Likelihood-based analysis of longitudinal data from outcome-related sampling designs.

Investigators commonly gather longitudinal data to assess changes in responses over time and to relate these changes to within-subject changes in predictors. With rare or expensive outcomes such as uncommon diseases and costly radiologic measurements, outcome-dependent, and more generally outcome-related, sampling plans can improve estimation efficiency and reduce cost. Longitudinal follow up of subjects gathered in an initial outcome-related sample can then be used to study the trajectories of responses over time and to assess the association of changes in predictors within subjects with change in response. In this article, we develop two likelihood-based approaches for fitting generalized linear mixed models (GLMMs) to longitudinal data from a wide variety of outcome-related sampling designs. The first is an extension of the semi-parametric maximum likelihood approach developed in Neuhaus, Scott and Wild (2002, Biometrika 89, 23-37) and Neuhaus, Scott and Wild (2006, Biometrics 62, 488-494) and applies quite generally. The second approach is an adaptation of standard conditional likelihood methods and is limited to random intercept models with a canonical link. Data from a study of attention deficit hyperactivity disorder in children motivates the work and illustrates the findings.

Partial likelihood ratio tests for the Cox model under complex sampling.

We develop an analog of the likelihood ratio test for Cox proportional hazards models fitted to sample survey data. We look at methods for computing the asymptotic distribution and at ways of improving the small-sample performance. The methods are illustrated with examples using data from the National Health and Nutrition Examination Survey and from a stratified case-cohort study.

Patterns of family doctor decision making in practice context. What are the implications for medical practice variation and social disparities?

Medical practice variation and social disparities in health are pervasive features of health care systems. But what impact might everyday clinical decision making have in shaping such aggregate patterns, and could this in turn be influenced by the immediate environment in which family doctors practise? We investigate this by studying inter-practitioner variation in clinical activity across four payment types in New Zealand, a "gatekeeper" primary care system. We do this for four measures of clinical activity by patient ethnic and socio-economic status in a 2001/2002 representative sample of 9272 encounters at 185 family practices. Initial analysis showed little variation in clinical activity either by patient status or by practice type. However, with the application of multi-level statistical techniques it was evident that, while there was still little systematic difference in practitioner activity rates by patient status, inter-practitioner variation was greater for patients of ethnic minority background and from socio-economically deprived areas. Furthermore, this variability was particularly marked in fee-for-service practice settings. Thus, to the extent that family doctor decision-making behaviour within practice context helps shape aggregate patterns of medical practice variation and social disparity, treatment differences are likely associated not with the level of service but with its variability.

A Hardy-Weinberg equilibrium test for analyzing population genetic surveys with complex sample designs.

Testing for deviations from Hardy-Weinberg equilibrium is a widely recommended practice for population-based genetic association studies. However, current methods for this test assume a simple random sample and may not be appropriate for sample surveys with complex survey designs. In this paper, the authors present a test for Hardy-Weinberg equilibrium that adjusts for the sample weights and correlation of data collected in complex surveys. The authors perform this test by using a simple adjustment to procedures developed to analyze data from complex survey designs available within the SAS statistical software package (SAS Institute, Inc., Cary, North Carolina). Using 90 genetic markers from the Third National Health and Nutrition Examination Survey, the authors found that survey-adjusted and -unadjusted estimates of the disequilibrium coefficient were generally similar within self-reported races/ethnicities. However, estimates of the variance of the disequilibrium coefficient were significantly different between the 2 methods. Because the results of the survey-adjusted tests account for correlation among participants sampled within the same cluster, and the possibility of having related individuals sampled from the same household, the authors recommend use of this test when analyzing genetic data originating from sample surveys with complex survey designs to assess deviations from Hardy-Weinberg equilibrium.

Using multiple frames in health surveys.

In this paper, we look at a unified approach to estimation in multiple frame surveys that enables the analysis to be handled with standard survey software. We investigate how the approach performs with data from the Auckland Diabetes, Heart and Health Survey, a dual frame survey looking at risk factors for cardiovascular disease with special attention to Maori and Pacific Island people and to older people. The results suggest that the use of multiple frames might be useful in other surveys where good estimates are wanted for both the whole population and particular subgroups at the same time.

Case-control analysis with a continuous outcome variable.

It is not uncommon for a continuous outcome variable Y to be dichotomized and analysed using logistic regression. Moser and Coombs (Statist. Med. 2004; 23:1843-1860) provide a method for converting the output from a standard linear regression analysis using the original continuous outcome Y to give much more efficient inferences about the same odds-ratio parameters being estimated by the logistic regression. However, these results apply only to prospective studies. This paper follows up Moser and Coombs by providing an efficient linear-model-based solution for data collected using case-control studies. Gains in statistical efficiency of up to 240 per cent are obtained even with small to moderate odds ratios. Differences in design efficiency between case-control and prospective sampling designs are found to be much smaller, however, when linear-model-based analyses are being used than they are when logistic regression analyses are being used.

Do hospital bed reduction and multiple system reform affect patient mortality?: A trend and multilevel analysis in New Zealand over the period 1988-2001.

