Document resource
WHAT IS ALREADY KNOWN ON THIS TOPIC The NHS best practice tariff pay-for-performance scheme is widely used in secondary care in England in different disease areas but little research exists about its effectiveness in improving patient care and outcomesWhether patients who meet the adult asthma best practice tariff criteria (receipt of a timely respiratory specialist review and a discharge bundle focused on improved asthma control) have a lower risk of readmission than those who do not has not been assessedWHAT THIS STUDY ADDS Patients whose care met the best practice tariff standard of care did not have a lower risk of readmission to hospitalReceiving a discharge bundle, a key element of the best practice tariff, was associated with reduced readmission to hospitalHOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE, OR POLICY These results highlight the importance of implementing the discharge bundle for patients admitted to hospital with acute asthmaIntroduction Asthma is a chronic respiratory disease characterised by airways inflammation, reversible airflow obstruction, bronchial hyper-responsiveness, and variable and recurrent symptoms, such as wheeze, breathlessness, a tight chest, and cough. 1 Asthma is estimated to affect 262 million people globally and is responsible for 20% of all disability adjusted life years caused by chronic respiratory diseases.2 3 Asthma attacks are a common feature of asthma, where symptoms become more severe over a short period of time. Although asthma attacks can often be self-treated by individuals with prescribed drug treatments, in some cases the asthma attacks result in a visit to the emergency department and, in more severe cases, hospital admission. As well as the often substantial burden asthma has on individual patients, emergency asthma care in the UK has been estimated to cost about £1.3bn (€1.5bn; $1.8bn) annually.4Many systems are used worldwide to encourage good quality care in hospital, including public reporting of performance data, performance feedback on clinician behaviour, and hospital accreditation.5 One such system is the pay-for-performance system, used extensively in the US, England, and France.6 In pay-for-performance schemes, providers are paid an additional payment if they meet specific criteria, or payment is withheld if criteria are not met. For example, in the US, the Centers for Medicare and Medicaid Services Hospital Acquired Conditions Reduction Programme withholds 1% of Medicare revenue from the worst performing quarter of hospitals according to a defined set of preventable inhospital adverse events.7The best practice tariff is a pay-for-performance incentive used by NHS England to encourage optimal patient care in hospital.8 The best practice tariff covers a wide range of diseases and procedures, ranging from Parkinson's disease to endoscopy. In England, providers are funded under the NHS payment scheme. NHS England pays a base price to hospitals for each admission and elective procedure, which varies according to disease, complexity, and length of stay. For diseases or procedures included within the best practice tariff, the tariff price is also paid to hospitals that meet specific criteria for a specific percentage of patient activity. These criteria can be data quality, care received, audit involvement (such as having a named clinical lead), and other relevant metrics. The use of the best practice tariff is intended to provide an incentive to hospitals to meet these criteria and improve overall patient care by "reducing unexplained variation and universalising best practice."9Introduced in 2021,10 the best practice tariff for adult asthma includes timely respiratory specialist review and a discharge bundle for patients that has elements focused on reducing readmission thospitalal. Hospitals that meet the best practice tariff receive an additional uplift of 9.6% as well as the base price reimbursement for adult patients with asthma, which corresponds to £71-434, depending on the complexity and length of stay of the hospital admission.11 Despite the best practice tariff covering 20 diseases and procedures in England, a recent systematic review identified only seven studies where the performance of the best practice tariff was assessed, with just two of the 20 areas in the best practice tariff (hip fracture and chronic obstructive pulmonary disease (COPD)) being included.12 At a time of increasing financial constraint and a need to streamline resources and optimise care, this study adds to the research on the best practice tariff by assessing the association between receiving best practice tariff standard of care and 30 day and 90 day readmission in adults admitted to hospital with an asthma attack.Methods Setting and data sources This study used hospital audit data from the National Respiratory Audit Programme, a continuous national audit of asthma and COPD in primary, secondary, and tertiary care in England and Wales. The National Respiratory Audit Programme was commissioned by the Healthcare Quality Improvement Partnership and is carried out by the Royal College of Physicians. This study used data from the 2022-23 adult asthma secondary care dataset. 13 The dataset has information on adult patients with acute asthma exacerbation discharged from hospital between 1 April 2022 and 31 March 2023. Hospitals submit clinical data on emergency admissions for asthma with an online tool, with patients then linked by their unique NHS number to Hospital Episode Statistics Admitted Patient Care, Patient Episode Dataset for Wales, and Office for National Statistics mortality data. Individual patient admissions were linked to their particular admission in Hospital Episode Statistics Admitted Patient Care and Patient Episode Dataset for Wales by admission date and discharge date within a tolerance of one day. Linkage is necessary to produce the outcomes report, which reports on 30 day and 90 day mortality and readmission for the index admission. Along with performance monitoring, the data collected are used to assess conformity with the best practice tariff.Study design This study followed a longitudinal cohort design with a short 90 day follow-up period. Patient variables of interest and confounding variables were collected from routine hospital care and entered into the audit during or shortly after admission. Outcome data were collected through linked data up to 90 days after patient discharge.Participants To be eligible for the adult asthma audit, patients must be aged ≥16 years and admitted to an adult ward with a primary diagnosis of asthma attack (recommended ICD-10 (international classification of diseases, 10th revision) codes 45.0 (predominantly allergic asthma), J45.9 (asthma, unspecified), and J46.X (status asthmaticus, which includes acute severe asthma)). An admission is defined as an episode where a patient is admitted and stays in hospital for ≥4 hours. Admission can be to emergency medicine centres, acute medical units, clinical decision units, short stay wards, or similar, but excludes patients treated transiently before discharge from the emergency department. 