Financial incentives and prescribing behavior in primary care.

Many healthcare systems prohibit primary care physicians from dispensing the drugs they prescribe due to concerns that this encourages excessive, ineffective or unnecessarily costly prescribing. Using data from the English National Health Service for 2011-2018, we estimate the impact of physician dispensing rights on prescribing behavior at the extensive margin (comparing practices that dispense and those that do not) and the intensive margin (comparing practices with different proportions of patients to whom they dispense). We control for practices selecting into dispensing based on observable (OLS, entropy balancing) and unobservable practice characteristics (2SLS). We find that physician dispensing increases drug costs per patient by 3.1%, due to more, and more expensive, drugs being prescribed. Reimbursement is partly based on a fixed fee per package dispensed and we find that dispensing practices prescribe smaller packages. As the proportion of the practice population for whom they can dispense increases, dispensing practices behave more like non-dispensing practices.

Patients' experience on pain outcomes after hip arthroplasty: insights from an information tool based on registry data.

BACKGROUND: Arthroplasty registries are rarely used to inform encounters between clinician and patient. This study is part of a larger one which aimed to develop an information tool allowing both to benefit from previous patients' experience after total hip arthroplasty (THA). This study focuses on generating the information tool specifically for pain outcomes. METHODS: Data from the Geneva Arthroplasty Registry (GAR) about patients receiving a primary elective THA between 1996 and 2019 was used. Selected outcomes were identified from patient and surgeon surveys: pain walking, climbing stairs, night pain, pain interference, and pain medication. Clusters of patients with homogeneous outcomes at 1, 5, and 10 years postoperatively were generated based on selected predictors evaluated preoperatively using conditional inference trees (CITs). RESULTS: Data from 6,836 THAs were analysed and 14 CITs generated with 17 predictors found significant (p < 0.05). Baseline WOMAC pain score, SF-12 self-rated health (SRH), number of comorbidities, SF-12 mental component score, and body mass index (BMI) were the most common predictors. Outcome levels varied markedly by clusters whilst predictors changed at different time points for the same outcome. For example, 79% of patients with good to excellent SRH and less than moderate preoperative night pain reported absence of night pain at 1 year after THA; in contrast, for those with fair/poor SHR this figure was 50%. Also, clusters of patients with homogeneous levels of night pain at 1 year were generated based on SRH, Charnley, WOMAC night and pain scores, whilst those at 10 years were based on BMI alone. CONCLUSIONS: The information tool generated under this study can provide prospective patients and clinicians with valuable and understandable information about the experiences of "patients like them" regarding their pain outcomes.

Quality-Adjusted Life Expectancy Norms for the English Population.

OBJECTIVES: The National Institute for Health and Care Excellence in England has implemented severity-of-disease modifiers that give greater weight to health benefits accruing to patients who experience a larger shortfall in quality-adjusted life-years (QALYs) under current standard of care than healthy individuals. This requires an estimate of quality-adjusted life expectancy (QALE) of the general population based on age and sex. Previous QALE population norms are based on nearly 30-year-old assessments of health-related quality of life in the general population. This study provides updated QALE estimates for the English population based on age and sex. METHODS: 5-level version of EQ-5D data for 14 412 participants from the Health Survey for England (waves 2017 and 2018) were pooled, and health-related quality of life population norms were calculated. These norms were combined with official life tables from the Office for National Statistics for 2017 to 2019 using the Sullivan method to derive QALE estimates based on age and sex. Values were discounted using 0%, 1.5%, and 3.5% discount rates. RESULTS: QALE at birth is 68.24 QALYs for men and 68.21 QALYs for women. These values are significantly lower than previously published QALE population norms based on the older 3-level version of EQ-5D data. CONCLUSION: This study provides new QALE population norms for England that serve to establish absolute and relative QALY shortfalls for the purpose of health technology assessments.

Impact of prevention in primary care on costs in primary and secondary care for people with serious mental illness.

