WEBVTT 00:00:00.000 --> 00:00:02.819 Imagine you've had major surgery, maybe a hip 00:00:02.819 --> 00:00:05.320 replacement. You get asked later, perhaps a year 00:00:05.320 --> 00:00:07.080 or two on, how you're doing via a questionnaire. 00:00:07.540 --> 00:00:09.640 Now this feedback, yours and thousands of others, 00:00:09.880 --> 00:00:12.339 helps gauge how successful these operations really 00:00:12.339 --> 00:00:15.300 are. Seems simple enough, right? But... What 00:00:15.300 --> 00:00:17.440 if the people who don't send back those forms, 00:00:17.519 --> 00:00:19.239 what if they aren't just, you know, randomly 00:00:19.239 --> 00:00:21.739 busy? What if they're actually a specific group 00:00:21.739 --> 00:00:24.420 and their absence completely changes the picture 00:00:24.420 --> 00:00:26.920 of success that the data shows us? That's the 00:00:26.920 --> 00:00:28.739 crucial question we're exploring today on the 00:00:28.739 --> 00:00:30.940 deep dive. We're looking closely at a particular 00:00:30.940 --> 00:00:32.920 research paper on patient -reported outcome measures, 00:00:33.039 --> 00:00:36.219 PROMs after total hip replacement, and the significant 00:00:36.219 --> 00:00:38.460 bias that can crop up when patients don't complete 00:00:38.460 --> 00:00:41.119 their follow -up. Now, these PROMs, they're really 00:00:41.119 --> 00:00:43.619 vital tools. Organizations like NICE here in 00:00:43.619 --> 00:00:46.960 the UK, or the FDA over in the US, use them to 00:00:46.960 --> 00:00:49.060 understand the real world impact of surgery beyond 00:00:49.060 --> 00:00:51.659 just, say, survival rates or basic complications. 00:00:52.219 --> 00:00:54.079 They're meant to capture the patient's own view 00:00:54.079 --> 00:00:56.899 on their quality of life, their recovery. And 00:00:56.899 --> 00:00:58.579 joining us to walk us through this paper and 00:00:58.579 --> 00:01:00.799 this, well, critical area of health care research 00:01:00.799 --> 00:01:04.260 is Professor Mo Imam. Professor Imam is a leading 00:01:04.260 --> 00:01:07.060 orthopedic surgeon and researcher, very involved 00:01:07.060 --> 00:01:09.260 in outcomes analysis, and importantly, one of 00:01:09.260 --> 00:01:11.420 the authors of the actual study we're discussing 00:01:11.420 --> 00:01:13.260 today. Welcome. Thank you. It's a pleasure to 00:01:13.260 --> 00:01:15.310 be here. So maybe let's set the scene with some 00:01:15.310 --> 00:01:17.510 headline numbers from your study. You tracked 00:01:17.510 --> 00:01:19.790 patients having total hip replacements, looked 00:01:19.790 --> 00:01:21.950 at their response rates to questionnaires over 00:01:21.950 --> 00:01:25.069 time. How did that response rate change from 00:01:25.069 --> 00:01:27.969 before the operation? Well, our initial capture 00:01:27.969 --> 00:01:30.390 rate preoperatively was actually excellent, almost 00:01:30.390 --> 00:01:33.430 99 percent. Oh, wow. OK. But yes, as time went 00:01:33.430 --> 00:01:38.069 on, we did see a steady decline at six weeks 00:01:38.069 --> 00:01:40.069 after the operation. It was down to about 90 00:01:40.069 --> 00:01:42.870 .6 percent. Right. By six months, it had dipped 00:01:42.870 --> 00:01:46.090 slightly further to 89%. Then one year out, it 00:01:46.090 --> 00:01:48.569 was 83%. And by the time we reached the two -year 00:01:48.569 --> 00:01:52.849 mark, we were down to 79 % response. 