Prevalence of complexity in primary care and its associated factors: A Singapore experience

ABSTRACT

Introduction: As the population ages, patient complexity is increasing, intensifying the demand for well-resourced, coordinated care. A deeper understanding of the factors contributing to this complexity is essential for optimising resource allocation. This study evaluates the prevalence of complex care needs in Singapore’s primary care settings and identifies the factors associated with these needs.

Method: Using a qualitative study design, we developed a patient complexity questionnaire to assess how Singapore family physicians recognise patient complexity. Sixty-nine experienced primary care physicians applied this tool to assess patient encounters, categorising each as “routine care” (RC), “medically challenging” (MC), or “complex care” (CC). We compared the care needs across these categories and used mixed-effects multinomial logistic regression to determine the independent predictors of complexity.

Results: Of the 4327 encounters evaluated, 15.0% were classified as CC, 18.5% as MC, and 66.4% as RC. In both CC and MC encounters, the most common medical challenges were polypharmacy (66.2% in CC, 44.9% in MC); poorly controlled chronic conditions (41.3% in CC, 24.5% in MC); and treatment interactions (34.4% in CC, 26.0% in MC). Non-medical issues frequently identified included low health literacy (32.6% in CC, 20.8% in MC); limited motivation for healthy lifestyle behaviours (27.2% in CC, 16.6% in MC); and the need for coordinated care with hospital specialists (24.7% in CC, 17.1% in MC). The top 3 independent predictors of complexity included mobility limitations requiring assistance (odds ratio [OR] for requiring wheelchair/trolley: 7.14 for CC vs RC, 95% confidence interval [CI] 4.74–10.74); longer consultation times with physicians (OR for taking >20 minutes for doctor’s consultation: 3.96 for CC vs RC, 95% CI 2.86–5.48); and low socioeconomic status (OR for living in 1- or 2-room HDB flats: 2.98 for CC vs RC, 95% CI 1.74–5.13).

Conclusion: High care needs, encompassing both CC and MC encounters, were prevalent in primary care interactions. These findings highlight that relying solely on chronic disease count is insufficient to capture the full spectrum of patient complexity.

CLINICAL IMPACT

What is New

• The prevalence of complexity in primary care was 15.0%, with another 18.5% of encounters deemed medically challenging.
• Complex care encounters were associated with longer doctor, nursing and allied health consultation time with higher overall cost of care.

Clinical Implications

• Patients with complex needs require adequate consultation time and an integrated, well-coordinated, team-based care approach.

---



As Singapore’s population ages, the incidence of non-communicable chronic diseases will rise in tandem.¹ There exists a subset of individuals, oftentimes elderly with multiple chronic conditions, whose care needs are particularly complex.² While there is no universal definition of a person with complex needs, these individuals have been found to consume more healthcare resources, yet often have poorer health outcomes.^3-5 There is a need to improve care for these individuals, but much is unknown about their prevalence, their needs and the ways to improve their outcomes, particularly in Asian settings.⁶

The prevalence of complexity in primary care varies greatly internationally, ranging from 3.98% to 42.5%.^7-11 These differences stem from discrepant healthcare contexts across localities, as well as varying definitions of complexity used.¹² For instance, a Spanish study¹¹ that found a complexity prevalence of 3.98% relied primarily on the clinical judgment of general practitioners among primary care patients over 14 years old with chronic conditions, while an English study¹⁰ that employed a model with a predetermined definition of complexity derived using a Delphi methodology arrived at a complexity prevalence of 41.6–42.5%.

In Singapore, polyclinics are large publicly-funded primary care clinics with a comprehensive suite of services including medical, nursing, allied health, laboratory and basic radiology. Many patients with chronic conditions choose polyclinics for their care due to low out-of-pocket costs and care accessibility.¹³ An estimated 6.9 million visits were made to polyclinics in 2023, and the top conditions were hyperlipidaemia, hypertension and diabetes mellitus.¹⁴ With the high volume of care for chronic conditions at polyclinics, we hypothesise that prevalence of complexity is also high.

In terms of definition of complexity, some studies and health systems equate complexity with multimorbidity.¹⁵ The Singapore Ministry of Health (MOH), for instance, uses number of chronic conditions to define complexity.¹⁶ However, most experts recognise that non-biomedical aspects like mental health, socioeconomic factors and patient preferences contribute to complexity and should be considered.¹² To better account for complexity, multiple frameworks^17-19 and measurement tools have been developed. For example, a Canadian group developed a complexity framework with 5 health dimensions: demographics, mental health, social capital, medical/physical health, and health and social experiences, that considers a patient’s sociopolitical and physical environmental contexts.¹⁹ While frameworks are helpful for understanding complexity conceptually, they are less useful for clinical implementation. Additionally, many frameworks highlight the influence of cultural, community and local healthcare system factors on patient complexity.^17,19 These factors differ from country to country, underscoring the importance of contextual evidence.²⁰ The paucity of Asian studies on this topic is a significant barrier to understanding and addressing these complex care (CC) needs in the local setting.^12,20,21

Some studies have developed various tools to quantify complexity, such as the Cumulative Illness Rating Scale (CIRS), INTERMED and Patient Centered Assessment Method (PCAM).^22-24 However, these tools come with their own limitations, such as lack of contextual adaptation to the local setting, omission of key factors, and insufficient granularity.

