcare providers and patients. Consequently, there is the potential for missing data should the patients attend both public and private sectors for medical consultations. Thirdly, this study was conducted in a single General Out-patient Clinic, generalisability of results hence could not be concluded from it. Fourthly, this study only analysed the public healthcare service utilisation by the number of visits for medical consultation and unplanned hospital admission episodes, and the actual public healthcare service utilisation should also include other parameters such as nursing and allied health service, duration of hospital admissions, prescriptions as well as diagnostic investigations. Lastly, HbA1c was the only diabetes

care goal that we included in our study, while actual management of diabetic patents is more complex, which entails other key indices including blood pressure, lipid profile, presence and degree of retinopathy, and others.

Implications and future research

This study evaluated the extent of multi-morbidity among diabetic patients attending a General Out-patient Clinic in Hong Kong, and the majority of these patients were Chinese. The result of this study should be useful in developing diabetic care guidelines and intervention in Primary Care Service in Hong Kong.

It is more often the case that a diabetic patient suffers from other co-morbid conditions, and this result underscores the importance of knowing the multimorbidity pattern in patients with Type 2 diabetes, and the need for a multidisciplinary approach to address their complex health care needs and ensure optimisation of their medical treatment.

This study tested the feasibility and identified challenges of such potential research work being conducted within the Hong Kong public healthcare setting. One challenge as already mentioned is that, despite the launching of eHRSS, there is still limitation of medical information sharing between the public and private healthcare providers. This consequently creates a major obstacle, should anyone wish to perform an epidemiological study on multi-morbidity in Hong Kong. Another limitation is the accuracy of medical record as the documentation and coding may vary among clinicians.

Using a single-disease framework to set management guideline for chronic diseases, such as diabetes, is rather out-dated. There is a need for further researches to examine the impact of multimorbidity on the achievement of diabetic care goals and the degree of healthcare service utilisation, which in turn will be useful for strategic resource planning and education.

We were aware that analysing the degree of multimorbidity by the total number of comorbid conditions would not give an account of the nature of the co-existing diseases. The majority of diabetic patients have 2 to 3 other co-existing conditions. Therefore, future multi-morbidity researches exploring commonly co-occurring disease combinations, are likely to provide more meaningful insights into the complex care needs of individuals with multiple chronic conditions.

Conclusion

Our study has illustrated again the fact that diabetes rarely occurs as a single chronic condition. The occurrence of multi-morbidity is common in Type 2 diabetic patients who are under the care of public primary care out-patient clinic in Hong Kong. Accordingly, it is imperative to use a multi-morbidity framework, not only to deliver a comprehensive patient-centred programme, but also to optimise diabetic care in order to reduce treatment burden and unplanned care.

Acknowledgement

We acknowledged the support from the research team of the Department of Family Medicine and Primary Health Care, New Territories West Cluster, the Hospital Authority of Hong Kong.

This study was approved by the Cluster Research Ethics Committee of New Territories West Cluster, the Hospital Authority of Hong Kong.

We do not have any conflict of interest to declare.