Article Text
PDF
Abstract
Objective To develop and internally validate a prognostic model to predict chronic pain after a new episode of acute or subacute non-specific idiopathic, non-traumatic neck pain in patients presenting to physiotherapy primary care, emphasising modifiable biomedical, psychological and social factors.
Design A prospective cohort study with a 6-month follow-up between January 2020 and March 2023.
Setting 30 physiotherapy primary care practices.
Participants Patients with a new presentation of non-specific idiopathic, non-traumatic neck pain, with a duration lasting no longer than 12 weeks from onset.
Baseline measures Candidate prognostic variables collected from participants included age and sex, neck pain symptoms, work-related factors, general factors, psychological and behavioural factors and the remaining factors: therapeutic relation and healthcare provider attitude.
Outcome measures Pain intensity at 6 weeks, 3 months and 6 months on a Numeric Pain Rating Scale (NPRS) after inclusion. An NPRS score of ≥3 at each time point was used to define chronic neck pain.
Results 62 (10%) of the 603 participants developed chronic neck pain. The prognostic factors in the final model were sex, pain intensity, reported pain in different body regions, headache since and before the neck pain, posture during work, employment status, illness beliefs about pain identity and recovery, treatment beliefs, distress and self-efficacy. The model demonstrated an optimism-corrected area under the curve of 0.83 and a corrected R2 of 0.24. Calibration was deemed acceptable to good, as indicated by the calibration curve. The Hosmer–Lemeshow test yielded a p-value of 0.7167, indicating a good model fit.
Conclusion This model has the potential to obtain a valid prognosis for developing chronic pain after a new episode of acute and subacute non-specific idiopathic, non-traumatic neck pain. It includes mostly potentially modifiable factors for physiotherapy practice. External validation of this model is recommended.
- REHABILITATION MEDICINE
- Musculoskeletal disorders
- PAIN MANAGEMENT
- Prognosis
Data availability statement
Data are available in a public, open access repository. Technical appendix, statistical code and dataset available from the Github repository: https://github.com/uashogeschoolutrecht/painr DOI: available upon acceptance.
This is an open access article distributed in accordance with the Creative Commons Attribution Non Commercial (CC BY-NC 4.0) license, which permits others to distribute, remix, adapt, build upon this work non-commercially, and license their derivative works on different terms, provided the original work is properly cited, appropriate credit is given, any changes made indicated, and the use is non-commercial. See: http://creativecommons.org/licenses/by-nc/4.0/.
Statistics from Altmetric.com
STRENGTHS AND LIMITATIONS OF THIS STUDY
Novel approach to determine an accurate sample size for prognostic model development, mitigating overfitting.
Inclusion of biomedical, psychological and social prognostic factors which are potentially modifiable by a physiotherapist.
Utilisation of three follow-up time points for chronic pain outcome assessment.
Introduction
Neck pain is a widespread and disabling health condition significantly impacting public health.1–3 It is ranked third in terms of years lived with disability in non-fatal diseases, with high costs due to extended work absence and healthcare utilisation.4 Chronic neck pain is particularly costly,5 and the prevalence has increased by 21% from 2005 to 2015, affecting approximately 358 million people worldwide.6 The estimated global number of neck pain cases is projected to be 269 million (219–322) by 2050, an increase of 32·5% (23·9–42·3) from 2020 to 2050.7
Physiotherapy is a common first-line treatment; however, its effectiveness in patients with chronic pain is often only moderate.8–10 Consequently, identifying prognostic factors to predict chronic pain is a top priority for neck pain research and clinical care.11 By identifying these prognostic factors, especially modifiable factors, physiotherapists can make more informed decisions, potentially target modifiable factors and prevent the development of chronic idiopathic neck pain.
The existing literature on prognostic models shows a low performance in predicting chronic neck pain.12 Moreover, the external validity of current prognostic models in terms of pain and recovery outcomes has not been proven in patients with acute and subacute neck pain.13 This may be attributed to the inclusion of heterogeneous groups of patients for the development of these prognostic models, characterised by varying pain duration (acute, subacute <12 weeks and chronic >3 months), clinical symptoms and prognosis. Furthermore, the varying definitions of the outcome, including persistent and/or recurrent pain groups, contribute to the low performance of these models. Additionally, much of the prognostic research has predominantly focused on non-modifiable factors, such as age, pain duration and sex, neglecting potentially modifiable factors.12 Incorporating modifiable factors has the potential to better tailor interventions to individual patients, which could enhance the model’s applicability and relevance in clinical practice.
