What patient-specific factors can potentially affect physiotherapy attendance of patients with knee OA at a local hospital in Singapore?

Introduction

Osteoarthritis (OA) contributes to disability levels substantially, with increased morbidity and reductions in the quality of life, substantially burdening the global healthcare systems.^1,2 In 2020, it was estimated that around 595 million people suffered from OA worldwide, or around 7.6% of the global population. This represents a significant increase of approximately 132.2% in OA cases since 1990.³ In Singapore, the number of people living with knee OA was estimated at 10.1% of the national population in 2007.⁴ Given Singapore’s ageing population and the increase in obesity rates over time, the prevalence of knee OA is expected to increase further, thus presenting a major problem for Singapore’s healthcare system to address.

Current care models available to help patients with knee OA in Singapore may be inadequate for their needs. A review of 37 models of care for OA internationally showed that a majority of them promoted self-management interventions in the primary care setting, and many were generic without individualisation.⁵ The care models, or OA management programmes, included in this review were heterogeneous, with some focussed on a patient’s first specialist clinic visit (4/37) while the others were rehabilitation pathways (10/37). However, it is more common for patients with knee OA seeking care in the public health system in Singapore to be referred on by their specialists to continue non-surgical treatment at outpatient physiotherapy. To address the existing care gaps in quality knee OA care in Singapore, the Collaborative Model of Care between Orthopaedics and Allied Healthcare Professionals (CONNACT) programme was developed to deliver evidence-based, individualised, multidisciplinary treatments for patients with knee OA in the community.⁶ The 12-week CONNACT programme aimed to improve the clinical outcomes of patients suffering from knee OA through non-surgical care. Strengths of the CONNACT programme were delivering care away from the tertiary healthcare settings and its targeted approach through individualisation. CONNACT patients’ psychological profile and body mass index (BMI) were assessed to determine whether psychology and dietetics intervention were included in their care programme. The results from this randomised controlled trial (RCT) have recently been published, and CONNACT participants had promising benefits in functional movements, dietary habits, perceived treatment effects and health-related productivity levels.⁷

While there is on-going work to improve the quality of care for patients with knee OA with individualised programmes in Singapore, localised evidence demonstrating the influence of patient-specific factors on treatment attendance behaviour is lacking. A systematic review of 20 high-quality international cohort studies explored the potential barriers to treatment attendance of patients with diverse health conditions at the physiotherapy outpatient clinics, none of which were based in the Asian settings.⁸ Despite this, this review found that physiotherapy adherence, commonly monitored via patient attendance, has multifactorial influences. These can include pain experienced during exercise, levels of physical and social activities, self-efficacy, presence of psychological disorders, family support and greater perceived barriers to exercise, amongst others. Poor treatment adherence can often lead to poorer health outcomes and higher healthcare costs, as the intended treatment is not being fully received by the patient.⁸ Thus, it is important that patients with knee OA have optimal treatment adherence to obtain the full benefits of their treatment, and this may be potentially achieved with individualised care models targeting patient-specific factors influential of patient attendance.

Therefore, it is important that we understand and quantify the magnitude (including directionality) of the impact of patient-specific factors on physiotherapy attendance behaviours in an Asian setting. Hence, this study aims to identify the patient-specific factors associated with attendance, and quantify the extent of impact, among patients with knee OA who attended physiotherapy at a local hospital in Singapore. We hypothesised that there are potential patient-specific factors that can be identified that are associated with physiotherapy attendance among the patients with knee OA. As patient attendance behaviours in care models in Singapore are not fully understood at present, we anticipate that the identification of such factors relevant for optimising the attendance behaviours can support the development of informed strategies for targeting the improvement of current care models.

Methods

Study design

A retrospective, cohort study using primary data from an RCT of the aforementioned Collaborative Model of Care between Orthopaedics and Allied Healthcare Professionals Trial [National Healthcare Group Domain Specific Review Board (DSRB) References: 2022/00088 and 2018/00408; Trial Registration No. NCT03809975] was conducted. In the CONNACT RCT, enrolled patients were randomised into the CONNACT or standard-care arm (Fig. 1).

