Bayesian inference for joint modelling of longitudinal continuous, binary and ordinal events

In medical studies, repeated measurements of continuous, binary and ordinal outcomes are routinely collected from the same patient. Instead of modelling each outcome separately, in this study we propose to jointly model the trivariate longitudinal responses, so as to take account of the inherent association between the different outcomes and thus improve statistical inferences. This work is motivated by a large cohort study in the North West of England, involving trivariate responses from each patient: Body Mass Index, Depression (Yes/No) ascertained with cut-off score not less than 8 at the Hospital Anxiety and Depression Scale, and Pain Interference generated from the Medical Outcomes Study 36-item short-form health survey with values returned on an ordinal scale 1-5. There are some well-established methods for combined continuous and binary, or even continuous and ordinal responses, but little work was done on the joint analysis of continuous, binary and ordinal responses. We propose conditional joint random-effects models, which take into account the inherent association between the continuous, binary and ordinal outcomes. Bayesian analysis methods are used to make statistical inferences. Simulation studies show that, by jointly modelling the trivariate outcomes, standard deviations of the estimates of parameters in the models are smaller and much more stable, leading to more efficient parameter estimates and reliable statistical inferences. In the real data analysis, the proposed joint analysis yields a much smaller deviance information criterion value than the separate analysis, and shows other good statistical properties too.

Developing Objective Metrics for Unit Staffing (DOMUS) study

OBJECTIVE

Safe midwifery staffing levels on delivery suites is a priority area for any maternity service. Escalation policies are tools that provide an operational response to emergency pressures. The aim of this study was to assess the feasibility of using a scoring system to contemporaneously assess the required staffing level based on demand and use this to determine delivery suite escalation level and utilise the information generated regarding clinical activity (Demand) and staffing levels (Capacity) to generate unit-specific calculation for the actual number of midwifery staff required.

SETTING

A maternity unit of a university-affiliated tertiary referral hospital.

DESIGN

Over a 12-month period, specifically designed scoring sheets were completed by delivery suite shift co-ordinators four times a day (04:00, 10:00, 16:00 and 22:00). Based on the dependency score (Demand) and the number of midwifery staff available (Capacity), an escalation level was determined for each shift. The 80th centile of the demand was used to determine optimal capacity.

RESULTS

A total of 1160 scoring sheets were completed. Average staff number throughout the year on any shift was 7 (range 3-11). Average dependency score was 7 (range 1-14). The 80th centile for demand was calculated to be 11.

CONCLUSIONS

This study stresses the importance and usefulness of a simple tool that can be used to determine the level of escalation on delivery suite based on an objective scoring system and can also be used to determine the appropriate staffing on delivery suite.

Increased risk of vascular disease associated with gout: a retrospective, matched cohort study in the UK clinical practice research datalink

Objectives

To determine whether gout increases risk of incident coronary heart disease (CHD), cerebrovascular (CVD) and peripheral vascular disease (PVD) in a large cohort of primary care patients with gout, since there have been no such large studies in primary care.

Methods

A retrospective cohort study was performed using data from the Clinical Practice Research Datalink (CPRD). Risk of incident CHD, CVD and PVD was compared in 8386 patients with an incident diagnosis of gout, and 39 766 age, sex and registered general practice-matched controls, all aged over 50 years and with no prior vascular history, in the 10 years following incidence of gout, or matched index date (baseline). Multivariable Cox Regression was used to estimate HRs and covariates included sex and baseline measures of age, Body Mass Index, smoking, alcohol consumption, Charlson comorbidity index, history of hypertension, hyperlipidaemia, chronic kidney disease, statin use and aspirin use.

Results

Multivariable analysis showed men were at increased risk of any vascular event (HRs (95% CIs)) HR 1.06 (1.01 to 1.12), any CHD HR 1.08 (1.01 to 1.15) and PVD HR 1.18 (1.01 to 1.38), while women were at increased risk of any vascular event, HR 1.25 (1.15 to 1.35), any CHD HR 1.25 (1.12 to 1.39), and PVD 1.89 (1.50 to 2.38)) but not any CVD.