14 The data then undergo a cleaning process to remove duplicates and records with logical inconsistencies. All scripts used in the cleaning process can be found at https://github.com/NationalAsthmaCOPDAudit. Inclusion criteria for this study were audit patients who were: present in the audit outcomes dataset (ie, index audit record was successfully linked with Hospital Episode Statistics or Patient Episode Dataset for Wales and Office for National Statistics); admitted to hospitals in England; recorded as male or female sex; and alive at discharge and not transferred to another hospital. Data for sex were taken from information in the National Respiratory Audit Programme rather than from patient reported gender. The patient's index admission was used for data analysis.Variables The primary variable of interest was meeting the best practice tariff standard of care for adult asthma recorded in the National Respiratory Audit Programme. 8 For a patient to meet the best practice tariff criteria, they must have received both a respiratory specialist review within 24 hours of admission and a discharge bundle with the required elements. A respiratory specialist review should be given by respiratory team members who "might be defined locally to include respiratory health professionals deemed competent at seeing and managing patients with acute asthma attack. These staff members might include a respiratory consultant, respiratory trainee of ST3 or above, respiratory specialist nurse or specific asthma nurse, or physiotherapist."8 The discharge bundle is "a group of evidence based items that should be implemented or checked and verified on discharge from hospital."8 The most recent 2023-25 best practice tariff guidelines erroneously refer to three mandatory elements and two optional elements, despite listing four elements with no mandatory or optional classification. The sensible interpretation of this guideline is that all four elements are required for the best practice tariff, and that the reference to optional and mandatory elements is from the previously well defined 2022-23 best practice tariff where five optional and mandatory elements were listed.15 The discharge bundle is produced by the British Thoracic Society and its most recent iteration, published in 2024, includes four elements (The Asthma 4, box 1).16Box 1The British Thoracic Society discharge bundle (The Asthma 4)16Medication review. Ensure all patients are prescribed an inhaled corticosteroid containing inhaler. All patients should be observed using their inhalers and coached to improve their technique as necessary. Adherence to inhaled corticosteroid containing inhaler should be assessed and importance of good adherence discussed.A personalised asthma action plan. All patients should receive a written (or electronic) self-management personalised asthma action plan that has been co-designed with the patient and is individualised for their situation, taking into account their health literacy, language, and appropriate treatment.Tobacco dependence advice and support for current smokers. All current smokers should be given very brief advice on tobacco dependence and referred to specialist support. Those vaping should have a plan in place to eventually stop vaping.Clinical review within four weeks. Clinical review within four weeks by a healthcare professional trained in asthma care to confirm the diagnosis of asthma and review oral steroid requirement, biomarkers, and the need for further follow-up.The audit collects data on whether a patient has received a respiratory review and the date and time, and the time when the patient arrived in hospital. Because admission time data are not collected, we used a metric of 24 hours from arrival to respiratory review instead of 24 hours from admission to respiratory review. We conducted a sensitivity analysis that accounted for the delay between arrival and admission, and allowed a respiratory specialist review within 28 hours of arrival (reflecting the target maximum time a patient should spend between arrival at hospital and admission) and within 32 hours of arrival (reflecting the median time between arrival and admission, as found in previous audits for COPD) to be defined as meeting the best practice tariff standard of care. The audit also collects data on whether the patient received a discharge bundle and the individual elements within the bundle. The data collected by the audit divides the medication review into three parts: drug treatment review, inhaler technique checked, and adherence checked. All three elements must be checked for a patient to meet the medication review criterion. The audit also collects data on whether community follow-up is requested within two working days as part of the older discharge bundle; this element satisfies the element, clinical review within four weeks.The primary variable of interest was defined as receiving best practice tariff standard of care (ie, respiratory specialist review within 24 hours, and a discharge bundle comprising a medication review, provision or review of a personalised asthma action plan, tobacco dependence support and advice if the patient was a current smoker, and clinical review within four weeks). Secondary variables of interest were receiving: individual elements of the best practice tariff; clustered elements of the best practice tariff as defined above (ie, medication review, any follow-up within four weeks); respiratory specialist review at any time; and none of the required elements of the discharge bundle.Outcomes The prespecified primary study outcomes were 30 day and 90 day readmission to hospital for an asthma attack (primary diagnosis codes J45.0, J45.9, and J46.X). Secondary study outcomes were 30 day and 90 day hospital readmission for any cause. Outcome variables were treated as binary variables in the analysis. Thirty day all cause emergency readmission is a standard metric measured by the NHS for healthcare performance monitoring because it gives an indication of whether a patient has been treated well in hospital before discharge. 