Study population

As per the CONNACT RCT, the study participants were recruited from the outpatient Orthopaedic Surgery Department in a tertiary referral centre in Singapore, Tan Tock Seng Hospital. The inclusion criteria for the CONNACT RCT are as follows: Participants: (i) had a diagnosis of knee OA from an orthopaedic specialist; (ii) were ≥ 45 years old; (iii) had activity-related knee pain without morning joint stiffness; (iv) had a Kellgren–Lawrence Score greater than 1; (v) had a Knee Injury and Osteoarthritis Outcome (KOOS${}_4)\le$75; and (vi) were community-ambulant. Exclusion criteria included patients with an alternative diagnosis to knee OA, patients with other forms of OA, patients unable to comply with the study protocol (e.g., cognitive impairment), patients with previous knee arthroplasty and wheelchair-bound patients or patients with medical conditions that will medically impede the study involvement. Further details of the CONNACT RCT can be found in Ref. 6. For this study, we focussed on the RCT participants who were randomised to standard care. This was because the attendance behaviour of standard-care participants was not intervened upon during the conduct of the RCT, and more reflective of the existing attendance patterns of patients with knee OA receiving routine care in the Singapore’s public health system. Thus, patient-specific factors of these patients were likely to be more influential on the observed number of physiotherapy sessions attended by them. This is in contrast with CONNACT participants, who were enrolled in the CONNACT programme with a pre-determined number of physiotherapy sessions (eight sessions over 12 weeks). Hence, we believed that focussing on standard-care participants would provide more informative knowledge on the understanding of patient-specific factors associated with physiotherapy attendance.

Outcomes

We examined the number of sessions attended by CONNACT RCT standard-care participants at the physiotherapy outpatient clinic of Tan Tock Seng Hospital during the 12-week follow-up period as our primary outcome (Fig. 1). Using physiotherapy visit date(s) information logged in the hospital appointment management administrative system, the physiotherapy attendance records of standard-care participants were checked and recorded by a research assistant. Baseline data on a variety of patient-specific factors were collected. The examination of existing literature identified several patient-specific factors theorised to influence the physiotherapy attendance behaviours of patients with knee OA.^8,9,10,11,12 The factors that were considered were as follows:

Except for BMI, physical function tests, data on the other specified factors were obtained via the primary data collected through questionnaire responses. Specific questionnaires or measurements used are elaborated on in the following, and more details can also be found in the CONNACT RCT protocol.⁶

Demographic factors

We collected specific aforementioned demographic information of the participants.

Medical factors

We measured participants’ height and weight to calculate their BMI score. We collected information on the patient’s co-morbidities and classified the medical conditions to determine their Charlson Co-morbidity Index score. We asked whether participants had previous knee joint injections to provide relief and reduce inflammation and pain, normally via visco-supplementation or corticosteroid injections.¹³ These injections must have been done at least three months prior to the baseline. We also recorded analgesia consumption using the Cumulative Analgesia Consumption Scale (CACS) for the participants’ knee OA condition.

Self-reported knee function

The KOOS questionnaire is a self-reported questionnaire used to assess knee function.¹⁴ In the CONNACT RCT, the KOOS₄ composite questionnaire was used. The KOOS₄ composite questionnaire is a modified version of KOOS, without the “Function in Sports and Recreation” domain. This was done as the target population were older adults who were unlikely to be involved in sports and recreation, thus, this domain was deemed to have lower relevance for this population. The other functional domains covered in the KOOS₄ questionnaire were the “Symptoms”, “Pain”, “Function in Daily Living” and “Quality of Life” domains. In our study, we focussed only on the KOOS₄ “Overall”, “Pain” and “Function in Daily Living” domains.

Physical function tests

Participants were administered two different physical tests to assess their functional capabilities. First, the 30-second Chair Stand Test (CST) was used to record both the number of sit-to-stand repetitions the participant achieved in 30s, and the time taken for five complete sit-to-stand repetitions.¹⁵ Next, the Stair Climb Test (SCT) was used to measure the overall time taken for a participant to climb up and down four steps.¹⁶ Additionally, for both CST and SCT, the participants had to rate the pain levels in their knee immediately after the respective tests. Through these functional assessments, we were able to assess the participants’ levels of physical knee function.