A largely unexplored part of the financial incentive for physicians to participate in preventive care is the degree to which they are the residual claimant from any resulting cost savings. We examine the impact of two preventive activities for people with serious mental illness (care plans and annual reviews of physical health) by English primary care practices on costs in these practices and in secondary care. Using panel two-part models to analyze patient-level data linked across primary and secondary care, we find that these preventive activities in the previous year are associated with cost reductions in the current quarter both in primary and secondary care. We estimate that there are large beneficial externalities for which the primary care physician is not the residual claimant: the cost savings in secondary care are 4.7 times larger than the cost savings in primary care. These activities are incentivized in the English National Health Service but the total financial incentives for primary care physicians to participate were considerably smaller than the total cost savings produced. This suggests that changes to the design of incentives to increase the marginal reward for conducting these preventive activities among patients with serious mental illness could have further increased welfare.

Inequality in quality-adjusted life expectancy by educational attainment in Norway: an observational study.

BACKGROUND: Health inequalities are often assessed in terms of life expectancy or health-related quality of life (HRQoL). Few studies combine both aspects into quality-adjusted life expectancy (QALE) to derive comprehensive estimates of lifetime health inequality. Furthermore, little is known about the sensitivity of estimated inequalities in QALE to different sources of HRQoL information. This study assesses inequalities in QALE by educational attainment in Norway using two different measures of HRQoL. METHODS: We combine full population life tables from Statistics Norway with survey data from the Tromsø study, a representative sample of the Norwegian population aged ≥ 40. HRQoL is measured using the EQ-5D-5L and EQ-VAS instruments. Life expectancy and QALE at 40 years of age are calculated using the Sullivan-Chiang method and are stratified by educational attainment. Inequality is measured as the absolute and relative gap between individuals with lowest (i.e. primary school) and highest (university degree 4 + years) educational attainment. RESULTS: People with the highest educational attainment can expect to live longer lives (men: + 17.9% (95%CI: 16.4 to 19.5%), women: + 13.0% (95%CI: 10.6 to 15.5%)) and have higher QALE (men: + 22.4% (95%CI: 20.4 to 24.4%), women: + 18.3% (95%CI: 15.2 to 21.6%); measured using EQ-5D-5L) than individuals with primary school education. Relative inequality is larger when HRQoL is measured using EQ-VAS. CONCLUSION: Health inequalities by educational attainment become wider when measured in QALE rather than LE, and the degree of this widening is larger when measuring HRQoL by EQ-VAS than by EQ-5D-5L. We find a sizable educational gradient in lifetime health in Norway, one of the most developed and egalitarian societies in the world. Our estimates provide a benchmark against which other countries can be compared.

Socioeconomic inequalities in HRQoL in England: an age-sex stratified analysis.

BACKGROUND: Socioeconomic status is a key predictor of lifetime health: poorer people can expect to live shorter lives with lower average health-related quality-of-life (HRQoL) than richer people. In this study, we aimed to improve understanding of the socioeconomic gradient in HRQoL by exploring how inequalities in different dimensions of HRQoL differ by age. METHODS: Data were derived from the Health Survey for England for 2017 and 2018 (14,412 participants). HRQoL was measured using the EQ-5D-5L instrument. We estimated mean EQ-5D utility scores and reported problems on five HRQoL dimensions (mobility, self-care, usual activities, pain/discomfort, anxiety/depression) for ages 16 to 90+ and stratified by neighbourhood deprivation quintiles. Relative and absolute measures of inequality were assessed. RESULTS: Mean EQ-5D utility scores declined with age and followed a socioeconomic gradient, with the lowest scores in the most deprived areas. Gaps between the most and least deprived quintiles emerged around the age of 35, reached their greatest extent at age 60 to 64 (relative HRQoL of most deprived compared to least deprived quintile: females = 0.77 (95% CI: 0.68-0.85); males = 0.78 (95% CI: 0.69-0.87)) before closing again in older age groups. Gaps were apparent for all five EQ-5D dimensions but were greatest for mobility and self-care. CONCLUSION: There are stark socioeconomic inequalities in all dimensions of HRQoL in England. These inequalities start to develop from early adulthood and increase with age but reduce again around retirement age.

Inequality of opportunity in a land of equal opportunities: The impact of parents' health and wealth on their offspring's quality of life in Norway.