17 Ninety day readmission looks at readmissions over a longer period of time, which can provide higher power if the 30 day readmission rate is low.Potential confounders The audit collects data on patient characteristics and care received in hospital, and also includes comorbidity data extracted from ICD-10 coded hospital records from linked Hospital Episode Statistics data. Variables that were a priori assessed as being possible confounders of the relation between meeting the best practice tariff criteria and readmission were adjusted for in the analyses, based on prior clinical knowledge and previous studies. 18 These variables were: sex; age at admission (continuous variable); index of multiple deprivation,19 a measure of relative deprivation based on the lower super output area of the patient's address, which was included as a categorical variable based on five groups, from 1 (most deprived) to 5 (least deprived), with a separate category for missing data (eg, if the patient was homeless or did not live in England); smoking status (never smoker, ex-smoker, current smoker, and not recorded); whether the patient had been prescribed three or more courses of rescue or emergency oral steroids in the 12 months before admission (yes, no, or not recorded); Charlson comorbidity index20 based on the updated weights from Quan et al,21 included as a categorical variable (0-1, 2, 3, 4, 5, and ≥6), indicating comorbidity severity ranging from less severe to more severe; and severity of asthma attack (classified as moderate, severe, or life threatening).The severity of the asthma attack was classified following guidance from the National Institute for Health and Care Excellence (NICE)22 with available physiological data only because data describing clinical signs were not available. Severe and life threatening exacerbations could not be reliably distinguished without clinical signs and therefore were combined. A severe or life threatening exacerbation was defined as having: oxygen saturation <92%, respiratory rate ≥25 breaths/min, heart rate ≥110 beats/min, per cent predicted peak flow ≤50%, respiratory rate <10 breaths/min, heart rate <30 beats/min, or a patient being too unwell to perform a peak flow (the latter three were used as proxies for signs of cardiorespiratory decompensation included in the definition of life threatening).Handling of quantitative variables Age was included as a continuous variable after graphical inspection of its relation with the outcome. Raw data for index of multiple deprivation data were provided as a rank, so the use of five groups for index of multiple deprivation as a categorical variable was appropriate. The Charlson comorbidity index was categorised because the distribution was highly skewed; most patients had a low Charlson comorbidity index and therefore making an assumption as to the underlying relation between the Charlson comorbidity index and outcome was not possible.Missing data The audit tool does not allow individuals to be submitted with incomplete data, but some variables include a not recorded category when information is not present in the patient notes. Because not recorded can often be synonymous with no in routinely collected data (eg, presence or absence of a disease code), data missing in this way were treated as a separate variable category in the analyses.Bias Selection bias Minimal selection criteria were applied to the audit dataset in deciding on the study population, so if selection bias is present, the likely cause is the process in which patients were included in the audit. The audit is designed to reflect the general population and in theory should capture all adult hospital admissions for asthma, although in practice case ascertainment is low (45.1%). 13 Selection bias could occur at a hospital level if better performing or resourced hospitals were more likely to take part in the audit or inputted more patients, resulting in a population that was not reflective of the general population. In this case, our results might not be generalisable to patients with asthma at hospitals with a lower engagement rate with the national audit.Collider bias Collider bias could be present in the study if the decision by a hospital to input a particular patient into the audit was associated with the patient's standard of care and also the patient's likelihood of readmission. Hospitals might be more likely to preferentially input data on patients who received the best standard of care, and they also might be less inclined to input data for patients who had a higher risk of readmission. This could lead to a spurious correlation between standard of care and readmission, although assessing the likelihood of this bias is difficult.Confounding Potential confounding variables were adjusted for in the model. Although the associations between the variables in the model and hospital admission are well established, 18 the relation to receiving best practice tariff care is not known. Given that all patients with asthma discharged alive should receive a timely specialist review and discharge bundle regardless of their characteristics, we do not believe that confounding would have a substantial effect on this study. To avoid the pitfalls of stepwise variable selection, however, we decided to include potentially confounding variables in the model a priori. We selected from the variables that were available to us, but if important variables were not available, then residual confounding will remain.Measurement error Data for this study were collected