Physical activity levels

We used the UCLA activity score, which is a validated scale used to measure physical activity.¹⁷ On a scale ranging from 1 to 10, the participants can indicate their activity levels, with lower levels representing lower activity levels and vice versa for higher activity levels. We further categorised the UCLA activity scores into the following segments: inactive (1–2), low physical activity level (3–4), moderate physical activity level (5–6), high physical activity level (7–8) and very high activity level (9–10), following a study that also focussed on the Asian population suffering from knee osteoarthritis.¹⁸

Psychological factors

To assess the participants’ psychological mindset, two different questionnaires were used: (1) PEG scale and (2) PHQ-4.^19,20 The PEG scale is a brief three-item scale used to assess pain intensity and pain interference experienced by the patients who suffer from musculoskeletal disorders.¹⁹ Higher PEG scores would be indicative of higher pain intensity and interference for the individual and therefore, they should be targeted for psychological intervention. The PHQ-4 is a brief screening scale used to assess anxiety and depression levels. Higher scores of PHQ-4 would be an indication that someone potentially suffers from either anxiety or depression.²⁰

Statistical analysis

No sample size calculation was done for this study, however, the expected number for this study based on the recruited CONNACT RCT participant numbers was around 50 per intervention arm. We computed the descriptive statistics for the participants’ physiotherapy attendance and outcomes collected for patient-specific factors that we identified previously for this study. Means and standard deviations (SDs) for continuous variables and percentages for categorical variables were tabulated. Multiple ordered logistic models were carried out to examine the associations between the particular patient-specific factors that we collected, separately, and physiotherapy attendance. In the first ordered logistic regression model that we ran for our primary outcome of physiotherapy attendance, we included adjustments for age, gender, BMI, KOOS₄ “Pain” and KOOS₄ “ADL” as the base model. These variables were selected as the base model adjustments as they are potential confounders influencing physiotherapy attendance and associated with other patient-specific factors, that are usually readily available to physiotherapists without physical testing requiring extensive equipment setup or training to assess the participants. After this, other patient-specific factors we wanted to specifically study were then added to the base model one by one separately, so that we could study their impact (magnitude and directionality), individually, on our outcome (physiotherapy attendance) as the exploratory variables. Factors with p-values less than 0.05 in the multivariate ordered logistic regression models performed were considered in our study.

Results

Participant characteristics

Among the 90 CONNACT RCT enrolees, 46 participants were randomised to receive standard care. Two were excluded from the analysis due to missing physiotherapy attendance data. On average, 44 standard-care participants attended 2.7 physiotherapy sessions ($\mathrm{\text{SD}}=1.4$) over the 12-week follow-up duration (Fig. 2).

The characteristics of the 44 standard-care participants in the CONNACT RCT included in this analysis, based on their demographics, medical factors, self-reported knee function, physical function tests, physical activity levels and psychological factors, are depicted in Fig. 3. We found that the participants were primarily females (35/44) and were on average around 72 years old. The average BMI of the participants was approximately 26.7 kg/m² ($\mathrm{\text{SD}}=4.8$). The average Charlson Co-morbidity Index score of the participants was 0.28 ($\mathrm{\text{SD}}=0.66$).

We found that five participants had received previous knee injection therapies. Around two-thirds (29/44) of the participants reported that they take analgesia for their knee OA condition, and amongst these participants, 27 and 21 of them were taking paracetamol and supplements, respectively.

Participants had the mean scores of 64.9 ($\mathrm{\text{SD}}=13.8$), 67 ($\mathrm{\text{SD}}=16$) and 58.5 ($\mathrm{\text{SD}}=10.3$) for the KOOS₄ “Pain”, “ADL” and “Overall” domains, and had the average pain levels of 3.6 ($\mathrm{\text{SD}}=3$) and 3.5 ($\mathrm{\text{SD}}=2.8$) for the functional tests namely CST and SCT, respectively. A majority of the participants (29/44) had the UCLA scores between 3 and 4, indicating low physical activity levels. Participants were also found to have the mean scores of 2.8 for PHQ-4 and 5.2 for PEG scale questionnaire, indicating no anxiety or depression symptom levels, but moderate pain interference levels.