79%. So 00:01:52.849 --> 00:01:55.170 quite a drop. Yes. A considerable proportion 00:01:55.170 --> 00:01:57.629 of that initial group was, well, no longer providing 00:01:57.629 --> 00:01:59.950 feedback at that point. So doing the maths quickly, 00:02:00.310 --> 00:02:02.530 that's roughly a fifth of the original patients 00:02:02.530 --> 00:02:04.250 who aren't represented in that two -year data. 00:02:04.510 --> 00:02:07.170 Now, was that drop off just random chance? Did 00:02:07.170 --> 00:02:09.580 it happen sort of? equally across all patient 00:02:09.580 --> 00:02:11.979 types? Or did your analysis find specific things 00:02:11.979 --> 00:02:15.500 linked to who didn't respond? Our analysis was 00:02:15.500 --> 00:02:17.900 very clear on this point. The non -response was 00:02:17.900 --> 00:02:20.360 absolutely not random. Right. We found quite 00:02:20.360 --> 00:02:22.580 significant associations between certain patient 00:02:22.580 --> 00:02:26.939 characteristics and their likelihood of responding 00:02:26.939 --> 00:02:28.979 to those follow -up questionnaires. OK. And that's 00:02:28.979 --> 00:02:31.240 really the crux of it, isn't it? So tell us, 00:02:31.259 --> 00:02:33.400 who were these patients who were less likely 00:02:33.400 --> 00:02:36.039 to respond? What did your study identify? Well, 00:02:36.039 --> 00:02:40.300 we found several key factors. One was age. Age, 00:02:41.039 --> 00:02:43.159 interesting. Yes, the non -responders were significantly 00:02:43.159 --> 00:02:45.479 younger than the responders, and this is true 00:02:45.479 --> 00:02:48.379 at all the follow -up points, with a p -value 00:02:48.379 --> 00:02:52.960 less than .001, so highly significant. For instance, 00:02:54.020 --> 00:02:55.919 at the two -year point, the median age of those 00:02:55.919 --> 00:02:59.340 who responded was about 69, but for the non -responders, 00:02:59.419 --> 00:03:02.990 the median age was 65. And why might that be? 00:03:03.129 --> 00:03:05.349 Does the paper suggest reasons? Well, the paper 00:03:05.349 --> 00:03:07.370 speculates a bit. It could be linked to younger 00:03:07.370 --> 00:03:09.449 patients perhaps having busier working lives, 00:03:09.530 --> 00:03:11.490 finding it harder to make time for questionnaires, 00:03:11.990 --> 00:03:14.629 or maybe feeling less connected to the health 00:03:14.629 --> 00:03:16.389 care system once that immediate post -op phase 00:03:16.389 --> 00:03:19.310 is over. OK. Younger patients not responding. 00:03:19.610 --> 00:03:21.490 That's quite an interesting finding, maybe a 00:03:21.490 --> 00:03:24.580 bit counterintuitive. What else stood out? Another 00:03:24.580 --> 00:03:26.400 really significant factor, and this is something 00:03:26.400 --> 00:03:29.000 we highlighted as quite a novel finding in this 00:03:29.000 --> 00:03:32.860 specific context, was the association with deprivation. 00:03:33.120 --> 00:03:35.039 Deprivation. How did you measure that? We used 00:03:35.039 --> 00:03:37.479 the income deprivation affecting children index, 00:03:37.879 --> 00:03:41.099 the IDACI score. It gives a measure of socioeconomic 00:03:41.099 --> 00:03:44.439 deprivation for an area here in the UK. And we 00:03:44.439 --> 00:03:47.039 saw that patients from more deprived areas, those 00:03:47.039 --> 00:03:50.740 with higher IDCI scores, were significantly less 00:03:50.740 --> 00:03:53.719 likely to respond. Again, this was a very robust 00:03:53.719 --> 00:03:57.060 finding, p -value less than .001, and it held 00:03:57.060 --> 00:03:59.159 true across all the time points we looked at. 00:03:59.259 --> 00:04:01.500 And the reasons for that? Likely multifaceted, 