BACKGROUND: The literature on Inequality of opportunity (IOp) in health distinguishes between circumstances that lie outside of own control vs. efforts that - to varying extents - are within one's control. From the perspective of IOp, this paper aims to explain variations in individuals' health-related quality of life (HRQoL) by focusing on two separate sets of variables that clearly lie outside of own control: Parents' health is measured by their experience of somatic diseases, psychological problems and any substance abuse, while parents' wealth is indicated by childhood financial conditions (CFC). We further include own educational attainment which may represent a circumstance, or an effort, and examine associations of IOp for different health outcomes. HRQoL are measured by EQ-5D-5L utility scores, as well as the probability of reporting limitations on specific HRQoL-dimensions (mobility, self-care, usual-activities, pain & discomfort, and anxiety and depression). METHOD: We use unique survey data (N = 20,150) from the egalitarian country of Norway to investigate if differences in circumstances produce unfair inequalities in health. We estimate cross-sectional regression models which include age and sex as covariates. We estimate two model specifications. The first represents a narrow IOp by estimating the contributions of parents' health and wealth on HRQoL, while the second includes own education and thus represents a broader IOp, alternatively it provides a comparison of the relative contributions of an effort variable and the two sets of circumstance variables. RESULTS: We find strong associations between the circumstance variables and HRQoL. A more detailed examination showed particularly strong associations between parental psychological problems and respondents' anxiety and depression. Our Shapley decomposition analysis suggests that parents' health and wealth are each as important as own educational attainment for explaining inequalities in adult HRQoL. CONCLUSION: We provide evidence for the presence of the lasting effect of early life circumstances on adult health that persists even in one of the most egalitarian countries in the world. This suggests that there may be an upper limit to how much a generous welfare state can contribute to equal opportunities.

Does regulation increase the rate at which doctors leave practice? Analysis of routine hospital data in the English NHS following the introduction of medical revalidation.

BACKGROUND: In 2012, the UK introduced medical revalidation, whereby to retain their licence all doctors are required to show periodically that they are up to date and fit to practise medicine. Early reports suggested that some doctors found the process overly onerous and chose to leave practice. This study investigates the effect of medical revalidation on the rate at which consultants (senior hospital doctors) leave NHS practice, and assesses any differences between the performance of consultants who left or remained in practice before and after the introduction of revalidation. METHODS: We used a retrospective cohort of administrative data from the Hospital Episode Statistics database on all consultants who were working in English NHS hospitals between April 2008 and March 2009 (n = 19,334), followed to March 2015. Proportional hazard models were used to identify the effect of medical revalidation on the time to exit from the NHS workforce, as implied by ceasing NHS clinical activity. The main exposure variable was consultants' time-varying revalidation status, which differentiates between periods when consultants were (a) not subject to revalidation-before the policy was introduced, (b) awaiting a revalidation recommendation and (c) had received a positive recommendation to be revalidated. Difference-in-differences analysis was used to compare the performance of those who left practice with those who remained in practice before and after the introduction of revalidation, as proxied by case-mix-adjusted 30-day mortality rates. RESULTS: After 2012, consultants who had not yet revalidated were at an increased hazard of ceasing NHS clinical practice (HR 2.33, 95% CI 2.12 to 2.57) compared with pre-policy levels. This higher risk remained after a positive recommendation (HR 1.85, 95% CI 1.65 to 2.06) but was statistically significantly reduced (p < 0.001). We found no statistically significant differences in mortality rates between those consultants who ceased practice and those who remained, after adjustment for multiple testing. CONCLUSION: Revalidation appears to have led to greater numbers of doctors ceasing clinical practice, over and above other contemporaneous influences. Those ceasing clinical practice do not appear to have provided lower quality care, as approximated by mortality rates, when compared with those remaining in practice.

Aggregate Distributional Cost-Effectiveness Analysis of Health Technologies.