Multivariate analysis

The results of the multiple ordered logistic regressions performed can be found in Table 1. Among the 17 patient-specific factors shortlisted from the univariate analysis, we found that six factors were also associated with physiotherapy attendance levels in our multivariate analysis (Fig. 4). First, females were 86% less likely to have higher physiotherapy attendance {$\mathrm{\text{Odds Ratio (OR)}}=0.14$ [95% $\mathrm{\text{confidence interval (CI)}}=0.031$–$0.68$]; $p=0.014$} compared to males. Second, an interquartile range (IQR) increase in BMI decreased the odds for higher attendance by 73% [$\mathrm{\text{OR}}=0.27$ (95% $\mathrm{\text{CI}}=0.123$–$0.587$); $p=0.001$] compared to those with BMI below 23.5kg/m². Third, participants who had previous joint injections were 91% less likely to have higher attendance [$\mathrm{\text{OR}}=0.09$ (95% $\mathrm{\text{CI}}=0.011$–$0.674$); $p=0.019$] compared to those who did not have previous knee injections. An IQR increase in the Chair Stand Test “Pain” scores decreased the odds for higher attendance by 86% [$\mathrm{\text{OR}}=0.14$ (95% $\mathrm{\text{CI}}=0.035$–$0.586$); $p=0.007$] compared to those with lower “Pain” scores. An IQR increase in the Stair Climb Test “Pain” scores decreased the odds for higher attendance by 71% [$\mathrm{\text{OR}}=0.29$ (95% $\mathrm{\text{CI}}=0.093$–$0.903$); $p=0.033$] compared to those with lower “Pain” scores. Lastly, higher anxiety and depression (an IQR increase in PHQ-4 score) were associated with 2.90 times higher odds for higher attendance [$\mathrm{\text{OR}}=2.90$ (95% $\mathrm{\text{CI}}=1.11$–$7.53$); $p=0.029$], compared to those with lower levels of anxiety and depression.

Discussion and Conclusion

We have identified several factors, across different aspects, that may potentially affect the physiotherapy attendance of patients with knee OA. Specifically, the physiotherapy attendance levels were lower for females, or those with higher BMI, previous joint injection therapy, consumption of analgesia, higher KOOS₄ pain scores or moderate physical activity levels. Additionally, we found that those with higher anxiety and depression symptom levels were more likely to attend more physiotherapy sessions.

Our finding regarding BMI and pain during physical exercise largely echoes the findings from other studies that focussed on patients with osteoarthritis or those who attended physiotherapy.^8,21,22 These studies reported that patients with higher BMI tend to have lower treatment attendance, while conversely those with lower BMI have higher attendance. It is hypothesised that high BMI can affect one’s mobility.²³ This can potentially explain why patients with high BMI have lower physiotherapy treatment attendance, as these participants may find exercise or travelling to the physiotherapy clinics more effortful, thus dissuading them from attending. Furthermore, we found that participants who experience higher levels of pain during physical exercise attend less physiotherapy sessions compared to those with lower levels of pain. This mirrors the findings from other studies, which reported that patients with OA who experienced improvements in pain had positive predicted exercise behaviours while those suffering from fibromyalgia who experienced increasing levels of upper body pain had worst management of aerobic activities.⁸ As participants who experience more pain during the exercise will not want to continue attending the exercise sessions in fear of the pain, this might explain why our patients who suffer from higher levels of pain have lower physiotherapy treatment attendance.