00:04:01.500 --> 00:04:03.800 I'd say. It could involve things like lower health 00:04:03.800 --> 00:04:06.479 literacy, perhaps, or maybe less stable housing 00:04:06.479 --> 00:04:09.939 or social situations, or simply having other 00:04:09.939 --> 00:04:12.240 more pressing concerns in their daily lives that 00:04:12.240 --> 00:04:14.699 mean completing and returning a form just isn't 00:04:14.699 --> 00:04:17.339 a priority. So, age and socioeconomic background 00:04:17.339 --> 00:04:19.740 are factors. What about their actual health status 00:04:19.740 --> 00:04:21.779 before the surgery? Did their starting point 00:04:21.779 --> 00:04:25.139 matter? Yes, it mattered profoundly. We found 00:04:25.139 --> 00:04:27.079 that patients who were less likely to respond 00:04:27.079 --> 00:04:30.139 later on had significantly lower preoperative 00:04:30.139 --> 00:04:33.740 scores on both the Oxford hip score and the EQ5D. 00:04:33.879 --> 00:04:36.259 Right, the Oxford hip score measures pain and 00:04:36.259 --> 00:04:38.879 function specifically for the hip. Exactly. So 00:04:38.879 --> 00:04:41.040 those who started with the worst baseline, meaning 00:04:41.040 --> 00:04:43.660 more pain, more functional limitation before 00:04:43.660 --> 00:04:46.339 their surgery, were actually less likely to engage 00:04:46.339 --> 00:04:48.759 with the follow -up process afterwards. To give 00:04:48.759 --> 00:04:50.740 you an example, patients who responded to all 00:04:50.740 --> 00:04:53.199 the post -op questionnaires had a median baseline 00:04:53.199 --> 00:04:56.079 OHS of 20. But those who missed all of them, 00:04:56.180 --> 00:04:58.600 they started lower, with a median OHS of just 00:04:58.600 --> 00:05:02.089 16. So that suggests that poor health -related 00:05:02.089 --> 00:05:04.810 quality of life right at the beginning is some 00:05:04.810 --> 00:05:07.189 kind of predictor for non -response later. It 00:05:07.189 --> 00:05:09.389 certainly seems to be, yes. And maybe most tellingly, 00:05:09.470 --> 00:05:12.029 you looked at whether a patient's previous experience 00:05:12.029 --> 00:05:14.410 or their score at an earlier point influenced 00:05:14.410 --> 00:05:16.129 whether they responded next time. What did that 00:05:16.129 --> 00:05:18.529 show about why people might not respond? Yes. 00:05:18.889 --> 00:05:21.250 This finding, I think, really strengthens the 00:05:21.250 --> 00:05:23.410 argument for disengagement being linked to the 00:05:23.410 --> 00:05:26.089 outcome itself. We found a strong correlation. 00:05:26.639 --> 00:05:30.180 If a patient had lower outcome scores or reported 00:05:30.180 --> 00:05:32.339 lower satisfaction at an earlier time point, 00:05:32.779 --> 00:05:35.379 they were significantly less likely to respond 00:05:35.379 --> 00:05:37.399 to the next questionnaire down the line. Oh, 00:05:37.480 --> 00:05:39.500 okay. So, for example, patients who didn't respond 00:05:39.500 --> 00:05:42.600 at the 24 -month mark, their median Oxford HIP 00:05:42.600 --> 00:05:45.259 score back at the 12 -month follow -up was 38. 00:05:46.160 --> 00:05:48.459 Compare that to those who did respond at 24 months. 00:05:48.959 --> 00:05:51.980 Their median OHS at 12 months was 43. That's 00:05:51.980 --> 00:05:53.779 a five -point difference. Is that significant? 