as part of the national audit with an online web tool. This tool has inbuilt validity checks and pop-up warning messages that reduce the likelihood of erroneous data being submitted to reduce the risk of random error. Non-differential measurement error, if present, will bias results towards the null, although we consider the risk of substantial measurement error to be low because of the method of data collection. If variables are systematically recorded incorrectly, bias can be introduced into the study. The study was a cohort study, with readmission determined with linked data after the collection of patient and confounding variables, meaning that differential measurement error is unlikely to bias the association between the variables and outcomes in this study.Sensitivity analyses To be eligible to be entered into the audit, patients must have a primary diagnosis of asthma. On inspecting the data, a substantial number of patients were found to not meet this criterion according to their linked Hospital Episode Statistics record. This finding could introduce bias if these patients were not given the best practice tariff and were more or less likely to be admitted to hospital than patients admitted with a primary diagnosis of asthma. To account for this potential for bias, we performed a sensitivity analysis restricted to patients who received a primary diagnosis of asthma in Hospital Episode Statistics.Patients who self-discharged will be less likely to receive discharge bundle elements and might be more likely to be readmitted to hospital. For this reason, a sensitivity analysis was undertaken excluding these patients. Furthermore, being administered or prescribed inhaled or oral steroids during admission or at discharge is likely to be associated with reduced hospital readmission, but prescription of steroids might lie on the causal pathway between receiving best practice tariff care and readmission. To take into account that this assumption might be incorrect and that prescription of steroids is unrelated to best practice tariff standard of care, we conducted a sensitivity analysis that also included receiving steroids as a confounder. Finally, a complete case analysis was performed to assess the effect of including a missing category in the regression model for some variables.Statistical methods Differences in patient characteristics between patients who met and did not meet the best practice tariff criteria were tested statistically with χ 2 tests for categorical variables and the Kruskal-Wallis test for continuous variables. Mixed effects logistic regression analyses were used to assess the relation between variables (best practice tariff criteria and individual elements) and readmission. All outcomes and variables of interest were modelled as binary variables. Adjusted models included each variable of interest with all potential confounders included as covariates. Potential confounders were included in the model a priori based on clinical judgment without using predictor selection methods. Unadjusted models included the variable of interest only. Both adjusted and unadjusted regression models included hospital as a random intercept to account for the hierarchical nature of the data, with patients grouped within hospitals. Coefficients are presented as (adjusted) odds ratios with 95% confidence intervals (CIs). The absolute risk reduction was calculated from the logistic regression models with the method of Austin.23 Number needed to treat was calculated as the inverse of the absolute risk reduction.Dominance analysis was used to identify the most important elements of the discharge bundle in the context of 30 day and 90 day readmission to hospital for asthma. Dominance analysis is a method used to identify which predictors in a model contribute the most to the fit of the model.24–27 This method is useful when predictors are highly correlated, as in our model, because discharge bundle elements are intrinsically correlated as part of the discharge bundle and elements might have complex interactive effects that cannot be sufficiently captured by our model. Online supplemental material has a brief description of the dominance analysis and how it works.SP110.1136/bmjmed-2025-001398.supp1Supplementary dataIn our analysis, we assessed the dominance of each discharge bundle element required to meet the best practice tariff criteria, with the exception of the smoking cessation element because this element was only relevant for smokers. For the fit statistic, we used the marginal R2GLMM as defined by Nakagawa and Schielzeth28 for use in generalised linear mixed effects models. Potentially confounding variables were included in the model, but their dominance was not assessed. Therefore, the standardised values of the dominance analysis do not refer to the proportion of the R2 explained by each discharge bundle element, but instead refer to the proportion of the R2 explained by each discharge bundle element after subtracting the R2 explained by the potentially confounding variables from the total R2 of the model. To account for the removal of the smoking cessation element from the dominance analysis, the smoking status variable was modified so that the current smoker category was divided into current smoker with cessation advice delivered and current smoker without cessation advice delivered.Statistical analysis was carried out with R version 4.4.1.29 Mixed effects models were performed with the glmer function within the lme4 package,30 with the R2 extracted with the glmm.r-square function from the MuMIn package.31 Dominance analysis was carried out with the domin function in the domir package.32 Tables were created with the finalfit package33 and plots were created with ggplot2.34Study size The study used all records present in the audit data that met the eligibility criteria. To avoid overfitting of regression models, a minimum of 10 outcome events was required for each predictor included in the model. The ratio of outcome events to predictors was >10 for all analyses, so we do not consider that overfitting was a concern in our study.Patient and public involvement Patients and the public were not directly involved in this piece of research because the study uses routinely