It is known that biomedical, psychological and social factors provide a comprehensive understanding of the neurophysiological changes involved in developing chronic pain.14 Consequently, there is a compelling need for a biopsychosocial approach that specifically focuses on modifiable prognostic factors to predict chronic pain after a new episode of non-specific idiopathic, non-traumatic neck pain. This study aimed to (1) identify which modifiable factors are independent prognostic factors of the development of chronic neck pain in patients with acute and subacute neck pain and (2) to develop and internally validate a model to predict chronic pain.
Methods
The methods of this study have been extensively described in the study protocol.15 Briefly summarised, the methods were as follows.
Study design
The present study is a prospective longitudinal cohort study that focuses on modifiable prognostic factors and follows the guidelines of the Prognosis Research Strategy (PROGRESS) framework and the Transparent Reporting of a multivariable prediction model for Individual Prognosis Or Diagnosis (TRIPOD) statement type 1b.16 17 This study adheres to the specific statistical recommendations for type 3 prognostic model research.16 The findings are reported according to the TRIPOD statement to ensure transparent reporting of the multivariable prediction model for individual prognosis (see online supplemental appendix 1).17
Supplemental material
Study setting
Participants were recruited from 30 Dutch physiotherapy primary care practices by 94 physiotherapists between 26 January 2020 and 31 August 2022. The study was completed in March 2023 (including reminders and time for response).
Patient and public involvement statement
None.
Participants
Patients were approached if they presented in one of the participating physiotherapy practices with a new episode of acute or subacute non-specific idiopathic, non-traumatic neck pain. Patients were included if they met the following criteria: age 18 years or older, a new presentation of neck pain no longer than 12 weeks after onset and the patient indicated on the body diagram that he/she experienced regional neck pain. If the patient had a previous episode of neck pain, the patient had to be relatively free from symptoms on the Numeric Pain Rating Scale (NPRS of <3) for at least 3 months prior to the present episode of neck pain. The exclusion criteria were: neck pain surgery in the past, cervical spine radiculopathy assessed with the upper limb neurodynamic test 1,18 widespread primary pain (International Classification of Diseased 11th Revision (ICD-11)) (diffuse musculoskeletal pain in at least four of five body regions (eg, shoulder or upper arm, wrist or hand, pelvis or ankle or food) and in at least three or more body quadrants (as defined by upper–lower/left-right side of the body) and axial skeleton (neck, back, chest and abdomen)),19 pain not caused by musculoskeletal origin (not located in the muscles, bones, joints or tendons)20 and inability to read or understand the Dutch language.
Baseline and follow-up procedure
During the first consultation, the physiotherapist informed eligible patients about the study purpose and expectations. Patients who verbally indicated they wanted to participate in the study signed an informed consent before completing the initial digital questionnaire at baseline (T0). Follow-up questionnaires were sent via email at 6 weeks (T1), 3 months (T2) and 6 months (T3), taking 20–40 minutes to complete. Participants were reminded to complete the questionnaires via email or telephone contact by their treating physiotherapist.
Outcome
The NPRS was used to quantify the presence of chronic pain. If the pain was present, defined as an NPRS ≥3, at all measurement moments (ie, 6 weeks, 3 months and 6 months), it was classified as chronic.15 21
Candidate prognostic factors
We included candidate prognostic factors to predict chronic pain or non-recovery identified in a previous systematic review and by neck pain experts in a Delphi study with >70% consensus in the first round.12 22 Details on candidate prognostic factors and their measurement are provided in our study protocol.12
Patient characteristics: sex and age.
Symptoms: pain intensity at baseline measured with the NPRS, duration of the acute or subacute neck pain in weeks, reported pain in different body regions (yes/no), accompanying headache (since the onset of neck pain and headache before the neck pain) and disability measured with the Pain Disability Index, where the sum score was divided by the entered items.23
Work-related factors: happiness at work, job satisfaction, and the potential to self-modify posture measured with a self-reported question.
