The goal of this clinical trial is to learn if personalized physiotherapy can lower neck pain in adults with recent (sub)acute neck pain who have a higher chance of the pain becoming long-lasting. The main questions it aims to answer are:

Does personalized physiotherapy lower pain compared to standard physiotherapy? Does personalized physiotherapy improve daily activities, quality of life, and pain interference compared to standard physiotherapy? Do physiotherapists deliver the personalized treatment as intended?

Researchers will compare each participant's own results during a standard care period to their results during a personalized care period, to see if the personalized approach works better than standard care.

Participants will:

Receive standard physiotherapy care for 3 weeks Then receive personalized physiotherapy, tailored to their individual needs, for several weeks (until their treatment goals are achieved) Answer short questions about their pain and daily functioning three times a week Take part in an interview about their experiences every 3 weeks

This study employs a replicated mixed-methods single-case experimental design (MMSCR) with a convergent, parallel approach, in which quantitative and qualitative data are collected concurrently and subsequently integrated. The protocol was developed in accordance with the Single-Case Reporting Guideline in Behavioural Interventions (SCRIBE) 2016 and the Mixed Methods Reporting in Rehabilitation & Health Sciences (MMR-RHS) guidelines.

Rationale Non-specific neck pain (NSNP) is a multifactorial condition in which physical, behavioral, cognitive, and contextual processes vary between patients, change over time, and continuously interact. A standardized treatment approach is therefore inherently limited. Personalized physiotherapy grounded in a Process-Based Therapy (PBT) framework aims to address this by continuously identifying which modifiable biopsychosocial processes are most relevant for an individual patient at a given moment, and selecting interventions that most efficiently target those processes. To make this approach feasible in daily primary care practice, a treatment protocol was developed together with a data-driven Clinical Compass, consisting of clinical indicators of modifiable biopsychosocial processes. This study is designed to obtain a first estimate of the preliminary effectiveness of this approach and to examine the fidelity with which it is delivered, prior to larger-scale evaluation.

Study design The quantitative component follows a nonconcurrent AB single-case experimental design (SCED) with repeated measurement, replicated systematically across 11 participants (11 SCEDs). Each participant undergoes a predetermined phase sequence: a baseline phase (A) followed by an intervention phase (B). The baseline phase spans three weeks with a minimum of nine measurement points; the intervention phase lasts a minimum of two weeks with a minimum of six measurement points, with actual duration depending on individual treatment progress. The number of measurements per phase was determined based on the discriminative power required to detect changes in outcome trajectories. Replication across 11 cases is achieved by delivering the intervention through different physiotherapists working in different geographic regions, all applying identical eligibility criteria.

Researcher blinding is not applied, as outcome measures are self-reported and the researcher is not involved in intervention delivery.

Study setting and therapist training The study will be conducted in nine primary care physiotherapy practices participating in the project consortium, purposefully selected to ensure variation in demographic characteristics, specialization, and practice size. Prior to study commencement, all participating physiotherapists complete a structured training programme targeting knowledge transfer, skill development, professional identity, and behavior change in practice. The programme consists of four sequential components: preparatory video modules with personal reflection, an online interactive case-based session, a live training session including role-play with a trained actor and guided reflection, and structured peer supervision sessions during the active treatment phase of the study to support implementation.

Baseline phase (Phase A) During the three-week baseline phase, participants receive standard care consisting of an intake and two treatment sessions in accordance with the Royal Dutch Society for Physical Therapy (KNGF) Clinical Practice Guideline for Neck Pain. During this phase, the physiotherapist delivers usual care without access to Clinical Compass results and without structured insight into the participant's modifiable biopsychosocial processes. Based on prior research in this population, an average reduction in pain intensity of approximately 35-45% is expected during the first six weeks due to standard care and spontaneous recovery. This anticipated recovery trajectory is explicitly modelled in the analysis, enabling differentiation between changes consistent with natural recovery or standard care and changes potentially attributable to the personalized intervention introduced in Phase B. All clinical activities during this phase are documented in the Electronic Health Record (EHR) to allow retrospective analysis.

Intervention phase (Phase B) During the intervention phase, treatment is continuously adapted to the modifiable biopsychosocial process(es) identified as most relevant, jointly mapped and evaluated with the patient throughout the treatment trajectory. Intervention sequencing is process-driven rather than time-driven. The approach follows five sequential steps: (1) joint identification of modifiable biopsychosocial processes using the Clinical Compass, a structured biopsychosocial anamnesis, and physical examination where relevant; (2) construction of a person-specific network model in which goals, problems, limitations, and influencing factors are represented as interconnected nodes, including feedback loops that may maintain the patient's complaints; (3) shared decision-making to prioritize the most central modifiable process, guided by six criteria (accessibility, centrality, therapist competence, risk, likelihood of change, and strategic positioning); (4) delivery of personalized interventions targeting the selected process, guided by a consensus-based framework linking processes to evidence-based intervention components within the physiotherapy domain, developed through a modified Delphi study among primary care physiotherapy experts; and (5) ongoing monitoring and adjustment, in which each session begins with shared reflection on changes in the targeted process, experiences in daily life, and barriers or facilitators, informing the decision to continue, adapt, or shift the intervention focus.