Recognising the limitations of existing tools and acknowledging primary care physicians’ knowledge of complexity—given their unique positions as firstline providers and coordinators of care for many individuals with complex needs,^7,11 we undertook a foundational approach. We decided to first gain a deeper understanding of the nature of complexity in Singapore’s primary care, rather than prematurely applying a measurement tool with limited scope. To this end, we previously undertook a qualitative study to understand how family physicians in Singapore recognise complexity.²⁵ Through this study involving physicians practising in both private and public primary care settings, we found 2 distinct groups of patients that required more clinical effort and resources (i.e. high care needs): patients with CC needs and those with medically challenging (MC) issues. In CC, 1 or more difficult medical issues were encountered, with concomitant issues in 1 or more of the following domains: psychological, socioeconomic and behavioural; and these factors adversely impacted medical care and outcomes. In contrast, with MC, issues were mainly limited to the medical domain. Patients not fitting the above characteristics were considered routine care (RC). The qualitative study findings suggested that while CC and MC patients could both be considered “complex”, CC patients had higher level care needs than MC patients. We were interested to identify the complexity-related factors that may differ between MC and CC.

This study aimed to further characterise complexity by assessing the prevalence of CC needs and identifying factors associated with high care needs (i.e. CC and MC) in a real-world primary care population. Understanding the magnitude (i.e. prevalence) of complexity and its associated factors is a critical first step in guiding future research and informing efforts to effectively manage complexity in primary care.

METHOD

Study design, population and recruitment

Experienced primary care physicians (PCPs) with postgraduate Family Medicine (FM) qualifications—Graduate Diploma in FM (GDFM), Master of Medicine in FM (MMed[FM]) or Fellowship of the College of Family Physicians Singapore (FCFP[S])—from National Healthcare Group Polyclinics (NHGP) were invited to participate in this cross-sectional study, where PCPs code patients that they had seen. We included physicians with FM training and experience, to mitigate the perception of complexity that may stem from underdeveloped professional competencies.¹²

NHGP has 6 polyclinics in central and northwest Singapore, providing subsidised primary care to an estimated population of 1.5 million residents, ranging from acute care services to chronic disease management and women and child health services.²⁶ Each polyclinic is a publicly-funded “one-stop” facility that houses physicians, nurses and allied healthcare professionals, with laboratory and radiological capabilities.²⁷

To determine the prevalence of patient encounters with high care needs in a representative population from NHGP, weighted random sampling was carried out from 24 August 2022 to 30 November 2022. Physicians were recruited from all 6 polyclinics according to the proportion of patients in NHGP seen at each clinic. PCPs were randomly selected using a random number generator and invited via email. Informed consent was obtained.

Patient complexity questionnaire

Based on the findings of a local qualitative study²⁵ and triangulating with other studies,^{3,7,10,12,17-20,28-30} a questionnaire on CC needs in Singapore primary care was formulated with inputs from an expert panel of experienced PCPs over a few meetings. The questionnaire was pre-tested (n=8) and pilot-tested (n=13), with revisions done iteratively before study commencement.

Participating PCPs assigned a complexity rating for each patient encounter, based on 3 categories: RC, MC and CC.  CC needs were also identified and were categorised according to the following domains: (1) medical (7 issues); (2) functional (3 issues); (3) mental health (9 issues); (4) social (4 issues); (5) behavioural (6 issues); as well as (6) resource utilisation (7 issues) (see Supplementary Material Annex A). An encounter was considered CC if there was 1 or more medical domain issues that interacted with a concomitant issue in another non-medical domain (2–6) impacting medical care and outcomes. MC encounters were mainly medical domain issues with little interacting issues in other domains that affected medical care or outcomes. RC encounters were routine encounters that were not considered CC or MC.

PCP participants attended training sessions—each lasting an hour and held virtually over a video-conferencing platform—on questionnaire-filling to standardise the categorisation of CC, MC and RC, using case studies to illustrate each category of patient encounters.