Conclusions

In this cohort of over 50s with gout, female patients with gout were at greatest risk of incident vascular events, even after adjustment for vascular risk factors, despite a higher prevalence of both gout and vascular disease in men. Further research is required to establish the reason for this sex difference.

Risk of vascular events in patients with polymyalgia rheumatica

BACKGROUND

Polymyalgia rheumatica is one of the most common inflammatory rheumatologic conditions in older adults. Other inflammatory rheumatologic disorders are associated with an excess risk of vascular disease. We investigated whether polymyalgia rheumatica is associated with an increased risk of vascular events.

METHODS

We used the General Practice Research Database to identify patients with a diagnosis of incident polymyalgia rheumatica between Jan. 1, 1987, and Dec. 31, 1999. Patients were matched by age, sex and practice with up to 5 patients without polymyalgia rheumatica. Patients were followed until their first vascular event (cardiovascular, cerebrovascular, peripheral vascular) or the end of available records (May 2011). All participants were free of vascular disease before the diagnosis of polymyalgia rheumatica (or matched date). We used Cox regression models to compare time to first vascular event in patients with and without polymyalgia rheumatica.

RESULTS

A total of 3249 patients with polymyalgia rheumatica and 12 735 patients without were included in the final sample. Over a median follow-up period of 7.8 (interquartile range 3.3-12.4) years, the rate of vascular events was higher among patients with polymyalgia rheumatica than among those without (36.1 v. 12.2 per 1000 person-years; adjusted hazard ratio 2.6, 95% confidence interval 2.4-2.9). The increased risk of a vascular event was similar for each vascular disease end point. The magnitude of risk was higher in early disease and in patients younger than 60 years at diagnosis.

INTERPRETATION

Patients with polymyalgia rheumatica have an increased risk of vascular events. This risk is greatest in the youngest age groups. As with other forms of inflammatory arthritis, patients with polymyalgia rheumatica should have their vascular risk factors identified and actively managed to reduce this excess risk.

Pain at multiple body sites and health-related quality of life in older adults: results from the North Staffordshire Osteoarthritis Project

Objectives. Number of pain sites (NPS) is a potentially important marker of health-related quality of life (HRQoL) but remains unexplored in older people. This cross-sectional study investigated whether, in older people including the oldest old, NPS was independently associated with poorer mental and physical HRQoL and if the association was moderated by age.

Methods. A postal questionnaire sent to a population sample of adults aged ≥50 years in North Staffordshire, UK, included the 12-item Short Form Health Survey (SF-12) mental component summary (MCS) and physical component summary (PCS), a blank body pain manikin, socio-demographic, health behaviour and morbidity questions. Participants shaded sites of pain lasting ≥1 day in the past 4 weeks on the manikin. OA consultation data were obtained for participants consenting to medical records review.

Results. A total of 13 986 individuals (adjusted response 70.6%) completed a questionnaire, of which 12 408 provided complete pain data. The median NPS reported was 4 [interquartile range (IQR) 0–8]. General linear models showed that an increasing NPS was significantly associated with poorer MCS (β = −0.43, 95% CI −0.46, −0.40) and PCS (β = −0.87, 95% CI −0.90, −0.84). Adjustment for covariates attenuated the associations but they remained significant (MCS: β = −0.28, 95% CI −0.31, −0.24; PCS: β = −0.63, 95% CI −0.66, −0.59). The association between NPS and MCS or PCS was moderated by age, but the strongest associations were not in the oldest old.

Conclusion. NPS appears to be a potentially modifiable target for improving physical and mental HRQoL in older people. Future analyses should investigate the influence of NPS on HRQoL over time in older people.

Health services changes: is a run-in period necessary before evaluation in randomised clinical trials?

BACKGROUND

Most randomised clinical trials (RCTs) testing a new health service do not allow a run-in period of consolidation before evaluating the new approach. Consequently, health professionals involved may feel insufficiently familiar or confident, or that new processes or systems that are integral to the service are insufficiently embedded in routine care prior to definitive evaluation in a RCT. This study aimed to determine the optimal run-in period for a new physiotherapy-led telephone assessment and treatment service known as PhysioDirect and whether a run-in was needed prior to evaluating outcomes in an RCT.