collected audit data. However, patients are directly involved in shaping the NRAP (National Respiratory Audit Programme) through its patient panel and results from this study will be disseminated as part of the NRAP newsletter.Results We identified 13 706 index admissions in the audit dataset, of whom 12 964 patients met the eligibility criteria and were included in the study. Of the patients who did not meet the eligibility criteria, 21 were not present in the outcomes dataset because of failed linkage, 643 were admitted to hospitals outside of England, 21 did not have a recorded sex of male or female, and 57 died in hospital or were transferred to another hospital. Online supplemental figure 1 shows a STROBE (Strengthening the Reporting of Observational Studies in Epidemiology) flowchart of the selection of the study population; 3627 (28.0%) patients received care that met the best practice tariff criteria. Patients who received care that met the best practice tariff criteria were more likely to be in the less deprived group (P<0.001), younger (median age 48 v 52 years), non-smokers (P<0.001), and prescribed inhaled and oral steroids in hospital or at discharge (P<0.001). These patients were less likely to have comorbid conditions (P<0.001) and to have had a more severe asthma attack (P<0.001), but were more likely to have been prescribed three or more courses of rescue steroids in the previous 12 months and have a record of this prescription. We found that 538 (4.1%) and 1077 (8.3%) patients were readmitted to hospital with asthma within 30 days and 90 days, respectively. Table 1 shows more details. Most patients (79.5%) who received a respiratory specialist review received a discharge bundle (8596/10 816) compared with 19.4% of those who did not receive a specialist review (417/2148).Table 1Characteristics of patients admitted to hospital with acute asthma who did or did not receive care that met the criteria for the best practice tariffCharacteristicsBest practice tariff standard of care met (n=3627, 28.0%)Best practice tariff standard of care not met (n=9337, 72.0%)All patients (n=12 964)P value*Median (IQR) age (years)48.0 (33.0 to 63.0)52.0 (35.0 to 67.0)51.0 (35.0 to 66.0)<0.001Men1080 (29.8)2754 (29.5)3834 (29.6)0.769Index of multiple deprivation (group 1=most deprived, group 5=least deprived) 1961 (26.5)2920 (31.3)3881 (29.9)<0.001 2801 (22.1)2175 (23.3)2976 (23.0) 3731 (20.2)1548 (16.6)2279 (17.6) 4559 (15.4)1376 (14.7)1935 (14.9) 5534 (14.7)1199 (12.8)1733 (13.4) Missing41 (1.1)119 (1.3)160 (1.2)Smoking status Never smoked1842 (50.8)4189 (44.9)6031 (46.5)<0.001 Ex-smoker991 (27.3)2150 (23.0)3141 (24.2) Current smoker712 (19.6)1890 (20.2)2602 (20.1) Not recorded82 (2.3)1108 (11.9)1190 (9.2)Patient prescribed inhaled steroids at discharge Yes3522 (97.1)8124 (87.0)11 646 (89.8)<0.001 No100 (2.8)1189 (12.7)1289 (9.9) Not prescribed for medical reasons5 (0.1)24 (0.3)29 (0.2)No of patients prescribed oral steroids for at least 5 days during admission or at discharge3481 (96.0)8080 (86.5)11 561 (89.2)<0.001Patient prescribed three or more courses of rescue or emergency oral steroids in the 12 months before admission Yes1418 (39.1)2081 (22.3)3499 (27.0)<0.001 No1882 (51.9)4148 (44.4)6030 (46.5) Not recorded327 (9.0)3108 (33.3)3435 (26.5)Severity of asthma attack Moderate1132 (31.2)3606 (38.6)4738 (36.5)<0.001 Severe and life threatening2495 (68.8)5731 (61.4)8226 (63.5)Charlson comorbidity index 0-13085 (85.1)7601 (81.4)10 686 (82.4)<0.001 2230 (6.3)568 (6.1)798 (6.2) 3218 (6.0)777 (8.3)995 (7.7) 458 (1.6)222 (2.4)280 (2.2) 522 (0.6)92 (1.0)114 (0.9) ≥614 (0.4)77 (0.8)91 (0.7)No of patients reviewed by a respiratory specialist within 24 hours of arrival3627 (100.0)2652 (28.4)6279 (48.4)No of patients reviewed by a respiratory specialist during admission3627 (100.0)7189 (77.0)10 816 (83.4)—No of patients discharged with a discharge bundle including all elements required to meet the best practice tariff3627 (100.0)2385 (25.5)6012 (46.4)—No of patients recorded as having received a discharge bundle3627 (100.0)5386 (57.7)9013 (69.5)—No of patients who received a review of their asthma medicine, adherence, and inhaler technique (ie, the medicine section of the Asthma 4 discharge bundle element)3627 (100.0)4477 (47.9)8104 (62.5)—No of patients who had their inhaler technique checked (discharge bundle element)3627 (100.0)5549 (59.4)9176 (70.8)—No of patients who had their inhaler drug treatment reviewed (discharge bundle element)3627 (100.0)6905 (74.0)10 532 (81.2)—No of patients who had their adherence to drug treatment checked (discharge bundle element)3627 (100.0)5077 (54.4)8704 (67.1)—No of patients who received a personalised asthma action plan or patient's plan was reviewed (discharge bundle element)3627 (100.0)3647 (39.1)7274 (56.1)—Community follow-up requested within 2 working days or specialist review requested within 4 weeks (discharge bundle element)3627 (100.0)5895 (63.1)9522 (73.4)—Community follow-up requested within 2 working days (discharge bundle element)2900 (80.0)3622 (38.8)6522 (50.3)<0.001Specialist review requested within 4 weeks (discharge bundle element)3302 (91.0)4924 (52.7)8226 (63.5)<0.001No of patients who did not receive any discharge bundle elements0 (0.0)1481 (15.9)1481 (11.4)—No of patients who smoked that received smoking cessation advice (discharge bundle element)712 (100.0)1120 (59.3)1832 (70.4)—No of patients readmitted for any cause within 30 days of discharge337 (9.3)967 (10.4)1304 (10.1)0.075No of patients readmitted with asthma within 30 days of discharge143 (3.9)395 (4.2)538 (4.1)0.491No of patients readmitted for any cause within 90 days of discharge697 (19.2)1826 (19.6)2523 (19.5)0.679No of patients readmitted with asthma within 90 days of discharge321 (8.9)756 (8.1)1077 (8.3)0.174Values are number (%) unless indicated otherwise.