Tracking low back pain outcomes in private clinics

Back pain is probably the most common thing a private clinic sees. Physios get a real feel for how each patient is progressing, and that clinical judgement counts for a lot. The harder part, in a busy clinic, is capturing that progress in a way you can look back on later. Scores often get recorded at the first session and become tricky to keep up after that.

I trained as a physio before moving into health data science, so I’ve seen this from the treatment room and from the data side. For this piece I’ve stuck to six papers that are free to read in full. Every number below comes from one of them, and the links are at the end if you want to check.

Let’s break it down into three questions: what to measure, when to measure it, and how to tell real change from noise.

What to measure

For a private clinic, the most useful example I found is in the Netherlands. The Dutch physiotherapy guideline for low back pain recommends three questionnaires, and the national registry has tested how they hold up when everyday primary care practices collect them (Meerhoff et al., 2021).

Across 865 practices, these questionnaires gave reliable and valid data in everyday care. Patients whose back pain had started within the last 12 weeks improved more than those with chronic pain, which is what you’d expect if the scores are picking up something real. On average, pain dropped by about 4 points out of 10 between the first and last measurement.

Research has its own standard. A core outcome set built for clinical trials recommends the Oswestry Disability Index 2.1a or the 24-item Roland-Morris for function, a numeric pain scale for average pain over the last week, and the SF-12 or PROMIS-10 for quality of life (Chiarotto et al., 2018). The team appraised 25 questionnaires in systematic reviews, then ran a two-round Delphi survey with researchers, clinicians and patients.

It wasn’t designed for clinics, though the authors suggest it may also suit routine practice. Two details from it are useful anyway. The ODI and SF-12 carry fees, which is why the free Roland-Morris and PROMIS-10 sit alongside them, and the Quebec scale was left out partly for the same fee reason. For patients who’ve had pain for less than a week, the authors also suggest adding a 24-hour version of the pain scale.

The two sets agree on the pain scale. For function they go different ways, and either is defensible. The Dutch set adds the patient’s own problem activities, while the trial set adds quality of life.

If you go with the Roland-Morris, one small detail matters for your data. The original asks patients to tick only the statements that apply, so a blank could mean no or could mean skipped. In interviews with 23 people with chronic back pain, a version with an explicit yes or no for every item was well received, and all of them found the questionnaire easy to understand and complete (Burbridge et al., 2020). The things they said back pain affected, from walking and stairs to sleep and mood, were covered by its items. Washing, driving and work weren’t asked about directly.

Whichever set you pick, the bigger win is using it at the same points, every time.

When to measure

The core set deliberately leaves timing open; the authors say time frames should fit each study’s goals (Chiarotto et al., 2018). So the best clues come from studies that followed patients over time.

A Danish study followed 2,777 adults who came to 10 private chiropractic clinics with a new episode of back pain (Hansen et al., 2025). They filled in the 23-item Danish version of the Roland-Morris, scored 0 to 100, at their first visit and again at 2, 13 and 52 weeks. Follow-ups went out by email, with reminders and a phone call for anyone who didn’t reply.

The patients split into nine different disability courses. Three improved, quickly or gradually, and stayed that way. Three started with moderate or high disability and barely moved. The last three, about one in ten patients, went up and down.

More pain sites, more comorbidities, prescription painkillers, anxiety, low mood, fear of movement and poorer general health all made a rockier course more likely. The more of these a patient carried, the stronger the pull. Even so, much of the variation stayed unexplained, and baseline information alone couldn’t tell you which path one person would take.

One detail stood out to me. Earlier studies like this found four to six courses, and the authors think part of the difference is that they checked in at two weeks, while earlier studies waited three or six months. An early check-in shows you patterns a single follow-up months later would miss.

Their advice to clinicians is worth repeating: reassure patients that most episodes go well, and be honest that individual courses vary a lot.

Here’s a schedule a busy clinic could realistically keep. It borrows the Danish time points; the rest is my suggestion, not something the study tested.

1. First session: function, pain and quality of life.

2. Around two weeks: function and pain, early enough to change the plan if things stall.

3. Discharge: the full set again, plus one question asking how the patient rates their overall change.

4. Around three months: function and pain by email or text, to see whether the gains held.

What counts as real change

A score going down doesn’t always mean a patient got better. Two things get in the way.

The first is noise. The reviews behind the core set found high quality evidence that the numeric pain scale’s measurement error was insufficient (Chiarotto et al., 2018). That’s a formal rating under the COSMIN standards, and in plain terms it means the noise in a single score can be bigger than the smallest change that matters. So a small drop in one patient’s pain score may not be progress at all.

The second is that what counts as meaningful depends on who you’re measuring. In 92 people with lumbar disc herniation treated with physio, researchers asked patients to rate their own change, then found the score change that best separated those who felt improved from those who didn’t, which they called the minimal clinically important change (Ghaderi Niri et al., 2024). The Dutch registry used minimal important change values taken from earlier research (Meerhoff et al., 2021).