METHODS

The PhysioDirect trial assessed whether PhysioDirect was as effective as usual care. Prior to the main trial, a run-in of up to 12 weeks was permitted to facilitate physiotherapists to become confident in delivering the new service. Outcomes collected from the run-in and main trial were length of telephone calls within the PhysioDirect service and patients' physical function (SF-36v2 questionnaire) and Measure Yourself Medical Outcome Profile v2 collected at baseline and six months. Joinpoint regression determined how long it had taken call times to stabilise. Analysis of covariance determined whether patients' physical function at six months changed from the run-in to the main trial.

RESULTS

Mean PhysioDirect call times (minutes) were higher in the run-in (31 (SD: 12.6)) than in the main trial (25 (SD: 11.6)). Each physiotherapist needed to answer 42 (95% CI: 20,56) calls for their mean call time to stabilise at 25 minutes per call; this took a minimum of seven weeks. For patients' physical function, PhysioDirect was equally clinically effective as usual care during both the run-in (0.17 (95% CI: -0.91,1.24)) and main trial (-0.01 (95% CI: -0.80,0.79)).

CONCLUSIONS

A run-in was not needed in a large trial testing PhysioDirect services in terms of patient outcomes. A learning curve was evident in the process measure of telephone call length. This decreased during the run-in and stabilised prior to commencement of the main trial. Future trials should build in a run-in if it is anticipated that learning would have an effect on patient outcome.

TRIAL REGISTRATION

Current Controlled Trials, ISRCTN55666618.

Increased cardiovascular mortality associated with gout: a systematic review and meta-analysis

BACKGROUND

Hyperuricaemia, the biochemical precursor to gout, has been shown to be an independent risk factor for mortality from cardiovascular disease (CVD), although studies examining the clinical phenomenon of gout and risk of CVD mortality report conflicting results. This study aimed to produce a pooled estimate of risk of mortality from cardiovascular disease in patients with gout.

DESIGN

Systematic review and meta-analysis.

METHODS

Electronic bibliographic databases were searched from inception to November 2012, with results reviewed by two independent reviewers. Studies were included if they reported data on CVD mortality in adults with gout who were free of CVD at time of entry into the study. Pooled hazard ratios (HRs) for this association were calculated both unadjusted and adjusted for traditional vascular risk factors.

RESULTS

Six papers, including 223,448 patients, were eligible for inclusion (all (CVD) mortality n = 4, coronary heart disease (CHD) mortality n = 3, and myocardial infarction mortality n = 3). Gout was associated with an excess risk of CVD mortality (unadjusted HR 1.51 (95% confidence interval, CI, 1.17-1.84)) and CHD mortality (unadjusted HR 1.59, 95% CI 1.25-1.94)). After adjusting for traditional vascular risk factors, the pooled HR for both CVD mortality (HR 1.29, 95% CI 1.14-1.44) and CHD mortality (HR 1.42, 95% CI 1.22-1.63) remained statistically significant, but none of the studies reported a significant association with myocardial infarction.

CONCLUSIONS

Gout increases the risk of mortality from CVD and CHD, but not myocardial infarction, independently of vascular risk factors.

Carbon monoxide expedites metabolic exhaustion to inhibit tumor growth

One classical feature of cancer cells is their metabolic acquisition of a highly glycolytic phenotype. Carbon monoxide (CO), one of the products of the cytoprotective molecule heme oxygenase-1 (HO-1) in cancer cells, has been implicated in carcinogenesis and therapeutic resistance. However, the functional contributions of CO and HO-1 to these processes are poorly defined. In human prostate cancers, we found that HO-1 was nuclear localized in malignant cells, with low enzymatic activity in moderately differentiated tumors correlating with relatively worse clinical outcomes. Exposure to CO sensitized prostate cancer cells but not normal cells to chemotherapy, with growth arrest and apoptosis induced in vivo in part through mitotic catastrophe. CO targeted mitochondria activity in cancer cells as evidenced by higher oxygen consumption, free radical generation, and mitochondrial collapse. Collectively, our findings indicated that CO transiently induces an anti-Warburg effect by rapidly fueling cancer cell bioenergetics, ultimately resulting in metabolic exhaustion.