*χ2 tests for categorical variables and Kruskal-Wallis test for continuous variables.IQR, interquartile range.After adjusting for potential confounders, patients meeting the best practice tariff criteria had similar odds of 30 day or 90 day readmission to hospital for asthma as those who did not meet the criteria (30 day readmission adjusted odds ratio 0.88, 95% CI 0.71 to 1.08; 90 day readmission adjusted odds ratio 1.01, 0.87 to 1.17) (figure 1 and online supplemental table 2). Table 2 shows the results for all model covariates. Online supplemental tables 1 and 2 show the unadjusted and adjusted analysis results for readmission for asthma and all causes. When the best practice tariff was divided into its constituent elements and assessed separately, receiving a discharge bundle was associated with a nearly 39% reduction in the odds of 30 day readmission for asthma (adjusted odds ratio 0.61, 95% CI 0.50 to 0.75; number needed to treat 68, absolute risk reduction 0.015) and a 23% reduction in the odds of 90 day readmission for asthma (adjusted odds ratio 0.77, 0.65 to 0.89; number needed to treat 67, absolute risk reduction 0.015). Patients who received a respiratory specialist review within 24 hours had similar odds of readmission as patients who did not (30 day readmission adjusted odds ratio 0.92, 95% CI 0.76 to 1.10; 90 day readmission adjusted odds ratio 1.01, 0.89 to 1.16), but patients who received a specialist review at any point during admission had a 30% lower odds of 30 day readmission for asthma (adjusted odds ratio 0.70, 0.55 to 0.89; number needed to treat 75, absolute risk reduction 0.013) (figure 1 and online supplemental table 2). All discharge bundle elements were more strongly associated with reduced 30 day than 90 day readmission for asthma.The sensitivity analysis, where receiving respiratory specialist review was defined as occurring within 28 hours of arrival to meet best practice tariff standard of care, was not significantly associated with readmission. When this definition was further increased to within 32 hours of arrival, a statistically significant positive association was found between receiving best practice tariff standard of care and 30 day readmission (adjusted odds ratio 0.80, 95% CI 0.65 to 0.98). Online supplemental table 3 has the full results.Table 2Adjusted odds ratios for association between receiving best practice tariff quality of care and 30 day and 90 day readmission to hospital for asthma, for adults admitted to hospital with acute asthmaModel covariateOdds ratio (95% CI) for 30 day readmission for asthmaOdds ratio (95% CI) for 90 day readmission for asthmaPatient meets best practice tariff quality of care (reference=no)—— Yes0.88 (0.71 to 1.08)1.01 (0.87 to 1.17)Index of multiple deprivation (reference=group 1)—— 20.90 (0.70 to 1.14)1.02 (0.86 to 1.22) 30.89 (0.68 to 1.16)0.92 (0.75 to 1.12) 41.04 (0.80 to 1.37)1.00 (0.81 to 1.22) 50.84 (0.62 to 1.14)0.94 (0.75 to 1.17) Missing0.72 (0.29 to 1.78)0.99 (0.55 to 1.78)Age (for every additional year of age)1.00 (0.99 to 1.00)0.99 (0.99 to 1.00)Sex (reference=men)—— Women1.26 (1.03 to 1.54)1.16 (1.00 to 1.34)Smoking status (reference=never smoked)—— Ex-smoker1.05 (0.85 to 1.30)1.12 (0.96 to 1.30) Current smoker0.93 (0.72 to 1.19)0.81 (0.68 to 0.98) Not recorded1.11 (0.81 to 1.51)0.97 (0.77 to 1.24)Patient prescribed three or more courses of rescue or emergency oral steroids in the 12 months before admission (reference=yes)—— No0.48 (0.39 to 0.59)0.42 (0.36 to 0.50) Not recorded0.59 (0.46 to 0.75)0.54 (0.46 to 0.65)Severity of asthma attack (reference=moderate)—— Severe and life threatening1.00 (0.83 to 1.20)1.15 (1.00 to 1.32)Charlson comorbidity index (reference=0-1)—— 21.30 (0.93 to 1.82)1.19 (0.92 to 1.54) 31.11 (0.80 to 1.54)1.42 (1.13 to 1.78) 40.69 (0.33 to 1.43)0.68 (0.39 to 1.18) 50.89 (0.32 to 2.46)1.11 (0.56 to 2.23) ≥60.54 (0.13 to 2.20)0.90 (0.39 to 2.08)CI, confidence interval.Figure 1Adjusted odds ratios for association between 30 day and 90 day readmission to hospital for asthma and receiving best practice tariff standard of care and associated elements, for adults admitted to hospital with acute asthma. The nesting structure of the elements within the best practice tariff standard of care is indicated by amount of indentation. Presented elements sometimes overlap and are not mutually adjusted for each other. Presented variables represent the association between a particular element and outcome, adjusted for age, sex, index of multiple deprivation group, smoking status, severity of asthma attack, Charlson comorbidity index, and a patient history of three or more courses of rescue or emergency oral steroid prescriptions in the 12 months before admission, with a clustering effect for hospitalFigure 2 shows the most commonly received combinations of discharge bundle elements (excluding the smoking cessation discharge bundle element which was only applicable to current smokers). We found that 6348 (49.0%) patients received all of the required elements of the discharge bundle whereas 1481 (11.4%) patients did not receive any of the discharge bundle elements. The results of the dominance analysis found that the most important elements of the discharge bundle were community follow-up within two days and checking adherence to drug treatment, which together accounted for 65% of the variation explained by the discharge bundle elements for 30 day readmission for asthma and 77% of the variation explained for 90 day readmission for asthma. The overall variance explained by the model was low (30 day asthma readmission R2=0.046; 90 day asthma readmission R2=0.053), with discharge bundle elements accounting for 21% and 8% of the explained variance for 30 day and 90 day readmission for asthma, respectively. Online supplemental table 4 shows the results of the dominance analysis.Figure 2Most common combinations of elements received as part of the discharge bundle for patients admitted to hospital with acute asthma. Row height is proportional to the number of patients who received that specific combination of discharge bundle elements. The smoking cessation element was not included because this element was only relevant for current smokersThe sensitivity analysis, where patients without a primary diagnosis of asthma in the linked Hospital Episode Statistics record were excluded, showed similar results to the main analysis, with coefficients for best practice tariff elements tending to show a slightly increased strength of association with readmission (online supplemental table 5). Online supplemental tables 5-8 show the results of the sensitivity analyses. Of the 2136 patients included in the audit