Using patient data to improve cancer waiting times May 2018
Summary Background We used patient-level data from Hospital Episode Statistics (HES) for 2016/17 to carry out detailed analysis of what drives waiting times for patients on cancer pathways and identify areas where trusts can make improvements to their performance against the 62-day cancer standard. By combining HES with many other data sources and applying econometric techniques, we have isolated the effect of each factor on cancer waiting times. This resource sets out our approach, how we conducted the analysis and the key drivers of waiting times. How can this be used operationally? The accompanying rapid improvement guide sets out practical measures trusts can take to address these areas. Key factors and their impact on cancer waiting times The two most important determinants are the number of appointments in the pathway, and the time to first appointment. Small delays at the early stage of the pathway lead to longer waits. Even for patients seen within two weeks, those who wait longer for their first appointment are more likely to wait more than 62 days for treatment. Cancellations and did not attends have a major effect, increasing the length of pathways by 20 to 30 days and the chance of breaching by 1. These not only affect the length of the pathway, but also the chance of breaching the 62-day standard. Patients who are not seen within the twoweek standard are much more likely to wait longer than the 62-day standard, and have much longer pathways overall. One-stop clinics reduced pathways by 13 days, and reduce the chance of breaching the 62-day standard by 1. 2
Linking cancer patient pathways across outpatient and admitted care We used patient-level data in HES to construct cancer pathways that follow the same cancer patient throughout their pathway. Using econometric analysis, we then identified which factors affected how long each patient waited for treatment, and whether they received treatment within 62 days. Cancer pathway data We used HES data from 2016/17 to track individual cancer patients along their entire pathway, from initial referral, through further diagnostics, to first definitive treatment. To do this, we created a unique patient pathway identifier that links all a patient s appointments and procedures along their whole pathway. Clock starts Referral Waiting time standards Entered onto patient tracking list Target: 62 days from referral to treatment 85% of patients (9 for screening referrals) should begin treatment within 62 days from referral for suspected cancer Two-week wait (2WW) 93% of patients to be seen by specialist within 14 days of urgent GP referral First outpatient appointment Forthcoming 28-day diagnostic standard Diagnostics / further appointments 31-day wait 96% of patients to be treated within 31 days from diagnosis Clock stops First definitive treatment Our sample To ensure the robustness of our analysis, we only include pathways where all data is completed accurately. This gives us a large dataset of over 250,000 cancer pathways across 81 trusts that started in 2016/17. 3
Using econometrics to analyse what affects the pathway length and the probability of meeting the 62-day target Our methods We used two different econometric methods to isolate the effects of a number of different factors on waiting times: negative binomial regressions: for all patients on a two-week wait referral for suspected cancer, to look at what affects pathway length logit regressions: for those patients who receive cancer treatment, to look at the likelihood of receiving treatment within 62 days. Analysis is split by tumour site: breast upper gastrointestinal lower gastrointestinal lung prostate Factors included in our model We are interested in identifying the effects of a number of operational factors on pathway length. But to correctly measure these effects, we need to account for other factors about the pathway that could also affect the time taken. These include patient characteristics, such as their age and gender, and details of the trust carrying out the treatment. Operational factors (subject of our analysis) Time to first outpatient appointment Total number of appointments Patient cancellations/dnas Hospital cancellations One-stop clinic Tele-appointments Multiple providers Non-operational factors (control variables) Patient characteristics Area characteristics Trust characteristics Capacity constraints 4
The number of appointments in the cancer pathway* is the most important determinant of waiting times This effect is based on all other characteristics of the pathway being the same. It applies to the probability of breaching the 62-day standard as well as the total length of the pathway. Effect of number of appointments on pathway length / probability of breaching +5% point +23% point increase in increase in probability of probability of breaching** breaching** 1 appointment 12 days (median) 2 appointments 3 appointments +29 days +19 days 4+ appointments +15 days 0 20 40 60 Additional pathway days relative to 1 appointment (all patients on 2WW pathway for suspected cancer) For example If a patient has three appointments before pathway ends (for treatment or otherwise), on average: their pathway is 19 days longer the likelihood of breaching increases by 5% points relative to if they had two appointments, all else being equal (for all cancers aggregated). This might seem obvious. However, we re able to show: robust evidence for the scale of the effect in days for patients how this differs for different cancer types how this also affects performance against the 62-day standard. Operational insight This supports a focus on streamlining pathways, where clinically appropriate. 5 * Refers to number of appointments between referral and removal from patient tracking list for cancer treatment or otherwise ** Relative to two appointments (treated patients only)