General factors: the lifestyle factors: smoking, alcohol, length and weight (body mass index), sleep quality measured with an adjusted sleep quality question from the Neck Disability Index,22 24 and physical activity measured by meeting the activity level according to the Dutch Healthy Exercise Norm (Yes/No).25
Psychological and behavioural factors: illness perceptions were assessed using the Dutch version of the Brief Illness Perception Questionnaire.26 Catastrophising was measured with the short version of the Pain Catastrophizing Scale.27 Depression and distress were assessed with the 21-item version of the Depression Anxiety Stress Scale.28 Kinesiophobia was measured using the 11-item version of the Tampa Scale for Kinesiophobia.29 Coping strategies were evaluated with the Pain Coping Inventory.30 31 Hypervigilance was assessed using the Pain Vigilance and Awareness Questionnaire,32 and self-efficacy in managing pain was measured with the 2-item version of the Pain Self-Efficacy Questionnaire.33
The remaining factors included, first, the ‘therapeutic relationship’, assessed through the self-reported question: ‘How much trust do you have in your healthcare provider/physiotherapist?’ Second, the ‘therapist’s orientation’, which could be either biomedical or biopsychosocial. The authors categorised this orientation based on open-ended and multiple choice questions about neck pain cases.15
Sample size
To ensure a sufficient sample size to reduce the effect of overfitting, the minimum number of events per candidate prognostic factor was calculated as recommended by Riley et al 2019.34 The expected value of the Cox-Snell R-squared of the new model was estimated at 0.23,22 35 36 and the estimated outcome event rate at 45%.12 The study considered 26 candidate prognostic factors, including four non-modifiable and 22 potentially modifiable prognostic factors. The a priori sample size calculation suggested a minimum of 598 participants for the prognostic model.
Statistical analysis methods and missing data
This study followed the PROGRESS framework type 3 research.16 The statistical software IBM SPSS (V.27) and R (V.4.2.2) were used for the statistical analysis.37 38 For the analysis, we extensively used the following R packages: tidyverse, MASS, pROC and Mice.39–42 The complete R script used in this study can be found on GitHub at https://github.com/uashogeschoolutrecht/painr (see online supplemental appendix 2 the table of contents).43
We used multiple imputation with fully conditional specification to impute incomplete records, assuming data to be at least missing at random.44 Predictive mean matching was used to impute continuous variables and logistic regression for categorical variables. After completing the data, the outcome variable (chronic pain) was determined for each participant. The factor ‘healthcare provider orientation’ exhibited significant missing data, which could not be imputed based on the patient-specific information. As a result, we had to proceed with the available data during the subsequent analysis, even though a significant portion was missing.
The predictive performance of each candidate prognostic factor of chronic pain was estimated using univariable logistic regression analysis. These analyses were not used to decide which prognostic factors would be included in the multivariable model.
Before multivariable modelling, we computed the variance inflation factor to assess multicollinearity. If this factor exceeded 10, the selection of candidate prognostic factors for modelling was guided by the clinical expertise of the authors of this study.
All candidate prognostic factors were entered into the multivariable model. To make the model more concise and identify the most significant prognostic factors, we applied backward elimination.
Model performance was quantified as its discriminative ability, using the area under the receiver operating characteristic curve (AUC), model calibration, using calibration plots and computing the Hosmer and Lemeshow goodness-of-fit test and as model fit, using Nagelkerke’s R2.
Bootstrap resampling with 1000 bootstrap samples was used for internal validation to calculate the optimism-corrected AUC and determine the shrinkage factor, thereby adjusting for overfitting by shrinking regression coefficients. After shrinking regression coefficients, we re-estimated the model intercept.
Results
A total of 2567 patients underwent eligibility assessment across 30 physiotherapy practices in the Netherlands. Among these patients, 1600 were excluded, primarily due to the fact that they already had chronic pain (lasting >12 weeks with an NPRS ≥3), cervical spine radiculopathy or widespread pain. Additionally, 307 patients refused to participate, citing disinterest, scheduling conflicts or stress at the time of invitation. Ultimately, 660 potential participants provided informed consent; however, 58 of them did not respond during the baseline measurement phase, resulting in the inclusion of 603 individuals in a period of 2.5 years (figure 1). Among them, 62 participants (10%) developed chronic pain, while 541 participants experienced recovery from their pain.
Flow-chart study. NPRS, Numeric Pain Rating Scale; N, number; T, time-point.
For the description of the participants’ characteristics, including candidate prognostic factors and the number of participants with missing data, see table 1. We included 397 women and 206 men. The mean pain intensity at baseline was 5.9 (SD 1.9), and the mean disability was relatively low, with a score of 2.7 (SD 2.1) on a 0–7 scale. Of our 603 participants, 92 (15.3 %) did not work. We included these participants as not working in all the work-related factors in our multivariable analyses.