Sample size

The estimated sample size was 3450 according to the following formula³¹: n=[Z²P(1-P)]/d², where Z is the statistic for level of confidence of 1.96, P is the expected prevalence of 42.5% (based on prior studies^7-10), d is precision of 2%,³² and after multiplying a variance inflation factor (VIF) of 1.47 to take into account the possible clustering of data within PCPs,³³ where VIF=1+ICC(cluster size-1), using a conservative intraclass correlation (ICC) of 0.01³⁴ and an estimated cluster size of 48 patients seen per PCP. Assuming each PCP would see 42–54 patients over 4 half-day sessions, 64–82 PCPs were required.

Data collection

Participating PCPs retrospectively assessed the complexity and care needs of patients they saw over 4 half-day sessions, referring to electronic case notes. To reduce recall bias, PCPs completed each questionnaire no later than 1 day after the clinic session. Questionnaires were hosted on REDCap.³⁵ To ensure the confidentiality of PCPs and their patients, participants entered a uniquely-assigned physician code and each patient’s masked identity number into the questionnaires, without any other identifiers. Each PCP participant received reimbursement of SGD60 (USD44) in grocery vouchers for each half-day clinic session of case notes review and questionnaire filling.

We obtained patient characteristics based on routinely-collected data in the NHGP database, including clinical and resource utilisation data in the preceding year. Time of consultation was extracted electronically based on time spent between the successful “call” of the patient’s queue number and the transfer of the patient’s queue number to the next provider. The medical conditions included were selected based on clinical judgment, local practice guidelines (Chronic Disease Management Programme¹⁶) and evidence on multimorbidity,^36,37 and extracted based on coding in electronic medical records (EMR). A member of the data analytics department who was not part of the study team extracted the data and clarified any missing fields on questionnaire with individual participants to ensure data completeness. Only de-identified data were sent to the study team for analysis.

Data analysis

We analysed the prevalence of complex patient encounters and CC needs. We compared EMR-derived sociodemographic, disease status and resource utilisation characteristics between the 3 groups of patients (RC, MC and CC), applying Kruskal-Wallis test for comparison of continuous variables, and Pearson’s chi-squared test or Fisher’s Exact test for categorical variables. Similarly, we compared the frequency of questionnaire complex factors between RC, MC and CC using Pearson’s chi-squared test or Fisher’s Exact test. We developed a mixed-effects multinomial logistic regression model to identify independent predictors of complexity (that were derived from EMR data), comparing RC with MC and CC, while accounting for potential clustering effects at the PCP level, fitted using the Mclogit R package version 0.9.6 (Elff, 2022). We also compared physician-identified complexity with the MOH definition of complexity,¹⁶ which is based on the number of chronic conditions present: simple being 1 chronic condition, moderate being 2 or 3 chronic conditions and complex being 4 or more chronic conditions.

For patients with 2 or more visits recorded, we only considered the first visit to eliminate multiple levels of clustering. We did not detect significant collinearity using VIF. All analyses were performed using R version 4.4.1 (R Core Team, 2021). The National Healthcare Group Domain Specific Review Board approved our study (2022/00226).

RESULTS

Sociodemographic characteristics

Seventy-two PCPs consented to participate (80.2% response rate). Eventually, 69 PCPs (95.8% completion rate) from 6 polyclinics assessed the complexity of 4327 unique patients, after excluding 24 second-visit encounters (see Supplementary Material Annex B). The mean age of patients was 61.2 years (standard deviation [SD] 17.4) with approximately equal proportion of genders. Among them, 73.7% were Chinese, 11.8% Malay, 8.8% Indian and 5.7% others. Socioeconomic status was represented by housing type: 3.6% lived in 1- or 2-room Housing & Development Board (HDB) government-subsidised flats,³⁹ 49.2% lived in 3- or 4-room HDB flats, and the remaining 47.2% in 5-room HDB, private housing and others (Table 1).

Table 1. Sociodemographic and encounter-related information of patients.

Prevalence of patient complexity

In terms of prevalence, 15.0% of patient encounters were CC, 18.5% deemed MC and 66.4% RC (Table 1).

Demographic, medical and resource factors

Demographic and encounter-related factors

Table 1 describes the demographic- and encounter-related factors between CC, MC and RC. Patients with CC encounters were more likely to be older (67.5 years for CC, 65.3 years for MC and 58.7 years for RC, P<0.001), residing in 1- or  2-room HDB flats (7.8% in CC, 2.2% in MC, 3.0% in RC, P<0.001), and require assistance for ambulation (27.7% in CC, 12.6% in MC and 4.4% in MC, P<0.001). Patients with CC encounters had statistically significantly more issues managed (4.06 for CC, 3.51 for MC, 2.19 for RC, P<0.001) and more medications dispensed at the encounter (5.32 for CC, 4.45 for MC, 2.87 for RC, P<0.001).