BACKGROUND: Health inequalities can be partially addressed through the range of treatments funded by health systems. Nevertheless, although health technology assessment agencies assess the overall balance of health benefits and costs, no quantitative assessment of health inequality impact is consistently undertaken. OBJECTIVES: To assess the inequality impact of technologies recommended under the NICE single technology appraisal process from 2012 to 2014 using an aggregate distributional cost-effectiveness framework. METHODS: Data on health benefits, costs, and patient populations were extracted from the NICE website. Benefits for each technology were distributed to social groups using the observed socioeconomic distribution of hospital utilization for the targeted disease. Inequality measures and estimates of cost-effectiveness were compared using the health inequality impact plane and combined using social welfare indices. RESULTS: Twenty-seven interventions were evaluated. Fourteen interventions were estimated to increase population health and reduce health inequality, 8 to reduce population health and increase health inequality, and 5 to increase health and increase health inequality. Among the latter 5, social welfare analysis, using inequality aversion parameters reflecting high concern for inequality, indicated that the health gain outweighs the negative health inequality impact. CONCLUSIONS: The methods proposed offer a way of estimating the health inequality impacts of new health technologies. The methods do not allow for differences in technology-specific utilization and health benefits, but require less resources and data than conducting full distributional cost-effectiveness analysis. They can provide useful quantitative information to help policy makers consider how far new technologies are likely to reduce or increase health inequalities.

Prospective payment systems and discretionary coding-Evidence from English mental health providers.

Reimbursement of English mental health hospitals is moving away from block contracts and towards activity and outcome-based payments. Under the new model, patients are categorised into 20 groups with similar levels of need, called clusters, to which prices may be assigned prospectively. Clinicians, who make clustering decisions, have substantial discretion and can, in principle, directly influence the level of reimbursement the hospital receives. This may create incentives for upcoding. Clinicians are supported in their allocation decision by a clinical clustering algorithm, the Mental Health Clustering Tool, which provides an external reference against which clustering behaviour can be benchmarked. The aims of this study are to investigate the degree of mismatch between predicted and actual clustering and to test whether there are systematic differences amongst providers in their clustering behaviour. We use administrative data for all mental health patients in England who were clustered for the first time during the financial year 2014/15 and estimate multinomial multilevel models of over, under, or matching clustering. Results suggest that hospitals vary systematically in their probability of mismatch but this variation is not consistently associated with observed hospital characteristics.

Multidimensional performance assessment of public sector organisations using dominance criteria.

Public sector organisations pursue multiple objectives and serve a number of stakeholders. But stakeholders are rarely explicit about the valuations they attach to different objectives, nor are these valuations likely to be identical. This complicates the assessment of their performance because no single set of weights can be chosen legitimately to aggregate outputs into unidimensional composite scores. We propose the use of dominance criteria in a multidimensional performance assessment framework to identify best practice and poor performance under relatively weak assumptions about stakeholders' preferences. We use as an example providers of hip replacement surgery in the English National Health Service and estimate multivariate multilevel models to study their performance in terms of length of stay, readmission rates, post-operative patient-reported health status and waiting time. We find substantial correlation between objectives and demonstrate that ignoring the correlation can lead to incorrect assessments of performance.

Use of large-scale HRQoL datasets to generate individualised predictions and inform patients about the likely benefit of surgery.

PURPOSE: The English NHS has mandated the routine collection of health-related quality of life (HRQoL) data before and after surgery, giving prospective patient information about the likely benefit of surgery. Yet, the information is difficult to access and interpret because it is not presented in a lay-friendly format and does not reflect patients' individual circumstances. We set out a methodology to generate personalised information to help patients make informed decisions. METHODS: We used anonymised, pre- and postoperative EuroQol-5D-3L (EQ-5D) data for over 490,000 English NHS patients who underwent primary hip or knee replacement surgery or groin hernia repair between April 2009 and March 2016. We estimated linear regression models to relate changes in EQ-5D utility scores to patients' own assessment of the success of surgery, and calculated from that minimally important differences for health improvements/deteriorations. Classification tree analysis was used to develop algorithms that sort patients into homogeneous groups that best predict postoperative EQ-5D utility scores. RESULTS: Patients were classified into between 55 (hip replacement) to 60 (hernia repair) homogeneous groups. The classifications explained between 14 and 27% of variation in postoperative EQ-5D utility score. CONCLUSIONS: Patients are heterogeneous in their expected benefit from surgery, and decision aids should reflect this. Large administrative datasets on HRQoL can be used to generate the required individualised predictions to inform patients.