Changes in coping behavior and the relationship to personality, health threat communication and illness perceptions from the diagnosis of diabetes: a 2-year prospective longitudinal study

Coping behavior is of critical importance in diabetes because of its impact upon self-care and hence eventual medical outcome. We examined how coping behavior and its relationship to personality, diabetes health threat communication (DHTC) and illness representations changes after diagnosis of diabetes. Newly diagnosed diabetic patients were assessed after diagnosis and at 6, 12 and 24 months using the DHTC, Illness Perceptions and Coping inventory questionnaires. Personality traits were assessed at baseline. Active coping, planning, positive reinterpretation and growth (PRG), seeking emotional and instrumental (social) support decreased over the 2 years from diagnosis while passive acceptance increased. Openness/intellect and conscientiousness traits were associated with active coping and seeking instrumental support. Openness/intellect also associated with planning and PRG. These relationships did not vary over time. Perceived threat and serious consequences were associated with active coping but the effect diminished over time. Illness coherence (understanding of diabetes), personal and treatment control were associated with active coping, planning and seeking instrumental support and did not change over time. The coping strategies most commonly employed by diabetic patients are adaptive. Coping behavior changes over the 2 years from diagnosis. Promoting better understanding of diabetes, perceptions of personal control and treatment effectiveness are more likely than perception of health threat to sustain adaptive problem focused coping behavior.

Changes in Coping Behavior and the Relationship to Personality, Health Threat Communication and Illness Perceptions from the Diagnosis of Diabetes: A 2-year Prospective Longitudinal Study

Coping behavior is of critical importance in diabetes because of its impact upon self-care and hence eventual medical outcome. We examined how coping behavior and its relationship to personality, diabetes health threat communication (DHTC) and illness representations changes after diagnosis of diabetes. Newly diagnosed diabetic patients were assessed after diagnosis and at 6, 12 and 24 months using the DHTC, Illness Perceptions and Coping inventory questionnaires. Personality traits were assessed at baseline. Active coping, planning, positive reinterpretation and growth (PRG), seeking emotional and instrumental (social) support decreased over the 2 years from diagnosis while passive acceptance increased. Openness/intellect and conscientiousness traits were associated with active coping and seeking instrumental support. Openness/intellect also associated with planning and PRG. These relationships did not vary over time. Perceived threat and serious consequences were associated with active coping but the effect diminished over time. Illness coherence (understanding of diabetes), personal and treatment control were associated with active coping, planning and seeking instrumental support and did not change over time. The coping strategies most commonly employed by diabetic patients are adaptive. Coping behavior changes over the 2 years from diagnosis. Promoting better understanding of diabetes, perceptions of personal control and treatment effectiveness are more likely than perception of health threat to sustain adaptive problem focused coping behavior.

Does life course socio-economic position influence chronic disabling pain in older adults? A general population study

BACKGROUND

Chronic pain is the leading cause of disability in developed countries. Prevalence is linked with socio-economic position (SEP), but little is known about the influence of SEP on disabling pain over the life course. We have investigated the influence of different life course trajectories of SEP on disabling pain ('pain interference') in postal surveys of adults aged ≥50 years sampled from the general population of adults registered with three UK general practices.

METHODS

Current pain interference was measured using the dichotomized 36-item Short-Form (SF-36) health survey. Three recalled SEP measures (age left school, longest job and current/most recent job) were dichotomized into low SEP (left school at or before minimum school leaving age; reported routine or manual occupations) and high SEP, from which eight life course SEP trajectories were constructed. Associations of (i) eight SEP trajectories and (ii) three individual SEP measures adjusted for each other, with pain interference, adjusted for potential confounders, were calculated using logistic regression.