Our findings relating to gender, psychological symptoms and previous knee injections were different from what had been reported by previous studies. In our study population, female participants attended fewer treatment sessions, compared to males, unlike other studies on physiotherapy and primary care patients, which found that males had lower attendance.^24,25 However, more recent studies focussing on physiotherapy treatment attendance had similar gender-based results as ours.^10,26 We speculate that for our participants with knee OA, who were predominantly non-working, elderly, females, many of them might be taking the homemaker role in their families, with less time to attend physiotherapy as they might be attending to domestic or household matters instead.²⁷ Also, we found that our study participants who had higher levels of anxiety and depression symptoms attended more physiotherapy sessions compared to those with lower symptom burden. This was again different from previous studies, which reported that high levels of these mental health symptoms for patients led to their non-attendance for the pelvic floor or weight management programmes.^28,29 Participants in this study with higher anxiety and depression symptoms might view their knee condition to be serious and be more inclined to attend physiotherapy treatment. Additionally, we found that participants who had previous knee injections had lower physiotherapy attendance compared to those who did not have knee injections. This finding differs from another study in the US which found that patients with osteoarthritis who had knee injections were more likely to attend physiotherapy, unlike those without knee injections.¹¹ This contrast may be attributed to several factors. First, our study was conducted among an Asian cohort, and health-related behaviours and choices of our participants might be very different from those in the other study, as the preference between active treatments (like physiotherapy) and passive modalities (like knee injections) could be culturally influenced. Second, the time frames (three months versus four months) considered and injectables (corticosteroid/visco-supplementation versus corticosteroid only) used were different in both studies. Corticosteroid injections can typically last for two months,³⁰ while visco-supplementation can last for up to six months,³¹ so the varying durations of effectiveness of the different knee injections may contribute to contrasting impacts on the attendance behaviours in these studies.

Future research, strengths and limitations

Research that can potentially improve the physiotherapy attendance of these patients and clinical outcomes should include targeting patients at risk of low physiotherapy attendance, to support them pro-actively in the rehabilitation journey for knee OA, and the design and evaluation of knee OA programmes that can be easily tailored to patient-specific factors linked to physiotherapy attendance.

We analysed multiple patient-specific factors from different theoretical domains, to understand the patient attendance behaviours from a broad perspective. From our results, we could explore the similarities and differences in physiotherapy attendance behaviours of patients in an Asian setting, compared to other Western settings, that the healthcare practitioners in Asian settings could find useful. Importantly, this study adds to the body of knowledge regarding physiotherapy attendance behaviours by bringing a unique cultural slant, representing the diversity of the knee OA population globally. However, this study is not without limitations. The small sample size, with only 44 patients studied, had limited choices of statistical methods deployed. Furthermore, our study only consisted of Chinese patients, therefore the findings in this study might not be a true representation of the broader Singaporean knee OA population. Lastly, we did not manage to collect other pre-morbid functions, like the history of falls, or socio-economic information, like income levels, which can also be important considerations for treatment attendance in our study setting.

Conclusion

We have identified potential patient-specific factors that may be influential on the physiotherapy attendance levels of patients with knee OA in Singapore. These factors were gender, BMI, pain during physical function, previous knee injections and psychological symptoms. With this knowledge, the healthcare practitioners and researchers alike may be able to develop programmes more suitable for the knee OA patient population in Singapore to adhere to, for better health outcomes.

Conflict of Interest

The authors have no conflicts of interest relevant to this paper.

Funding/Support

This study is funded by the National Medical Research Council of Singapore — Health Services Research New Investigator Grant.

Author Contributions

Michelle Jessica Pereira and Bryan Yijia Tan contributed to the conception and design of the study. Michelle Jessica Pereira and Chien Joo Lim assisted in the acquisition of data, analysis and data interpretation. The contributions of Chi Min Ryan Cheok are in Data Analysis, Manuscript drafting and Revision. Michelle Jessica Pereira, Bryan Yijia Tan, Chien Joo Lim and Yong Hao Pua assisted in the revision of the paper.

ORCID

Chi Min Ryan Cheok  https://orcid.org/0009-0007-9978-4560

Chien Joo Lim  https://orcid.org/0000-0001-7234-9787

Bryan Yijia Tan  https://orcid.org/0000-0002-2794-703X

Yong Hao Pua  https://orcid.org/0000-0003-2313-9665

Michelle Jessica Pereira  https://orcid.org/0000-0002-4251-6086