Table 2. Disease status and resource utilisation data of patients.

Medical conditions

The top 5 medical conditions experienced by patients with high care needs were: hyperlipidaemia (81.4% in CC, 80.3% in MC), hypertension (75.9% in CC, 70.7% in MC), diabetes (60.7% in CC, 51.8% in MC), chronic kidney disease (23.2% in CC, 17.4% in MC) and ischaemic heart disease (19.8% in CC, 17.0% in MC) (Table 2).

Resource utilisation data

Patients with CC encounters utilised significantly more resources across all types of resources studied, including number of medications, cost of care, number and duration of visits to doctors, nurses, medical social workers, psychologists and financial counsellors (Table 2). Doctors spent a mean duration of 17.8 (SD 13.5) minutes attending to CC patients per consultation, compared to 14.9 (SD 9.8) minutes for MC patients, and 12.7 (SD 11.8) minutes for RC patients.

Table 3. Frequency of questionnaire complexity factors.

Prevalence of complex care needs

CC needs in CC and MC encounters were categorised as medical, functional, mental health, social, behavioural and related to resource utilisation, and are described in Table 3.

Medical care needs

The most common medical domain issues among CC and MC patients were polypharmacy (66.2% in CC, 44.9% in MC), poor control of chronic conditions (41.3% in CC, 24.5% in MC) and treatment interactions (34.4% in CC, 26.0% in MC).

Non-medical care needs

The most common non-medical domain issues faced by CC and MC patients were largely from the behavioural and resource utilisation domains. They were poor health literacy (32.6% in CC, 20.8% in MC); poor motivation for healthy lifestyle behaviours (27.2% in CC, 16.6% in MC); required coordination of care with hospitalists (24.7% in CC, 17.1% in MC); refused or non-adherent to medications (22.4% in CC, 10.4% in MC); and required multidisciplinary team intervention (19.2% in CC, 9.7% in MC). 

Table 4. Multinomial logistic regression for predictors of complexity.

Independent indicators of complexity and care needs

Based on our mixed-effects multinomial logistic regression model (Table 4), factors that independently predicted for high care needs (both CC and MC) included: requiring aid for ambulation and/or requiring wheelchair/trolley (odds ratio [OR] 2.63 for MC versus [vs] RC, 95% confidence interval [CI] 1.77–3.91; OR 7.14 for CC vs RC, 95% CI 4.74 –10.74); a greater number of issues managed during the encounter (OR 1.64 for MC vs RC, 95% CI 1.52–1.77; OR 2.16 for CC vs RC, 95% CI 1.96–2.38); longer duration of doctor consultation (OR for taking >20 minutes for doctor’s consultation was 2.03 for MC vs RC, 95% CI 1.56–2.66; OR for taking >20minutes for doctor’s consultation was 3.96 for CC vs RC, 95% CI 2.86–5.48); and higher cost of care over the past 1 year. Factors that independently predicted for CC (but not MC) included living in 1- or 2-room HDB flats (OR 2.98 for CC vs RC, 95% CI 1.74–5.13); a greater number of medications dispensed during the encounter (OR 1.05 for CC vs RC, 95% CI 1.01–1.09); longer total nursing time; and total allied health time. The number of chronic conditions did not independently predict for complexity or high care needs.

DISCUSSION

Our study categorised complexity in 4327 primary care encounters. The prevalence of complexity in our polyclinic sample was 15.0%, with another 18.5% deemed medically challenging. Different demographics, biopsychosocial characteristics and resource utilisation patterns were associated with each group. The independent indicators of complexity included lower socioeconomic status (by housing type)³⁸; requiring assistance for mobility; a greater number of issues managed and medications dispensed during the encounter; longer duration of doctor; nursing and allied health consultation time; and higher cost of care over the past year.

Sociodemographic data of the sample population were reasonably representative of the general population in Singapore. In terms of ethnicity, 73.7% of study sample vs 74.1% of the general population were Chinese; 11.8% of study sample vs 13.6% of the general population were Malay; 8.8% of study sample vs 9.0% of the general population were Indian; and 5.7% of the study sample vs 3.3% of the general population were “others” in 2022.³⁹ For housing status, 3.6% of the study sample vs 6.7% of the general population lived in 1- or 2-room HDB flats; 49.2% of the study sample vs 48.6% of the general population lived in 3- or 4-room HDB flats; and 47.2% of the study sample vs 44.4% of the general population lived in 5-room HDB flats or private housing in 2022.³⁹ Unsurprisingly, the sample population had more chronic conditions than that of the general population—with 69.4% of the study sample vs 31.9% of the general population with hyperlipidaemia; 60.3% of the study sample vs 37.0% of the general population having hypertension; and 39.2% of the study sample vs 8.5% of the general population diagnosed with diabetes.⁴⁰