RESULTS

A total of 2533 individuals provided data on all three SEP measures. A consistently low life course SEP trajectory was significantly associated with current pain interference compared with a high trajectory [odds ratio (OR) = 2.76, 95% confidence interval (CI): 2.19-3.47], even after adjustment for age and gender. Further adjustment reduced the association but it remained significant (OR = 2.04; 95% CI: 1.55-2.68). In the model with individual measures, low age left school (OR = 1.45; 95% CI: 1.15-1.82) and manual longest job (OR = 1.47; 95% CI: 1.13-1.91) were independently associated with pain interference.

CONCLUSIONS

Our results highlight the potential for reducing chronic disabling pain in later life by addressing inequalities in both childhood education and adult occupational opportunities.

Is cancer associated with polymyalgia rheumatica? A cohort study in the General Practice Research Database

Objective

To investigate the incidence of new cancer diagnoses in a community sample of patients with polymyalgia rheumatica (PMR).

Methods

All incident cases of PMR in the UK General Practice Research Database (GPRD) (1987–99), without pre-existing cancer or vascular disease and treated with corticosteroids (n=2877) were matched with up to five age, sex and GP practice patients without PMR (n=9942). Participants were followed up until first cancer diagnosis, death, transfer out of the database or end of available records.

Results

The mean age of the sample was 71.6 years (SD 9.0), 73% were female. Median follow-up time was 7.8 years (IQR 3.4, 12.3). 667 (23.2%) people with a PMR diagnosis developed cancer compared with 1938 (19.5%) of those without PMR. There was an interaction between PMR status and time. In the first 6 months after diagnosis, those with a PMR diagnosis were significantly more likely to receive a cancer diagnosis (adjusted HR (95% CI): 1.69 (1.18 to 2.42)). The number of events was small, but occurrences of prostate, blood, lymph nodes, female reproductive and nervous system cancers may be more common in those with PMR in the first 6 months after PMR diagnosis.

Conclusions

An increase in the rate of cancer diagnoses was noted in the first 6 months of observation, but we were unable to determine whether the cancer incidence in PMR was different from controls, beyond this time point. Clinicians should ensure they fully exclude cancer as a cause of PMR-like symptoms and monitor patients for possible malignancies.

Statin use and clinical osteoarthritis in the general population: a longitudinal study

BACKGROUND

One hypothesis has posited whether abnormal lipid metabolism might be a causal factor in the pathogenesis of osteoarthritis (OA). Routine statin use in clinical practice provides the basis for a natural experiment in testing this hypothesis.

OBJECTIVE

To test the hypothesis that statins reduce the long-term occurrence of clinically defined OA.

DESIGN

Cohort design with a 10-year follow-up.

PARTICIPANTS

16,609 adults cardiovascular disease cohorts aged 40 years and over from the UK General Practice Research Database with data available to 31 December 2006.

INTERVENTION

Statins were summarised as annual mean daily dose and dose change over two-year time periods.

MAIN MEASURES

Incident episode of clinically defined osteoarthritis was assessed within 2 years, and at 4-year and 10-year follow-up time periods, using Cox and discrete time survival analysis. Covariates included age, gender, deprivation, body mass index, cholesterol level, pain-modifying drug co-therapies, and duration and severity of cardiovascular disease.

KEY RESULTS

Higher therapeutic dose of statin, with a treatment duration of at least 2 years was associated with a significant reduction in clinical OA compared to non-statin users in the follow-up time period. The estimated adjusted rate ratios were as follows: lowest statin dose quartile 1: 2.5 (95 % CI 2.3, 2.9); quartile 2: 1.3 (1.1, 1.5); quartile 3: 0.8 (0.7, 0.95); and highest statin dose quartile 4: 0.4 (0.3, 0.5). The largest statin dose increments were associated with significant reductions estimated at 18 % in OA outcome within 2 years and 40 % after 4 years, compared to non-statin users.