BACKGROUND: The impact of hospital and system restructuring on the quality and pattern of care is an important issue of public policy concern. OBJECTIVE: To assess the effect on patterns of care and patient outcomes of a substantial reduction in public hospital bed availability and multiple reorganizations in New Zealand through the 1990s. RESEARCH DESIGN: Trend analysis using both tabular and multilevel techniques. SUBJECTS: Access to discharge data, amounting to 6,639,487 records, was secured for all 34 major public hospitals in New Zealand over the period 1988-2001. OUTCOME MEASURES: Number of discharges, admission rate, access levels, mean length of stay, unplanned readmission rate, and 60-day postadmission mortality rate. RESULTS: Although the number of inpatient beds in use declined by one-third over the period and the national population grew by nearly one-fifth, discharge volumes increased significantly and rates of inpatient admission were maintained, as were access levels for vulnerable groups. These changes were accompanied by workload adjustments (a halving in length of stay and an increase by a quarter in readmission rates). Yet age-adjusted postadmission patient mortality decreased by a quarter over the period of study, a rate of decline that was slowed by the major workload adjustments but not by reform phase. CONCLUSIONS: Other things being equal, a substantial reduction in inpatient bed availability can be effected in national public hospital systems, while largely maintaining access and quality of care. However, the workload adjustments that are required may slow improvements in patient outcomes.

On the Breslow-Holubkov estimator.

Breslow and Holubkov (J Roy Stat Soc B 59:447-461 1997a) developed semiparametric maximum likelihood estimation for two-phase studies with a case-control first phase under a logistic regression model and noted that, apart for the overall intercept term, it was the same as the semiparametric estimator for two-phase studies with a prospective first phase developed in Scott and Wild (Biometrica 84:57-71 1997). In this paper we extend the Breslow-Holubkov result to general binary regression models and show that it has a very simple relationship with its prospective first-phase counterpart. We also explore why the design of the first phase only affects the intercept of a logistic model, simplify the calculation of standard errors, establish the semiparametric efficiency of the Breslow-Holubkov estimator and derive its asymptotic distribution in the general case.

Quality of hospital care for Maori patients in New Zealand: retrospective cross-sectional assessment.

BACKGROUND: New Zealand has a substantial indigenous minority--the Maori--that has considerably worse health status than the majority population. We aimed to assess possible disparities in quality of hospital care for Maori with data on preventable adverse events as an indicator of suboptimum treatment. METHODS: We undertook a nationally representative cross-sectional survey of admissions to general public hospitals with more than 100 beds providing acute care. A sample of 6579 patients admitted in 1998 to 13 hospitals was selected by stratified systematic list sample. We did a two-stage retrospective assessment of records by structured implicit review. Outcome measures were occurrence, effect, and preventability of adverse events. FINDINGS: Maori accounted for just greater than 15% of admissions and were on average younger, were more likely to be from from deprived areas, had a different case mix, and were in hospital for a shorter stay compared with patients of non-Maori/non-Pacific origin. Overall, after age standardisation, 14% of admissions for Maori were associated with an adverse event, compared with 11% for non-Maori/non-Pacific patients (p=0.01 for difference between groups). For preventable, in-hospital events, this disparity persisted after controlling for age, other sociodemographic factors, and case mix (adjusted odds ratio 1.47; p=0.05). Analysis of potential causal factors showed no markedly or consistently different pattern between the groups. INTERPRETATION: Despite a predominantly publicly funded hospital system, our findings suggest that hospital care received by Maori is marginally poorer than that received by New Zealand citizens of non-Maori/non-Pacific origin. Although no cause specific to Maori was evident, various policy and system issues can be addressed.

Modeling eligibility under national systems of compensation for treatment injury.

How can different schemes of compensation for treatment injury be evaluated? This article offers an empirical approach to assessment based on the simulation of alternative models of eligibility using real-world data. It draws on information about adverse events generated from a representative survey of public hospital admissions in New Zealand and classifies these under a range of eligibility criteria for different possible compensation systems. These are then evaluated according to a number of policy design considerations, using variables available from the New Zealand study.

Secondary analysis of case-control data.

We extend the discussion of Lee et al. and others on methods for performing secondary analyses of case-control sampled data and carry out an extensive investigation of efficiency and robustness. We find that, with the exception of the 'analyse-the-controls-only' strategy for populations in which cases are rare, ad hoc methods in common usage often lead to extremely misleading conclusions and that it is not possible to tell in advance when this will happen. Weighted likelihood and semi-parametric maximum likelihood methods are justified theoretically. We find that semi-parametric maximum likelihood can be as much as twice as efficient as the weighted method, but is subject to bias in estimating parameters of interest when the nuisance models this method requires have been mis-specified. The weighted method needs no nuisance models and thus is robust in this regard, but we cannot tell when it is going to be very inefficient without sophisticated modelling as through the SPML method. Practitioners should routinely use both methods and will often have to weigh up the practical consequences of severe inefficiency and lack of robustness in the context of their enquiries.

Adverse events in New Zealand public hospitals II: preventability and clinical context.