00:05:54.009 --> 00:05:57.310 It is, yes. It's clinically meaningful and statistically 00:05:57.310 --> 00:06:00.670 significant. It strongly implies that if a patient 00:06:00.670 --> 00:06:02.470 felt they weren't achieving the outcome they'd 00:06:02.470 --> 00:06:05.050 hoped for, or perhaps wasn't satisfied with their 00:06:05.050 --> 00:06:07.550 result or the service received, they were then 00:06:07.550 --> 00:06:09.790 less inclined to participate in future data collection. 00:06:10.990 --> 00:06:12.990 It's almost like they've voted with their feet 00:06:12.990 --> 00:06:15.449 by simply not responding. That's a really powerful 00:06:15.449 --> 00:06:18.149 insight. It suggests it's not always just about 00:06:18.149 --> 00:06:20.370 practical barriers, but maybe a psychological 00:06:20.370 --> 00:06:22.769 disengagement based on how well they feel they've 00:06:22.769 --> 00:06:25.740 done. Precisely. Were there any factors you perhaps 00:06:25.740 --> 00:06:28.279 expected to be linked to non -response but actually 00:06:28.279 --> 00:06:30.639 weren't, according to your study? Well, interestingly, 00:06:30.819 --> 00:06:33.519 in our particular cohort, we didn't find a significant 00:06:33.519 --> 00:06:36.879 link between gender or whether the patient experienced 00:06:36.879 --> 00:06:39.540 surgical or medical complications after the operation 00:06:39.540 --> 00:06:42.519 and their likelihood of responding. Really? No 00:06:42.519 --> 00:06:44.980 link to complications? Not a statistically significant 00:06:44.980 --> 00:06:47.779 one in our data set, no. And the paper does note 00:06:47.779 --> 00:06:50.100 that the finding on gender actually differs from 00:06:50.100 --> 00:06:52.459 some other studies out there. Which just highlights, 00:06:52.600 --> 00:06:54.720 I suppose, that the factors influencing non -response 00:06:54.720 --> 00:06:57.220 can vary a bit, depending on the specific patient 00:06:57.220 --> 00:06:59.339 group and the setting you're looking at. Right. 00:06:59.480 --> 00:07:02.420 Context matters. OK. So we've got a clearer picture 00:07:02.420 --> 00:07:04.759 now of who tends to be missing from this follow 00:07:04.759 --> 00:07:08.100 -up data. Younger patients, those from more deprived 00:07:08.100 --> 00:07:10.740 areas, and crucially, those who started with 00:07:10.740 --> 00:07:13.519 lower scores or had poorer outcomes or satisfaction 00:07:13.519 --> 00:07:17.000 earlier on. So let's bridge that now to the core 00:07:17.000 --> 00:07:19.740 problem this paper really tackles, the bias. 00:07:20.459 --> 00:07:23.060 If that specific group is less likely to respond, 00:07:23.439 --> 00:07:25.339 what does it fundamentally mean for the outcome 00:07:25.339 --> 00:07:28.259 data that we do see reported? Well, the fundamental 00:07:28.259 --> 00:07:31.060 consequence, the big issue, is that the reported 00:07:31.060 --> 00:07:33.779 outcomes from these data sets or registries that 00:07:33.779 --> 00:07:36.500 suffer from significant loss to follow -up are 00:07:36.500 --> 00:07:39.740 inherently biased. Biased how? Because the group 00:07:39.740 --> 00:07:42.699 that is disproportionately not responding are 00:07:42.699 --> 00:07:45.160 the very patients with the worst scores, perhaps 00:07:45.160 --> 00:07:48.189 less improvement or lower satisfaction. So, the 00:07:48.189 --> 00:07:51.050 group that does respond is therefore skewed. 00:07:51.470 --> 00:07:53.790 It's weighted towards patients who are, generally 00:07:53.790 --> 00:07:57.189 speaking, doing well and are satisfied. So, the 00:07:57.189 --> 00:07:59.810 picture of success painted by those who reply 00:07:59.810 --> 00:08:02.730 is potentially overly optimistic compared to 00:08:02.730 --> 00:08:05.209 what the entire group of patients actually experienced. 