data without a primary diagnosis of asthma, 1343 (62.9%) had an asthma code present in another diagnosis position. The most common primary diagnosis ICD-10 codes in those without a primary diagnosis of asthma were J18.1 (lobar pneumonia, unspecified), J10.1 (influenza with other respiratory manifestations, seasonal influenza virus identified), and J22 (unspecified acute lower respiratory infection). We found that patients without a primary diagnosis of asthma were significantly more likely to receive a discharge bundle than those who had a primary diagnosis of asthma (P<0.001). Online supplemental table 9 shows the comparison between patients with and without a primary diagnosis of asthma.Few patients in the dataset self-discharged, and exclusion of these patients from the analysis did not substantially affect the results (online supplemental tables 10-14). Including receiving steroids in hospital and at discharge as covariates in the regression model did not substantially affect the results (online supplemental tables 15 and 16). The complete case sensitivity analysis, where patients who had missing data for index of multiple deprivation, history of oral steroids, and smoking history were excluded, showed similar results to the main analysis, with coefficients for best practice tariff elements tending to show a slightly increased strength of association with readmission (online supplemental table 17). Online supplemental tables 18-26 show results grouped by sex.Discussion Principal findings We found that although receiving best practice tariff standard of care was not associated with a statistically significant reduction in 30 day or 90 day readmission for asthma, the discharge bundle element of the best practice tariff was associated with a reduction in readmission. This finding is in contrast with a previous study assessing the role of the best practice tariff in COPD care (which similarly consisted of timely specialist review and provision of a discharge bundle), which did not find an effect of the best practice tariff or its constituent elements on 30 day mortality or readmission. 35Context and implications The British Thoracic Society discharge care bundle was developed in response to the 2014 National Review of Asthma Deaths report. 36 37 Evidence indicates that the inclusion of individual elements in the bundle (eg, having a personalised asthma action plan) is associated with a reduced number of attendances at emergency departments.38 No studies so far, however, have examined the association between the successful delivery of an asthma discharge care bundle and patient outcomes in adult asthma. The importance of measuring patient level data when investigating the effect of care bundles is evident from the COPD literature. When COPD discharge care bundles were first introduced, hospitals reporting their use did not see reduced readmissions,39 despite a meta-analysis of randomised controlled trials suggesting otherwise.40 A likely explanation was that the rate of provision of the care bundle in these hospitals was extremely low, and interventions to improve this situation have been shown to be effective.41Our study provides evidence that is consistent with the effectiveness of the discharge bundle in reducing readmission for adults with asthma, and underlines the importance of its delivery in secondary care. Community follow-up within two working days was found to have the strongest association with reduced readmission, although this element has been removed from the most recently published guidance on discharge bundles because of its unfeasibility.16 Our results indicate that timely follow-up is important and likely to complement the care delivered in hospital during a stressful time for the patient, but although the breakdown of the individual elements of the discharge bundle is interesting, receiving a discharge bundle showed the strongest relation with reduced readmission, regardless of the individual elements within it that were recorded as received. This finding shows the holistic benefits of the discharge bundle as a whole, and might also indicate that the individual elements are not always recorded consistently. Moreover, data were not linked with primary care, so we could not assess which patients received the review that they were recommended or referred to in hospital. Our study found that receiving a discharge bundle was associated with a greater reduction in 30 day than 90 day readmission. A reduced effect over time is a common occurrence in multiple behavioural interventions42 43 and highlights the importance of primary care in ensuring that people with asthma are not readmitted to hospital.Strengths and limitations of this study One key strength of our study was that we analysed delivery of the asthma care bundles to individual patients in a nationwide cohort and directly assessed their effect based on linked data, giving us a large sample size with longitudinal follow-up that negates the possibility of reverse causality or recall bias. Despite the specifications of the best practice tariff, the audit only collects data on arrival time, not admission time. This approach means that a more strict criterion of 24 hours from arrival to review was applied to classify a patient as meeting the best practice tariff, rather than 24 hours from admission, arguably a more beneficial metric at the patient level to ensure timely care. To account for this difference, we performed a sensitivity analysis that allowed for review within 28 and 32 hours of arrival. We found a significant association between meeting the best practice tariff and reduced 30 day readmission for asthma when receiving a review was extended to within 32 hours of arrival. We do not believe that this limitation affects our conclusions that the discharge bundle rather than timeliness of respiratory review was the core element of the best practice tariff responsible for preventing readmission. Rather, we consider that the extended time limit reflects the changing focus from earlier specialist review, focused on acute initial assessment and stabilisation, to later review, focused on optimisation and future risk reduction of asthma attacks later in the admission. The extended specification might also