Same questionnaire, two different thresholds: 14.5 points in one study and 20 in the other. Treat these figures as guides, not hard cut-offs.

In the disc herniation group, the ODI and Quebec scale picked up improvement well and the Roland-Morris less so. If your caseload leans towards disc problems, that’s a reason to favour the ODI.

What this really means for a clinic is simple. Don’t judge one patient on a single before and after difference. Look at the trend across several time points, check it against the patient’s own rating of change, and save firm conclusions for patterns across your whole caseload.

Where it falls apart in practice

The questionnaires hold up. Collecting them consistently is the hard part.

The best real-world test I found is the Dutch national physiotherapy registry (Meerhoff et al., 2021). Across 865 practices and 6,560 physios, the back pain questionnaire data were reliable, valid, and good enough to tell practices apart once case mix was accounted for. But across the whole registry only 59% of patients agreed to share their data, and among the back pain patients only about 42% had a before and after score.

There’s a subtler problem in the detail. Patients without a repeat score had fewer sessions on average, 6.0 against 8.1. The authors suggest that when someone recovers quickly, the follow-up questionnaire feels pointless to the physio or the patient and gets skipped. If that’s right, the quickest recoveries are the ones most likely to drop out of the data, which could make a clinic’s results look worse than they really are.

These were also volunteer practices. The authors note they were probably early adopters who already favoured questionnaires, so most clinics are likely starting from a lower base.

I see the same pattern when I talk to clinics. They aren’t starting from zero. Some data gets recorded, just not consistently, and pain scores in particular are often left out.

Both papers suggest making follow-up easier to complete. The Dutch authors suggest letting patients fill in questionnaires on their phones, and the Danish study kept a year of follow-up going with emailed questionnaires, reminders and phone calls (Hansen et al., 2025).

Where it can go wrong is treating an app as the treatment itself. A review of five randomised trials with 894 patients compared self-management apps, used without any therapist contact, against usual care or no treatment (de Melo Santana et al., 2023). Some trials reported benefits, but pooled together there was no significant difference in pain or disability, and the authors rated the evidence as low certainty. So use apps to collect data and keep patients engaged between sessions. Don’t sell them as the care.

What this means for your clinic

If you run a clinic, here’s the short version.

• Starting from scratch, the Dutch set is a sound default: the Quebec scale and the 0 to 10 pain scale, plus the Patient Specific Functioning Scale for the activities patients struggle with most. Check the licence first, as the Quebec scale carries a fee. If you already use the ODI or Roland-Morris, keep it, because switching breaks your before and after comparisons.

• If you use the Roland-Morris, ask for a yes or no on every item so a blank never hides missing data.

• Measure at the first session, around two weeks, at discharge and around three months.

• Judge change against a sensible threshold and the patient’s own view, not a single drop in score.

• Automate follow-ups, especially for patients who recover fast and stop coming.

• Expect around one in ten patients to follow an up and down course. That’s how back pain behaves, not proof your care failed.

None of this needs new software or a research team. It needs the same few questions asked at the same points, every time.

Once that’s in place, your own data starts answering questions no paper can. Which of your back pain patients do best, how early you can tell, and where the plan needs to change. That’s the work I do with clinics: getting the right data recorded first, then reading what it says.

Low back pain is the first condition I’m working through like this. Others will follow, the same way.

References

Burbridge, C. et al. (2020) Measuring the impact of chronic low back pain on everyday functioning: content validity of the Roland Morris disability questionnaire. Journal of Patient-Reported Outcomes, 4(1), 70. Available at: https://doi.org/10.1186/s41687-020-00234-5

Chiarotto, A. et al. (2018) Core outcome measurement instruments for clinical trials in nonspecific low back pain. Pain, 159(3), pp. 481-495. Available at: https://doi.org/10.1097/j.pain.0000000000001117

de Melo Santana, B. et al. (2023) Efficacy of mHealth interventions for improving the pain and disability of individuals with chronic low back pain: systematic review and meta-analysis. JMIR mHealth and uHealth, 11, e48204. Available at: https://doi.org/10.2196/48204

Ghaderi Niri, H. et al. (2024) Oswestry Disability Index, Roland-Morris Disability Questionnaire, and Quebec Back Pain Disability Scale: responsiveness and minimal clinically important changes in Iranian people with lumbar disc herniation following physiotherapy. The Archives of Bone and Joint Surgery, 12(1), pp. 58-65. Available at: https://doi.org/10.22038/ABJS.2023.72246.3366

Hansen, L.A. et al. (2025) Variation of disability courses in Danish chiropractic patients with low back pain: a one-year cohort study with four follow-ups. BMC Musculoskeletal Disorders, 27(1), 19. Available at: https://doi.org/10.1186/s12891-025-09365-6

Meerhoff, G.A. et al. (2021) Reliability, validity and discriminability of patient reported outcomes for non-specific low back pain in a nationwide physical therapy registry: a retrospective observational cohort study. PLoS One, 16(6), e0251892. Available at: https://doi.org/10.1371/journal.pone.0251892