Treatment duration and frequency are tailored to the individual patient's goals and presenting problems. The intervention concludes once agreed treatment goals are achieved, or if no improvement occurs after six weeks; this decision is explicitly discussed with the patient beforehand. Participants may continue to receive any other usual care they would otherwise receive during the study.

Data collection In parallel with quantitative monitoring, two additional qualitative data sources are collected: semi-structured interviews conducted with participants every three weeks to explore their experienced change and treatment attribution, and session data comprising audio recordings of treatment sessions together with session-level data extracted from the EHR.

Fidelity assessment Intervention fidelity is assessed using three complementary approaches, all based on audio recordings and/or EHR data: (1) a global rating of intervention delivery during the B-phase, based on a purpose-developed rating scale assessing the presence of core PBT elements, the absence of inappropriate elements, and quality of execution; (2) a parallel global rating applied to baseline-phase recordings to confirm the expected absence of PBT-consistent delivery during standard care, supporting intervention differentiation; and (3) a qualitative, deductive analysis of EHR session data examining the sequence, content, and rationale of delivered intervention components, and how these align with patient characteristics and adjustments made over the treatment trajectory. A minimum of 20% of sessions per participant will be reviewed. Two trained raters, blinded to treatment phase, will independently code recordings after a calibration period; inter-rater reliability will be calculated using an intraclass correlation coefficient. Fidelity results are not fed back to physiotherapists during the study.

Data analysis

Quantitative analysis: Individual and cross-participant effects are evaluated using visual analysis (following the Visual Aid Implying an Objective Rule, VAIOR) complemented by multivariate piecewise linear regression, consistent with established SCED analytic approaches. Because the intervention is expected to produce a delayed or progressive rather than immediate effect, visual analysis focuses primarily on the final measurements of the intervention phase. Improvement direction is predefined for each outcome domain prior to analysis. Consistency of effects across the 11 replications is evaluated against a predefined replication criterion (a positive effect in at least 8 of 11 cases). Two trained researchers conduct visual analyses independently, with inter-rater reliability reported.

Participants who complete the baseline phase and at least two weeks of the intervention phase are included in the analysis; missing data are handled using the Expectation-Maximization algorithm.

Sample size was informed by an a priori power analysis for SCED designs, based on a conservative effect size estimate derived from comparable prior research in the same population, and will be re-evaluated after the first five participants.

Qualitative analysis: Interview transcripts are analyzed using descriptive open coding within three predefined analytical domains: experienced change, the relationship between experience and intervention, and the relationship between experience and quantitative measurements. Analysis is conducted independently by two researchers with consensus meetings to resolve differences, supported by peer debriefing from a third researcher not involved in data collection. A narrative pattern description is developed for each participant, and an audit trail and reflexive memos are maintained throughout.

Integration: For each participant, quantitative outcome trajectories (level, trend, and variability) are systematically compared with the qualitative pattern description through triangulation, assessing congruence, divergence, and complementarity to arrive at individual-level meta-inferences.

A cross-case overview subsequently compares all 11 cases on quantitative pattern, experienced trajectory, congruence/divergence, treatment attribution, fidelity scores, and intervention content and sequence, aiming to identify recurring patterns, explanatory variation between cases, and potential subgroups.

Inclusion Criteria:

• Age 18 years or older
• New episode of neck pain with duration ≤12 weeks since onset
• Neck pain located in the region shown in Figure 2
• ≥19% risk of developing chronic complaints, as determined using a prognostic model
• If a previous episode of neck pain occurred, the current episode must have been preceded by ≥3 consecutive months without persistent pain (defined as absence of pain associated with activity limitation or healthcare consultation)

Exclusion Criteria:

• Prior neck surgery
• Cervical radiculopathy as identified using the Upper Limb Neurodynamic Test (ULNT1)
• Widespread primary pain according to ICD-11 criteria (diffuse musculoskeletal pain in ≥4 of 5 body regions, across ≥3 body quadrants, involving the axial skeleton)
• Pain of non-musculoskeletal origin
• Insufficient proficiency in Dutch to read or understand study instructions

After study completion, de-identified quantitative participant data and associated statistical code will be made available via DataverseNL. Data will be anonymized and reviewed by a data steward prior to sharing, in accordance with FAIR principles and applicable data protection regulations. Access restrictions may apply where necessary to protect participant privacy. Qualitative data (interview transcripts and audio recordings) will not be shared publicly due to the risk of participant re-identification.

De-identified quantitative data, statistical code, and the study protocol will become available following publication of the primary study results, expected in 2028. Data will remain available for at least 5 years after publication via DataverseNL, in accordance with FAIR data principles.

Data will be shared via DataverseNL following review by a data steward to ensure compliance with GDPR and participant privacy protections. De-identified quantitative data and analytic code will be accessible to researchers upon reasonable request for legitimate secondary research purposes. Qualitative data (interview transcripts, audio recordings) will not be shared publicly due to re-identification risk. Access may be restricted where necessary to protect participant confidentiality.

Only the raters evaluating the audio recordings for intervention fidelity are blinded to the phase