Addressing Missing Data in Patient-Reported Outcome Measures (PROMS): Implications for the Use of PROMS for Comparing Provider Performance.

Patient-reported outcome measures (PROMs) are now routinely collected in the English National Health Service and used to compare and reward hospital performance within a high-powered pay-for-performance scheme. However, PROMs are prone to missing data. For example, hospitals often fail to administer the pre-operative questionnaire at hospital admission, or patients may refuse to participate or fail to return their post-operative questionnaire. A key concern with missing PROMs is that the individuals with complete information tend to be an unrepresentative sample of patients within each provider and inferences based on the complete cases will be misleading. This study proposes a strategy for addressing missing data in the English PROM survey using multiple imputation techniques and investigates its impact on assessing provider performance. We find that inferences about relative provider performance are sensitive to the assumptions made about the reasons for the missing data.

Location, quality and choice of hospital: Evidence from England 2002-2013.

We investigate (a) how patient choice of hospital for elective hip replacement is influenced by distance, quality and waiting times, (b) differences in choices between patients in urban and rural locations, (c) the relationship between hospitals' elasticities of demand to quality and the number of local rivals, and how these changed after relaxation of constraints on hospital choice in England in 2006. Using a data set on over 500,000 elective hip replacement patients over the period 2002 to 2013 we find that patients became more likely to travel to a provider with higher quality or lower waiting times, the proportion of patients bypassing their nearest provider increased from 25% to almost 50%, and hospital elasticity of demand with respect to own quality increased. By 2013 average hospital demand elasticity with respect to readmission rates and waiting times were - 0.2 and - 0.04. Providers facing more rivals had demand that was more elastic with respect to quality and waiting times. Patients from rural areas have smaller disutility from distance.

Comparison of pre- and post-operative health-related quality of life and length of stay after primary total hip replacement in matched English and German patient cohorts.

PURPOSE: We compare pre- and post-operative health-related quality of life (HRQoL) and length of stay after total hip replacement (THR) in matched German and English patient cohorts to test for differences in admission thresholds, clinical effectiveness and resource utilisation between the healthcare systems. METHODS: German data (n = 271) were collected in a large orthopaedic hospital in Munich, Germany; English data (n = 26,254) were collected as part of the national patient-reported outcome measures programme. HRQoL was measured using the EuroQoL-5D instrument. Propensity score matching was used to construct two patient cohorts that are comparable in terms of preoperative patient characteristics. RESULTS: Before matching, patients in England showed lower preoperative EQ-5D scores (0.35 vs 0.52, p < 0.001) and experienced a larger improvement in HRQoL (0.43 vs 0.33, p < 0.001) than German patients. Patients in the German cohort were more likely to report no or only moderate problems with mobility and pain preoperatively than their English counterparts. After matching, improvements in HRQoL were comparable (0.32 vs 0.33, p = 0.638); post-operative scores were slightly higher in the German cohort (0.82 vs 0.85, p = 0.585). Length of stay was substantially lower in England than in Germany (4.5 vs 9.0 days, p < 0.001). CONCLUSIONS: Our results highlight differences in preoperative health status between countries, which may arise due to different admission thresholds and access to surgery. In terms of quality of life, THR surgery is equally effective in both countries when performed on similar patients, but hospital stay is shorter in England.

Comparing the performance of the Charlson/Deyo and Elixhauser comorbidity measures across five European countries and three conditions.