The combined prevalence of high care needs (CC and MC) of 33.6% in our study more closely mirrored the midpoint of prevalence of complexity reported in other western studies, which ranged from 3.98% to 42.5%.^5,10-12 The differences likely stemmed from different definitions used for complexity. Our study is the first to differentiate complex care from medically challenging encounters. We did so because our qualitative study with PCPs had previously differentiated CC from MC, identifying that persons with CC required a multidisciplinary approach with longer time for consultation.²⁵ The requirement for increased consultation time and resource utilisation was elucidated in our study, with CC encounters requiring a mean duration of 17.8 minutes for their doctor’s consultation, compared to 14.9 minutes for MC patients and 12.7 minutes for RC patients. It is uncertain if the consultation time in our study truly reflected the ideal time required, as the time pressures of a busy polyclinic could have constrained the time that doctors could dedicate to CC and MC encounters. Additionally, longer nursing and allied health consultation times were independent predictors for complexity, highlighting the importance of a multidisciplinary approach in managing complexity.

Asian studies on complexity were limited. They had largely focused on inpatients,^41-43 specific conditions (e.g. alcohol misuse),⁴⁴ validation of specific complexity tools⁴⁵ or equated multimorbidity with complexity.^21,46 Among Asian studies that studied complexity in primary care contexts, 2 were qualitative studies regarding primary care physicians’ perceptions of complexity.^25,47 There was only 1 study⁴⁸ that showed a medium positive relationship between patient complexity (identified via PCAM) and healthcare costs. Our study adds to the literature describing how the use of an intuitive questionnaire could identify individuals with complex needs—who were more likely to have incurred higher cost and consumed more resources. This questionnaire could be easily administered in a form suited to any resource setting—integrated seamlessly with electronic medical records systems, or filled manually via pen and paper. On top of higher resource utilisation, our study further identified other independent indicators of complexity, including lower socioeconomic status, requiring assistance for mobility and a greater number of issues managed during the encounter, providing an opportunity for the pre-emptive identification of these individuals for earlier intervention and prior to resource consumption.

Our study found that while 62.5% of CC and 56.1% of MC patients had 4 or more conditions, using multimorbidity alone to identify complexity would miss a large proportion of patients who required additional care. Rather than the number of chronic conditions, we discovered that complexity was better predicted by the number of issues addressed during the encounter,²⁵ many of which were non-medical in nature. This echoed other studies showing that chronic disease count did not equate to complexity or medical difficulty.^3,7,29,49 Even among RC patients, two-thirds had 2 or more chronic conditions. It appeared that many PCPs considered the management of multimorbidity as routine, reflecting the high prevalence of multimorbidity.^50,51 Policymakers in Singapore tend to identify complexity through multimorbidity, which could lead to inaccurate projections of resource requirements. Our findings suggest that funding models for primary care can be adjusted to account for complexity rather than focusing solely on multimorbidity. Such adjustments would better reflect resource utilisation, including appropriately capturing doctor time required for effective management and funding for multidisciplinary teams. This is particularly relevant in the era of population health and national enrolment with primary care providers under the nationwide Healthier SG initiative.⁵² Our study further identified common complex care needs in CC and MC encounters, yet these needs are often not coded in routine clinical care. More efficient and precise methods of identifying CC and MC encounters using electronic medical records are required to support accurate resource projections and provider reimbursements.

We found that living in 1- or  2-room HDB flats independently predicted for physician-defined complexity, agreeing with previous studies showing similar associations with lower socioeconomic status.^7,19,29 Requirement of assistance for mobility was an independent indicator of physician-defined complexity, perhaps because these individuals require functional support in the form of caregivers or community resources such as day care/rehabilitation centres. The mobility aids could have also served as visual cues for PCPs to explore the impact of their patients’ functional status on their medical care.

The most common non-medical domain issues in CC encounters included poor health literacy, low motivation for healthy behaviours, the need for coordination with hospitalists, medication non-adherence and requiring multidisciplinary intervention. Given the high prevalence of complexity in primary care, family medicine education, particularly at the postgraduate level, must emphasise skill development for managing complex care needs. Findings from our study can inform the identification of specific knowledge, skills, attitudes and partnerships essential for complexity management in the local context, thereby guiding curriculum design with the explicit aim of building learners’ confidence in caring for individuals with complex needs.⁵³ This curriculum should involve broad-based transdisciplinary training, such as enhancing communication skills to tackle health literacy issues, practising motivational interviewing, and adopting team-based care approaches.