AIMS: To assess the preventability and clinical context of adverse events identified in New Zealand public hospitals. METHODS: Two-stage retrospective review of 6579 medical records randomly sampled from admissions for 1998 in 13 generalist hospitals providing acute care. Initial screening and medical review according to a standardised protocol. RESULTS: Eight hundred and fifty adverse events were identified, of which over one third (315) were preventable to a significant degree. Preventability of events increased with age of patient, and was more characteristic of certain diagnostic categories. Half of all events (413/850) were both preventable and occurred in hospital, giving an occurrence rate of 6.3%. This rate increased with patient age, as did impact on patients and extended stay. There were also variations by diagnostic category. Over half of the events were associated with surgery (489) and one third with medicine (303); operative incidents were predominant in the former, drug-related in the latter, and system issues in both. Events in medicine had greater patient impact, and were less likely to have occurred in a public hospital. System errors featured prominently as an area for prevention, being associated with nearly half of all preventable in-hospital events. Consultation and education were also important, resources less so. CONCLUSIONS: Preventable adverse events have a major impact on patient outcomes and extended hospital stay. A substantial proportion of these are system related and, hence, in principle susceptible to quality improvement.

How much variation in clinical activity is there between general practitioners? A multi-level analysis of decision-making in primary care.

OBJECTIVES: There is considerable policy interest in medical practice variation (MPV). Although the extent of MPV has been quantified for secondary care, this has not been investigated adequately in general practice. Technical obstacles to such analyses have been presented by the reliance on ecological small area variation (SAV) data, the binary nature of many clinical outcomes in primary care and by diagnostic variability. The study seeks to quantify the extent of variation in clinical activity between general practitioners by addressing these problems. METHODS: A survey of nearly 10 000 encounters drawn from a representative sample of general practitioners in the Waikato region of New Zealand was carried out in the period 1991-1992. Participating doctors recorded all details of clinical activity for a sample of encounters. Measures used in this analysis are the issuing of a prescription, the ordering of a laboratory test or radiology examination, and the recommendation of a future follow-up office visit at a specified date. An innovative statistical technique is adopted to assess the allocation of variance for binary outcomes within a multi-level analysis of decision-making. RESULTS: As expected, there was considerable variability between doctors in levels of prescribing, ordering of investigations and requests for follow up. These differences persisted after controlling for case-mix and patient and practitioner attributes. However, analysis of the components of variance suggested that less than 10% of remaining variability occurred at the practitioner level for any of the measures of clinical activity. Further analysis of a single diagnostic group--upper respiratory tract infection--marginally increased the practitioner contribution. CONCLUSIONS: The amount of variability in clinical activity that can definitively be linked to the practitioner in primary care is similar to that recorded in studies of the secondary sector. With primary care doctors increasingly being grouped into larger professional organisations, we can expect application of multi-level techniques to the analysis of clinical activity in primary care at different levels of organisational complexity.

Co-morbidity and health outcomes in three Auckland hospitals.

AIMS: To establish the burden of co-morbid disease using the Charlson Index among hospital inpatients and its relationship to key health outcomes. METHODS: An initial screen was carried out on 1575 medical records selected by systematic list sample from admissions for 1995 in three public hospitals in the Auckland region. In the course of the administration of the instrument, screeners were required to record the occurrence of co-morbid disease using the Charlson Index. RESULTS: A third of patients had co-morbid disease, of which chronic pulmonary disease and congestive heart failure were the most frequently recorded. While the Charlson Index was associated with age of patient, length of stay, inpatient mortality, and adverse event status, the simple presence or absence of co-morbidity was as an effective predictor as the extended index. Co-morbidity was more likely to be recorded for Maori, for patients from deprived areas, and for circulatory or respiratory diagnoses. Specific co-morbid conditions were predictive of health outcomes. CONCLUSIONS: Levels of co-morbid disease established for patients using the Charlson Index in three Auckland public hospitals are similar to those recorded internationally. Co-morbidity is predictive of key health outcomes that are of clinical and managerial significance. Controlling for co-morbidity will be important in making comparisons of the quality of care.

Adverse events in New Zealand public hospitals I: occurrence and impact.

AIM: To assess the occurrence and impact of adverse events in New Zealand public hospitals. METHODS: Two-stage retrospective review of 6579 medical records, selected by systematic list sample from admissions for 1998 in 13 generalist hospitals providing acute care. After initial screening, medical records were reviewed by trained medical practitioners using a standardised protocol. RESULTS: Except for hospital stay, the sample appeared to be closely representative of New Zealand public hospital admissions for 1998 on key demographic and clinical criteria. The proportion of hospital admissions associated with an adverse event was 12.9% (incidence rate, 11.2%), of which nearly one fifth had occurred outside a public hospital (mainly doctor s rooms, patient s home, rest home, or private hospital). Most adverse events had minor patient impact, with less than 15% associated with permanent disability or death. Hospital workload was strongly affected, however, with adverse events adding an average of over nine days (median 4 days) to the expected hospital stay. There was limited evidence of patterning by diagnostic category. The elderly were disproportionately affected. CONCLUSIONS: The study provides representative base parameters that can contribute to the wider understanding, and potential improvement, of patient safety and the quality of care in New Zealand public hospitals.