00:08:05.410 --> 00:08:07.870 Precisely. And our study provides some quite 00:08:07.870 --> 00:08:10.449 concrete evidence for this bias. We look at the 00:08:10.449 --> 00:08:12.129 actual improvement in the Oxford HIPPS score, 00:08:12.470 --> 00:08:15.100 from baseline to follow -up. At 12 months, the 00:08:15.100 --> 00:08:16.660 median improvement for patients who responded 00:08:16.660 --> 00:08:18.939 was three points higher than for those who didn't 00:08:18.939 --> 00:08:20.839 respond at that point. Three points higher. And 00:08:20.839 --> 00:08:23.220 by 24 months, that difference had actually widened 00:08:23.220 --> 00:08:25.660 slightly to four points. Four points. Now, does 00:08:25.660 --> 00:08:27.259 that difference, three or four points, sound 00:08:27.259 --> 00:08:30.160 small or is it meaningful? While it might sound 00:08:30.160 --> 00:08:32.960 small numerically, it's statistically significant 00:08:32.960 --> 00:08:35.679 and it represents a tangible skew in the reported 00:08:35.679 --> 00:08:38.350 outcome data. If you only look at the responders, 00:08:38.909 --> 00:08:41.309 you are seeing results that are, on average, 00:08:41.809 --> 00:08:43.870 notably better than what the non -responders 00:08:43.870 --> 00:08:50.450 actually experienced. Is it unique to hip replacements 00:08:50.450 --> 00:08:53.429 or PROMs data? Or is it a much wider concern 00:08:53.429 --> 00:08:55.750 in health care data and registries generally? 00:08:56.090 --> 00:08:58.610 Oh, it's a much broader concern. This phenomenon 00:08:58.610 --> 00:09:01.509 where loss to follow up isn't random, where it 00:09:01.509 --> 00:09:03.549 represents a subset with different and often 00:09:03.549 --> 00:09:06.009 less favorable outcomes. Well, that's a critical 00:09:06.009 --> 00:09:07.950 consideration that really extends far beyond 00:09:07.950 --> 00:09:10.730 just hip PROMs. It's relevant for, I'd say, virtually 00:09:10.730 --> 00:09:13.610 any study or registry data set that relies on 00:09:13.610 --> 00:09:15.789 ongoing patient participation or follow up over 00:09:15.789 --> 00:09:19.340 time. So any long term tracking? Yes. The paper 00:09:19.340 --> 00:09:21.659 touches on its importance for survival analysis 00:09:21.659 --> 00:09:24.059 in joint replacement registries, for instance, 00:09:24.480 --> 00:09:27.179 and it acknowledges that the profile of non -responders 00:09:27.179 --> 00:09:29.259 might differ depending on the procedure. For 00:09:29.259 --> 00:09:32.299 example, with hip resurfacing, you might actually 00:09:32.299 --> 00:09:34.659 lose younger, more active patients to follow 00:09:34.659 --> 00:09:37.220 up precisely because they're doing so well and 00:09:37.220 --> 00:09:39.559 have just got back to their busy lives. That's 00:09:39.559 --> 00:09:41.460 a different dynamic to what we saw in total hip 00:09:41.460 --> 00:09:44.440 replacements. But the key point remains, the 00:09:44.440 --> 00:09:47.669 missing data is rarely, if ever, random. You 00:09:47.669 --> 00:09:50.490 also made a really fascinating comparison in 00:09:50.490 --> 00:09:53.669 the paper between the UK's national PROMS data 00:09:53.669 --> 00:09:56.710 and the Swedish HIP Arthroplasty Register. What 00:09:56.710 --> 00:09:58.690 did that comparison tell you about the impact 00:09:58.690 --> 00:10:01.409 of response rates? Ah, yes. That comparison offers 00:10:01.409 --> 00:10:04.370 quite powerful sort of external support for our 00:10:04.370 --> 00:10:06.809 findings on bias. The Swedish Register, you see, 00:10:07.049 --> 00:10:08.750 consistently achieves a significantly higher 00:10:08.750 --> 00:10:11.730 response rate, typically around 90%. 