Our study corroborated the findings of numerous studies, affirming that complexity is a multidomain concept.^12,18-20 The complexity questionnaire data clearly demonstrated that the issues faced by patients with high care needs are highly heterogeneous, highlighting the necessity of an individualised, rather than one-size-fits-all, approach. This underscores the importance of a holistic, multidisciplinary team-based care approach,²⁶ as well as the necessity of allocating sufficient resources to primary care practices. Such team-based care may need to extend beyond traditional team members, such as nurses, medical social workers and psychologists, to include community partners and social service agencies. For instance, scheduled physical or virtual multidisciplinary case discussions could be organised for patients with particularly complex psychosocial needs, allowing for a more holistic and coordinated approach to care. These resources and collaborations are essential for delivering the comprehensive and customised care required to effectively manage individuals with high care needs.

Strengths and limitations

To our knowledge, this is the first local study investigating the prevalence and associated factors of complexity in primary care. It builds on previous work and provides rich data on an important but previously unidentified group of patients in Singapore. We ensured the robustness of the estimate of prevalence via sample size calculation and undertook measures to standardise the definition of complexity through compulsory briefing sessions for participants. The study was adequately powered and the study sample was fairly representative of the general population sociodemographically. PCPs were randomly selected to mitigate potential biases—at the level of individual physician in recognising complexity, and systemic biases arising from variations in the patient profiles seen by different doctors. We also performed mixed-effects multinomial logistic regression to further account for PCP-level clustering due to variability in the training and experience of individual PCPs.

There are several limitations. First, this is a cross-sectional study and causation cannot be determined. We were unable to investigate patients’ progression and longer-term outcomes. Second, this study involved only PCPs from a public healthcare organisation in Singapore, precluding the generalisability of our findings to other primary care contexts like private general practices in Singapore or primary care practices in other countries. Different settings might have different resources, styles of practice, patient profiles, cultural norms and healthcare infrastructure that would affect the perception of complexity. Third, the identification of issues was limited to what the PCP was aware of or had time to ask about. Fourth, we were unable to obtain the perspectives of patients or other healthcare providers. Fifth, we acknowledge that some of the independent variables used in our multinomial logistic regression model (e.g. ambulation status, housing type, number of chronic conditions) may be closely related to the definition of our outcome measures (i.e. MC/CC, which were based on medical, functional, social and resource utilisation domains being affected). This overlap introduces complexity in fully disentangling the relationships between predictors and outcomes using the current regression framework.

Future research

To better model the complex interplay between predictors and the definition of complexity (MC/CC), a follow-up study that employs advanced analytical techniques, such as Structural Equation Modelling (SEM) can be undertaken. In addition to validating this questionnaire and replicating this study in other family medicine settings, other potential areas of future research could also include cohort studies that examine the relationship between complexity factors and clinical or resource utilisation outcomes prospectively. This questionnaire could also be utilised in interventional studies to identify patients with complex needs and demonstrate effectiveness of clinical programmes.

CONCLUSION

The prevalence of complexity in Singapore primary care was 15.0%, and an additional 18.5% was medically challenging. Our study identified key areas for improving care for patients with complex care needs: adequate consultation time, multidisciplinary teamwork, complexity training in medical education, and refined methods for identifying complexity.

Suppplementary Material Annex A and B

Acknowledgements
This research was supported by seedcorn fund from the Centre for Primary Health Care and Innovation (CPHCRI), a joint initiative between the Lee Kong Chian School of Medicine and the National Healthcare Group (Reference Code: CPHCRI 8.1/009). Funding source was not applied in study design, participant recruitment, data collection, data analysis or data interpretation. We would like to acknowledge Mr Lim Hai Thian’s support in data anonymisation and extraction, as well as Mr Ng Xinyao and Ms Koh Hui Li for assisting with the administrative aspects of the study.

