|Home | About | Journals | Submit | Contact Us | Français|
To review studies of the use of mortality data in quality and safety improvement in general practice.
Search of Medline, Embase and CINAHL for articles reporting mortality monitoring or mortality reviews in general practice. The included articles were reported in English and of any study design, excluding case reports and comment pieces. Studies of palliative care and bereavement, and of primary care programmes in developing countries, were excluded.
229 articles were identified in the searches, 65 were identified as potentially relevant and 53 were included in the review. The studies addressed the impact of primary care provision on mortality rates, methods of monitoring mortality, and the role of audit and death registers in quality and safety improvement. General practitioners were interested in using mortality data but reported difficulties in obtaining complete information. There were no experimental studies of the impact of the use of mortality data, and little evidence of long‐term systematic initiatives to use mortality data in quality and safety improvement in general practice.
Mortality data are not used systematically in general practice although general practitioners appear interested in the potential of this information in improving quality and safety. Improved systems to provide complete data are needed and experimental studies required to determine the effectiveness of use of the data to improve general practice care.
Data about mortality in general practice populations have not been routinely used to monitor performance or plan practice policies and services (in this paper we use the term general practice to include family practice, and general practitioners to include family practitioners and family physicians). The data may not have been readily available to general practitioners, or may have been regarded as difficult to interpret in the context of small populations. However, in the UK increased attention is being paid to mortality monitoring following the discovery that the doctor Harold Shipman unlawfully killed around 236 of his general practice patients.1,2 The inquiry set up to investigate what happened recommended the development of a national system for monitoring general practice mortality rates, that practices should keep death registers, and that health authorities should undertake reviews of samples of records of deceased patients.3
Despite the interest in monitoring mortality in general practice, concerns have been raised about the practicality of mortality monitoring at the level of the practice4 and the government has initiated a review of this issue along with other of the inquiry's recommendations. Thus, the role of monitoring in informing practice policies and promoting quality improvement remains unclear. Monitoring may have potential to support improvements in quality and safety within practices, but whether and how this potential may be realised is uncertain. Therefore, we undertook a review to investigate the potential of mortality monitoring in general practice. The specific aims of the review were to identify what methods have been used to review mortality data in general practice and how the data have been used.
We searched MEDLINE (1966 to end 2005), EMBASE (1980 to end 2005) and CINAHL (1982 to end 2005) for papers reporting studies of mortality monitoring or mortality reviews in general practice. Search terms included relevant subject headings supplemented by appropriate free‐text terms for general practice and primary health care and death and mortality. Reference lists of papers were scanned to identify additional papers (copies of the search strategies are available on request from the authors).
We included papers published in English reporting experimental and non‐experimental studies of methods of compiling and maintaining information on deaths in general practice, reviews of deaths for quality improvement, health service provision and targeting of health initiatives, and studies of monitoring of mortality rates to detect illegal or aberrant behaviour by general practitioners. We defined general practice populations as people with primary healthcare services available to them from general or family practice or other primary healthcare providers, whether or not they were currently using these services. People in these groups may be identified from population registers or registers of patients maintained by primary care services, depending on the systems adopted in different countries. Studies of patients identified because they had used other services such as hospitals were excluded. We excluded case reports, letters and comment pieces. We also excluded studies of the care and experiences of terminally ill patients, palliative care, euthanasia, bereavement counselling, studies of disease epidemiology, and evaluations of primary care programmes in developing countries. Abstracts of articles identified in the searches were reviewed, potentially relevant articles being obtained. In view of the types of studies included in the review, a quantitative analysis was not appropriate, and therefore we summarised articles in a table and present a narrative review.
We identified 229 articles in the searches, of which 65 were assessed as potentially relevant. Of these, 53 were eventually included (see table 11),), 17 of which involved studies to assess the impact of provision primary care services on mortality, 8 studies and reports of systems to monitor patient mortality rates in general practice, and 28 studies or reports of general practice registers, audits and confidential inquiries in single practices or groups of practices.
Most of the studies of the impact of primary care provision used regression modelling of data extracted from existing datasets, and had been undertaken in the US. The US studies indicated an association between the supply of primary care physicians and lower mortality5,6 even when socioeconomic status and ethnicity were accounted for.7,8 The finding held for infant mortality,9 stroke mortality,10 all‐cause, heart disease and cancer mortality,11,12 cervical cancer rates and mortality,13 and the magnitude of the effect varied between black and white ethnic groups.14,15 There was also an association between level of development of primary care and all‐cause mortality in a study involving 18 wealthy countries.16 In a comparison of primary care physicians and specialists in the role of personal doctor, patients of family physicians experienced lower mortality rates, after adjusting for demographic and disease status.17 However, in another US study, no relationship was found between falls in mortality rates and the provision of additional primary care services in physician shortage areas.18
The relationship between primary care provision and population mortality was not found in two UK studies. In a study of standardised mortality ratios and infant mortality in 99 English health authorities, the relation between GP supply and mortality failed to reach statistical significance after adjusting for deprivation, ethnicity, social class and long‐term limiting illness.19 In a subsequent analysis, lower mortality was found to be weakly associated with increasing practice size.20 In a study of the impact of systematic management of hypertension in comparison with usual care over a 15 year period, although there were fewer cases of left ventricular hypertrophy and angina in the systematic treatment group, there was no difference in mortality.21
The second group of studies related to monitoring systems to detect excess mortality in general practice patient populations. All of these were from the UK and concerned with creating a monitoring system. In one Scottish health district, observed and expected numbers of deaths for each practice were compared, the expected being age, sex and deprivation adjusted estimates based on the district population.22 Practices were sent reports that showed how they compared with other practices and whether their rates were higher than expected. A high number of patients in nursing homes was suggested as an explanation for excess mortality. In a scheme involving 114 general practices in Northern Ireland, cross‐sectional control charts were used and practices provided with feedback and workshops.23 Care for patients in nursing homes as well as poor data quality were proposed as the reason for most cases of excess mortality in a study evaluating the feasibility of monitoring using cumulative sum charts.24 More detailed adjustment for case mix was also recommended, although the charts were judged to be potentially useful for monitoring deaths in primary care. Control charts were evaluated as a monitoring tool in another study,25 and others have highlighted the inevitability of detecting practices with excess mortality for innocent reasons26,27—false positives—and the time‐consuming nature of investigations to determine the cause of excessive mortality.28 In a report of a process for investigating practices with excess mortality, cumulative sum plots were used to highlight the association between mortality and nursing homes.29 However, the studies of monitoring systems were all concerned with exploring methods and feasibility rather than the systematic evaluation of the impact of monitoring. There were no experimental studies.