CONCLUSIONS

This longitudinal study from a national clinical practice setting provides evidence that higher statin dose and larger statin dose increments were associated with a reduction in clinically defined OA outcome.

Point-of-care prognosis for common musculoskeletal pain in older adults

IMPORTANCE

Many site-specific, multivariable risk models for predicting the outcome of musculoskeletal pain problems have been published. The overlapping content in these models suggests a common set of generic indicators suitable for use in primary care.

OBJECTIVE

To investigate whether a brief set of generic prognostic indicators can predict the outcome of musculoskeletal pain in older patients presenting to general practitioners.

DESIGN, SETTING, AND PARTICIPANTS

A prospective observational cohort study conducted from September 1, 2006, through March 31, 2007, of consecutive patients 50 years or older presenting with noninflammatory musculoskeletal pain to 1 of the 5 participating general practices in the United Kingdom.

MAIN OUTCOME MEASURES

During consultation, the treating physician assessed and recorded 5 brief generic items (duration of present pain episode, current pain intensity, pain interference with daily activities, presence of multiple-site pain, and ultrashort depression screen) and recorded their overall prognostic judgment. The primary outcome was patient-rated improvement, which was measured 6 months after consultation and cross-validated with repeated measures up to 3 years.

RESULTS

A total of 194 (48.1%) of 403 participants were classified as having an unfavorable outcome at 6 months. Inclusion of 3 generic prognostic indicators (duration of present pain episode, pain interference with daily activities, and presence of multiple-site pain) in the prognostic model improved on reliance on physicians' prognostic judgment alone (C statistic = 0.72 vs 0.62; net reclassification index = 0.136; proportion correctly classified = 69%). The improvement in prognostic accuracy was attributable to correcting physicians' tendency toward overoptimistic expectations of outcome.

CONCLUSIONS AND RELEVANCE

Three easy-to-obtain pieces of information followed by systematic recording of the general practitioners' prognostic judgment provide a simple generic assessment of prognosis at point of care in older persons presenting with musculoskeletal problems to primary care practices in the United Kingdom. Such an assessment offers a common foundation for investigating the usefulness of prognostic stratification for guiding management in the consultation across a range of common painful conditions.

The development and validation of an internet-based training package for the management of perineal trauma following childbirth: MaternityPEARLS

BACKGROUND

Birth-related perineal trauma has a major impact on women's health. Appropriate management of perineal injuries requires clinical knowledge and skill. At present, there is no agreement as to what constitutes an effective clinical training programme, despite the presence of sufficient evidence to support standardised perineal repair techniques. To address this deficiency, we developed and validated an interactive distance learning multi-professional training package called MaternityPEARLS.

METHOD

MaternityPEARLS was developed as a comprehensive e-learning package in 2010. The main aim of the MaternityPEARLS project was to develop, refine and validate this multi-professional e-learning tool. The effect of MaternityPEARLS in improving clinical skills and knowledge was compared with two other training models; traditional training (lectures + model-based hands on training) and offline computer lab-based training. Midwives and obstetricians were recruited for each training modality from three maternity units. An analysis of covariance was done to assess the effects of clinical profession and years of experience on scoring within each group. Feedback on MaternityPEARLS was also collected from participants. The project started in January 2010 and was completed in December 2010.

RESULTS

Thirty-eight participants were included in the study. Pretraining and post-training scores in each group showed considerable improvement in skill scores (p<0.001 in all groups). Mean changes were similar across all three groups for knowledge (3.24 (SD 5.38), 3.00 (SD 3.74), 3.30 (SD 3.73)) and skill (25.34 (SD 8.96), 22.82 (SD 9.24), 20.7 (SD 9.76)) in the traditional, offline computer lab-based and e-learning groups, respectively. There was no evidence of any effect of clinical experience and baseline knowledge on outcomes.

CONCLUSIONS

MaternityPEARLS is the first validated perineal trauma management e-learning package. It provides a level of improvement in skill and knowledge comparable to traditional methods of training. However, as an e-learning system, it has the advantage of ensuring the delivery of a standardised, continuously updated curriculum that has global accessibility.