have allowed more patients to reach best practice tariff standard of care, increasing the power to find a true effect of the delivery of a respiratory review at any time and the discharge bundle. We found that those who received a specialist review were four times more likely to receive a discharge bundle, and receiving a specialist review at any point during admission was associated with a reduction in 30 day readmission for asthma, indicating the importance of review as a whole. Our results raise a question about the importance of receiving a respiratory specialist review within 24 hours over receiving a specialist review at any point, but with the caveat that timely care is likely focused on acute outcomes, such as length of stay, that have not been measured in this study.We did not censor patients who died after discharge when performing the regression analyses for readmission, but the number of patients not captured in the readmission data was small (<0.5% of all patients) so we do not believe that this methodology substantially affected our results. As with all observational studies, confounding could explain part or all of the association between our variables of interest and outcome. We believe that the included patient characteristic variables in our regression model should suitably adjust for confounding between our variables and outcome, but residual confounding is possible, and the variance explained by the regression models was low. If an unmeasured confounder is significantly associated with elements of the best practice tariff and also associated with readmission, then the results of our analysis might indicate a spurious association. In what direction unmeasured confounding might occur is unclear. Those who received the best practice tariff tended to have asthma that was clinically more severe on admission and less well controlled overall. But those who received the best practice tariff tended to be in the less deprived group and have fewer comorbid conditions, and therefore unmeasured risk factors that are generally associated with poor health could confound the observed relation.We used a missing category for several variables, which can introduce bias. The similarity between the no and not recorded coefficient estimates for the variable, prescribed three courses of oral steroids in the past 12 months, makes it likely that not recorded is synonymous with no for this variable and that the inclusion of a missing category is not introducing bias in this case. Similarly, the not recorded category coefficients for smoking status did not differ significantly from 1 (reference=never smoked), which might indicate that not recorded also means no for this variable. Finally, only a few patients (1.2%) had missing data for a category for index of multiple deprivation, so including a missing category is unlikely to have substantially affected the results, regardless of the missing data mechanism for these patients. The complete case analysis, where patients with missing data were excluded, showed similar results to the main analysis.Generalisability Receiving elements of the best practice tariff standard of care was associated with improved patient outcomes. At the hospital level, however, we cannot say whether implementation of the best practice tariff payment system has improved hospital care by motivating hospitals to try harder to meet the best practice tariff. Despite widespread use of pay-for-performance schemes in secondary care globally, 44 little evidence exists to suggest that these schemes are effective.6 12 45 One such programme in the US is the Hospital Value Based Purchasing programme, where top performing hospitals in a range of process, outcome, patient satisfaction, and cost effectiveness measures receive bonus payments whereas poor performers are penalised with payments withheld. An observational difference-in-difference study found that changes in mortality rates of providers enrolled in the Hospital Value Based Purchasing programme were similar to those who were not enrolled, including subgroups such as poor performers at baseline.46In France, Incitation financière à l'amélioration de la qualité (Financial Incentive to Quality Improvement) provides bonus payments to hospitals that attain a good score generated by achievement and improvement of 10 domains focused on quality of care and recording. A semi-randomised preliminary study undertaken before the rollout of Incitation financière à l'amélioration de la qualité did not find a change in the scores of hospitals that received payments compared with those that did not.47 In Taiwan, a nationwide matched cohort study assessing a pay-for-performance scheme for total knee arthroplasty care found mixed results; although patients enrolled in the scheme had a lower risk of revision surgery, no significant difference in postoperative infection was found in those who were and were not enrolled.48Although the effectiveness of a particular pay-for-performance initiative should ideally be first assessed in cluster randomised controlled trials,49 this is rarely done and instead observational studies, with their known limitations, are conducted after implementation. With the best practice tariff currently covering 20 diseases and procedures in secondary care in England and the broad use of pay-for-performance globally, our study highlights the importance of assessing the effectiveness of the interventions chosen for inclusion within the best practice tariff and the effectiveness of the best practice tariff as a whole.Conclusions In this study in adults admitted to hospital with acute asthma, although we found no significant association between a record of receiving best practice tariff care and 30 day or 90 day readmission to hospital for asthma, key elements of the best practice tariff relating to discharge seemed to be associated with reduced readmission rates. This study highlights the importance of the discharge bundle in hospital in preventing readmission, with respiratory review as the main vehicle for delivery of the discharge bundle elements. Our results highlight the importance of randomised controlled trials in assessing the effectiveness of pay-for-performance schemes before their implementation nationwide.