90 %? That's 00:10:11.730 --> 00:10:14.809 high. It is. Compare that to the national UK 00:10:14.809 --> 00:10:17.250 PROMS data, which was around 80 % at the time 00:10:17.250 --> 00:10:19.409 we did our study. Now, what's really striking 00:10:19.409 --> 00:10:22.590 is that despite presumably performing similar 00:10:22.590 --> 00:10:25.509 surgeries with similar patients, the Swedish 00:10:25.509 --> 00:10:27.750 registry reported a lower patient satisfaction 00:10:27.750 --> 00:10:30.730 rate, about 84 % satisfied. That's compared to 00:10:30.730 --> 00:10:33.649 a much higher rate, over 95 % reporting improvement 00:10:33.649 --> 00:10:36.269 in the UK PROMS data with its lower response 00:10:36.269 --> 00:10:38.529 rate. So higher response rate, lower reported 00:10:38.529 --> 00:10:40.730 satisfaction. How does that support the bias 00:10:40.730 --> 00:10:43.419 idea? Well, our interpretation is that because 00:10:43.419 --> 00:10:45.860 the Swedish registry captures a larger proportion 00:10:45.860 --> 00:10:48.259 of the total patient population, including more 00:10:48.259 --> 00:10:51.139 of those patients with less ideal outcomes, their 00:10:51.139 --> 00:10:53.440 reported satisfaction rate is actually a more 00:10:53.440 --> 00:10:55.980 accurate reflection. It's naturally lower because 00:10:55.980 --> 00:10:57.879 it includes more of the less satisfied people. 00:10:58.320 --> 00:11:00.580 It really underscores how the response rate itself 00:11:00.580 --> 00:11:02.639 directly impacts the picture of success that 00:11:02.639 --> 00:11:05.919 gets reported. That makes sense. Given this significant 00:11:05.919 --> 00:11:07.980 bias that can creep in, what are some potential 00:11:07.980 --> 00:11:10.399 ways forward? What strategies or efforts does 00:11:10.399 --> 00:11:12.980 the paper mention to try and reduce this loss 00:11:12.980 --> 00:11:14.820 to follow up and get a more complete picture? 00:11:15.220 --> 00:11:17.480 The paper really emphasizes that substantial 00:11:17.480 --> 00:11:20.220 effort is needed first to understand and then 00:11:20.220 --> 00:11:22.879 to overcome the barriers stopping patients from 00:11:22.879 --> 00:11:25.509 following up. This could involve practical things 00:11:25.509 --> 00:11:28.809 beyond just sending out standard postal questionnaires, 00:11:28.809 --> 00:11:31.149 offering different ways to respond, maybe email 00:11:31.149 --> 00:11:33.809 or telephone calls, perhaps even app -based systems. 00:11:33.889 --> 00:11:36.990 Using technology more? Potentially, yes. And 00:11:36.990 --> 00:11:39.570 also considering things like language barriers 00:11:39.570 --> 00:11:42.009 or providing extra assistance for patients who 00:11:42.009 --> 00:11:44.509 might have lower literacy levels or perhaps less 00:11:44.509 --> 00:11:47.289 access to digital technology. It's really about 00:11:47.289 --> 00:11:49.509 making the process as easy and accessible as 00:11:49.509 --> 00:11:52.250 we can for all patient groups, particularly those 00:11:52.250 --> 00:11:54.370 we've now identified. as being less likely to 00:11:54.370 --> 00:11:56.389 respond otherwise. Right, targeting the effort. 