Time to first appointment is the second most important determinant of waiting times Time to first appointment is obviously key to meeting the two-week wait standard. However, we have been able to show that even small delays during the early stages are not made up later on in the pathway. This means those patients not only wait longer overall, but are also increasingly likely to wait longer than the 62-day standard to receive their treatment. Effect of time to first appointment on pathway length / probability of breaching For example Additional pathway days 1 40 35 30 25 20 15 10 5 25 20 15 10 5 Percentage point increase in probability of breaching 62-day standard 2 If a patient waits 10 to 12 days for their first appointment, on average: their pathway is 10 days longer the likelihood of breaching increases by 5% points relative to if they waited 1-5 days, all else equal (for all cancers aggregated). 0 1-5 (baseline) 6-9 10-12 13-14 15+ Days to first appointment 0 Operational insights If a patient waits longer for their first appointment but is still seen within 14 days, they do not catch up and the probability that they will wait longer than the 62-day target increases. If they do not have their first appointment within 14 days, there is a much greater impact on both their total pathway length and the probability of breaching. Overall: trusts should focus on reducing time to first appointment to meet all standards, not just the two-week wait target. 6 1 All patients; 2 Treated patients only
Other important determinants of pathway length and likelihood of breaching Attending a one-stop clinic where a patient attends multiple appointments in a single day significantly reduces both pathway length and probability of breaching. Appointments cancelled or not attended, and patients attending multiple providers increase waiting times and breaches. There is a small but significant positive link between GP quality (measured by cancer Quality and Outcomes Framework points) and meeting the 62-day standard. Factor Effect on pathway length Effect on probability of breaching 62-day standard (% points) One-stop clinic 13 days 9% DNA/patient cancellation 33 days 11% Hospital cancellation 20 days 9% Total cancer referrals by trust, diagnostic performance, GP performance and region also affect performance. Multiple providers 4 days 2 Referring GP achieving full QOF points for cancer - 2% Operational insights Increasing roll-out of one-stop across tumour sites and trusts could significantly reduce total pathway length and probability of breaching the 62-day target. Improving processes for reducing DNAs and cancellations, and streamlining transfers of patients between multiple trusts, could also markedly reduce waiting times. 7
Differences in the number of pathway points across different tumour sites All suspected cancer patients 10 6 4 2 10 6 4 2 10 6 4 2 Breast Lung Prostate 10 6 4 2 10 6 4 2 Upper GI Lower GI 10 Most breast and lung cancer pathways contain only one appointment. Upper and lower GI pathways are most likely to have two appointments before pathway ends, which could reflect a more challenging diagnostic process that needs to be spread over several days. Variation in average number of points (episodes/spells) per pathway by cancer type Only patients who receive treatment 6 4 2 10 6 4 2 Breast Lung 10 6 4 2 10 6 4 2 10 6 4 2 Upper GI Lower GI Prostate 8
Detailed differences across tumour sites Our main results reported above are for all cancers, but we have also run our analysis separately for each of the tumour sites in our sample. These results are reported below. Additional days in pathway by cancer type Additional probability of breaching by cancer type Breast Upper GI Lower GI Lung Prostate All cancers Breast Upper GI Lower GI Lung Prostate All cancers Days to first appointment 6-9 (versus 1-5) 5 5 6 6 4 8% 9% 4% 10-12 (versus 1-5) 8 6 10 12 11 10 2% 12% 12% 5% 13-14 (versus 1-5) 11 11 13 16 15 14 3% 11% 14% 7% 15+ (versus 1-5) 19 42 35 41 37 37 12% 22% 26% 8% 18% 2 No. of appointments 2 (versus 1) 20 25 20 35 43 29 3 (versus 1) 26 53 46 56 51 48 12% 9% 16% 5% 5% 4+ (versus 1) 29 82 62 79 64 63 14% 35% 26% 3 31% 23% One-stop clinic -7-23 -17-17 -15-13 -5% -11% -7% -9% -16% -9% DNA/patient cancelled 12 45 36 40 28 33 5% 12% 6% 17% 16% 11% Hospital cancelled 8 31 26 27 13 20 6% 12% 8% 19% 13% 9% Telephone appointment -5-7% >1 provider involved 13 4 4% 18% 18% 14% 31% 21% Full QOF -3-2% Observations 116,707 29,242 48,509 18,453 38,470 257,379 4,991 1,153 3,615 1,240 4,384 15,580 Key differences between cancers The impact of all factors on pathway length and probability of breaching is generally smaller for breast cancer than for the other tumour sites, reflecting a higher degree of pathway standardisation. Time to first appointment is less important for lung and prostate pathways: only waits of 15 days or more have a significant effect on the probability of breaching for these tumour sites. One-stop clinic is more important for tumour sites other than breast (especially upper GI): this is likely to be because onestop clinic is already standard practice in breast cancer, and GI cancers tend to require more diagnostic tests. Use of telephone appointments is associated with shorter breast cancer pathways and lower probability of breaching in prostate, which does not show up in the all-cancer analysis. Transferring from one provider to another has the biggest effect on prostate cancer pathways. 9
Contact us: NHS Improvement Wellington House, 133-155 Waterloo Road, London, SE1 8UG 0300 123 2257 enquiries@improvement.nhs.uk improvement.nhs.uk @NHSImprovement This publication can be made available in a number of other formats on request. NHS Improvement May 2018 Publication code: IT 04/18