BACKGROUND: The Charlson and Elixhauser comorbidity measures are commonly used methods to account for patient comorbidities in hospital-level comparisons of clinical quality using administrative data. Both have been validated in North America, but there is less evidence of their performance in Europe and in pooled cross-country data, which are features of the European Collaboration for Healthcare Optimization (ECHO) project. This study compares the performance of the Charlson/Deyo and Elixhauser comorbidity measures in predicting in-hospital mortality using data from five European countries in three inpatient groups. METHODS: Administrative data is used from five countries in 2008-2009 for three indicators commonly used in hospital quality comparisons: mortality rates following acute myocardial infarction, coronary artery bypass graft surgery and stroke. Logistic regression models are constructed to predict mortality controlling for age, gender and the relevant comorbidity measure. Model discrimination is evaluated using c-statistics. Model calibration is evaluated using calibration slopes. Overall goodness-of-fit is evaluated using Nagelkerke's R(2) and the Akaike information criterion. All models are validated internally by using bootstrapping and externally by using the 2009 model parameters to predict mortality in 2008. RESULTS: The Elixhauser measure has better overall predictive ability in terms of discrimination and goodness-of-fit than the Charlson/Deyo measure or the age-sex only model. There is no clear difference in model calibration. These findings are robust to the choice of country, to pooling all five countries and to internal and external validation. CONCLUSIONS: The Elixhauser list contains more comorbidities, which may enable it to achieve better discrimination than the Charlson measure. Both measures achieve similar calibration, so for the purpose of ECHO we judged the Elixhauser measure to be preferable.

Comparing hospital performance within and across countries: an illustrative study of coronary artery bypass graft surgery in England and Spain.

OBJECTIVE: To assess the feasibility, strengths and weaknesses of using administrative data to compare hospital performance across countries, using mortality after coronary artery bypass graft (CABG) surgery as an illustrative example. METHODS: Country specific and pooled models using individual-level data and logistic regression methods assess individual hospital performance using funnel plots accounting for multiple testing. Outcomes are adjusted for age, sex, comorbidities and indicators of patient severity. Data includes patients from all publicly funded hospitals delivering CABG surgery in England and Spain. Inpatient hospital-level standardized mortality rates within 30 days of CABG surgery are calculated for 83 999 CABG patients between 2007 and 2009. RESULTS: Unadjusted national mortality rates are 5% in Spain and 2.3% in England. Country-specific models identified similar patterns of excess mortality 'alerts' and 'alarms' in hospitals in Spain or England. Pooling data from both countries identifies larger numbers of alerts and alarms in Spanish hospitals, and risk-adjustment increased the already large national mortality difference. This was reduced but not eliminated by accounting for lower volume in Spanish hospitals. CONCLUSION: Cross-national comparisons potentially add value by providing international performance benchmarks. Hospital-level analysis across countries can illuminate differences in hospital performance, which might not be identified using country-specific data or incomplete registry data, and can test hypotheses that may explain national differences. Difficulties of making data comparable between countries, however, compound the usual within-country measurement problems.

Socioeconomic inequality in hip replacement in four European countries from 2002 to 2009--area-level analysis of hospital data.

BACKGROUND: Cross-country comparisons of socioeconomic equity in health care typically use sample survey data on general services such as physician visits. This study uses comprehensive administrative data on a specific service: hip replacement. METHODS: We analyse 651 652 publicly funded hip replacements, excluding fractures and accidents, in adults over 35 in Denmark, England, Portugal and Spain from 2002 to 2009. Sub-national administrative areas are split into socioeconomic quintile groups comprising approximately one-fifth of the national population. Area-level Poisson regression with Huber-White standard errors is used to calculate age-sex standardised hip replacement rates by quintile group, together with gaps and ratios between richest and poorest groups (Q5 and Q1) and the middle group (Q3). RESULTS: We find pro-rich-area inequality in England (2009 Q5/Q1 ratio 1.35 [CI 1.25-1.45]) and Spain (2009 Q5/Q1 ratio 1.43 [CI 1.17-1.70]), pro-poor-area inequality in Portugal (2009 Q5/Q1 ratio 0.67 [CI 0.50-0.83]) and no significant inequality in Denmark. Pro-rich-area inequality increased over time in England and Spain but not significantly. Within-country differences between socioeconomic quintile groups are smaller than between-country differences in general population averages: hip replacement rates are substantially lower in Portugal and Spain (8.6 and 7.4 per 10 000 in 2009) than England and Denmark (20.2 and 27.8 per 10 000 in 2009). CONCLUSION: Despite limitations regarding individual-level inequality and area heterogeneity, analysis of area-level data on publicly funded hospital activity can provide useful cross-country comparisons and longitudinal monitoring of socioeconomic inequality in specific health services. Although this kind of analysis cannot provide definitive answers, it can raise important questions for decision makers.