The articles dealing with registers and audit were all concerned with the potential of reviews of deaths to contribute directly to improvements in quality or safety in practices. They included descriptions of methods for creating death registers either in individual general practices30,31 or for groups of practices.32 In studies published more than 20 years ago, practices had simply reported the numbers and causes of deaths among their patients.33,34 Several of the more recent studies highlighted the difficulties practices had in obtaining timely and accurate data.35,36 Death registers were reported as facilitating communication in primary care teams, bereavement follow‐up and practice audit.37 Information obtained from collection of information about deaths had been used to help practices understand the healthcare needs of their practice populations and enable comparisons between practices.38 One team reported use of the information to investigate nursing involvement in terminal care at home,39 three others to investigate place of death,40,41,42 three more potentially preventable deaths,43,44,45 and others age at death and smoking46 and referrals for autopsy.47 In another practice, information was sought about the involvement of the general practitioner in the care of patients before death.48 In the only study to attempt to directly relate clinical practice to mortality, the collection of information about deaths and clinical performance over 25 years enabled the practice to reflect on the impact of its activities in comparison with other practices.49,50 Information on the numbers and causes of deaths had also been used in multipractice audit to promote reflection on potentially preventable deaths.51
Confidential inquiries have been used in specialist settings to identify common failings in clinical practice and monitor the impact of initiatives to improve care. We identified an example of critical incident reviews undertaken by primary care teams,52 and also examples of locality schemes involving central data collection and analysis combined with feedback to practices, including programmes to investigate deaths due to asthma, stroke and suicides.53,54,55,56,57 All the articles dealing with audit and registers involved exploration of methods; there were no experimental studies designed to investigate the impact of audit or inquiries into deaths.
We have undertaken a review of studies of use of mortality data in quality improvement activities in general practice, and found evidence of increasing interest in such activity. Two decades ago interest was limited and restricted to descriptions of numbers and causes of deaths, but in the last 10 years exploration has begun of the role of registers, monitoring, audit and critical incident review.
Mortality data were used in three ways. In the first, involving principally US studies, interest centred on the provision of evidence that increased primary care physician supply is associated with lower mortality. The findings point to an association in the US, but it would be premature to accept that the relationship is causative because it could be explained by unknown characteristics of either primary care physicians or localities that lead to the concentration of primary care physicians in areas with low population mortality. The association was also identified in an international comparison, but variation in health system characteristics argue for caution in drawing firm conclusions. Evidence is required from experimental or quasi‐experimental studies such as time series to confirm whether increasing the provision of primary care services in developed countries does reduce population mortality rates.
The second group of studies were concerned with the identification of practitioners and practices with higher than expected patient mortality rates. These studies were all from the UK and prompted by the Shipman case mentioned earlier. The findings indicate that improved data quality is required, and that the investigation of practices that signal with high mortality rates will often identify innocent explanations. The potential of monitoring to inform quality and safety initiatives has not been investigated in these studies, but the potential is suggested by the third group of studies involving the development of registers and conduct of audit and incident reviews. These indicated that practices had difficulty in obtaining complete information about deaths and found the routine provision of this information helpful in supporting bereaved relatives and facilitating practice‐based educational discussions. Practitioners appeared interested in the data but reported difficulties in obtaining complete information and data that enabled them to compare mortality from their own practice with similar practices. Some practices had also undertaken audits to identify potentially preventable causes of deaths, and others had taken part in either practice or locality based critical incident reviews. Critical incident reviews are increasingly common in the UK, and the analysis of reports of reviews of deaths at the locality level could have a role to play in improving patient safety in primary care. There were no experimental studies of these quality improvement methods.
The limitations of the review should be noted. We used a broad search strategy and believe that most relevant articles were identified, but studies of the role of mortality data in audit and quality improvement published in languages other than English were omitted. There may have been progress in some countries which we have not been able to report. The studies were undertaken in a limited number of countries, and the findings may not be applicable in countries with different healthcare systems. Given the heterogeneous nature of the included studies and the absence of experimental studies, we have undertaken a narrative review only. Nevertheless, with this qualification, our findings do indicate that the development of methods to use mortality data to improve the quality and safety of general practice has begun. Further research is now needed to develop systems to provide data routinely to enable primary care professionals to explore associations between the processes and outcomes of care, and to evaluate the potential of monitoring with feedback and mortality reviews to improve patient safety and promote improved clinical policies.
This study was supported by a grant from the Leicestershire Medical Research Foundation (Medisearch).
Competing interests: None declared.