Baseline characteristics of the study population
There was some loss to follow-up at various follow-up moments. However, only 78 participants did not complete any follow-up measurement. At the 6-week measurement, 154 participants failed to submit the required forms. This number changed to 224 at the 3-month follow-up and to 211 at the 6-month mark. The Little’s Missing Completly at Random (MCAR) test yielded a p-value >0.05, supporting the appropriateness of multiple imputations.44
The interventions most frequently applied were (1) joint mobilisation, manipulation, traction and nerve mobilisation techniques, with an application rate of 85.4%, and (2) information and advice, with an application rate of 86.7%. Exercise and massage were applied to 58.1% and 54.7% of the study population. For a detailed overview of the interventions applied across the study population, see online supplemental appendix 3.
Univariable prognostic factors of development of chronic pain
The univariable analyses (see figure 2) revealed significant positive associations between the following candidate prognostic factors and chronic pain: being female, higher pain intensity at baseline, longer duration of neck pain, experiencing pain in different body regions, the onset of headache since the neck pain began, higher disability scores, unemployment, higher scores on catastrophising, illness beliefs about recovery (concerned and duration), depression, distress and lower treatment beliefs. Some of these factors were identified with broad CIs. For most factors not showing significant associations, the ORs were close to 1, indicating a lack of a clinically meaningful association.
Univariable logistic regression analysis: unadjusted association between each candidate prognostic factor and the outcome of chronic pain. The first figure displays the continuous variables, while the second illustrates the categorical and dichotomous variables, and OR and corresponding CIs are presented. BMI denotes body mass index, W represents weight (kg) and H stands for height (m). P-values are indicated as follows: *for 0.01<p≤0.05, **for 0.001<p≤0.01 and ***for p≤0.001.
Multivariable modelling
The inclusion of ‘work status’ as a category among the work-related prognostic factors resulted in multicollinearity within the following factors: happiness and satisfaction at work and the ability to change posture during work. To mitigate this issue, we decided to include only the factor ‘ability to change posture at work’ in our final model. This decision was based on the distinct conceptual domain of this factor, which differs from the psychological construct already well represented by the other included factors. The candidate prognostic factor ‘work status’ is thus also referred to as the ability to change posture at work in the analysis. Following this adjustment, multicollinearity was no longer observed.
Several prognostic factors were identified from the multivariable logistic regression analysis. These included sex (female), higher pain intensity at baseline, reported pain in different body regions, headache since the onset of neck pain, headache(s) before the neck pain, an inability or neutral score on self-modify posture during work, not working, lower scores pain identity and treatment beliefs, higher scores in beliefs regarding recovery (duration and concerns) and higher scores on distress and self-efficacy. The ORs including 95% CIs are presented and visualised in figure 3. Of all prognostic factors, not working showed the strongest association (OR 4.87). The combined prognostic model showed an AUC of 0.86 (95% CI 0.82 to 0.90) and a Nagelkerke’s R2 of 0.31 (figure 4). The Hosmer–Lemeshow test yielded a p value of 0.7167, indicating a good model fit. The calibration plot (figure 4) revealed acceptable to good calibration over the range of predicted probabilities. The Brier score was 0.077, indicating solid performance.
Adjusted multivariable logistic regression model.
Area under the Receiver Operating Characteristic (ROC) and calibration curve. The tick marks at the bottom of the Calibration curve represent the distribution of predicted probabilities. Each tick mark indicated a predicted probability for an individual observation. A dense cluster of tick marks indicated more observations with that specific predicted probability. This distribution occurs within the dataset.
Internal validation prognostic model
The bootstrap validation yielded a shrinkage factor of 0.83, which was then used to multiply the regression coefficients. The resulting model, including the re-estimated intercept, is in figure 3. The AUC after correction for optimism was 0.83. The optimism-corrected Nagelkerke’s R2 was 0.24.
The intermezzo section highlights a detailed patient profile to clarify the applicability and interpretation of our findings in a practical context. The supplemental figure presents an interactive visualisation depicting the varied pain trajectories among participants within our cohort, alongside the linear predictor and the probabilities of chronic pain derived from our multivariable prognostic model. This visualisation illustrates the complexity and variability of pain progression over time. For a comprehensive visualisation of all participants, see the web application: https://rstudio-connect.hu.nl/painr-app/.