---

References

1. Malhotra R, Bautista MAC, Müller AM, et al. The Aging of a Young Nation: Population Aging in Singapore. Gerontologist 2019;59:401-10.
2. Upshur RE, Tracy S. Chronicity and complexity: is what’s good for the diseases always good for the patients? Can Fam Physician 2008;54:1655-8.
3. Peek CJ, Baird MA, Coleman E. Primary care for patient complexity, not only disease. Fam Syst Health 2009;27:287-302.
4. Maciejewski ML, Powers BJ, Sanders LL, et al. The intersection of patient complexity, prescriber continuity and acute care utilization. J Gen Intern Med 2014;29:594-601.
5. Hong CS, Atlas SJ, Ashburner JM, et al. Evaluating a Model to Predict Primary Care Physician-Defined Complexity in a Large Academic Primary Care Practice-Based Research Network. J Gen Intern Med 2015;30:1741-7.
6. Upshur RE. Understanding Clinical Complexity the Hard Way: A Primary Care Journey. Healthc Q 2016;19:24-8.
7. Grant RW, Ashburner JM, Hong CS, et al. Defining patient complexity from the primary care physician’s perspective: a cohort study. Ann Intern Med 2011;155:797-804.
8. Barcones MF, Amatriain E, Lobo E, et al. Complex patients. J Psychosom Res 2019;121:140.
9. Oliveira CA, Weber B, Dos Santos JLF, et al. Health complexity assessment in primary care: A validity and feasibility study of the INTERMED tool. PLoS One 2022;17:e0263702.
10. Salisbury C, Lay-Flurrie S, Bankhead CR, et al. Measuring the complexity of general practice consultations: a Delphi and cross-sectional study in English primary care. Br J Gen Pract 2021;71:e423-e431.
11. Hernansanz Iglesias F, Martori Cañas JC, Limón Ramírez E, et al. Clustering Complex Chronic Patients: A Cross-Sectional Community Study From the General Practitioner’s Perspective. Int J Integr Care 2021;21:4.
12. Nicolaus S, Crelier B, Donzé JD, et al. Definition of patient complexity in adults: A narrative review. J Multimorb Comorb 2022;12:26335565221081288.
13. Tan CC, Lam CSP, Matchar DB, et al. Singapore’s health-care system: key features, challenges, and shifts. Lancet 2021;398:1091-104.
14. Healthhub.sg. Polyclinic Attendance (Top 4 Conditions). https://www.healthhub.sg/a-z/health-statistics/top-4-conditions-of-polyclinic-attendances. Accessed 8 October 2024.
15. Holzhausen M, Fuchs J, Busch M, et al. Operationalizing multimorbidity and autonomy for health services research in aging populations–the OMAHA study. BMC Health Serv Res 2011;11:47.
16. Ministry of Health, Singapore. Chronic Disease Management Programme Handbook for Healthcare Professionals (Singapore, 2022).
17. Grembowski D, Schaefer J, Johnson KE, et al. A conceptual model of the role of complexity in the care of patients with multiple chronic conditions. Med Care 2014;52:S7-S14.
18. Safford MM, Allison JJ, Kiefe CI. Patient complexity: more than comorbidity. the vector model of complexity. J Gen Intern Med 2007;22:382-90.
19. Schaink AK, Kuluski K, Lyons RF, et al. A scoping review and thematic classification of patient complexity: offering a unifying framework. J Comorb 2012;2:1-9.
20. Zullig LL, Whitson HE, Hastings SN, et al. A Systematic Review of Conceptual Frameworks of Medical Complexity and New Model Development. J Gen Intern Med 2016;31:329-37.
21. Kaneko H, Hanamoto A, Yamamoto-Kataoka S, et al. Evaluation of Complexity Measurement Tools for Correlations with Health-Related Outcomes, Health Care Costs and Impacts on Healthcare Providers: A Scoping Review. Int J Environ Res Public Health 2022;19:16113.
22. Linn BS, Linn MW, Gurel L. Cumulative illness rating scale. J Am Geriatr Soc 1968;16:622-6.
23. de Jonge P, Huyse FJ, Stiefel FC, et al. INTERMED–a clinical instrument for biopsychosocial assessment. Psychosomatics 2001;42:106-9.
24. Pratt R, Hibberd C, Cameron IM, et al. The Patient Centered Assessment Method (PCAM): integrating the social dimensions of health into primary care. J Comorb 2015;5:110-9.
25. Quek JS, Lee ES, Low LL, et al. How family physicians in Singapore recognise complexity during consultations: a qualitative study. BMC Prim Care 2024;25:134.
26. Chong PN, Tang WE. Transforming primary care-the way forward with the TEAMS2 approach. Fam Pract 2019;36:369-70.
27. Mun VL, Chuan TN, Hong CM. Chronic Disease Management In Singapore Polyclinics. Malays Fam Physician 2006;1:94-6.
28. Manning E, Gagnon M. The complex patient: A concept clarification. Nurs Health Sci 2017;19:13-21.
29. Mount JK, Massanari RM, Teachman J. Patient care complexity as perceived by primary care physicians. Fam Syst Health 2015;33:137-45.
30. Loeb DF, Binswanger IA, Candrian C, Bayliss EA. Primary care physician insights into a typology of the complex patient in primary care. Ann Fam Med 2015;13:451-5.