00:11:56.509 --> 00:11:58.629 And just briefly, what were some of the limitations 00:11:58.629 --> 00:12:01.090 of your own study that you acknowledged? Important 00:12:01.090 --> 00:12:04.549 for context? Yes, of course. Like any research, 00:12:04.730 --> 00:12:06.350 there were limitations. We didn't have access 00:12:06.350 --> 00:12:08.730 to really comprehensive socio -demographic details 00:12:08.730 --> 00:12:10.929 for all patients, things like education level 00:12:10.929 --> 00:12:14.350 or specific language needs. Nor did we have detailed 00:12:14.350 --> 00:12:16.870 clinical data on comorbidities or mental health 00:12:16.870 --> 00:12:19.269 conditions, which could certainly also influence 00:12:19.269 --> 00:12:22.320 response. Okay. Also, it's worth noting a low 00:12:22.320 --> 00:12:24.500 hip score isn't always solely due to the hip 00:12:24.500 --> 00:12:27.460 surgery itself. It can be influenced by other 00:12:27.460 --> 00:12:30.840 coexisting health conditions. And while our institutional 00:12:30.840 --> 00:12:33.840 data capture rate was relatively high compared 00:12:33.840 --> 00:12:36.440 to the national average at the time, these findings 00:12:36.440 --> 00:12:38.120 would definitely benefit from being validated 00:12:38.120 --> 00:12:40.299 in other populations and different health care 00:12:40.299 --> 00:12:43.279 settings. Transparency about limitations is always 00:12:43.279 --> 00:12:47.799 good practice. Okay, let's switch gears slightly 00:12:47.799 --> 00:12:50.019 lightning round. Quick questions, sharp insights. 00:12:50.620 --> 00:12:52.879 Professor, what's the single most crucial takeaway 00:12:52.879 --> 00:12:55.659 for anyone looking at patient feedback data? 00:12:56.059 --> 00:12:58.919 Assume non -responders are different. And assume 00:12:58.919 --> 00:13:01.480 their outcomes are likely less favorable than 00:13:01.480 --> 00:13:04.279 those who did respond. Never take the data you 00:13:04.279 --> 00:13:06.299 have as the whole picture unless your follow 00:13:06.299 --> 00:13:08.710 -up is near perfect. Good point. If you could 00:13:08.710 --> 00:13:11.370 implement one change tomorrow to improve PROM's 00:13:11.370 --> 00:13:13.730 data quality based purely on this study, what 00:13:13.730 --> 00:13:16.450 would it be? I'd shift focus and importantly 00:13:16.450 --> 00:13:19.070 resources towards actively understanding why 00:13:19.070 --> 00:13:21.610 certain patients aren't responding and then putting 00:13:21.610 --> 00:13:23.889 targeted strategies in place to try and re -engage 00:13:23.889 --> 00:13:26.509 them or capture their feedback effectively. Makes 00:13:26.509 --> 00:13:29.409 sense. And finally, what's the one essential 00:13:29.409 --> 00:13:31.950 question professionals should always ask themselves 00:13:31.950 --> 00:13:34.429 when they see published outcome data that's based 00:13:34.429 --> 00:13:37.769 on patient feedback? Always ask. What was the 00:13:37.769 --> 00:13:40.190 follow -up rate? And, truthfully, was any attempt 00:13:40.190 --> 00:13:42.850 made to characterize the non -responders? Because 00:13:42.850 --> 00:13:45.009 the lower that response rate, the greater the 00:13:45.009 --> 00:13:47.730 potential for bias, and the more cautious you 00:13:47.730 --> 00:13:50.330 absolutely must be about the conclusions being 00:13:50.330 --> 00:13:52.289 drawn from it. Excellent. Very clear points. 00:13:53.129 --> 00:13:55.389 So just to recap some of the most crucial takeaways 00:13:55.389 --> 00:13:57.690 from our deep dive today for everyone listening. 