Intermezzo
The patient (participant 110), a male, describes his neck pain intensity as 6 on the NPRS and also reports low back pain. Since the onset of neck pain, he has also developed headaches, which were not present before the neck pain. Despite being employed, he finds it impossible to modify his posture during work. He anticipates the duration of his symptoms to be quite long, assessing it at 9 out of 10. Despite this, his concern for his condition is relatively minimal, with a score of 2 out of 10. His confidence in the therapy is high, rated at 8 on a 0–10 scale. Stress is absent in his case, evidenced by a score of 0 out of 21. While he admits to only a moderate understanding of his pain, scoring a 6 out of 10, he shows a high level of self-efficacy, achieving a full score of 12 on a 0–12 scale.
The patient (participant 914), a female, reports experiencing a pain intensity level of 6 on the NPRS. She notes pain in other regions of her body as well. Since developing neck pain, she has also begun to experience headaches, which she did not have prior to the neck pain. Currently, she is not employed. She anticipates her symptoms will persist, rating the anticipated duration as 10 on a scale from 0 to 10, indicating a long-term expectation of symptoms. She expresses moderate concern about her neck pain, with a concern level of 5 on a 0–10 scale. Her confidence in the effectiveness of her therapy is also moderate, rated a 5 on a 0–10 scale. She reports experiencing a moderate level of stress, scoring 12 on a 0–21 scale. Her self-reported understanding of her pain is 6 on a 0–10 scale, and scores a moderate self-efficacy, with a score of 6 on a 0–12 scale.
Linear predictor (LP)
The linear predictor (LP) is given by:
LP=−5.782
+(0.468×sex(female=1))
+(0.227×pain intensity)
+(0.734×pain in different body regions)
+(0.726×headache(s) since the neck pain)
−(0.070×headache(s) before the neck pain)
+(0.384×potential to self-modify posture at work)
+(1.311×work status)
+(0.184×duration beliefs)
+(0.108×concerns)
−(0.204×treatment beliefs)
+(0.083×distress)
−(0.142×identity beliefs)
+(0.109×self-efficacy)
Probability of chronicity
Probability of chronicity
Participant 110
LP calculation for patient X yields LP=−1.88, resulting in:
Participant 914
LP calculation for patient X yields LP=0.98, resulting in:
Discussion
In this prospective cohort study, we (1) identified which (modifiable factors) are independent prognostic factors of the development of chronic neck pain, and we (2) developed and internally validated a prognostic model for predicting chronic pain after a new episode of acute or subacute non-specific idiopathic, non-traumatic neck pain. We found several significant associations between non-modifiable and modifiable factors and chronic pain: being female, higher pain intensity at baseline, longer duration of neck pain, experiencing pain in different body regions, the onset of headache since the neck pain began, higher disability scores, unemployment, higher scores on catastrophising, illness beliefs about recovery (concerned and duration), depression, distress and lower treatment beliefs.
The internally validated prognostic model demonstrates good prognostic performance, underscored by an optimism-corrected AUC of 0.83. The calibration indicates a solid performance, as indicated by the calibration curve, alongside a commendable Brier score. The Hosmer–Lemeshow test, with a p value of 0.717, affirms a good model fit. Nonetheless, the model’s corrected R2 of 0.24 suggests that the model provides a meaningful but limited explanation of the probability distribution of the outcome of chronic pain. The model comprises 12 variables, four non-modifiable and eight potentially modifiable by physiotherapists. The non-modifiable factors include sex, reported pain in different body regions, longer existing headaches and employment status (not working). Potentially modifiable factors encompass baseline pain intensity, self-efficacy, headache onset concurrent with neck pain, the ability to self-modify posture at work, illness beliefs regarding recovery (including concerns and expected duration) and beliefs about neck pain identity and treatment.
When comparing our individual prognostic factors and those included in our prognostic model with existing prognostic studies in musculoskeletal pain, several common factors emerge, including age, work status, reported pain in different body regions (including headache), baseline pain identity and self-efficacy.45–49 In our study, not working showed a high OR in both univariable and multivariable analyses. A physiotherapist cannot directly modify this factor; however, attention could be given to potentially modifiable factors associated with unemployment, such as physical disability and mental health.50 51 In addition, in our study, a higher score on the Pain Self-Efficacy Questionnaire 2-item version was associated with higher odds of chronic neck pain. Notably, this association was characterised by a low regression coefficient and OR and was insignificant with a small CI. Moreover, this outcome may be biased using this short questionnaire, where the largest group of our population scored above 10 on a 0–12 point scale for self-efficacy, exhibiting a known ceiling effect.52 This notable outcome might, therefore, be questioned.