Determinants of hospital length of stay for people with serious mental illness in England and implications for payment systems: a regression analysis.

BACKGROUND: Serious mental illness (SMI), which encompasses a set of chronic conditions such as schizophrenia, bipolar disorder and other psychoses, accounts for 3.4 m (7 %) total bed days in the English NHS. The introduction of prospective payment to reimburse hospitals makes an understanding of the key drivers of length of stay (LOS) imperative. Existing evidence, based on mainly small scale and cross-sectional studies, is mixed. Our study is the first to use large-scale national routine data to track English hospitals' LOS for patients with a main diagnosis of SMI over time to examine the patient and local area factors influencing LOS and quantify the provider level effects to draw out the implications for payment systems. METHODS: We analysed variation in LOS for all SMI admissions to English hospitals from 2006 to 2010 using Hospital Episodes Statistics (HES). We considered patients with a LOS of up to 180 days and estimated Poisson regression models with hospital fixed effects, separately for admissions with one of three main diagnoses: schizophrenia; psychotic and schizoaffective disorder; and bipolar affective disorder. We analysed the independent contribution of potential determinants of LOS including clinical and socioeconomic characteristics of the patient, access to and quality of primary care, and local area characteristics. We examined the degree of unexplained variation in provider LOS. RESULTS: Most risk factors did not have a differential effect on LOS for different diagnostic sub-groups, however we did find some heterogeneity in the effects. Shorter LOS in the pooled model was associated with co-morbid substance or alcohol misuse (4 days), and personality disorder (8 days). Longer LOS was associated with older age (up to 19 days), black ethnicity (4 days), and formal detention (16 days). Gender was not a significant predictor. Patients who self-discharged had shorter LOS (20 days). No association was found between higher primary care quality and LOS. We found large differences between providers in unexplained variation in LOS. CONCLUSIONS: By identifying key determinants of LOS our results contribute to a better understanding of the implications of case-mix to ensure prospective payment systems reflect accurately the resource use within sub-groups of patients with SMI.

Methods to Quantify the Importance of Parameters for Model Updating and Distributional Adaptation.

PURPOSE: Decision models are time-consuming to develop; therefore, adapting previously developed models for new purposes may be advantageous. We provide methods to prioritize efforts to 1) update parameter values in existing models and 2) adapt existing models for distributional cost-effectiveness analysis (DCEA). METHODS: Methods exist to assess the influence of different input parameters on the results of a decision models, including value of information (VOI) and 1-way sensitivity analysis (OWSA). We apply 1) VOI to prioritize searches for additional information to update parameter values and 2) OWSA to prioritize searches for parameters that may vary by socioeconomic characteristics. We highlight the assumptions required and propose metrics that quantify the extent to which parameters in a model have been updated or adapted. We provide R code to quickly carry out the analysis given inputs from a probabilistic sensitivity analysis (PSA) and demonstrate our methods using an oncology case study. RESULTS: In our case study, updating 2 of 21 probabilistic model parameters addressed 71.5% of the total VOI and updating 3 addressed approximately 100% of the uncertainty. Our proposed approach suggests that these are the 3 parameters that should be prioritized. For model adaptation for DCEA, 46.3% of the total OWSA variation came from a single parameter, while the top 10 input parameters were found to account for more than 95% of the total variation, suggesting efforts should be aimed toward these. CONCLUSIONS: These methods offer a systematic approach to guide research efforts in updating models with new data or adapting models to undertake DCEA. The case study demonstrated only very small gains from updating more than 3 parameters or adapting more than 10 parameters. HIGHLIGHTS: It can require considerable analyst time to search for evidence to update a model or to adapt a model to take account of equity concerns.In this article, we provide a quantitative method to prioritze parameters to 1) update existing models to reflect potential new evidence and 2) adapt existing models to estimate distributional outcomes.We define metrics that quantify the extent to which the parameters in a model have been updated or adapted.We provide R code that can quickly rank parameter importance and calculate quality metrics using only the results of a standard probabilistic sensitivity analysis.