00:13:58.330 --> 00:14:00.629 Firstly, Patient response rates in long -term 00:14:00.629 --> 00:14:02.909 studies like PROMs often drop off significantly, 00:14:03.009 --> 00:14:05.570 which leaves quite a substantial gap in the data 00:14:05.570 --> 00:14:08.730 we collect. Secondly, and this is critical, this 00:14:08.730 --> 00:14:11.590 non -response isn't just random. Studies, like 00:14:11.590 --> 00:14:13.210 the one we've discussed, show that the patients 00:14:13.210 --> 00:14:15.909 less likely to respond are often younger, come 00:14:15.909 --> 00:14:18.690 from more deprived areas, and significantly had 00:14:18.690 --> 00:14:21.570 lower baseline scores or experienced poorer outcomes 00:14:21.570 --> 00:14:24.529 and satisfaction earlier on. Thirdly, this leads 00:14:24.529 --> 00:14:28.259 directly to a major issue. Bias. The reported 00:14:28.259 --> 00:14:30.659 outcome data, based only on those patients who 00:14:30.659 --> 00:14:33.220 do respond, is likely painting an overly positive 00:14:33.220 --> 00:14:35.320 picture compared to the true average outcome 00:14:35.320 --> 00:14:37.720 for all patients because those less successful 00:14:37.720 --> 00:14:40.600 results are disproportionately missing. Fourthly, 00:14:40.639 --> 00:14:43.159 this isn't just an orthopedics issue. This problem 00:14:43.159 --> 00:14:45.480 of non -random loss to follow -up and the bias 00:14:45.480 --> 00:14:48.120 it creates is a vital thing to consider for anyone 00:14:48.120 --> 00:14:50.000 working with patient reported outcomes, registry 00:14:50.000 --> 00:14:52.279 data, or really any research relying on ongoing 00:14:52.279 --> 00:14:54.759 patient feedback over time. And finally, getting 00:14:54.759 --> 00:14:57.360 a genuinely accurate understanding of how well 00:14:57.360 --> 00:15:01.299 treatments are truly working requires real, dedicated 00:15:01.299 --> 00:15:04.360 effort. Effort to understand and reduce the barriers 00:15:04.360 --> 00:15:06.940 to follow -up, making sure the voices and experiences 00:15:06.940 --> 00:15:09.360 of patients who had less favorable outcomes are 00:15:09.360 --> 00:15:12.159 also included so we get a more complete and much 00:15:12.159 --> 00:15:15.299 less biased picture. That was a really essential 00:15:15.299 --> 00:15:17.519 deep dive into a critical aspect of healthcare 00:15:17.519 --> 00:15:20.340 data validity. Understanding this potential for 00:15:20.340 --> 00:15:22.299 bias and patient follow -up is just fundamental 00:15:22.299 --> 00:15:24.580 for anyone making decisions based on these kinds 00:15:24.580 --> 00:15:26.840 of metrics. And if you found this deep dive valuable, 00:15:27.139 --> 00:15:28.759 please do take a moment to rate and share the 00:15:28.759 --> 00:15:30.940 show. It really helps other professionals discover 00:15:30.940 --> 00:15:33.620 insights like these. Yes. I think this research 00:15:33.620 --> 00:15:35.980 really does underscore that the methodology of 00:15:35.980 --> 00:15:39.320 how we collect data profoundly impacts the reliability 00:15:39.320 --> 00:15:41.840 of the conclusions we can draw about clinical 00:15:41.840 --> 00:15:44.620 effectiveness. It pushes us all, I hope, to be 00:15:44.620 --> 00:15:47.100 more critical consumers of outcome data. Absolutely. 00:15:47.820 --> 00:15:50.120 So perhaps a final thought for you to consider 00:15:50.120 --> 00:15:52.639 as you look at outcome data in your own professional 00:15:52.639 --> 00:15:55.490 world. Always pause and ask yourself who might 00:15:55.490 --> 00:15:57.789 not be represented in these dumpers and what 00:15:57.789 --> 00:16:00.330 might their story and their outcome truly be?