Our model incorporated four illness perception factors: beliefs about recovery (including concerns and duration), identity and treatment beliefs. Longitudinal studies on low back pain have yielded similar findings, illustrating individual associations between illness beliefs (eg, duration and treatment beliefs) and negative clinical outcomes over various time periods.53–55 However, in prognostic multivariable models, the contribution of illness perceptions to the robustness of a prognostic model varies.55 56 Notably, illness beliefs are often excluded from the candidate prognostic factors in the models developed and externally validated for neck pain models.12 57–59 Recent research has shown that modifying illness beliefs related to identity and concerns can mediate outcomes, specifically disability and pain, within physiotherapy primary care practices.60 Consequently, further research into the modification of illness perception factors and their influence on the development of chronic pain, is imperative. Such studies are crucial to ascertain if physiotherapy interventions can effectively alter patients’ outcomes.
Furthermore, it is important to note that several psychological factors, such as depression, kinesiophobia, catastrophising and poor coping skills, are commonly recognised as associated with and prognostic for chronic pain.14 61 These factors were not retained in our final prognostic model. Although these factors showed an association in our univariable analysis, they did not improve the predictive accuracy of our model. Notably, our baseline measurements indicated a distinctly non-normal distribution for these psychological factors, contrasting with studies in chronic pain patients where these factors are more prevalent.62 Despite their exclusion from our final model, screening for these factors during the initial pain phase and ongoing monitoring during recovery remain important. This is particularly noteworthy considering the body of evidence indicating that treatments targeting psychological factors, such as catastrophising, depression and distress, have shown favourable outcomes when addressed by healthcare providers. However, it is essential to highlight that these studies have primarily focused on patients with chronic musculoskeletal pain.49 62–64 In contrast, it is important to note that most studies involving patients with acute and subacute musculoskeletal pain have mainly focused on pain and disability as outcomes. However, these studies, which investigate the effectiveness of treating physiological factors, should also examine whether identified changes in these psychological factors contribute to the reduction in pain intensity or disability observed in their study population.49 65 66
The incidence of chronic pain in our participants differed from our systematic review findings. Our preliminary sample size calculation assumed a 45% chronicity rate for neck pain, which divided the number of patients by the non-recovery cases.12 This disparity can be attributed to our definition of chronic pain and the definition of the measurement approach. Unlike most studies that use single time point assessment (eg, 3 months, 6 months or 12 months) with specific pain score threshold,67 including those in our review,12 our study used a more comprehensive approach. This approach provides a precise representation of chronic pain as a continuous experience. Using this methodology, we excluded the recurrent pain group, which includes pain-free or mild time periods, diverging from the ICD-11 broader definition of chronic pain.19 We hypothesise that distinguishing between continuous and recurrent pain will lead to a more effective prognostic model, acknowledging the distinct pain experiences of these groups.
Limitations
The calibration curve suggests a substantial overestimation of higher risks; this estimation was based on only a few patients, as most had a relatively low estimated risk of chronic pain.
In the initial sample size calculation, we assumed a 45% incidence of chronic pain based on our systematic review.12 This calculation allowed for 26 candidate prognostic variables among a cohort of 598 participants.34 However, this study yielded a lower-than-expected incidence of chronic pain, with only 10% of participants, indicating an underpowered and potentially inadequate sample size. However, the increased risk of overfitting and the potential for overly optimistic model performance seems to be minimal, as suggested by our internal validation analysis, which revealed a shrinkage factor close to 1.
Chronic primary pain, as described by the ICD-11, is accompanied by significant emotional distress or functional disability. We used a threshold of ≥3 to define chronic pain based on the observation that mild pain typically does not entail marked emotional distress or functional disability.68 69 However, the literature indicates that establishing a definitive cut-off point for mild and moderate pain, especially regarding pain-related interference with functioning and emotions, is complex.69–71 Therefore, choosing a threshold of 3 is debatable, and selecting a different threshold could yield different study results.