31. Hajian-Tilaki K. Sample size estimation in epidemiologic studies. Caspian J Intern Med 2011;2:289-98.
32. Naing L, Nordin RB, Abdul Rahman H, et al. Sample size calculation for prevalence studies using Scalex and ScalaR calculators. BMC Med Res Methodol 2022;22:209.
33. Rutterford C, Copas, A, Eldridge S. Methods for sample size determination in cluster randomized trials. Int J Epidemiol 2015;44:1051-67.
34. Adams G, Gulliford MC, Ukoumunne OC, et al. Patterns of intra-cluster correlation from primary care research to inform study design and analysis. J Clin Epidemiol 2004;57:785-94.
35. Harris PA, Taylor R, Minor BL, et al. The REDCap consortium: Building an international community of software platform partners. J Biomed Inform 2019;95:103208.
36. Tyagi S, Koh V, Koh GC, et al. Defining and measuring multimorbidity in primary care in Singapore: Results of an online Delphi study. PLoS One 2022;17:e0278559.
37. Fortin M, Almirall J, Nicholson K. Development of a research tool to document self-reported chronic conditions in primary care. J Comorb 2017;7:117-123.
38. Housing & Development Board (HDB). Types of Flats, 2023. https://www.hdb.gov.sg/residential/buying-a-flat/finding-a-flat/ types-of-flats. Accessed 3 December 2024.
39. Department of Statistics, Singapore. SingStat Table Builder. https://tablebuilder.singstat.gov.sg/. Accessed 3 December 2024.
40. Ministry of Health, Singapore. National Population Health Survey (NPHS) 2022 Report. https://www.moh.gov.sg/others/resources-and-statistics/nphs-2022. Accessed 3 December 2024.
41. Yoshida S, Matsushima M, Wakabayashi H, at al. Validity and reliability of the Patient Centred Assessment Method for patient complexity and relationship with hospital length of stay: a prospective cohort study. BMJ Open 2017;7:e016175.
42. Yoshida S, Matsushima M, Wakabayashi H, et al. Correlation of patient complexity with the burden for health-related professions, and differences in the burden between the professions at a Japanese regional hospital: a prospective cohort study. BMJ Open 2019;9:e025176.
43. Yokokawa D, Shikino K, Kishi Y, et al. Does scoring patient complexity using COMPRI predict the length of hospital stay? A multicentre case-control study in Japan. BMJ Open 2022;12:e051891.
44. Sugiyama Y, Matsushima M, Yoshimoto H. Association between alcohol consumption/alcohol use disorders and patient complexity: a cross-sectional study. BMJ Open 2020;10:e034665.
45. Aoki T, Inoue M, Nakayama T. Development and validation of the Japanese version of Primary Care Assessment Tool. Fam Pract 2016;33:112-7.
46. Sugiyama Y, Mutai R, Aoki T, et al. Multimorbidity and complex multimorbidity, their prevalence, and associated factors on a remote island in Japan: a cross-sectional study. BMC Prim Care 2022;23:258.
47. Wong PYA, Chan FYS, Ong L, et al. A qualitative study of challenges and enablers faced by private general practitioners providing primary care to patients with complex needs in Singapore. BMC Prim Care 2022;23:14.
48. Sugiyama Y, Mutai R, Matsushima M. Association between patient complexity and healthcare costs in primary care on a Japanese island: a cross-sectional study. BMJ Open 2023;13:e068497.
49. Weiner SJ, Schwartz A, Sharma G, et al. Patient-centered decision making and health care outcomes: an observational study. Ann Intern Med 2013;158:573-9.
50. Lee ES, Lee PSS, Xie Y, et al. The prevalence of multimorbidity in primary care: a comparison of two definitions of multimorbidity with two different lists of chronic conditions in Singapore. BMC Public Health 2021;21:1409.
51. Low LL, Kwan YH, Ko MSM, et al. Epidemiologic Characteristics of Multimorbidity and Sociodemographic Factors Associated With Multimorbidity in a Rapidly Aging Asian Country. JAMA Netw Open 2019;2:e1915245.
52. Foo C, Chia HX, Teo KW, et al. Healthier SG: Singapore’s multi-year strategy to transform primary healthcare. Lancet Reg Health West Pac 2023;37:100861.
53. Osborn J, Raetz J, Huntington J, et al. A Curriculum on Care for Complex Patients: Resident Perspectives. Fam Med 2016;48:35-43.

 

Ethics statement

This study was approved by the National Healthcare Group Domain Specific Review Board (2022/00226).

Declaration

The author(s) declare there are no affiliations with or involvement in any organisation or entity with any financial interest in the subject matter or materials discussed in this manuscript.

Correspondence

Dr Jing Sheng Quek, Woodlands Polyclinic, National Healthcare Group Polyclinics, 3 Fusionopolis Link, Nexus@one-north, South Tower, #05-10, Singapore 138543. Email: [email protected]