Furthermore, in our study’s protocol discussion, we noted that our study did not influence the therapies participants received; however, these therapies could potentially affect both the outcomes and the accuracy and generalisability of the developed model. Participants were treated according to the Dutch Physiotherapy Guideline for neck pain, which might modify our candidate prognostic factors and potentially reduce chronicity risks. Given the diversity of factors, the variety of modalities used by physiotherapists and the therapists’ varied backgrounds, we considered the impact of these therapies on our study results minimal. Ideally, these therapies would either not be applied or should have been analysed within the multivariable prognostic model to assess their impact; however, this was not feasible due to sample size constraints.
Our final prognostic model retained the factor ‘self-modifying posture during work’. This factor was measured subjectively using a non-validated question, which poses a limitation as it may not distinguish between perceived and actual behaviour during work. The limitation of this subjective measurement lies in its inability to clearly distinguish whether individuals perceive that they can change positions during work or are changing their positions. Additionally, this type of questioning prevents us from confirming the accuracy of reports, such as whether a patient who claims they cannot change positions is indeed unable to do so. Establishing the validity and discriminative ability of the different concepts being tested is important to investigate.
Clinical application and further research
The development of this prognostic model has identified several potential modifiable factors. In clinical practice, a physiotherapist can use this model to gain insight into a patient’s probability of experiencing chronic neck pain. Furthermore, assessing and intervening on the modifiable factors in our model can be beneficial. However, we must be aware that although they have been validated for their prognostic value in our 1b prognostic study, it does not mean that modifying these factors will necessarily reduce the risk of developing chronicity. It is highly recommended to evaluate the performance of our model in an external validation study. If the model is found adequate, a prognostic model impact study is required, to quantify the effect on physiotherapist decision-making in patients with acute or subacute non-specific idiopathic, non-traumatic neck pain (TRIPOD statement).17
Conclusion
This model has the potential to obtain a valid prognosis for developing chronic pain after a new episode of acute or subacute non-specific idiopathic, non-traumatic neck pain. It includes mostly potential modifiable factors for physiotherapy practice. External validation of this model is recommended.
Data availability statement
Data are available in a public, open access repository. Technical appendix, statistical code and dataset available from the Github repository: https://github.com/uashogeschoolutrecht/painr DOI: available upon acceptance.
Ethics statements
Patient consent for publication
Ethics approval
This study involves human participants, but the Medical Research Ethics Committee Utrecht declared that the Medical Research Involving Human Subjects Act (WMO) does not apply to this study (protocol number 19-766/C), exempted this study. Participants gave informed consent to participate in the study before taking part. Participants who gave informed consent were assigned a unique code to allow anonymous data collection, facilitated through the secure Formdesk data transfer system.72
Acknowledgments
The authors would like to express their gratitude to all the physiotherapists who facilitated the inclusion of patients and specifically to all the patients who participated in this study.
References
Supplementary materials
Supplementary Data
This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.
Footnotes
X @smeets1964
Contributors All authors materially participated in this research. Their main contribution to the manuscript is described below. MJV is the guarantor, substantial contribution to study conception, study design, data analysis, data interpretation, drafting and revising the manuscript, and significant involvement in conceptualising the web application and GitHub repository; HW: substantial contribution to study conception, study design, data analysis, data interpretation, drafting and revising the manuscript; FM: contribution to study conception, study design, data interpretation and revising the manuscript; SMJvK: substantial contribution to the study design, data analysis and data interpretation, drafting and revising the manuscript; MT: substantial contribution to the data analysis and data interpretation, revising the manuscript and key architect of the web application and GitHub repository; Professor RJEMS: contribution to study conception, data analysis, data interpretation,drafting and revising the manuscript.
Funding This work was supported by the Institute of Movement Studies and partly by the Utrecht University of Applied Sciences research voucher. The funding concerns an internal promotion voucher of the University of Applies Sciences. The funders had no role in the study design, data collection, analysis, decision to publish or manuscript preparation.
Competing interests None declared.
Patient and public involvement Patients and/or the public were not involved in the design, conduct, reporting or dissemination plans of this research.
Provenance and peer review Not commissioned; externally peer reviewed.
Supplemental material This content has been supplied by the author(s). It has not been vetted by BMJ Publishing Group Limited (BMJ) and may not have been peer-reviewed. Any opinions or recommendations discussed are solely those of the author(s) and are not endorsed by BMJ. BMJ disclaims all liability and responsibility arising from any reliance placed on the content. Where the content includes any translated material, BMJ does not warrant the accuracy and reliability of the translations (including but not limited to local regulations, clinical guidelines, terminology, drug names and drug dosages), and is not responsible for any error and/or omissions arising from translation and adaptation or otherwise.